economics
“This impressive collection of papers is a fitting tribute to one of the greatest economists of the late twentieth century.” —Michael Bordo, Department of Economics, Rutgers University Carmen M. Reinhart is Professor of Economics at the University of Maryland. Carlos A. Végh is Professor of Economics at the University of Maryland. Andrés Velasco, on leave as Sumitomo Professor of International Finance and Development at Harvard’s John F. Kennedy School of Government, is currently serving as Chile’s Minister of Finance. All three are Research Associates at the National Bureau of Economic Research.
“Guillermo Calvo’s work has always been rigorous in its theoretical treatment of the most important macroeconomic phenomena of our time. This fine collection of papers is very much in the spirit of his work, tackling critical problems with new theories and careful empirical analysis. It includes important new research by the leaders in the profession whose papers deal with financial crises, monetary regimes, policy rules, and determinants of economic growth. This is a masterful achievement, one that no serious scholar of international finance can afford to be without.” —Andrew K. Rose, Rocca Professor, Haas School of Business, University of California, Berkeley
Contributors Cover photograph by Pilar Bilecky.
Leonardo Auernheimer Fabrizio Coricelli Padma Desai Allan Drazen Sebastian Edwards Roque B. Fernández Stanley Fischer Ricardo Hausmann Bostjan Jazbecˇ Peter Isard
Graciela L. Kaminsky Michael Kumhof Amartya Lahiri I. Igal Magendzo Enrique G. Mendoza Frederic S. Mishkin Igor Masten Pritha Mitra Alejandro Neut Maurice Obstfeld Edmund S. Phelps
The MIT Press Massachusetts Institute of Technology Cambridge, Massachusetts 02142 http://mitpress.mit.edu
Assaf Razin Carmen M. Reinhart Francisco Rodriguez Efraim Sadka Ratna Sahay Rajesh Singh Evan Tanner Rodrigo Wagner Carlos A. Végh Andrés Velasco
978-0-262-18266-9
Money, Crises, and Transition
Essays in Honor of Guillermo A. Calvo edited by Carmen M. Reinhart Carlos A. Végh Andrés Velasco
Money, Crises, and Transition Essays in Honor of Guillermo A. Calvo edited by Carmen M. Reinhart, Carlos A. Végh, and Andrés Velasco
Guillermo A. Calvo, one of the most influential macroeconomists of the last thirty years, has made pathbreaking contributions in such areas as time inconsistency, lack of credibility, stabilization, transition economies, debt maturity, capital flows, and financial crises. His work on macroeconomic issues relevant for developing countries has set the tone for much of the research in this area and greatly influenced practitioners’ thinking in Latin America, Eastern Europe, Asia, and elsewhere. In Money, Crises, and Transition, leading specialists in Calvo’s main areas of expertise explore the themes behind this impressive body of work. The essays take on the issues that have fascinated Calvo most as an academic, a senior advisor at the International Monetary Fund, and as the chief economist at the Inter-American Development Bank: monetary and exchange rate policy (both in theory and practice); financial crises; debt, taxation, and reform; and transition and growth. A final section provides a behind-the-scenes look at Calvo’s career and intellectual journey and includes an interview with Calvo himself.
Money, Crises, and Transition
Guillermo A. Calvo. Photograph by Pilar Bilecky.
Money, Crises, and Transition Essays in Honor of Guillermo A. Calvo
edited by Carmen M. Reinhart, Carlos A. Ve´gh, Andre´s Velasco
The MIT Press Cambridge, Massachusetts London, England
( 2008 Massachusetts Institute of Technology All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from the publisher. MIT Press books may be purchased at special quantity discounts for business or sales promotional use. For information, please e-mail
[email protected] or write to Special Sales Department, The MIT Press, 55 Hayward Street, Cambridge, MA 02142. This book was set in Palatino on 3B2 by Asco Typesetters, Hong Kong and was printed and bound in the United States of America. Library of Congress Cataloging-in-Publication Data Money, crises, and transition : essays in honor of Guillermo A. Calvo / edited by Carmen M. Reinhart, Carlos A. Ve´gh, Andre´s Velasco. p. cm. Includes bibliographical references and index. ISBN 978-0-262-18266-9 (hbk. : alk. paper) 1. Monetary policy. 2. Foreign exchange rates. 3. Financial crises. 4. International finance. I. Calvo, Guillermo A. II. Reinhart, Carmen M. III. Ve´gh Gramont, Carlos A., 1958– IV. Velasco, Andre´s. HG230.3.M682 2008 2007039868 332 0 .042—dc22 10 9 8 7
6 5 4 3
2 1
Contents
List of Contributors Acknowledgments Introduction xiii
ix xi
1
I
Monetary and Exchange Rate Policy in Theory
1
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate
3
Maurice Obstfeld 2
Optimal Exchange Rate Regimes: Turning Mundell-Fleming’s Dictum on Its Head 21 Amartya Lahiri, Rajesh Singh, and Carlos A. Ve´gh
3
Monetary Policy Rules, the Fiscal Theory of the Price Level, and (Almost) All that Jazz: In Quest of Simplicity 41 Leonardo Auernheimer
II Monetary and Exchange Rate Policy in Practice 4
69
Can Inflation Targeting Work in Emerging Market Countries?
71
Frederic S. Mishkin 5
Why Should Emerging Economies Give up National Currencies? A Case for ‘‘Institutions Substitution’’ 95 Enrique G. Mendoza
6
Hard Currency Pegs and Economic Performance Sebastian Edwards and I. Igal Magendzo
121
vi
Contents
III Financial Crises 7
157
Asset Prices and Self-Fulfilling Macroeconomic Pessimism
159
Andre´s Velasco and Alejandro Neut 8
The Center and the Periphery: The Globalization of Financial Turmoil
171
Graciela L. Kaminsky and Carmen Reinhart 9
Why Do Some Countries Recover More Readily from Financial Crises?
217
Padma Desai and Pritha Mitra IV Debt, Taxation, and Reforms
247
10 Government Debt: A Key Role in Financial Intermediation
249
Michael Kumhof and Evan Tanner 11 Capital Income Taxation in the Globalized World
279
Assaf Razin and Efraim Sadka 12 Can Public Discussion Enhance Program ‘‘Ownership’’?
295
Allan Drazen and Peter Isard V Transition and Growth
325
13 Sources and Obstacles for Growth in Transition Countries: The Role of Credit 327 Fabrizio Coricelli, Bostjan Jazbecˇ, and Igor Masten 14 Growth in Transition Economies: Domestic Policies, External Assistance, and Institution Building 349 Stanley Fischer and Ratna Sahay 15 Growth Collapses
377
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
Contents
vii
VI The Man Behind the Mind 16 The Columbia Years
429
431
Edmund S. Phelps 17 The Practitioner
437
Roque B. Ferna´ndez 18 In His Own Words: An Interview with Guillermo Calvo Enrique G. Mendoza Publications of Guillermo A. Calvo Index 471
463
445
Contributors
Leonardo Auernheimer Texas A&M University
Michael Kumhof International Monetary Fund
Fabrizio Coricelli University of Siena and University of Ljubljana
Amartya Lahiri University of British Columbia
Padma Desai Columbia University Allan Drazen University of Maryland Sebastian Edwards University of California, Los Angeles Roque B. Ferna´ndez Universidad del CEMA Stanley Fischer Bank of Israel Ricardo Hausmann Harvard University Peter Isard International Monetary Fund Bostjan Jazbecˇ University of Ljubljana and Slovenian National Bank Graciela L. Kaminsky George Washington University
I. Igal Magendzo Central Bank of Chile Igor Masten University of Ljubljana Enrique G. Mendoza University of Maryland Frederic S. Mishkin Columbia University Pritha Mitra International Monetary Fund Alejandro Neut BBVA Economic Research Maurice Obstfeld University of California, Berkeley Edmund S. Phelps Columbia University Assaf Razin Tel Aviv University and Cornell University
x
Carmen Reinhart University of Maryland Francisco Rodriguez Wesleyan University Efraim Sadka Tel Aviv University Ratna Sahay International Monetary Fund Rajesh Singh Iowa State University Evan Tanner International Monetary Fund Carlos A. Ve´gh University of Maryland Andre´s Velasco Harvard University Rodrigo Wagner Harvard University
Contributors
Acknowledgments
This festschrift grew out of a conference in honor of Guillermo Calvo, held at the International Monetary Fund’s headquarters April 15–16, 2004. We are grateful to the Fund—and, in particular, to the Research Department—for its hospitality and generosity in sponsoring this event. While putting together the conference program, we were fortunate to have the support of the Research Department’s staff, particularly Patricia Medina, who worked tirelessly for many months on conference logistics and participants’ travel arrangements. Numerous conference participants—including many who endured a long journey from Guillermo’s native Argentina—contributed to the conference’s success as presenters, lunch and dinner speakers, discussants, and session chairs. In addition to all of this volume’s contributors, we are thankful to Michael Bordo, Eduardo Borensztein, Ricardo Caballero, Agustin Carstens, Gerardo Della Paolera, Michael Dooley, Martin Eichenbaum, Andrew Feltenstein, Raquel Fernandez, Ronald Findlay, Robert Flood, Pablo E. Guidotti, Alejandro Izquierdo, Olivier Jeanne, Moshin Khan, Leonardo Leiderman, Ashoka Mody, Pablo Andres Neumeyer, Juan Pablo Nicolini, Guilllermo Perry, Eswar Prasad, Raghu Rajan, Anoop Singh, Federico Sturzenegger, Lars Svensson, Ernesto Talvi, John Taylor, and Martin Uribe for making this such a special occasion. Finally, we are grateful to Elizabeth Murry for her early and enthusiastic support for this project on behalf of The MIT Press and to John Covell, Lianna Kong, and Sandra Minkkinen for helping us take it to completion.
Introduction
Most of the chapters in this volume were prepared for a conference in honor of Guillermo Calvo, organized by the International Monetary Fund’s Research Department and held at Fund headquarters in Washington, DC, on April 15–16, 2004. At the editors’ request, a couple of chapters were specially prepared after the conference for inclusion in this volume. The Fund was a natural and gracious host since Guillermo had a distinguished affiliation with the Fund’s Research Department from 1987 to 1994. Under his intellectual leadership, the Research Department carried out path-breaking research on, among other issues, capital flows, debt maturity, and inflation stabilization. Guillermo also made important contributions to the internal discussion and formulation of Fund policies, particularly in Eastern Europe, the former Soviet Union, and Latin America. The conference brought together renowned academics and policy makers who have had the privilege of being associated with Guillermo throughout his illustrious career. Some of Guillermo’s former colleagues at Columbia University— where he began his academic career in the early 1970s—also gave fascinating behind-the-scenes accounts of Guillermo’s early professional blossoming. One such account, by Edmund Phelps, recent winner of the Nobel Prize in economic sciences, is included in this volume. Delivering the opening remarks, Agustin Carstens (at the time IMF Deputy Managing Director) struck a theme that would resonate throughout the conference. In Carstens’s words, ‘‘Guillermo’s ability to reduce complex problems to its essential elements has taught us that complex models are for lesser minds—for those who cannot grasp the essential elements out of a given reality. In Guillermo’s hands, the chaos of reality has always yielded simple and illuminating models.’’ Indeed, in an era in which increasingly complicated paradigms have become the rule, Guillermo’s simple—but never simplistic—models are a constant reminder that intellectual elegance and policy relevance can indeed go hand in hand.
xiv
Introduction
A second theme that pervaded the conference was Guillermo’s crucial role in bringing Latin American policy issues to the forefront of the academic discussion. In fact, Guillermo is arguably the single most important person in bringing modern economics to bear on the problems of nations south of the Rio Bravo. He brought rigor, discipline, and intellectual clarity to the difficult task of illuminating some of the region’s most pressing economic problems. As a token of gratitude, numerous academics and policymakers crossed the Rio Bravo to be present at the conference. Among them was Roque Fernandez—former Argentine finance minister—who spoke on Guillermo’s contributions to recent Argentine economic policy and whose remarks are also part of this volume. This volume is divided into six parts. The first five parts cover many of the issues that have been dear to Guillermo’s heart over his thirty-year career: monetary and exchange rate policy, financial crises, debt, taxation, reform, growth, and transition. The last part sheds light on the mind behind the man. In one capacity or another, all the chapters’ authors have been closely associated with Guillermo and have had the privilege of learning and benefiting from Guillermo’s seemingly infinite wisdom. Since some of Guillermo’s most influential contributions have been in monetary and exchange rate policy, it seems only appropriate for the volume to open with theoretical advances in this area. A perennial and important question in open economy macroeconomics is: should countries fix or float? In other words, should an open economy peg its currency against a strong world currency or should it let the currency float? The standard response, based on the Mundell-Fleming model, is that the optimal exchange rate regime should depend on the type of shock hitting the economy. If shocks are predominantly of real origin, flexible exchange rates are optimal as they allow a quicker adjustment of relative prices through changes in the nominal exchange rate. If shocks are predominantly monetary, however, then fixed (or predetermined) exchange rates are preferred because adjustments to the nominal money supply are automatically carried out by the central bank. As Guillermo has put it, this is a result that ‘‘every well-trained economist carries on the tip of her tongue.’’ The first two chapters in the book deal with challenges to this basic tenet. In chapter 1, Maurice Obstfeld takes issue with a recent finding by Michael Devereux and Charles Engel, which holds that even if an economy is hit by real shocks, fixed exchange rates may be preferable if the nominal prices of both exports and imports are preset in the domestic currency. The key behind the Devereux-Engel result is that, under such a pricing assumption, flexible rates cease to be a useful tool to effect adjustments in relative prices. Obstfeld argues that a minor modification of the Devereux-Engel set up—adding nontradable goods to their model—restores the optimality of flexible exchange rates under
Introduction
xv
real shocks, though for different reasons than those emphasized by the traditional Mundell-Fleming model. In Obstfeld’s set up, flexible exchange rates become optimal again because the nominal interest rate is freed as a monetary stabilization instrument. In chapter 2, Amartya Lahiri, Rajesh Singh, and Carlos Ve´gh approach the same problem from a very different angle by departing altogether from the traditional Mundell-Fleming set up. Based on the idea that asset market imperfections are likely to be as—if not more—important than goods market frictions in developing countries, they present a model of flexible prices where some economic agents do not have access to capital markets (in other words, there is asset market segmentation). In such a world, the authors show that the Mundell-Fleming result is turned on its head: flexible rates become optimal if monetary shocks dominate, while fixed rates are preferable if real shocks dominate. The key to understanding these results is that asset market segmentation critically affects the standard adjustment mechanism that operates under pegged exchange rates as it does not allow some agents to adjust their nominal money balances. In contrast, adjustments in the nominal exchange rate hit all agents equally. The chapter’s punchline is thus that the choice of the optimal exchange rate regime should depend not only on the type of shock (real versus monetary), but also on the type of friction (goods market versus asset market). In chapter 3, Leonardo Auernheimer goes back to basics and revisits some of the themes that Guillermo has focused on in many of his contributions: the relevance of the government budget constraint, the concern with levels versus rates of change, and the nominal interest rate as a policy instrument. The central question addressed in this chapter is the well-known (but controversial) fiscal theory of the price level. As Auernheimer points out, the literature on this topic is extensive, occasionally obscure, and not lacking in sound and fury (witness Willem Buiter’s claim that ‘‘the fiscal theory of the price level is fatally flawed’’). As though guided by Guillermo’s own invisible hand, Auernheimer sets up an extremely simple model and takes the reader on an illuminating journey into the fiscal theory of the price level (and some related issues), stressing some of its truths and debunking some of its myths. The second part of the book turns to practical and empirical aspects of monetary and exchange rate policy. In chapter 4, Frederic Mishkin notes that Guillermo has been rightly skeptical of the adoption of inflation targeting schemes in emerging markets. Guillermo has convincingly argued in various pieces that, given weak monetary and fiscal institutions, high degrees of dollarization, and financial vulnerability, the consequences of giving policy makers too much discretion could be disastrous. While taking Guillermo’s concerns seriously, Mishkin presents a strong case for inflation targeting in emerging markets provided that countries
xvi
Introduction
can find ways of dealing effectively with large exchange rate fluctuations. He illustrates his arguments by looking closely at the cases of Chile and Brazil. In chapter 5, Enrique Mendoza goes a step further and forcibly argues that, all things considered, emerging market countries would be better off giving up their national currency altogether. In his view, while giving up the national currency would certainly not eliminate business cycle fluctuations or financial problems associated with loose fiscal policies, it would considerably reduce an emerging market’s vulnerability to abrupt changes in capital flows and enable it to freeride on some of the credibility already gained by the country issuing the hard currency. But how do countries without a national currency fare in practice? There is virtually no hard evidence. In chapter 6, Sebastian Edwards and Igal Magendzo take a novel approach to analyzing the inflation and growth performance of countries with no currency of their own. To this end, they address an important methodological problem: joining a common-currency area (or dollarizing) may itself be an endogenous decision that could be related to economic performance. Their innovative answer is to resort to a ‘‘treatment effects model’’ (a technique often used in the labor literature) that jointly estimates the probability of being a common currency country and macroeconomic performance equations. Their main finding is that countries with no currency of their own have had significantly lower inflation and enjoyed higher growth, but suffered from higher macroeconomic volatility. The jury is thus still out on the net benefits of giving up a national currency. Ever since the now (in)famous Mexican crisis of December 1994 (aptly referred to as ‘‘the first financial crisis of the twenty-first century’’), Guillermo’s research on financial crises has set the agenda for much of the academic work in this area. The three chapters in part III provide new insights into this critical topic. In chapter 7, Andre´s Velasco and Alejandro Neut show how, in a simple world with capital market imperfections (that cause borrowing to be collateralized), pessimism can lead to self-fulfilling crises. To understand the basic idea, suppose that, due to a sudden burst of pessimism, stock market prices are expected to fall. If investment is subject to collateral constraints, the fall in stock market prices reduces the value of firms and hence investment. But lower investment leads to smaller returns in the future which, in turn, reduce stock market prices, thus validating the initial pessimism. Hence, bouts of pessimism can be self-fulfilling (Keynes’ animal spirits?), a scary thought indeed. Velasco and Neut proceed to analyze policies that might make this possibility less likely, such as expansionary monetary and fiscal policy, debt restructuring, and financial and legal reform. But in practice, once a crisis erupts, how does it spread out? And what determines the speed with which countries recover? In chapter 8, Graciela Kaminsky and Carmen Reinhart tackle the first question, analyzing how financial turbulence
Introduction
xvii
in emerging countries can spread across borders. They conclude that a financial crisis will spread globally only if it affects some major financial center; otherwise, it spreads, at most, regionally. For example, during the Asian crisis, Japanese banks’ exposure to Thailand—and their subsequent curtailment of lending to other Asian countries—played a critical role in spreading the crisis. The second question is tackled by Padma Desai and Pritha Mitra in chapter 9. They find that export performance before a crisis appears to be a critical variable in predicting the speed of recovery. Through investor expectations, export performance explains most of the difference in postcrisis interest rate and exchange rate movements. In contrast, the fiscal position and national savings do not seem to matter as much. This casts some serious doubts on the desirability of contractionary fiscal policies, typically advocated by the International Monetary Fund. Some of Guillermo’s most influential contributions have been in the area of debt, taxation, and reform. In joint work with Pablo Guidotti, Guillermo showed that, under certain conditions, unanticipated increases in government spending are optimally financed with unanticipated inflation, which, in the model, imposes no costs. But does unanticipated inflation really have few costs? In chapter 10— the opening chapter of part IV—Michael Kumhoff and Evan Tanner argue that the reason why policy makers avoid unanticipated levies on government debt (through outright default or sudden burst of high inflation) is that government debt plays a key role in financial intermediation. Hence, defaulting or inflating it away would be akin to a negative technological shock. To make an empirical case for this argument, they show that in developing countries (1) government domestic debt is now larger than external debt, and growing relative to external debt; (2) banks voluntarily hold a very large fraction of their assets in domestic government debt; and (3) a deep and stable government bond market is critical for further development of domestic financial markets. In this light, explicit or implicit defaults on government debt would have real costs that would need to be traded off against the benefits of lower distortionary taxation. As countries develop, they increasingly rely on (personal and capital) income taxes. But how feasible is it to rely on capital income taxation in an increasingly globalized economy? As globalization makes it easier for firms to either locate in low-tax countries and/or shift operations across countries to avoid taxation, will there be a race to the bottom? In chapter 11, Assaf Razin and Efraim Sadka build a political economy model to assess how feasible it is for governments to impose capital income taxes. They conclude that a race to the bottom will indeed take place and argue that available evidence for European countries bears this out. It is easy for theory to suggest this or that policy. But in practice implementation of different policies is fraught with difficulties. In particular, the IMF’s track record in helping developing countries achieve successful reforms is, at best,
xviii
Introduction
open to question. In chapter 12, Allan Drazen and Peter Isard argue that an oftcited reason IMF programs get off track is a lack of so-called program ‘‘ownership,’’ which loosely refers to a country’s commitment to a given set of reforms and stabilization policies specified in an IMF program. Drazen and Isard provide a theoretical framework for thinking about the role of public discussion in building and demonstrating ownership. In turn, this framework points to various directions in which IMF programs can be strengthened. During his seven years at the IMF’s Research Department (1987–1994), Guillermo was witness to one of the most extraordinary economic events of the twentieth century: the transformation of the former communist countries into market economies. Guillermo’s work in this area was very influential both within and outside the IMF. In particular, in joint work with Fabrizio Coricelli, Guillermo forcefully argued that one of the main reasons behind the massive output collapse at the outset of the transition was a severe credit crunch. In chapter 13, Fabrizio Coricelli, along with Bostjan Jazbecˇ and Igor Masten, takes this line of research one step further. The authors study the macroeconomic performance of the central European countries in the process of entering the European Union. They conclude that the relative underdevelopment of credit markets in these countries is partly responsible for low output growth and high volatility. Joining the European Union should thus allow these economies to further build their financial systems and to grow faster. Like Guillermo, Stanley Fischer was a privileged witness to the economic transformation in the former communist countries, first as the World Bank’s Chief Economist (1988–1990) and later as the IMF’s Deputy Managing Director (1994– 2001). In fact, Fischer (with various collaborators) wrote some of the most influential pieces on the macroeconomic aspects of the transition. In chapter 14, Fischer teams up once again with one of his main collaborators, Ratna Sahay, to examine the role of institutions and initial conditions in the transition process. They find that, while adverse conditions constrained growth in the initial years, their effect wore off with time. Fischer and Sahay also reject the charge that international financial organizations—and mainstream economists—paid inadequate attention to institution building. They argue instead that institution building— typically with substantial assistance from international financial institutions— complemented policy and structural reforms in promoting growth. In recent years, Guillermo (with various collaborators) has focused much of his energy on the consequences of sudden stops in capital inflows—particularly on output—and how countries appear to recover quickly from such crises without much internal or external credit. In chapter 15, Ricardo Haussman, Francisco Rodriguez, and Rodrigo Wagner take this analysis a step further and tackle the broader question: why do developing economies fall into deeper and longer reces-
Introduction
xix
sions than developed ones? Clearly, as the authors note, the phenomenon of deep and prolonged recessions presents a challenge to standard macroeconomic theory by suggesting that the typical growth picture of economic fluctuations around an increasing trend does not quite apply to the developing world. By focusing on a large sample, the authors conclude that while countries fall into these growth crises for multiple reasons—including wars, export collapses, sudden stops, and political transitions—most of these variables do not help predict the episode’s duration. They do find, however, that a measure of the density of a country’s export base is significantly associated with lower crisis duration. The volume closes with three short chapters that provide a rare and fascinating behind-the-scenes look at Guillermo’s career and intellectual journey from his early days in Buenos Aires through his doctorate at Yale, his years at Columbia, and his tenure at the IMF’s Research Department, to his more recent and highly influential role as the Inter-American Development Bank’s Chief Economist. In chapter 16, Edmund Phelps (recent Nobel Prize winner in Economics and Guillermo’s colleague at Columbia University for many years) first reminisces about the years at Columbia (where a macroeconomic ‘‘dream team’’ had assembled). In chapter 17, Roque Fernandez (former Central Bank President and Finance Minister in Argentina) reveals Guillermo’s advisory role at a critical juncture in Argentina’s recent economic history. Finally, in chapter 18, Enrique Mendoza conducts a probing interview with Guillermo, which offers an incisive and penetrating look into Guillermo’s intellectual evolution and his views about the major ideas that have shaped the profession, both academically and from a public policy point of view.
I
Monetary and Exchange Rate Policy in Theory
1
Pricing-to-Market, the InterestRate Rule, and the Exchange Rate Maurice Obstfeld
The beauty of economics as an intellectual pursuit is its position at the intersection of formal theory, statistical analysis, and human events—coupled with its ultimate potential to improve peoples’ every day lives. A master economist must assume away the distracting inessential details of a situation in the interest of mathematical clarity. At the same time, he or she must see how relevant subtleties may affect the interpretation of data and the applicability of different models in the real world. Because the policy decisions at stake are so complex, with such vast potential to do harm or good, economics (and especially macroeconomics) is perpetually unsettled, subject to constant questioning and reassessment. Guillermo Calvo must be ranked as one of the great masters of economics, and one of the most unsettling. Time and again, starting with his classic work on the dynamic inconsistency problem, he has uncovered the hidden crux of a major scientific issue and forced the profession to rethink conventional beliefs. One area that has undergone extensive rethinking of late is the classic Milton Friedman case for exchange-rate flexibility, according to which floating exchange rates are helpful in cushioning national economies from real idiosyncratic shocks. Paradoxically, one influential assault on the Friedman case originates in the idea that the pass-through of exchange-rate changes to domestic prices may be high, another in the idea that pass-through may be very low. According to Calvo and Reinhart (2002), one factor behind the reluctance of emerging markets to allow large swings in nominally floating exchange rates is a relatively rapid pass-through of those swings to consumer prices. Rapid passthrough of this kind has two implications. Exchange-rate changes will have a greater potential in the short term to affect domestic inflation, and thereby to impede the pursuit of an inflation target. At the same time, rapid pass-through to all the prices consumers face blunts the exchange rate’s impact on international relative prices, and thereby reduces its potential expenditure-switching effects. On both counts, the costs of exchange-rate volatility are higher in emerging markets compared with the benefits that Friedman claimed.
4
Maurice Obstfeld
Another way the expenditure-switching effects of the exchange rate can be eliminated is if there is zero pass-through—both to domestic and import prices. This is the polar opposite of the case Calvo and Reinhart emphasize, but it would prevail if both domestic firms and foreign producers of a country’s imports were to preset their prices in terms of the local currency. Devereux and Engel (2003) analyze a formal model that includes this type of local-currency pricing. In their model, they show that welfare-maximizing monetary policies may entail fixed exchange rates. This theoretical analysis is viewed as a major challenge to the Friedman case, and one that is applicable to industrial rather than emerging economies.1 A foreign-based exporter presetting its price in an emerging-market currency would implicitly be acquiring a contingent asset denominated in that currency while issuing a contingent liability denominated in goods. This practice would therefore contradict the observation of ‘‘original sin,’’ which restricts emerging borrowers to issuing liabilities indexed to international currencies.2 As a result, local-currency pricing of imports is not expected to characterize emerging economies. The modest goal of the present paper is to reinstate Friedman’s case in the industrial-country setting while retaining the Devereux-Engel local-currency pricing framework.3 A minor modification of the Devereux-Engel model—the introduction of nontraded goods—is enough to restore the need for exchange flexibility, even when all shocks are real. In my modified model, exchange-rate changes still have absolutely no expenditure-switching effects in goods markets. They are necessary, however, to allow countries to pursue independent interestrate policies in a world of international capital mobility. That is, the rationale for exchange flexibility does not necessarily originate in goods markets, as in Friedman, but in asset markets. Divergent interest-rate movements may be needed, in turn, to support the divergent consumption movements implied by idiosyncratic national technology shocks in the presence of nontraded goods. Exchange-rate movements can also enhance risk sharing when there are nontraded goods. A byproduct of my argument is an analysis of equilibria and optimal policies in terms of interest-rate rules rather than money-supply rules—instead, the money supply is endogenous below. 1.1
The Model
I adopt the basic setup outlined by Devereux and Engel (2003) but modify it in two ways. First, I model monetary policy as a choice by the central bank of the nominal interest rate (rather than a monetary aggregate). Second, and more importantly, I introduce nontraded goods in order to illustrate the scope for an independent interest-rate policy in the Devereux-Engel local-currency pricing (LCP)
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate
5
framework. What is the intuition for this last effect? With nontraded goods and flexible prices, a national productivity shock has a disproportionate effect on home consumption, introducing an ex post asymmetry between countries. To mimic this response under sticky prices—thereby achieving the best possible (second-best) ex post allocation—authorities must apply a disproportionate interest-rate stimulus in their domestic economies. The basic setup of the model is as follows: there are two (ex ante) symmetrical countries, Home and Foreign. Each country produces a continuum of tradable goods (Home’s are indexed by ½0; 1Þ, Foreign’s by ½1; 2) and a continuum of nontradable goods (indexed by ½0; 1Þ). Each Home representative agent is an atomistic yeoman producer of one differentiated tradable good i and one differentiated nontradable good i, and also supplies labor.4 The producer of generic goods i maximizes ( " #) 1r y X t ðCt ðiÞÞ kLt ðiÞ ; U0 ðiÞ ¼ E0 b 1r t¼0 where C is a consumption index, L is labor supply, r > 0 and b A ð0; 1Þ. Because of the assumption that the monetary instrument is the nominal interest rate, and that the money supply adjusts endogenously, there is no need to model explicitly the demand for money (see Woodford 2003), and I will assume that any moneydemand effects on welfare are negligibly small. A critical assumption in the model is that of market segmentation between Home and Foreign. A Home producer of tradables can practice third-degree price discrimination, charging distinct same-currency prices in the Home and Foreign markets. By assumption, Home and Foreign consumers (who are also producers of other goods) face prohibitively high costs of arbitraging the resulting international price differentials. Let Qt be nominal marketable wealth at the start of period t, Pt the nominal price of consumption during the period, Tt transfer payments from the government, and Rtþ1 the nominal ex post return on the agent’s portfolio. Furthermore, let Yj ðiÞ be the level of output that Home producer i supplies to the Home tradables market ð j ¼ hÞ and to the Home nontradables market ð j ¼ nÞ; let Pj ðiÞ be the corresponding domestic-currency price charged. To the Foreign market, Home producer i supplies Yh ðiÞ at Foreign-currency price Ph ðiÞ. Then the flow budget constraint for producer i is Qtþ1 ðiÞ ð1 þ Rtþ1 ðiÞÞQt ðiÞ ¼ Ph; t ðiÞYh; t ðiÞ þ Et Ph; t ðiÞYh; t ðiÞ þ Pn; t ðiÞYn; t ðiÞ þ Tt Pt Ct ðiÞ;
ð1:1Þ
6
Maurice Obstfeld
where E is the Home-currency price of Foreign currency (the nominal exchange rate). There are isomorphic intertemporal maximization problems for the Foreign agents (whose supplies are denoted by asterisks). To maximize utility, each producer must grasp the nature of Home and Foreign demand, which depend in turn on the form of the consumption index. As in Obstfeld and Rogoff (2000), overall consumption depends on consumption of tradables and nontradables, C¼
Ctg Cn1g g g ð1 gÞ 1g
;
where the tradables subindex depends on consumption of Home- and Foreignproduced tradables, Ct ¼ 2Ch1=2 Cf1=2 : In turn, Ch , Cf , and Cn are CES functions of the available varieties, Cj ¼
ð 1
Cj ðiÞ
ðy1Þ=y
y=ðy1Þ di
;
0
with substitution elasticity y.5 Based on these assumptions, demands for the goods produced by individual i take the forms Ch ðiÞ ¼
g Ph ðiÞ y Ph 1 Pt 1 C; Pt P 2 Ph
Ch ðiÞ ¼
g Ph ðiÞ y Ph 1 Pt 1 C ; Pt P 2 Ph
Cn ðiÞ ¼ ð1 gÞ
Pn ðiÞ Pn
y
Pn P
and
1 C:
The exact price indexes entering the price indexes that Home consumers face are defined as follows: P ¼ Ptg Pn1g ; Pt ¼ Ph1=2 Pf1=2 ;
and
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate
(Ð1 ½ Pj ðiÞ 1y di1=ð1yÞ ; Pj ¼ Ð 02 ½ 1 Pj ðiÞ 1y di1=ð1yÞ ;
7
j ¼ h; n; j ¼ f:
There are complete markets in claims on future money payments. As in Backus and Smith (1993), and given the ex ante symmetry of the two countries, the resulting ex post allocation satisfies the condition Cr ðC Þr t ¼ t Pt Et Pt
ð1:2Þ
in all dates and states, where C is Foreign consumption and P is the Foreign price level measured in Foreign currency. Since purchasing power parity need not hold ex post in this model, the preceding condition does not generally equalize marginal utilities of consumption internationally. Production functions for every variety are given by the generic form Yh ¼ ALh ;
Yn ¼ ALn
in Home and Yf ¼ A Lf ;
Yn ¼ A Ln
in Foreign, so that A and A are economy-wide productivity shocks. Letting lowercase letters (except for interest rates) denote natural logarithms, I write the stochastic processes followed by the productivity shocks as at ¼ lat1 þ ut ;
at ¼ lat1 þ ut ;
ð1:3Þ
where l A ½0; 1 and the shocks u and u are normally distributed with means of zero and a common variance of su2 . Finally, the economy’s nominal anchor is provided by the nominal interest-rate setting rule followed by the central bank, logð1 þ it Þ ¼ i þ cpt ah ut ah ut ;
ð1:4Þ
where it is the nominal interest earned between dates t and t þ 1. (See also Benigno and Benigno 2008.) Foreign’s central bank has a corresponding rule, logð1 þ it Þ ¼ i þ cpt af ut af ut :
ð1:5Þ
It would be possible to add ‘‘noise’’ to these reaction functions by positing that central banks observe with error some of the variables to which they respond, but I will not pursue that generalization at this stage.
8
1.2
Maurice Obstfeld
The Flexible-Price Equilibrium
Consider next the model’s equilibrium when all prices are flexible and the central banks do not respond to productivity innovations (that is, the a coefficients in the interest-rate rules are all zero). Under flexible prices, producers set domesticmoney prices at a fixed gross markup, y=ðy 1Þ, over nominal marginal cost (equal to W=A in Home and W =A in Foreign), where W and W are the Home and Foreign nominal wage rates. Using the conditions for the optimal laborconsumption tradeoff, W r W C ¼ k ¼ ðC Þr ; P P
ð1:6Þ
along with the price-index definitions, one can derive the flex-price levels of overall consumption for the two countries: C¼
y 1 1g=2 g=2 1=r ðA Þ ; A yk
1=r y1 1g=2 g=2 C ¼ A : ðA Þ yk
ð1:7Þ
Observe that C ¼ C always in the flex-price equilibrium when all goods are tradable (that is, when g ¼ 1). But the equality need not hold when g < 1. In the latter case, a country’s flex-price consumption depends disproportionately on its own productivity shock. The reason is simple: that shock affects the nontradable as well as the tradable sector. The formulas in equation 1.7 suggest already that a differential response of national interest rates to global and national productivity shocks—and hence, exchange-rate flexibility—will be necessary under sticky prices to mimic the flexible-price consumption responses to productivity shocks. Using the price-index definitions and equation 1.6, one can also establish that in the flex-price equilibrium, Ph A Ph ¼ : ¼ Pf A Pf Thus, despite discriminatory price setting, in this flexible-price setting Home and Foreign consumers face the same international relative prices in equilibrium, and the global allocation of resources is constrained efficient (subject to monopoly and the nontradability of nontradable goods). Of course, Pn ¼ Ph , and the same is true in Foreign. Equilibrium real interest rates must be consistent with the path of expected consumption growth described by equations 1.3 and 1.7. Nominal interest rates and
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate
9
the resulting path for the overall money price level must, in turn, be consistent with the required path of equilibrium real interest rates. Nominal interest rates have their relevant impact on the economy through the intertemporal Euler equation for nominal bond holdings, r C Cr t ¼ ð1 þ it ÞbEt tþ1 : Pt Ptþ1
ð1:8Þ
(There is a parallel equation for Foreign.) Taking logs of the preceding equality and noting that consumption is lognormally distributed, I derive rct pt ¼ logð1 þ it Þ þ log b rEt ctþ1 Et ptþ1 þ
r2 2 1 2 s þ s þ rscp : 2 c 2 p
The variances above are endogenous, but because they will be constant over time, it is simple to compute them once we have solutions for the equilibrium levels of c and p in terms of current shocks and the means and variances of future variables. After substituting the policy rule in equation 1.4 for logð1 þ it Þ above, we obtain a difference equation with the unique stable price-level solution: y X 1 sþ1t 1 r2 2 1 2 pt ¼ log b þ i þ sc þ sp þ rscp : rðEt fcsþ1 cs gÞ 2 1þc c 2 s¼t
ð1:9Þ
Above, consumption can be expressed in terms of the underlying technology shocks using equation 1.7, allowing one to directly compute the equilibrium values of sc2 , sp2 , and scp . This computation is straight-forward and I therefore omit it here. At this point I observe only that higher current and expected future consumption growth rates are associated with a higher price level today. The reason is that higher consumption growth requires higher real interest rates. A higher price level raises the real interest rate through a policy channel—a higher nominal interest rate—and through a lowering of inflation expectations. Once the nominal price levels P and P are determined, nominal wages and the nominal exchange rate, which is given by E ¼ W=W in this model, are likewise pinned down. 1.3
The Model with Preset Nominal Prices
In the sticky-price version of the model, producers of tradables set their domestic and export prices one period in advance of sales, and must meet all demand that materializes at that price. Prices can be reset fully after one period, but again must be maintained for a period thereafter. Exporters set prices in the purchaser’s currency—there is LCP as in Devereux and Engel (2003). Nontradables producers
10
Maurice Obstfeld
simply set prices in their respective domestic currencies. While these price dynamics would be oversimplified for many purposes, they do allow us to consider the qualitative stabilization roles of interest and exchange rates in a usefully transparent setting. Let’s consider the price-setting problem of a generic Home producer i who sets prices for date t on date t 1. Because the decision has no repercussions beyond date t, we may imagine that the producer chooses prices Ph; t , Ph; t , and Pn; t so as to maximize ( ) Ct ðiÞ 1r Et1 kLt ðiÞ 1r subject to equation 1.1, Lt ðiÞ ¼
Yh; t ðiÞ þ Yh ; t ðiÞ þ Yn; t ðiÞ ; At
and the demand equations Yh; t ðiÞ ¼
Yh; t ðiÞ
g Ph; t ðiÞ y Ph; t 1 Pt; t 1 Ct ; Pt; t Pt 2 Ph; t
g Ph; t ðiÞ ¼ 2 Ph; t
Yn; t ðiÞ ¼ ð1 gÞ
!y
Ph; t Pt; t
!1
Pt; t Pt
1
Ct ;
and
Pn ðiÞ y Pn 1 Ct : P Pn
In performing this maximization the atomistic producer takes macro quantities and price indexes (not indexed by i above) as given. Because all domestic prices are preset and, I assume, known as of date t 1, the first-order condition for Ph; t ðiÞ, for example, is Ph; t ðiÞ ¼
y kPt Et1 fCt =At g : y 1 Et1 fCt ðiÞr Ct g
(I am making an assumption of common knowledge on the part of price setters.) All producers are symmetric, so Ct ðiÞ ¼ Ct in equilibrium, for all i, and therefore all set their Home tradable prices equal to Ph; t ¼
y kPt Et1 fCt =At g : y1 Et1 fCt1r g
ð1:10Þ
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate
11
Similar reasoning leads to the Home firms’ pricing formulas for exports and nontradables, respectively: Ph; t ¼
y kPt Et1 fCt =At g y kP Et1 fCt =At g ; t ¼ r y 1 Et1 fEt Ct Ct g y 1 Et1 fðCt Þ 1r g
ð1:11Þ
Pn; t ¼
y kPt Et1 fCt =At g ¼ Ph; t : y1 Et1 fCt1r g
ð1:12Þ
The second equality in (1.11) above is derived using equation 1.2. These three pricing equations have isomorphic counterparts for Foreign producers. From equations 1.10–1.12, the relative tradables prices Home consumers face are Ph; t Et1 fCt =At g : ¼ Pf; t Et1 fCt =At g (Recall that Pf; t is preset, in domestic-currency terms, by Foreign exporters to Home.) Correspondingly Foreigners face relative tradables prices Ph; Et1 fCt =At g t : ¼ Pf; t Et1 fCt =At g
The pricing formulas, equations 1.10–1.12, also yield useful information about expected consumption levels. Because Pn ¼ Ph , the overall price level is P ¼ 1ðg=2Þ g=2 Ph Pf . Thus, 1.10 can be written as Ph; t g=2 y kEt1 fCt =At g ¼ : Pf; t y 1 Et1 fCt1r g
ð1:13Þ
Likewise, using the formula for the Foreign exporter’s price, Pf , one finds that
Pf; t Ph; t
1g=2 ¼
y kEt1 fCt =At g : y 1 Et1 fCt1r g
ð1:14Þ
Combining the last two expressions to eliminate relative prices yields
yk 1¼ y1
1=gð2gÞ
ðEt1 fCt =At gÞ1=2g ðEt1 fCt =At gÞ1=2ð2gÞ ðEt1 fCt1r gÞ1=gð2gÞ
:
One similarly derives the corresponding expression involving expected Foreign consumption,
12
Maurice Obstfeld
yk 1¼ y1
1=gð2gÞ
ðEt1 fCt =At gÞ1=2g ðEt1 fCt =At gÞ1=2ð2gÞ ðEt1 fðCt Þ 1r gÞ1=gð2gÞ
:
Lognormality implies the following solution for the expected logarithm of Home consumption: Et ctþ1
þ scu 1 y1 gð2 gÞ Et atþ1 þ scu Et atþ1 su2 r 2 þ 1 sc ¼ log þ 2g 2ð2 gÞ 2r r yk r 2 ð1:15Þ
(with a parallel solution for Et ctþ1 ). This expression is a critical ingredient in the welfare analysis of monetary policies, because it contributes (via the consumption Euler equation, 1.8) to the contemporaneous innovation in consumption, ct , and hence to the variance of consumption and its covariance with technology shocks. In the present setting the overall price level is known a period in advance, so the log of the Euler equation 1.8 is
ct ¼ Et ctþ1
1 r2 log b þ logð1 þ it Þ ð ptþ1 pt Þ þ sc2 : 2 r
ð1:16Þ
To solve for the price level now, substitute the interest-rate rule, equation 1.4, into 1.16, and take date t 1 expectations to derive a difference equation for pt ¼ Et1 pt , pt ¼
1 r2 : Et1 ptþ1 þ rðEt1 ctþ1 Et1 ct Þ log b þ i þ sc2 2 1þc
I am now allowing the a coefficients in the interest-rate rule to differ from zero, but because the price level for date t is determined a period earlier, the a values do not enter its solution, which is pt ¼
y X 1 sþ1t 1 r2 log b þ i þ sc2 : rðEt1 fcsþ1 cs gÞ 2 1þc c s¼t
This is the natural extension of equation 1.9 to the case in which pt is a function only of information dated t 1 or earlier. Using 1.3 and 1.15 to substitute for the expected consumption terms above, one finds that pt ¼
glð2 gÞðl 1Þ at1 at1 1 r2 þ log b þ i þ sc2 : 2g 2ð2 gÞ 2 1þcl c
ð1:17Þ
A complete solution of the model requires an expression for realized consumption, ct , in terms of the date-t shocks. That expression, in turn, allows computation
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate
13
of the equilibrium values of the key moments sc2 , scu , and scu . Combination of equation 1.16 with 1.3, 1.4, 1.15, and 1.17 yields cl 2g g 1 ct ¼ at þ at þ ðah ut þ ah ut Þ þ W; ð1:18Þ 1þcl 2r 2r r where W is a function only of date t 1 (or earlier) information. Equations 1.7 and 1.18 disclose the key difference in consumption dynamics between the flexible and fixed-price cases. In the flexible-price case, assuming that ah ¼ ah ¼ 0, the responses of consumption to technology shocks are given by dct 2 g ; ¼ dat 2r
dct g ¼ : dat 2r
ð1:19Þ
With sticky prices, however, 1.18 shows that the responses of consumption are muted whenever l < 1. Why? For l ¼ 1, technology follows a random walk and so does log consumption; according to equation 1.18, current consumption therefore can adjust fully with no change in the real rate of interest. When l < 1, however, consumption is mean-reverting, and current consumption can adjust to its flex-price level only if the real interest rate falls. In the flexible-price case, pt indeed does fall, creating a lower real interest rate both through higher expected inflation and through the associated policy-induced fall in the nominal interest rate it . In contrast, if pt is rigid in the short run, the required real interest rate response is muted and so is the rise in ct . By appropriate choice of the policy response coefficients ah and ah in equation 1.4, however, the central bank can induce the full flex-price consumption responses, and I show later that it will wish to do so. That result also holds in the Devereux-Engel (2003) model with no nontraded goods, as the authors show. Because g ¼ 1 in their setting, however, flex-price consumption responses to technology shocks are symmetrical, and so central banks’ policy responses are absolutely symmetrical as well. That is not the case when g < 1, for then a relatively more forceful Home interest rate intervention is needed to mimic the flexible-price consumption response. Variable international interest-rate differentials imply exchange-rate variation, however, even though the exchange rate has no expenditure switching effects between Home and Foreign goods in this model. As a last step before a formal welfare analysis of policy rules, I derive the endogenous covariances entering into the model. To simplify the algebra, let us assume that the productivity shocks are independent, so that suu ¼ 0. From equation 1.18, ( ) cl 2g ah 2 cl g ah 2 2 2 þ þ þ ð1:20Þ su ; sc ¼ r r 1þcl 2r 1 þ c l 2r
14
Maurice Obstfeld
scu ¼
cl 1þcl
2g ah 2 s ; þ r u 2r
ð1:21Þ
and scu ¼
cl 1þcl
g a þ h su2 : 2r r
ð1:22Þ
The expressions involving Foreign consumption c are analogous but involve the Foreign interest-rate policy coefficients, af and af . Those coefficients do not enter the Home covariances because of the highly insulating role of exchange-rate changes in this particular LCP setting. Exchange-rate changes merely facilitate independent monetary policies in a world of international capital mobility. 1.4
Welfare and Optimal Monetary Policy Rules
To assess welfare, observe that the Home labor supply must be consistent with Et Ltþ1
g ¼ 1 2 g ¼ 1 2
Ph; tþ1 Ptþ1
1 1 C Ctþ1 g Ph; tþ1 þ Et Et tþ1 Atþ1 Atþ1 2 Ptþ1
Pf; tþ1 Ph; tþ1
!1g=2 g=2 Ctþ1 Ctþ1 g Pf; tþ1 þ : Et E t Atþ1 Atþ1 2 Ph; tþ1
Using equation 1.13 and the Foreign analog of 1.14 to eliminate the relative price terms above, one finds that y1 g g 1r Et Ltþ1 ¼ 1 Et fCtþ1 g þ Et fðCtþ1 Þ 1r g : yk 2 2 As a result, period Home expected utility can be written as ( Et
1r Ctþ1 kLtþ1 1r
)
(
) 1r Ctþ1 y1 g g 1r 1r 1 Et fCtþ1 g þ Et fðCtþ1 Þ g ¼ Et 1r y 2 2 y 1 2g ðy 1Þð1 rÞ g y1 1r Þ 1r g ¼ Et fðCtþ1 Et fCtþ1 g 2 y ð1 rÞy ( ) y 1 2g ðy 1Þð1 rÞ ð1 rÞ 2 2 sc ¼ exp ð1 rÞEt ctþ1 þ 2 yð1 rÞ ( ) g y1 ð1 rÞ 2 2 sc : exp ð1 rÞEt ctþ1 þ 2 2 y
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate
15
I have already noted that the distribution of Foreign consumption, C , does not depend on the Home interest-rate rule. Therefore, in considering Home’s optimal interest-rate rule, I need only consider maximization of the first summand in the last equation with respect to the feedback coefficients ah and ah . Moreover, using equation 1.15, it is sufficient to maximize þ scu ð1 rÞ 2 1 y1 gð2 gÞ Et atþ1 þ scu Et atþ1 Et ctþ1 þ þ sc ¼ log þ 2g 2ð2 gÞ 2 r yk r
su2 sc2 ; 2r 2
or even more simply, to maximize V1
ð2 gÞscu gscu sc2 þ : 2r 2r 2
Equations 1.20–1.22 imply that cl 2g cl g V z ð2 gÞ þ ah þ g þ ah 1þcl 2 1þcl 2 ( 2 2 ) cl 2g cl g þ ah þ þ ah : 1þcl 2 1þcl 2 Maximization with respect to the policy parameters yields the procyclical (positive) responses 2g cl ah ¼ 1 and 2 1þcl g cl ah ¼ 1 : 2 1þcl The Foreign response coefficients in 1.5 are analogous,6 with g cl af ¼ 1 and 2 1þcl 2g cl 1 : af ¼ 2 1þcl A comparison of equations 1.18 and 1.19 reveals that these policy responses yield a consumption response to innovations in technology that is identical to the
16
Maurice Obstfeld
flex-price response. They make the variance of consumption equal to its flex-price variance, and induce the flex-price covariances with the shocks to technology. But interest-rate intervention alone cannot bring the world economy to the flex-price consumption levels. Policy optimization thus yields a strictly second-best allocation, with welfare below the flexible-price level. In particular, as Devereux and Engel (2003) note, consumers will in general face the wrong relative prices in the preset-price equilibrium, prices that do not reflect true levels of relative economic scarcity. 1.5
The Need for Exchange-Rate Flexibility
A key point about the preceding second-best interest-rate rules is that they predict asymmetric national responses to technology shocks—except in the special case of no nontradables ðg ¼ 1Þ that Devereux and Engel analyze. A useful way to think about this asymmetry is to define the mutually orthogonal global and idiosyncratic shocks uw 1
u þ u 2
ud 1
u u : 2
and
Then one can express the second-best interest-rate rules for Home and Foreign, respectively, in the simple forms cl cl logð1 þ it Þ ¼ i þ cpt 1 uw; t ð1 gÞ 1 ud; t and 1þcl 1þcl cl cl logð1 þ it Þ ¼ i þ cpt 1 uw; t þ ð1 gÞ 1 ud; t : 1þcl 1þcl The countries respond identically to the global shock in all cases, but have oppositely signed responses to the idiosyncratic shock when there are nontradable goods and, consequently, g < 1. I noted the intuition for this result in the introduction: when g ¼ 1, productivity shocks in either country have perfectly symmetrical consumption effects in the flex-price equilibrium, so internationally symmetrical interest-rate responses always suffice to induce the flex-price response to any shock. That is, when all goods are tradable, it is optimal for central banks to respond only to global shocks and to respond with equal interest-rate changes. Nontradables change this. Because a domestic technology shock has a stronger effect on domestic than on foreign consumption, a relatively more forceful domestic interest-rate response may be required.
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate
17
This asymmetry has implications for exchange rates because, given an interestrate parity condition, divergent interest-rate movements will call for exchangerate changes. To see this formally, observe that the Home and Foreign bond Euler equations 1.8 of a Home investor may be combined to yield the exchangerate equation Et ¼
r 1 þ it Et fEtþ1 Ctþ1 g : 1 þ it Et fCr tþ1 g
After taking logs and substituting the optimal interest-rate rules, the international risk-sharing condition, and the equations 1.18 for ex post consumption levels, one concludes that, apart from additive predetermined terms, the log exchange rate under optimal monetary policies is given by cl 2ð1 gÞc cl Et etþ1 et ¼ 2ð1 gÞ 1 : ud; t þ ad; t þ 1þc 1þcl 1þc 1þcl This expression makes it clear that idiosyncratic technology shocks will induce exchange-rate movements through the asymmetric response of consumption, something that does not occur in this model when g ¼ 1 and all goods are tradable.7 It is not an expenditure-switching function of exchange rates in commodity markets that gives them a role in optimal second-best monetary policies. Instead, the rationale for exchange-rate flexibility lies in the asset markets. Exchange-rate adjustment makes room for expenditure-changing interest-rate policies, and it does so by offsetting the incipient expected return differentials that divergent interest-rate movements would otherwise cause. Exchange-rate movements can also enhance risk sharing. To enjoy the benefits of both activist monetary policy and open capital markets, governments must allow the exchange rate to move. 1.6
Conclusion
Even when the exchange rate plays no expenditure-switching role, as in the Devereux-Engel model, countries may wish to have flexible exchange rates in order to free the domestic interest rate as a stabilization tool. This can be true even when all shocks are real. Why does no need for exchange flexibility arise in the Devereux-Engel model with exclusively tradable goods? There, national consumptions move in a perfectly synchronized fashion when all shocks are real, whether prices are flexible or preset. Optimal monetary policy simply raises the stickyprice consumption response to its flexible-price level, a job that can be accomplished through globally symmetric monetary policies that maintain asset-market equilibrium without the need for exchange-rate changes. In contrast, nontraded
18
Maurice Obstfeld
goods make national consumption responses to asymmetric real shocks asymmetric themselves. In that case, optimal monetary policy requires a relatively greater monetary stimulus in the country experiencing the shock, and very possibly, a change in the international nominal interest-rate differential and in the exchange rate. These results may strike the reader as abstract, but they are at the crux of monetary policy decisions and institutions with first-order impacts on peoples’ welfare. Guillermo Calvo has repeatedly uncovered the connections between the seemingly abstruse theorem and the strikingly relevant policy conclusion. As John Maynard Keynes famously put it, ‘‘The master-economist . . . must be mathematician, historian, statesman, philosopher. . . . He must understand symbols and speak in words. . . . He must study the present in the light of the past for the purposes of the future.’’8 Even by Keynes’s exacting standard, Calvo easily qualifies as one of the great master economists of our day. Notes Prepared for the International Monetary Fund conference in honor of Guillermo A. Calvo, April 15–16, 2004. I thank Ricardo Caballero, Mathias Hoffmann, Lorenz Ku¨ng, and conference participants for helpful comments. 1. See Engel (2002) for further discussion. Obstfeld and Rogoff (2000) argue that while the industrialcountry retail or consumer prices of imports indeed appear sticky in domestic currency, wholesale import prices do exhibit some exchange-rate pass-through. 2. Eichengreen and Hausmann (1999). Calvo and Reinhart (2002) identify the widespread foreigncurrency denomination of liabilities as another key factor behind ‘‘fear of floating.’’ 3. After writing this paper, I learned that Duarte (2004) had independently made essentially the same point about the need for exchange-rate flexibility (albeit in a somewhat different setting). See also Duarte and Obstfeld (2008). These two papers are based on money supply rather than nominal interest rate policy feedback rules. On some implications of this difference, see note 7 below. 4. Foreign producer i supplies tradable 1 þ i and nontradable i. 5. For j ¼ f, the integration is over the interval ½1; 2. 6. It is easily verified that these coefficients also maximize the equal-weights ‘‘world planner’’ welfare function 12 U þ 12 U , as Devereux and Engel (2003) also find. Thus, the Nash equilibrium in a policyrule-setting game between the countries is efficient—there is no coordination failure in this model, though that is a model-specific result (Obstfeld and Rogoff 2002, Benigno and Benigno 2003). It may also be checked that at this optimum, the value that c (the response to the price level) takes in ð0; yÞ is irrelevant for welfare. That is, the optimal choice of the a coefficients fully offsets any welfare effect of c. 7. Notice from the exchange-rate equation that even when l ¼ 1, and productivity shocks therefore are permanent, the exchange rate will move in response to ad; t ¼ ad; t1 þ uD; t . When l ¼ 1, the difference equation governing the exchange rate reduces to the simpler form et ¼
2ð1 gÞc Et etþ1 ; ad; t þ 1þc 1þc
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate
19
which has the standard no-bubbles solution. The exchange rate can change even at an unchanged international interest-rate differential because relative money supplies adjust endogenously, essentially to mimic the monetary rule given by Duarte and Obstfeld (2008). In the case l ¼ 1, as shown earlier, there is no need for an interest-rate change to produce the flexible-price response of consumption. The latter is automatic. Because overall consumer price levels are rigid, however, a globally asymmetric consumption response implies that an exchange-rate change is still needed to maintain the Backus-Smith risk sharing conditions. In the Devereux-Engel (2003) paper all shocks are permanent. In the absence of an appropriate feedback policy rule, however, consumption responses may not equal flexible-price responses, in contrast to the finding of the present paper. The reason is that it is the money supply rather than the nominal interest rate that is the policy instrument in the analysis of Devereux and Engel, and the endogenous adjustment of the interest rate, given money supplies, influences the size of the consumption response. 8. Keynes 1963, p. 141.
References Backus, David K., and Gregor W. Smith. 1993. ‘‘Consumption and Real Exchange Rates in Dynamic Economies with Non-Traded Goods.’’ Journal of International Economics 35: 297–316. Benigno, Gianluca, and Pierpaolo Benigno. 2003. ‘‘Price Stability in Open Economies.’’ Review of Economic Studies 70: 743–764. ———. 2008. ‘‘Exchange Rate Determination under Interest Rate Rules.’’ Journal of International Money and Finance 27, in press. Calvo, Guillermo A., and Carmen M. Reinhart. 2002. ‘‘Fear of Floating.’’ Quarterly Journal of Economics 117: 379–408. Devereux, Michael B., and Charles Engel. 2003. ‘‘Monetary Policy in the Open Economy Revisited: Price Setting and Exchange-Rate Flexibility.’’ Review of Economic Studies 70: 765–783. Duarte, Margarida. 2004. ‘‘Monetary Policy and the Adjustment to Country-Specific Shocks.’’ Federal Reserve Bank of Richmond Economic Quarterly 90: 21–40. Duarte, Margarida, and Maurice Obstfeld. 2008. ‘‘Monetary Policy in the Open Economy Revisited: The Case for Exchange-Rate Flexibility Restored.’’ Journal of International Money and Finance 27, in press. Eichengreen, Barry, and Ricardo Hausmann. 1999. ‘‘Exchange Rates and Financial Fragility.’’ In New Challenges for Monetary Policy, 329–368. Kansas City, MO: Federal Reserve Bank of Kansas City. Engel, Charles. 2002. ‘‘Expenditure Switching and Exchange-Rate Policy.’’ In NBER Macroeconomics Annual 2002, ed. Mark Gertler and Kenneth Rogoff, 231–272. Cambridge, MA: MIT Press. Keynes, John Maynard. 1963. ‘‘Alfred Marshall.’’ In Essays in Biography, 125–217. New York: W. W. Norton & Company. Obstfeld, Maurice, and Kenneth Rogoff. 2000. ‘‘New Directions for Stochastic Open Economy Models.’’ Journal of International Economics 50: 117–153. ———. 2002. ‘‘Global Implications of Self-Oriented National Monetary Rules.’’ Quarterly Journal of Economics 117: 503–535. Woodford, Michael. 2003. Interest and Prices. Princeton, NJ: Princeton University Press.
2
Optimal Exchange Rate Regimes: Turning MundellFleming’s Dictum on Its Head Amartya Lahiri, Rajesh Singh, and Carlos A. Ve´gh
2.1
Introduction
One of the most influential results in open economy macroeconomics—which derives from the Mundell-Fleming model (see Mundell 1968 and Fleming 1962)— holds that the choice of the exchange rate regime should depend on the type of shock hitting the economy. If shocks are predominantly of real origin, then flexible exchange rates are optimal. Instead, if shocks are mainly monetary, fixed (or, more generally, predetermined) exchange rates are optimal. In fact, as Calvo (1999, 4) aptly puts it, this is ‘‘a result that every well-trained economist carries on her tongue.’’ Calvo (1999) himself offers a simple derivation of this result in a model in which the policy maker’s objective is to minimize output variability. The intuition is simple enough: in the Mundell-Fleming world of sticky prices and perfect capital mobility, real shocks require an adjustment in relative prices which, in the presence of sticky prices, can most easily be effected through changes in the nominal exchange rate. In contrast, monetary shocks require an adjustment in real money balances that can be most easily carried out through changes in nominal money balances (which happens endogenously under fixed exchange rates). By and large, this key result has remained unscathed in modern variations of the Mundell-Fleming model. For instance, Cespedes, Chang, and Velasco (2004) incorporate liability dollarization and balance sheets effects and conclude that the standard prescription in favor of flexible exchange rates in response to real shocks is not essentially affected. But rather than tweaking at the margin with variations of the traditional Mundell-Fleming model, it could be argued that one should take issue with its most critical assumption: imperfection in goods markets (i.e., sticky prices) but undistorted capital markets (in other words, perfect capital mobility). Is this the world we necessarily live in? Far from it. In developing countries, in particular, asset market frictions appear to be equally, if not more, important than goods market frictions. In fact, a large segment of the population does not seem to have
22
Amartya Lahiri, Rajesh Singh, and Carlos A. Ve´gh
access to asset markets.1 In this light, it seems worth revisiting the MundellFleming question in a model with flexible prices but segmented asset markets. This type of model posits that while a fraction of the population (referred to as traders) has access to asset markets, the rest of the population (nontraders) does not. In an earlier paper (Lahiri, Singh, and Ve´gh 2007), we examined this issue in the context of a stochastic model in which traders have access to incomplete markets. In contrast, this paper develops a much starker, perfect-foresight version of the model that, by avoiding myriad technical complications, allows the essential mechanisms and intuition to shine through. The paper’s punchline is that—contrary to the Mundell-Fleming prescription mentioned above—if shocks are real, fixed exchange rates are optimal, whereas if shocks are monetary, flexible exchange rates are optimal. Intuitively, flexible exchange rates allow for a costless adjustment to monetary shocks by altering the real value of existing nominal money balances. In contrast, under fixed rates, asset-market segmentation prevents nontraders from rebalancing real money balances by accessing asset markets, which affects the consumption path. Under real shocks, fixed rates allow purchasing power to be transferred across periods, which results in some consumption smoothing. Under flexible rates, on the other hand, nontraders are forced to consume their current endowment. We thus conclude that the optimal exchange rate regime should depend not only on the type of shock (real versus monetary)—as rightly emphasized by Mundell-Fleming models—but also on the type of distortion (goods markets versus asset markets frictions).2 These ideas can be succinctly summarized in the following 2 2 matrix: Table 2.1 Optimal exchange rate regime
Real shock Monetary shock
Goods market friction
Asset market friction
Flexible Fixed
Fixed Flexible
The optimal exchange rate regime thus becomes an empirical issue that depends both on the type of shock hitting a particular economy and on the relative distortions present in goods and asset markets. The paper proceeds as follows. Section 2.2 develops the main model—a perfectforesight version of Lahiri, Singh, and Ve´gh (2007)—and solves it for the cases of both flexible and fixed exchange rates. Section 2.3 compares the two regimes for fluctuating output and velocity paths. Section 2.4 contains some brief concluding remarks. Some technical issues are relegated to appendices.
Optimal Exchange Rate Regimes
2.2
23
The Model
Consider a discrete-time model of a small open economy perfectly integrated with the rest of the world in goods markets. There are two types of agents: traders (who have access to capital markets) and nontraders (who do not have access to capital markets). The fraction of traders is l while that of nontraders is 1 l. There is no uncertainty in the model and agents are blessed with perfect foresight. The law of one price holds for the only good; hence, Pt ¼ Et Pt , where Pt is the nominal price of the good, E is the nominal exchange rate (in units of domestic currency per unit of foreign currency), and Pt is the foreign nominal price. Foreign inflation is assumed to be zero and, for simplicity, Pt is taken to be unity. Hence, Pt ¼ Et . Both traders and nontraders are subject to a cash-in-advance constraint. For the case of traders, we follow Lucas’s (1982) timing and assume that asset markets open first (say, in the morning) followed by goods markets (in the afternoon). By assumption, of course, nontraders do not have access to asset markets and hence only visit goods markets.3 There are two types of shocks: real and monetary. Both traders and nontraders face identical shocks. Real shocks are captured by fluctuations in the endowment of the only good, y. Following Alvarez, Lucas, and Weber (2001), we capture monetary—or velocity—shocks by allowing both traders and nontraders to access a fraction v of current period sales ðvt Pt yt Þ and letting vt fluctuate over time. To fix ideas, it proves useful to keep in mind the following scenario regarding the model’s timing conventions. Households consist of two individuals: a shopper and a seller. As is standard, households do not consume their own endowment. As goods markets are about to open in the afternoon, the seller and the shopper part and, in the standard model, do not see each other until the end of the day. In other words, the seller stays in the store selling the endowment to other households’ shoppers and the shopper visits other stores to purchase goods. In the standard model, then, the shopper does not return to the store until after goods markets close and therefore has no access to the money balances accrued to the seller from the sale of the current-period endowment ðPt yt Þ. In the current paper, we depart from the standard model by allowing the shopper to come back to the store once during the goods market session, empty the cash register, and go back to shopping. We assume that the amount of money in the cash register at the time the shopper comes back to the store is vt Pt yt , where 0 < v < 1. Finally, both traders (T) and nontraders (NT) have identical preferences, given by Ui ¼
y X
b t uðcti Þ;
i ¼ T; NT
t¼0
where c i denotes consumption of an agent of type i.
ð2:1Þ
Amartya Lahiri, Rajesh Singh, and Carlos A. Ve´gh
24
2.2.1
Nontraders
Nontraders do not have access to asset markets and hence hold only money. Their flow budget constraint is given by NT Mtþ1 ¼ MtNT þ Et yt Et ctNT ;
ð2:2Þ
where MtNT denotes end-of-period t 1 (and hence beginning of period t) nominal money balances in the hands of nontraders. The initial level of nominal money balances, M0NT , is given. Nontraders are subject to a cash-in-advance constraint of the form MtNT þ vt Et yt b Et ctNT :
ð2:3Þ
The nominal money balances that nontraders can use to purchase goods consist of the nominal money balances that they bring into period t, MtNT , and a fraction vt of current-period sales (recall that, by assumption, 0 < vt < 1). We will only consider equilibrium paths along which the cash-in-advance constraint binds.4 If the cash-in-advance binds, then we can solve for ctNT from equation 2.3 to obtain ctNT ¼
MtNT þ vt Et yt ; Et
t b 0:
ð2:4Þ
To find out how much money balances nontraders will carry on to the next period, substitute 2.4 into 2.2 to obtain NT Mtþ1 ¼ ð1 vt ÞEt yt :
ð2:5Þ
When the cash-in-advance constraint binds, the nontraders’ problem becomes completely mechanical. In other words, their opportunity set consists of only one point in every period—given by equation 2.4—and there is thus no need to carry out any maximization. Intuitively, nontraders begin their life with a given level of nominal money balances, M0NT . They augment these cash balances with a fraction v0 of period 0 sales, v0 E0 y0 . Since the cash-in-advance binds, they spend all of their money balances, M0NT þ v0 E0 y0 , on consumption in period 0. Their end-ofperiod cash balances consist of the cash proceeds from selling their endowment, E0 y0 , minus the amount of period 0 sales spent in period 0, v0 E0 y0 . They thus enter period 1 with M1NT ð¼ ð1 v0 ÞE0 y0 Þ and the process begins anew.
Optimal Exchange Rate Regimes
2.2.2
25
Traders
Traders have access to asset markets and thus behave like consumers in any standard model with perfect capital mobility. The only difference is that, like nontraders, they have access to a fraction vt of current-period sales. Let us first look at the flow constraint for the asset market. Traders enter the asset market with a certain amount of nominal money balances, MtT , and a certain amount of bonds, bt . Once in the asset market, they receive/pay interest on the bonds they carried into the asset markets, Et rbt ; receive transfers from the government, T; and buy/sell bonds in exchange for money.5 Traders exit the asset mar^ T of nominal money balances and btþ1 of bonds. The flow ket with a quantity M t constraint for the asset market is thus ^ T ¼ M T þ Et ð1 þ rÞbt þ Tt : Et btþ1 þ M t t l
ð2:6Þ
Traders are subject to a cash-in-advance constraint: ^ T þ vt E t yt b E t c T : M t t
ð2:7Þ
What will traders’ nominal money balances be at the end of period t? Traders will have the money brought from the asset market plus the proceeds from the sale of their endowment ðEt yt Þ minus the money balances used to purchase goods ðEt ct Þ: T ^ T þ Et yt Et c T : Mtþ1 ¼M t t
ð2:8Þ
By substituting 2.8 into 2.6, we obtain the traders’ flow constraint for period t as a whole: T Et btþ1 þ Mtþ1 ¼ MtT þ Et ð1 þ rÞbt þ Et yt þ
Tt Et ctT : l
ð2:9Þ
2.2.2.1 Utility Maximization For the purposes of the maximization—and by substituting equation 2.6 into 2.7—we can rewrite the cash-in-advance constraint as MtT þ Et ð1 þ rÞbt þ
Tt Et btþ1 þ vt Et yt b Et ctT : l
ð2:10Þ
Traders thus maximize lifetime utility subject to the flow budget constraint 2.9 and the cash-in-advance constraint 2.10, for given values of M0T and b0 . The Lagrangian is then given by
Amartya Lahiri, Rajesh Singh, and Carlos A. Ve´gh
26
y X
b t uðc T Þ þ
t¼0
þ
y X
b
t
Tt T b t ht MtT þ Et ð1 þ rÞbt þ þ Et yt Et ctT Et btþ1 Mtþ1 l t¼0
y X
Ct MtT
t¼0
Tt T þ Et ð1 þ rÞbt þ Et btþ1 þ vt Et yt Et ct ; l
where ht and Ct denote the multipliers associated with constraints 2.9 and 2.10, respectively. T The first-order conditions with respect to c T , Mtþ1 , and btþ1 are given, respectively, (assuming, as usual, that bð1 þ rÞ ¼ 1) by u 0 ðctT Þ ¼ Et ðht þ Ct Þ; ht ¼ bðhtþ1 þ Ctþ1 Þ;
ð2:11Þ and
Etþ1 ðhtþ1 þ Ctþ1 Þ ¼ Et ðht þ Ct Þ:
ð2:12Þ ð2:13Þ
The first-order condition with respect to ht naturally recovers the flow constraint 2.9. Finally, the Kuhn-Tucker condition for Ct recovers 2.10 and requires the complementary slackness condition Tt T T Mt þ Et ð1 þ rÞbt þ Et btþ1 Et ct Ct ¼ 0: ð2:14Þ l Combining first-order conditions 2.11 and 2.13 yields T u 0 ðctT Þ ¼ u 0 ðctþ1 Þ:
As in standard cash-in-advance models with Lucas’s (1982) timing, traders will fully smooth consumption over time. 1 Combining first-order conditions 2.12 and 2.13 (using b ¼ 1þr ) yields Etþ1 ht ð1 þ rÞ 1 ¼ Ct : ð2:15Þ Et Perfect capital mobility (for traders) implies that the interest parity condition holds: 1 þ it ¼ ð1 þ rÞ
Etþ1 ; Et
ð2:16Þ
which enables us to rewrite condition 2.15 as h t it ¼ Ct :
ð2:17Þ
Optimal Exchange Rate Regimes
27
Since, at an optimum, ht > 0, equation 2.17 says that if it > 0, then Ct > 0, which implies from the complementary slackness condition 2.14 that the cash-inadvance constraint binds. Since we will only consider equilibria in which the nominal interest rate is positive, the cash-in-advance constraint will always bind and traders’ end-of-period money balances can be obtained by combining 2.7 and 2.8:6 T Mtþ1 ¼ ð1 vt ÞEt yt :
2.2.3
ð2:18Þ
Government
The government’s flow constraint is given by Et htþ1 ¼ ð1 þ rÞEt ht þ Mtþ1 Mt Tt ;
ð2:19Þ
where ht denotes net foreign bonds held by the government. 2.2.4
Equilibrium Conditions
Money market equilibrium implies that Mt ¼ lMtT þ ð1 lÞMtNT :
ð2:20Þ
NT T Equations 2.5 and 2.18 imply that Mtþ1 ¼ Mtþ1 . Together with the money market equilibrium condition 2.20, this implies that
Mt ¼ MtNT ¼ MtT : Since there are no differences across agents in terms of the endowment, all agents hold the same amount of money (on a per-capita basis). Hence, 2.5 and 2.18 together with the money market equilibrium condition 2.20 yield a quantity theory equation: Mtþ1 ¼ ð1 vt ÞEt yt ;
t b 0:
ð2:21Þ
To make it directly comparable to the quantity theory equation found in textbooks (typically written as MV ¼ Py, where V denotes velocity), we can rewrite this last equation as Mtþ1 ¼ Et yt ; 1 vt
t b 0:
Amartya Lahiri, Rajesh Singh, and Carlos A. Ve´gh
28
Velocity is therefore given by 1=ð1 vt Þ. Thus, a higher v captures an increase in the velocity of circulation, which rationalizes our term ‘‘velocity shocks’’ when referring to changes in v. To obtain the economy’s flow constraint, multiply the nontraders’ flow constraint (equation 2.2) by 1 l and the traders’ flow constraint (equation 2.9) by l and then add them, taking into account the government’s flow constraint (equation 2.19) and the money market equilibrium condition (equation 2.20): ktþ1 kt ¼ rkt þ yt ½lctT þ ð1 lÞctNT ; where k 1 ht þ lbt denotes the economy’s per-capita net foreign assets. Iterating forward and imposing the transversality condition limt!y we obtain the resource constraint ð1 þ rÞk 0 þ
y X t¼0
b t yt ¼
y X
b t ½lctT þ ð1 lÞctNT :
ð2:22Þ ktþ1 ð1þrÞ t
¼ 0,
ð2:23Þ
t¼0
In what follows, we will assume that k 0 ¼ 0.7 2.2.5
Equilibrium Consumption
We will now derive expressions for consumption of both traders and nontraders. To obtain nontraders’ consumption, substitute the quantity theory equation 2.21 into 2.4 (recall that Mt ¼ MtNT ) to obtain 8 < M 0 þ v0 y0 ; t¼0 E0 ctNT ¼ ð1v ð2:24Þ ÞE y þv E y t1 t1 t1 t t t : ; t b 1: Et This expression will prove useful when dealing with fixed exchange rates. When dealing with flexible exchange rates, however, it will prove convenient to use 2.21 to rewrite 2.24 as ctNT ¼ yt
Mtþ1 Mt ; Et
t b 0:
ð2:25Þ
To obtain traders’ consumption, substitute 2.24 into 2.23 and solve for the constant level of c T , denoted by c T , to obtain " !# y X 1l p r M0 t ð1 vt1 ÞEt1 yt1 þ vt Et yt p T c ¼y þ þ v0 y0 þ b y ; Et l 1 þ r E0 t¼1 ð2:26Þ
Optimal Exchange Rate Regimes
29
where y p 1 ð1 bÞ
y X
b t yt
t¼0
denotes permanent income. Alternatively, substitute 2.25 into 2.23 and iterate to obtain cT
y r 1 lX t Mtþ1 Mt : ¼y þ b Et 1 þ r l t¼0 p
ð2:27Þ
Equations 2.25 and 2.27 make clear the redistributive role that monetary policy plays in this model. If, say, money supply is constant, then nontraders consume their endowment ðctNT ¼ yt Þ and traders their permanent income ðc T ¼ y p Þ. An increase in the money supply ðMtþ1 > Mt Þ implies a transfer from nontraders to traders. The reverse is true in the case of a reduction in the money supply. 2.2.6
Flexible Exchange Rates
Consider a flexible exchange rate regime in which the monetary authority sets a constant path of the nominal money supply:8 Mt ¼ M;
ð2:28Þ
t b 0:
Substituting 2.28 into 2.25, we obtain nontraders’ consumption: ctNT ¼ yt ;
ð2:29Þ
t b 0:
Two observations are worth making. First, consumption of nontraders will fluctuate one-to-one with fluctuations in the endowment. Flexible exchange rates provide no insulation whatsoever for nontraders from output fluctuations. Second, consumption of nontraders is not affected by velocity shocks. Substituting 2.28 into 2.27, we obtain traders’ consumption c T ¼ y p:
ð2:30Þ
Let us now derive the path of the nominal exchange rate. From the quantity theory equation 2.21, we obtain Et ¼
M ; ð1 vt Þyt
It follows that
t b 0:
ð2:31Þ
Amartya Lahiri, Rajesh Singh, and Carlos A. Ve´gh
30
Etþ1 ð1 vt Þ yt ¼ : Et ð1 vtþ1 Þ ytþ1
ð2:32Þ
When output increases (that is, ytþ1 > yt )—and for constant velocity—the nominal exchange rate will fall (i.e., the domestic currency appreciates). Intuitively, higher output increases real money demand and hence leads to a fall in the price level (in other words, in the nominal exchange rate). On the other hand, when there is an increase in velocity (i.e., vtþ1 > vt )—and for constant output—the nominal exchange rate will increase (i.e., the domestic currency depreciates). Intuitively, an increase in velocity implies that more money is available to purchase the same level of output, which will lead to a higher price level (that is, a higher nominal exchange rate). Finally, the path of the nominal interest rate follows from combining the interest parity condition 2.16 with 2.32: 1 þ it ¼ ð1 þ rÞ 2.2.7
ð1 vt Þ yt : ð1 vtþ1 Þ ytþ1
ð2:33Þ
Fixed Exchange Rates
Consider now a fixed exchange rate regime in which the monetary authority sets a constant value of the nominal exchange rate: Et ¼ E: To ensure that initial conditions under fixed rates are consistent with those under flexible rates (in the sense that they generate the same initial level of real-money balances as in the case of flexible exchange rates), we take initial nominal money holdings to be M0NT ¼ M0T ¼ M0 ¼ M. Further, we assume that the exchange rate is fixed at the level given by E ¼ M=ð1 v0 Þy0 . Under these assumptions, initial real money balances M0 =E are given by ð1 v0 Þy0 , as is the case under flexible rates (recall 2.31). Under a fixed exchange rate, we can use equation 2.24 to obtain consumption of nontraders: ( M0 þ v0 y0 ; t ¼ 0; ctNT ¼ E ð2:34Þ ð1 vt1 Þ yt1 þ vt yt ; t b 1: Since M0 =E ¼ ð1 v0 Þy0 , it follows that c0NT ¼ y0 . By the same token, using 2.26, consumption of traders is given by ( !) y X 1l r t p p T c ¼y þ b ½ð1 vt1 Þ yt1 þ vt yt : y y0 þ l 1þr t¼1
ð2:35Þ
Optimal Exchange Rate Regimes
31
Let us now derive the path of the nominal money supply, which is endogenous under fixed exchange rates. M0 ¼ M, as remarked earlier. The path of Mt for t b 1 then follows from the quantity theory equation 2.21: Mtþ1 ¼ ð1 vt ÞEyt ; 2.3
t b 0:
Comparing Flexible versus Fixed Exchange Rates
We are now ready to ask our main question: which exchange regime is better? 2.3.1
Velocity Shocks Only
Suppose that there are only velocity shocks (i.e., set yt ¼ y p ). Then, under flexible rates, consumption of nontraders is completely flat and equal to y p (as follows from equation 2.29). Further, as equation 2.30 indicates, traders’ consumption is also equal to permanent income. Clearly, this equilibrium corresponds to the first-best. Both traders and nontraders perfectly smooth consumption over time. Under fixed rates, it follows from equation 2.34 that consumption of nontraders is given by ctNT ¼
y p; y p ð1 þ vt vt1 Þ;
t ¼ 0; t b 1:
In turn, consumption of traders is given by (from 2.21 and 2.27) " # y 1 lX t p T c ¼ y 1 b ðvt vt1 Þ : l t¼0
ð2:36Þ
ð2:37Þ
In what follows, it will prove useful to define a ‘‘permanent’’ velocity shock, v p , as v p 1 ð1 bÞ
y X
b t vt :
t¼0
Under the assumption that v0 ¼ v p , it follows that (see appendix 5.2)9 y X
b t ðvt vt1 Þ ¼ 0:
t¼1
Substituting 2.38 into 2.37, we obtain traders’ consumption: c T ¼ y p:
ð2:38Þ
Amartya Lahiri, Rajesh Singh, and Carlos A. Ve´gh
32
Traders’ consumption is therefore the same under flexible and fixed exchange rates and they are thus indifferent between the two regimes. As for nontraders, it follows from 2.36 and 2.38 that the present discounted value of nontraders’ consumption under fixed rates is the same as under flexible exchange rates. As a result, nontraders are clearly better off under flexible exchange rates, in which case they have a flat path of consumption. Since traders are indifferent, we conclude that flexible exchange rates dominate. What is the underlying intuition? The key lies in the role of the exchange rate as a shock absorber in the presence of velocity shocks. If velocity increases, the nominal exchange rate also increases (a nominal depreciation of the domestic currency), thus offsetting the shock. Under fixed exchange rates, the natural adjustment mechanism (the agents’ ability to recompose their nominal money balances through the central bank) is not fully operative because nontraders cannot access asset markets. Hence, fluctuations in velocity lead to fluctuations in consumption. Specifically, an increase in velocity ðvt > vt1 Þ implies that more money balances are available for consumption; a decrease in velocity ðvt < vt1 Þ implies that fewer money balances are available for consumption. 2.3.2
Output Shocks Only
Suppose that there are only output shocks (i.e., set vt ¼ v > 0). Then under flexible rates, consumption of nontraders and traders continues to be given by 2.29 and 2.30. Nontraders absorb the full variability of the endowment path. Under fixed exchange rates, consumption of nontraders follows from 2.34: y0 ; t ¼ 0; ctNT ¼ ð2:39Þ yt þ ð1 vÞð yt1 yt Þ; t b 1: Under the assumption that y0 ¼ y p , it follows that (see appendix 5.3) y X
b t ð yt1 yt Þ ¼ 0:
ð2:40Þ
t¼1
From 2.39 and 2.40, it follows that the present discounted value of ctNT under fixed rates will be the same as under flexible rates. Consumption of traders follows from 2.35 and 2.40: c T ¼ y p:
ð2:41Þ
As is the case under velocity shocks, traders’ consumption is the same under flexible and fixed exchange rates. Traders are therefore indifferent between the two regimes.
Optimal Exchange Rate Regimes
33
For expositional clarity, it is useful to rewrite nontraders’ consumption as t ¼ 0; y0 ; NT ð2:42Þ ct ¼ vyt þ ð1 vÞ yt1 ; t b 1; which makes clear that from t ¼ 1 onward, nontraders’ consumption is an average of this period’s and last period’s output. Clearly, consumption of nontraders will fluctuate under both flexible and fixed exchange rates but will fluctuate less under fixed rates. Since, as already shown, the present discounted value of ctNT is the same under both regimes, nontraders’ welfare will be higher under fixed exchange rates. Intuitively, 2.42 states that today’s consumption is a weighted average of last period’s and this period’s real sales revenues. Fixed exchange rates allow purchasing power to be transferred across periods which, as equation 2.42 makes clear, results in some consumption smoothing over time. In contrast, under flexible rates a constant money supply implies that the real value of last period’s sales is equal to current output. As a result, current consumption depends solely on current output. We conclude that since traders are indifferent between the two regimes and nontraders are better off under a fixed exchange rate, social welfare will be maximized if, in response to output shocks, a fixed exchange rate is adopted. 2.4
Concluding Remarks
We have shown that the influential Mundell-Fleming result—that the choice of the optimal exchange rate regime should depend on the type of shock hitting the economy—critically depends on the assumption that while there are frictions in goods markets (in other words, sticky prices), asset markets are frictionless. If we reverse these assumptions—frictionless goods markets and segmented asset markets—we turn the famous Mundell-Fleming dictum on its head: flexible rates are called for in the presence of monetary shocks whereas fixed exchange rates are optimal in the presence of real shocks. We thus conclude that the optimal exchange rate depends not only on the type of shock (monetary versus real) but also on the type of friction (goods market versus asset market). A more modern approach to exchange rate regimes would view fixed and flexible exchange rate regimes as two particular cases of a more general monetary policy rule, which could in turn incorporate a response to current (if observable) and past shocks. In Lahiri, Singh, and Ve´gh (2006), we follow this more general approach and show how an optimal monetary policy rule would actually involve responding to contemporaneous shocks. It is only in the absence of output shocks
Amartya Lahiri, Rajesh Singh, and Carlos A. Ve´gh
34
(that is, a world with only velocity shocks) that a ‘‘pure’’ flexible exchange rate— as studied in this paper—would be optimal. 2.5
Appendixes
2.5.1
Conditions for a Binding Cash-in-Advance
This appendix derives the conditions needed for the cash-in-advance to bind for both nontraders and traders and then provides an example of the restrictions that need to be imposed on the output and velocity processes. 2.5.1.1 When Does the Cash-in-Advance Bind for Nontraders? NT y gt¼0 to maximize lifetime utility equation 2.1 subject Nontraders choose fctNT ; Mtþ1 to the sequence of flow constraints given by 2.2 and the sequence of cash-inadvance constraints given by 2.3 for a given M0NT . In terms of the Lagrangian: L¼
y X
b t uðctNT Þ þ
t¼0
þ
y X
y X
NT b t lt ðMtNT þ Et yt Et ctNT Mtþ1 Þ
t¼0
b t Ct ðMtNT þ vt Et yt Et ctNT Þ;
t¼0
where lt and Ct are the multipliers associated with constraints 2.2 and 2.3, respectively. NT The first-order conditions for ctNT and Mtþ1 are given by u 0 ðctNT Þ ¼ Et ðlt þ Ct Þ and
ð2:43Þ
bðltþ1 þ Ctþ1 Þ ¼ lt :
ð2:44Þ
The Kuhn-Tucker condition for Ct is given by MtNT þ vt Et yt b Et ctNT ;
Ct b 0;
ðMtNT þ vt Et yt Et ctNT ÞCt ¼ 0: Suppose that Ct > 0; that is, the cash-in-advance constraint binds. Then, it follows from 2.43 and 2.44 that NT u 0 ðctþ1 Þ 1 Etþ1 lt : ¼ u 0 ðctNT Þ b Et lt þ Ct
Optimal Exchange Rate Regimes
35
Hence, for the cash-in-advance to bind, it must be the case that u 0 ðctNT Þ > b
Et 0 NT u ðctþ1 Þ: Etþ1
ð2:45Þ
If the cash-in-advance binds, it means that nontraders prefer not to carry over nominal money balances from one period to the next even though doing so would provide more consumption tomorrow. In other words, money balances are not used for saving purposes. In this case—and as condition 2.45 indicates—the consumer is unwilling to save and therefore today’s marginal utility will be higher than tomorrow’s, adjusted by the discount factor and the return on money. To fix ideas, consider the case of logarithmic preferences. Condition 2.45 then reduces to NT ctNT < ctþ1
1 Etþ1 : b Et
ð2:46Þ
Using the quantity theory (equation 2.21), we can rewrite this equation as ctNT 1 1 vt yt : ð1 þ m < Þ tþ1 NT 1 vtþ1 ytþ1 b ctþ1
ð2:47Þ
Flexible Exchange Rates Consider the case of flexible exchange rates with a constant money supply. In this case, ctNT ¼ yt and mtþ1 ¼ 0. Equation 2.47 then reduces to b<
1 vt : 1 vtþ1
ð2:48Þ
As long as this condition holds (which implies a restriction on the variability of the velocity shocks), the cash-in-advance constraint will bind. Clearly, since this condition involves exogenous variables, one can always choose parameters such that it will hold. Fixed Exchange Rates Consider the case of fixed exchange rates. In this case, Etþ1 ¼ Et ¼ E. Use condition 2.46, taking into account 2.34, to obtain b<
ð1 vt Þ yt þ vtþ1 ytþ1 : ð1 vt1 Þyt1 þ vt yt
ð2:49Þ
Again, since this condition involves only exogenous variables, one can always choose b and output and velocity processes such that it holds.
Amartya Lahiri, Rajesh Singh, and Carlos A. Ve´gh
36
Intuition To understand the intuition as to why the cash-in-advance (CIA) may bind for nontraders, consider the case of flexible exchange rates and no velocity shocks (that is, output shocks only). In that case, condition 2.48 will always hold because, by assumption, b < 1. Intuitively, suppose that yt > ytþ1 and consider the nontrader’s choice at time t. Based on the consumption-smoothing motive, nontraders would want to save in order to consume more next period when output will be low. However, given that mt ¼ 0, periods of high output will coincide with periods in which the real return on nominal money balances is low. To see this, notice that using the cash-in-advance the gross real return on holding money is given by Et ytþ1 ¼ : Etþ1 yt Since yt > ytþ1 , then Et =Etþ1 < 1, which means a negative real return on money. Hence, with logarithmic preferences, the nontrader’s desire to dissave based on the negative real return on money more than offsets the desire to save based on consumption-smoothing motives. 2.5.1.2 When Does the Cash-in-Advance Bind for Traders? For the CIA to bind for traders, we just need to ensure that the nominal interest rate is positive. The restrictions needed for this depend on the exchange rate regime. Flexible Exchange Rates From the interest parity condition 2.16, a positive nominal interest rate requires that Etþ1 1 > : Et 1þr Using the quantity theory equation 2.21, it follows that Etþ1 1 vt1 yt1 ¼ : Et 1 vt y t Combining the last two equations—and recalling that bð1 þ rÞ ¼ 1—it follows that if b<
1 vt1 yt1 ; 1 v t yt
ð2:50Þ
then the nominal interest rate will always be positive and the CIA will always bind for traders as well.
Optimal Exchange Rate Regimes
37
Fixed Exchange Rates Under fixed exchange rates, the interest parity condition 2.16 indicates that the nominal interest rate will always be positive since 1 þ i ¼ 1 þ r. 2.5.1.3 An Example Let us illustrate the restrictions necessary to ensure a binding cash-in-advance constraint for cases of only one shock at a time (the case studied in the text). Suppose b ¼ 0:96. Output Shocks Only Suppose that vt ¼ v ¼ 0:2 > 0 and that yt alternates between 1.04 and 1. For nontraders, 2.48 holds since b < 1 and condition 2.49 becomes (assuming the most restrictive case, which is yt1 ¼ 1:04, yt ¼ 1, and ytþ1 ¼ 1:04): b<
ð1 vÞyt þ vytþ1 ; ð1 vÞyt1 þ vyt
which reduces to b < 0:977 and hence holds. For traders, 2.50 is satisfied since b < yt =ytþ1 ¼ 0:962, and hence the CIA binds under both flexible and fixed exchange rates. Velocity Shocks Suppose that yt ¼ ytþ1 ¼ y p . The velocity variable alternates between two values: 0.20 and 0.22. Assume first that vt1 ¼ 0:2, vt ¼ 0:22, and vtþ1 ¼ 0:2. Then, for nontraders under flexible rates, it must be the case that b<
1 vt ; 1 vtþ1
ð2:51Þ
which holds since b < 0:975. Under fixed rates, it must the case that b<
1 vt þ vtþ1 ; 1 vt1 þ vt
ð2:52Þ
which holds—since under the most restrictive case in which vt1 ¼ 0:2, vt ¼ 0:22, and vtþ1 ¼ 0:2, then b < 0:961. For traders, the cash-in-advance always holds. 2.5.2
Proof that
Rewrite y X t¼1
Py
t¼1
PT
tF1
b t (vt C vt1 ) F 0 if v0 F v p
b t ðvt vt1 Þ as
b t ðvt vt1 Þ ¼ bv0 þ ð1 bÞ
y X t¼1
b t vt :
Amartya Lahiri, Rajesh Singh, and Carlos A. Ve´gh
38
But, by definition of v p and given that v0 ¼ v p , y X
b t vt ¼
t¼1
vp b vp ¼ v p: 1b 1b
Hence, y X
b t ðvt vt1 Þ ¼ bðv p v0 Þ ¼ 0
t¼1
as v0 ¼ v p . 2.5.3
Proof that
PT
tF1
b t ( yt1 C yt ) F 0 if y0 F y p
Replace v with y in section 5.2. Notes This chapter was originally prepared for the conference in honor of Guillermo Calvo, held at the International Monetary Fund in April 2004. We are grateful to Martin Eichenbaum and conference participants for helpful comments and suggestions. 1. Mulligan and Sala-i-Martin (2000) report that even for the United States, 59 percent of the population (as of 1989) did not hold interest-bearing assets. One would conjecture that this figure is even higher for developing countries. 2. It is worth noting that our results are in the spirit of an older literature that focused on the pros and cons of alternative exchange rate regimes in models with no capital mobility (see, for instance, Fischer (1977) and Lipschitz (1978)). See also Ching and Devereux (2003) for a related analysis in the context of optimal currency areas. 3. Asset market segmentation could be endogenized by assuming that there is a fixed cost of accessing asset markets. With idiosyncratic fluctuations in endowment, the number of agents that choose to gain access to asset markets would be endogenously determined. 4. Appendix 2.5.1 derives sufficient conditions for the cash-in-advance constraint to bind. Contrary to what our intuition would first tell us—that the cash-in-advance constraint would rarely bind because nontraders would like to save some money for low endowment periods—the cash-in-advance may bind under very weak conditions because unspent money balances have an opportunity cost that is positively related to the state of the economy (that is, the opportunity cost is higher in good times). In good times, therefore, nontraders would like to save for consumption smoothing motives but dissave for financial reasons. 5. Given the open economy nature of the model, the private sector as a whole must always be able to exchange money for foreign bonds (and vice versa) in the asset market (even under flexible rates), and bonds for goods (and vice versa) in the goods market. One can imagine a trading agency that is in charge of such activities or, alternatively, that the household has a third member, a foreign trader, whose job is to put aside some of the household’s money or bonds during the asset market and transact with foreigners during the goods market.
Optimal Exchange Rate Regimes
39
6. Appendix 2.5.1 derives the restrictions needed to ensure a positive nominal interest rate. 7. This assumption just ensures that the present discounted value of income is identical across traders and nontraders when the money supply or the exchange rate is fixed. 8. We will consider only the extreme cases of a constant money supply (under flexible rates) and a fixed exchange rate (as opposed to time-varying paths of the exchange rate). For an extension of our main results to more general rules involving a fixed rate of growth of either the money supply or the exchange rate, see Lahiri, Singh, and Ve´gh (2006). 9. In a stochastic version of the model, the equivalent assumption would be that velocity shocks are white noise.
References Alvarez, Fernando, Robert Lucas, Jr., and Warren Weber. 2001. ‘‘Interest Rates and Inflation.’’ American Economic Review 91: 219–225. Calvo, Guillermo. 1999. ‘‘Fixed versus Flexible Exchange Rates: Preliminaries of a Turn-of-Millenniun Rematch.’’ Mimeo, University of Maryland. Cespedes, Luis, Roberto Chang, and Andres Velasco. 2004. ‘‘Balance Sheets and Exchange Rate Policy.’’ American Economic Review 94: 1183–1193. Ching, Stephen, and Michael B. Devereux. 2003. ‘‘Mundell Revisited: A Simple Approach to the Costs and Benefits of a Single Currency Area.’’ Review of International Economics 11: 674–691. Fischer, Stanley. 1977. ‘‘Stability and Exchange Rate Systems in a Monetarist Model of the Balance of Payments.’’ In The Political Economy of Monetary Reforms, ed. Robert Z. Aliber, 59–73. Fleming, J. Marcus. 1962. ‘‘Domestic Financial Policies Under Fixed and Flexible Exchange Rates.’’ IMF Staff Papers 9: 369–379. Lahiri, Amartya, Rajesh Singh, and Carlos A. Ve´gh. 2006. ‘‘Optimal Monetary Policy under Asset Market Segmentation.’’ Mimeo, Iowa State University. ———. 2007. ‘‘Segmented Asset Markets and Optimal Exchange Rate Regimes.’’ Journal of International Economics 72: 1–21. Lipschitz, Leslie. 1978. ‘‘Exchange Rate Policies for Developing Countries: Some Simple Arguments for Intervention.’’ IMF Staff Papers 25: 650–675. Lucas, Robert E., Jr. 1982. ‘‘Interest Rates and Currency Prices in a Two-Country World.’’ Journal of Monetary Economics 10: 335–359. Mulligan, Casey, and Xavier Sala-i-Martin. 2000. ‘‘Extensive Margins and the Demand for Money at Low Interest Rates.’’ Journal of Political Economy 5: 961–991. Mundell, Robert A. 1968. International Economics. New York: MacMillan.
3
Monetary Policy Rules, the Fiscal Theory of the Price Level, and (Almost) All that Jazz: In Quest of Simplicity Leonardo Auernheimer
Guillermo Calvo’s work has invariably been distinguished not only by deep relevance, but by elegant simpleness. Since imitation (successful or not) is the sincerest form of admiration, it seems fitting for this paper to reexamine a group of closely related questions in the same spirit of simplicity. Readers of Calvo’s work will also recognize the connection between some of the points touched on in this paper and some of his contributions—most notably, the relevance of the government budget constraint (Calvo 1985), the concern with levels versus rates of change (Calvo 1981), and the study of interest rate pegging policies (Calvo and Ve´gh 1990). The central question I address is the so-called fiscal theory of the price level (FTPL), and some closely related topics, such as the pegging of the nominal interest rate and the existence not only of an equilibrium, but of mechanisms that under different assumptions will or will not allow the attainment of such equilibrium. The relevant literature during the last fifteen years or so has been both vast and diverse, with the ‘‘eye of the storm’’ being the proposition, first advanced in its most stark form by Woodford (1995), that under certain conditions, and for a given level of nominal government debt (bonds cum fiat money), the path of the price level is determined exclusively by fiscal policy (in other words, the anticipated path of the government primary surplus and of government debt). Indispensable references in this discussion are, among others, Sims (1994, 1999), Cochrane (2001, 2002), Kocherlakota and Phelan (1999), McCallum (1997, 2001), Christiano and Fitzgerald (2000), Buiter (2002, 2005), Leeper (1991), Bassetto (2002), Niepelt (2004), Dupor (2000), and Daniel (2001) for the case of the open economy and, of course, Woodford (especially 1994, 1995, and 1998). Although the natural benchmark for the beginning of the discussion is Woodford (1995), there are previous pieces anticipating some of the same propositions (Begg and Haque 1984, Auernheimer and Contreras 1990, 1995). As is usually the case, there
42
Leonardo Auernheimer
are also classical works touching on some of the very basic concepts—Metzler (1951) being, of course, the primary reference, as well as Tobin (1974) and, more recently, the already classic Sargent and Wallace’s ‘‘unpleasant monetarist arithmetic’’ (1981). This paper is certainly not a survey, but the revisitation of the question along the lines of a very simple framework first presented in Auernheimer and Contreras (1990), and the evaluation of some of the propositions for and against the basic FTPL claims. Even a casual reader will notice that the literature is extensive, in some instances a bit unfocused, and at times full of sound and fury.1 Rather than addressing the various propositions and counterpropositions, we will proceed with the analysis of the simple model, derive some results, and indicate, in each case, the extent to which these results are or are not in agreement with those advanced by the various contenders. We consider an extremely simple model (and will not hesitate to use graphical representations when helpful), which is rather standard in the literature and, in fact, basically coincides with models used in most of the discussion. At variance with most if not all the literature on the subject, we use continuous time, for two related reasons.2 First, the analysis is simpler and less cluttered. Second, ambiguities and potential confusion among dates and beginning- or end-of-the-period questions are removed, and the distinction between levels and rates of change is sharpened—this last consideration being particularly important in this case.3 There is still another, more compelling reason for our choice of continuous time analysis, which will become clear as we proceed. In the class of problems we discuss, the fundamental questions are about initial conditions at time t ¼ 0: in the absence of anticipated future changes in policy or in behavioral parameters, the whole path of all variables, including of course the price level, is determined at that initial time. More often than not, and depending on the pre-initial conditions at time t ¼ 0 , those initial conditions require discrete changes in some variables—the most important one in our case, although not the only one, being the price level. There is quite a lot of economic analysis and market mechanisms that can be discussed concerning the realization of those changes, and in continuous time there is a very sharp distinction between the mechanics of those discrete changes happening at once (as if time could be ‘‘stopped’’ at t ¼ 0 ) and the ensuing evolution of the variables once time is ‘‘reinitiated’’ at t ¼ 0þ . In section 1 we present the bare-bones simple model, and elaborate on some basic points to be addressed. Section 2 considers the case of a ‘‘monetary rule,’’ in which the government sets the path of the money supply, and section 3 compares the model’s conclusions to some of the propositions in the literature. Section 4 discusses the case of a central bank pegging the nominal interest rate, and section 5 concludes.
Monetary Policy Rules, the Fiscal Theory of the Price Level
3.1
43
The Model
Individuals are identical, infinitely lived, and at each point in time maximize the present value of their lifetime utility, which depends on their consumption and the services of their real money stock. We assume the utility function to be separable in the two arguments. Time is continuous, there is perfect foresight, and prices are perfectly flexible. Then, at each initial point t ¼ 0, the typical individual maximizes ðy
UðcðtÞ; mðtÞÞert dt
ð3:1Þ
0
where r is the positive, constant rate of time preference, and c and m are the flow of consumption and the stock of real cash balances, respectively. Individuals receive a constant flow endowment of an instantaneously perishable single commodity, which we call ‘‘real income,’’ and hold two assets: fiat money, which yields no interest, and government bonds. The exact characterization of these government bonds is crucial for the questions at hand, and requires some elaboration.4 Here, we define them as ‘‘call bonds,’’ paying the prevailing nominal interest rate but with instantaneous maturity: in other words, very much as sight deposits. As deposits, they are denominated in terms of the monetary unit (say, one dollar), and changes in the nominal interest rate carry no capital gains or losses. This is, in its discrete-time equivalent form of ‘‘one-period bonds,’’ the assumption used in much of the literature.5 The stock of nominal money is denoted by M, with P being the money price of the single good (the price level), so m 1 M=P is the real money stock. The typical individual’s flow budget constraint expressed in nominal terms, an identity at each point time t, is then yPðtÞ þ BðtÞiðtÞ 1 cðtÞPðtÞ þ vðtÞPðtÞ þ dBðtÞ=dt þ dMðtÞ=dt where y is the constant level of real income, B the level of government bonds, i the nominal interest rate, and v the flow of a real head tax (subsidy, if negative). Defining b 1 B=P as the real stock of government bonds and p as the rate of inflation, the constraint can be expressed in real terms as y þ bðtÞr ¼ cðtÞ þ vðtÞ þ dbðtÞ=dt þ dmðtÞ=dt þ mðtÞpðtÞ:
ð3:2Þ
Notice that in writing 3.2 we have used the assumption that the equilibrium Fisher condition, iðtÞ ¼ r þ pðtÞ;
ð3:3Þ
44
Leonardo Auernheimer
where r is the (constant) real interest rate, holds at all times. Because 3.3 is an equilibrium condition, then 3.2 is no longer a strict identity, but an equality that will hold at all times. The government (the fiscal authority cum the central bank) prints fiat money at zero cost, borrows or lends via the sale or purchase of bonds, imposes neutral head taxes (transfer, if negative), and spends on goods which are thrown into the ocean—a real flow denoted by g. The flow government budget constraint in nominal terms, again an identity, is then dB=dt 1 BðtÞi þ gðtÞPðtÞ vðtÞPðtÞ dM=dt: Using 3.3 once more, the constraint can be written in real terms as db=dt ¼ bðtÞr þ gðtÞ vðtÞ ðdm=dtÞ mðtÞpðtÞ ¼ bðtÞr sðtÞ ðdm=dtÞ mðtÞpðtÞ
ð3:4Þ
where sðtÞ ¼ vðtÞ gðtÞ is the government primary surplus (deficit when negative). This is again an equality that will hold at all times, as in the case of 3.2. Addition of the two budget constraints yields the resource constraint y ¼ cðtÞ þ gðtÞ:
ð3:5Þ
In what follows we assume that government expenditures remain constant, so that any changes in the primary surplus s take place via a change in the tax (subsidy). What this means is that the level of consumption will also remain constant. Maximization of 3.1 by the typical individual, subject to its budget constraint 3.2, yields the familiar Euler necessary conditions for optimality: dc=dt ¼ ðUc =Ucc Þðr rÞ
ð3:6Þ
and dc=dt ¼ ðUc =Ucc Þðr þ pÞ ðUm =Ucc Þ:
ð3:7Þ
Expressions 3.6 and 3.7 imply ðUm =½r þ pÞ ¼ Uc ;
ð3:8Þ
and the condition r ¼ r, in turn, implies that dc=dt ¼ 0;
ð3:9Þ
that is, consumption is piecewise constant.6 Note, again, that given the assumption of constant levels of government expenditures and real income, then consumption becomes a constant parameter that, being piecewise constant, can
Monetary Policy Rules, the Fiscal Theory of the Price Level
45
change a finite number of times, as any parameter can—in this case, for example, as a result of exogenous changes in the endowment. In addition, optimal plans are to satisfy the transversality conditions7 lim bðtÞert ¼ 0 as t ! y and lim mðtÞert ¼ 0
as t ! y:
We are not constraining government debt to be nonnegative—that is, we are allowing b < 0, so that government can lend to the public and hold positive net assets.8 We are also ruling out mðtÞ < 0. Notice next that expression 3.8 is the demand for money. Since in all of what follows we will specify a constant level of real income and of government expenditures (with whatever changes take place in the primary surplus assumed to be via a change in head taxes), then consumption will be constant. For a real interest rate that is then also constant, for convenience we can write expression 3.8 as m ¼ LðiÞ ¼ Lðr þ pÞ
L 0 < 0:
ð3:8 0 Þ
Next, and most importantly, ‘‘helicopter drops’’ (or ‘‘annihilations’’ a` la Hume) of fiat money as part of a deliberate policy rule are precluded.9 In particular, we assume that any discrete change in the stock of nominal money takes place via an equal change in the stock of nominal bonds, with the opposite sign—in other words, at any point in time t DMjt 1 DBjt
ð3:10Þ
This is, of course, an assumption at the heart of both the fiscal theory and the correct specification of a constant nominal interest rate pegging. Assumption 3.10 is the equivalent of specifying that all changes in the nominal money stock need to be accounted for in the government budget constraint’s other asset—the stock of bonds. Still another assumption will be that the system operates in a range at which jðqLðÞ=qpÞðp=LðÞÞj < 1;
ð3:10 0 Þ
that is, an increase in the inflation rate will increase the inflation tax—an assumption that not only conforms to empirical estimates (except for some very dramatic, short-lived hyperinflationary episodes) but that is necessary to make sense of the analysis.
46
Leonardo Auernheimer
Omitting the time arguments in order to avoid clutter and defining p ¼ L1 ðmÞ r ¼ lðmÞ, where L1 is the inverse of function 3.8 0 , the simple model can be summarized in the two expressions db=dt ¼ br s mm
ð3:11Þ
or its equivalent db=dt ¼ br s dm=dt mp;
ð3:11 0 Þ
and dm=dt ¼ mðm lðmÞÞ
ð3:12Þ
where m 1 ðdM=dtÞð1=MÞ is the rate of nominal money growth. Notice that, in this construction, 3.8 holds at all times at which the system is engaged in the optimal paths implied by the first order conditions.10 We will now consider the first of two cases, in the context of the model, of a well-defined policy rule—a monetary rule (the second being an interest rate rule, to be considered later). In every case we will ask the following three questions: (1) Is there a unique path of the price level satisfying the transversality conditions? (2) If there is such a unique path, is it accessible for any initial conditions? (3) If the answer to the two previous questions is in the affirmative, is there a market mechanism so that agents are induced to reach such a unique path? This third question is directly related to the general point raised by Lucas (1978) in a different context, but very relevant here nonetheless: One would feel more comfortable . . . with rational expectations equilibria if these equilibria were accompanied by some form of ‘‘stability theory’’ which illuminated the forces which move an economy toward equilibrium. (1429)
We will refer to this as the existence, or lack of existence, of a market-induced stability mechanism, and our answers to the question will be more on the side of conjectures than on the side of proofs. Finally, notice that in this model, for an exogenous level of the government surplus and for the control of nominal aggregates, a unique path of the price level is sufficient for determining a unique path of the level of real government debt and of the real money stock. 3.2
A Monetary Rule
In the system 3.11–3.12, assume that government fixes a preannounced path of the money supply. For simplicity, we assume that such path implies a constant rate of monetary growth; that is, MðtÞ ¼ M0 expðmtÞ
tb0
ð3:12 0 Þ
Monetary Policy Rules, the Fiscal Theory of the Price Level
47
Figure 3.1
In addition, we also assume that the anticipated and effective fiscal policy is to keep constant the level of the primary budget surplus, s. This is, of course, a typical non-Ricardian policy, as is defined in the literature on the FTPL.11 The system in expressions 3.11 and 3.12 can then be represented in the phase diagram in figure 3.1, where b ¼ ðs þ mmÞ=r
ð3:13Þ
lðmÞ ¼ m
ð3:14Þ
are the values of b and m satisfying db=dt ¼ 0 and dm=dt ¼ 0, respectively, with b and m being the long-run steady values satisfying both 3.13 and 3.14. Notice that, of course, 3.14 is not the equilibrium ‘‘money supply equal to money demand’’ (which is given by 3.8 0 and implied in 3.14), but the condition for the rates of inflation and monetary growth to be equal. Since at any point in time the stocks of both nominal bonds ðBÞ and nominal money ðMÞ are given, a discrete increase (decrease) in the price level will decrease (increase) the levels of both real debt and real money balances by the same proportion.12 In terms of the graph in figure 3.1, this can be represented by movements along the ray passing through the initial pair b, m. Note, though, that since the price level is not a state variable, an initial point is strictly defined as a pair Bð0Þ, Mð0Þ—the nominal stock of bonds and money—so that in this graphical representation an initial position is characterized as any point along a ray with a slope Bð0Þ=Mð0Þ. Notice also that the graph in figure 3.1 is drawn for the more general case of a positive rate of monetary expansion; in other words, m > 0. The special but theoretically important case in which the nominal money stock is constant— m ¼ 0—is depicted in the graph of figure 3.2.
48
Leonardo Auernheimer
Figure 3.2
Consider first the case m > 0. The answer to our first question (is there a unique path of the price level satisfying the transversality conditions?) is in the affirmative. This is the path mðtÞ ¼ m and bðtÞ ¼ b , and the path of prices implied by 3.12 0 . The answer to the second question (if there is a unique such path, is it accessible for any initial conditions?) is in the negative. An initial ratio Bð0Þ=Mð0Þ ¼ b =m is a necessary condition for reaching the unique equilibrium m , b , via a discrete change in the price level. For any other initial nominal levels of debt and money there is no path of the price level satisfying the transversality and the nonnegative money conditions. Notice, nevertheless, that such a general conclusion would not hold if a nonnegative restriction is imposed on nominal debt (that is, if government is not allowed to lend and hold positive assets). In this case, any initial ratio of nominal debt and nominal money could result in paths of the price level for which both real debt and real money would approach zero. The third, and most interesting, question is whether there is a market mechanism that, if the necessary initial condition Bð0Þ=Mð0Þ ¼ b =m is satisfied, will generate a price level P such that Bð0Þ=P ¼ b and Mð0Þ=P ¼ m . The answer to this question is given here as the sketch of a conjecture. Consider a pre-initial case at which the price level is P , and given the initial levels of nominal money and bonds the system rests at its steady state equilibrium b , m , with the inflation rate equal to the rate of monetary growth, p ¼ m. Perform now the usual experiment of assuming that all of a sudden, for no other reason than a mysterious uniform marking-up of prices, the price level increases to P1 > P . The usual story for how an instantaneous equilibrium is restored is (implicitly assuming that chronological time is stopped) that the rise in prices has decreased real cash balances, with the result that now, at the same inflation rate, agents have ‘‘too little money’’—m1 < Lðr þ mÞ—with the stock of real bonds
Monetary Policy Rules, the Fiscal Theory of the Price Level
49
having also fallen to b1 < b .13 They will then decrease their spending, but since aggregate output is fixed, successive rounds of ‘‘instantaneous’’ price adjustments (decreases) will take place, with the price level ending up at the original value P , with the economy back at the initial equilibrium. This story is the usual one in models where money is the only asset. What about our case, in which agents also hold government bonds? Wouldn’t individuals, in this instantaneous process, also seek to perform a portfolio adjustment, exchanging bonds for money? There are two problems with this version of the story. First, it is true that individuals will be able to withdraw their call bonds (i.e., not to continue part or all of their rollover), but it is also true that if government is committed to the preannounced path of the nominal money supply, it will immediately restore its previous level—it would take only an infinitely small increase in the nominal interest rate to persuade depositors to keep the initial level of those call bonds. The second problem is that an exchange of bonds for money will not restore the previous equilibrium. In this conjecture, it seems that the ‘‘monetary story’’ of intended changes in expenditures is rather persuasive.14 The alternative to the ‘‘instantaneous’’ adjustment process outlined above is, of course, that at the point b1 , m1 the inflation rate would increase to a level p1 > m for which m1 ¼ Lðr þ p1 Þ, and for which the system would engage in the dynamics of a path violating the transversality conditions—a path not only formally inadmissible but one which makes little economic sense. For initial values outside the ray Bð0Þ=Mð0Þ ¼ b =m there is no price level that would allow the system to attain the unique steady-state equilibrium. A particular instance of this proposition is the Sargent-Wallace case of an ‘‘unpleasant monetarist arithmetic’’ (Sargent and Wallace 1981) in which, starting from an initial equilibrium m0 , b0 , government lowers the rate of monetary expansion, keeping the same level of the government primary surplus.15 It is easy to show that, under assumption 3.10 0 , a fall in the rate of monetary growth will result in an equilibrium at m > m0 , b < b0 , and to show (and to verify by mere inspection from the graph in figure 3.3, in which A is the initial equilibrium, and point A0 the new steady state after the change in policy) that from the original equilibrium at m0 , b0 , there is no possible change in the price level that would generate an immediate adjustment to the new unique equilibrium. Two things can then happen: there is no anticipated further change in either fiscal policy (the primary surplus) or monetary policy (the rate of monetary expansion), or there is the generalized anticipation that at a future time some of those changes will occur so that a new long-run equilibrium is achieved. In the first case a long-run feasible equilibrium will simply not exist. In the second case, a future corrective policy change is anticipated—the possibility envisioned by Sargent and Wallace (1981) being an increase in the rate of money growth. Either this correction or an appropriate rise in the fiscal surplus
50
Leonardo Auernheimer
Figure 3.3
would generate a unique and feasible adjustment, ultimately leading to a unique steady-state equilibrium. Another possibility is, of course, a temporary violation of the monetary rule, with a once-and-for-all increase in the nominal money stock, leading to a corresponding once-and-for-all increase in the price level—a temporary, short-lived hyperinflation. It is easy to show (even in simple graphical terms in the context of figure 3.3), that this can be accomplished either via a ‘‘rain of money’’ (a temporary violation of assumption 3.10) or a purchase of bonds. These possibilities are depicted in figure 3.3, where a one-time gift or a rain of money results in a movement from point A to point D, and a purchase of bonds in a movement from A to C, with a one-time increase in prices then allowing to access the new long-run equilibrium at point A0 . Notice, parenthetically, that there are two distinct effects of the increase in the money stock, not always clearly distinguished in the literature. One is the obvious ‘‘inflating away’’ part of real debt via a fall in the real value of bonds; the other, the purchase of interest bearing bonds in exchange for non-interest-bearing money. That is why a rain of money requires a higher increase in the nominal money stock, and a higher one-time increase in prices, than the purchase of bonds—in the first case, only the first of these effects is at work. If debt is denominated in real terms, as in Sargent and Wallace’s analysis, then the required increase in the money stock via purchase of bonds can also do the trick, with the highest necessary increase in money and the highest increase in prices (point B in figure 3.3); here, a rise in prices will not decrease real debt, so that the entire adjustment needs to take place via the second effect, and a rain of money will not work at all. This temporary hyperinflation ‘‘solution’’ is indeed the manner in which some governments have at times tried to deal with the ‘‘unpleasant arithmetic’’ problem.16
Monetary Policy Rules, the Fiscal Theory of the Price Level
51
Another, less spectacular, possibility is for government to implement a ‘‘soft’’ monetary rule, in which the rate of growth of the money supply is fixed, but in which discrete, discretional changes in nominal money are part of the policy. With the existence of call bonds, it is even possible, if there is a market-induced stability mechanism, for the public to take the initiative via changes in those deposits, with government passively allowing for those endogenous changes. We do not pursue this topic here (although we will refer to it again in the concluding section 5), except to note first that such a scenario would be quite similar (in fact, observationally equivalent) to a nominal interest rate pegging or targeting. This is not surprising: as we will indicate later, a nominal interest rate pegging determines, at each point in time, the rate of change (but not the level) of prices, and the possibility alluded to in this paragraph determines at each point in time the rate of change of nominal money, but not its level. Consider now the particular case m ¼ 0, a constant nominal money stock, which is the one considered in much of the literature on the FTPL. Now the two system equations 3.13–3.14 become b ¼ s=r lðmÞ ¼ m;
ð3:13 0 Þ ð3:14Þ
which are shown in the graph of figure 3.2. This is a case that, as hinted before, is of especial theoretical interest, and has been analyzed first by Obstfeld and Rogoff (1983) in the pure monetary context, and in the context of the fiscal theory of the price level by, among others, Carlstrom and Fuerst (2000), Christiano and Fitzgerald (2000), Kocherlakota and Phelan (1999), and McCallum (2001). Notice that in this case the link between equations 3.11 and 3.12 would seem to be broken, since changes in the real money stock brought about by changes in the price level do not affect the seigniorage component of the government budget constraint, simply because there is no inflationary finance. But the link still exists for nominal debt, with discrete changes in the price level affecting real debt, and hence the government’s fiscal position. In this case, the answer to our first question (is there a unique path of the price level satisfying the transversality conditions?) is in the negative: there is a welldefined pair of values m , b such that bðtÞ ¼ b and mðtÞ ¼ m for all t, but there is also a set of infinite possible paths of the price level (and the inflation rate) which originate for any initial pair of values bð0Þ ¼ b , mð0Þ < m , along which the transversality conditions are met and the real value of money converges to zero. The result is the same as in the FTPL literature cited above, but with a caveat. Note that, starting from the initial steady state equilibrium b , m , since we have nominal debt, such a point bð0Þ ¼ b , mð0Þ < m cannot be generated by the
52
Leonardo Auernheimer
usual ‘‘inflationary bad dream’’ story, which would result in an immediate increase in the price level and therefore in a point bð0Þ < b , mð0Þ < m for which the transversality conditions would not hold. The condition for hyperinflation, with bð0Þ ¼ b , mð0Þ < m , would result, even in the case of the dream, if debt is in real terms or, for either nominal or real debt, if a sudden ‘‘annihilation’’ of money would take place. 3.3
The Fiscal Theory of the Price Level in the Case of a Monetary Rule
We proceed now to compare our results for the case of a monetary rule with some of the tenets of the FTPL and some of the propositions of its critics. As indicated above, we have performed our analysis in what can be considered a nonRicardian policy par excellence, with a fixed constant primary budget surplus and a fixed path of the money supply—a non-Ricardian policy being, of course, the case where, according to the FTPL, the theory becomes relevant as an alternative to the monetary explanation. Before proceeding, some general comments are in order: first, that in terms of our simple model, the basic proposition of the FTPL is that the price level is determined by equation 3.11, or, more specifically, by equation 3.13—that is, equation 3.11 when real debt is not changing, and independent of any monetary considerations. In turn, expression 3.8 0 , the demand for money, is considered to be more useful when viewed, for an exogenous monetary policy, as determining the interest rate—in other words, the rate of inflation (Woodford 1995, 12, 18). Second, that a considerable part of the discussion on the FTPL has focused on whether expression 3.11, or, more specifically, its equilibrium version 3.13, is a budget constraint or an equilibrium condition. In our view, much of this discussion has been less than illuminating, and has quickly reached the stage of decreasing (or maybe even negative) returns.17 A reference to the theological discussion of how many angels can dance on the head of a pin is irresistible. We take the view that 3.11 is a flow budget constraint (and, more specifically, resulting from an identity), while 3.13 is one of the necessary equilibrium conditions, together with 3.14. Finally, we should note that a consistent feature of the FTPL is the treatment of nominal money and nominal bonds as an aggregate (‘‘total nominal debt’’), and although there is a careful specification of bonds being interest-bearing debt and money being, in general, non-interest-bearing debt, we find the practice slightly misleading. In fact, the neatest result of the FTPL is under the assumption that money effectively becomes an interest-bearing obligation, via a policy of returning to the private sector the revenues from money creation—a policy that we analyze below. Closely related to this point is the definition of an ‘‘adjusted fiscal surplus’’
Monetary Policy Rules, the Fiscal Theory of the Price Level
53
as the usual pure primary surplus plus the revenues from money. Let us consider now some of our conclusions and their correspondence with some of the propositions in this literature. We discuss first, because it is the most straightforward, the case of the hyperinflation solution discussed by Carlstrom and Fuerst (2000), Christiano and Fitzgerald (2000), Kocherlakota and Phelan (1999), and McCallum (2001), resulting from an ‘‘annihilation,’’ or a one-time confiscation of part of the nominal money stock. For the more general case in which m > 0, as we have shown above, such a hyperinflationary solution does not satisfy the transversality conditions if we assume the possibility of negative government debt, as Kocherlakota and Phelan (1999) do. As shown by Obstfeld and Rogoff (1983), such a solution is indeed an equilibrium in a world without nominal government debt, as it would be in our model, with indexed government debt (i.e., specified in real terms). The criticism of the FTPL as implying such a hyperinflationary alternative solution is then rather curious when in fact it is the presence of the fiscal element, with nominal government debt, that allows us to rule out hyperinflation as an equilibrium when m > 0. In the case in which m ¼ 0, considered in most of the literature, hyperinflation is indeed an equilibrium, but we note that the monetarist alternative will not yield a unique non-hyperinflationary equilibrium either in the presence of a nonRicardian policy of an exogenous money stock and a fixed surplus.18 For the monetarist adjustment to work, what is needed is an annihilation of all nominal assets (money and bonds) by the same proportion. Whether the adjustment back to the steady-state equilibrium takes place along the lines sketched by Woodford (1995, 12), in which the adjustment is motivated by an increase in ‘‘total government debt,’’ or via a discrepancy between the demand for money and the actual money stock (as described in the previous section, or as in Kocherlakota and Phelan 1999) is almost immaterial. A tenet of the FTPL is that the price level adjusts to assure db=dt ¼ 0, or br ¼ s þ mm, meaning the price level is determined by expression 3.13—or by 3.13 0 , with m ¼ 0, in which case the above equation reduces to br ¼ s. Interpreted in this form, it would seem clear from the analysis that this would not be the case—such an adjustment, if it takes place, will in general result in paths of the price level that will not satisfy the transversality conditions. The only exception is, of course, the case in which, in terms of our graphical interpretation, the initial ratio of nominal bonds to nominal money happens to be such that Bð0Þ=Mð0Þ ¼ b =m . The FTPL is often cast in an apparently slightly, but fundamentally different, form. For the case in which p ¼ m, expression 3.13, the government’s flow budget constraint that needs to hold for db=dt ¼ 0 can be written as b ¼ ðs þ mpÞ=r;
54
Leonardo Auernheimer
so that B=P ¼ ðs þ ðM=PÞði rÞÞ=r ¼ ðs þ ðM=PÞðiÞ ðM=PÞðrÞÞ=r and ðB þ MÞ=P ¼ ðs þ miÞ=r
ð3:15Þ
where ðs þ miÞ is the ‘‘adjusted fiscal surplus,’’ with the term mi rather than the usual ‘‘inflation tax’’ term mp, to account for the ‘‘interest savings on the government’s monetary liabilities’’ (Woodford 1995, 10). In expression 3.15 it is clear that the price level is uniquely determined by the total of government liabilities ðB þ MÞ and the adjusted fiscal surplus as posited by the FTPL. But notice that equation 3.15 is nothing else than the final reduced form of our overall system 3.11– 3.12, since it assumes not only db=dt ¼ 0 but also p ¼ m; that is, dm=dt ¼ 0. It is then not surprising for the price level to depend on these fiscal magnitudes. It is trivial that in a system in which there is more than one nominal asset, a correspondence needs to exist between the total of those nominal assets, or an arbitrary subset of them, and the price level for the economy needs to exhibit neutrality in the presence of perfect price flexibility. Another problem, of course, is with the definition of the fiscal surplus as being ‘‘adjusted’’ for the revenues from money creation; as it is clear from expression 3.15, for a given unadjusted primary fiscal deficit s, the nominal interest rate (and hence the inflation rate) will depend on the exogenous rate of money growth. As will be mentioned later, an interest rate pegging policy, by fixing the term mi, eliminates this objection, and goes a long way to explain why the FTPL appears as ‘‘especially useful’’ in such a regime. (Woodford 1995, 4). Finally, consider the following argument by one of the main proponents of the FTPL, which in our view is more substantial and goes to the core of the theory. Furthermore, it is even arguable that the expected path of the money supply does not matter for price-level determination, except through its consequences for the government’s budget. (Woodford 1995, 15; his emphasis)
One argument against this proposition could be the following: consider an economy without government finances, in which the only role of government is to print fiat money, which is distributed via head subsidies and increases over time at a fixed proportional rate (which of course could be zero). Will a one-time change (expected or unexpected) in either the stock or its rate of change have consequences for the price level? The answer is of course in the affirmative. But such an argument would be misplaced, amounting to throwing away the baby with the bath water—proponents of the FTPL would immediately reply that in such an economy all possible policies are Ricardian, and their point would be well
Monetary Policy Rules, the Fiscal Theory of the Price Level
55
taken. What is then that makes money to appear to become so unimportant once nominal government debt is introduced? In our view, more than a question of whether an expression such as equation 3.13 determines the price level, the answer rests on the fact that in the presence of nominal government debt, all monetary actions have a fiscal component and a fiscal effect, via the effects of changes in prices on real debt, unless those effects are neutralized via changes in the purely fiscal magnitudes, such as the debt or the primary fiscal surplus (what would, of course, amount to a Ricardian policy for which the monetarist adjustment becomes relevant). This point can be still better understood if we consider Woodford’s discussion of a regime which would ‘‘neutralize the fiscal effects of a change in the money supply’’ (1995, 15). Consider, in terms of the notation of our model, a regime in which the ‘‘revenues from money creation,’’ defined in Woodford as mi, were returned to the public—of course, on a per head basis, unrelated to money holdings. In this case monetary changes have no fiscal effects, and the relevant equations for the government’s flow budget constraint can be written as db=dt ¼ br þ mi s dm=dt mp db=dt þ dm=dt ¼ br þ mr s and, calling w ¼ b þ m ‘‘total government debt,’’ then dw=dt ¼ wr s: In this case it is obvious that all that matters for the determination of the price level (and for the overall global adjustment, for that matter) is the nominal aggregate W 1 B þ M. This is probably the purest exhibit of the FTPL, a case in which exchanges of money for bonds and vice versa are of no consequence. But notice first that it is derived, as it were, by a policy that in fact converts money into bonds, and second, that since there is a well-defined demand for money, equation 3.14 is still relevant and needs to be satisfied. In fact, the system becomes identical to the case analyzed in the previous section for the case m ¼ 0, even with the same geometrical interpretation in figure 3.3, with the term v instead of b. 3.4
A Nominal Interest Rule
A policy of pegging the nominal interest rate to a predetermined path or single value has a long history of controversy in the profession, and for quite a time had a consistently bad reputation. In the seventies and early eighties, in the era of monetarism with adaptive expectations, the argument was that such policy results in explosive behavior. After the advent of rational expectations, the
56
Leonardo Auernheimer
argument was, and still is in many quarters, that it would result in price-level indeterminacy: if the monetary authority pegs the nominal interest rate, then the nominal money stock needs to react passively in response to changes in the price level so as to keep constant the real money corresponding to the demand for money, pinned down by the exogenous value of the rate—meaning that the central bank would need to validate any price level. Explicit consideration of the government budget constraint and the absence of rains of money (that is, changes in the money stock unmatched by corresponding changes of opposite sign on government assets, such as government debt) allows the elimination of indeterminacy—though this is not an opinion shared by most critics of the FTPL. There are reasons why the validity of an interest rate pegging is usually discussed within the context of the FTPL, mainly because in both cases the determinacy or indeterminacy of the price level is at stake, and in both cases the government budget constraint, and in particular the connection among its components, is at the center of the analysis. Yet this need not be necessarily the case.19 In this section we use the elementary framework of section 1 to analyze the implications of what we can in general call an ‘‘interest rate rule.’’ We will first present the mechanics involved, perform some experiments, and then discuss some of the features and implications of a nominal interest rate pegging. We will argue that not only prices are uniquely determined, but that the policy yields some attractive features and plausible results, both in the current context and in the context of an open economy. At the outset, we need to stress a distinction that, although often made explicit in some of the literature, is sometimes not kept in mind. The distinction is between a nominal interest rate ‘‘pegging’’ (or ‘‘fixing’’), and a policy of interest rate ‘‘targeting.’’ The two yield in general the same results, and are therefore observationally equivalent. But, even at the theoretical level (and perhaps still more so), they imply very different types of policies. We characterize a pegging policy as one in which the central bank is willing to lend and borrow, irrespective of quantities, at a fixed nominal interest rate. The central bank is then completely passive, and the actual purchase or sales of bonds (in our case, call bonds) are entirely left to the initiative of the private sector, with the government simply supplying (or demanding) whatever quantities the public demands (or supplies). The analogy with the pegging or fixing of the exchange rate is immediate. A policy of targeting a fixed, preannounced level of the interest rate, even when (as in our very simple model) the targeting is always perfect, is characterized by an active central bank policy of open market operations so as to keep the rate at the targeted level. The reason why these observationally equivalent policies are quite different is that there may not exist a market-induced stability mechanism inducing individuals,
Monetary Policy Rules, the Fiscal Theory of the Price Level
57
despite the fixed level of the interest rate, to take the initiative and engage in exchanges between money and bonds that would lead to a unique rational expectations equilibrium. In what follows we will simply refer to an interest rate rule as encompassing both of these observationally equivalent cases. Consider then the model presented and analyzed in the previous sections, which results in the reduced form equations 3.11 0 and 3.12, which we reproduce here for convenience: db=dt ¼ br s dm=dt mp dm=dt ¼ mðm lðmÞÞ
ð3:11 0 Þ ð3:12Þ
where, as before, government debt is in nominal terms—more specifically, as call bonds. Suppose now that the central bank stands prepared to borrow and lend at a fixed nominal interest rate i , and that such commitment is believed by the public and effectively implemented. Since the nominal interest rate is fixed, so is the real money stock, say, at a level m ¼ Lði Þ;
ð3:16Þ
and, with the Fischer equation holding at all times, arbitrage will assure that p ¼ i r:
ð3:16 0 Þ
Using these values in 3.11 0 and 3.12 yields then the extremely simple formulation db=dt ¼ br s m ði rÞ
ð3:17Þ
and 3.16. A graphical representation, provided in figure 3.4, is immediate and almost embarrassingly simple. Note that although the real money stock will at all chronological times be constant at the level m (since during the instantaneous adjustment process taking place at a date at which we stop chronological time in order to follow the logical steps of the same tatonnement process we considered in section 1), we find it useful to use the plane m, b to follow such a process, rather than in the real line m ¼ m . From 3.17, there is then only one level of the real government debt at which db=dt ¼ 0, and this is b ¼
s þ m ði rÞ ; r
ð3:17 0 Þ
with values of the real debt above this level resulting in rising levels of debt, and vice versa.
58
Leonardo Auernheimer
Figure 3.4
It is obvious from 3.17 that in the absence of anticipated future changes in parameters and in order to satisfy the transversality conditions the adjustment needs to be instantaneous. This is reflected in the arrows showing the laws of motion in figure 3.4. The answer to our first question (is there a unique equilibrium satisfying the transversality conditions?) is in the affirmative. In the context of figure 3.4, for example, this is point A, at b , m . Notice that, for any given level of the nominal aggregate ðB þ MÞ (‘‘government total liabilities,’’ as consistently called by proponents of the FTPL), there is a unique mix of the two components in the aggregate (that is, a unique ratio B=M) and, given such a ratio, a unique level of prices, for which the unique steady-state equilibrium is realized. The right B=M ratio and the right price level are each of them necessary conditions, and, when holding jointly, they become sufficient. This simple reasoning sheds light on the reason why the nominal interest rate rule, when the government budget constraint is taken seriously (in other words, considering the aggregate ðB þ MÞ as given at any point in time), does not lead to price indeterminacy. At the same time, notice that, as a formal matter, the sole specification of the aggregate ðB þ MÞ is not sufficient, per se, for determining the equilibrium price level, despite claims by the FTPL. This is the same problem that we mentioned while considering the case of the monetary rule; that is, that the condition dm=dt ¼ 0 is consistently neglected by expositors of the FTPL. In the case of the interest rate rule, it is clear that any aggregate ðB þ MÞ will determine a unique level of prices for which db=dt ¼ 0 (in other words, equation 3.17 0 will hold), but in general this will not be the price level for which dm=dt ¼ 0 will also be satisfied.20 What is necessary for both conditions to hold is for the ratio of debt to money to be the right one, and an interest rate rule will generate such a ratio via open market operations. This is so, of course, for the fundamental reason that money is a non-interest-bearing government ‘‘liability.’’
Monetary Policy Rules, the Fiscal Theory of the Price Level
59
Our second question is, as in other cases, whether the adjustment leading to that unique equilibrium is feasible. Consider, in the same figure 3.4, a pre-initial point anywhere in the b, m plane—say, point B. Given the pegged nominal interest rate policy, the unique equilibrium can be reached via an open market operation (a sale of bonds to government in exchange for money), leading to point D, and from there on a fall in the price level, which will translate to an increase in the real value of both bonds and money leading to the equilibrium point A. Another alternative amounting to the same is a fall in prices leading to point C, and from there on a transformation of bonds into money via exchanges with the central bank, leading again to the equilibrium A. There are, obviously, infinite possible sequences of falls in the price level and open market operations leading, in this instantaneous adjustment process, to the unique, rational steady-state equilibrium. Our third question (is there a market-induced stability mechanism such that agents would motivate the required exchanges of money for bonds and vice versa, as well as the required changes in prices?) is more complex, and our answer will be tentative. If the mechanism exists, then the central bank’s passive pegging will be sufficient. Individuals hold both money and bonds, so that at each point in time they decide on both the composition of their portfolio and their consumption expenditures. Thus, we have here two possible adjustment mechanisms, one which changes the ðB=MÞ ratio and another that changes prices. It seems reasonable to assume that agents increase their consumption expenditures (leading to a rise in prices, since aggregate consumption is constant) when they hold ‘‘too much money’’ ðm > m Þ and ‘‘too many bonds’’ ðb > b Þ, and that they exchange money for bonds, or vice versa, in order to achieve a ratio of their nominal assets equal to ðB=MÞ ¼ ðb =m Þ. If this is the case, then there are reasons to believe, at the intuitive level, that a market-induced stability mechanism exists, and we take this to be the case. If this mechanism falters, then the central bank can always engage in active open market operations in order to complement it—marking a transition from pegging to targeting. We perform now three experiments, which should be sufficient for providing both a sense of the mechanism involved in an interest rate rule as well as some arguments in favor of a rather positive evaluation of the rule in terms of plausible results and policy consequences. Consider, first, an unexpected permanent increase in the level of the pegged nominal interest rate, a case described in the graph of figure 3.5 (of course, the same discussion can be performed in analytical terms). Suppose, in order to avoid clutter, that the initial position was at point A, and that the interest rate increase results in a new equilibrium at pont C. The adjustment, in this case, requires an exchange of money for bonds; that is, an open market sale of bonds, which moves
60
Leonardo Auernheimer
Figure 3.5
Figure 3.6
the system from point A to point B, and from there, a fall in the level of prices resulting in a movement from point B to point C. Notice that in this case the result is consistent with the usual initial contractionary effect of a rise in the interest rate—which in this model, since aggregate output and consumption are constant, translates in a fall of the price level. And this is so despite the fact that, given the constant real interest rate, the inflation rate will immediately increase to a permanently higher level (expression 3.16 0 ).21 Next, take the case in which an increase in the level of the pegged interest rate is announced, at time t0 , to take place at a future time t1 , the announcement being both believed and later implemented. This is depicted in figure 3.6, where point A shows the initial equilibrium b , m at the time of the announcement. With efficient markets, a requirement in the adjustment is that at the time of the effective increase of the interest rate no discrete changes in the price level are to take place (the usual ‘‘asset price continuity principle’’), meaning any changes at time t1 will
Monetary Policy Rules, the Fiscal Theory of the Price Level
61
Figure 3.7
be limited to open market operations. The qualitative features of the analysis are then easy to characterize: an initial fall in the price level, together with a purchase of bonds by the public at the time of the announcement (immediately positioning the economy at a point such as C), followed by a gradual increase in the level of real government debt during the interim period (from C to D) and, at time t1 (at which the increase in the interest rate is consummated), an additional purchase of government bonds by the public. The effects are qualitatively similar to the previous case of a contemporaneous, unanticipated change, but with a smaller initial increase in the level of prices. Finally, consider the case of a one-time, unanticipated fall in the level of the fiscal primary surplus, s. This is an interesting experiment, because it is a policy change of the ‘‘unpleasant monetarist arithmetic’’ type: a rise in the primary deficit, with no change in the rate of inflation. This case is depicted in figure 3.7, where point A corresponds to the initial equilibrium b , m . A fall in the primary surplus determines a lower long-run level of the real government debt, say, at b < b , with a new equilibrium at point A0 —notice that the arrows indicating the law of motion are drawn for this new lower level b . The instantaneous adjustment involves both a discrete increase in the price level (from A to B) and an open market operation (a sale of government debt by the public), resulting in a movement from point B to point A0 . Some comments on the results of these conceptual experiments are in order. First, as indicated above, notice that the results of an unanticipated increase in the nominal interest rate give a rather clear account of the contractionary effects of an increase in the interest rate and, as a corollary, an explanation for the short- versus long-run consequences of such an increase: an initial deflationary effect, followed by a permanently higher rate of inflation. This is not a result that is easy to generate in models with microeconomic foundations.
62
Leonardo Auernheimer
Second, the adjustment following a fall in the primary surplus implies an automatic response to an occurrence particularly prevalent in developing economies. In those economies, it is common to encounter situations in which the primary deficit is incompatible with the inflation rate implied by either the rate of money growth or, in the case of an open economy, an exogenous path of the nominal exchange rate with a ‘‘too-low’’ rate of devaluation and inflation. In all those episodes the eventual unavoidable accommodation (usually in the central bank’s rather than in the fiscal authority’s behavior) results in a loss of credibility for the monetary authority. With a nominal interest rate rule, which requires no change in the preannounced monetary policy, the credibility question is placed at the door of the fiscal authority, where it belongs. An additional comment is also in order, related to the point made in the previous paragraph concerning the implications of a nominal interest rate rule for the case of an open economy. It is easy to show that the conclusions in this paper can be extended, with the same qualitative results, to the case of a small open economy, and in particular for the nominal interest rate rule. Consider the case of such a small open economy, with perfect capital mobility. In that case, the nominal exchange rate becomes the price level, and the rate of devaluation the inflation rate. Suppose then that the nominal interest rate is pegged at a level i . Now expression 3.3 becomes the uncovered interest rate parity condition iðtÞ ¼ r þ ^eðtÞ; with r being the rest of the world’s interest rate in terms of foreign currency and eˆ the rate of devaluation (for e being the nominal exchange rate). Then, in the same way that the closed economy pegging the nominal interest rate indirectly pegs the inflation rate but not the price level, in this case it yields control over the rate of devaluation, but not on the level of the exchange rate. Notice the important difference between this system and the case of an exchange rate rule, in which both the level and the rate of change of the nominal exchange rate are exogenously controlled. Suppose now, for example, that there is an unexpected increase in the primary fiscal deficit (that is, a fall in the fiscal surplus). The adjustment then implies a devaluation, but one which takes place automatically, as part of the normal mechanism of the rule, rather than one decreed by the central bank when following an exchange rate rule. An additional comment for the case of the open economy is that an increase in the fixed level of the nominal interest rate results in an initial fall of the nominal exchange rate—in an appreciation of the currency. This is, of course, equivalent to the initial fall in the price level for the case of the closed economy. It is a widely accepted result among both economists and policy makers, but one which is, again, not easy to generate in models with microeconomic foundations.22
Monetary Policy Rules, the Fiscal Theory of the Price Level
3.5
63
Concluding Remarks
We have presented a very simple continuous-time model with microeconomic foundations and used it to analyze some of the questions related to the fiscal theory of the price level and to a policy of pegging the nominal interest rate. We have shown that the hyperinflationary solution as an alternative rational expectations path does not exist (in the context of an exogenous monetary rule) for the quite general case of a positive rate of monetary expansion. Indeed, we see no reason for the usual statement that such a path is the FTPL alternative to the monetary explanation resulting in a steady-state, noninflationary path of prices and real variables. In fact, the hyperinflationary possibility has been shown to exist long before the formulation of the FTPL, and it is the presence of nominal government debt (fiscal theory or not) that allows the elimination of such a path when the rate of money growth is positive. The strict formulation of the FTPL ignores the requirement that in the steady state, in a stationary economy, the rate of inflation needs to be the same as the rate of monetary growth—in terms of our expressions for the case of a monetary rule, not only expression 3.13 needs to hold in the steady state, but so does 3.14. The problem is eliminated under a scheme in which interest on money is transferred to the private sector as a lump-sum transfer, in which case, in the steady state, the price level depends exclusively on the aggregate ðB þ MÞ, but this is done only at the cost of essentially transforming money into bonds. In an economy with one or more nominal assets, in which a rational expectations unique equilibrium is characterized by a unique set of real values of those assets, it will be true that for each of those assets (or arbitrary group of those assets) their real value will be their nominal value divided by the price level. One can then write such an expression, but to assert that from such an equation it follows that the price level is determined by the level of that nominal asset or any combination (sum) of those assets would be fallacious and highly misleading. Of course, nowhere does any proponent of the FTPL assert that, but the use of a single equation (the long-run equilibrium government budget constraint) conveys the impression that at times some elements or variations of the fallacy may be present in the argument. We find that an interest rate pegging rule not only delivers a unique, welldetermined price level, but that some simple experiments generate results which are compatible with, and provide a theoretical explanation of, empirical observations. We have also argued that such a rule seems to have some attractive features as a policy. We have explored a couple of other issues not strictly pertaining to the main purpose of the paper, but which were closely related. In particular, for the case of
64
Leonardo Auernheimer
a monetary rule, we have characterized the short-run temporary hyperinflationary ‘‘solution’’ that has been the response in some countries at various times. Here, we have distinguished clearly—with even the graphic representation contributing to that clarity—the two effects that a one-time increase in the nominal money has on government debt: inflating debt away, when in nominal terms, and the purchase of real debt that is always possible for a positive price of money, including the case of indexed debt. The latter is often neglected, in particular when considering the government’s incentives to orchestrate a surprise hyperinflation. Finally, a short comment is in order regarding levels and rates-of-change considerations involved in various policy rules. It is clear that the usual strict monetary rule implies an exogenous path of nominal money involving both its level and its rate of change. In the case of an open economy, an analogue is the standard exchange rate rule, again fixing both the level and the rate of change (in other words, the rate of devaluation) of the nominal exchange rate.23 In contrast, a nominal interest rate rule, indirectly and via the Fisher equation, fixes the rate of change, but not the level, of prices. In section 2, when considering the case of the usual monetary rule (which sets the path of both the level and the rate of change of the nominal money stock), we made reference to the possibility of a modified monetary rule, fixing only the rate of change of the nominal money stock. In this case, in the context of our model, at each point in time the level of both the money supply and government debt could be decided by the public, and in principle such a rule could yield results very similar to the nominal interest rate rule, provided either that there would be a market incentive mechanism for individuals to engage in open market operations, or that the central bank would take the initiative. What seems to be interesting about this characterization is that both an interest rate rule and the modified monetary rule, as fixing exogenous rates of change or either prices or money, leave a degree of freedom for the market to achieve a steady-state rational expectations unique equilibrium that may not, and in some cases will definitely not, be feasible with policies which exogenously fixed both levels and rates of change. Acknowledgments I am deeply indebted to Carlos A. Ve´gh for many insightful comments and, more importantly, for his interest in my work and our invigorating exchanges for many years. At various times I benefitted from conversations on this topic with Stanley Fischer and Ben McCallum. The discussion in section 3.4 relies on the analysis first presented in Auernheimer and Contreras 1990, 1995.
Monetary Policy Rules, the Fiscal Theory of the Price Level
65
Notes 1. Witness some of the statements in the discussion: ‘‘The thesis of this paper is that the ‘fiscal theory of the price level’ is fatally flawed’’ (Buiter 2002, 459) and, in the abstract of the same paper, ‘‘This paper argues that the ‘fiscal theory of the price level’ has feet of clay.’’ Niepelt (2004) paper title: ‘‘The Fiscal Myth of the Price Level.’’ 2. For a notable exception, see Daniel (2001). 3. It is of course true that a correctly constructed discrete time model should have, as a limit when the interval between dates approaches zero, an equivalent continuous time version, and vice versa, but this is not always easy to secure. See, for example, Carlstrom and Fuerst (2001), where, in the context of a different question, it is shown how ‘‘seemingly minor modifications’’ (whether beginning-of-the-period or end-of-the-period real money balances should enter the utility function) can have the consequence of ‘‘dramatic differences’’ in the conclusions. The authors interpret their results as suggesting that ‘‘as monetary theorists we must be more careful in writing down the basics of our models’’ and that ‘‘there are concerns with continuous-time analyses that simply sweep this fundamental timing issue under the rug.’’ We certainly agree with the first of these conclusions, but much less so with the second. 4. See, for example, Cochrane (2001) for a discussion of the importance of debt maturity in the context of the questions at hand. 5. See Buiter (2002), McCallum (2001), Kocherlakota and Phelan (1999), and Woodford (1995) among others. 6. The equality r ¼ r is not an arbitrary assumption. If behind the assumption of a ‘‘fixed, instantaneously perishable flow endowment’’ there is a fixed, nondepreciable, nonreproduceable aggregate capital stock—the most sensible economic assumption—capital can be exchanged in the market and will have a price in terms of its output. Given the fixed rate of time preference, and a fixed marginal product of capital, the price of capital will adjust so that the interest rate (the ratio of the marginal product of capital and its price in terms of output) turns out to be equal to the rate-of-time preference so as to assure the optimality of the constant level of consumption, dictated by the overall resource constraint. 7. See McCallum (2001), 21. 8. The usual assumption in this literature is b b 0, and this has consequences for the set of admissible adjustment paths. We see no reason for this limitation, which is at variance with much of the literature on the open macroeconomy, for example, in which government can hold positive net assets. Kocherlakota and Phelan (1999) also allow for negative government debt. 9. This will not necessarily preclude discrete changes in either the nominal money supply or nominal bonds as ‘‘conceptual experiments.’’ 10. See (note 13), though, for a subtle qualification to this statement. 11. More specifically (Kocherlakota and Phelan 1999), a Ricardian policy is one in which government sets a fiscal and monetary policy compatible with any possible path of prices. A non-Ricardian policy is one in which government’s setting of the monetary and fiscal variables is compatible with only one path of prices. A typical example of the latter is the current specification of a monetary rule; a typical example of the former would be, in the current case, the setting of the rate of money growth to whichever value is necessary for the long-run equilibrium to obtain, for any possible level of the fiscal surplus and current government debt. For this last case proponents of the FTPL concede the relevance of the traditional monetary analysis. 12. This would not be so, of course, in the case of indexed government debt (i.e., debt denominated in real terms), a case to which we will make reference later.
66
Leonardo Auernheimer
13. The possibility of this instantaneously temporary difference between the demand and the supply of money is the ‘‘subtle qualification’’ alluded to in note 10. 14. The idea of an adjustment process taking place at a single point in chronological time can be interpreted, of course, as describing an adjustment during a very short interval, in a manner that all agents understand as taking place separately from the laws of motion implied by the overall equations—in our case, equations 3.11–3.12. 15. Sargent and Wallace (1981) analyze the case of indexed government debt—that is, denominated in real terms—but their results are also relevant in the case of nominal debt. 16. Although initially triggered via a 150 percent devaluation, followed by the increase in the money supply (rather than an increase in money followed by a rise in all prices, including the exchange rate, as in our analysis), readers familiar with the Latin American experience will probably recognize the analogy of the analysis to the Rodrigazo episode in Argentina in June 1975—the name referring to Celestino Rodrigo, the economics minister at the time. 17. To have an idea of the ‘‘slipperiness’’ of the issue, see, for example, the ‘‘change of heart’’ in the case of one of the proponents of the FTPL in different versions of the same paper (Cochrane 2002). 18. Notice that Kocherlakota and Phelan (1999), in arguing for the plausibility of the monetarist solution (meaning a one-time fall in the price level by the same proportion as the nominal money annihilation), assume a departure from the non-Ricardian policy, with government changing the fiscal surplus to adjust for the increase in real debt as a result of the price fall. 19. In fact, our early work (Auernheimer and Contreras 1990, 1995), which together with Begg and Haque’s (1984) has been credited by a generous colleague as being ‘‘the original exponents of the FTPL’’ (Buiter 2002), was exclusively motivated by the analysis of an interest rate policy pegging, without explicit references to a general theory of price determination. 20. Woodford (1995) recognizes this problem, but dismisses it as irrelevant for the proposition that the aggregate ðB þ MÞ is sufficient to determine the path of prices. 21. It is easy to show that, given assumption 3.10 0 (i.e., the economy is in the range of the monetary demand for which higher inflation translates into higher revenue from money creation), then the new equilibrium point C will always be above the line connecting points A and B, meaning that a fall in prices will be required. 22. Calvo and Ve´gh (1990) also generate this result, through a very different mechanism. 23. Notice that this is equivalent, for the close economy, to the government fixing the price level, and its rate of change, by committing to buy or sell a representative bundle of goods at a preannounced fixed price.
References Auernheimer, L., and Contreras, B. 1990. ‘‘Control of the Interest Rate with a Government Budget Constraint: Price Level Determinacy, and other Results.’’ Mimeo., Texas A&M University. ———. 1995. ‘‘Control de la tasa de interes con restriccion presupuestaria: determinacion de los precios y otros resultados.’’ El Trimestre Economico 62, no. 3: 381–396. Bassetto, M. 2002. ‘‘A Game-Theoretic View of the Fiscal Theory of the Price Level.’’ Econometrica 70, no. 6: 2167–2195. Begg, D., and Haque, B. 1984. ‘‘A Nominal Interest Rate Rule and Price Level Indeterminacy Reconsidered.’’ Greek Economic Review 6: 31–46.
Monetary Policy Rules, the Fiscal Theory of the Price Level
67
Buiter, W. H. 2002. ‘‘The Fiscal Theory of the Price Level: A Critique.’’ Economic Journal 112: 459–480. ———. 2005. ‘‘New Developments in Monetary Economics: Two Ghosts, Two Eccentricities, a Fallacy, a Mirage and a Mythos.’’ Economic Journal 115: C1–C31. Calvo, G. A. 1981. ‘‘Devaluation: Levels versus Rates.’’ Journal of International Economics 11, no. 2: 165– 172. ———. 1985. ‘‘Macroeconomic Implications of the Government Budget: Some Basic Considerations.’’ Journal of Monetary Economics 15, no. 1: 95–112. Calvo, G. A., and Ve´gh, C. A. 1990. ‘‘Interest Rate Policy in a Small Open Economy.’’ IMF Staff Papers 37, no. 4: 753–776. Carlstrom, C. T., and T. S. Fuerst. 1998. ‘‘Real Indeterminacy in Monetary Models with Nominal Interest Rate Distortions: The Problem with Inflation Targets.’’ Working Paper No. 9818, Federal Reserve Bank of Cleveland. ———. 2000. ‘‘The Fiscal Theory of the Price Level.’’ Federal Reserve Bank of Cleveland Economic Review Q1: 22–32. ———. 2001. ‘‘Timing and Real Indeterminacy in Monetary Models.’’ Journal of Monetary Economics 47: 285–298. Christiano, L. J., and Fitzgerald, T. 2000. ‘‘Understanding the Fiscal Theory of the Price Level.’’ Federal Reserve Bank of Cleveland Economic Review 36, no. 2: 1–37. Cochrane, J. 2001. ‘‘Long-Term Debt and Optimal Policy in the Fiscal Theory of the Price Level.’’ Econometrica 69, no. 1: 69–116. ———. 2002. ‘‘Money as Stock: Price Level Determination With No Money Demand.’’ National Bureau of Economic Research Working Paper 7498, Cambridge, MA. Daniel, B. 2001. ‘‘The Fiscal Theory of the Price Level in an Open Economy.’’ Journal of Monetary Economics 48: 293–308. Dupor, B. 2000. ‘‘Exchange Rates and the Fiscal Theory of the Price Level.’’ Journal of Monetary Economics 45, no. 3: 613–630. Kocherlakota, N., and C. Phelan. 1999. ‘‘Explaining the Fiscal Theory of the Price Level.’’ Federal Reserve Bank of Minneapolis Quarterly Review 23, no. 4: 14–23. Leeper, E. 1991. ‘‘Equilibria Under Active and Passive Monetary and Fiscal Policies.’’ Journal of Monetary Economics 27: 129–147. Lucas, R. E. 1978. ‘‘Asset Prices in an Exchange Economy.’’ Econometrica 46, no. 6: 1429–1425. McCallum, B. T. 1997. ‘‘Crucial Issues Concerning Central Bank Independence.’’ Journal of Monetary Economics 39: 99–112. ———. 2001. ‘‘Indeterminacy, Bubbles and the Fiscal Theory of the Price Level.’’ Journal of Monetary Economics 47: 19–30. Metzler, L. A. 1951. ‘‘Wealth, Saving, and the Rate of Interest.’’ Journal of Political Economy 59, no. 2: 93– 116. Niepelt, D. 2004. ‘‘The Fiscal Myth of the Price Level.’’ Quarterly Journal of Economics 119: 277–300. Obstfeld, M., and Rogoff, K. 1983. ‘‘Speculative Hyperinflations in Maximizing Models: Can We Rule Them Out?’’ Journal of Political Economy 91, no. 4: 675–687.
68
Leonardo Auernheimer
Sargent, T., and N. Wallace. 1981. ‘‘Some Unpleasant Monetarist Arithmetic.’’ Federal Reserve Bank of Minneapolis Quarterly Review 5, no. 3. Sims, C. 1994. ‘‘A Simple Model for Study of the Determination of the Price Level and the Interaction of Monetary and Fiscal Policy.’’ Economic Theory 4: 381–399. ———. 2001. ‘‘Fiscal Consequences for Mexico of Adopting the Dollar.’’ Journal of Money, Credit an Banking 33, no. 2: 597–616. Tobin, J. 1974. ‘‘Friedman’s Theoretical Framework.’’ In Milton Friedman’s Monetary Framework, ed. R. J. Gordon, 77–89. Chicago: University of Chicago Press. Woodford, M. 1994. ‘‘Monetary Policy and Price Level Determinacy in a Cash-In-Advance Economy.’’ Economic Theory 4: 345–389. ———. 1995. ‘‘Price Level Determinacy Without Control of a Monetary Aggregate.’’ Carnegie-Rochester Series on Public Policy 43: 1–46. ———. 1998. ‘‘Doing Without Money: Controlling Inflation in a Post-Monetary World.’’ Review of Economic Dynamics 1: 173–219.
II
Monetary and Exchange Rate Policy in Practice
4
Can Inflation Targeting Work in Emerging Market Countries? Frederic S. Mishkin
An important theme in Guillermo Calvo’s work has been that emerging-market economies are very different from advanced economies, and this has important implications that need to be factored in when designing macroeconomic policies. For example, Guillermo (e.g., see Calvo 1999a and 2001 and Calvo and Mendoza 2000) has been quite skeptical of inflation targeting as a monetary policy strategy for emerging-market countries.1 He worries that allowing policy makers too much discretion in the weak institutional environment of emerging-market countries might lead to poor macroeconomic outcomes. I am a known advocate of inflation targeting, but I think that Guillermo’s concerns about inflation targeting’s efficacy in emerging-market countries need to be taken seriously. Before starting it is important to make clear what an inflation targeting regime is all about. It comprises five elements: (1) the public announcement of mediumterm numerical targets for inflation; (2) an institutional commitment to price stability as the primary goal of monetary policy, to which other goals are subordinated; (3) an information-inclusive strategy in which many variables, and not just monetary aggregates or the exchange rate, are used for deciding the setting of policy instruments; (4) increased transparency of the monetary policy strategy through communication with the public and the markets about the plans, objectives, and decisions of the monetary authorities; and (5) increased accountability of the central bank in attaining its inflation objectives. The list should clarify one crucial point about inflation targeting: it entails much more than a public announcement of numerical targets for inflation for the year ahead. This is important in the context of emerging-market countries because many of them routinely report numerical inflation targets or objectives as part of their government’s economic plan for the coming year, and yet their monetary policy strategy should not be characterized as inflation targeting, which requires the other four elements for it to be sustainable over the medium term. In this paper, I explore what additional issues need to be addressed in emerging-market countries to make inflation targeting work for them. I start by
72
Frederic S. Mishkin
outlining why emerging-market economies are so different from advanced economies and then discuss why developing strong fiscal, financial, and monetary institutions is so critical to the success of inflation targeting. Then I look at two emerging-market countries that illustrate what it takes to make inflation targeting work well, Chile and Brazil. I then address a particularly complicated issue for central banks in emerging-market countries who engage in inflation targeting: how they deal with exchange rate fluctuations. The next topic focuses on the IMF’s role in promoting the success of inflation targeting in emerging-market countries. The conclusion from this analysis is that inflation targeting is more complicated in emerging market countries and is thus not a panacea. However, inflation targeting done correctly can be a powerful tool to help promote macroeconomic stability in these countries, as is supported by recent empirical research that finds that inflation performance in emerging-market countries improves after the adoption of inflation targeting, with no adverse effect on output (Battini and Laxton, forthcoming, and Mishkin and Schmidt-Hebbel 2007). 4.1
Why Emerging Market Economies Differ from Advanced Economies
In our paper on the choice of exchange rate regimes (Calvo and Mishkin 2003), Guillermo and I have outlined five fundamental institutional differences for emerging-market countries that must be taken into account to derive sound theory and policy advice. These are •
Weak fiscal institutions
Weak financial institutions, including government prudential regulation and supervision
•
•
Low credibility of monetary institutions
•
Currency substitution and liability dollarization
•
Vulnerability to sudden stops (of capital inflows)
Advanced countries are not immune to problems with their fiscal, financial, and monetary institutions, but there is a major difference in the degree of the problem in emerging-market countries. Such weak institutions make emergingmarket countries very vulnerable to high inflation and currency crises, so that the real value of money cannot be taken for granted. As a result, emerging-market countries face the threat of domestic residents switching to a foreign currency, leading to currency substitution (Calvo and Ve´gh 1996). Currency substitution is likely to be due not only to past inflationary experience, but also to the fact that a currency like the U.S. dollar is a key unit of account for international transactions. This phenomenon has induced the monetary authority to allow banks to offer for-
Can Inflation Targeting Work in Emerging Market Countries?
73
eign exchange deposits. In this fashion, a sudden switch away from domestic and into foreign money need not result in a bank run, since in the presence of foreign exchange deposits such a portfolio shift could be implemented by simply changing the denomination of bank deposits. Otherwise, deposits would be drawn down to purchase foreign exchange, resulting in a bank run. Foreign exchange deposits induce banks—partly for regulatory reasons that prevent them from taking exchange rate risk—to offer loans denominated in foreign currency, usually dollars, leading to what is called liability dollarization. As pointed out in Mishkin (1996) and Calvo (2001), liability dollarization is what leads to an entirely different impact of currency crises on the economy in emerging market versus advanced countries. In emerging-market countries, a sharp real currency depreciation raises the value of liabilities in local currency, thus causing the net worth of corporations and individuals to fall, especially those whose earnings come from the nontradables sector. This serious negative shock to corporations’ and individuals’ balance sheets then increases asymmetric information problems in credit markets, leading to a sharp decline in lending and an economic contraction. Thus, liability dollarization (where the currency mismatch takes place in corporate and household balance sheets) may become a major problem for economies that are relatively closed and highly indebted (this has typically been the case in several emerging-market countries after the capital-inflow episode in the first half of the 1990s; see Calvo, Izquierdo, and Talvi 2002). Under those circumstances, the monetary authority is likely to display ‘‘fear of floating’’ (see Calvo and Reinhart 2002)—a reluctance to allow free fluctuations in the nominal exchange rate—placing an additional constraint on emerging market countries’ monetary policy.2 It should be noted, however, that not all emergingmarket countries suffer from liability dollarization in a serious way (for example, Chile and South Africa. See Eichengreen, Hausmann, and Panizza 2002). A dominant phenomenon in emerging-market countries is a sudden stop, a large negative change in capital inflows, which, as a general rule, appears to contain a large unanticipated component (see Calvo and Reinhart 2000). This phenomenon is mostly confined to emerging market countries. It likely hits only such countries because of their weak fiscal and financial institutions, and it is only recently that sudden stops have been subject to systematic empirical analysis. Preliminary evidence suggests that there is a high degree of bunching of sudden stops across emerging market countries. This is especially evident after the 1998 Russian crisis and the recent Wall Street scandals of Enron and others. This leads to the conjecture that, to a large extent, sudden stops have been a result of factors somewhat external to emerging market countries as a group. They might have been triggered by crisis in one emerging market country, as the Russian crisis illustrates, but contagion across emerging-market countries was likely due to difficulties
74
Frederic S. Mishkin
experienced outside these countries, especially in world financial centers (for evidence about that, see Kaminsky, Reinhart and Ve´gh 2003). To illustrate, a possible contagion mechanism is a liquidity crunch in a financial center, triggered by margin calls following unexpected capital losses, which leads the financial center to dump emerging market securities or to at least not bid for new debt instruments issued by emerging market countries. Calvo (1999b) conjectures that this mechanism could explain the large negative impact that the Russian crisis had on all emerging markets. The effect of sudden stops on individual countries is by no means uniform. In Latin America, for example, Argentina suffered a very serious dislocation, while neighboring Chile escaped relatively unscathed (although Chile did see its growth rate fall by more than 50 percent). In Asia, Korea has had a strong recovery while Indonesia is still reeling from the shock. Tentative analysis suggests that these different outcomes have much to do with initial conditions. Chile had low debt relative to Argentina and did not suffer from liability dollarization, while at the time of crisis most debt instruments in Argentina were denominated in U.S. dollars. On the other hand, even though both Korea and Indonesia suffered from foreign exchange debt, the former was able to socialize much of the financial problem (as a result, Korea’s debt climbed from about 12 to about 33 percent of GDP in 1996– 1998). Once again, debt and currency mismatch appear to have played a crucial role in determining the depth of crisis. In the following sections, we will see how these institutional features make inflation targeting a more complicated exercise in emerging-market economies than in advanced economies. 4.2
Developing Strong Fiscal and Financial Institutions
Fiscal stability is a fundamental necessary condition for inflation control, and hence inflation targeting. A key lesson from the ‘‘unpleasant monetarist arithmetic’’ discussed in Sargent and Wallace (1981) and the recent literature on fiscal theories of the price level (Woodford 1994 and 1995) is that irresponsible fiscal policy puts pressure on the monetary authorities to monetize the debt, thereby producing rapid money growth and high inflation. If fiscal imbalances are large enough, monetary policy eventually becomes subservient to fiscal considerations (so-called fiscal dominance), and an inflation target would have to be abandoned or seriously modified. Fiscal policy is also found to be highly procyclical in emerging market countries, which can also contribute to procyclical monetary policy (Kaminsky, Reinhart, and Ve´gh 2004), and both can contribute to cyclical inflation fluctuations that make an inflation target harder to achieve.
Can Inflation Targeting Work in Emerging Market Countries?
75
Similarly, a safe and sound financial system is also a necessary condition for the success of an inflation-targeting regime. A weak banking system is particularly dangerous. Once a banking system is in a weakened state, a central bank cannot raise interest rates to sustain the inflation target because this will likely lead to a financial system collapse. Not only can this directly cause a breakdown of the inflation targeting regime, but it can also lead to a currency collapse and a financial crisis, which will also erode the control of inflation. When markets recognize the weakness of the banking system, there will be a reversal of capital flows out of the country (a sudden stop) that will result in a sharp depreciation of the exchange rate, which leads to upward pressures on the inflation rate. Moreover, as a result of the currency devaluation that most likely accompanies the monetary expansion, the debt burden of domestic firms denominated in foreign currency rises, while the assets, which are denominated in domestic currency, do so at a much slower pace, thus leading to a decline in net worth. As described in Mishkin (1996), this deterioration in balance sheets then increases adverse selection and moral hazard problems in credit markets, leading to a sharp decline in investment and economic activity, and ultimately a complete collapse of the banking system. The subsequent bailout of the banking system leads to a huge increase in government liabilities, which will have to be monetized in the future (Burnside, Eichenbaum, and Rebelo 2001), thus undermining the inflation-targeting regime. Unfortunately, the scenario outlined here has happened all too often in recent years as evidenced by the twin crises (currency and financial) in Chile in 1982, in Mexico in 1994–1995, in East Asia in 1997, in Ecuador in 1999, and in Turkey in 2000–2001. Fiscal imbalances can also lead to banking and financial crises that will blow out any monetary regime to control inflation. As outlined in Mishkin and Savastano (2001), large budget deficits may force the government to confiscate assets, particularly those in the banking system. This has indeed happened often in Latin America. The suspicion that this might occur would then cause depositors and other creditors to pull their money out of the banking system, and the resulting banking crisis would then cause a contraction of lending and the economy. This has happened several times in Argentina’s checkered history, with the most recent variant occurring in 2001. The Argentine banking system was generally in quite good shape until 2000, even though the economy had been in a recession for several years. The strength of the Argentine banking system was the result of a sophisticated prudential regulatory and supervisory regime put into place after the tequila crisis that made Argentina’s prudential supervision one of the best in the emerging-market world (see Calomiris and Powell 2000). Large budget deficits forced the Argentine government to look for a new source of funds from the
76
Frederic S. Mishkin
banking system, which was primarily foreign-owned. After Domingo Cavallo became minister of the economy in April 2001 and the central bank president, Pedro Pou, was forced to resign, prudential supervisory standards were weakened and banks were both encouraged and coerced into purchasing Argentine government bonds. With the bonds’ value declining as the likelihood of default on this debt increased, banks’ net worth plummeted. The likely insolvency of the banks then led to a classic run on the banks and a full-scale banking crisis by the end of 2001. The result was a collapse of currency, a devastating depression, and an initial surge in inflation. The particular inflation control problems arising from fiscal and financial sector imbalances are of course not unique to emerging-market countries and are also a concern in advanced economies. However, these problems are of a different order of magnitude for emerging-market countries and so must be addressed at the outset if an inflation-targeting regime is to keep inflation under control. Fiscal reforms, which increase transparency of the government budget, and budget rules, which help keep budget deficits from spinning out of control, are needed to prevent the fiscal imbalances that can lead to a collapse of an inflation targeting regime.3 Avoiding financial instability requires several types of institutional reform. First, prudential regulation of the banking and financial system must be strengthened in order to prevent these types of financial crises.4 Second, the safety net provided by the domestic government and the international financial institutions set up by Bretton Woods might need to be limited in order to reduce the moral hazard incentives for banks to take on too much risk.5 Third, currency mismatches need to be limited in order to prevent currency devaluations from destroying balance sheets. Although prudential regulations requiring that financial institutions match up any foreign-denominated liabilities with foreign-denominated assets may help reduce currency risk, they do not go nearly far enough. Even when the banks have equal foreign-denominated (dollar) assets and liabilities, if the banks’ dollar assets are loans to companies who themselves are unhedged, then banks are effectively unhedged against currency devaluations because the dollar loans become nonperforming when the devaluation occurs.6 Thus, limiting currency mismatches may require government policies to limit liability dollarization or at least reduce the incentives for it to occur.7 Fourth, policies to increase the openness of an economy may also help limit the severity of financial crises in emerging-market countries. The reason why openness may affect financial fragility is that businesses in the tradable sector have balance sheets that are less exposed to negative consequences from a currency devaluation when their debts are denominated in foreign currency. Because the goods they produce are traded internationally, they are more likely to be priced in foreign currency. Then a de-
Can Inflation Targeting Work in Emerging Market Countries?
77
valuation that raises the value of their debt in terms of domestic currency is also likely to raise the value of their assets, thus insulating their balance sheets from the devaluation. Moreover, as argued in Calvo, Izquierdo, and Talvi (2002), the more open the economy, the smaller the required real currency depreciation following a sudden stop. Therefore, although firms in the nontradable sector are exposed to balance sheet shocks if they are liability dollarized, the size of the shock is smaller when the economy is more open. One view is that these fiscal and financial reforms must be in place before inflation targeting can even be attempted (Masson, Savastano, and Sharma 1997). However, although fiscal and financial stability are necessary conditions for inflation control, I think the view that these reforms are prerequisites for attempting an inflation-targeting regime in emerging-market countries is too strong. Indeed, Batini and Laxton (forthcoming) find that adoption of inflation targeting in emerging market countries is followed by better inflation performance even when the fiscal and financial reforms are not yet in place. Because inflation targeting commits the government to keeping inflation low, it can be argued (Brash 2000 and Bernanke, Laubach, Mishkin, and Posen 1999) that inflation targeting can help promote fiscal and financial reforms because it becomes clearer that the government must support these reforms if the inflation-targeting regime is to be successful. Also, a commitment to inflation control by the government makes it harder for the government to advocate loose and procyclical fiscal policy as it is clearly inconsistent with the inflation target. However, instituting an inflationtargeting regime by no means insures fiscal and financial reforms. If an inflationtargeting regime is to be sustainable, a commitment to and work on these reforms is required. In fact, after adopting inflation targeting, emerging-market countries do seem to pursue many of the necessary reforms (Batini and Laxton, forthcoming). 4.3
Developing Strong Monetary Institutions
Two underlying monetary institutions are necessary for monetary authorities to keep inflation under control. The first is a public and institutional commitment to price stability as the overriding long-run goal of monetary policy. The institutional commitment can be written into law, as it is in central bank legislation in many countries (most prominent among which are the Economic and Monetary Union [EMU] countries). This is clearly helpful in mandating that the central bank control inflation. Inflation targeting takes this institutional commitment even further by defining price stability—more specifically, by providing a numerical inflation target as the objective for monetary policy. However, it is important to recognize that laws may matter less than the general public’s and politicians’
78
Frederic S. Mishkin
commitment to support price stability. Here past history matters. Many emergingmarket countries have had a history of poor support for the price stability goal and since laws are easily overturned in these countries, it is not clear that laws will be sufficient. The second institutional arrangement necessary for the success of inflation targeting is a public and institutional commitment to the instrument independence of the central bank.8 Instrument independence means that the central bank is prohibited from funding government deficits and must be allowed to set the monetary policy instruments without interference from the government. Additionally, the members of the monetary policy board must be insulated from the political process by giving them long-term appointments and protection from arbitrary dismissal. There is a large literature on central bank independence and the forms that it takes (see Cukierman 1992 and the surveys in Forder 2000 and Cukierman 2006), but again what is written down in the law may be less important than the political culture and history of the country, a point emphasized in Cukierman (2006). The contrast between Argentina and Canada is instructive here. Legally, the central bank of Canada does not appear to be very independent because the government has the ultimate responsibility for the conduct of monetary policy. In the event of a disagreement between the Bank of Canada and the government, the minister of finance can issue a directive that the bank must follow. However, because the directive must be specific and in writing, and because the Bank of Canada is a trusted public institution, a government override of the bank is likely to be highly unpopular and will rarely occur. Thus, in practice, the Bank of Canada is highly independent. In contrast, the central bank of Argentina was highly independent from a legal perspective. However, this did not stop the Argentine government from forcing the resignation of the highly respected central bank president Pedro Pou in April 2001 and replacing him with a president who would do the government’s bidding. Indeed, it is unimaginable in countries like Canada and the United States or in Europe that the public would tolerate the removal of the head of the central bank in such a manner, and indeed I do not know of any case of this happening in recent history.9 Thus a strong legal commitment to central bank independence without genuine public and political support for this independence may not be enough to ensure monetary policies that will focus on inflation control in many emerging market countries. An important advantage of inflation targeting is that it allows the monetary authorities some discretion and flexibility to use monetary policy to cope with shocks to the domestic economy. Indeed, as argued by Fraga, Goldfajn, and Minella (2003), this flexibility is even more important in emerging-market countries because they are subject to larger shocks. Inflation-targeting regimes typically have built-in flexibility to allow them to achieve their inflation target over longer hori-
Can Inflation Targeting Work in Emerging Market Countries?
79
zons, and are better characterized as ‘‘flexible inflation targeting’’ as described in Svensson (1997) and Bernanke, Laubach, Mishkin, and Posen (1999). This advantage of inflation targeting is also its weakness. Critics of inflation targeting, of whom Calvo has been a prominent one (for example, see Calvo 1999a, Calvo 2001, and Calvo and Mendoza 2000), have argued that inflation targeting may allow too much discretion to monetary policy and so may not provide a sufficient nominal anchor. However, as pointed out in Bernanke and Mishkin (1997), inflation targeting should be seen as a regime of ‘‘constrained discretion.’’ Transparent discussions of the conduct of monetary policy can make it more difficult for the central bank to follow overly expansionary monetary policy, while accountability means that the central bank pays a high price if it engages in discretionary policy that leads to high inflation. However, Calvo is quite correct that inflation targeting is no panacea. In order for inflation targeting to work to constrain discretion, it has to be supported by the public and the political process. Inflation targeting can help focus the public debate so that it supports a monetary policy focus on long-run goals such as price stability, as has occurred in many inflation-targeting countries (Bernanke, Laubach, Mishkin, and Posen 1999). However, these benefits require even more transparency and excellent communication skills on the part of the central bank in the more politically complicated environment of emerging market countries. Even excellent communication may not be enough if the political environment is incapable of supporting an independent central bank that focuses on inflation control. 4.4 Two Case Studies of Inflation Targeting in Emerging Market Countries: The Chilean and Brazilian Examples Despite the challenges in getting inflation targeting to work, inflation targeting has been successful in promoting macroeconomic stability in a number of emerging market countries. Here we will look at two case studies, Chile and Brazil, which illustrate that institutional development along the lines discussed earlier has been critical to inflation targeting’s success in the emerging market context.10 Chile is the poster child for inflation-targeting regimes in emerging market countries because it has had great success in lowering inflation to levels found in advanced countries, while at the same time experiencing very rapid economic growth.11 From 1991, when the inflation-targeting regime started, to the present, Chile has been able to lower inflation rates from above 20 percent to around 3 percent. Over the same period, output growth has been very high, averaging 6 percent per year from 1991 to 2002. However, it is important to emphasize that the success of Chile’s inflation-targeting regime has been based on development of the requisite institutions we have discussed here. Over the 1991–2002 period
80
Frederic S. Mishkin
Chile’s budget surplus averaged a little under one percent of Gross Domestic Product (GDP). In addition, due largely to the measures taken in the aftermath of the severe banking crisis in the early 1980s (costing over 40 percent of GDP), Chile developed banking regulation and supervisory practices that are among the best in the emerging market world (Caprio 1998) and are viewed as comparable to those found in advanced countries. As a result, even during the tequila crisis of 1995, the Russian meltdown of 1998, and the current difficulties in Latin America, the soundness of Chile’s financial system has never come into question. The controls on short-term capital inflows have also been cited as another important factor behind the relative stability of the Chilean economy in the 1990s. However, these controls are very controversial and it is not at all clear that they have made a significant contribution to Chile’s success (de Gregorio, Edwards, and Valdes 2000 and Edwards 1999). Chile also has worked on developing strong monetary institutions. In 1989 Chile passed new central bank legislation (which took effect in 1990, just before the start of the inflation-targeting regime) that gave independence to the central bank and mandated price stability as its primary objective. Indeed, as pointed out in Landerretche, Morande´, and Schmidt-Hebbel (1999), Chile only gradually hardened up its inflation-targeting regime over time, with the announced inflation objective initially being interpreted more as official inflation projections rather than formal or ‘‘hard’’ targets. Only after the central bank already had some success with disinflation, by 1994, did the inflation projections become hard targets, with the central bank now accountable for meeting them. Furthermore, until August 1999 Chile had an exchange-rate band around a crawling peg, and in 1998 it came close to fixing its exchange rate for a time by narrowing the exchange-rate band sharply (though this was a mistake that will be discussed in the next section and was reversed later).12 Thus Chile has exhibited some ‘‘fear of floating’’ along the lines described by Calvo and Reinhart (2002). Indeed, Chile’s inflation-targeting regime should be seen as very evolutionary, with Chile going to a full-fledged inflation-targeting regime only in May 2000 (see Mishkin and Savastano 2002). Brazil implemented inflation targeting shortly after the real collapsed in January 1999. After much confusion, in early February the new central bank president, Arminio Fraga, announced that Brazil would soon be adopting an inflationtargeting strategy. At the same time, Fraga put his money where his mouth was by increasing the interbank policy interest rate by 600 basis points, to 45 percent. On June 21, the president of Brazil issued a decree instituting an inflationtargeting framework. This framework included all the features of a full-fledged inflation-targeting regime, including (1) the announcement of multiyear inflation targets (with explicit numerical targets for the twelve-month rate of inflation in
Can Inflation Targeting Work in Emerging Market Countries?
81
the years 1999, 2000, and 2001, and a commitment to announce the targets for 2002 two years in advance; (2) assigning the National Monetary Council the responsibility for setting the inflation targets and tolerance ranges based on a proposal by the minister of finance; (3) giving the central bank full power to implement the policies needed to attain the inflation targets; (4) establishing procedures to increase the central bank’s accountability (specifically, if the target range is breached, the central bank president would have to issue an open letter to the minister of finance explaining the causes of the deviation, the measures that would be taken to eliminate it, and the time it would take to get inflation back inside the tolerance range); and (5) taking actions to improve the transparency of monetary policy (concretely, the central bank was requested to issue a quarterly inflation report modeled after that produced by the Bank of England). Brazil’s adoption of inflation targeting was not preceded by prior development of all the fiscal, financial, and monetary reforms discussed earlier. First, the 1999 collapse of the real can be attributed to the inability of the Brazilian government to put its fiscal house in order: the currency crisis was prompted by a moratorium on debt payments instituted at the beginning of January by the governor of the Brazilian state of Minas Gerais. Brazil did then take steps to improve its fiscal balances under its program with the IMF, but there were still doubts about the durability of its fiscal reforms (see Mishkin and Savastano 2002.) On the other hand, Brazil did have a strong banking system because it had undergone a major restructuring following the banking crisis of 1994–1996 (see Caprio and Klingebiel 1999.) The independence of Brazil’s central bank and the commitment to price stability, however, were not clear cut. Both were based on a presidential decree and confidence in a superb central bank president, but not on a more formal commitment based on legislation. To the surprise of many, the inflation-targeting strategy seemed to work. The initial inflation targets were set at 8 percent for 1999, 6 percent for 2000, and 4 percent for 2001, with a tolerance range of G2 percent. There was a remarkably small pass-through from the large depreciation of the real (which fell by 45 percent on impact and thereafter stabilized at 30 percent below its pre-devaluation level). For several months, the output contraction was contained (in fact, annual GDP grew by almost 1 percent in 1999), Brazil was not cut off from external financing—though there was some arm twisting involved—and there were no major bank runs. By March 1999, asset prices had started to recover, the real appreciated, and the central bank found room to lower interest rates—which it did, quite aggressively (from a high of 45 percent to below 20 percent in a sevenmonth period). Inflation and the exchange rate remained subdued through October, when the monthly inflation rate rose to 1.2 percent, the largest monthly increase since June 1996, and the exchange rate crossed, briefly, the then-critical
82
Frederic S. Mishkin
mark of R$2.00 per U.S. dollar. In the event, inflation in 1999 reached 8.9 percent, above the 8 percent target for the year but well within the 2 percent tolerance range; during 2000 inflation continued falling, and closed the year right at the 6 percent mid-point target set by the central bank in mid-1999. However, in 2001, the inflation rate exceeded the 4 percent inflation target by more than the 2 percent tolerance range, ending up at 7.7 percent. The weakness of some aspects of the institutional framework for fiscal and monetary policy did come back to haunt the inflation-targeting regime in Brazil. In the run-up to the presidential election in October 2002, the market had concerns that the front-runner, Lula, would weaken fiscal and monetary institutions. Lula had made statements seeming to indicate that once in office he would encourage fiscal policy to be highly expansionary and would not take steps to prevent a possible default on Brazil’s foreign debt. He also indicated that he would not reappoint the highly respected president of the central bank, Arminio Fraga. Thus Lula’s commitment to the independence of the central bank, to price stability, and to the inflation-targeting regime was far from clear. Not surprisingly, the lack of market confidence in Lula, who was elected President, led to a sharp depreciation of the Brazilian real and a sharp upward spike in inflation to 12.5 percent, which substantially overshot the inflation target of 3.5 percent for 2002. The impact of the Brazilian election on the inflation-targeting regime illustrates that weak fiscal and monetary institutions are likely to create severe problems for an inflation-targeting regime. However, the response of the Brazilian government and central bank to the overshoots of the inflation targets illustrates that inflation targeting can help keep inflation under control in the face of big shocks like the 2002 real depreciation. As noted earlier, under the presidential decree that gave birth to inflation targeting, the Banco Central do Brasil was required to submit an open letter to the ministry of finance explaining the causes of the breach of the inflation target and what steps would be taken to get the inflation rate back down again. Given the large overshoot of the inflation targets, central bank credibility was on the line and the Banco Central do Brasil handled it well. It explained how it would modify its inflation targets and what inflation path it would shoot for in the future with a very high degree of transparency (see Fraga, Goldfajn, and Minella 2003). First it explained why the exchange rate had overshot, and made explicit estimates of the size of the shocks and their persistence. It estimated the regulated-price shock to be 1.7 percent and estimated the inertia from past shocks to be 4.2 percent, of which two-thirds was to be accepted, resulting in a further adjustment of 2.8 percent. Then the central bank added these two numbers to the previously announced target of 4 percent to get an adjusted inflation target for 2003 of 8.5 percent (¼ 4% þ 1.7% þ 2.8%). The adjusted target was then announced in the
Can Inflation Targeting Work in Emerging Market Countries?
83
open letter sent to the minister of finance in January 2003, which also explained that getting to the nonadjusted target of 4 percent too quickly would entail far too high a loss of output. Specifically, the letter indicated that an attempt to achieve an inflation rate of 6.5 percent in 2003 would be expected to entail a decline of 1.6 percent of GDP, while trying to achieve the nonadjusted target of 4 percent would be expected to lead to an even larger decline of GDP of 7.3 percent. The procedure followed by the Banco Central do Brasil in this instance was a textbook case for central bank response to shocks in emerging-market countries. First, the procedure had tremendous transparency, both in articulating why the initial inflation target was missed, but also in how the central bank was responding to the shock and its plans to return to its longer-run inflation goal. This degree of transparency helped minimize the loss of credibility and the need to adjust the short-term inflation target. Second, the central bank recognized that not adjusting the inflation target was just not credible because the market and the public clearly recognized that inflation would overshoot the initial target. Thus adjusting the target was absolutely necessary to retain credibility, because to do otherwise would have just signaled to the markets that the central bank was unwilling to be transparent. Third, by discussing alternative paths for the inflation rate and why the particular path using the adjusted target was chosen, the central bank was able to demonstrate that it was not what Mervyn King (1996) has referred to as an ‘‘inflation nutter’’ who only cares about controlling inflation and not about output fluctuations. By outlining that lower inflation paths would lead to large output losses, the Banco Central do Brasil demonstrated that it was not out of touch with the concerns of the public because it indeed does care about output losses, just as the public and the politicians do. This procedure is then likely to promote public and political support for central bank independence. The outcome from this particular episode has been favorable. After the initial spike, Brazil’s inflation rate and interest rates have been coming down rapidly. From its level of 12.5 percent in 2002, the inflation rate fell to 9.3 percent by the end of 2003, which is within the central bank tolerance range for the adjusted inflation target of 8.5 percent. Market expectations of inflation also dropped dramatically and suggested that the market expected the central bank to meet its inflation targets in 2004. The interbank policy interest rate also fell from 26.5 percent in June 2003 to 16.3 percent in February 2004. Also, after the initial decline in GDP, the Brazilian economy began growing again. The Brazilian central bank was not able to do this on its own. As seen in this discussion, an important component of a successful response to an inflation shock of this magnitude is growing confidence in government fiscal policy. President Lula surprised many of his detractors by supporting measures to maintain fiscal discipline. He supported and got legislation passed in August 2003 to reform the
84
Frederic S. Mishkin
public pension system to make it fiscally sustainable. His government also pursued conservative spending policies that resulted in a primary budget surplus in 2003 of 4.3 percent of GDP, which was above the target of 4.25 percent requested by the IMF. Of course, there are still concerns about how Brazil will proceed in the future. A new central bank law that would strengthen the Banco Central do Brasil’s independence and provide a firmer institutional basis for inflation targeting has not yet been passed. Also, given the pressures on the Lula government from the left, continuing fiscal responsibility is not completely assured. The very high real interest rates in Brazil, which are the highest in all of the inflation-targeting countries, suggests that the credibility of the inflationtargeting regime in Brazil is still far from complete and so the central bank needs to pursue a highly restrictive monetary policy in order to keep inflation under control. The examples of Chile and Brazil illustrate that inflation targeting is indeed feasible in emerging-market economies, despite their more complicated political and economic environment. Inflation targeting has been able to provide a strong nominal anchor that can keep inflation expectations in check. However, this requires not only a focus on good communication and transparency by the central bank, but also supportive policies to develop strong fiscal, financial, and monetary institutions. The Brazilian case (and the previously cited evidence in Batini and Laxton, forthcoming) suggests that these policies do not have to be fully in place when inflation targeting is adopted for it to produce good macroeconomic outcomes, but it is important to have a strong commitment to develop these institutions and a continuing vigilance that fiscal, financial, and monetary institutions will continue to be strengthened. 4.5
Dealing with Exchange Rate Fluctuations
The special features of emerging-market economies outlined in the first section of the paper suggests that emerging-market countries are likely to have greater concerns about exchange rate fluctuations than advanced countries do. Because of their past history and lower credibility in their ability to keep inflation under control, emerging-market countries are more likely to find that depreciations lead to a rise in inflation as a result of the pass-through from higher import prices and greater demand for exports. In addition, liability dollarization means that depreciations are particularly dangerous because they can trigger a financial crisis along the lines suggested in Mishkin (1996, 1999). These countries have much of their debt denominated in foreign currency and when the currency depreciates, this increases the debt burden of domestic firms. Since assets are typically denominated in domestic cur-
Can Inflation Targeting Work in Emerging Market Countries?
85
rency and so do not increase in value, there is a resulting decline in net worth. This deterioration in balance sheets then increases adverse selection and moral hazard problems, which leads to financial instability and a sharp decline in investment and economic activity. This mechanism explains why the currency crises in Chile in 1982, Mexico in 1994–1995, East Asia in 1997, Ecuador in 1999, Turkey in 2000–2001 and Argentina in 2001–2002 pushed these countries into full-fledged financial crises, which had devastating effects on their economies. The potential devastating impact of currency depreciation on the financial system and the possibility that it will lead to a burst of inflation provides a reason for central banks in emerging-market countries not to pursue benign neglect of exchange rates even when they are inflation targeting. In order to prevent sharp depreciations of their currencies, which can destroy balance sheets and trigger a financial crisis, central banks in these countries may have to smooth exchange rate fluctuations. However, there is a danger that monetary policy, even under an inflation-targeting regime, may put too much focus on limiting exchange rate movements. The possibility of ‘‘fear of floating’’ (Calvo and Reinhart 2002) when emerging-market countries engage in inflation targeting is a real one. The first problem with too strong a focus on limiting exchange rate movements is that it runs the risk of transforming the exchange rate into a nominal anchor that takes precedence over the inflation target. This has been a problem recently in Hungary, a transition country that has an exchange rate target as part of its inflation-targeting regime ( Jonas and Mishkin 2005). For example, when the Hungarians adopted inflation targeting in July 2001, they retained an exchange rate band of G15 percent. Pursuing two nominal objectives could result in a situation where one objective will need to be given preference over the second, but without a clear guidance on how such conflict would be resolved. This is likely to make monetary policy less transparent and hinder the achievement of the inflation target. Indeed, this is what has happened in Hungary. In January 2003, the forint appreciated to the upper end of the band, and speculation about the revaluation of the parity resulted in a sharp acceleration of capital inflows, forcing the National Bank of Hungary to respond by cutting interest rates by two percentage points and intervening heavily in the foreign exchange market. The National Bank of Hungary is reported to have bought more than five billion euros, increasing international reserves by 50 percent and base money by 70 percent.13 Even though the National Bank of Hungary subsequently began to sterilize this huge injection of liquidity, market participants then assumed that maintaining the exchange rate band would have a priority over the inflation target and expected inflation in 2003 to exceed the National Bank of Hungary’s inflation target by 5 percentage points.14 The outcome was that in 2003 Hungary did indeed overshoot its target, but by a somewhat smaller 2.2 percentage points.15
86
Frederic S. Mishkin
The second problem resulting from too strong a focus on limiting exchange rate fluctuations is that the impact of changes in exchange rates on inflation and output can differ substantially depending on the nature of the shock that causes the exchange rate movement. Different monetary policy responses therefore depend on the nature of the shock. If the domestic currency depreciates because of a pure portfolio shock, inflation is likely to rise and the appropriate response to keep it under control is to tighten monetary policy and raise interest rates. In emergingmarket countries that have substantial liability dollarization, tightening monetary policy to prevent a sharp depreciation may be even more necessary to avoid financial instability (for the reasons mentioned). On the other hand, if the exchange rate depreciation occurs because of real shocks, the impact is less likely to be inflationary and a different monetary policy response is warranted, but even here it depends on the nature of the shock. A negative terms-of-trade shock, which lowers demand for exports, reduces aggregate demand and is thus likely to be deflationary. In this situation, the correct interest rate response is to lower interest rates to counteract the drop in aggregate demand, and not to raise interest rates. The mistakes that the Chilean central bank made in 1998 illustrate how serious the second problem can be. Chile’s inflation-targeting regime also included a focus on limiting exchange rate fluctuations by having an exchange rate band with a crawling peg that was (loosely) tied to lagged domestic inflation. Instead of easing monetary policy in the face of the negative terms-of-trade shock, the central bank raised interest rates sharply and even narrowed its exchange rate band. In hindsight, these decisions were a mistake: the inflation target was undershot and the economy entered a recession for the first time in the 1990s.16 With this outcome, the central bank came under strong criticism for the first time since it had adopted its inflation-targeting regime in 1990, weakening support for the independence of the central bank and its inflation-targeting regime. During 1999, the central bank did reverse course, easing monetary policy by lowering interest rates and allowing the peso to decline, and in May 2000 it revised its inflation-targeting regime to reduce its focus on exchange rates. The conclusion from the discussion here is that there is a rationale for central banks in emerging-market countries to smooth exchange rates, but they can go to far. To cope with potential problems of financial instability but preserve the focus on inflation control, central banks could increase the transparency of any intervention in the foreign exchange market by making it clear to the public that the intervention’s purpose is to smooth excessive exchange rate fluctuations and not to prevent the exchange rate from reaching its market-determined level over longer horizons. However, continuing exchange market interventions, particularly unsterilized ones, are likely to be counterproductive because they are not transparent. Instead, exchange rate smoothing via changes in the interest rate instrument will be more transparent and indicate that the nominal anchor continues
Can Inflation Targeting Work in Emerging Market Countries?
87
to be the inflation target, not the exchange rate. Central banks could also explain to the public the rationale for exchange rate intervention in a manner analogous to that for interest-rate smoothing—that is, as a policy aimed not at resisting market-determined movements in an asset price, but at mitigating potentially destabilizing effects of abrupt and sustained changes in that price. It is also important for central banks to recognize that the pass-through from exchange rate changes to inflation is likely to be regime dependent. After a sustained period of low inflation engineered by an inflation-targeting regime, the effect of the exchange rate on the expectations-formation process and price-setting practices of households and firms in the economy is likely to fall.17 Thus, inflation targeting is likely to help limit the pass-through from exchange rates to inflation and thus the view that a currently high pass-through is a barrier to successful inflation targeting is unwarranted. 4.6
How Can the IMF Help?
A theme in both this and my recent paper with Calvo (Calvo and Mishkin 2003) is that developing strong fiscal, financial, and monetary institutions is the key to successful macroeconomic policy. One way the IMF can help emerging-market countries who choose to inflation target is to provide them with better incentives to develop these institutions. Instead of trying to impose a large number of conditions on countries, as when the IMF lending program is put into effect, the IMF can provide the right incentives by being more willing to extend programs ex ante to countries who are making a serious attempt at reform, while saying no to governments who are unwilling to do so. In addition, the IMF conditions for evaluating monetary policy under its programs can be modified to reflect the special features of inflation-targeting regimes. In the past, a key element of IMF conditionality was ceilings on the growth rate of net domestic assets. Unfortunately, net domestic assets conditionality, which is derived under the IMF’s financial programming framework, is based on an outdated theory, the monetary approach to the balance of payments (see Mussa and Savastano 1999), which requires that the growth rate of monetary aggregates is closely linked to inflation. However, the link between monetary aggregates and inflation is almost always found to be very weak when inflation rates are reasonably low, as is the case for emerging market countries that have adopted inflation targeting. As a result, targets for net domestic asset targets are likely to lead to inappropriate setting of monetary policy instruments and are likely to decrease monetary policy transparency. In an inflation-targeting regime, it seems natural to replace net domestic asset conditionality with assessment of the country’s inflation performance. Indeed, this is what the IMF moved to in evaluating monetary policy under its program
88
Frederic S. Mishkin
for Brazil when inflation targeting was adopted in 1999. The IMF program has conducted quarterly reviews on how well Brazil has done in meeting its inflation targets, but there is still a problem in that the IMF evaluation is essentially backward-looking (Blejer, Leone, Rabanal, and Schwartz 2001). Inflation targeting is inherently forward-looking, so the issue arises as to how IMF conditionality might be modified to be more forward-looking. One approach would be for the IMF to monitor monetary policy institutions. Specifically, the IMF conditions could focus on the degree of central bank independence, whether the central bank mandate focuses on price stability as the long-run overriding goal of monetary policy, and whether transparency and accountability of the central bank is high. As part of this monitoring, the IMF could conduct a careful assessment of central bank procedures including the legitimacy of its forecasting process and whether it provides adequate explanations for misses of its inflation targets. In a sense this shift in approach is similar to the shift in approach that has occurred in bank supervision in recent years. In the past, bank supervision was also quite backward-looking in that it focused on the current state of banks’ balance sheets. However, this backward-looking approach is no longer adequate in today’s world, in which financial innovation has produced new markets and instruments that make it easy for banks and their employees to make huge bets easily and quickly. In this new financial environment, a bank that is quite healthy at a particular point in time can be driven into insolvency extremely rapidly by trading losses, as forcefully demonstrated by the failure of Barings in 1995. Thus bank examinations have now become far more forward-looking and place much greater emphasis on evaluating the soundness of a bank’s management processes with regard to controlling risk. Similarly, the IMF could shift its conditionality to focus on the management processes in central banks to keep inflation under control. 4.7
Concluding Remarks
It has been my honor to have Guillermo Calvo as a close friend for over twenty years, ever since he helped recruit me to come to Columbia University. Guillermo has always cared deeply about applying economic analysis to promote better outcomes in emerging-market countries and his work has been an inspiration to me. This paper expands on many themes in Guillermo’s research. The bottom line from the analysis here is that inflation targeting can be an effective tool for emerging market countries to manage their monetary policy. However, to ensure that inflation targeting produces superior macroeconomic outcomes, emerging market-countries would benefit by focusing even more attention on institutional development, while international financial institutions like the IMF can help by providing these countries with better incentives to engage in this development.
Can Inflation Targeting Work in Emerging Market Countries?
89
The hope is that this will lead to better economic performance in these countries over the coming years. Notes 1. Although I suspect that Guillermo may be more favorable now, but you will have to ask him. 2. The main concern is fluctuations in the real exchange rate but, as is well known (see Mussa 1986), the real exchange rate shows high correlation with the nominal exchange rate. 3. For a discussion of fiscal reforms, see Perry, Whalley, and McMahon (2000) and Tanzi (2000). 4. See Mishkin (2003) for a description of which financial policies help prevent financial crises in emerging market countries. 5. For example, recent research, summarized in Demirgu¨c¸-Kunt and Kane (2002), suggests that the spread of deposit insurance in emerging market countries has made banking crises more likely. Other—rather mixed—evidence about the incidence of moral hazard at the international level is provided by Dell’Ariccia, Go¨dde, and Zettelmeyer (2000) and Jeanne and Zettelmeyer (2001). 6. See Mishkin (1996) for how this occurred in Mexico and Garber (1999) for a discussion of how the prudential regulations for Mexican banks requiring a matched book did not protect them from currency risk. 7. Recent papers by Caballero and Krishnamurthy (2002) and Jeanne (2002) suggest that dedollarization may require a major overhaul of the domestic financial sector. 8. There is an important distinction between goal and instrument independence (Debelle and Fischer 1994 and Fischer 1994). Instrument independence is the ability of the central bank to set monetary policy instruments without government interference, while goal independence means that the monetary authorities set the goals for monetary policy. The standard view in the literature is that central banks should have instrument but not goal independence. 9. However, except for the Great Depression, advanced countries have not been hit by equally large shocks as in Argentina. Thus the stability of the central bank in advanced countries may be partly explained by the size of the shocks and, in particular, the general absence of sudden stops. 10. For an overall comparison of the inflation targeting experience in Chile and Brazil, see SchmidtHebbel and Werner (2002). 11. For those who don’t know this expression, a poster child is a beautiful but disadvantaged child put on a poster (or advertisement) for a charity in order to stimulate donations. 12. See Mishkin and Savastano (2002). 13. See Morgan (2003). 14. Analysts have interpreted this as an evidence that the National Bank of Hungary is determined to maintain the currency band even at the cost of temporary higher inflation. See IMF (2002). 15. Israel also had a fairly tight exchange rate band during the early years of its inflation targeting regime and this caused it to experience similar difficulties. See Bernanke, Laubach, Mishkin, and Posen (1999) and Leiderman (2000). 16. Given its location in Latin America, Chile’s central bank did have to worry about a loss of credibility from exchange rate deprecation. Also, because Chile encountered a sudden stop of capital inflows at the time, the ability of the Chilean central bank to pursue countercyclical policy was limited.
90
Frederic S. Mishkin
However, although lowering interest rates in 1998 may not have been as attractive an option as it was in a country like Australia, which responded to the terms-of-trade shock by easing monetary policy, the sharp rise in the policy interest rate in 1998 was clearly a policy mistake. 17. See Muinhos (2001), Frenkel (2002), and Minella, de Freitas, Goldfajn, and Muinhos (2002, 2003).
References Batini, Nicoletta, and Douglas Laxton. Forthcoming. ‘‘Under What Conditions Can Inflation Targeting Be Adopted? The Experience of Emerging Markets.’’ In Monetary Policy Under Inflation Targeting, eds. Frederic S. Mishkin and Klaus Schmidt-Hebbel. Santiago: Central Bank of Chile. Bernanke, B., T. Laubach, F. Mishkin, and A. Posen. 1999. Inflation Targeting: Lessons from the International Experience. Princeton, NJ: Princeton University Press. Bernanke, Ben S., and Frederic S. Mishkin. 1997. ‘‘Inflation Targeting: A New Framework for Monetary Policy?’’ Journal of Economic Perspectives 11, no. 2: 97–116. Blejer, Mario I., Alfredo M. Leone, Pau Rabanal, and Gerd Schwartz. 2001. ‘‘Inflation Targeting in the Context of IMF-Supported Adjustment Programs.’’ Working Paper no. 01/31, IMF, Washington, D.C. Brash, Donald T. 2000. ‘‘Inflation Targeting in New Zealand, 1988–2000.’’ Speech to the Trans-Tasman Business Cycle, Melbourne, Australia, February 9. Burnside, Craig, Martin Eichenbaum, and Sergio Rebelo. 2001. ‘‘Prospective Deficits and the Asian Currency Crisis.’’ Journal of Political Economy 109, no. 6: 1155–1197. Caballero, Ricardo J., and Arvind Krishnamurthy. 2002. ‘‘Excessive Dollar Debt: Financial Development and Underinsurance.’’ Mimeo., MIT, Cambridge, MA. Calomiris, Charles, and Andrew Powell. 2000. ‘‘Can Emerging Market Bank Regulators Establish Credible Discipline? The Case of Argentina, 1992–1999.’’ Working Paper No. 7715, NBEK, Cambridge, MA. Calvo, Guillermo A. 1999a. ‘‘Capital Markets and the Exchange Rate.’’ Mimeo., University of Maryland. Calvo, Guillermo A. 1999b. ‘‘Contagion in Emerging Markets: When Wall Street is the Carrier.’’ Mimeo., University of Maryland. Calvo, Guillermo A. 2001. ‘‘Capital Markets and the Exchange Rate: With Special Reference to the Dollarization Debate in Latin America.’’ Journal of Money, Credit, and Banking 33, no. 2: 312–334. Calvo, Guillermo A., Alejandro Izquierdo, and Ernesto Talvi. 2002. ‘‘Sudden Stops, the Real Exchange Rate and Fiscal Sustainability: Argentina’s Lessons.’’ Working Paper No. 469, IDB, Washington, D.C. Calvo, G., and E. Mendoza. 2000. ‘‘Capital-Market Crises and Economic Collapse in Emerging Markets: An Informational-Frictions Approach.’’ Mimeo., Duke University and University of Maryland. Calvo, Guillermo, and Frederic S. Mishkin. 2003. ‘‘The Mirage of Exchange Rate Regimes for Emerging Market Countries.’’ Journal of Economic Perspectives 17, no. 4: 99–118. Calvo, Guillermo A., and Carmen M. Reinhart. 2000. ‘‘When Capital Flows Come to a Sudden Stop: Consequences and Policy.’’ In Reforming the International Monetary and Financial System, eds. Peter B. Kenen and Alexander K. Swoboda, 395–427. Washington, D.C.: IMF. Calvo, Guillermo A., and Carmen M. Reinhart. 2002. ‘‘Fear of Floating.’’ Quarterly Journal of Economics 117, no. 2: 379–408.
Can Inflation Targeting Work in Emerging Market Countries?
91
Calvo, Guillermo A., and Carlos A. Ve´gh. 1996. ‘‘From Currency Substitution to Dollarization and Beyond: Analytical and Policy Issues.’’ In Money, Exchange Rates, and Output, ed. Guillermo A. Calvo, 153–175. Cambridge, MA: MIT Press. Caprio, Gerald, Jr. 1998. ‘‘Banking on Crises: Expensive Lessons from Recent Financial Crises.’’ Working Paper No. 1979, World Bank, Washington, D.C. Caprio, Gerald Jr., and D. Klingebiel. 1999. ‘‘Episodes of Systemic and Borderline Financial Crises.’’ Mimeo., World Bank, Washington, D.C. Cukierman, Alex. 1992. Central Bank Strategy, Credibility, and Independence: Theory and Evidence. Cambridge, MA: MIT Press. Cukierman, Alex. 2006. ‘‘Central Bank Independence and Monetary Policymaking Institutions: Past, Present, and Future.’’ Working Paper No. 360, Central Bank of Chile. Debelle, Guy, and Stanley Fischer. 1994. ‘‘How Independent Should a Central Bank Be?’’ In Goals, Guidelines, and Constraints Facing Monetary Policymakers, Conference Series 38, ed. Jeffrey C. Fuhrer, 195–221. Boston: Federal Reserve Bank of Boston. De Gregorio, Jose, Sebastian Edwards, and Rodrigo O. Valdes. 2000. ‘‘Controls on Capital Inflows: Do They Work?’’ Working Paper No. w7645, NBER, Cambridge, MA. Dell’Ariccia, Giovanni, Isabel Go¨dde, and Jeromin Zettelmeyer. 2000. ‘‘Moral Hazard and International Crisis Lending: A Test.’’ Mimeo., International Monetary Fund, Washington, D.C. Demirgu¨c¸-Kunt, Asli, and Edward J. Kane. 2002. ‘‘Deposit Insurance around the Globe: Where Does it Work?’’ Journal of Economic Perspectives 16, no. 2: 175–195. Edwards, Sebastian. 1999. ‘‘How Effective are Controls on Capital Inflows? An Evaluation of Chile’s Experience.’’ Mimeo., UCLA. Eichengreen, Barry, Ricardo Hausmann, and Ugo Panizza. 2002. ‘‘Original Sin: The Pain, the Mystery, and the Road to Redemption.’’ Paper presented at the conference Currency and Maturity Matchmaking: Redeeming Debt from Original Sin, IDB, Washington, D.C., November 21–22. Fischer, Stanley. 1994. ‘‘Modern Central Banking.’’ In The Future of Central Banking: The Tercentenary Symposium of the Bank of England, eds. Forrest Capie, Charles A. E. Goodhart, Stanley Fischer, and Norbert Schnadt. Cambridge, MA: Cambridge University Press. Forder, James. 2000. ‘‘Central Bank Independence and Credibility: Is There a Shred of Evidence? Review.’’ International Finance 3, no. 1: 167–185. Fraga, Arminio, Goldfajn Ilan, and Andre´ Minella. 2003. ‘‘Inflation Targeting in Emerging Market Economies.’’ In NBER Macroeconomics Annual 2003, eds. Mark Gertler and Kenneth Rogoff, 365–400. Cambridge, MA: MIT Press. Frenkel, Jacob. 2002. ‘‘The Transition to Inflation Targeting.’’ In Stabilization and Monetary Policy: The International Experience, Banco de Mexico, 103–111. Mexico City: Banco de Mexico. Garber, Peter. 1999. ‘‘Hard-Wiring to the Dollar: From Currency Board to Currency Zone.’’ In Global Markets Research 2, no. 1: 67–76. Jeanne, Olivier. 2002. ‘‘Why Do Emerging Economies Borrow in Foreign Currency?’’ Paper presented at the conference Currency and Maturity Matchmaking: Redeeming Debt from Original Sin, IDB, Washington, D.C., November 21–22. Jeanne, Olivier, and Jeromin Zettelmeyer. 2001. ‘‘International Bailouts, Moral Hazard, and Conditionality.’’ Economic Policy 33: 409–432.
92
Frederic S. Mishkin
Jonas, Jiri, and Frederic S. Mishkin. 2005. ‘‘Inflation Targeting in Transition Countries: Experience and Prospects.’’ In Inflation Targeting, Michael Woodford, ed., 353–413. Chicago: University of Chicago Press. Kaminsky, Graciela L., Carmen M. Reinhart, and Carlos A. Ve´gh. 2003. ‘‘The Unholy Trinity of Financial Contagion.’’ Journal of Economic Perspectives 17, no. 4: 51–74. Kaminsky, Graciela L., Carmen M. Reinhart, and Carlos A. Ve´gh. 2004. ‘‘When it Rains It Pours: Procyclical Macropolicies and Capital Flows.’’ In NBER Macroeconomics Annual 2004, eds. Ben S. Bernanke and Mark Gertler, 11–53. Cambridge, MA: NBER. King, Mervyn. 1996. ‘‘How Should Central Banks Reduce Inflation? Conceptual Issues.’’ In Achieving Price Stability, Federal Reserve Bank of Kansas City, MO. 53–91. Landerretche, Oscar, Felipe Morande´, and Klaus Schmidt-Hebbel. 1999. ‘‘Inflation Targets and Stabilization in Chile.’’ Working Paper No. 55, Central Bank of Chile. Leidermann, Leonardo, and Gil Byfman. 2000. ‘‘Inflation Targeting Under a Crawling Band Exchange Rate Regime: Lessons from Israel.’’ In Inflation Targeting in Practice: Strategic and Operational Issues and Application to Emerging Market Economies, eds. Mario I. Blejer, Alainlze, Alfredo M. Leone, and Sergio Werlang, 70–79. Washington, D.C.: International Monetary Fund. Masson, P., M. Savastano, and S. Sharma. 1997. ‘‘The Scope for Inflation Targeting in Developing Countries.’’ Working Paper No. 97/130, IMF, Washington, D.C. Minella, Andre´, Paulo Springer de Freitas, Ilan Goldfajn, and Marcelo Kfoury Muinhos. 2002. ‘‘Inflation Targeting in Brazil: Lessons and Challenges.’’ Working Paper 53, Banco Central do Brazil. ———. 2003. ‘‘Inflation Targeting in Brazil: Constructing Credibility under Exchange Rate Volatility.’’ Working Paper 77, Banco Central do Brazil. Mishkin, Frederic S. 1996. ‘‘Understanding Financial Crises: A Developing Country Perspective.’’ In Annual World Bank Conference on Development Economics, eds. Michael Bruno and Boris Pleskovic, 29– 62. Washington, D.C.: World Bank. Mishkin, Frederic S. 1999. ‘‘Lessons from the Asian Crisis.’’ Journal of International Money and Finance 18, no. 4: 709–723. Mishkin, Frederic S. 2003. ‘‘Financial Policies and the Prevention of Financial Crises in Emerging Market Countries.’’ In Economic and Financial Crises in Emerging Market Countries, ed. Martin Feldstein, 93– 130. Chicago: University of Chicago Press. Mishkin, Frederic S., and Miguel Savastano. 2001. ‘‘Monetary Policy Strategies for Latin America.’’ Journal of Development Economics 66, no. 2: 415–444. Mishkin, Frederic S., and Miguel Savastano. 2002. ‘‘Monetary Policy Strategies for Emerging Market Countries: Lessons from Latin America.’’ Comparative Economic Studies 44, no. 2: 45–83. Mishkin, Frederic S., and Klaus Schmidt-Hebbel. 2007. ‘‘Does Inflation Targeting Make a Difference?’’ In Monetary Policy Under Inflation Targeting, eds. Frederic S. Mishkin and Klaus Schmidt-Hebbel, 291– 372. Santiago: Central Bank of Chile. Morgan, J. P. 2003. Emerging Europe, Middle East & Africa Weekly, Newsletter, January 31. Muinhos, Marcelo Kfoury. 2001. ‘‘Inflation Targeting in an Open Financially Integrated Emerging Economy: The Case of Brazil.’’ Working Paper 26, Banco Central do Brazil. Mussa, Michael. 1986. ‘‘Nominal Exchange Rate Regimes and the Behavior of Real Exchange Rates: Evidence and Implications.’’ In Real Business Cycles, Real Exchange Rates, and Actual Policies, eds. Karl
Can Inflation Targeting Work in Emerging Market Countries?
93
Brunner and Allan Meltzer, Carnegie-Rochester Conference Series on Public Policy 25, 117–213. Amsterdam: North-Holland. Mussa, Michael, and Miguel Savastano. 1999. ‘‘The IMF Approach to Economic Stabilization.’’ In NBER Macroeconomics Annual 1999, eds. Ben S. Bernanke and Julio Rotemberg, 79–121. Cambridge, MA: MIT Press. Perry, Guillermo, John Whalley, and Gary McMahon. 2000. Fiscal Reform and Structural Change in Developing Countries, Vol. 2. New York: Palgrave Macmillan. Sargent, Thomas, and Neil Wallace. 1981. ‘‘Some Unpleasant Monetarist Arithmetic.’’ Federal Reserve Bank of Minneapolis Quarterly Review 5, no. 3: 1–17. Schmidt-Hebbel, K., and A. M. Werner. 2002. ‘‘Inflation Targeting in Brazil, Chile, and Mexico: Performance, Credibility, and the Exchange Rate.’’ Economia 2, no. 2: 30–89. Svensson, Lars O. 1997. ‘‘Inflation Forecast Targeting: Implementing and Monitoring Inflation Targets.’’ European Economic Review 41: 1111–1146. Tanzi, Vito. 2000. ‘‘Rationalizing the Government Budget or Why Fiscal Policy Is so Difficult.’’ In Economic Policy Reform: The Second Stage, ed. Anne Krueger, 435–452. Chicago: University of Chicago Press. Woodford, Michael. 1994. ‘‘Monetary Policy and Price Level Determinacy in a Cash-in-Advance Economy.’’ Economic Theory 4: 345–380. Woodford, Michael. 1995. ‘‘Price Level Determinacy without Control of a Monetary Aggregate.’’ Carnegie-Rochester Conference Series on Public Policy 43: 1–46.
5
Why Should Emerging Economies Give up National Currencies? A Case for ‘‘Institutions Substitution’’ Enrique G. Mendoza
5.1
Introduction
The severity and contagion of emerging markets crises during the last twelve years are an unprecedented phenomenon particular to the era of fast-moving, globalized financial markets. Why have emerging economies—many of which embarked in far-reaching programs of stabilization and market-oriented reform during the 1990s—fared so poorly? A simplistic explanation is ‘‘bad government,’’ which results in corrupt, incompetent, or dysfunctional political and legal institutions. Indeed, many emerging economies are plagued by problems such as rampant tax evasion, crony capitalism, non-functioning or nonexistent bankruptcy procedures, widespread financial fraud, corrupt judiciary systems, acute political strife, and so on. Yet the findings of the growing body of research on emergingmarkets crises show that even ‘‘good’’ governments may have a high degree of vulnerability to large, sudden reversals of international capital flows. At the core of this vulnerability lie two fundamental elements: the lack of credibility of domestic policy-making institutions in emerging economies and the imperfections of the globalized capital markets. The lack of credibility of economic policies in emerging countries can be attributed again to bad government, but in these countries even the policies of wellintentioned and benevolent governments would be questioned. The reason is that, for all the emphasis on new features of capital markets crises in the era of financial globalization, one key aspect of these crises is old news for citizens of emerging countries: government policy displays a high degree of time inconsistency. The currency pegs and managed exchange rates announced and implemented in the late 1980s and early 1990s were offered as commitments to stabilization policies that would result in sustainable low inflation. As time passed, however, the policy adjustments needed to stick to these commitments became too costly to bear, so adjustments were put off, leading to unsustainable
96
Enrique G. Mendoza
levels of exchange rates and financial crises. These crises were then resolved by government policies that engineered large redistributions of wealth within the private sector, and from the private sector to the public sector and/or the foreign sector. Governments have done this several times before the recent emerging markets crises (witness, for a short list of examples, Chile 1982, Mexico 1976 and 1982, or Argentina 1990). At the end of each of these debacles, the repenting government promised not to do it again, and to show its commitment it started over with a new set of monetary arrangements it swore to comply with—until the next crisis. The adverse macroeconomic effects caused by lack of policy credibility and by imperfect world capital markets are the subject of two well-established but largely disconnected strands of the international macroeconomics literature. Recently, however, some of the literature on emerging markets crises began to focus on the interaction between these two elements, showing how they combine to create a transmission mechanism that can greatly amplify the effects of adverse exogenous shocks. In an environment in which economic policy is not credible, policy reforms like privatization of public enterprises and financial institutions, or stabilization policies like the widespread use of exchange rate management to reduce inflation, may have fueled the period of bonanza that in most cases preceded balance-of-payments crises (see Calvo and Mendoza 1996 and Mendoza and Uribe 2000). Financial globalization may have created the possibility of large swings in capital flows driven by self-fulfilling expectations, imperfect and/or costly information, and other similar contagion-prone frictions that can cause large international movements of financial capital despite a country’s ‘‘strong fundamentals’’ (see Cole and Kehoe 1996, Calvo and Mendoza 2000a and 2000b, Mendoza 2002 and 2005, and Chang and Velasco 2000). The interaction between lack of credibility and capital market imperfections is reflected in the fact that a noncredible government is typically an implicit feature of the mechanisms that drive emerging-markets crises caused by financial frictions. The mechanisms used in many theoretical models to date assume that emerging economies feature short-term dollar-denominated debt, unhedged currency risks, collateral or liquidity requirements limiting the ability to contract foreign debt, implicit government guarantees offered to domestic banks or borrowers, and macroeconomic policy environments that are highly unstable and costly to evaluate. These are all features that can be attributed, at least in part, to a government’s lack of credibility at home and in world capital markets. Examples of how the interaction between financial frictions and lack of credibility can create a business cycle transmission mechanism capable of reproducing the sudden stop phenomenon are provided in Mendoza (2001, 2002, and 2006). Mendoza (2002) examines an economy with incomplete insurance markets in
Why Should Emerging Economies Give up National Currencies?
97
which uncertain duration of economic policy has real effects via a mechanism akin to a stochastic tax distortion, and foreign creditors impose liquidity requirements on domestic borrowers. If the economy has a sufficiently large external debt, a sudden policy reversal (even if only a low-probability event) triggers a large reversal in capital inflows, a suddenly binding foreign borrowing constraint, and sharp downward adjustments in economic activity and domestic relative prices. Mendoza (2001) develops a monetary variant of this model and shows that adopting a hard currency can yield large welfare gains by eliminating the risk of devaluation (in other words, the risk of a reversal of exchange rate policy) and by relaxing liquidity requirements on foreign borrowing. The main point of this article is to argue that abandoning national currencies to adopt a hard currency can be an effective policy for emerging economies to deal simultaneously with the lack of credibility of domestic financial policies and the imperfections of globalized capital markets. This two-prong approach to the problem contrasts with several of the policy recommendations that have been put forward to deal with emerging-markets crises, which typically aim to tackle either the weaknesses of domestic policy-making institutions or the imperfections of world capital markets. Those that view weak domestic institutions as the culprit tend to favor policies that can end pervasive moral hazard problems, including policies favoring floating exchange rate regimes, committing international financial institutions to refrain from providing large bail-outs, and allowing for orderly default procedures by sovereign lenders (see Lerrick and Meltzer 2001). Those that emphasize the imperfections of global capital markets tend to support policies that provide coordinated financial assistance and minimize the risk of financial contagion by aiming to keep countries at their sustainable levels of debt, or by supporting emerging-market debt prices at or above nonfundamental crash levels (see Calvo 2002 and Calvo, Izquierdo, and Talvi 2002). This article builds its case for the adoption of hard currencies in emerging economies by putting together, in an intuitive manner, arguments developed elsewhere in the literature, particularly by Calvo (1998, 2000, and 2002), Calvo and Mendoza (2000a and 2000b), Mendoza (2001, 2002, 2005, and 2006), and Mendoza and Smith (2006). The article is intentionally short on the technical details contained in these papers. The central idea is that abandoning national currencies and adopting a hard currency, when seen from the perspective of emerging economies facing noncredible policy-making institutions and imperfect globalized capital markets, can be very beneficial because it does away with the lack of credibility of monetary and exchange rate policies, and it reduces the informational frictions behind several mechanisms of financial contagion and sudden reversals of capital flows (by rendering unnecessary the costly investment of knowledge and resources that goes into assessing domestic monetary policies).
98
Enrique G. Mendoza
These arguments deviate from conventional arguments in the debate on the adoption of hard currencies or the formation of currency areas, which tend to focus on whether it is wise to give up the use of independent monetary policy as a tool for stabilization policy. (These often boil down to arguments similar to those in the protracted debate on the significance of nominal rigidities and the effectiveness of monetary policy tools.) The initial premise here is precisely that independent monetary policy in the class of countries that concern this article is a cause of concern, either because policy-making institutions in these countries are not efficient vehicles for stabilization policy, or because even if they were, the global capital market may fail to assess domestic policies with precision and differentiate them from those of other countries in which policies are mismanaged. The aim is thus to think of abandoning national currencies as a vehicle for ‘‘institutions substitution,’’ a paraphrase of the ‘‘imports substitution’’ approach to economic development that was dominant in developing countries during the 1960s. The rest of the paper is organized as follows. Section 5.2 documents key features of emerging-markets crises that point to the central role of noncredible policy and imperfect capital markets in causing the crises. Section 5.3 reviews some analytical results that suggest that the globalization of (imperfect) financial markets had endogenous mechanisms that increased the vulnerability of emerging economies to sudden stops. Section 5.4 summarizes an analytical framework for studying economic fluctuations in a small open economy that features a transmission mechanism linking imperfections in world capital markets with lack of credibility, or uncertain duration, of domestic economic policy. Section 5.5 draws policy conclusions. 5.2 Sudden Stops and Contagion: Facts and Lessons from Emerging-Markets Crises Despite heated debate in the aftermath of the 1994 Mexican crash, there is wide agreement now that the emerging-markets crises of the 1990s signaled the dawn of a new era in capital-markets crises (see Calvo and Mendoza 1996 and 2000a). This understanding was reached after observing two phenomena common to these crises: the sudden stop phenomenon and the phenomenon of financial contagion. A sudden stop consists of a sharp and sudden reversal in capital inflows, a corresponding abrupt adjustment in the current account, and sharp declines in production, absorption, and the relative prices of goods and financial assets. Sudden stops occurred in all of the emerging-markets crises, perhaps with the exception of the 1999 Brazilian crisis. Financial contagion took place when financial markets that were seemingly unrelated to developments in a struggling emerging
Why Should Emerging Economies Give up National Currencies?
99
economy were nevertheless affected and suffered severe effects in terms of price corrections and liquidity. One example was the ‘‘Tequila Effect,’’ by which the Mexican 1994 crises triggered a sudden stop in Argentina. The most prominent cases, however, are the Asian crisis of 1997 and the Russian crisis of 1998. Contagion in the Russian case nearly crashed financial markets in industrial countries and triggered a liquidity crisis that forced the Federal Reserve to cut interest rates and coordinate the collapse of hedge fund long-term capital management (LTCM). Studies like those of Calvo, Izquierdo, and Talvi (2002), Calvo and Reinhart (1999), and Milesi-Ferretti and Razin (2000) document in detail the features of the sudden stops observed in the emerging countries that suffered financial crises in the 1990s and the last two years. A document by the International Monetary Fund (1999) describes the collapses in equity prices and the increase in their volatility. Mendoza (2006) and Parsley (2001) show evidence of sharp changes in the relative price of nontradable goods for Hong Kong, Korea, and Mexico. Figures 5.1 to 5.3 provide a summary view of the stylized facts of sudden stops for the cases of Argentina, Korea, Mexico, Russia, and Turkey. Figure 5.1 shows recent time series data for each country’s current account as a share of GDP. Sudden stops are displayed in these plots as sudden, large swings of the current account that in most cases exceeded five percentage points of GDP. Figure 5.2 shows data on consumption growth as an indicator of real economic activity. These plots show that sudden stops are associated with a sharp collapse in the real sector of the economy. Figure 5.3 provides information on two key financial indicators for each country, the price of domestic equity (valued in U.S. dollars) and the spread of the yield in J. P. Morgan’s Emerging Markets Bond Index Plus (EMBIþ) for each country relative to U.S. Treasury bills. Sudden stops feature large declines in equity prices and sudden, sharp increases in EMBIþ spreads, with equity prices often leading the surge of the spreads at the monthly frequency. Empirical studies of contagion disagree on how to define and measure financial contagion, and on whether there is evidence in the data that contagion was common in emerging-markets crises (see, for example, Kaminsky and Reinhart 2000 and Rigobon 2002). Nevertheless, casual observation of figure 5.3—keeping in mind the timing of the different crises that took place during the period plotted in the chart—suggests that there was indeed some degree of financial contagion. Moreover, Rigobon’s work shows that even when statistically accurate concepts and measures of contagion are adopted, the data indicate that there were components of contagion across financial markets during the Mexican and Russian crises.
100
Enrique G. Mendoza
Figure 5.1 Current Account Balances in Percent of Gross Domestic Product. Source: International Financial Statistics, IMF.
Underlying economic causes of the sudden stop and contagion phenomena cannot be extracted from descriptions of the stylized facts like the one just summarized. In the next two sections, this paper reviews analytical arguments making the case that the uncertain duration of economic policy in emerging economies and the imperfections of globalized capital markets could be the culprits. An important necessary condition for this hypothesis to hold is that global capital inflows into emerging markets show an important component driven by factors external to emerging economies. Calvo, Leiderman, and Reinhart (1996) showed that, indeed, an important statistical predictor of the surge of capital inflows into
Why Should Emerging Economies Give up National Currencies?
101
Figure 5.2 Annual Growth Rates in Real Private Consumption Expenditures. Source: World Development Indicators, World Bank.
emerging markets in the first half of the 1990s was the decline in U.S. interest rates. More recently, one can observe in figure 5.4 two important changes in the amount and composition of net private capital inflows into emerging economies (see also Razin, Sadka, and Yuen 1998). First, there was a large and persistent reduction in total net inflows after the Asian and Russian crises. Total net inflows were seven times higher at the peak of the capital inflows boom in 1996 than in 2000. Second, the composition of net inflows changed dramatically. Bank loans never surged and became increasingly negative since 1997, while foreign direct investment (FDI) flows actually increased modestly and remained fairly stable. A deeper analysis would show two additional key facts. First, that an important element of the fall in capital inflows was the protracted retrenchment of the emerging-economies bond market induced by the Russian crisis and Ecuador’s default on its Brady bonds. Second, that the change in composition of the inflows
102
Figure 5.3 Equity Prices and Country Risk. Source: JP Morgan.
Enrique G. Mendoza
Why Should Emerging Economies Give up National Currencies?
103
Figure 5.3 (continued)
reflected also changes in the country distribution of inflows, as countries like China or Mexico, recipients of large FDI inflows, fared much better than countries relying on portfolio or bank loan flows (like Argentina or Ecuador). The Asian and Russian crises naturally could account for a temporary fall in capital inflows into the countries affected by these crises, but the trends shown in figure 5.4 affected all emerging economies and seem to be persistent. From this perspective, the drying up of capital inflows after Russia 1998 would seem to a country like Argentina as exogenous as the boom that Calvo, Leiderman, and Reinhart (1996) attributed to the fall in U.S. interest rates in the early 1990s. Thus, emerging countries are clearly vulnerable to large and potentially long-lasting swings in their ability to access global capital markets that are driven to a nontrivial extent by forces beyond their direct control.
104
Enrique G. Mendoza
Figure 5.4 Net Private Capital Flows to Emerging Markets
5.3
Rational Contagion in Global Capital Markets
Making the case that contagion and sudden stops are new phenomena in the era of global capital markets requires an argument for justifying that financial globalization can indeed trigger large swings in international capital flows that may not be justified by a country’s fundamentals. Calvo and Mendoza (2000a and 2000b) presented three economic models supporting this argument. The first model is a model of rational contagion in which global investors are less willing to pay for relevant country-specific information as global capital market integration progresses, and hence are more likely to react to rumors. In this model, the global capital market consists of a large number of identical meanvariance-optimizing investors who can choose whether or not to pay a fixed cost and eliminate the idiosyncratic uncertainty of investment in a particular emerging economy. For simplicity, all investment opportunities are ex ante identical and their returns are independently and identically distributed. How does the incentive to pay the fixed information cost (meaning the gain of paying and designing an optimal portfolio using the updated information versus investing blindly on the basis of the ex ante distribution of returns) vary as the
Why Should Emerging Economies Give up National Currencies?
105
number of emerging economies in the global capital market grows? We showed that as long as the size of short positions that can be taken on specific countries is limited, the incentive to pay the fixed information cost always reaches a point at which it becomes a decreasing function of the number of countries in the market (up to a lower bound at which it becomes independent of market size). The intuition is that the return to blind investment is becoming risk-free as the number of countries increases, while taking advantage of learning for sure about a low return on a particular country requires taking an increasingly large short position on that country. The second model that Calvo and Mendoza proposed makes an argument for rational herding among mutual fund managers or investment bankers. Assume that these managers and bankers are again identical mean-variance optimizers, but now consider that they face an incentive structure that rewards them when they produce above-market returns and punishes them when the opposite occurs. We showed that if the marginal reward is larger than the marginal punishment, there is a range of multiple optimal portfolios. Within this range, all investors move in a herd to mimic whichever portfolio becomes the day’s preferred portfolio. The reason is that holding this portfolio is optimal because investing more (less) would yield a marginal cost higher (lower) than the marginal gain. Moreover, increasing the number of countries in which to invest (in the same fashion as in the first example) widens the multiplicity range up to a point at which it becomes independent of the size of the global capital market. The third model starts where the fixed-information-cost analysis ended. That is, given that information is costly and an individual investor’s incentives to pay for it fall as the market grows, it makes sense to expect that the market will organize itself into a group of ‘‘sophisticated’’ informed traders and a group of uninformed agents that invest their funds with these sophisticated traders. Consider then a set-up in which the uninformed try to extract information about a country’s fundamentals from the noisy information they get about the specialists’ trades. Informed investors take an action, say buying Argentine bonds, that is observable to uninformed investors but that represents a combination of two variables, one that is an accurate signal of Argentine fundamentals (so that the choice to buy Argentine bonds is increasing on this signal) and another variable that reflects factors particular to the informed traders and has a negative effect on the choice to buy Argentine bonds (like margin calls informed traders may get from their own creditors). For simplicity, the observed variable is modeled as the difference of the fundamentals signal minus the investor-specific shock. Uninformed individuals know the unconditional distributions of both (which are standard normal distributions) while specialists know their exact values.
106
Enrique G. Mendoza
A standard signal extraction problem emerges in this setting: upon observing the specialists’ trade, the uninformed compute the distribution of the fundamentals signal conditional on the observed trade, and this conditional distribution has a mean that is decreasing in the noise-to-signal ratio (meaning the ratio of the variances of the noise variable relative to the fundamentals signal). Hence, if this ratio is low—that is, if the variance of fundamentals is high relative to the variance of the noise variable—uninformed investors are more likely to act on the observed trade because their conditional expectation that the trade reflects strong fundamentals is higher. In a limiting case in which the noise-to-signal ratio approaches zero even though the variance of the noise per se is large, uninformed investors bet that observed trades always signal fundamentals even though they in fact have an important ingredient of noise. Unfortunately for emerging markets, a low noise-to-signal ratio is very likely because of the relatively high cyclical variability of their fundamentals indicators, such as the terms of trade or gross domestic product (see Mendoza 1995), compared with the rare occurrence of widespread margin calls and liquidity crunches a` la the Russian crisis. The contagion of the Russian default that drove investors into liquidating positions across emerging markets worldwide can be interpreted then as the outcome of a low-probability, nonfundamentals shock—in other words, large margin calls, which induced a sizable cut in the observed positions of specialized investors in emerging markets that was then largely misinterpreted by uninformed investors as a shock to the fundamentals of emerging economies. Informational frictions are crucial for the three models reviewed. For instance, if all investors are equally and perfectly informed, there would always be investors willing to purchase the securities that investors who are facing margin calls are liquidating at prices that cannot deviate much from the fundamentals prices. Asset prices would then collapse only in the case of a true negative shock to fundamentals. Short-selling constraints and margin calls (which themselves can be interpreted as resulting from informational frictions) are just as crucial. If unlimited short selling were possible, for example, investors would never have reduced incentives to acquire country-specific information at a fixed cost. Moreover, there are reasons to believe that informational frictions are more pervasive in the context of global capital flows to emerging markets than in other asset markets. This is because of the amount and diversity of the information that needs to be gathered on a large set of economic, social, and political factors, and the expertise required to process and interpret it. There is also little room for economies of scale in this information-gathering process: collecting information on determinants of asset returns in Korea may be of some help in predicting returns in southeast Asia but is of little help in fully understanding returns prospects in Indonesia and is of no use in assessing returns for Latin America.
Why Should Emerging Economies Give up National Currencies?
5.4
107
Noncredible Policy, Financial Frictions, and Sudden Stops
The macroeconomic implications of the temporariness of noncredible government policies is the subject of a large literature in international macroeconomics (Rebelo and Vegh 1996 and Calvo and Vegh 1999 provide detailed surveys). The starting point of this literature was the observation that in countries in which noncredible stabilization programs were perceived as temporary in the 1970s and 1980s, consumption boomed in the early stages of the programs and then collapsed at or around the time at which the programs were abandoned. Calvo (1986) argued that this could be explained via intertemporal substitution: if the public knows with full certainty that inflation stabilization will be temporary, it is optimal to substitute consumption away from future high-inflation periods into present lowinflation periods. Helpman and Razin (1987) approached the issue focusing on wealth effects resulting from the fiscal implications of temporary inflation stabilizations, modeling these wealth effects as intergenerational wealth, redistributions that break the Ricardian equivalence principle. A temporary currency peg that lowers current inflation redistributes wealth, favoring the generations that are alive while the peg is in place, and working against those that arrive at the time the peg collapses and in the periods that follow. Mendoza and Uribe (2000) developed a stochastic model in which a noncredible currency peg is the source of uninsurable devaluation risk, which in turn is the driving force of business cycles in a small open economy perfectly integrated into international capital markets. Their analysis incorporates mechanisms similar to Calvo’s intertemporal substitution effect and Helpman and Razin’s fiscalinduced wealth effect, but in the context of an equilibrium business cycle model in which the private sector formulates optimal plans with regard to consumption of traded and nontraded goods, investment, labor, and money holdings. The optimal decisions of the private sector in the Mendoza-Uribe model are distorted by uninsurable devaluation risk in the following manner. First, uncovered, risk-adjusted nominal interest parity implies that the domestic nominal interest rate features an endogenous time- and state-dependent risk premium relative to the world nominal interest rate (determined as a multiple of the rate at which the public expects the currency to be devalued). The size of this premium depends on the expected decline in wealth caused by the increase in seigniorage associated with the collapse of the currency and the surge in inflation that follows. Second, fluctuations in the nominal interest rate act as a random tax on current labor income, future capital income, and the return on saving, because domestic money is a risky asset that helps agents economize transactions costs in purchasing consumption and investment goods (a similar outcome obtains in Mendoza 2001, assuming instead that money enters directly in the utility function).
108
Enrique G. Mendoza
In the Mendoza-Uribe setup, a time-varying J-shaped probability of the abandonment of a currency peg yields macroeconomic dynamics consistent with several stylized facts of exchange-rate-based stabilizations. A J-shaped pattern of devaluation probabilities captures the observation that these stabilization plans start with a very uncertain future (so devaluation probabilities are high when they are introduced). Credibility then improves gradually (so devaluation probabilities decline) until a certain point in time during which the survival of the plan is increasingly questioned (so devaluation probabilities begin to rise). Blanco and Garber (1986) and Klein and Marion (1997) provide econometric evidence in favor of this argument. The currency risk premium implies that a J-shaped pattern of devaluation probabilities yields a similarly shaped time path of domestic nominal interest rates for the economy attempting to stabilize with imperfect credibility. In turn, the declining interest rate in the early stages of the program favors booms in consumption, investment, and employment; a widening current account deficit; a reduction in the expenditure velocity of circulation of money; and a real appreciation of the currency (meaning an increase in the relative price of nontradable goods in terms of tradable goods). These are low-tax periods in terms of the random taxes mentioned earlier, and they are also periods in which, each date that the peg survives, the public realizes that actual wealth is higher than expected the date before (since the surge in seigniorage did not materialize). Hence, the model’s intertemporal tax distortions and fiscal-induced wealth effect both work to favor the economic boom and real appreciation. As the nominal interest rate touches bottom and starts to increase, the tax-like distortions work in the opposite direction and they can outweigh the fiscal-induced wealth effect (in which case there is a recession that predates the collapse of the peg, as observed, for example, before the devaluations in Mexico in 1994 and in Argentina in 2002). It is crucial to note that, in contrast with the perfect-foresight studies common in the imperfect-credibility or policy-temporariness literature, the Mendoza-Uribe model features the above cyclical dynamics regardless of whether the currency is devalued ex post or not. Thus, the essential element of the analysis is the fact that the stabilization policy anchored on the currency peg is not fully credible, and hence the public expects that a devaluation may happen with some probability, in an environment in which currency risk goes unhedged. The findings of the Mendoza-Uribe model suggest that the analysis of macroeconomic policies of uncertain duration has gone some way in accounting for the stylized facts of temporary or noncredible exchange-rate-based stabilizations. As a framework that can account for sudden stops, however, it has an important shortcoming: stabilizing emerging economies are assumed to have access to a perfect, frictionless international capital market in which they can borrow as much as
Why Should Emerging Economies Give up National Currencies?
109
they want as long as they refrain from running Ponzi games. As will be illustrated later, it is not possible in this environment to obtain sudden reversals of the current account in response to either a policy reversal or other exogenous shocks of foreign or domestic origin. Consider now the possibility that as a result of the informational frictions discussed in section 5.2, international capital markets are highly imperfect. The recent literature on the sudden stops phenomenon has explored the potential of a large variety of capital-market imperfections to act as triggers of sudden stops (see the survey by Arellano and Mendoza 2003). For simplicity, follow Mendoza (2001 and 2002) and Mendoza and Smith (2006) in assuming that these imperfections take the form of foreign borrowing constraints that impose liquidity requirements or collateral constraints on borrowers, and debt contracts that can only be denominated in units of tradable goods for economies with a large nontradablegoods industry (a phenomenon referred to as ‘‘liability dollarization’’). A liquidity requirement is used by lenders as a criterion that helps them manage default risk by requiring borrowers to pay a fraction of their current obligations out of current income, or equivalently by requiring them not to let their total debt-to-income ratio exceed a certain level. A classic case is the scoring criteria used in mortgage lending, by which qualified borrowers are required to satisfy specific ratios of nonmortgage debt payments and total debt payments as a fraction of gross income. What happens with the predictions of the devaluation risk framework of Mendoza and Uribe (2000) when the structure of international capital markets is modified to introduce liquidity requirements? There are two essential changes. First, the liquidity requirement introduces an occasionally binding constraint on foreign borrowing, by which debt must be less than or equal to a certain fraction of the value of domestic income in units of tradable goods. Second, whether this constraint binds or not at a particular date and state of nature is a combined result of the underlying exogenous shocks driving the model, the endogenous dynamics of foreign debt, income generated in the tradables and nontradables sectors of the economy, and the domestic relative price of nontradables. In this economy with borrowing constraints, sudden stops occur when the country’s external debt is sufficiently large and a sufficiently adverse combination of shocks shifts the economy from a situation in which the liquidity requirement was not binding to a situation in which it binds. The sufficiently adverse stock of debt and sufficiently adverse shocks are those such that the amount by which domestic agents would like to enlarge their foreign debt, in the absence of a liquidity constraint by foreign lenders, exceeds the level that this constraint allows. The constraint will then be met by a sudden current account reversal and a collapse in private absorption. Moreover, the adjustments in consumption and the current
110
Enrique G. Mendoza
account are magnified if they trigger a collapse in domestic income and/or in the relative price of nontradable goods (that is, the real exchange rate) because this implies that in addition to the liquidity constraint becoming suddenly binding, the constraint is tightened further by the endogenous collapses of income and prices (since debt is denominated in units of tradables but partially leveraged on the sizable income generated by the nontradables sector). How strong can these effects be? To answer this question, consider some of the quantitative simulation results reported in Mendoza (2002). These simulations apply to Mexico as an example of a representative small, open, emerging economy. Mexico has large industrial sectors producing and generating income in tradable and nontradable goods, and both goods are also consumed by the domestic private sector. Three exogenous random shocks are introduced to drive Mexico’s business cycles: domestic productivity shocks, shocks to the world real interest rate, and shocks that reflect the uncertain duration of economic policy (modeled as regime-switching shocks affecting the rate of depreciation of the currency or a set of direct and indirect tax rates). Mexican national accounts data are used to calibrate the model. The calibration also examines historical data on Mexico’s foreign asset position to pin down parameter values that allow the model to mimic Mexico’s average foreign debt-to-output ratio and a reasonable value for the liquidity requirement coefficient (in other words, the maximum debt-to-GDP ratio in units of tradables). Figure 5.5 shows the effects of a shift from a state with high productivity, low world interest rate, and low policy distortions to a state with the opposite features for different levels of initial net foreign assets, or the negative of foreign debt. The plots show the impact effects for an economy assumed to have access to a perfect credit market and for the economy that faces the liquidity requirement. The results shown in figure 5.5 have three key implications. First, if there are no imperfections in international credit markets, there are no sudden stops. Even at very high debt levels, the adverse shocks trigger relatively smooth adjustments in the model’s endogenous macroeconomic aggregates. Second, for very low or very high debt levels there is also no room for sudden stops. With very high debt the borrowing constraint binds regardless of whether the shocks are favorable or unfavorable. With very low debt the constraint never binds and the economy’s response to the shocks is nearly identical to that observed in the case in which no credit-market imperfection is present. These are tranquil times in which shifts across good and bad shocks produce relatively smooth business cycles and the economy’s access to world capital markets is not compromised. Third, in the range of debt positions just above the level at which the range of debt positions with nonbinding liquidity constraints ends, adverse shocks trigger a suddenly binding borrowing constraint. This is the sudden stop range. Here, the equilibrium
Why Should Emerging Economies Give up National Currencies?
Figure 5.5 Impact Effects of a Shift from ‘‘Best’’ to ‘‘Worst’’ State (as a function of the foreign asset grid)
111
112
Enrique G. Mendoza
response of the economy to the same size of adverse shocks to productivity, world interest rate, and policy distortions as in the other ranges features a sharp reversal of the current account and a major collapse in domestic consumption and production. For part of this range there is a large drop in the relative price of nontradables. (The price can also rise because the liquidity constraint induces supplyand-demand effects in the market of nontradables, and depending on which effect is stronger, the price can rise or fall.) The simulations shown in figure 5.5 are an imperfect approximation to reality and they follow from a stylized model with many caveats and unrealistic assumptions. Yet the magnitude of the sudden stop adjustments that it produces is striking. Moreover, some of the caveats and unrealistic assumptions of this setting do not apply to other recent quantitative studies that yield the same main prediction: capital market imperfections are a powerful mechanism for producing sudden stops. The literature on quantitative applications of models of emerging-markets crises has examined alternatives that consider the margin constraints and trading costs (Mendoza and Smith 2006 and Cavallo, Kisselev, Perri, and Roubini 2002), costly monitoring of borrowers (Cespedes, Chang, and Velasco 2000), collateral constraints (Paasche 2001), working capital (Oviedo 2002, Uribe and Yue 2006, and Neumeyer and Perri 2006), default risk (Arellano 2005), and current account targets (Valderrama 2002). A closer look at the analysis of Mendoza and Smith (2006) can help highlight the interaction between the information costs that distort international capital markets (as argued in section 5.2), the determination of emerging markets asset prices, and the business cycle transmission mechanism that drives sudden stops. This analysis also brings into play a key feature of credit contracts among players in international capital markets: collateral constraints in the form of margin requirements. Factual descriptions of the contagion of the Russian crisis and its connection with the collapse of LTCM describe how widespread margin calls played a central role in precipitating the systemic sale of emerging countries’ assets in global capital markets (see Dunbar 2000 and the Wall Street Journal article series published September 22–24, 1998). Consider a representative foreign securities firm that specializes in trading a small open economy’s equity. This firm incurs recurrent transactions costs as well as transactions costs that are a quadratic function of the size of the trades it undertakes. These two transactions costs represent the informational costs alluded to in the second section. The per-trade costs are important because they imply an elasticity in the foreign traders’ demand for emerging-markets equity that is inversely related to the per-trade transactions costs. At the limits, if the per-trade cost is zero the trader’s demand is infinitely elastic; if the per-trade cost is infinite the demand is perfectly inelastic. The recurrent costs are important because they capture
Why Should Emerging Economies Give up National Currencies?
113
the idea that trading in emerging-market assets requires continuous investment in gathering and processing country-specific information. These costs need to be incurred just to be prepared to trade on a particular emerging country’s assets, regardless of whether there is actually any trade, and the costs themselves are unrelated to the size of the trades. For example, understanding Chile’s inflationtargeting system under its framework of widespread indexation, or Mexico’s unique setup for conducting monetary policy by adjusting cortos (or ‘‘shorts’’) among banking institutions, is a complex, information-intensive task and its cost does not depend on how much a proposed trade in Chilean or Mexican bonds is worth. In the absence of recurrent trading costs, foreign traders would be willing to hold on to a given position on emerging-markets assets if and only if the price equals the fundamentals price (defined as the expected present value of dividends discounted at the world real interest rate). They would be willing to buy more if the price is lower than the fundamentals level, and less if the price is above the fundamentals level. In the presence of recurrent costs, however, traders hold on to a given emerging-markets asset position only if the price is below fundamentals. Thus, recurrent costs are a necessary condition for the price of emergingmarkets assets to remain lower than justified by fundamentals even in the long run (when asset positions reach a stationary equilibrium). The foreign traders trade equity with the corresponding emerging country. Lenders in the global credit market that agents in this country has access to demand to hold, as collateral, a certain fraction of the market value of the domestic agents’ equity holdings. This margin constraint is a collateral constraint with two unique features. First, custody of the collateral is surrendered to lenders when the debt is contracted, so the typical issues that arise with other collateral constraints, related to whether lenders can or want to take over the collateral after a defaulting borrower declares default, are irrelevant. Second, a fall in asset prices tightens the constraint because it reduces the market value of the assets offered as collateral, thereby allowing the lender to make a margin call on the borrower to make up the difference. The margin call is automatic and does not require legal proceedings beyond the margin clause already agreed to in the debt contract. Some margin requirements exist as a result of government regulation of the financial industry but they are also widely used as explicit clauses in credit contracts and can take different shapes (value-at-risk collateralization, for example, is a form of margin requirement in which the lender makes margin calls based on its estimate of the potential losses that it could face if a worst-case scenario for the particular asset exposure in question materializes). In this setting, again, a sufficiently large initial debt and a mix of sufficiently adverse shocks to productivity, world interest rates, and domestic policies shifts the
114
Enrique G. Mendoza
economy from a state in which the margin constraint did not bind to one in which it does. But now the process that follows has an extra ingredient: a modernized version of the debt-deflation mechanism pioneered by Irving Fisher (1933). A debt-deflation process starts when the sufficiently adverse shocks hit and agents in the emerging economy get an initial margin call. They then rush to ‘‘fire-sale’’ their equity in the global capital market. However, they meet there with foreign traders that have a less-than-infinitely-elastic demand for equity because of the informational costs they incur, and so are only willing to buy the extra equity at a reduced price. If there were no informational costs, the margin call would simply result in an equity reallocation from home agents to foreign traders with no change in emerging-markets equity prices. But with trading costs the asset price falls, and the lenders see the value of the assets they hold as collateral shrink. This triggers the margin clauses of debt contracts again, and a new round of margin calls takes place. The process is repeated until the equilibrium price settles at a level that satisfies the margin constraint (Mendoza and Smith 2006 ensure that it does by assuming, as was done in the second section, that agents cannot take unlimited short positions). A mechanism for financial contagion also emerges from this model. Consider that a Russian-style default triggers a ‘‘world liquidity shock,’’ which takes the form of a large increase in the world interest rate. An emerging economy sitting in South America and totally unrelated to Russia may experience a sudden stop and a collapse in the prices of its internationally traded assets as a result of the combination of margin constraints and trading costs. An interest-rate shock is a change in fundamentals and as such it should be reflected in what happens to an emerging country’s asset prices and economic conditions, but the point is that there can be substantial overreaction to this change in fundamentals because of the imperfections of the world’s capital market. 5.6
Policy Implications
It may seem odd that a paper on abandoning national currencies reaches this point without having said much about money and exchange rates (except for the discussion of devaluation risk as a driving force of business cycles). This is because the problems that were described with lack of credibility of government policy, and imperfect international capital markets can exist regardless of the monetary institutions and currency arrangements. For example, Kumhof, Li, and Yan (2006) showed that, if fiscal fundamentals are off track, balance-of-payments crises occur even if a managed exchange rate is replaced with a floating rate under inflation targeting. It also follows from this reasoning, however, that giving
Why Should Emerging Economies Give up National Currencies?
115
up the national currency to adopt a hard currency does not rule out financial crises that can look like a sudden stop (consider, for example, the Great Depression). When it comes to recent sudden stops in emerging economies, however, monetary and exchange rate policies are at the core of government policy’s lack of credibility and the informational frictions behind global capital market imperfections. An emerging country that completely gives up its national currency for the dollar or the euro, and so reduces its central bank to a bank-supervision agency, stands to make several gains. 1. The devaluation risk that has played a central role in the chronic boom-bust cycle associated with the introduction and collapse of managed exchange rates in this class of countries would be greatly reduced. It can never be fully eliminated, inasmuch as a sovereign nation can never credibly commit to not try to reverse course and reintroduce a national currency in the future. 2. Foreign investors that have a long history in investing resources and building expertise to track monetary policy in Europe and the United States would no longer need to ponder whether or not to pay for gathering and processing costly information on the monetary policy of each emerging economy they were interested in. This can be interpreted as a decline in information costs, which would translate into better-informed investors and reduced vulnerability to herd behavior. 3. Reduced information costs also increase the demand elasticity for emergingmarkets equity of foreign traders, which limits the size of asset price declines and Fisherian debt-deflations that could occur because of frictions like collateral constraints and margin calls. 4. Financial assets and liabilities would be matched in terms of currency denomination, ending the liability dollarization problem. Again, a problem with a similar flavor will emerge whenever a sudden, large relative price collapse takes place, but it would no longer be possible for the lack of credibility of domestic monetary or exchange rate policy to generate it. 5. Considering the contracting environment from which liquidity requirements and collateral constraints emerge, enhanced credibility and reduced informational frictions could result in better access to international capital markets in terms of reduced liquidity coefficients and margin requirements. We know little about how optimal credit contracts would or should react to the adoption of a hard currency in replacement of a national currency, but what was argued here on the basis of findings from the literature is that if liquidity requirements and collateral constraints did fall, vulnerability to sudden stops and contagion would be greatly reduced.
116
Enrique G. Mendoza
Regardless of these advantages, emerging countries will not rush to abandon their currencies. A national currency is a valued symbol of national identity. It is also a valued and powerful tool for governments to have in difficult times in which it is a lot easier, and more discreet, to transfer wealth from one set of economic agents to another by simply printing money rather than by proposing an explicit policy to do it. There is also the loss of seigniorage and sovereignty involved in giving up monetary policy-making powers to a foreign nation, and the potential loss of not being able to conduct independent monetary policy to smooth the pains of domestic business cycles. The latter is the less-relevant argument, even if perhaps it is the one that often occupies academic and policy debates on exchange rate regimes. The reason is put bluntly in Robert Lucas Jr.’s 1996 Nobel lecture: ‘‘Central bankers and even some monetary economists talk knowledgably of using interest rates to control inflation but I know of no evidence from even one economy linking these variables in a useful way . . .’’ (666). Even if we had such scientific knowledge, the recent historical record of many emerging economies shows that their policy-making institutions have performed poorly at using independent monetary and exchange-rate policy to preserve price stability and avoid large business cycles. Moreover, even if we had both knowledge and institutions for monetary policy to be useful, the small potential benefit of finetuning business cycles would have to be matched up against the five listed gains of giving up national currencies. If abandoning national currencies seems a good but unrealistic idea, what else can be done to address the underlying causes of emerging-markets crises? One set of alternatives considers the possibility of establishing support schemes for the prices of emerging-markets assets (see Lerrick and Meltzer 2001, Calvo 2002, and Durdu and Mendoza 2006). The premise of these proposals is that either because of moral hazard problems (Lerrick and Meltzer) or because of capital market imperfections (Calvo), asset prices of emerging countries can fall sharply below levels warranted by fundamentals, and this is a central ingredient of the process driving sudden stops. Hence the international financial organizations could be redesigned to set up facilities that prevent prices from reaching those crash levels. Another alternative is to deepen the internationalization of the financial system, and either secure support for local bank subsidiaries from their foreign owner banks, or limit the extent to which those local banks can provide fuel for the liability dollarization process by imposing on them high reserve requirements (i.e., by moving to narrow banking). These indirect means of achieving the same goals that adopting a hard currency can achieve much more easily seem unduly costly to impose on the citizens of emerging countries, many of whom suffer grave economic consequences as a result of recurrent episodes of sudden stops and financial contagion. Abandoning
Why Should Emerging Economies Give up National Currencies?
117
national currencies does seem a radical idea, but so did the euro, the European Union, and the North American Free Trade Agreement not so long ago. Of course giving up national currencies is not a panacea. It cannot fix many fundamental economic and institutional problems plaguing emerging countries, or eliminate forever all forms of financial crises. It is, however, unlike any other currency arrangement in that it ties as tightly as possible the government’s hands so as to prevent it from exercising its confiscatory powers via monetary policy, and in that it simplifies greatly the task of assessing domestic financial policies that is so critical for driving global capital inflows into emerging economies. Acknowledgments This paper borrows heavily from joint work with Guillermo Calvo and Katherine Smith. Comments and suggestions from Cristina Arellano and Alejandro Izquierdo are also gratefully acknowledged. References Arellano, Cristina. 2005. ‘‘Default Risk and Aggregate Fluctuations in Emerging Economies.’’ Mimeo., University of Minnesota. Arellano, Cristina, and Enrique G. Mendoza. 2003. ‘‘Credit Frictions and Sudden Stops in Small Open Economies: An Equilibrium Business Cycle Framework for Emerging Markets Crises.’’ In Dynamic Macroeconomic Analysis, eds. Sumru Altug, Jagjit S. Chadha, and Charles Nolan, 335–405. New York: Cambridge University Press. Blanco, Herminio, and Peter M. Garber. 1986. ‘‘Recurrent Devaluation and Speculative Attacks on the Mexican Peso.’’ Journal of Political Economy 94: 148–166. Calvo, Guillermo A. 1986. ‘‘Temporary Stabilization: Predetermined Exchange Rates.’’ Journal of Political Economy 94: 1319–1329. ———. 1998. ‘‘Capital Flows and Capital-Market Crises: The Simple Economics of Sudden Stops.’’ Journal of Applied Economics 1: 35–54. ———. 2000. ‘‘Balance-of-Payments Crises in Emerging Markets: Large Capital Inflows and Sovereign Governments.’’ In Currency Crises, ed. Paul Krugman, 71–98. Chicago: University of Chicago Press. ———. 2002. ‘‘Globalization Hazard and Delayed Reform in Emerging Markets.’’ Mimeo., Center for International Economics, University of Maryland. Calvo, Guillermo A., Alejandro Izquierdo, and Ernesto Talvi. 2002. ‘‘Sudden Stops, the Real Exchange Rate and Fiscal Sustainability: Argentina’s Lessons.’’ Mimeo., Research Department, Inter-American Development Bank, Washington, D.C. Calvo, Guillermo A., Leonardo Leiderman, and Carmen M. Reinhart. 1996. ‘‘Inflows of Capital to Developing Countries in the 1990s.’’ Journal of Economic Perspectives 10, no. 2: 123–140. Calvo, Guillermo A., and Enrique G. Mendoza. 1996. ‘‘Mexico’s Balance of Payments Crises: A Chronicle of a Death Foretold.’’ Journal of International Economics 41: 235–264.
118
Enrique G. Mendoza
———. 2000a. ‘‘Capital-Market Crises and Economics Collapse in Emerging Markets: An Informational-Frictions Approach.’’ American Economic Review 90, no. 2: 59–64. ———. 2000b. ‘‘Rational Contagion and the Globalization of Securities Markets.’’ Journal of International Economics 51, no. 1: 79–113. Calvo, Guillermo A., and Carmen M. Reinhart. 1999. ‘‘When Capital Inflows come to a Sudden Stop: Consequences and Policy Options.’’ Mimeo., Center for International Economics, University of Maryland. Calvo, Guillermo A., and Carlos A. Vegh. 1999. ‘‘Inflation Stabilization and BOP Crises in Developing Countries.’’ In Handbook of Macroeconomics, Vol 1C, eds. John B. Taylor and Michael Woodford. Amsterdam: North-Holland. Cavallo, Michelle, Kate Kisselev, Fabrizio Perri, and Nouriel Roubini. 2002. ‘‘Exchange Rate Overshooting and the Cost of Floating.’’ Mimeo., Stern School of Business, New York University. Cespedes, Luis, Roberto Chang, and Andre´s Ve´lasco. 2000. ‘‘Balance Sheets and Exchange Rate Policy.’’ Mimeo., Department of Economics, New York University. Chang, Roberto, and Andres Velasco. 2000. ‘‘Banks, Debt Maturity and Crises.’’ Journal of International Economics 51, no. 1: 169–194. Cole, Harold L., and Timothy J. Kehoe. 1996. ‘‘A Self-Fulfilling Model of Mexico’s 1994–95 Debt Crisis.’’ Journal of International Economics 41, no. 3: 309–330. Dunbar, Nicholas. 2000. Inventing Money: The Story of Long Term Capital Management and the Legends Behind It. New York: Wiley and Sons. Durdu, C. Bora, and Enrique G. Mendoza. 2006. ‘‘Are Asset Price Guarantees Useful for Preventing Sudden Stops?’’ Journal of International Economics 69, no. 1: 84–119. Fisher, Irving. 1933. ‘‘The Debt-Deflation Theory of Great Depressions.’’ Econometrica 1: 337–357. Helpman, Elhanan, and Assaf Razin. 1987. ‘‘Exchange Rate Management: Intertemporal Tradeoffs.’’ American Economic Review 77: 107–123. International Monetary Fund. 1999. International Capital Markets. Washington, D.C.: International Monetary Fund. Kaminsky, Graciela L., and Carmen M. Reinhart. 2000. ‘‘On Crises, Contagion, and Confusion.’’ Journal of International Economics 51: 145–168. Klein, Michael W., and Nancy P. Marion. 1997. ‘‘Explaining the Duration of Exchange-Rate Pegs.’’ Journal of Development Economics 54: 387–404. Kumhof, Michael, Shujing Li, and Isabel Yan. 2007. ‘‘Balance of Payments Crises Under Inflation Targeting.’’ Journal of International Economics 72, no. 1: 242–264. Lerrick, Adam, and Allan H. Meltzer. 2001. ‘‘Blueprint for and International Lender of Last Resort.’’ Mimeo., Carnegie Mellon University, Pittsburgh, PA. Lucas Jr., Robert E. 1996. ‘‘Nobel Lecture: Monetary Neutrality.’’ Journal of Political Economy 104: 661– 682. Mendoza, Enrique G. 1995. ‘‘The Terms of Trade, The Real Exchange Rate and Economic Fluctuations.’’ International Economic Review 36: 101–137. ———. 2001. ‘‘The Benefits of Dollarization when Stabilization Policy Lacks Credibility and Financial Markets are Imperfect.’’ Journal of Money, Credit and Banking 33, no. 2: 440–474.
Why Should Emerging Economies Give up National Currencies?
119
———. 2002. ‘‘Credit, Prices, and Crashes: Business Cycles with a Sudden Stop.’’ In Preventing Currency Crises in Emerging Markets, eds. Jeffrey Frankel and Sebastian Edwards, 335–390. Chicago: University of Chicago Press. ———. 2005. ‘‘Real Exchange Rate Volatility and the Price of Nontradables in Sudden-Stop-Prone Economies.’’ Economia 6, no. 1: 103–148. ———. 2006. ‘‘Lessons from the Debt-Deflation Theory of Sudden Stops.’’ American Economic Review 96, no. 2: 411–416. Mendoza, Enrique G., and Katherine A. Smith. 2006. ‘‘Quantitative Implications of a Debt-Deflation Theory of Sudden Stops and Asset Prices.’’ Journal of International Economics 70, no. 1: 82–114. Mendoza, Enrique G., and Martin Uribe. 2000. ‘‘Devaluation Risk and the Business Cycle Implications of Exchange Rate Management.’’ Carnegie-Rochester Conference Series on Public Policy 53: 239–296. Milesi-Ferreti, Gian Maria, and Assaf Razin. 2000. ‘‘Current Account Reversals and Currency Crises: Empirical Regularities.’’ In Currency Crises, ed. Paul Krugman. Chicago: University of Chicago Press. Neumeyer, Pablo, and Fabrizio Perri. 2005. ‘‘Business Cycles in Emerging Economies: The Role of Interest Rates.’’ Journal of Monetary Economics 52: 345–380. Oviedo, P. Marcelo. 2004. ‘‘Intermediation of Capital Inflows: The Macroeconomic Implications of Neoclassical Banks & Working Capital.’’ Mimeo., Iowa State University. Paasche, Bernhard. 2001. ‘‘Credit Constraints and International Financial Crises.’’ Journal of Monetary Economics 28: 623–650. Parsley, David. 2001. ‘‘Accounting for Real Exchange Rate Changes in East Asia.’’ Working Paper No. 6/2001, Hong Kong Institute of Monetary Research. Razin, Assaf, Efraim Sadka, and Chi-Wa Yuen. 1998. ‘‘A Pecking Order Theory of Capital Flows and International Tax Principles.’’ Journal of International Economics 44, no. 1: 45–68. Rebelo, Sergio, and Carlos A. Ve´gh. 1996. ‘‘Real Effects of Exchange-Rate-Based Stabilization.’’ In NBER Macroeconomics Annual 1996, Cambridge, MA: MIT Press. Rigobon, Roberto. 2002. ‘‘Contagion: How to Measure It?’’ In Preventing Currency Crises in Emerging Markets, eds. Jeffrey Frankel and Sebastian Edwards, 269–329. Chicago: University of Chicago Press. Uribe, Martin, and Zhanwei Vivian Yue. 2006. ‘‘Country Spreads and Emerging Countries: Who Drives Whom?’’ Journal of International Economics 69: 6–36. Valderrama, Diego. 2002. ‘‘The Impact of Financial Frictions on a Small Open Economy: When Current Account Borrowing Hits a Limit.’’ PhD dissertation, Department of Economics, Duke University. Wall Street Journal. 1998. ‘‘Markets Under Siege,’’ Sept. 22–24, p. A1.
6
Hard Currency Pegs and Economic Performance Sebastian Edwards and I. Igal Magendzo
‘‘. . . almost all independent countries choose to assert their nationality by having, to their own inconvenience and that of their neighbors, a peculiar currency of their own . . .’’ —John Stuart Mill, Principles of Political Economy
6.1
Introduction
The recurrence of currency crises in emerging countries has generated an intense debate on exchange rate policies. Pegged-but-adjustable exchange rate regimes have rapidly lost adepts, while hard pegs and freely floating rates have gained in popularity (See Summers 2000 and Fischer 2001). Recently, a number of economists have gone as far as arguing that (many) emerging nations should completely give up their national currencies, and join a currency union.1 Indeed, in a series of highly influential papers, Guillermo Calvo and his associates have made some of the most persuasive arguments with respect to the adoption of ‘‘common currency’’ monetary regimes (see, for example, Calvo 1999, 2000, and 2001; Calvo and Reinhart 2002; and Calvo, Izquierdo, and Talvi 2003). In principle, currency unions can take two forms. First, a country may adopt another nation’s currency as its own. When the other nation is an advanced country, this monetary arrangement has come to be known by the general name of dollarization.2 Under dollarization the country in question completely gives up monetary independence, and monetary policy is run by the advanced nation’s central bank. Countries can dollarize in a unilateral fashion—in which case they will lose the revenue from seignorage—or they can sign a monetary treaty with the advanced country and share seignorage.3 Under the second type of currency union, a group of countries creates a new currency that is common to the group.4 Under this option, monetary policy is run by a common central bank, the members of the currency union share seinorage, and the common currency’s exchange rate may float relative to other currencies. In the rest of this paper we will refer to
122
Sebastian Edwards and I. Igal Magendzo
this type of monetary regime as an independent currency union or ICU.5 Panama and Monaco are good examples of dollarized countries. The East Caribbean Currency Area (ECCA) and the Communante Financiere de l’Afrique (CFA), on the other hand, are good representatives of what we have called an independent currency union. In the last few years some emerging countries have decided to give up their currencies and officially dollarize their economies. In 2000, for example, in the midst of a major crisis, Ecuador gave up its currency, the sucre, and adopted the U.S. dollar. El Salvador adopted the dollar during 2001, and in May 2001, the dollar became legal tender in Guatemala. In other countries, however, politicians have systematically refused to consider dollarization, even in the face of major and costly financial crises. This was the case of Argentina, for instance, during late 2001 and early 2002. A number of authors have argued that countries—and in particular, emerging countries—that give up their currency will tend to outperform countries with a currency of their own. According to this view, not having a domestic currency will have two major positive effects on economic performance. First, inflation will be lower in common-currency countries than in nations with their own currency. Alesina and Barro (2001, 382), for instance, have argued that adopting an advanced nation’s currency ‘‘eliminates the inflation-bias problem of discretionary monetary policy.’’ Second, countries that give up their currency will tend to grow faster than countries with a domestic currency. This growth effect is supposed to take place through two channels: (1) A common currency will tend to result in lower interest rates, higher investment, and faster growth (Dornbusch 2001). (2), by eliminating exchange rate volatility, a common currency is supposed to encourage international trade; this, in turn, will result in faster growth. Rose (2000) and Rose and Van Wincoop (2001), among others, have emphasized this trade channel.6 Other authors, however, have voiced skepticism regarding the alleged benefits of common-currency regimes. According to an alternative view that goes back at least to Meade (1950), countries with a hard peg will have difficulty accommodating external shocks, including terms of trade and world interest rate disturbances. This, in turn, will be translated into greater instability and, under some circumstances, will lead to lower economic growth (Fischer 1976, Parrado and Velasco 2002). Frankel (1999) has taken a more nuanced view, and has argued that there is no unique recipe on exchange rate policy; while some countries will benefit from hard pegs, for other countries a floating regime will be more appropriate. And according to Eichengreen (2001), the evidence on the relationship between monetary regimes and growth is inconclusive, and does not support the claim that dollarization—or any exchange rate regime, for that matter—is an important determinant of growth.
Hard Currency Pegs and Economic Performance
123
Surprisingly, until recently there have been very few formal empirical studies on the economic consequences of common currencies. In particular, international comparative studies on alternative exchange rate and monetary regimes have traditionally ignored common-currency countries. For instance, the comprehensive study on exchange rate regimes, growth, and inflation by Ghosh, Gulde, Ostry, and Wolf (1995) does not include nations that do not have a currency of their own. Likewise, the IMF (1997) study on alternative exchange rate systems excludes common-currency countries, and the recent paper by Levy-Yeyeti and Sturzenegger (2001) on exchange rates and economic performance excludes nations that do not have a central bank of their own. This lack of empirical evidence means that countries contemplating giving up their currency have very little information on how other countries have historically performed under this monetary regime. Most existing evidence on dollarization is based on the experience of a single country: Panama, which has used the U.S. dollar as legal tender since 1904.7 Recently, Andrew Rose and a series of collaborators have analyzed in great detail the effects of common currencies on the volume of international trade (see, for instance, Rose 2000, Engel and Rose 2002, and Frankel and Rose 2002). This interesting and increasingly influential research has concluded that, with other things given, countries with a common currency tend to trade among themselves more intensively than countries that have a domestic currency. These analyses, however, do not make a distinction between the two types of common-currency regimes discussed above: strictly dollarized and independent currency unions (ICU). For instance, an inspection of the data sets used by Engel and Rose (2002) and Frankel and Rose (2002) indicates that they treat dollarized and ICU nations as a homogeneous group. Moreover, their sample is tilted toward ICU countries, and has relatively few observations on strictly dollarized nations. From a policy perspective, however, it is important to make a distinction between these two common-currency regimes. The reason for this is that the two regimes have important differences in terms of independence of monetary policy, seignorage, and capacity to absorb external shocks. Making a distinction between dollarized and ICU countries is also important from a political economy point of view. As Frieden (2003) has argued, adopting another country’s currency is usually perceived as giving up sovereignty, and has serious political costs. These political costs may be reduced, however, if the country becomes a partner in an ICU. It is even possible that by joining an ICU the country reaps most of the benefits of a common currency without incurring the political costs associated with this measure. The purpose of this paper is to analyze whether common currency countries— both dollarized and ICU countries—have outperformed countries that have a currency of their own. The paper is empirical and proceeds in steps: we first analyze
124
Sebastian Edwards and I. Igal Magendzo
the behavior of all common currency countries and compare them with countries with domestic currencies. We then turn to a comparison of the dollarized and ICU countries. Performing these type of international comparisons, however, is not easy. The problem is how to define an appropriate control group with which to compare the common-currency nations. Since the adoption of a common currency is not a natural experiment, using a broad control group of all countries with a domestic currency is likely to result in biased estimates. In this paper we tackle this issue by using a treatment effects model that estimates jointly the probability of being a common-currency country and outcome equations for GDP growth, inflation, and volatility (Maddala 1983; Heckman, Hidehiko, and Todd 1997; Green 2000; Wooldridge 2002). Some authors—most notably Alesina and Barro (2000b, 2001)—have analyzed the conditions under which a (small) economy would benefit from giving up its currency. In contrast, we are interested in finding out how countries with a long experience with a common-currency regime have performed relative to countries with a currency of their own. Before proceeding, it is useful to point out the ways in which our analysis differs from other related work in this general area. First, we have made an effort to include data on both dollarized and ICU countries. This has not been easy, as most strictly dollarized countries are very small and their data are not included in readily available data sets. After significant effort we were able to obtain data on GDP per capita growth and inflation for twenty strictly dollarized countries. We also use data on thirty-two countries that are members of an independent currency union. Our data set, then, is significantly more general than the data set used by other researchers. Second, we focus directly on the most important macroeconomic variables—real GDP per capita growth, inflation, and growth volatility. Other studies, in contrast, have analyzed performance in an indirect fashion, and have focused on ancillary variables such as the level of international trade and/or interest rates. For instance, Edwards (2001b) and Powell and Sturzenegger (2003) have investigated the way in which the exchange rate/monetary regime affects interest rate behavior, and the cost of capital. On the other hand, Frankel and Rose (2002) have analyzed the way in which currency unions affect bilateral trade and, through this channel, economic growth.8 Third, we use a ‘‘treatment effects model’’ to estimate the way in which dollarization affects the macroeconomic variables of interest. And fourth, we make a distinction between strictly dollarized and ICU countries. The rest of the paper is organized as follows: In section 6.2 we provide a preliminary analysis of historical experiences with common currencies. In section 6.3 we use treatment regressions to analyze the effects of common currencies on a group of macroeconomic variables. In section 6.4 we go a step further and disaggregate the common currency countries into strictly dollarized and ICU countries. In sec-
Hard Currency Pegs and Economic Performance
125
tion 6.5 we undertake a robustness analysis, and we analyze how different samples and estimation techniques affect the results. In particular, we report results on comparative performance obtained from an analysis that uses ‘‘matching estimators’’ techniques, and from using an instrumental variables version of treatment regressions. Finally in section 6.6 we provide some concluding remarks. 6.2 Common-Currency Experiences During 1970–1998: A Preliminary Analysis 6.2.1
Some Background Discussion
Table 6.1 presents a list of fifty-two common-currency countries and territories with available data for the period 1970–1998.9 In compiling this list we have excluded countries that, while having a currency of their own, have had a long tradition of fixed exchange rates, such as Ireland before 1979 and Bermuda.10 We have divided our sample into two groups: (1) countries that use an advanced country’s currency as legal tender, or strictly dollarized countries (in our sample twenty countries satisfy this criterion); (2) other common-currency countries, which we call ICUs. The majority of the ICUs use a currency that is common to the area, but is not issued by any of the individual countries. Five of them, however, use the currency of another emerging country as legal tender. It is important to note that our data set is much larger than that used by other studies. For instance, in the influential studies by Rose and Engel (2002) and Frankel and Rose (2002), there are only twenty-six countries that have data on real GDP per capita. Of these, only seven use another nation’s currency, and only two—Panama and Puerto Rico—use a convertible currency as legal tender and are, thus, strictly dollarized. The countries and territories that have had a strictly dollarized monetary system are very small indeed. Many are city-states well integrated into their neighbors’ economies—Monaco, Liechtenstein, and Andorra are good examples. The largest strictly dollarized countries are Liberia, Panama, and Puerto Rico. However, only Panama and Puerto Rico remain dollarized today; Liberia abandoned the system in the 1980s, when the government of President Samuel Doe decided to issue local currency as a way of avoiding the constraints imposed by the dollarized system.11 An important characteristic of many strictly dollarized economies is that they are extremely open. In most of them there are no controls on capital mobility or on any type of international financial transactions—so much so that in 2001 six out of the thirteen independent dollarized nations were in the Organisation for Economic Co-operation and Development (OECD) list of ‘‘Unfair Tax Havens,’’ or countries whose lax financial regulations allow individuals and
126
Sebastian Edwards and I. Igal Magendzo
Table 6.1 Common Currency Countries with Available Macroeconomics Data CFA Franc Zone
France
Italy
Benin
Andorra (also Spanish Peseta)D
San MarinoD
Burkina Faso
French GuyanaD
Cameroon
French Polynesia
Central African Republic
GuadeloupeD
Chad
MartiniqueD
Comoros
MonacoD
Congo Cote d’Ivoire
New CaledoniaD ReunionD
East Africa
ECCA
Tanzania (until 1979) Uganda (until 1979)
Equatorial Guinea Gabon Guinea-Bissau Mali Niger Senegal Togo
Antigua and Barbuda
Bhutan
St. Kitts and Nevis
Singapore
St. Lucia
Brunei
South Africa
Puerto RicoD
Kenya (until 1979)
India
Liberia (until 1989)D Marshall IslandsD Palau PanamaD
NauruD TuvaluD
Grenada Montserrat
St. Vincent and the Grenadines
D
KiribatiD
Dominica
USA
Micronesia, Fed. States of D
Australia
Lesotho Namibia Swaziland New Zealand Cook IslandsD
Denmark GreenlandD Switzerland LiechtensteinD Belgium LuxembourgD
Note: D : Corresponds to a dollarized country. We provide in bold either the name of the ICU or the name of the country whose currency the common-currency countries have adopted.
corporations to evade taxes. Many dollarized territories in table 6.1—French Guyana, Martinique, for example—have extremely close economic links with the ‘‘home country,’’ including labor mobility and free trade both in goods and in financial claims. These characteristics of the dollarized economies—very small and extremely open—suggest that using a broad control group of all nondollarized countries, which are much larger and not as open, may indeed generate biased results.12 The ICU countries included in our sample are larger than the dollarized nations. These ICU nations have a common central bank that, in principle, can engage in independent monetary policy. The two most important currency unions in table 6.1 currently have a pegged exchange rate relative to a convertible cur-
Hard Currency Pegs and Economic Performance
127
rency: the ECCA has had a fixed exchange rate with respect to the U.S. dollar since 1975. The CFA franc, on the other hand, was pegged to the French franc at the time of its inception in 1948. In January 1994 the CFA franc was devalued and repegged with respect to the French franc, and in January 1999, when the euro was launched, the CFA franc became pegged to the euro. Notice that only a handful of countries in table 6.1 have adopted a common-currency monetary regime within the timeframe of our sample. This means that it is not possible to undertake a diffs-in-diffs analysis. 6.2.2
Unconditional and Unadjusted Comparative Analysis: 1970–1998
In table 6.2 we present comparative data on inflation, per capita GDP growth, and the standard deviation of growth for our common currency countries.13 In order to put things in perspective we also present data on these three variables for an unadjusted control group that includes all countries with a currency of their own. The table contains three panels: panel A includes data on all fifty-two commoncurrency countries; panel B contains data on the strictly dollarized countries, or countries that have adopted a convertible currency; and panel C includes the ICU countries. In each panel we include data on the mean and median for the three outcome macroeconomic variables. In column 3 we present data on mean and median differences between the common-currency countries and the ‘‘with currency’’ control group. The numbers in parentheses are t-statistics for the significance of these differences. The test for the means differences is a standard tstatistic, while the medians differences test is a t-test obtained using a bootstrapping procedure. In making the computations for inflation differentials we have followed Rose and Engel (2002) and have excluded countries with hyperinflations.14 However, excluding these observations only affects the calculation of the means difference (quantitatively, but not qualitatively); it has no discernible effect on the computation of median differences. The results reported in this table indicate that the difference in inflation means is quite sizable and statistically significant; on average, inflation in commoncurrency countries as a group (panel A) has been nine percentage points lower than in countries with their own currency.15 The difference in inflation medians is also negative, much smaller (3.7 percentage points), and still statistically significant. These results also show that the rate of inflation in the strictly dollarized countries (panel B) has been almost one-half that of the ICUs (panel C). In the latter group, however, inflation has still been significantly lower than in countries in the control group. In terms of real per capita GDP growth, the unadjusted comparisons in table 6.2 show that there are no significant differences in the means across any of the two
128
Sebastian Edwards and I. Igal Magendzo
Table 6.2 Inflation, Growth, and Volatility in Common-Currency Countries and Countries with Domestic Currency (1) Dollarized and union countriesa
(2) Other countriesb
(3) Difference* (1) (3)
8.94 (10.13) 3.73 (11.38)
A. All Common-Currency Countries versus Control Group Inflation Mean
7.26
16.20
Median
5.30
9.03
Mean
1.29
1.19
0.10 (0.48)
Median
1.29
1.90
0.61 (4.01)
Mean
4.74
4.17
0.57 (1.91)
Median
3.48
2.89
0.59 (2.67)
Per capita GDP growth
Volatility of Growth
B. Strictly Dollarized versus Control Group Inflation Mean
4.24
16.20
11.96 (7.92)
Median
3.58
9.03
5.45 (13.52)
Mean
1.25
1.19
0.06 (0.17)
Median
1.26
1.90
0.64 (5.06)
Mean
4.77
4.17
Median
3.34
2.89
0.60 (1.40) 0.45 (1.53)
Per capita GDP growth
Volatility of Growth
Hard Currency Pegs and Economic Performance
129
Table 6.2 (continued) (1) Monetary Unionsa
(2) Other countriesb
(3) Difference* (1) (3)
Mean
8.73
16.20
Median
6.94
9.03
7.47 (6.95) 2.09 (4.08)
Mean
1.33
1.19
0.14 (0.52)
Median
1.29
1.90
0.61 (1.63)
Mean
4.71
4.17
0.54 (1.54)
Median
3.65
2.89
0.76 (2.32)
C. ICUs versus Control Group Inflation
Per capita GDP growth
Volatility of Growth
a
Number of observations with data for inflation is 760, of which 249 are strictly dollarized and 511 belong to an ICU. There are 1,332 observations with data for per capita growth, of which 526 belong to a strictly dollarized country and 806 to an ICU. b Number of observations with data on inflation is 2,732 and there are 3,907 observations with data for per capita GDP growth. * Numbers in parentheses are t-statistics.
common-currency groups and the control group. The results also indicate, however, that the medians difference is significantly negative: the median rate of growth in the two common-currency groups has been significantly lower—in a statistical sense—than in the control group of countries with a currency of their own. Finally, our results show that both groups of common-currency countries have experienced greater growth volatility than the control group: both means and medians differences are significantly positive. Although the comparisons reported in table 6.2 are informative, they are subject to two potential limitations. First, these are unconditional comparisons, as no effort has been made to control for other factors potentially affecting macroeconomic performance. Second, the control group may not be the appropriate one. If this is the case, the results presented in table 6.2 may be subject to a treatment bias.16 We address both of these problems in the econometric analysis reported in sections 6.3 through 6.5 of this paper.
130
6.3
Sebastian Edwards and I. Igal Magendzo
Common Currencies and Macroeconomic Performance
We are interested in investigating whether countries with a common-currency regime have had a better macroeconomic performance than countries with a currency of their own. We focus on three dimensions of performance: GDP per capita growth, inflation, and growth volatility. In principle, the exchange rate regime will affect the growth process—both the first and second moments— through three potential channels. First, a lower cost of capital—usually associated with ‘‘hard peg’’ economies—will result in a higher rate of physical capital accumulation and a higher growth rate of potential output. Second, a high level of international trade—which, as Rose and his coauthors have persuasively shown, is associated with common-currency regimes—is likely to have a positive effect on total factor productivity (TFP) growth, and on the growth rate of potential output. This effect has been emphasized in a number of endogenous growth models, and operates through the effect of openness on the accumulation of knowledge. And third, since the exchange rate regime will affect the country’s ability to accommodate external terms of trade shocks, it will affect growth volatility and, possibly, average growth. Indeed, if as Meade (1950) and Corden (2002), among others, have argued, countries with a hard peg—and in particular countries with a common currency—have more difficulty accommodating external disturbances, they will tend to exhibit a more volatile rate of growth than countries with a domestic currency. The relationship between common currencies and inflation is rather straightforward. A strictly dollarized system will tend to solve the inflationary bias associated with discretionary monetary policy, allowing a small country to share the anchor country’s rate of inflation. A number of studies on Panama—the strictly dollarized country par excellence—provide support for this view. Indeed, Panama’s mean rate of inflation for 1970–1997 was 3.4 percent per year; the median rate during this period was 1.9 percent. Neither of these figures is significantly different from U.S. inflation during that period. As Alesina and Barro (2000b) have argued, the discretionary inflation bias will only be eliminated if the local currency is (perceived to be) linked irrevocably to the anchor currency, and it will tend to have a significantly lower rate of inflation than countries with a currency of their own. 6.3.1
The Empirical Model
Our main interest is to undertake a comparative analysis of the conditional effect of a common currency on macroeconomic performance. In order to do this, we estimate jointly an outcome equation and the probability of being a common-
Hard Currency Pegs and Economic Performance
131
currency country. As in the preceding section, we consider three outcome variables: GDP per capita growth, inflation, and volatility. In the estimation of the inflation model we use unbalanced panel data for 169 countries, covering 1971– 1998. For the growth model we used two panel data sets: the first one is comprised of five-year averages; the second data set is a panel with yearly data. Finally, for the volatility model we use averages of five-year periods to calculate the standard deviation of GDP growth.17 The empirical treatment effects model may be written as follows: yjt ¼ xjt b þ gdj þ m jt djt ¼
1; 0;
if djt > 0 otherwise
djt ¼ wjt a þ e jt :
ð6:1Þ ð6:2Þ ð6:3Þ
Equation 6.1 is the macroeconomic performance equation, where yjt stands for each of the macroeconomic outcome variables of interest in country j and period t; xjt is a vector of covariates that captures the role of traditional determinants of economic performance; djt is a dummy variable (the treatment variable) that takes a value of 1 if country j in period t is a common-currency country, and 0 if the country has a currency of its own. Accordingly, g is the parameter of interest: the effect of the treatment on the outcome. The decision to have a common currency is assumed to be the result of an unobserved latent variable djt , described in equation 6.2. djt , in turn, is assumed to depend linearly on vector wjt . Some of the variables in wjt may be included in xjt (Maddala 1983, 120).18 b and a are parameter vectors to be estimated. m jt and e jt are error terms assumed to be bivariate normal, with a zero mean and a covariance matrix given by s v : ð6:4Þ v 1 If the performance and common-currency equations are independent, the covariance term v in equation 6.4 will be 0. Under most plausible conditions, however, it is likely that this covariance term will be different from 0. Greene (2000) has shown that if equation 6.1 is estimated by least squares, the treatment effect will be overestimated. Traditionally, this problem has been tackled by estimating the model using a two-step procedure (Maddala 1983). In the first step, the treatment equation 6.2 is estimated using probit regressions. From this estimation a hazard is obtained for each j t observation. In the second
132
Sebastian Edwards and I. Igal Magendzo
step, the outcome equation 6.1 is estimated with the hazard added as an additional covariate. From the residuals of this augmented outcome regression, it is possible to compute consistent estimates of the variance-covariance matrix 6.4. Instead of the traditional two-stage method, in this paper we use a more efficient maximum likelihood procedure to estimate the model in equations 6.1 through 6.4 jointly.19 As shown by Greene (2000), the log likelihood for observation k is given by equations 6.5 and 6.5 0 : ( ) pffiffiffiffiffiffi wk a þ ðyk xk b dÞv=s 1 yk xk b d 2 pffiffiffiffiffiffiffiffiffiffiffiffiffi Lk ¼ log F log 2ps; ð6:5Þ 2 s 1 v2 if dk ¼ 1 (
wk a ð yk xk bÞv=s pffiffiffiffiffiffiffiffiffiffiffiffiffi Lk ¼ log F 1 v2
)
1 2
yk xk b s
2 log
pffiffiffiffiffiffi 2ps;
ð6:5 0 Þ
if dk ¼ 0: The model in equations 6.1–6.4 will satisfy the consistency and identifying conditions of mixed models with latent variables if the outcome variable yjt is not a determinant (directly or indirectly) of the treatment equation—that is, if y is not one of the variables in w in equation 6.3.20 For the cases of per capita GDP growth and volatility this is a reasonable assumption. Although the level of GDP per capita may affect the probability of having a common currency, its rate of change, or the second moment of its rate of change, is unlikely to have an impact on the decision to have a domestic currency. This consistency and identifying restriction is also met in the case of the inflation model. Indeed, in every one of the countries in our sample, the decision to use a common currency can be traced historically to variables that are structural in nature, including the country’s size and its cultural and political relation with the anchor country. Moreover, what may affect the decision to dollarize is the propensity to have a high inflation rate. This propensity, however, is indeed captured by some of the variables in the wjt vector in equation 6.3. However, in order to check for the robustness of the results obtained from the estimation of the model in equations 6.1 through 6.5, in section 5 we present results obtained from the instrumental variables estimation of a treatment effects model for inflation. In the estimation of the model 6.1–6.5, we also impose some exclusionary restrictions; that is, a number of the wjt covariates included in equation 6.3 are not included in the outcome equation 6.1. These exclusionary restrictions are not required for identification of the parameters, but they are generally recommended as a way of addressing issues of collinearity.21
Hard Currency Pegs and Economic Performance
6.3.2
133
Basic Results for Common-Currency Countries as a Group
In this section we report the results obtained from the estimation of the treatment effects model given by equations 6.1–6.4. The ‘‘treatment group’’ is defined as all countries without a currency of their own. That is, the dummy variable dtj takes a value of 1 if in period t country j does not have a domestic currency; no distinction is made, at this point, between strictly dollarized and ICU countries (see, however, the results in section 6.4). The data set covers 1970–1998, and includes 199 counties and territories. The number of observations varies depending on the outcome variable considered. There are 3,122 observations on inflation and 5,233 observations on growth per capita. When using five-year averages—both in the growth and volatility models—the panel has 950 observations.22 6.3.2.1 The Treatment Equation In a highly influential article, Mundell (1961, 181) argued that the ‘‘optimum currency area is the region.’’ By this he meant that regional considerations— geographical proximity and the existence of factor mobility, among others—were more important than national (or sovereign) considerations in determining optimal currency areas. This region-based approach has been present in most subsequent work on the subject of optimal currency areas.23 Following Mundell’s insight, we include a number of regional variables in our empirical analysis of the probability of being a common-currency country. More specifically, in the specification of the treatment equation 6.3, we included the following covariates that encapsulate the importance of the region: 1. A dummy variable that takes the value of 1 if the economy in question is an independent nation and 0 if it is a territory. Since factor mobility is much lower across independent nations than between a dependent territory and the home country, we expect the coefficient of this variable to be negative in the estimation of equation 6.3. 2. A dummy variable that takes the value of 1 if the country in question is an island or archipelago. Since island-archipelago countries are relatively isolated, they tend to be self-contained regions. We expect this variable to have a negative coefficient in 6.3. 3. A dummy variable that takes the value of 1 if the country has a common border with a nation whose currency is defined, by the IMF, as a ‘‘convertible currency.’’ We call this variable ‘‘border,’’ and we expect its estimated coefficient to have a positive sign in equation 6.3. In addition to these regional variables,24 the following covariates were also included in the specification of the treatment equation 6.3:
134
Sebastian Edwards and I. Igal Magendzo
4. The log of population, measured in millions of people, as an index of the country’s size. We expect the estimated coefficient of this variable to be negative, indicating that larger countries are less likely to use another nation’s currency. 5. The log of initial (1970) GDP, taken as a measure of the country’s economic size. We also expect the coefficient of this variable to be negative. 6. An indicator of the degree of openness of the economy. For the majority of countries and years we used the Sachs and Warner (1995) openness index. We used data from a variety of sources to supplement the Sachs-Warner index for those countries and years not covered in their sample.25 As Frieden (2003) has argued, its estimated coefficient is expected to be positive. 7. A variable that measures the (log of the) distance between each country and the global markets; in defining this ‘‘distance variable’’ we followed Leamer (1999). We expect its estimated coefficient to be positive, indicating that countries that are less integrated into world markets will have a lower probability of being common-currency countries. 6.3.2.2 The Outcome Equations We specified (and estimated) three different models, for GDP per capita growth, inflation, and growth volatility. A difficulty we faced in undertaking this analysis is that many common-currency countries have limited data availability. For instance, very few of the strictly dollarized countries have data on education attainment or on some other variables traditionally included in growth empirical analyses (Barro and Sala-i-Martin 1995, Barro 1996). Indeed, popular data sets, such as the World Bank World Development Indicators (WDI), the International Financial Statistics (IFS), or the Barro-Lee (1996) data set, include data on only a handful—three or four—strictly dollarized countries. Nevertheless, and after searching in a number of alternative data sources, we were been able to include a number of covariates in the outcome equations 6.1 for per capita growth, inflation, and volatility.26 In the estimation of the GDP growth model we included, as customary, initial GDP, a measure of openness, a variable that captures the country’s geographical location, regional dummies (no dummies were included for Asia, which is defined as the reference region),27 and the common-currency dummy. As Sachs (2000), among others, has argued, countries located close to the equator tend to grow more slowly, after controlling for other factors, than nations in other parts of the world. Our geography variable—which we call ‘‘tropics’’—is defined as the (normalized) absolute distance from each country to the equator. We expect its coefficient to be negative, capturing the fact that, with other things constant, countries closer to the tropics will tend to grow at a slower rate than countries in other geographical areas.28
Hard Currency Pegs and Economic Performance
135
In the inflation model, in addition to the common-currency dummy, we included openness, lagged inflation (as a measure of persistence), inflation lagged twice, the tropics geography variable, distance, and the regional dummies. Finally, in the volatility model we include the following covariates in the outcome equation: initial GDP, openness, the regional dummies, and the common-currency dummy. As in the other models, in some of the specifications we introduced regional variables. In the volatility model we expect that the estimated coefficient of both openness and initial GDP will be negative. 6.3.2.3 Results In table 6.3 we summarize the results obtained from the estimation of the treatment effects model for GDP per capita growth. Table 6.4 contains the results for inflation, and table 6.5 those for growth volatility. Each of these tables contains two panels. The upper panel includes the results from the outcome equation; the lower panel contains the estimates for the treatment equation. Probability of Being a Common-Currency Country As may be seen from these tables, the results are similar across models and are quite satisfactory. The vast majority of the coefficients have the expected signs and are statistically significant at conventional levels. They clearly indicate that the probability of being a common-currency country is higher for very small, not independent countries (or territories). Being an island reduces the probability of having a common currency, as does greater distance from world markets. The estimated coefficient of ‘‘border’’ is not significant at conventional levels and has a negative sign (see, however, the discussion in section 6.4).29 GDP per Capita Growth In table 6.3 we present the results obtained from the estimation of the growth model. We report results from six systems: the first three were estimated using annual data, while the last three were estimated using fiveyear averages. As may be seen, the traditional regressors have the expected signs and are significant at conventional levels. In terms of the monetary regime, these results show that the coefficient of the common-currency dummy is positive and statistically significant for all specifications. Its point estimate ranges from 0.749 to 1.204. This suggests that, during the period under consideration and after controlling for other factors, countries with a common-currency regime experienced a higher rate of growth of GDP per capita than countries with a currency of their own. These results suggest that the growth advantage of common-currency countries amounted, on average, to approximately one percentage point per year. Notice that these results are quite different from the simple means differences reported in table 6.2: while according to those results there have been no
Table 6.3 Common Currencies and GDP Growth: A Treatment Effects Model* (Maximum Likelihood)
Log(GDP0 )
Model 1
Model 2
Model 3
Model 4
Model 5
Model 6
0.399 (5.87)
0.449 (5.24)
0.506 (5.82)
0.438 (4.57)
0.464 (3.85)
0.532 (3.85)
2.573 (9.86)
2.536 (9.55)
2.465 (9.26)
2.587 (7.04)
2.509 (6.82)
2.429 (6.61)
OPEN TROPIC
—
3.668 (3.70)
— 0.872 (2.71)
—
0.754 (2.15) 0.747 (2.27)
1.021 (2.86) 1.813 (4.15)
LAC
—
0.136 (0.46)
0.234 (0.79)
MENA
—
0.493 (1.30)
NORTHAM
—
SASIA
—
AFRICA
—
—
4.308 (3.11)
—
—
0.816 (1.94) 0.682 (1.48)
1.204 (2.17) 1.918 (3.16)
—
0.145 (0.35)
0.266 (0.64)
0.849 (2.17)
—
0.608 (1.13)
1.010 (1.84)
0.126 (0.16)
0.416 (0.52)
—
0.176 (0.16)
0.460 (0.41)
0.877 (1.73) 1.630 (5.40)
0.536 (1.04) 1.509 (4.98)
1.106 (1.54) 1.520 (3.52)
0.728 (1.00) 1.362 (3.15)
3.297 (6.86)
4.234 (6.93)
3.861 (6.24)
3.361 (4.91)
4.112 (4.77)
3.659 (4.19)
Log(GDP0 )
0.462 (29.38) 0.129 (6.88)
0.463 (29.47) 0.130 (6.91)
0.463 (29.46) 0.129 (6.88)
0.490 (12.93) 0.122 (2.73)
0.493 (13.12) 0.123 (2.73)
0.492 (13.06) 0.121 (2.70)
INDEP
1.025 (12.98)
1.026 (12.97)
1.025 (12.95)
0.554 (3.47)
0.535 (3.34)
0.539 (3.37)
0.116 (1.43) 0.115 (1.57)
0.118 (1.45) 0.116 (1.58)
0.118 (1.45) 0.115 (1.57)
0.112 (0.43) 0.240 (1.42)
0.113 (0.65) 0.241 (1.43)
0.122 (0.64) 0.237 (1.40)
0.992 (14.01)
1.006 (14.27)
1.005 (14.25)
0.848 (5.09)
0.898 (5.45)
0.889 (5.40)
DISTANCE
0.524 (7.16)
0.521 (7.13)
0.523 (7.16)
0.749 (4.50)
0.739 (4.43)
0.748 (4.48)
Constant
3.531 (4.73)
3.574 (4.78)
3.552 (4.75)
1.390 (0.83)
1.536 (0.92)
1.431 (0.85)
Dummy EUROPE
Constant
1.079 (2.17)
— —
Treatment equation Log(POP)
BORDER OPEN ISLAND
Number of observations
5233
5233
5233
950
950
950
LR chi2
5.58
5.70
5.44
3.54
3.54
3.54
Prob > chi2
0.018
0.016
0.020
0.060
0.060
0.060
* The upper panel contains the outcome equation. The lower panel contains the estimation of the treatment equation, or equation on the probability of being a common-currency country. The numbers in parentheses are t-statistics.
Hard Currency Pegs and Economic Performance
137
differences in average rates of growth across the two groups of countries, the estimates in table 6.3 indicate that common-currency countries have grown at a significantly faster rate than countries with a currency of their own. The chi square test for the independence of the treatment and outcome equations indicates that in all specifications the null hypothesis of independence across the equations is rejected at conventional levels.30 Inflation The results for the inflation model are reported in table 6.4. As may be seen from the outcome equation in the upper panel, the common-currency dummy is negative and significant in every one of the specifications. The point estimates range from 11.99 to 14.43, not only confirming that inflation has historically been lower in the common-currency countries, but also indicating that the common-currency advantage is somewhat larger than what the simple mean differences results reported in table 6.2 suggest. The other covariates in the inflation regressions reported in table 6.4 have the expected signs, and are statistically significant. In particular, these results indicate that more open countries have lower inflation, as do countries that are geographically closer to the global markets. Inflation appears to have some degree of persistence, and the regional dummies indicate that, relative to the benchmark (Asia), Latin America and Africa have had a significantly higher rate of inflation. As the w 2 statistics show, the null hypothesis of independent equations is rejected at conventional levels. In order to investigate the robustness of these results and deal with potential endogeneity problems, we also estimated the inflation model using an instrumental variables treatment approach. The results are presented in section 6.5. Volatility Table 6.5 contains the results for the volatility models. The null hypothesis of independent equations is rejected at conventional levels—the w 2 statistics range from 29.0 to 34.7—and the dummy variables for common currency are significantly positive, indicating that countries without a domestic currency have experienced a higher degree of growth volatility than countries with a currency of their own. Openness reduces volatility—a result that is in line with a number of theoretical results in international economics.31 In addition, our estimates indicate that, after controlling for other factors, countries that are closer to the equator have exhibited a higher degree of volatility. Also, countries with a higher initial level of GDP per capita have had a somewhat higher degree of volatility (the point estimate of this coefficient is, however, rather low). According to these results, growth in the countries of the Middle East and North Africa (MENA) has been particularly volatile. In what appears to be a counterintuitive result, the coefficient for the Europe dummy is positive, although not significant. The reason for this apparent anomaly is that the Eastern and Central European nations are part
138
Sebastian Edwards and I. Igal Magendzo
Table 6.4 Common Currencies and Inflation: A Treatment Effects Model* Model 1
Model 2
Model 3
8.591 (9.14)
9.089 (9.31)
8.059 (7.84)
1.575 (1.82)
3.256 (2.62)
6.941 (3.77)
INFLATIONT1
0.089 (18.57)
6.252 (1.89) 0.089 (18.44)
9.501 (2.47) 0.085 (17.93)
INFLATIONT2
0.003 (2.41)
0.003 (2.38)
0.003 (2.58)
Dummy
12.233 (7.99)
11.991 (7.80)
14.434 (9.47)
OPEN DISTANCE TROPIC
—
10.226 (3.88)
EUROPE
—
—
LAC
—
—
MENA
—
—
10.259 (8.72) 3.997 (2.08)
NORTHAM
—
—
2.546 (0.80)
SASIA
—
—
1.974 (1.03)
AFRICA
—
—
10.631 (8.76)
4.678 (0.62)
11.381 (1.00)
11.381 (1.00)
0.585 (23.75) 0.320 (9.31)
0.585 (23.75) 0.320 (9.31)
0.585 (23.75) 0.320 (9.31)
INDEP
1.319 (9.06)
1.319 (9.06)
1.319 (9.06)
BORDER
0.671 (5.29)
0.671 (5.29)
0.671 (5.29)
0.146 (1.41)
0.146 (1.41)
0.146 (1.41)
1.511 (14.73)
1.511 (14.73)
1.511 (14.73)
DISTANCE
0.350 (3.03)
0.350 (3.03)
0.350 (3.03)
Constant
14.843 (4.73)
14.843 (4.73)
14.843 (4.73)
Constant Treatment equation Log(POP) Log(GDP0 )
OPEN ISLAND
Hard Currency Pegs and Economic Performance
139
Table 6.4 (continued)
Number of observations
Model 1
Model 2
Model 3
3122
3122
3122
LR chi2
6.75
6.75
6.75
Prob > chi2
0.027
0.027
0.027
* The upper panel contains the outcome equation. The lower panel contains the estimation of the treatment equation, or equation on the probability of being a common-currency country. The numbers in parentheses are t-statistics.
of the World Bank European region. Finally, notice that the estimated coefficients for the common-currency dummy are significantly larger than the unadjusted mean differences in volatility presented in table 6.2. 6.4 Strict Dollarization and Independent Currency Unions: Is There a Difference? The results reported in the preceding section grouped all common-currency countries together, implicitly assuming that strict dollarization and ICU are equivalent monetary regimes. This, however, need not be the case. As was argued in section 1, there are a number of differences between them in terms of independence of monetary policy, seignorage revenue, and capacity to absorb external shocks. In this section we break up the common-currency countries into strictly dollarized countries and ICUs. We then estimate treatment effects models for GDP growth, inflation, and volatility for each of these two groups separately.32 The specification of both the treatment and outcome equations are similar to those reported in tables 6.3–6.5 for the common-currency nations. The results for GDP growth are in table 6.6. Those for inflation are in table 6.7 and those on volatility are in table 6.8. In each of these tables we report a limited number of equations. The results for alternative specifications were similar, and are available on request. An inspection of the results suggest the following patterns: There are important differences in the results for the treatment equation for ICU and strictly dollarized countries. More specifically, the results for the probability of not having a domestic currency differ in the following respects for these two samples: first, while the coefficient of ‘‘border’’ is positive (and significant at either the 5 or 10 percent level) in the strict dollarization models in table 6.7 (models 1 and 3), it is significantly negative in the estimates for the ICU countries (table 6.7, models 2 and 4). This indicates that geographical proximity to a
•
Table 6.5 Common Currencies and Growth Volatility: A Treatment Effects Model Model 1
Model 2
Model 3
0.026 (2.78)
0.034 (3.24)
0.024 (2.08)
2.471 (6.78)
2.392 (6.55)
2.148 (5.92)
1.468 (1.65)
3.289 (2.51)
2.178 (3.27)
Log(GDP0 ) OPEN TROPIC
— 2.456 (5.83)
EUROPE
—
—
2.457 (3.80) 0.660 (1.14)
LAC
—
—
1.021 (1.01)
MENA
—
—
2.334 (4.45)
NORTHAM
—
—
0.871 (0.81)
SASIA
—
—
AFRICA
—
Dummy
1.031 (1.49) 0.092 (0.23)
— 2.353 (3.47)
2.225 (3.27)
2.225 (3.27)
0.464 (12.18) 0.151 (3.44)
0.468 (12.26) 0.159 (3.59)
0.465 (12.10) 0.160 (3.65)
1.336 (6.82)
1.341 (6.82)
1.398 (7.11)
OPEN
0.090 (0.48) 0.184 (1.96)
0.078 (0.42) 0.186 (1.36)
0.075 (0.40) 0.196 (1.31)
ISLAND
1.170 (7.13)
1.172 (7.12)
1.139 (6.91)
DISTANCE
0.586 (3.58)
0.541 (3.23)
0.571 (3.36)
Constant
3.514 (2.07)
4.016 (2.31)
3.752 (2.12)
Constant Treatment equation Log(POP) Log(GDP0 ) INDEP BORDER
Number of observations LR chi2 Prob > chi2
950
950
950
34.71
28.98
32.99
0.0
0.0
0.0
* The upper panel contains the outcome equation. The lower panel contains the estimation of the treatment equation, or equation on the probability of being a common-currency country. The numbers in parentheses are t-statistics.
Table 6.6 Independent Currency Unions, Strict Dollarization, and GDP Growth: A Treatment Effects Model
Log(GDP0 ) OPEN TROPIC Dummy ICU
Model 1
Model 2
Model 3
Model 4
0.566 (6.46) 3.227 (10.66) 3.041 (3.03) 2.041 (4.31)
0.545 (4.40) 3.319 (7.66) 3.299 (2.31) 2.723 (3.68)
0.501 (5.18) 2.658 (9.68) 2.601 (2.47)
0.591 (4.72) 3.281 (7.60) 3.300 (2.31)
—
—
— 1.644 (3.70) 0.050 (0.16) 0.861 (2.17) 0.640 (0.81) 0.329 (0.64) 2.113 (6.46) 4.415 (6.93)
— 1.688 (3.78) 0.150 (0.34) 1.070 (1.90) 0.719 (0.64) 0.454 (0.62) 2.184 (3.68) 4.315 (4.71)
0.267 (0.63) 1.606 (3.57) 0.102 (0.32) 0.863 (2.14) 0.477 (0.81) 0.413 (0.77) 1.690 (4.92) 4.181 (6.52)
Constant
0.345 (16.62) 0.105 (5.32) 0.808 (7.73) 1.176 (7.64) 3.831 (8.40) 0.744 (8.47) 1.240 (11.90) 4.826 (4.70)
0.386 (7.83) 0.095 (1.94) 0.238 (1.70) 1.032 (3.05) 3.654 (3.23) 0.692 (3.30) 1.407 (5.83) 6.345 (2.69)
0.524 (22.71) 0.030 (0.80) 1.253 (12.78) 0.169 (1.61) 1.047 (10.35) 0.561 (5.68) 0.347 (2.79) 4.003 (3.06)
0.542 (9.44) 0.080 (0.90) 1.333 (5.77) 0.388 (1.54) 1.276 (5.49) 0.89 (2.01) 0.581 (2.09) 2.434 (0.83)
Number of observations LR chi2 Prob > chi2
4707 4.16 0.040
856 2.68 0.102
4427 3.97 0.050
856 4.23 0.038
Dummy Dollarization EUROPE LAC MENA NORTHAM SASIA AFRICA Constant
0.532 (0.84) 1.666 (2.61) 0.098 (0.21) 0.998 (1.73) 0.630 (0.55) 0.594 (0.87) 1.580 (3.19) 4.181 (6.52)
Treatment equation Log(POP) Log(GDP0 ) INDEP BORDER OPEN ISLAND DISTANCE
* The upper panel contains the outcome equation. The lower panel contains the estimation of the treatment equation, or equation on the probability of being a common-currency country. The numbers in parentheses are t-statistics. Models 1 and 3 annual data. Models 2 and 4 five year averages.
142
Sebastian Edwards and I. Igal Magendzo
Table 6.7 Independent Currency Unions, Strict Dollarization, and Inflation: A Treatment Effects Model Model 1
Model 2
10.137 (8.26)
8.337 (7.55)
6.983 (1.94)
6.223 (3.07)
INFLATIONT1
11.231 (2.76) 0.084 (17.96)
11.031 (2.59) 0.084 (16.75)
INFLATIONT2
0.003 (2.42)
0.003 (2.46)
Dummy ICU
20.970 (9.52)
OPEN DISTANCE TROPIC
Dummy Dollarization
— 10.026 (4.79)
—
LAC
9.760 (1.31) 10.724 (7.42)
9.413 (3.33) 10.776 (8.45)
MENA
3.690 (2.00)
3.800 (1.86)
NORTHAM
3.467 (1.05)
2.802 (0.83)
SASIA
2.107 (1.04)
1.169 (0.56)
12.232 (8.76) 51.319 (2.92)
12.167 (8.62) 46.204 (2.92)
0.492 (16.30) 0.250 (6.85)
0.881 (14.78) 0.433 (5.82)
0.260 (1.44)
1.967 (9.80)
0.146 (1.41)
2.418 (11.05)
1.581 (12.76)
1.352 (7.63)
DISTANCE
0.180 (1.29)
0.031 (0.41)
Constant
7.443 (4.85)
15.533 (4.85)
EUROPE
AFRICA Constant Treatment equation Log(POP) Log(GDP0 ) INDEP OPEN ISLAND
Hard Currency Pegs and Economic Performance
143
Table 6.7 (continued)
Number of observations
Model 1
Model 2
2925
2676
LR chi2
6.15
4.29
Prob > chi2
0.02
0.04
* The upper panel contains the outcome equation. The lower panel contains the estimation of the treatment equation, or equation on the probability of being a common-currency country. The numbers in parentheses are t-statistics.
convertible-currency nation is only important in determining the probability of being a dollarized country. Second, the coefficient for openness is significantly positive, as expected, in the dollarized models (table 6.7); it is negative in the ICU models. As we argue in section 6.5, the CAF countries largely drive this result. These important differences in the probit results support the hypothesis that it is not correct to consider—as most authors have done until now—the two types of common-currency regimes as a homogeneous group. In the GDP growth models, the dummy for strict dollarization is never significant, indicating that after controlling for simultaneity and other factors, there is no discernible difference in growth performance between dollarized countries and countries with a currency of their own.
•
In the GDP growth models, the dummy for ICUs is significantly positive, with a point estimate ranging from 1.36 to 2.72. This suggests that, with other things given, ICU countries have grown at a significantly faster rate than countries with a currency of their own. Moreover, according to our point estimates the growth effect of an ICU monetary regime appears to have been quite large.
•
In the inflation models the dummies for both the strictly dollarized and the ICU countries are significantly negative, confirming that both types of commoncurrency countries have been able to have a significantly lower rate of inflation than the with-domestic-currency countries.
•
In the inflation models the point estimates for the dummies are quite different for the two groups of countries. In fact, these results suggest that the low inflation advantage is greater for the ICU countries. This contrasts with the unadjusted comparisons reported in table 6.2, which suggested that strictly dollarized countries had a lower average inflation rate than ICU countries.
•
The results for the volatility model show that the dummies’ coefficients are significantly positive for both groups of countries. This supports the idea that superhard-peg countries’ regimes tend to result in higher real volatility.
•
Table 6.8 Independent Currency Unions, Strict Dollarization, and Volatility: A Treatment Effects Model Model 1
Model 2 0.287 (2.22) 2.549 (7.00) 4.389 (3.15)
Dummy ICU
0.321 (2.70) 1.771 (4.36) 1.933 (1.54) 2.271 (3.62)
Dummy Dollarization
—
Log(GDP0 ) OPEN TROPIC
—
0.654 (1.11) 0.053 (1.21) 2.567 (4.87) 1.115 (1.05) 0.569 (0.83) 0.607 (1.39) 2.035 (2.35)
2.106 (4.07) 0.644 (1.07) 0.087 (0.21) 2.330 (4.34) 0.880 (0.82) 1.403 (1.95) 0.305 (0.66) 3.756 (4.37)
Constant
0.370 (7.59) 0.133 (2.91) 1.115 (4.26) 0.824 (2.29) 3.795 (3.36) 0.969 (4.73) 1.130 (4.79) 2.928 (1.25)
0.548 (9.50) 0.139 (1.55) 1.608 (6.44) 0.294 (1.17) 1.167 (4.81) 0.689 (2.86) 0.338 (1.11) 5.569 (1.75)
Number of observations LR chi2 Prob > chi2
855 14.19 0.0
806 14.78 0.0
EUROPE LAC MENA NORTHAM SASIA AFRICA Constant Treatment equation Log(POP) Log(GDP0 ) INDEP BORDER OPEN ISLAND DISTANCE
* The upper panel contains the outcome equation. The lower panel contains the estimation of the treatment equation, or equation on the probability of being a common-currency country. The numbers in parentheses are t-statistics.
Hard Currency Pegs and Economic Performance
6.5
145
Robustness Analysis and Further Results
In this section we deal with some extensions, we investigate the robustness of the results, we address potential endogeneity problems in the inflation equation, and we inquire as to what is behind our reported results. 6.5.1
Nonparametric Methods
It is possible that the specification forms chosen for the outcome equations affect our reported results. In particular, the linearity of these equations may affect the estimates of the treatment coefficient. In order to investigate whether this is an important factor, we undertook a nonparametric analysis based on ‘‘matching estimators’’ (see Blundell and Costa Dias 2000). A general advantage of this nonparametric method is that no particular specification of the underlying model has to be assumed. Matching estimators pair each common-currency country with countries from the with-domestic-currency group.33 If the sample is large enough, for each treated (common-currency) observation we should find, in principle, at least one untreated observation with exactly the same characteristics. Each of these properly selected untreated observations provides the required counterfactual for our comparative analysis.34 The problem is that under most general conditions it is not possible to find an exact match between a treated and untreated observation. The matching estimator method focuses on estimating an average version of the parameter of interest.35 That is, the matching estimator consists of obtaining the difference in outcome as an average of the differences with respect to similar—rather than identical—untreated outcomes. Rosenbaum and Rubin (1983) have shown that an efficient and simple way to perform this comparison is to rely on a propensity score, defined as the probability of participation or treatment: PðxÞ ¼ ProbðD ¼ 1=xÞ: In our case, this is the probability of a country being a common-currency country. This reduces a multidimensional problem to a one-dimensional problem, provided that we can estimate PðxÞ. Instead of matching countries directly on all of their characteristics, we can compare countries with similar probabilities of being a common-currency country. An additional advantage of this method is that the estimation of the treatment on the treated is not affected by the lack of data on some of the other variables affecting the outcome. In that regard, then, this method provides us with an elegant way of handling potential problems emanating from omitted variables in the outcome equations.36
146
Sebastian Edwards and I. Igal Magendzo
Table 6.9 Common Currencies, ICUs, and Dollarization: Matching Estimators, Mean Differences* ICUs
Dollarized
All common currency
6.48 (8.75)
7.06 (11.84)
6.38 (10.98)
B. GDP per capita growth
0.53 (1.97)
1.11 (2.84)
0.91 (.88)
C. Growth Volatility
0.86 (2.95)
1.18 (1.98)
0.27 (1.88)
A. Inflation
* The results reported in this table correspond to the five nearest neighbors without replacement. The numbers in parentheses are t-statistics.
In this section we report results obtained from using a simple-average nearestneighbor estimator. According to this method, for each treated observation, we select a predetermined number of untreated nearest neighbors. The nearest neighbors of a particular treated observation i are defined as those untreated observations that have the smallest difference in propensity score with respect to i. If we choose to use nn nearest neighbors, we set Wij ¼
1 nn
for the observations that have been selected; for other observations we set Wij ¼ 0. We applied this method to both one nearest neighbor and five nearest neighbors. The results we report in table 6.9 were obtained using a multiple treatments procedure that assumes that at any point in time there are two possible (and alternative) ‘‘treatments,’’ a dollarization treatment, and an ICU treatment. The untreated group is comprised of all those countries with a currency of their own.37 The matching results in table 6.9 may be summarized as follows: first, we confirm that both types of hard peg monetary regimes have resulted in lower inflation than regimes with a domestic currency. Second, the GDP growth estimates confirm that ICU countries have grown at a faster rate than countries with a currency of their own; the opposite is the case for dollarized countries. In fact, these results indicate that countries with a dollarized regime have grown at a significantly lower rate than countries with a domestic currency. This result contrasts with the regression results reported in table 6.7, which suggested that there had been no difference in the rates of growth in the two groups. And third, the estimates for volatility indicate that in both hard peg regimes volatility has been
Hard Currency Pegs and Economic Performance
147
higher than in countries with a currency of their own. These mean differences are only significant when five nearest neighbors are used. Overall, we interpret the results from the matching exercise as providing broad support to the findings reported in the preceding section: countries with both hard peg regimes have had lower inflation than countries with a currency, only ICU countries have had higher growth, and both hard pegs appear to have had a more volatile real economy than countries with a domestic currency. 6.5.2
Redefining ‘‘Very Rapid Inflation’’
In our inflationary analysis we excluded countries with extremely rapid inflation, or ‘‘hyperinflation countries.’’ In the estimates reported in tables 6.4 and 6.7 the sample excluded countries with a rate of inflation in excess of 200 percent per year. It is possible, however, that by still allowing highly inflationary countries in the sample, the estimates obtained are being driven by extreme or outlier observations. In order to investigate this issue we re-estimated the inflationary equation under alternative definitions of ‘‘extremely rapid inflation.’’ More specifically, in the alternative estimates we first excluded observations with an annual rate of inflation in excess of 100 percent; we then repeated the exercise excluding observations with inflation in excess of 50 percent per year. The results obtained when these new samples were used confirmed those reported earlier in the sense that inflation is significantly lower in commoncurrency countries. Interestingly, however, under these new definitions of very rapid inflations the stability advantage of ICU countries appears to be even greater than that reported in section 6.4. 6.5.3
Instrumental Variables Estimation of the Inflation Model
The results presented in the preceding sections assumed that the treatment did not depend on the outcome variable.38 That is, in estimating the treatment models in sections 6.3 and 6.4 we assumed that the treatment (common-currency) dummy was a strictly exogenous variable. It is possible to argue that this assumption is not valid in the inflation model; according to this line of argument, countries that choose a common-currency regime are countries with high inflation rates. In order to deal with this possible source of bias, we also estimated the inflation model using an instrumental variables (IV) version of the treatment approach. Under general conditions the outcome equation 6.1 can be rewritten as: y ¼ a þ gd þ xb0 þ dðx EðxÞÞb1 þ e0 þ dðe1 e0 Þ;
ð6:10 Þ
148
Sebastian Edwards and I. Igal Magendzo
where the subscript indicates the value of d and e0 and e1 are zero mean, normally distributed error terms, conditional on X and a set of instruments Z. We assume that the treatment depends on the instruments Z—that is, that the vector w in equation 6.3 contains variables that are excluded from the outcome equation and that are orthogonal to the error terms. Woodridge (2002) shows that a consistent estimator of g, even when d is endogenous, is obtained by a two-step procedure. In the first step we estimate the parameters in equation 6.3 using a probit regression of d on X and Z; from this estimation we then obtain the hazard for each observation. In the second step we estimate the parameters in 6.10 by ordinary least squares (OLS), including the dummy variable, the variables in X, the hazard function, and the interaction of the dummy variable with X EðXÞ, where we use the sample average of X as an estimate of the expected value. The hazard—estimated using the instruments Z—together with the interaction terms play the role of a control function that controls for possible selection bias.39 In the estimation of 6.1 0 for inflation, the following instruments were used: log of population, log of initial GDP, an independent dummy, an index of the degree of openness, the ‘‘tropics’’ variable, the regional dummies, the log of distance, the ‘‘border’’ variable, the ‘‘island’’ variable, and the average of inflation in the five years prior to our first data point. The results obtained from this IV treatment model confirm those presented in sections 6.3 and 6.4 above. More specifically, the point estimate of the dummy for strict dollarization is 12.9 and was significant. The dummy for the ICU was also significant and its point estimate was 18.8. The detailed results from these IV treatment estimates (not reported here due to space considerations) are available from us on request. 6.5.4
What is Really Behind these Results? The Role of the CFA and the ECCA
From a policy perspective—and in particular from a ‘‘lessons’’ point of view—an interesting question is whether a specific group of countries is behind the results reported in the preceding sections. We are particularly interested in understanding whether there is any pattern behind the results suggesting that, while strictly dollarized countries have grown at a rate similar to those countries with a currency of their own, ICU countries have grown significantly faster. In order to investigate this issue we analyzed the residuals from our regression analysis, and inquired as to the characteristics of our ICU data set. As may be seen in table 6.1, the ICU data set may be divided into three groups of countries: seven members of the East Caribbean Currency Area (ECCA), fifteen members from the CFA Franc Zone in Africa, and countries from different parts of the world. A detailed inspection of the raw data suggests that there are very sub-
Hard Currency Pegs and Economic Performance
149
stantial differences in terms of economic growth between the ECCA nations, on the one hand, and the rest of the ICU countries. Indeed, for the period under consideration, average yearly GDP per capita growth in the ECCA countries has been 3.16 percent. In the other ICU countries it has only been 0.79 percent. The comparison of medians yields a similar result, with the median growth for ECCA countries at 3.6 percent and that for the rest of the ICUs at 0.72 percent. This unconditional comparison suggests that the ECCA nations’ performance is behind our findings (table 6.6) that, with other things given, ICU nations grow at a faster rate than countries with a domestic currency. In order to investigate this, we estimated a separate treatment effects model for GDP per capita growth for ECCA and other ICU nations.40 The estimated coefficient for the ECCA commoncurrency dummy variable was 2.3, with a t-statistics of 4.46, confirming that ECCA nations have outperformed by a wide margin countries with a currency of their own. The results for the non-ECCA ICU countries were quite different, with a statistically insignificant estimated coefficient for the treatment dummy of 0.5.41 These results, then, confirm the notion that the driving force behind the apparent superior growth performance of common-currency countries, reported in the preceding sections, is fully driven by the group of seven ECCA nations. 6.6
Concluding Remarks
In the last few years a number of economists have argued that the emerging economies should give up their domestic currencies. Interestingly, there have been very few systematic comparative studies on the performance of countries that, indeed, do not have a currency of their own. Moreover, there has been no effort in the empirical literature to make a distinction between strictly dollarized countries and independent currency union countries. Most of the literature on the subject has, in fact, been based on case studies of Panama, or on indirect performance analyses of groups of countries strongly dominated by ICUs. This lack of empirical analysis has resulted in policy debates that, until now, have been based on conjecture and not on hard historical evidence. This has particularly been the case when the issue under discussion relates to the merits of strict dollarization. The purpose of this paper has been to analyze, from a comparative perspective, economic performance in economies that do not have a currency of their own. We have argued that the main difficulty in performing this type of comparison is defining the correct control group with which to compare the performance of the common-currency countries. We tackle this issue by using the treatment effects model developed in the labor economics literature; this method allows us to jointly estimate the probability of being a common-currency country and the effect of
150
Sebastian Edwards and I. Igal Magendzo
having this particular monetary regime on specific macroeconomic outcomes. Estimation using this technique yields results that are substantially different from those obtained from simple comparisons using a large control group of all withdomestic-currency countries. More specifically, we found that both ICU and dollarized countries have had a significantly lower rate of inflation than withcurrency ones. We found that macroeconomic volatility has been higher in both dollarized and ICU economies than in with-currency countries. We believe that our results are particularly interesting with respect to GDP growth. The estimations reported in table 6.6 suggest that while strictly dollarized countries have had a statistical rate of GDP per capita growth that is not different from that of countries with a currency of their own, ICU countries have grown faster than with-currency nations. When we investigated these findings further, we found that the East Caribbean Currency Area countries were the driving force behind this estimated superior performance. Indeed, once these seven countries were excluded from the sample, we found no statistical difference in GDP per capita growth between the rest of the ICU countries and countries with a currency of their own. The ECCA countries constitute a very special group: they are very small indeed, with an average population of fewer than 100,000 inhabitants. They are all islands geographically close to major markets. Their main industry is tourism, and they have very close economic and cultural ties with the United Kingdom. We believe that their experience with ICU may not be entirely useful for larger countries planning to reform their exchange rate and monetary regime. This does not mean, however, that other emerging nations would not benefit from giving up their currencies. Indeed, as Mundell (1961) argued four decades ago, it is perfectly possible that countries with certain characteristics will benefit from giving up their currency and either dollarizing or joining a currency union. In that regard the experiences of Ecuador and El Salvador—two recent dollarizers—will be very useful in assessing this question in the future. Naturally, the euro experience will provide a wealth of information on the consequences of common currencies in advanced nations. Acknowledgments This paper was prepared for presentation at a conference in honor of Guillermo Calvo, held at the International Monetary Fund April 15–16, 2004. In our work on common currencies we have benefited from discussions with John Cochrane, Barry Eichengreen, Eduardo Engel, Ed Leamer, Robert Barro, and Andy Rose. We thank our discussants Lars Svensson and Hang Genberg for many helpful comments. Marcela Aurelio provided wonderful assistance.
Hard Currency Pegs and Economic Performance
151
Notes 1. See Alesina, Barro, and Tenreyro (2002) for a discussion on the conditions under which emerging countries will benefit from giving up their currency. 2. Although this option is known as ‘‘dollarization,’’ the advanced country’s currency need not be the dollar. It could be any other convertible currency. 3. In early 2000 Florida’s then-senior senator Connie Mack introduced legislation into the U.S. Senate aimed at sharing seignorage with countries that decided to adopt the U.S. dollar as legal tender. The bill, however, did not move in the legislative process. 4. The euro is perhaps the best-known example of this type of currency union. 5. Strictly speaking there is a third type of currency union: when a small country adopts a nonconvertible currency from another country as legal tender. In this case the ‘‘credibility’’ effect of having monetary policy run by an advanced nation’s central bank will not be present. Thus, in the empirical analysis that follows we group this third category with what we have called independent currency unions. 6. On analytical aspects of dollarization see Calvo (1999) and Eichengreen and Haussman (1999); for an interesting theoretical analysis see Cooper and Kempf (2001). 7. Goldfjan and Olivares (2001) use econometrics to evaluate Panama’s experience with dollarization. Moreno-Villalaz (1999) provides a detailed analysis of the Panamanian system. Bogetic (2000) describes several aspects of dollarization in a number of countries. As far as we know, Rose and Engel (2000) and Edwards (2001a) are the first two papers to provide a statistical and econometric analysis of economic performance in dollarized countries and/or currency unions. See also the papers in Levy-Yeyati and Stuzenegger (2003). 8. See Klein (2002) for a discussion on dollarization and trade, including a comprehensive bibliography on the subject. 9. These countries have data for a long enough period for at least one of two variables: GDP per capita or inflation. In the rest of the paper we will use the term ‘‘countries’’ to refer both to independent countries and to territories. 10. Rose and Engel (2002), in contrast, consider both Ireland and Bermuda as common-currency countries. Ireland’s currency was the Irish pound, or ‘‘punt.’’ From September 1928 through March 1979 it was linked to the British pound at parity. Bermuda’s currency is the Bermuda dollar. The Bermuda Monetary Authority issues BD$ notes in several denominations (BD$ 2, 5, 10, 20, 50, and 100). The BD$ is linked to the U.S. dollar at parity. It should be noted that if these countries are added to the list of common currency countries in table 6.1, the results reported in this paper remain basically unaffected. 11. It is not easy to date unequivocally Liberia’s abandonment of the dollarized system. In July 1974 the National Bank of Liberia (NBL) was opened. In 1982 the NBL began issuing five-dollar coins, and in 1989 it began issuing five-dollar notes. On Liberia’s dollarization experience see Barret (1995) and Berkeley (1993). 12. The median population of all nondollarized emerging nations is over 100 times larger than that of the dollarized economies. 13. Our volatility measure is the standard deviation of growth, calculated over a five-year period. When alterative time frames were used (i.e., seven years) the results did not change in any significant way.
152
Sebastian Edwards and I. Igal Magendzo
14. More specifically, we excluded from the control group those observations with a rate of inflation in excess of 200 percent per year. This resulted in eighty observations being dropped from the control group of countries with a currency of their own. See section 6.5 for results under alternative definitions of ‘‘high inflation.’’ 15. When hyperinflation countries are not excluded the means difference in inflation is a staggering 62 percent. 16. See Maddala (1983). 17. Volatility of GDP growth is measured as the standard deviation of growth in subperiods of five years. The subperiods correspond to the years 1974–1978, 1979–1983, 1984–1988, 1989–1993 and 1994–1998. Accordingly, the covariates were averaged out for the same subperiods. The union, ICU, and dollarized dummies were set to one for a particular subperiod if the country belonged to the respective group for at least four of the five subperiod years. In particular, Equatorial Guinea belongs to the CFA union only since 1985, but it was assigned a value of 1 for the ICU (and the union) dummy for the subperiod 1984–1988. On the contrary, even though Liberia was dollarized till 1981, it was not considered a dollarized country for the subperiod 1979–1983 (neither was it included in the control group). On the other hand, the West African Currency Union was dissolved in 1980, so Kenya, Tanzania, and Uganda are not considered part of an ICU (nor as part of the control group) for the subperiod 1979– 1983. 18. It is assumed, however, that djt does not depend on yjt . Otherwise, as discussed later, the model cannot be identified. 19. The two-steps estimates yield similar results, and are available from the authors on request. 20. Details on identification and consistency of models with mixed structures can be found in Maddala (1983). See also Heckman (1978), Angrist (2001), and Wooldridge (2002). 21. Wooldridge (2002). 22. Remember that the volatility variable is measured as the standard deviation of growth over a fiveyear period. When alternative seven-year periods were used, there were no significant changes in the results. 23. See Alesina, Barro, and Tenreyro (2002) for a recent discussion on the subject. 24. Unfortunately, there are no available data on other regional variables of interest, including a generalized index of factor mobility or of synchronicity of shocks, for all countries in our sample. 25. See the original Sachs-Werner (1995) article for a specific list of requirements for a country to qualify as ‘‘open.’’ 26. The data are available from the authors on request. 27. These variables are expected to capture the effect of some variables for which there were no data. 28. See Leamer (1997) and Venables (2002) for discussions on the role of geography and distance on economic growth. 29. In addition to the variables in tables 6.3–6.6, we considered additional covariates. In particular, we constructed an index on whether the country in question was a member of a ‘‘deep’’ trading area. This index, however, identifies almost fully the common-currency countries, reducing the spirit of the analysis. 30. An important question is whether these results—as well as those in tables 6.4–6.5—are subject to an ‘‘omitted variables’’ bias, stemming from the fact that there are no data on some of the traditional
Hard Currency Pegs and Economic Performance
153
regressors for the common-currency countries. We address this issue in some detail in section 6.4 of this paper. 31. See Frenkel and Razin (1987). 32. In doing this we excluded the other common-currency group from the with-domestic-currency sample. That is, the with-domestic-currency category is strictly made of countries that do have a currency of their own. 33. If we estimate the equation above using all nontreated observations, the selection bias is given by: BðxÞ ¼ Eðu0 =x; D ¼ 1Þ Eðu0 =x; D ¼ 0Þ: 34. In order to guarantee that all treated agents have such a counterpart in the population (not necessarily in the sample), we also need to assume that 0 < ProbðD ¼ 1=xÞ < 1: 35. This averaged version is given by Ð Eð y1 y0 =x; D ¼ 1Þ dFðx=D ¼ 1Þ Ð ; MðSÞ ¼ S S dFðx=D ¼ 1Þ where S is a subset of the support of x given D ¼ 1. 36. This assumes that there are no omitted variables problems in the estimation of the propensity used to select the nearest neighbors. 37. See Lechner (2002) for a discussion on matching methods with multiple treatments. The results in table 6.9 correspond to the case when ‘‘no replacement’’ is allowed in selecting the nearest neighbors in the control group. See Edwards and Magendzo (2002, 2006) for details. 38. This assumption is sometimes referred as the ‘‘ignorability-of-treatment assumption.’’ (Wooldridge 2002). 39. For details on these procedures see Woodridge (2002). 40. In the estimation we had to respecify the treatment equation. The reason is that some of the regressors (islands, for example) fully predicted the probability of being a currency union country. 41. These results are from a specification that includes regional dummies in the GDP growth outcome equation. If these dummies are excluded the ICU dummy becomes negative with a t-statistic of 1.46.
References Alesina, A., and R. J. Barro. 2000a. ‘‘Currency Unions.’’ Working Paper No. 7927, NBER, Cambridge, MA. ———. 2000b. ‘‘One Country, One Currency?’’ In Currency Unions, eds. A. Alesina and R. J. Barro, 11– 19. Stanford, CA: Hoover Institution Press. ———. 2001. ‘‘Dollarization.’’ American Economic Review 91, no. 2: 381–385. Alesina, A., R. J. Barro, and S. Tenreyro. 2002. ‘‘Optimal Currency Areas.’’ Working Paper No. 9072, NBER, Cambridge, MA. Angrist, J. 2001. ‘‘Estimation of Limited Dependent Variable Models with Dummy Endogenous Regressors: Simple Strategies for Empirical Practice.’’ Journal of Business and Economic Statistics 19, no. 1: 2–16.
154
Sebastian Edwards and I. Igal Magendzo
Barret, L. 1995. ‘‘The Economic Trials of Liberia.’’ West Africa, March 27–April 2, 461–463. Barro, R. J. 1996. ‘‘Determinants of Economic Growth: A Cross-Country Empirical Study.’’ Working Paper No. 5698, NBER, Cambridge, MA. Barro, R. J., and Lee, J. 1996. ‘‘International Data on Educational Attainment: Updates and Implications.’’ Working Paper No. 42, CID, Harvard University, Cambridge, MA. Barro, R. J., and X. Sala-i-Martin. 1995. Economic Growth. New York: McGraw Hill. Berkeley, B. 1992. ‘‘Liberia’s Warring Currencies.’’ Institutional Investors, September 1992. Blundell, R., A, and Costa Dias, M. 2000. ‘‘Evaluation Methods for Non-Experimental Data.’’ Fiscal Studies 21: 427–468. Bogetic, Z. 2000. ‘‘Official Dollarization: Current Experiences and Issues.’’ Cato Journal 20, no. 2: 179– 213. Calvo, Guillermo A. 1999. ‘‘Fixed versus Flexible Exchange Rates.’’ Mimeo., University of Maryland. ———. 2000. ‘‘Balance of Payments Crises in Emerging Markets: Large Capital Inflows and Sovereign Governments.’’ In Currency Crises, ed. Paul Krugman, 71–98. Chicago: University of Chicago Press. ———. 2001. ‘‘Economic Policy in Stormy Waters: Financial Vulnerability in Emerging Economies.’’ Journal of Applied Economics 4, no. 1: 1–25. Calvo, Guillermo A., Alejandro Izquierdo, and Ernesto Talvi. 2003. ‘‘Sudden Stops, the Real Exchange Rate, and Fiscal Sustainability: Argentina’s Lessons.’’ Working Paper No. 9828, NBER, Cambridge, MA. Calvo, Guillermo A., and Carmen M. Reinhart. 2002. ‘‘Fear of Floating.’’ Quarterly Journal of Economics 117, no. 2: 379–408. Cooper, R. W., and H. Kempf. 2001. ‘‘Dollarization and the Conquest of Hyperinflation in Divided Societies.’’ Federal Reserve Bank of Minneapolis Quarterly Review 25, no. 3: 3–12. Corden, W. M. 2002. Too Sensational: On the Choice of Exchange Rate Regimes. Cambridge, MA: MIT Press. Dornbusch, R. 2001. ‘‘Fewer Monies, Better Monies.’’ American Economic Review 91, no. 2: 238–242. Edwards, S. 2001a. ‘‘Dollarization: Myths and Realities.’’ Journal of Policy Modeling 23, no. 3: 249–265. ———. 2001b. ‘‘Exchange Rate Regimes, Capital Flows, and Crisis Prevention.’’ In Economic and Financial Crises in Emerging Market Economies, ed. Martin, Feldstein, 31–78. Chicago: University of Chicago Press. Edwards, S., and I. I. Magendzo. 2002. ‘‘Dollarization, Inflation and Growth.’’ Working Paper No. 8671, NBER, Cambridge, MA. Edwards, S., and I. I. Magendzo. 2006. ‘‘Strict Dollarization and Economic Performance.’’ Journal of Money, Credit and Banking 38, no. 1: 269–282. Eichengreen, B. 2001. ‘‘What Problems Can Dollarization Solve?’’ Journal of Policy Modeling 23, no. 3: 267–277. Eichengreen, B., and R. Haussmann. 1999. ‘‘Exchange Rates and Financial Fragility.’’ Paper presented at the IDB-OECD Forum, November. Fischer, S. 1976. ‘‘Stability and Exchange Rate Systems in a Monetarist Model of the Balance of Payments.’’ In The Political Economy of Monetary Reform, ed. Robert Z. Aliber, 59–73. New York: McMillan.
Hard Currency Pegs and Economic Performance
155
———. 2001. ‘‘Exchange Rate Regimes: Is the Bipolar View Correct?’’ Journal of Economic Perspectives 15, no. 2: 3–24. Frankel, J. A. 1999. ‘‘No Single Currency Regime is Right for All Countries at at All Times.’’ Working Paper No. 7338, NBER, Cambridge, MA. Frankel, J. A., and A. K. Rose. 2002. ‘‘An Estimate of the Effect of Common Currencies on Trade and Income.’’ Quarterly Journal of Economics 117, no. 2: 437–466. Frenkel, J. A., and A. Razin. 1987. Fiscal Policies and Growth in the World Economy. Cambridge, MA: MIT Press. Frieden, J. 2003. ‘‘The Political Economy of Dollarization.’’ In Dollarization, eds. E. Levy-Yeyati and F. Sturzenegger, 305–333. Cambridge, MA: MIT Press. Ghosh, A., A. Gulde, J. Ostry, and H. Wolf. 1995. ‘‘Does the Nominal Exchange Rate Regime Matter?’’ Working Paper No. 95/121, IMF, Washington, D.C. Goldfajn, I., and G. Olivares. 2000. ‘‘Is Adopting Full Dollarization the Solution? Looking at the Evidence.’’ Working Paper No. 416, Pontificia Universidade Cato´lica do Rio de Janeiro. Greene, W. H. 2000. Econometric Analysis. New York: Macmillan. Heckman, J. 1978. ‘‘Dummy Endogenous Variables in a Simultaneous Equation System.’’ Econometrica 46, no. 4: 931–959. Heckman, J. J., I. Hidehiko, and P. E. Todd. 1997. ‘‘Matching as an Econometric Evaluation Estimator: Evidence from Evaluating a Job Training Programme.’’ Review of Economic Studies 64, no. 4: 605–654. IMF. 1997. ‘‘Exchange Rate Arrangements and Economic Performance in Developing Countries.’’ In World Economic Outlook. Washington, D.C.: IMF. Klein, M. W. 2002. ‘‘Dollarization and Trade.’’ Working Paper No. 8879, NBER, Cambridge, MA. Leamer, E. E. 1997. ‘‘Access to Western Markets and Eastern Effort.’’ In Lessons from the Economic Transition: Central and Eastern Europe in the 1990s, ed. Salvatore Zecchini, 503–526. Dordrecht: Kluwer Academic Publishers. Lechner, M. 2002. ‘‘Program Heterogeneity and Propensity Score Matching: An Application to the Evaluation of Active Labor Market Policies.’’ Review of Economics and Statistics 84, no. 2: 205–220. Levy-Yeyeti, E., and F. Sturzenegger. 2001. ‘‘To Float or to Trail: Evidence on the Impact of Exchange Rate Regimes.’’ Working Paper No. 1/01, Universidad Torcuato di Tella. Levy-Yeyeti, E., and F. Sturzenegger. 2003. Dollarization. Cambridge, MA: MIT Press. Maddala, G. S. 1983. Limited-Dependant and Qualitative Variables in Econometrics. New York: Cambridge University Press. Meade, J. E. 1950. The Balance of Payments. New York: Oxford University Press. Moreno-Villalaz, J. L. 1999. ‘‘Lessons from the Monetary Experience of Panama: A Dollar Economy with Financial Integration.’’ Cato Journal 18, no. 3: 421–439. Mundell, R. A. 1961. ‘‘A Theory of Optimum Currency Areas.’’ American Economic Review 51, no. 4: 657–665. Parrado, E., and A. Velasco. 2002. ‘‘Optimal Interest Rate Policy in a Small Open Economy.’’ Working Paper No. 8721, NBER, Cambridge, MA.
156
Sebastian Edwards and I. Igal Magendzo
Powell, A., and F. Sturzenegger. 2000. ‘‘Dollarization: The Link Between Devaluation and Default Risk.’’ Working Paper, Universidad Torcuato di Tella, Buenos Aires, Argentina. Rose, A. 2000. ‘‘One Money, One Market: Estimating The Effect of Common Currencies on Trade.’’ Economic Policy 15, no. 30: 7–46. Rose, A. K., and C. Engel. 2000. ‘‘Currency Unions and International Integration.’’ Working Paper No. 7872, NBER, Cambridge, MA. ———. 2002. ‘‘Currency Unions and International Integration.’’ Journal of Money, Credit, and Banking 34, no. 4: 1067–1089. Rose, A. K., and E. Van Wincoop. 2001. ‘‘National Money as a Barrier to International Trade: The Real Case for Currency Union.’’ American Economic Review 91, no. 2: 386–390. Rosenbaum, P. R., and D. B. Rubin. 1983. ‘‘The Central Role of the Propensity Score in Observational Studies for Causal Effects.’’ Biometrika 70, no. 1: 41–55. Sachs, J., and A. Warner. 1995. ‘‘Economic Reform and the Process of Global Integration.’’ Brookings Papers on Economic Activity 1995, no. 1: 1–188. Sachs, J. D. 2000. ‘‘Globalization and Patterns of Economic Development.’’ Weltwirtschaftliches Archiv/ Review of World Economics 136, no. 4: 579–600. Summers, L. H. 2000. ‘‘International Financial Crises: Causes, Prevention, and Cures.’’ American Economic Review 90, no. 2: 1–16. Venables, A. J., H. Overman, and S. Redding. 2002. ‘‘The Economic Geography of Trade, Production and Income.’’ Working Paper No. 0508, CEPR, London. Wooldridge, J. M. 2002. Econometric Analysis of Cross-Section and Panel Data. Cambridge, MA: MIT Press.
III
Financial Crises
7
Asset Prices and Self-Fulfilling Macroeconomic Pessimism Andre´s Velasco and Alejandro Neut
7.1
Introduction
The volatility in asset prices (especially stock prices) over the last few years, both in the United States and elsewhere, has caused economists to wonder about the effects of movements in these prices on aggregate demand and economic activity. One standard link runs from stock prices to private wealth, and then to consumption demand and income. Feeling suddenly poor after a bear market, consumers can curtail their spending, pushing the economy even further down. But several econometric studies have looked at the link between financial assets and consumption only to conclude that the elasticities involved are relatively small.1 As long as fundamentals are in place, volatility in asset prices will have a manageable effect on output. Another possible effect runs from asset prices to investment demand. As Guillermo Calvo famously emphasized (and foretold) in 1994, a sudden loss of investor confidence can reduce the value of assets and trigger pernicious dynamics that drive a small open economy like Mexico into recession. This line of reasoning is not applicable only to small open economies. In a closed economy the link between asset prices and investment is also present. And, as this paper shows, it may generate multiple equilibria. A bear market reduces the value of the collateral held by households and firms, which in turn cuts their ability to borrow and invest, again pushing output down. But that need not be the end of the story. Changes in activity affect future returns, which in turn affect current stock prices. Circular causation can occur, which inevitably raises the question of whether movements in asset prices and economic activity are based on self-fulfilling beliefs. Can it be the case that any event (a terrorist attack? a political crisis? a poor night’s sleep?) causes a bout of depression and a fall in the stock market, which in turn triggers a fall in investment and a recession, which justifies the initial pessimism? This paper provides an extremely simple model that delivers such a result.
Andre´s Velasco and Alejandro Neut
160
Multiple equilibria under rational expectations can arise in very different settings. In Calvo’s seminal paper (1988), the denomination of government bonds was at the root of the problem. In this case the difficulty lies with a particular financial market imperfection: the need to collateralize borrowing. This difference has policy implications. In Calvo (1988), interest rate caps could rule out the bad equilibrium. Here, interest rate caps or subsidies may help improve the good equilibrium but do nothing to eliminate the bad outcome. By contrast, policies that increase aggregate demand and output may succeed in ruling out the undesirable equilibrium. 7.2
The Model
The model has two periods, one good, and two kinds of people, capitalists and workers. Workers supply labor and consume. Capitalists own the factors of production other than labor, which they rent to firms and also consume. They finance investment in excess of their own resources by borrowing from workers. The key aspect of the model is that such borrowing is constrained by the need to put up collateral, and the value of this collateral in turn depends on asset prices. 7.2.1
Domestic Production
Production is carried out by competitive firms. Each firm has access to the CobbDouglas technology Yt ¼ Kta Ltg Nt1ag ;
0
ð7:1Þ
where K denotes capital, L land and N labor. Without loss of generality, we assume capital depreciates completely in one period. Land L is fixed, and we normalize Lt ¼ 1 from now on. In each one of the two periods cost-minimizing firms choose capital, land, and labor according to Yt RtK ¼ a ð7:2Þ Kt Yt L Rt ¼ g ð7:3Þ Lt Yt Wt ¼ ð1 a gÞ ð7:4Þ Nt
Asset Prices and Self-Fulfilling Macroeconomic Pessimism
161
where RtK , RtL , and Wt are the factor returns to capital, land, and labor. The nume´raire in this economy is the price of output Y, which is used for consumption and investment in capital for the following period. 7.2.2
Workers
Workers consume and they supply one unit of labor inelastically (so that N ¼ 1 by assumption from now on), for which they are paid labor’s marginal return. As consumers, they maximize a standard two-period utility function subject to the budget constraint C1 þ
C2 W2 ¼ D þ W1 þ ; 1þr 1þr
ð7:5Þ
where D is accumulated wealth saved by workers (and consequently borrowed by capitalists) in the past and r is the market rate of interest for bond transactions between these two groups. Period 1 savings by workers are also channeled to capitalists, who invest in either land or capital. Using equations 7.1 and 7.4, budget constraint 7.5 can be written as C1 þ
C2 ð1 a gÞK2a ¼ D þ ð1 a gÞK1a þ ; 1þr 1þr
ð7:6Þ
where K2 is obtained through investment in capital in the first period. Capital K1 and D (assets for the workers, debts for the capitalists) are given by history. The solution to the maximization problem faced by workers boils down to the savings function S1 ¼ f ½r; ð1 a gÞK1a þ B; ð1 a gÞK2a :
ð7:7Þ
From now on, assume the ‘‘normal’’ case in which f is increasing in r.2 7.2.3
Capitalists
Capitalists are the key players in the model: they finance spending partly with loans, and their borrowing is subject to frictions. They consume in the closing period only. Their objective is to maximize the utility from such consumption, which boils down to maximizing the amount consumed. They only own land and capital, but the latter depreciates completely once used. Qt denotes the market value of land in period t (after returns are paid). In this simple two-period model, Q1 is endogenous and Q2 ¼ 0.
Andre´s Velasco and Alejandro Neut
162
At the beginning of each period, capitalists collect the income from capital and land and repay debt (to workers). In the first period, their net resources available for investment are N1 1 R1K K1 þ R1L L D ¼ ða þ gÞY1 D
ð7:8Þ
where the second equality comes from 7.2 and 7.3. Although individual capitalists have additional resources in the form of land, the value of land Q1 does not enter equation 7.8. The reason is that land is only traded among capitalists; hence there is no additional aggregate net value that can be diverted to acquire capital. Notice that N1 is exogenous because D is given and so is Y1 : aggregate land and capital are inherited, and labor is inelastically supplied. The capital stock available in the second period is K2 , equal to investment in period 1.3 Capitalists can invest in additional capital subject to the budget constraint K2 ¼ N1 þ B1
ð7:9Þ
where B1 is the amount borrowed by capitalists in period 1. A crucial assumption is that, because of limitations in contract enforcement and the like, in the case of nonpayment lenders can seize at most the value and returns to land, net of enforcement costs, equal to L. Hence workers will not lend at the initial time an amount generating obligations larger than the value of collateral:4,5 B1 a maxf0; Q1 Lg: 7.2.4
ð7:10Þ
Market Clearing
Capitalists can only borrow from workers, so that B1 ¼ S1 :
ð7:11Þ
Equations 7.7, 7.9, and 7.11 together yield K2 ða þ gÞK1a þ A ¼ f ½r; ð1 a gÞK1a þ D; ð1 a gÞðK1 þ K2 Þ a :
ð7:12Þ
This is an implicit expression for r as a function of K2 and a pair of exogenous variables r ¼ fðK2 ; K1 ; DÞ:
ð7:13Þ
It is easy to show that, under the assumptions made on f , f is increasing in K2 and D. The partial relationship between the interest rate and K1 is ambiguous.
Asset Prices and Self-Fulfilling Macroeconomic Pessimism
7.2.5
163
Equilibria
Next we define three schedules that jointly determine the possible equilibria of the model. 1. LK schedule. Equating the marginal returns of land and capital implies R2L ¼ R2K : Q1 Using 7.2 and 7.3 in 7.14 one obtains g K2 : Q1 ¼ a
ð7:14Þ
ð7:15Þ
This schedule, which we term LK (for no arbitrage between land and capital), gives the equilibrium price of land today as a function of investment in capital today. As K2 rises, tomorrow’s capital-land ratio goes up, increasing the marginal product of land and hence raising Q1 . 2. KB schedule. If capitalists are not financially constrained and can borrow as much as they want, they maximize their next-period consumption by choosing an amount of capital investment such that R2K ¼ 1 þ r:
ð7:16Þ
Using 7.2 and 7.13, this equation implies K2 ½1 þ fðK1 ; K2 ; DÞ 1=ð1aÞ ¼ a 1=ð1aÞ
ð7:17Þ
where the LHS is unambiguously increasing in K2 . We denote this schedule KB because it corresponds to no-arbitrage between capital and bonds. Let K2 be the level of capital that solves 7.17. We assume K2 > N1 , so that capitalists will indeed want to borrow. But financial constraints may not let capitalists borrow the resources to invest as much capital as they want, in which case capital-bonds arbitrage does not obtain. In those constrained situations the amount of investment is bounded from above as follows: K2 ½1 þ fðK1 ; K2 ; DÞ 1=ð1aÞ a a 1=ð1aÞ :
ð7:18Þ
3. FC schedule. The fact that borrowing is constrained can lead investment to be constrained. Combining 7.10 and 7.9 one has K2 a maxfN1 ; Q1 L þ N1 g:
ð7:19Þ
Andre´s Velasco and Alejandro Neut
164
Rearrange 7.19 to read Q1 b K 2 þ L N1 ;
if K2 b N1 :
ð7:20Þ
This inequality shows that for every level of planned investment K2 , the price of land Q1 must be sufficiently high for that investment to be feasible. We term this the FC (financial constraint) schedule.6 7.3
Outcomes with and without Crises
The model can be solved quite simply using a diagrammatic representation in Q1 , K2 space. Notice that LK always holds because there are no rigidities or constraints in arbitraging between land and capital. Therefore, to solve for K2 and Q1 one needs one additional equation. There are two candidate inequalities, depicted by schedules KB and FC. At least one of these inequalities must hold with equality. The reason is straightforward: if KB is not binding, the return to capital is greater than r. For this to be the case, capitalists must be financially constrained—in other words, the FC schedule is binding. On the other hand, if FC is not binding, this means that capitalists are financially unconstrained, and therefore KB holds with equality. We therefore have the following possible cases: Case where KB is binding: This is a situation where LK and KB together yield the level of investment K2 that capitalists would like to undertake if unconstrained (meaning as long as the inequality in schedule FC still holds). We show this case in figures 7.1 and 7.2.
•
Case where FC is binding: In this case firms do not have enough collateral to obtain any additional credit. Schedules LK and FC determine an equilibrium as long as capitalists are willing
•
Figure 7.1
Asset Prices and Self-Fulfilling Macroeconomic Pessimism
165
Figure 7.2
Figure 7.3
to invest more under those circumstances (that is, as long as the inequality in KB holds). We show this case in figures 7.2 and 7.3. Figures 7.1 and 7.3 involve a unique equilibrium, at point A in each case. Figure 7.2 involves multiple equilibria at points A and B. The bad equilibrium implies no borrowing (K2 ¼ N1 ) and lower asset prices than does the good equilibrium. A key exogenous variable is N1 . The lower N1 , the farther to the left is FC, possibly taking the economy from the one-equilibrium case in figure 7.1 to the twoequilibria case in figure 7.2. A necessary and sufficient condition for a bad equilibrium with no borrowing to exist (whether unique as in figure 7.3, or non-unique as in figure 7.2) is that the FC schedule be above the LK schedule at K2 ¼ N1 . That is to say Lb
g N1 : a
ð7:21Þ
Andre´s Velasco and Alejandro Neut
166
Hence, if enforcement costs L are sufficiently small, the no-borrowing bad equilibrium disappears. Notice that this condition need not be too restrictive, since given the definition of net worth (N1 ¼ ða þ gÞY1 D), it could be near zero or even negative if inherited debts are large. The intuition of multiple equilibria is simple: there is feedback between asset prices and aggregate demand. A higher land price allows the capitalists to borrow and invest more in capital. But at the same time, higher capital investment raises the marginal product of land and therefore its price. For some parameter values, the circularity opens the door for multiplicity and self-fulfiling beliefs. 7.4
Policy Alternatives
This model is too simple to say much about policy, but a few points are suggestive. Focus on the case of multiple equilibria only, and look for ways to eliminate the bad outcome. Expansionary Fiscal and Monetary Policies To the extent they can raise current income Y1 and consequently raise N1 , these policies can move FC to the right until condition 7.21 no longer holds.7 A sufficiently large expansion will cause the FC curve to be below the LK schedule at K2 ¼ N1 , eliminating the bad equilibrium in figure 7.5. The intuition is that with larger current output the capitalists’ gross resources available for investment rise relative to debt due, so they can afford to invest more for every price of land. In addition, higher Y1 raises asset prices for any level of investment. The combination can rule out self-fulfilling pessimistic animal spirits. Debt Forgiveness or Rescheduling A reduction in D (debt could be written down, paid by the government, or involuntarily reprogrammed to period 2) also increases N1 , moving the vertical portion of the FC schedule to the right while leaving LK unchanged. A sufficiently large cut in the current debt burden would rule out the bad equilibrium by causing FC to be below LK at K2 ¼ N1 . This case also corresponds to figure 7.4, with a unique equilibrium at point A. The intuition is the same as in the previous case: with improved cash flow, the capitalists can afford to invest even if asset prices are low. Financial and Legal Reform A reduction in L brings about a downward movement in FC, as in figure 7.5. For a sufficiently small L, the bad equilibrium disappears. The single equilibrium can
Asset Prices and Self-Fulfilling Macroeconomic Pessimism
167
Figure 7.4
Figure 7.5
be either unconstrained (as drawn) or constrained. In the new single equilibrium, capitalists can borrow and invest as much as they would like. There are other policies that may have advantages even though they cannot eliminate a bad equilibrium when one exists. Consider, for instance, subsidies to savings. To the extent such a policy lowers the value of r ¼ fðK1 ; K2 ; DÞ for any level of K2 , it shifts up the KB curve (see figure 7.6). But FC is unchanged, so there is no interest rate that can free the economy from the possibility of a bad equilibrium. An alternative way to see this is to note that r does not enter condition 7.21, so changes in interest rates alone cannot affect the number of feasible equilibria. The intuition is that a bad equilibrium is the consequence of asset prices that are so low capitalists end up financially constrained when they do not borrow anything. In that situation, the present value of their collateral is negative.8 Therefore, no matter how low interest rates go, the present value of that collateral remains negative and the bad equilibrium does not disappear.
Andre´s Velasco and Alejandro Neut
168
Figure 7.6
7.5
Conclusions
The model in this chapter involves perfect competition and well-functioning markets in all respects but one: there is an imperfection in the capital market that causes borrowing to be collateralized. Since the value of the collateral depends on current asset prices, and asset prices depend on current expenditure and future profits, this opens the door to self-fulfilling expectations. But in certain conditions, policies can rule out the bad outcome, if one exists. The mechanism in the paper is quite specific. But the result that small deviations from the perfect financial markets paradigm can generate more than one equilibrium is not. Other asset prices—for instance, the exchange rate in an open economy—can play the same role. For an example see Velasco (2001). Notes An earlier version of this paper was circulated under the title ‘‘The Macroeconomics of Terrorism.’’ Generous financial support from the National Science Foundation and the Harvard Center for International Development is acknowledged with thanks. We are grateful to the members of LIEP at Harvard University for useful comments. 1. Housing wealth has a significantly larger effect on consumption as estimated by Case, Quigley, and Shiller (2001). 2. That is, assume the substitution and wealth effects of movements in interest rates are greater than the income effect. 3. Recall capital depreciates completely. Introducing a depreciation rate for capital less than one would change nothing, but would make the algebra more cumbersome. 4. This is as in Kiyotaki and Moore (1997), Krugman (1999), and Aghion, Bachetta, and Banerjee (2001), among many others.
Asset Prices and Self-Fulfilling Macroeconomic Pessimism
169
5. A possible objection is that the collateral constraint should depend on period 2 variables—that is to say, on the ‘‘stuff’’ lenders can seize in the event of nonpayment. The same results, but with somewhat more cumbersome algebra, would obtain if we adopted that specification. For instance, the case in which lenders can seize the total returns earned by the capitalist in period 2. The constraint would be ð1 þ rÞB1 a maxf0; ða þ gÞY2 Lg: See note 6 for details. 6. If instead we had assumed the constraint ð1 þ rÞB1 a maxf0; ða þ gÞY2 Lg; the FC schedule would be Q1 b u½ðK2 N1 Þð1 þ fðK2 ; K1 ; BÞÞ þ L 1=a
if K2 > N1 ;
where u is a positive constant and the RHS of the inequality is unambiguously increasing in K2 . The only difference would be that the FC is no longer linear in K2 . 7. Of course, extending the model to allow for such Keynesian effects would require labor supply to be endogenous and prices to be sticky. Both can be added without much complication. 8. This is the case for any ‘‘bad’’ equilibrium; for instance, the equilibrium in figure 7.3.
References Aghion, P., P. Bacchetta, and A. Banerjee. 2001. ‘‘Currency Crises and Monetary Policy in an Economy with Credit Constraints.’’ European Economic Review 45, no. 7: 1121–1150. Calvo, G. 1988. ‘‘Servicing the Public Debt: The Role of Expectations.’’ American Economic Review 78: 647–661. Calvo, G. 1994. ‘‘Comments on Dornbusch-Werner.’’ Brookings Papers on Economic Activity, no. 2: 299– 303. Case, K., J. Quigley, and R. Shiller. 2001. ‘‘Comparing Wealth Effects: The Stock Market versus the Housing Market.’’ Working Paper No. 8606, NBER, Cambridge, MA. Kiyotaki, N., and J. Moore. 1997. ‘‘Credit Cycles.’’ Journal of Political Economy 105, no. 2: 211–248. Krugman, P. 1999. ‘‘Balance Sheets, the Transfer Problem and Financial Crises.’’ In International Finance and Financial Crises: Essays in Honor of Robert Flood, eds. P. Isard, A. Razin, and A. Rose, 33–44. Norwell, MA: Kluwer Academic Publishers. Velasco, A. 2001. ‘‘The Impossible Duo? Globalization and Monetary Independence in Emerging Markets.’’ Brookings Trade Forum 2001: 69–99.
8
The Center and the Periphery: The Globalization of Financial Turmoil Graciela L. Kaminsky and Carmen Reinhart
The first springs of great events, like those of great rivers, are often mean and little. —Jonathan Swift
8.1
Introduction
A succession of crises in emerging-market economies during the 1990s ignited a debate in academic and policy-making circles about the transmission of shocks across national borders. The spreading market strain surrounding the Mexican peso crisis of 1994, the Asian credit crunch of 1997, and the Russian devaluation and Long-Term Capital Management’s (LTCM) implosion, both in 1998, have spawned a body of work that can be summarized under the heading ‘‘contagion.’’ These episodes have also resuscitated the interest in codes and standards, monetary arrangements, the role of international institutions, and securities law— summarized as the international financial architecture—to construct a barrier to prevent contagion. The academic literature on contagion—less pejoratively described as spillovers—and the international propagation of shocks has progressed along two roads in recent years. Early studies attempted to document the existence of contagion. More recent papers have primarily sought to discriminate among the possible transmission channels of disturbances—that is, whether shocks propagate through channels established by trade patterns, geography, commonalities among lenders, or from other sources. In our view, much of this literature suffers from three serious drawbacks. First, most studies have not discriminated between the origins of shocks. One expects, a priori, that the global or regional consequences of a disturbance may depend importantly on whether the shock—to borrow terms from Sir Arthur Lewis—originates in the periphery or in the center (Lewis 1977). Were the
172
Graciela L. Kaminsky and Carmen Reinhart
regional consequences of the Thai crisis so severe because of Thailand’s direct links with other countries in the region or because that shock affected the region’s largest economy—Japan? Were the paralysis of the bond markets in many parts of the globe and the persistent equity market volatility due to the Russian default or to concerns that LTCM’s reach was wider than understood, and that other firms in other financial centers of the world shared similar failings? And what about the dog that did not bark: why did Ecuador’s 1999 default not have greater international consequences? In contrast, this paper attempts to capture the origins of systemic turbulences and measure the direct and indirect linkages between national markets. Second, terms like contagion and spillover can be quite slippery. Some authors seek to learn about potential linkages by examining correlation patterns across markets using long time series. Others focus on market behavior during specific episodes dated a priori from other sources. Our approach is data driven. We define financial turmoil in terms of the behavior of financial prices and let the data determine when there were episodes of spillover. In particular, we define days of turmoil as days with extreme returns.1 Using information on the daily behavior of stock-market price indexes for thirty-five emerging-to-mature market economies from January 1997 to August 1999, we examine what happens in stock markets around the world on days of turmoil in financial centers (Germany, Japan, and the United States) and in crisis-prone emerging economies (Brazil, Russia, and Thailand). Third, most of the studies of financial spillovers rely on an examination of contemporaneous and lagged correlations. But correlations alone cannot tell those systemic turbulences due to a common shock (say, a large change in oil prices or the announcement of election results in an important country) from true spillover (a change in one national market directly related to extreme price movements in another market). To learn about the determinants of systemic financial turmoil, we turn to newspapers and summarize the key world events associated with significant price changes. In many cases, this allows us to identify the source of the shock—the center or the periphery—and to understand better the temporal dimension of the financial market adjustment. To be more specific about spillovers, we borrow from the literature on financial market efficiency to distinguish between ‘‘weak-form globalization’’ and ‘‘strongform globalization’’ of turmoil. Weak-form globalization occurs when country j experiences anomalous returns in days of extreme returns in country i, where anomalous behavior is interpreted as a change in the distribution of returns assessed by a nonparametric procedure. This definition does not require the countries suffering the spillovers to have extreme returns (that is, to be in the 5th and 95th percentile). Strong-form globalization occurs when country j experiences tur-
The Globalization of Financial Turmoil
173
moil when country i has extreme returns in the stock market. That is, it is a statement about simultaneity of extreme returns. Using these definitions, we construct two indices of globalization and examine the patterns of spillover among crisisprone emerging markets and financial centers. While the analysis of more episodes is clearly necessary, one preliminary conclusion we draw is that financial shocks often traverse a circuitous route. Problems occur synchronously in many emerging markets on the periphery because a shock in one of them first influenced a financial center. If the shock never reaches the center, it is doubtful it can become systemic, irrespective of the definition of systemic that is used. For example, in the case of the Asian crisis, Japanese bank exposure to Thailand—and their subsequent retrenchment from lending to other Asian countries—played a prominent role in the spread of the crisis. The role played by the center ( Japan) in this episode was much the same as that played by U.S. banks in the 1980s during the Latin American debt crisis. Similarly in 1998, Russia’s default triggered a pervasive widening of spreads that hobbled the weakened LTCM and led to a generalized withdrawal of risk taking. Thus, financial centers serve a key role in propagating financial turmoil. When financial centers remain safe, problems in an emerging market stop at the region’s border. The rest of the paper is organized as follows: section 2 presents a brief discussion of some of the analytical issues relevant to our analysis of the globalization of financial turmoil. Section 3 constructs the two indices of globalization of financial turmoil and examines the pattern of spillover within and across regions. Section 4 discusses the origins of high degree of contagion. Concluding remarks are presented in section 5. 8.2
Analytical Issues
For the purposes of our analysis, we divide the world into center and periphery countries. The former consist of the countries that house the largest financial centers (such as New York, London, Berlin, and Tokyo) while the latter comprise everyone else. We distinguish among three patterns in the propagation of shocks. First, there is the transmission of shocks from one periphery country to another periphery country, which can take place if the two countries are directly linked through bilateral trade or finance (figure 8.1). Recent examples of this type of transmission mechanism include the adverse impact of the 1997–1998 Asian crisis on Chilean exports and the contractionary impact on Argentina of the Brazilian devaluation in January 1999. This transmission channel may also be operative if there are bilateral finance links. For instance, Costa Rican banks were borrowing from Mexican banks on the eve of the Mexican crisis (see Calvo and Reinhart 1996), but when Mexican banks ran into trouble this source of funds disappeared.
174
Graciela L. Kaminsky and Carmen Reinhart
Figure 8.1 The Transmission of Shocks from One Periphery Country to Another
Figure 8.2 The Transmission of Shocks from one Periphery Country to Another Through a Center Country
Second, there is the transmission of shocks from one periphery country to another via a center country (as shown in figure 8.2). There are several prominent examples of this type of transmission mechanism in the literature. Corsetti, Pesenti, Roubini, and Tille (1998) model trade competition among the periphery countries in a common third ‘‘center’’ market. For instance, Thailand and Malaysia export many of the same goods to Japan, Hong Kong, and Singapore. Hence, when Thailand devalued in mid-1997, the crisis spread to Malaysia, who lost some of its competitive edge in the common third markets. Calvo (1998) suggests that Wall Street may have been the carrier of the ‘‘Russian virus’’ in the fall of 1998; he focuses on asymmetric information and liquidity problems in the financial centers. So, when Russia (a periphery country) defaulted on its bonds, the leveraged investors that held those bonds in the center country faced margin calls and needed to raise liquidity. The margin calls caused them to sell their asset holdings (the bonds and stock of other countries in their portfolio) to an uninformed counterpart. Because of information asymmetries, a ‘‘lemons problem’’ arises and the assets are sold at a fire sale price.2 A variant of this financial center story concerns open-end fund portfolio managers who need to raise liquidity in anticipation of future redemptions. As before, the strategy would be to sell other assets held in the portfolio. The sell-off
The Globalization of Financial Turmoil
175
depresses the asset prices of other countries and the original disturbance spreads across markets. Frankel and Schmukler (1998) find evidence suggesting that the crisis in Mexico in late 1994 spread to other equity markets in Latin America through New York rather than directly. Kaminsky, Lyons, and Schmukler (2001), who examine the behavior of the mutual fund industry in international equity markets, support this venue of spillover. They find that in the aftermath of the Thai crisis, the largest mutual fund withdrawals affected Hong Kong and Singapore, which have the most liquid financial markets. Kaminsky and Reinhart (2000 and 2001) focus instead on the role of commercial banks lenders in the center country. They stress that following the initial losses due to a crisis in a periphery country, a bank’s need to rebalance the overall risk of its asset portfolio can lead to a marked reversal in commercial bank lending across the markets where the bank has exposure. By calling loans and drying up credit lines to the crisis country, center banks deepen the original crisis. Through the act of calling loans elsewhere, they propagate the crisis to other countries. The debt crisis in the early 1980s and the Asian crisis in 1997 provide two clear examples of this mechanism. Following Mexico’s default in 1982, U.S. banks with extensive exposure to Mexico spread the crisis across Latin America. In 1997, Japanese banks, heavily exposed to Thailand, played the same role in spreading the crisis throughout Asia. Third, there is the transmission of symmetric shocks from the center country to the periphery (figure 8.3). This is the type of shock stressed in several papers by Calvo, Leiderman, and Reinhart (1993, 1996), who analyze the effect of changes in U.S. interest rates on capital flows to Latin America in the early part of the 1990s. While an obvious example of this type of shock is changes in interest rates in a financial center country, more subtle ones may include the kinds of regulatory changes in the financial centers or structural shifts in financial markets. As an example of the latter, the closure of Salomon Brothers’ bond arbitrage desk on July 6, 1998 is thought to have been a factor contributing to the loss of liquidity in the market for emerging-market bonds, making the markets less resilient and impairing LTCM’s prospects.
Figure 8.3 Symmetric Shocks from Center to Periphery
176
8.3
Graciela L. Kaminsky and Carmen Reinhart
Financial Globalization: Measures and Determinants
The crises of the 1990s triggered an immense interest in understanding extreme events. The literature in the late 1990s focused mostly on extreme events in the exchange market as captured by exchange rate devaluations, foreign exchange reserve losses, and in some cases, spikes in overnight interest rates.3 The goal was to examine whether deteriorating fundamentals were at the root of these crises. But when a variety of countries started to topple like dominoes, many authors began to focus on the characteristics of contagion.4 This literature also deals with issues of systemic risk. But systemic risk is not just connected to currency crises. Systemic risk may also be triggered by shocks in banking and stock or bond markets. Our goal in this section is to present a measure of systemic events triggered by turmoil in a financial market in one country.5 We understand financial turmoil as an extreme event in a financial market, be it a rally or a crash. That is, we confine our definition of extreme events to the tail of the distributions of returns by looking at returns in the 5th and 95th percentile of the distribution. Because our interest is in systemic events, we have to consider a substantially large number of markets. Sometimes these systemic events are not long-lasting (for example, the worldwide stock market crash in October 1987), implying a need to use high-frequency data. This puts some restrictions on the markets we can examine. To accommodate these needs, we focus on daily returns in stock markets. Our data set spans the period beginning on January 1, 1997 through August 31, 1999. We focus on the daily return on equities in the local currency, based on the available local bourse indices. The countries in our sample cover mature- and emerging-market economies, thereby excluding countries with less-developed capital markets and a significant extent of financial repression. We can classify the sample into five somewhat arbitrary seven-country groupings: the G7 countries, which are comprised of Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States; and the transition economies, which include Czech Republic, Estonia, Hungary, Poland, Russia, Slovakia, and Ukraine. The remaining three groups are primarily sorted by region. The Asian cluster includes Hong Kong, Indonesia, Malaysia, the Philippines, Singapore, South Korea, and Thailand. The other European group, which excludes those countries that are part of the G7, includes Finland, Greece, Holland, Norway, Spain, Sweden, and Turkey. Finally, the Latin American sample consists of the larger economies in the region, Argentina, Brazil, Chile, Colombia, Mexico, Peru, and Venezuela. Table 8.1 provides summary information on the stock markets examined. Not surprisingly, the degree of instability of stock returns varies considerably across countries. Thus, our definition of extreme events is country-dependent. For example, a drop of 1.8 percent is classified as an extreme event in the United
The Globalization of Financial Turmoil
177
Table 8.1 Stock Market Returns in Domestic Currency: Summary Statistics Percentiles Countries
Mean
5th
95th
Hong Kong
1.40
3.39
3.17
Indonesia
1.62
3.94
4.51
Korea Malaysia
1.96 1.72
4.43 3.63
5.32 3.78
Philippines
1.41
3.26
3.46
Singapore
1.30
2.89
3.00
Thailand
1.78
3.49
4.74
Greece
1.59
3.32
3.62
Finland
1.31
2.79
2.82
Holland
1.16
2.42
2.46
Norway
1.00
2.29
2.18
Spain
1.03
2.13
2.29
Sweden Turkey
0.98 2.49
2.16 5.75
2.00 5.82
Canada France
0.71 0.90
1.63 2.02
1.49 2.01
Italy
1.26
2.61
2.88
Germany
1.20
2.54
2.27
Japan
0.93
1.99
2.20
UK
0.86
1.90
1.86
USA
0.87
1.80
1.90
Argentina
1.60
4.31
3.41
Brazil
2.13
4.84
4.31
Chile
1.03
2.24
2.42
Colombia
0.83
2.01
2.17
Mexico Peru
1.35 1.01
2.78 2.15
3.23 2.33
Venezuela
1.55
3.98
3.48
Czech Republic
0.94
2.30
2.10
Estonia
1.84
4.00
4.54
Hungary
1.63
3.48
3.48
Poland
1.35
3.02
3.03
Russia
2.49
5.10
6.48
Slovakia
0.97
2.49
2.38
Ukraine
2.07
5.18
5.42
Notes: The sample extends from January 1, 1997 to August 31, 1999. Mean is the average of one-day percent returns in absolute values.
178
Graciela L. Kaminsky and Carmen Reinhart
States but it takes a 3.94 percent downfall to qualify as an extreme event in Indonesia. Figure 8.4 provides the first glimpse of the bunching of turmoil that we are trying to explain. This figure reports the percentage of countries with simultaneous extreme changes in stock prices (those movements in the 5th and 95th percentiles) around the globe or in various regions. The top panel reports the proportion of countries worldwide simultaneously in the 5th or 95th percentile of the distribution. The five other panels show the same evidence by region. The globalization of turmoil is quite evident during the last few days of October 1997, following the collapse of the Hong Kong stock market. However, the phenomenon was short lived, underscoring the importance of daily data. The simultaneous turbulences in the fall of 1998 were more persistent. The events triggering these systemic crashes originated in Russia, starting on June 1, 1998 with the suspension of trading in future markets, and were fueled further by the failed auction of Russian GKO government bills on August 27, 1998. On that day, 74 percent of the stock markets around the world collapsed. Turbulence persisted until October as the crisis in Russia was followed by dislocation in G7 financial markets as LTCM spiraled downward. The final episode was very short and was linked to the Brazilian crisis in January 1999. Regional turmoil is far more frequent, with the last half of 1997 plagued by problems in most Asian countries. In the first half of 1998, the turbulence spread to Latin America, with turmoil in Brazil at its core. Fragility in mature markets was mostly concentrated in the fall of 1998. In the next section we use two definitions of globalization to examine the determinants of world and regional instability. 8.3.1
Weak-Form Globalization
This definition focuses on whether turmoil in one country (returns in the 5th or 95th percentile of the distribution) triggers anomalous behavior in other countries. Anomalous behavior is defined as a change in the distribution of returns. In particular, we estimate the frequency distribution of returns in country j on days of turmoil in country i, and the frequency distribution of returns in country j on days of no turmoil in country i. We compare these two distributions using the Kolmogorov-Smirnov test of equality of distributions. We classify a country as affected by extreme crashes or rallies in another country when we reject the null hypothesis of equality of the distributions at a five percent significance level or less. We call this phenomenon weak-form globalization from country i to country j because it does not impose simultaneous occurrence of returns in the tails for globalization to occur. We first examine the degree of weak-form globalization triggered by turmoil in three financial centers: Germany, Japan, and the United States. Table 8.2 reports
Figure 8.4 The globalization of turmoil Note: Numbers in the y-axis represent the percentage of countries experiencing turmoil. Turmoil is defined as those observations in the 5th and 95th percentiles.
0.93
0.86
0.87
Japan
UK
USA
1.26
1.20
Italy
Germany 2.20 1.86 1.90
1.99
1.90
1.80
2.88 2.27
2.61
2.54
1.49 2.01
1.63
2.02
0.71
0.90
Canada
France
5.82
5.75
2.49
Turkey
2.29 2.00
2.13 2.16
1.03 0.98
Spain Sweden
2.46
1.00
2.18
1.16
Holland
Norway
2.42
3.62 2.82
3.32
2.79
1.59
1.31
Greece
Finland
2.29
4.74
3.49
1.78
Thailand
3.46 3.00
1.41
1.30
Philippines
Singapore
3.26
5.32 3.78
4.43
3.63
2.89
1.96
1.72
Korea
3.17 4.51
3.39 3.94
95th
1.40 1.62
Mean 5th
Malaysia
Hong Kong Indonesia
Countries
Percentiles
On all days
1.39
1.79
1.20
3.57
2.59
2.22
1.36
3.74
2.40 2.18
2.49
2.77
3.07
2.06
2.09
1.95
2.23
2.38
2.22
2.08 2.55
3.42
3.25
3.07
5.87
5.63
4.29
3.72
10.12
6.03 4.34
5.39
5.21
5.67
6.13
4.05
4.66
5.54
4.80
4.97
4.92 5.24
Mean 5th
3.10
3.32
2.84
5.17
5.29
4.45
2.75
9.26
5.12 3.21
5.05
4.89
5.05
3.60
6.10
4.23
4.21
5.89
6.35
5.07 7.41
95th
Percentiles
0.18
***
**
***
***
***
***
0.06
*** ***
***
***
***
0.06
0.69
0.18
***
**
0.98
0.44 **
0.97
1.23
2.95
1.88
2.02
1.42
0.93
3.56
1.58 1.74
1.68
1.75
2.20
2.17
2.38
2.27
2.00
2.63
2.81
2.60 2.74
2.06
2.96
4.15
4.67
5.23
3.94
2.04
10.12
4.79 4.07
4.95
3.85
4.85
6.12
3.94
3.93
4.95
5.37
5.52
6.61 4.76
2.12
2.67
4.37
4.21
4.48
3.47
1.67
8.78
4.01 3.28
5.05
3.80
4.67
5.07
8.16
7.00
5.83
5.71
7.26
7.41 7.30
On days of turmoil in USA
0.28
**
***
**
**
***
0.07
0.12
** ***
**
**
***
0.10
0.47
***
**
0.06
***
** ***
2.60
1.19
1.19
1.90
1.61
1.23
1.83
3.45
1.31 1.47
1.53
1.69
2.11
2.39
1.83
1.85
1.84
2.58
2.41
2.40 2.17
3.73
3.06
2.79
5.35
4.33
3.34
3.72
8.85
3.69 3.71
4.64
4.54
4.79
5.57
3.54
3.62
4.86
6.15
5.45
3.92 4.84
4.02
3.00
2.77
3.94
4.96
3.11
2.90
9.89
3.78 3.31
4.44
3.33
5.10
4.93
7.52
3.85
4.31
9.09
6.87
7.01 7.12
Percentiles K&S p-value Mean 5th 95th
On days of turmoil in Japan
Percentiles K&S p-value Mean 5th 95th
On days of turmoil in Germany
Empirical Distribution of Stock Market Returns
Table 8.2 Turmoil in Financial Centers: How Does It Spread?
***
**
0.62
***
**
0.15
***
**
0.30 **
**
**
**
0.06
0.21
0.23
0.51
0.52
0.76
** ***
K&S p-value
180 Graciela L. Kaminsky and Carmen Reinhart
2.07
Ukraine
3.03 6.48 2.38 5.42
3.02
5.10 2.49
5.18
2.98
3.15
4.73 1.09
2.34
3.48
2.80
1.62
2.25
1.85
0.89 2.67
1.71
3.73
12.20
10.80 3.13
6.02
10.03
8.35
4.28
7.51
5.75
2.38 6.05
5.04
9.77
10.44 5.76
6.06
14.98 2.35
5.28
4.94
8.25
2.82
4.32
4.58
2.30 5.29
3.23
8.07
0.06
0.08
*** 0.71
***
***
0.16
***
0.26
***
0.79 **
**
**
2.14
3.10
3.82 0.96
2.42
3.47
2.97
1.18
1.72
1.29
1.04 1.98
1.05
2.73
6.40
11.07
10.11 2.56
6.41
10.76
11.31
2.99
5.09
3.02
2.28 4.79
2.79
9.71
4.94
8.09
7.46 1.82
4.81
8.63
9.09
2.09
2.32
2.88
3.93 4.20
2.39
4.72
0.52
0.13
0.10 0.30
0.09
**
0.09
0.14
0.06
0.29
0.11 0.09
0.43
0.32
3.51
3.02
4.19 0.83
2.90
3.26
2.38
1.31
2.04
1.66
0.88 2.95
1.78
4.76
9.17
8.63
12.49 2.58
6.60
10.76
8.35
3.48
6.20
4.70
2.42 6.05
4.69
10.08
8.13
7.04
14.67 1.94
6.53
9.10
5.36
2.33
3.77
3.97
1.95 7.17
3.55
10.34
***
0.26
0.07 0.37
***
***
0.43
**
0.13
**
0.88 ***
**
***
Notes: Turmoil is defined as those observations in the 5th and 95th percentiles. Mean is the average of one day percent returns in absolute values. The Kolmogorov-Smirnov test evaluates whether the frequency distribution on days of turmoil in the corresponding financial center is different from the frequency distribution on all other days. 5th and 95th percentiles report the values of stock market returns at those percentiles. The sample extends from January 1, 1997 to August 31, 1999. ***, ** represent the significance of the Kolmogorov-Smirnov test at 1 and 5 percent level respectively.
1.35
2.49 0.97
Poland
4.54 3.48
4.00
3.48
1.84
1.63
Estonia
Hungary
Russia Slovakia
2.10
2.30
Czech Republic
0.94
1.55
2.33
1.01
Peru
Venezuela
3.48
1.03
0.83 1.35
Chile
Colombia Mexico
2.15
2.42 2.17 3.23
2.24
2.01 2.78
2.13
3.98
3.41 4.31
4.31
4.84
1.60
Argentina
Brazil
The Globalization of Financial Turmoil 181
182
Graciela L. Kaminsky and Carmen Reinhart
the spillover of extreme events with a country-by-country detail. To get a highresolution picture of anomalous behavior in the stock market, we report the 5th and 95th percentiles of the distribution of returns for all observations in the sample and for the observations on days of market turmoil in each of the financial centers. For example, the 5th and 95th percentile returns for Argentina for the whole sample are 4.31 and 3.41. When there is turmoil in the United States, the 5th and 95th percentile returns for Argentina become 9.37 and 8.13. As shown in this table, the Kolmogorov-Smirnov test rejects the null hypothesis of similar distribution of stock market returns in Argentina on days of financial turmoil in the United States and the distribution on all other days. Thus, we catalog Argentina as suffering weak-form globalization from turmoil in the United States. In the event, extreme movements in equity markets in the United States are transmitted instantaneously to most Latin American countries—the only exceptions being Colombia and Venezuela. Indeed, 70 percent of the countries in Latin America are, according to our measure, affected by toil events in the United States. In contrast, turmoil in the United States triggers an anomalous behavior in only 29 percent of Asian stock markets. The pattern of the problem-spreading in Japan is in sharp contrast to that observed for the United States. In this case, Latin American markets do not react at all to turmoil in Japan, but 71 percent of the Asian countries experience anomalous returns when Japan posts an outsized return. Table 8.3 summarizes these results. Overall, shocks in financial centers are transmitted instantaneously to basically all (70 percent) mature markets (G7 and European countries), whether the shock occurs in Germany, Japan, or the United States. These results are suggestive of the higher degree of integration of those markets. The regional characteristics of the transmission of shocks to emerging Table 8.3 Weak-Form Globalization of Turmoil: Regional and World Effects Percentage of Countries with Anomalous Returns when Turmoil in Regions
Germany
Japan
USA
Asia
43
71
29
Europe
71
71
71
100
75
75
Latin America
43
0
71
Transition economies
57
14
43
World
59
44
56
G7
Notes: Turmoil is defined as those observations in the 5th and 95th percentiles. An anomalous return is interpreted as a change in the distribution of returns in country j on days of turmoil in country i.
The Globalization of Financial Turmoil
183
economies are, however, different. U.S. shocks are strongly transmitted across Latin America; the shocks in Germany simultaneously affect stock markets in Eastern Europe, Latin America, and Asia; while Japanese turbulences mostly affect other Asian countries. Interestingly, this pattern of transmission matches the pattern of exposure of financial institutions in Germany, Japan, and the United States to emerging economies as examined by van Rijckenghem and Weder (2000). These authors classify bank lending to emerging economies by area of loan origin. They find that European banks are the largest creditors in all regions, with North American banks concentrating their lending in Latin America and Japanese banks mostly lending to other Asian countries. In particular, at the onset of the Asian crisis, 32 percent of all the international loans to Asian countries originated in Japan, 44 percent originated in Europe, and just 10 percent originated in North America. Also, during 1997 and 1998, most lending to Eastern European countries (including Russia) originated in Western Europe (80 percent), while lending to Latin America originated in Western European banks (60 percent) and North American banks (30 percent). Rijckenghem and Weder (2000) also examine the shifts in portfolios of European, North American, and Japanese banks during the Asian and Russian crises. Japanese banks consistently withdrew from Asia, reducing their lending from $124 billion in mid-1997 to $86 billion by the end of 1998. North American banks mainly shifted their lending among emerging markets during the Asian crisis (from Asia to Latin America and Europe), while they reduced their positions in all three regions during the Russian crisis. European banks continued to build up their positions in all regions even after the onset of the Asian crisis. Only during the first half of 1998 did they reduce their holdings in Asia while increasing them in Latin America and Eastern Europe. The Russian crisis triggered the end of this expansionist investment strategy of European banks in emerging markets, with all banks reducing their exposure to all three regions by about $20 billion. Table 8.4 examines whether turmoil is transmitted from one country in the periphery to another country in the periphery or to financial centers. In particular, it examines the pattern of spillovers on days of turmoil in three crisis-prone countries in our sample—Brazil, Russia, and Thailand—on a country-by-country basis. Table 8.5 summarizes the information. The patterns of globalization are similar for Brazil and Russia. Turbulence in those countries coincides with abnormal movements around the globe, with the sole exception of Asia. Extreme movements in Thailand are not so far-reaching, in that they spill over only to other Asian economies. This evidence begs for an answer as to through which channels these crisis-prone countries with small asset markets have such far-reaching effects. To answer this question, we examine whether the days of turbulence in a
0.93
0.86
0.87
Japan
UK
USA
1.26
1.20
Italy
Germany 2.20 1.86 1.90
1.99
1.90
1.80
2.88 2.27
2.61
2.54
1.49 2.01
1.63
2.02
0.71
0.90
Canada
France
5.82
5.75
2.49
Turkey
2.29 2.00
2.13 2.16
1.03 0.98
Spain Sweden
2.46
1.00
2.18
1.16
Holland
Norway
2.42
3.62 2.82
3.32
2.79
1.59
1.31
Greece
Finland
2.29
4.74
3.49
1.78
Thailand
3.46 3.00
1.41
1.30
Philippines
Singapore
3.26
5.32 3.78
4.43
3.63
2.89
1.96
1.72
Korea
3.17 4.51
3.39 3.94
95th
1.40 1.62
Mean 5th
Malaysia
Hong Kong Indonesia
Countries
Percentiles
On all days
1.72
1.29
1.37
2.01
2.02
1.38
1.32
3.73
1.61 1.69
1.99
1.89
2.14
2.20
2.18
2.13
2.09
2.52
2.60
2.41 2.50
3.32
3.13
3.00
5.54
4.33
3.34
3.32
9.25
4.74 3.61
4.95
5.00
4.85
4.99
4.05
3.58
5.28
4.73
5.65
4.82 5.15
Mean 5th
3.70
3.00
3.76
3.94
4.96
2.94
2.29
8.78
3.78 3.17
4.57
3.38
4.36
5.50
7.52
5.92
4.80
9.42
6.76
7.41 6.77
95th
Percentiles
***
***
0.24
**
**
***
***
***
** ***
***
***
***
**
0.31
0.09
0.12
0.68
0.37
0.32 **
1.31
1.52
1.18
2.05
2.08
1.51
1.19
4.29
1.67 1.61
2.11
1.82
2.42
2.23
2.05
1.94
2.19
2.85
2.18
2.05 2.32
2.80
3.14
2.55
5.68
5.63
4.29
3.18
10.99
6.03 4.07
5.23
5.21
5.59
6.02
3.51
3.30
6.01
4.53
4.97
3.26 5.15
4.02
2.69
3.03
3.51
3.22
2.55
2.46
9.86
3.06 2.09
3.92
2.68
4.50
3.85
5.96
7.00
3.98
11.80
4.91
6.84 6.47
0.23
***
0.61
***
***
***
**
***
*** ***
***
***
***
***
0.69
0.24
***
**
0.49
0.17 0.30
1.01
0.99
1.25
1.58
1.39
0.99
0.94
3.12
1.16 1.22
1.61
1.32
1.71
1.89
5.71
2.35
2.46
2.81
2.35
2.45 2.89
2.13
2.03
3.22
4.32
3.91
3.08
1.90
8.23
3.09 2.79
4.73
3.15
4.81
5.53
6.33
3.84
4.84
5.86
5.66
4.52 5.03
2.18
2.89
3.56
3.35
2.86
2.20
2.13
7.98
2.54 2.09
3.88
2.50
3.31
4.02
10.42
7.62
6.80
6.91
7.39
6.84 8.00
0.13
0.11
0.25
0.38
0.33
0.69
**
0.65
0.58 0.31
0.08
0.29
0.74
0.26
***
**
**
0.12
0.87
*** **
K&S p-value
On days of turmoil in Thailand
Percentiles K&S p-value Mean 5th 95th
On days of turmoil in Russia
Percentiles K&S p-value Mean 5th 95th
On days of turmoil in Brazil
Empirical Distribution of Stock Market Returns
Table 8.4 Turmoil in Emerging Markets: How Does It Spread?
184 Graciela L. Kaminsky and Carmen Reinhart
2.07
Ukraine
3.03 6.48 2.38 5.42
3.02
5.10 2.49
5.18
4.58
4.37
4.75 0.97
2.61
3.65
3.04
1.61
2.63
1.93
1.10 3.21
2.25
7.67
10.44
10.94
12.49 3.04
5.74
10.51
10.49
3.70
7.51
5.64
3.74 6.05
5.04
10.09
8.51
8.99
13.85 2.68
6.00
9.32
7.01
2.81
3.85
3.84
2.02 8.43
4.34
12.19
***
***
** 0.94
***
***
***
***
***
***
** ***
***
***
3.32
3.27
9.74 1.19
2.45
3.61
3.54
1.83
2.03
1.63
0.91 2.58
1.58
3.89
10.44
11.56
17.49 3.10
6.27
10.76
10.49
3.93
6.65
5.11
3.11 5.57
5.04
10.08
7.41
8.37
16.71 2.52
4.66
5.91
8.64
3.13
3.08
3.78
1.88 6.40
2.53
7.66
**
0.16
*** **
***
***
***
***
**
**
** **
0.16
**
2.23
2.37
3.46 1.14
1.98
2.18
2.37
0.98
1.85
1.21
0.88 1.80
1.28
2.56
4.15
6.24
7.35 3.07
5.00
6.02
6.98
2.21
4.55
2.47
2.08 3.76
2.86
5.19
6.09
8.56
8.56 2.90
3.75
3.47
5.74
2.06
3.67
3.51
2.47 4.59
3.36
6.72
0.12
0.96
0.31 0.47
0.06
0.48
0.37
0.96
0.32
**
0.40 0.26
0.09
0.20
Notes: Turmoil is defined as those observations in the 5th and 95th percentiles. Mean is the average of one day percent returns in absolute values. The Kolmogorov-Smirnov test evaluates whether the frequency distribution on days of turmoil in the corresponding emerging market is different from the frequency distribution on all other days. 5th and 95th percentiles report the values of stock market returns at those percentiles. The sample extends from January 1, 1997 to August 31, 1999. ***, ** represent the significance of the Kolmogorov-Smirnov test at 1 and 5 percent level respectively.
1.35
2.49 0.97
Poland
4.54 3.48
4.00
3.48
1.84
1.63
Estonia
Hungary
Russia Slovakia
2.10
2.30
Czech Republic
0.94
1.55
2.33
1.01
Peru
Venezuela
3.48
1.03
0.83 1.35
Chile
Colombia Mexico
2.15
2.42 2.17 3.23
2.24
2.01 2.78
2.13
3.98
3.41 4.31
4.31
4.84
1.60
Argentina
Brazil
The Globalization of Financial Turmoil 185
186
Graciela L. Kaminsky and Carmen Reinhart
Table 8.5 Weak-Form Globalization of Turmoil: Regional and World Effects Percentage of Countries with Anomalous Returns when Turmoil in Regions Asia Europe G7 Latin America
Brazil
Russia
Thailand
14
29
100
100
67 0
83
67
17
100
86
14
Transition economies
86
83
0
World
76
73
18
Notes: A turmoil is defined as those observations in the 5th and 95th percentiles. An anomalous return is interpreted as a change in the distribution of returns in country j on days of turmoil in country i.
particular crisis-prone emerging market were also days of turbulence in a financial center with which that particular country was associated. We then examine whether problems in crisis-prone emerging markets not associated with turmoil in financial centers also have wide spillover effects. We chose financial centers according to the evidence discussed in the literature. We pair Brazil with the United States, Russia with Germany, and Thailand with Japan. Table 8.6 examines the periphery-to-periphery and periphery-center-periphery connections. Days of turmoil in crisis-prone emerging economies are divided into two samples, those on which the corresponding financial center was also roiled, and those on which the corresponding financial center was not. The results are dramatically different. Turbulence in Brazil accompanied by turbulence in the United States is transmitted around the world, with Asia the only untouched region. In contrast, turbulence in Brazil unaccompanied by turbulence in the United States only affects stock markets in Latin America. Turmoil in this case only has a regional reach. The same picture of propagation of shocks is observed in the case of Russian jitters. Turbulence becomes global if a financial center is affected, but remains regional when the stock market in the financial center is calm. The evidence from Thailand is somewhat different. Again, simultaneous turmoil in the financial center ( Japan) and in Thailand triggers a broader propagation of shocks. But here this propagation is only regional in nature. There is not even regional propagation when turbulence affects only the stock market in Thailand. The regional characteristics of some the turbulences in stock markets agree with the evidence from currency crises.6 The question is, what causes this regional pattern of spillovers? Strong bilateral trade patterns may provide one explanation. For example, Kaminsky and Rein-
The Globalization of Financial Turmoil
187
hart (2000) point to the strong bilateral trade among Mercosur countries, but caution that turmoil in Brazil is still transmitted rapidly to non-Mercosur Latin American countries. Similarly, shocks from Russia are strongly transmitted to most of the transition economies even though bilateral trade links among transition economies diminished drastically in importance following the collapse of the communist regimes in Eastern Europe in 1989–1991. Third-party trade links may provide another explanation. For example, Malaysia and Thailand sell similar goods to Japan and the United States, explaining the contagion from Thailand to Malaysia following the Thai devaluation in July 1997. But the Mexican crisis in 1994 strongly affected Argentina and Brazil and these countries do not compete with Mexico in third markets. Again, financial links may help to explain regional contagion. For example, Kaminsky, Lyons, and Schmukler (2004) examine investment strategies of U.S.-based mutual funds specializing in Latin America and find that they were a key element in explaining the reach of the Tequila crisis: as investors stampeded out of mutual funds specializing in Latin America following the Mexican devaluation, managers (under the pressure of the massive redemptions) had to sell not just Mexican stocks, but also stocks from Argentina and Brazil. Table 8.7 summarizes these results by region and examines the null hypothesis of financial center irrelevance versus the alternative hypothesis that a financial center has to be affected for the turmoil to become systemic. To examine this hypothesis, we construct the Wilcoxon, or rank-sum, test. To construct this test, we look at the results from the Kolmogorov-Smirnov test and construct two samples. The first sample captures the weak-form globalization pattern following turmoil in a crisis-prone emerging market coinciding with turmoil in a financial center. For each j country in the sample, we assign a value equal to 1 if turmoil in the pair of crisis-prone emerging market and financial center triggers anomalous behavior in country j, and 0 otherwise. The second sample captures the weak-form globalization pattern following turmoil in a crisis-prone emerging market not coinciding with turmoil in a financial center. Again, for each j country in the sample, we assign a value equal to 1 if turmoil in just the crisis-prone emerging market triggers anomalous behavior in country j, and 0 otherwise. Denote the observations from the first sample by fXg and the observations from the second sample by fYg. The null hypothesis of financial-center irrelevance implies that PðX > kÞ ¼ PðY > kÞ for all k. We are interested in the one-sided alternative that X is stochastically larger than Y; that is, PðX > kÞ > ðY > kÞ for all k. To construct the rank-sum test, we rank all the observations without regard to the sample to which they belong. Then the Wilcoxon test statistic is formed as the sum of the ranks in the first sample:
Table 8.6 Financial Turmoil in Emerging Markets and Financial Centers: How Does It Spread? Empirical Distribution of Stock Market Returns On all days
On days of turmoil in Brazil With financial center
Percentiles Countries
Mean
5th
Without financial center
Percentiles 95th
Mean
5th
95th
K&S p-value
Percentiles Mean
5th
95th
K&S p-value
Hong Kong
1.40
3.39
3.17
2.41
4.82
7.41
0.32
1.67
4.38
5.53
0.74
Indonesia
1.62
3.94
4.51
2.50
5.15
6.77
**
2.00
4.02
6.26
0.34
Korea
1.96
4.43
5.32
2.60
5.65
6.76
0.37
2.23
5.65
6.76
0.82
Malaysia
1.72
3.63
3.78
2.52
4.73
9.42
0.68
2.49
3.75
14.94
0.92
Philippines
1.41
3.26
3.46
2.09
5.28
4.80
0.12
1.68
4.60
3.11
0.12
Singapore
1.30
2.89
3.00
2.13
3.58
5.92
0.09
1.73
3.21
4.92
0.36
Thailand
1.78
3.49
4.74
2.18
4.05
7.52
0.31
1.97
3.95
6.36
0.77
Greece
1.59
3.32
3.62
2.20
4.99
5.50
**
1.64
4.00
3.47
0.13
Finland
1.31
2.79
2.82
2.14
4.85
4.36
***
1.59
4.49
2.83
**
Holland
1.16
2.42
2.46
1.89
5.00
3.38
***
1.57
2.78
2.92
0.10
Norway
1.00
2.29
2.18
1.99
4.95
4.57
***
1.66
4.04
3.47
**
Spain
1.03
2.13
2.29
1.61
4.74
3.78
**
1.43
4.83
2.49
0.15
Sweden
0.98
2.16
2.00
1.69
3.61
3.17
***
1.41
3.11
2.58
0.16
Turkey
2.49
5.75
5.82
3.73
9.25
8.78
***
2.88
8.04
7.09
0.26
Canada
0.71
1.63
1.49
1.32
3.32
2.29
***
0.71
1.81
1.51
0.98
France
0.90
2.02
2.01
1.38
3.34
2.94
***
1.18
2.95
2.20
0.19
Italy
1.26
2.61
2.88
2.02
4.33
4.96
**
1.83
4.19
3.38
0.11
Germany
1.20
2.54
2.27
2.01
5.54
3.94
**
1.52
3.55
2.15
0.41
Japan
0.93
1.99
2.20
1.37
3.00
3.76
0.24
1.36
2.41
3.87
0.20
UK
0.86
1.90
1.86
1.29
3.13
3.00
***
1.12
2.13
2.59
0.09
USA
0.87
1.80
1.90
1.72
3.32
3.70
***
0.93
1.77
1.68
0.19
Argentina
1.60
4.31
3.41
4.58
10.44
8.51
***
3.48
8.82
6.55
***
Brazil
2.13
4.84
4.31
7.67
10.09
12.19
***
6.68
9.69
8.81
***
Chile
1.03
2.24
2.42
2.25
5.04
4.34
***
1.95
3.87
4.08
***
Colombia
0.83
2.01
2.17
1.10
3.74
2.02
**
1.14
3.97
2.23
**
Mexico
1.35
2.78
3.23
3.21
6.05
8.43
***
2.19
3.45
5.12
0.12
Peru
1.01
2.15
2.33
1.93
5.64
3.84
***
1.44
5.34
3.37
**
Venezuela
1.55
3.98
3.48
2.63
7.51
3.85
***
2.33
7.19
3.93
**
Czech Republic
0.94
2.30
2.10
1.61
3.70
2.81
***
1.44
3.71
2.85
**
Estonia
1.84
4.00
4.54
3.04
10.49
7.01
***
2.78
8.05
8.57
0.22
Hungary
1.63
3.48
3.48
3.65
10.51
9.32
***
2.46
6.35
7.99
0.29
Poland
1.35
3.02
3.03
2.61
5.74
6.00
***
1.84
3.79
4.36
0.26
Russia
2.49
5.10
6.48
4.75
12.49
13.85
**
3.58
6.54
8.44
0.13
Slovakia
0.97
2.49
2.38
0.97
3.04
2.68
0.94
0.91
3.07
2.93
0.94
Ukraine
2.07
5.18
5.42
4.37
10.94
8.99
***
3.72
10.94
8.99
0.08
Notes: Turmoil is defined as those observations in the 5th and 95th percentiles. Mean is the average of one-day percent returns in absolute values. The Kolmogorov-Smirnov test evaluates whether the frequency distribution on days of turmoil in the corresponding emerging market (with or without turmoil in a financial center) is different from the frequency distribution on all other days. 5th and 95th percentiles report the values of stock market returns at those percentiles. The sample extends from January 1, 1997 to August 31, 1999. ***, ** represent the significance of the Kolgomorov-Smirnov test at 1 and 5 percent respectively.
Empirical Distribution of Stock Market Returns On days of turmoil in Russia With financial center Percentiles Mean 5th
95th
On days of turmoil in Thailand Without financial center
Percentiles K&S p-value Mean 5th 95th
With financial center Percentiles K&S p-value Mean 5th 95th
Without financial center Percentiles K&S p-value Mean 5th 95th
K&S p-value
2.05
3.26
6.84
0.17
1.78
3.10
6.95
0.11
2.45
4.52
6.84
***
1.96
4.51
5.55
0.13
2.32
5.15
6.47
0.30
1.79
3.44
5.83
0.56
2.89
5.03
8.00
**
2.13
4.28
5.75
0.72
2.18
4.97
4.91
2.85
4.53
11.80
0.49
2.14
3.94
5.08
0.13
2.35
5.66
7.39
0.87
2.00
5.66
7.39
0.69
**
2.74
4.68
9.98
**
2.81
5.86
6.91
0.12
2.15
4.05
4.36
0.48
2.19
6.01
3.98
***
1.79
4.49
3.95
0.58
2.46
4.84
6.80
**
1.99
4.43
5.97
0.34
1.94
3.30
7.00
0.24
1.62
3.30
6.22
0.15
2.35
3.84
7.62
**
1.80
3.82
5.56
0.36
2.05
3.51
5.96
0.69
2.29
3.47
7.44
0.77
5.71
6.33
10.42
***
5.59
6.76
10.74
***
2.23
6.02
3.85
***
1.71
3.86
3.95
0.70
1.89
5.53
4.02
0.26
1.76
5.84
3.67
0.31
2.42
5.59
4.50
***
1.51
3.17
3.95
0.11
1.71
4.81
3.31
0.74
1.38
2.66
3.06
0.47
1.82
5.21
2.68
***
1.14
2.53
1.86
0.27
1.32
3.15
2.50
0.29
1.24
1.98
2.93
0.49
2.11
5.23
3.92
***
1.49
2.96
3.89
0.07
1.61
4.73
3.88
0.08
1.31
2.77
3.51
0.19
1.67
6.03
3.06
***
0.86
1.78
1.96
0.68
1.16
3.09
2.54
0.58
0.99
2.30
2.39
0.20
1.61
4.07
2.09
***
1.10
2.77
2.06
0.12
1.22
2.79
2.09
0.31
0.98
2.36
2.02
0.34
4.29
10.99
9.86
***
3.56
6.66
9.87
0.18
3.12
8.23
7.98
0.65
3.13
8.22
7.96
0.42
1.19
3.18
2.46
**
0.88
2.20
2.03
0.12
0.94
1.90
2.13
**
0.92
1.92
2.21
0.08
1.51
4.29
2.55
***
0.84
2.08
1.30
0.39
0.99
3.08
2.20
0.69
0.90
2.03
2.34
0.99
2.08
5.63
3.22
***
1.46
2.90
3.05
0.23
1.39
3.91
2.86
0.33
1.16
2.33
2.78
0.10
2.05
5.68
3.51
***
1.09
2.31
2.10
0.31
1.58
4.32
3.35
0.38
1.44
2.58
3.62
0.82
1.18
2.55
3.03
0.61
1.05
1.93
3.06
0.59
1.25
3.22
3.56
0.25
0.69
1.50
1.87
**
1.52
3.14
2.69
***
1.15
2.11
2.35
0.10
0.99
2.03
2.89
0.11
0.89
1.65
2.73
0.40
1.31
2.80
4.02
0.23
1.04
1.83
3.49
0.79
1.01
2.13
2.18
0.13
0.99
2.11
2.16
0.16
3.32
10.44
7.41
**
2.63
4.85
8.49
0.13
2.23
4.15
6.09
0.12
2.18
4.74
6.53
0.33
3.89
10.08
7.66
**
3.24
8.46
10.79
0.51
2.56
5.19
6.72
0.20
2.57
5.12
7.20
0.24
1.58
5.04
2.53
0.16
1.27
3.07
3.73
0.63
1.28
2.86
3.36
0.09
1.31
2.99
3.53
0.28
0.91
3.11
1.88
**
0.85
3.79
1.94
0.18
0.88
2.08
2.47
0.40
0.91
1.98
2.51
0.37
2.58
5.57
6.40
**
1.91
3.75
5.89
0.50
1.80
3.76
4.59
0.26
1.56
3.66
3.62
0.61
1.63
5.11
3.78
**
1.25
2.83
3.84
0.47
1.21
2.47
3.51
**
1.16
2.29
2.65
**
2.03
6.65
3.08
**
1.50
4.19
3.07
0.24
1.85
4.55
3.67
0.32
1.95
4.17
4.12
0.12
1.83
3.93
3.13
***
1.43
2.33
3.07
0.06
0.98
2.21
2.06
0.96
0.99
2.18
2.35
0.99
3.54
10.49
8.64
***
3.42
9.09
9.31
***
2.37
6.98
5.74
0.37
1.76
6.16
4.63
0.76
3.61
10.76
5.91
***
2.42
7.22
4.22
0.14
2.18
6.02
3.47
0.48
1.69
3.49
3.12
0.72
2.45
6.27
4.66
***
1.86
4.38
3.86
**
1.98
5.00
3.75
0.06
1.51
4.28
2.83
0.20
9.74
17.49
16.71
***
9.37
17.85
15.71
***
3.46
7.35
8.56
0.31
3.08
7.29
8.32
0.18
1.19
3.10
2.52
**
1.11
3.06
2.79
0.40
1.14
3.07
2.90
0.47
1.31
4.01
3.19
0.23
3.27
11.56
8.37
0.16
3.04
7.04
8.83
0.19
2.37
6.24
8.56
0.96
1.65
3.68
8.28
0.99
14 891
24 0.00
67
1452
73
83
86
759
12
33
0
0
14 0
Without financial center
0.00
W p-value
1188
18
0
14
17
67 0
With financial center
Thailand
1023
6
0
14
14
0 0
Without financial center
0.00
W p-value
Notes: The financial center is respectively USA for Brazil, Germany for Russia, and Japan for Thailand. Turmoil is defined as those observations in the 5th and 95th percentiles. An anomalous return is interpreted as a change in the distribution of returns in country j on days of turmoil in country i.
1320
World
Wilcoxon statistic (W)
86 76
Transition economies
83
0
83 100
G7
Asia Europe
Latin America
0 29
14 100
Regions 29 100
With financial center
W p-value
With financial center
Without financial center
Russia
Brazil
Percentage of countries with anomalous returns when turmoil in
Table 8.7 Weak-Form Globalization of Turmoil: Regional and World Effects
190 Graciela L. Kaminsky and Carmen Reinhart
The Globalization of Financial Turmoil
W¼
n X
191
ð8:1Þ
R1i
i¼1
where n is the number of countries in each sample. Under the null hypothesis, the average rank of an observation in sample 1 should equal the average rank of an observation in sample 2. Using Fisher’s principle of randomization, it is straightforward to verify that EðWÞ ¼
nð2n þ 1Þ 2
and
VarðWÞ ¼
ns 2 2
ð8:2Þ
where s is the standard deviation of the combined ranks ri for both samples: s2 ¼
2n 1 X ðri rÞ 2 : 2n 1 i¼1
ð8:3Þ
The last row of table 8.7 shows the Wilcoxon test statistic for each sample and the one-sided p value for the null hypothesis of financial-center irrelevance. For example, for the case of Brazil, the proportion of all countries affected when both Brazil and the United States experience turmoil is 76 percent, and the proportion of countries affected when just Brazil experiences turmoil is 24 percent. For these two samples, the Wilcoxon p-value under the null hypothesis of financial-center irrelevance is less than 0.01, leading us to reject the null hypothesis of financialcenter irrelevance. The results for the other two emerging markets are similar. In all cases, the tests reject the null hypothesis of financial-center irrelevance in favor of the alternative hypothesis that a financial center has to be affected for turmoil to become systemic. 8.3.2
Strong-Form Globalization
In the previous section we examined whether turmoil in one country triggers anomalous behavior in stock markets around the world, with anomalous behavior defined as a change in the distribution of returns. Under this definition of globalization, other stock markets do not have to experience extreme returns in response to extreme returns in one stock market for globalization to occur. A more stringent concept of the globalization of turmoil would be one of simultaneous turmoil. We call this definition of the globalization of turmoil ‘‘strong-form globalization.’’ A globalization index in this case will just be the proportion of countries with simultaneous extreme events. Our task in this section is to examine the determinants of this index.
192
Graciela L. Kaminsky and Carmen Reinhart
To examine the causes of systemic events, we use a multinomial logit approach. We also estimated the model using order logit techniques. The results are quite similar, so we do not report them to save space. Since we are interested in explaining the degree of globalization, our left-hand variable will be a dummy variable that can take three values: low, medium, and high globalization.7 Low globalization occurs when less than 25 percent of the countries in the sample experience turmoil; medium globalization occurs when there are more than 25 percent, but less than 50 percent, of the countries in turmoil. Finally, high globalization occurs when 50 percent or more of all the countries experience turmoil. Our explanatory variables are dummy variables capturing days of turmoil in financial centers, days of turmoil in crisis-prone countries on days of turmoil in financial centers, and days of turmoil in crisis-prone countries when financial centers are not affected. These dummies will take a value of 1 on days of turmoil and 0 otherwise. Equation 8.4 is the multinomial logit equation to be estimated. , ! j1 X 0 0 Pðy ¼ iÞ ¼ expðx bi Þ 1þ x bi ð8:4Þ i¼1
The variable y is the globalization index, and the vector x includes the dummy variables capturing turmoil in the various countries. The variable Pðy ¼ iÞ is the probability associated with outcome i. The index j refers to the number of outcomes in our estimation: low, medium, and high globalization. The vector b is the vector of coefficients to be estimated. As is usual in this type of estimations, for each explanatory variable we estimate j 1 parameters. The probability that there is low globalization is our base case and it is equal to , ! j1 X 0 Pðy ¼ lowÞ ¼ 1 1þ expðx b i Þ : ð8:5Þ i¼1
The estimation of equation 8.4 is somewhat problematic because not all the markets are open at the same time. Thus, a shock leading to turmoil in Brazil can affect all Latin American economies the same day, European economies the same day or the following day depending on the time at which the shock occurs, and Asian countries only on the following day. Similarly, if a shock occurs in Russia, the index of globalization on the left-hand side has to include countries in turmoil in Europe, the G7, and Latin America on the same day and countries in turmoil in Asia the next day, but if the turmoil originates in Thailand, the index of globalization on the left-hand side has to include the number of countries in turmoil in all the regions the same day of the shock.
The Globalization of Financial Turmoil
193
We deal with this problem in two different ways. First, we estimate equation 8.4 using only turmoil originating in shocks from one time zone at a time. In this case, the left-hand-side variable is constructed depending on the origin of the shock, and we estimate three separate versions of equation 8.4 for financial centers and three separate versions of equation 8.4 for crisis-prone emerging markets. The shortcoming of this type of estimation is that we cannot evaluate jointly the effects of extreme events in the various crisis-prone countries and financial centers. Second, to account for the effect of turbulence in the three crisis-prone countries jointly, we perform panel estimations. To deal with the different time zones, the index of globalization on the left-hand side accounts for low, medium, and high globalization by region. For each region, we align the explanatory variables on the right-hand side according to the region they may affect. Since we estimate the regression for all the regions at the same time, the parameters b provide a somewhat different measure of the effects of turmoil in the various countries on globalization. For example, the episodes of high globalization are more confining in the sense that they require all the regions to have a high degree of globalization simultaneously. This was not the case in the nonpanel estimation. Finally, within the panel regression estimates, we jointly evaluate the effects of coincidence of multiple shocks in emerging markets and financial centers. We construct two dummy variables. The first one captures days of turmoil in emerging markets coinciding with turmoil in financial centers. This variable can take four values, 0 to 3. If this variable takes the value 3, it means that the three crisisprone emerging economies experience turmoil and so do their respective financial centers. The second explanatory variable in this regression will capture the number of crisis-prone emerging markets in turmoil when there is no turmoil in financial centers. This variable also takes four values, 0 to 3. Tables 8.8 and 8.9 examine the effects of turmoil originating in one time zone at a time. Table 8.8 concentrates on turmoil originating in financial centers. The first equation has as its explanatory variable a dummy variable equal to 1 when Germany experiences turmoil, and 0 otherwise. The second regression has as its explanatory variable a dummy variable equal to 1 when Japan experiences turmoil, and 0 otherwise. Finally, the third equation has as its explanatory dummy variable a dummy variable equal to 1 when United States experiences turmoil, and 0 otherwise. Table 8.9 uses the same methodology to evaluate the degree of globalization following jitters in one turmoil cluster at a time: Brazil-U.S., RussiaGermany, and Thailand-Japan. For each turmoil cluster, the regression has two explanatory dummy variables. One dummy variable is equal to 1 on days of turbulences in the emerging market coinciding with days of turmoil in the corresponding financial center, and 0 otherwise. The second explanatory dummy variable is equal to 1 on days of turbulences in the emerging market not accompanied
194
Graciela L. Kaminsky and Carmen Reinhart
Table 8.8 P2 expðx 0 bi ÞÞ Strong-Form Globalization: Multinomial Logit Estimation Pð y ¼ iÞ ¼ expðx 0 bi Þ=ð1 þ i¼1 Coefficients Degree of globalization
Turmoil in Germany
Turmoil in Japan
Turmoil in USA
Medium
2.51*** (7.88)
1.78*** (5.49)
1.25*** (3.74)
High
4.71*** (7.14)
2.56*** (5.72)
2.45*** (6.85)
Pseudo R2 Number of observations
0.19 694
0.08 694
0.10 694
Probabilities conditional on Degree of globalization
Turmoil in Germany
Turmoil in Japan
Turmoil in USA
Low
40
58
52
Medium High
36 23
26 16
22 26
Notes: Numbers in parentheses represent z statistics. ***, **, * represents the significance of the coefficient at the 1, 5, and 10 percent levels. Turmoil is defined as those observations in the 5th and 95th percentiles. The left-hand-side variable captures the degree of globalization. There are three possible degrees of globalization: low (when less than 25 percent of the countries in the sample experience turmoil), medium (when more than 25 percent but less than 50 percent of the countries experience turmoil), high (when 50 percent or more of all countries in the sample experience turmoil). In order to be able to estimate our model, coefficients for the low globalization are set equal to zero (that is our base case). Interpretation of the reported coefficients has to be done with respect to the base case. Our model was estimated with a constant but constant coefficients are not reported here for expositional purposes. Probabilities are given in percent terms and are derived from the multinomial logit estimation shown in the top panel.
by turmoil in the corresponding financial center, and 0 otherwise. To evaluate jointly the contribution of these clusters to the globalization of turmoil, we estimate a multivariate turmoil-cluster panel regression. We estimate the model using panel data because of the time-zone problem. The results are reported in table 8.10. Finally, table 8.11 reports the panel estimation evaluating the effects of multiple coincidence of turmoil in the three crisis-prone emerging markets. The top panels in all these tables report the estimated coefficients, while the bottom panels show the conditional probabilities of globalization obtained from the estimations shown in the top panels. As for the results, table 8.8 shows that turmoil in financial centers triggers turbulences around the world, with the explanatory power (as captured by the pseudo R 2 ) ranging from 8 percent for turmoil originating in Japan to 19 percent for turmoil originating in Germany. Again, this pattern could be explained, in part,
p-values
0.13
Pseudo R2
21
21 57
Low
Medium High
27 17
56
Without financial center 10 2
87
No turmoil
***
0.28
0.15
1.18 (1.07)
1.36*** (3.61)
27 50
23
With financial center 23 2
75
Without financial center
Turmoil in Russia
Pseudo R2
5.53*** (7.87)
2.70*** (4.33)
7 1
92
No turmoil
***
*
p-values
0.03
0.37 (0.63)
0.91** (2.3)
Without financial center
40 13
47
With financial center
17 4
80
Without financial center
Turmoil in Thailand
Pseudo R2
2.18*** (2.6)
2.32*** (4.03)
With financial center
Turmoil in Thailand
8 3
90
No Turmoil
*
**
p-values
Notes: Numbers in parentheses represent z statistics. ***, **, * represents the significance of the coefficient at the 1, 5, and 10 percent levels. Turmoil is defined as those observations in the 5th and 95th percentiles. The left-hand-side variable captures the degree of globalization. There are three possible degrees of globalization: low (when less than 25 percent of the countries in the sample experience turmoil), medium (when more than 25 percent but less than 50 percent of the countries experience turmoil), high (when 50 percent or more of all countries in the sample experience turmoil). In order to be able to estimate our model, coefficients for the low globalization had to equal zero (that is our base case). Interpretation of the reported coefficients has to be done with respect to the base case. Our model was estimated with a constant but constant coefficients are not reported here for expositional purposes. P column reports p-values for test of equality between parameters estimated with and without turmoil in financial centers. The financial center is, respectively, USA for Brazil, Germany for Russia, and Japan for Thailand. Number of observations for our sample was 694. Probabilities are given in percent terms and are derived from the multinomial logit estimation shown in the top panel.
With financial center
Degree of globalization
Turmoil in Brazil
Probabilities conditional on
2.47*** (4.85)
4.64*** (8.44)
High
1.41*** (3.61)
2.14*** (3.62)
Without financial center
With financial center
Without financial center
With financial center
Medium
Degree of globalization
Turmoil in Russia
Turmoil in Brazil
Coefficients
Table 8.9 P2 expðx 0 b i ÞÞ Strong-Form Globalization Multinomial Logit Estimation Pð y ¼ iÞ ¼ expðx 0 b i Þ=ð1 þ i¼1
The Globalization of Financial Turmoil 195
p-values
2.67*** (9.22)
High
0.85** (2.11)
0.58*** (2.12)
0.07 3469
***
***
77 21 1
Low
Medium High
36 13
51 13 3
84
24 15
61
With financial center
Without financial center
With financial center
***
0.54
Turmoil in Russia
0.37 (0.99)
0.17 (1.07) 2.14*** (4.68)
0.63** (2.34)
28 2
70
Without financial center
p-values
With financial center
Thailand
***
*
p-values
31 9
60
With financial center
22 2
76
Without financial center
Turmoil in Thailand
0.41 (1.15)
0.05 (0.31)
Without financial center
Notes: Numbers in parentheses represent z statistics. ***, **, * represents the significance of the coefficient at the 1, 5, and 10 percent levels. Turmoil is defined as those observations in the 5th and 95th percentiles. The left-hand-side variable captures the degree of globalization. There are three possible degrees of globalization: low (when less than 25 percent of the countries in the sample experience turmoil), medium (when more than 25 percent but less than 50 percent of the countries experience turmoil), high (when 50 percent or more of all countries in the sample experience turmoil). In order to be able to estimate our model, coefficients for the low globalization had to equal zero (that is our base case). Interpretation of the reported coefficients has to be done with respect to the base case. Our model was estimated with a constant but constant coefficients are not reported here for expositional purposes. P column reports p-values for test of equality between parameters estimated with and without turmoil in financial centers. The financial center is, respectively, USA for Brazil, Germany for Russia, and Japan for Thailand. Probabilities are given in percent terms and are derived from the multinomial logit estimation shown in the top panel.
No Turmoil
Degree of globalization
2.71*** (8.49)
0.35 (1.31)
Without financial center
Turmoil in Brazil
Probabilities conditional on
Pseudo R2: Number of observations:
0.93*** (4.23)
With financial center
Without financial center
With financial center
Medium
Degree of globalization
Russia
Brazil
Coefficients
Table 8.10 P2 expðx 0 b i ÞÞ Strong-Form Globalization: Multinomial Logit Panel Estimation Pð y ¼ iÞ ¼ expðx 0 bi Þ=ð1 þ i¼1
196 Graciela L. Kaminsky and Carmen Reinhart
77 21 1
Low
Medium
High
13
31
56
With financial center
2
22
76
Without financial center
62
20
19
With financial center
4
22
75
Without financial center
Turmoil in two emerging markets
***
***
p-value
94
4
2
With financial center
6
21
73
Without financial center
Turmoil in three emerging markets
Notes: Numbers in parentheses represent z statistics. ***, **, * represents the significance of the coefficient at the 1, 5, and 10 percent levels. Turmoil is defined as those observations in the 5th and 95th percentiles. The left-hand-side variable captures the degree of globalization. There are three possible degrees of globalization: low (when less than 25 percent of the countries in the sample experience turmoil), medium (when more than 25 percent but less than 50 percent of the countries experience turmoil), high (when 50 percent or more of all countries in the sample experience turmoil). In order to be able to estimate our model, coefficients for the low globalization had to equal zero (that is our base case). Interpretation of the reported coefficients has to be done with respect to the base case. Our model was estimated with a constant but constant coefficients are not reported here for expositional purposes. P column reports p-values for test of equality between parameters estimated with and without turmoil in financial centers. The explanatory variable is emerging market. Such variable could equal 0, 1, 2, 3 depending on how many emerging markets (Brazil, Russia, Thailand) experienced turmoil concurrently. Probabilities are given in percent terms and are derived from the multinomial logit estimation shown in the top panel.
No turmoil
0.57*** (2.71) 0.07 3469
0.01 (0.11)
Without financial center
Turmoil in one emerging market
Probabilities conditional on
2.64*** (14.64) Pseudo R2 #Observations
0.65*** (4.76)
With financial center
Degree of globalization
High
Medium
Degree of globalization
Coefficients Emerging markets
Table 8.11 P2 expðx 0 b i ÞÞ Strong-Form Globalization: Multinomial Logit Panel Estimation Pð y ¼ iÞ ¼ expðx 0 bi Þ=ð1 þ i¼1
The Globalization of Financial Turmoil 197
198
Graciela L. Kaminsky and Carmen Reinhart
by the higher worldwide exposure of German banks to emerging markets in Asia, Latin America, and transition economies during the late 1990s. The bottom panel provides another metric to capture the reach of turmoil in the three financial centers: the probability of low, medium, and high globalization, conditional on turmoil in each financial center and derived from the multinomial estimation. Again the likelihood of medium-to-high globalization on days of turmoil in Germany is about 60 percent, but only about 40 percent when shocks originate in Japan, with market jitters in the United States triggering medium-to-high globalization with a probability of almost 50 percent. Table 8.9 reports the results for turmoil originating in emerging markets. The first three columns report the estimates for Brazil, the next three for Russia, and the last three for Thailand. As we did when we evaluated weak forms of globalization, we pay particular attention to whether days of turmoil in the three crisisprone emerging markets coincide with days of turmoil or with days of no turmoil in financial centers. Financial centers would be irrelevant in explaining high degrees of globalization of turmoil if the coefficient b attached to the dummy capturing turmoil in the emerging market–financial center cluster is not statistically different from the coefficient b attached to the dummy capturing turmoil in just the emerging market. This hypothesis is tested in the third column for each emerging market. In all cases, we reject this hypothesis at all conventional significance levels. To better understand the effects of turmoil in the various countries, the bottom panel of table 8.9 also reports the conditional probabilities of low, medium, and high globalization obtained from the estimation of equation 8.4. The results for Brazil indicate that low globalization is the most likely outcome when turbulence in Brazil does not coincide with turbulence in a financial center. In contrast, when the financial center is also experiencing an extreme event, high globalization becomes the most likely event, with the probability reaching 57 percent. Interestingly, if there is no turmoil in Brazil or the United States, the likelihood of a high clustering of countries with turmoil collapses to 2 percent. The results for Russia are quite similar. Again, the necessary ingredient for high globalization is the coincidence of turmoil in Russia and Germany. In this case, the average probability of high globalization is about 50 percent, but declines to 2 percent when only Russia experiences turbulence. The results for Thailand indicate that the reach of the Asian crisis was limited in scope. Still, jitters in Thailand are transmitted to other Asian countries only on days of jitters in Japan. The explanatory power of these shocks, as captured by the pseudo R 2 , is high for Brazil and Russia but, as expected, quite small for Thailand. To evaluate jointly the effects of turmoil in the three emerging markets and financial centers, we report the panel estimates in tables 8.10 and 8.11. As shown in table 8.10, each of the three emerging market–financial center clusters contributes
The Globalization of Financial Turmoil
199
to trigger financial turmoil worldwide, as captured by the statistically significant b coefficients of the three crisis-cluster dummies for the high globalization event. Still, the contribution of the Thailand–Japan cluster is somewhat smaller. Our panel estimation, though, has a smaller predictive power than the nonpanel estimations because of the restrictions imposing similar effects of turmoil of the various center-periphery clusters across the five regions. Our more stringent definition of high-globalization episodes is also reflected in lower probabilities of high globalization following turbulences in the three center-periphery clusters. Finally, the results in table 8.11 bring to the spotlight the magnification effect of simultaneous turbulences in several center-periphery clusters. Note that the probability of high globalization now increases to 94 percent when the three crisis clusters experience turbulences, but just 13 percent when one crisis cluster is in turmoil. Note that the probability of high globalization on days of no turmoil in any of the crisis clusters is just 1 percent. 8.4
The Origins of Globalization
In the previous section, we evaluated the odds of simultaneous turbulence around the world when crisis-prone emerging markets and financial centers were experiencing turmoil. We did not explain the origin of these turbulences. To do that, we have to bring in information beyond that of daily movements in equity prices. Our source is the written record: we used reports from Bloomberg.com, the Financial Times, and the Wall Street Journal to construct a chronology of news in those days. We limited our search to days on which at least 50 percent of countries in one region had stock market jitters. This chronology is reported in the appendix table (located at the end of this chapter). The first column dates the days of regional and global turmoil. The next six columns report the proportion of countries, worldwide and by region, with stock market turmoil. For clarity, we just report the proportion of markets in turmoil when it reaches more than 50 percent of the countries worldwide or in each region. The last column reports the news. To study the onset and propagation of turmoil, it is important to collect all news, local and foreign, that triggers jitters. This news can be about the state of the economy, financial institutions, or policies, or may just be rumors. The appendix table does not report all the news events that moved markets on a particular day, but reports the most common source of market jitters in the region or around the world. As shown in this chronology, the first day of worldwide turmoil is October 27, 1997, with 57 percent of the countries around the globe experiencing turmoil. The tension started to build up toward the end of August. Until that time, while several Asian countries experienced turbulences, they did not spread to other
200
Graciela L. Kaminsky and Carmen Reinhart
countries in that region. But on August 28, 1997, financial markets in Indonesia, Malaysia, the Philippines, Singapore, and Hong Kong collapsed amid a deepening loss of confidence in the ability of governments to tackle their severe economic problems. On October 22 turmoil reached Hong Kong and spread in Asia, with about 60 percent of the Asian countries experiencing market crashes. The crisis in Hong Kong deepened and on October 23, it triggered a global sell off in Europe, the G7 countries, and Latin America. By October 27, worldwide globalization reached about 60 percent of the countries in the sample. This time around, the globalization of the turmoil was short-lived and within two days markets rebounded, with massive rallies around the world. December 11 is the next day of significant interregional spillover, with Korea at the center of the debacle in Asia and Europe. Still, repercussions in the G7 countries were minor. Another day of interregional turmoil was January 12, 1998. At the heart of the jitters was the collapse of Peregrine (Hong Kong), one of Asia’s largest investment banks. The next cluster of global instability started toward the end of May 1998, with Russian tension spreading to Latin America, transition economies, Asia, Europe, and the G7 countries. The degree of globalization rapidly rose, reaching about 50 percent of countries worldwide by June 15. Rumors of devaluation in China and the weakness of the Japanese economy and the yen also contributed to the buildup of skittishness. The degree of globalization reached 60 percent on August 11. On August 21, shares of German banks heavily exposed to Russia collapsed, triggering downfalls in other G7 countries. On August 27, the failed auction of Russian GKOs reignited fears of financial collapse, bringing major downturns in 75 percent of countries worldwide. Financial turmoil griped Latin American markets following Moody’s downgrade of Brazilian and Venezuelan foreign debt. Moody’s also put Argentina’s debt and its eleven banks on review for a possible downgrade on September 3. While markets in some regions rebounded during the first week of September, financial concerns, brought again to the limelight by Standard & Poor’s downgrade of Spain’s second-largest bank (with heavy exposure to Argentina) and of Argentina’s two largest banks on September 10, together with LTCM’s collapse and bail-out on September 24, triggered stock market crashes around the world. This episode of worldwide financial instability came to an end with news of credit easing in financial centers; this was related to the intermeeting reduction in the federal funds interest rate on October 15 in the United States. The last episode of financial instability in our sample occurred around the time of the devaluation of the Brazilian real, which was extremely short-lived. Only on January 13 did financial markets around the world collapse.
The Globalization of Financial Turmoil
201
Table 8.12 summarizes our findings about the news that rocked financial markets. But first, the top panel shows the proportion of days of rallies and days of crashes during episodes of high regional and world globalization (at least 50 percent of countries affected by turmoil). Note that 85 percent of the episodes of high world globalization involve stock market crashes. Episodes of high regional globalization are more balanced. With the exception of the Asian region, in which days of joint rallies outnumber days of simultaneous crashes, about 60 percent of the days of high regional globalization consist of crashes. The middle panel classifies the days of high globalization, both at a regional level and worldwide, according to the type of news that seems to have triggered the spillover. Financial concerns from bankruptcies of large banks or adverse shocks in one or more asset markets in a center country seem to be at the core of high worldwide globalization (40 percent of the episodes). Only 20 percent of the days of high spillovers seem to be driven by economic, political, and monetary news at the center. Another important source of instability is concerns about financial fragility in the periphery (25 percent of the episodes). In contrast, financial worries in center countries only account for 26 percent of the episodes of high regional globalization. Financial fragility in the periphery seems to be at the heart of regional turbulences (31 percent of the episodes). Finally, international agreements also contribute to regional turbulences. One final aspect of globalization that we have still not addressed is whether high globalization occurs when the magnitude of the shocks in the stock market is larger. The bottom panel addresses this question. We first divide extreme returns in three categories according to their size: large (within the 1-percent critical region on both tails), medium (between the 1-percent and 3-percent critical regions on both tails), and small (between the 3-percent and 5-percent critical regions). Afterward, we estimate the average size of the returns for all the countries in turmoil for each episode of low, medium, and high world globalization. The bottom panel in table 8.12 shows the proportion of episodes of low, medium, and high world globalization with small, medium, and large returns. Larger (in absolute values) returns are more common on days of high globalization: all the shocks in episodes of high globalization are clustered in, at the most, the 3-percent critical region, while during episodes of low globalization 46 percent of the shocks are relatively small (between the 3-percent and 5-percent critical regions). 8.5
Concluding Comments
This paper presents a new approach to measuring and understanding systemic financial turbulences. We defined two measures of systemic disturbances—weak-
202
Graciela L. Kaminsky and Carmen Reinhart
Table 8.12 Days of Globalization: Asymmetries, Origins, and Size of Shocks A. Asymmetries Days of high globalization Regions
Proportion of crashes
Proportion of rallies
World
85
15
Asia Latin America
29 69
71 31
Europe
61
39
G7
56
44
Transition economies
61
39
B. News on Days of High Globalization Proportion of Days with News about: Economy and politics
Financial sector Banking Regions Regional World
Center
Monetary policy
International agreements
Other Periphery Center
Periphery Center
Periphery Center
PeriphPeriphery Center ery
8
8
18
23
11
10
11
3
10
10
30
15
10
10
10
5
2
7
C. Degree of Globalization and Size of Returns Degree of world globalization
Returns Small
Medium
Large
Low
46
48
5
Medium
12
86
2
0
92
8
High
Notes: Numbers in the above tables are in percent. In panel C, small returns C are returns between the 3rd (97th) and 5th (95th) percentiles. Medium returns are returns between the 1st (99th) and 3rd (97th) percentiles. Large returns are returns in the 1st (99th) percentile. The first cell of this panel indicates that 46 percent of the days of low globalization had countries experiencing on average a small return.
The Globalization of Financial Turmoil
203
and strong-form globalization—and created the corresponding indices of globalization. These indices allowed us to capture the routes through which market jitters in one country reach other countries in the same region, or even worldwide. They also allowed us to estimate the likelihood of low-to-high globalization following a variety of shocks in crisis-prone emerging markets and financial centers. One of the preliminary conclusions we draw from this exercise is that financial centers are at the core of systemic problems: the worldwide globalization of the turbulences in Asia in fall 1997 only occurred after the stock market crash in the United States on October 27, while the Russian downfall spread around the globe only after it triggered fragilities in German banks and helped to provoke LTCM’s bankruptcy. Without distress in a financial center, disturbances spread at most regionally, with the ‘‘silk road’’ of regional financial distress related in part to trade links, but also to financial linkages. For example, as documented in Kaminsky, Lyons, and Schmukler (2004), the 1994 Mexican crisis spread so rapidly to Argentina and Brazil via the massive mutual fund (specialized in Latin America) withdrawals from those two countries. Finally, our evidence indicates that collapses and not rallies are at the heart of high-globalization episodes, suggesting the need for models with asymmetries to explain systemic turmoil. Our research has focused on explaining the geographical extent of financial turmoil. Still, the temporal dimension of high-globalization episodes of turbulences varies as much, with some episodes lasting just a couple of days (sell-off in stock markets around the world following the Hong Kong collapse in October 1997) while others, such as the turmoil during fall 1998, lingered much longer. Also, our research, like most of the previous literature, has just focused on a particular asset market. But the degree of systemic problems should not only be understood as synchronized jitters across a particular asset market in a variety of countries, but also as simultaneous turmoil across markets in a particular country. Future research should inquire into these differences too. Acknowledgments We have received insightful comments and suggestions from Fernando Broner, Guillermo Calvo, and Takatoshi Ito, as well as participants at presentations held at the 2002 Third Joint Central Bank Research Conference on Risk Measurement and Systemic Risk held in Basel, Switzerland; the LACEA 2002 meeting in Madrid, Spain; the Bank of England, Cornell University, George Washington University, the London School of Economics, the University of Maryland, and the University of West Virginia. We are grateful to Amine Mati for superb research assistance.
204
Graciela L. Kaminsky and Carmen Reinhart
Appendix Table The Globalization of Financial Turmoil: Chronology of News January 1, 1997 to August 31, 1999 DAY
ALL
ASIA
7 Apr 97
EUR
G7
TRA
57
28 Aug 97
71
22 Oct 97
57
23 Oct 97
NEWS EUR: European and US stocks up, benefiting from comments of EU finance ministers who indicated the single currency will begin on time. (FSOTHER, CENTER) ASIA: Investors flee the region amid a deepening loss of confidence in the ability of governments to tackle their severe economic problems. (E&PN, PERIPHERY) ASIA: Share prices fall sharply in Asian markets due to a sharp drop in futures prices in Singapore and fears about higher interest rates and currency stability in Hong Kong. (FS-OTHER, PERIPHERY)
57
71
57
G7, EUR, LA: Hong Kong Monetary Authority was forced to sell US dollars to support the currency, triggering interest rate hikes that prompted a global sell off. (FS-OTHER, PERIPHERY)
71
86
86
EUR, G7, LA: A $600 billion sell-off shut down the US market for the first time since 1981. The sell-off was triggered by Southeast Asia’s shaky economies and by a jump in interest rates, as well as by a stream of weak earning reports. Panic grips other regions as US market crashes, especially after the Hong Kong declines of the past week. (E&PN, PERIPHERY); (FS-OTHER, CENTER)
71
27 Oct 97
57
28 Oct 97
80
100
86
86
29 Oct 97
63
57
71
86
30 Oct 97
LA
71
57
57
ASIA, EUR, TRA, G7: still reacting to US market crash and Hong Kong crash. (FSOTHER, CENTER) LA: Markets soar as US market rallies. (FS-OTHER, CENTER)
71
EUR, G7, TRA: Markets soared as US soared the day before. (FS-OTHER, CENTER) ASIA: Asian markets finish lower as investors fear another steep drop in US markets. (FS-OTHER, CENTER) EUR: Stocks soared after Greenspan eased concern that inflation could be on the rise. (MP, CENTER)
The Globalization of Financial Turmoil
205
Appendix Table (continued) DAY
ALL
ASIA
EUR
G7
LA
TRA
NEWS LA: Strong declines in the region stemmed from contagion in Asia. Fears about Brazil’s real currency and liquidity crunch of its banking system. (FSOTHER, PERIPHERY); (FS-BANKING, PERIPHERY)
3 Nov 97
57
ASIA: Stocks rally as a financial aid package to Indonesia restores calm to the region. China also eases credit. (IA, PERIPHERY); (MP, PERIPHERY)
7 Nov 97
57
12 Nov 97
71
17 Nov 97
86
24 Nov 97
57
1 Dec 97
11 Dec 97
G7: The US dollar surges, reaching a sixmonth high as concerns increased in the market over the ability of the Japanese government to revive the country’s economy. (E&PN, CENTER)
57
57
LA: Concern about fiscal austerity package announced by Brazil. Markets also fall after steep declines in Asian markets. (E&PN, PERIPHERY), (FSOTHER, PERIPHERY) TRA: Stocks fall after major drops in Asian markets. (FS-OTHER, PERIPHERY) EUR, G7: Stocks up as Japan PM hints that public spending may be used to stimulate the economy and protect depositors following the collapse of the nation’s largest bank. US reports low inflation measures. (FS-BANKING CENTER); (E&PN, CENTER) EUR: Shares fall after the collapse of Japan’s fourth-largest brokerage firm, Yamaichi Securities. (FS-BANKING, CENTER)
57
100
71
G7: Stock markets rally on gains in Asian markets overnight. (FS-OTHER, PERIPHERY) ASIA: Stocks slumped as Moody’s cut rating of South Korea’s currency. (E&PN, PERIPHERY) EUR: Stocks down amid a new wave of selling in Asian markets and signs of weakness in the US economy. (FSOTHER, PERIPHERY); (E&PN, CENTER)
206
Graciela L. Kaminsky and Carmen Reinhart
Appendix Table (continued) DAY
ALL
ASIA
EUR
G7
9 Jan 98
LA
TRA
71
12 Jan 98
71
NEWS LA: Asian turmoil, especially concerns about Indonesia, causes market declines. Central Bank of Chile raises key interest rate. (FS-OTHER, PERIPHERY); (MP, PERIPHERY)
71
EUR, TRA: Peregrine, one of Asia’s largest investment banks (Hong Kong), files for liquidation, raising concerns about emerging markets in general. (FSBANKING, PERIPHERY)
13 Jan 98
71
ASIA: Stocks rose on optimism about IMF-backed reforms for the region. (IA, PERIPHERY)
14 Jan 98
86
ASIA: Stocks continued to rise on optimism about IMF-backed reforms for the region. (IA, PERIPHERY)
19 Jan 98
100
ASIA: Indonesia signaled commitment to the much-awaited bank reform. Camdessus issues statement of confidence about Malaysia and countries in the region. (FS-BANKING, PERIPHERY); (IA, PERIPHERY)
22 Jan 98
57
2 Feb 98
71
ASIA: The plunging Indonesian rupiah dragged the rest of Asia into a downward spiral. (FS-OTHER, PERIPHERY) ASIA: Stocks up as value-oriented funds flooded back into Asia from Europe and US. Strength driven by liquidity even though nothing changed in the fundamentals front. (FS-OTHER, PERIPHERY)
27 Apr 98
86
26 May 98
27 May 98
1 Jun 98
86
EUR, G7: Concern US will raise interest rates to fight inflation. (MP, CENTER) 57
LA: Concerns about a potential devaluation in Russia affecting Brazil and other emerging markets. (FS-OTHER, PERIPHERY)
57
EUR: Speculation about Russian devaluation of the ruble caused fall in stock prices. (FS-OTHER, PERIPHERY) 57
TRA: Russian stock prices plummeted while the main market for Russian futures announced that it was suspending trading indefinitely. Unfulfilled expectations of foreign aid to Russia contributed to the declines. (FS-OTHER, PERIPHERY); (IA, PERIPHERY)
The Globalization of Financial Turmoil
207
Appendix Table (continued) DAY
ALL
15 Jun 98
51
ASIA
EUR
86
57
17 Jun 98
71
18 Jun 98
100
G7
LA 57
ASIA: Countries in the region still reacting to the US and Japan coordinated actions to prop up the yen. (IA, CENTER)
86
71
57
13 Aug 98
57
20 Aug 98
54
71
71
TRA: Russia would receive 22.6 billion dollars from IMF and other bilateral donors. (IA, PERIPHERY) EUR, G7, LA: Foreign investors seemed to be the main driving force behind the market drop. Fears of a weaker yen, and the prospect of devaluation in China, sent shock waves throughout the world. (FS-OTHER, CENTER); (FS-OTHER, PERIPHERY)
71
18 Aug 98
21 Aug 98
ASIA: Japanese government announced that GDP contracted for a second consecutive quarter. (E&PN, CENTER) ASIA, LA, G7, TRA: Loss of confidence in emerging markets in general as Russian market tumbled for a seventh straight day. (FS-OTHER, PERIPHERY)
71
60
NEWS
ASIA: US and Japan coordinated actions to sell US dollars and buy Japanese yen. Markets soared due to the stronger yen. (IA, CENTER)
14 Jul 98
11 Aug 98
TRA
TRA: Russian shares fell more than 10 percent early on growing fears of a liquidity crisis among Russian banks. (FS-BANKING, PERIPHERY) EUR: Gains in European markets following a major Wall Street advance (FS-OTHER, CENTER)
71
LA: Concern Russian banks may fail and Venezuela may devalue (FSBANKING, PERIPHERY); (FS-OTHER, PERIPHERY)
71
LA: Concern about imminent currency devaluation in Venezuela. (FS-OTHER, PERIPHERY) EUR, G7, LA: Russia’s Central Bank stated that some Russian banks could go bankrupt accentuating the Russian financial crisis. In Germany (a major lender to Russia) stocks plunged, triggering downfalls in London and Paris. (FS-BANKING, PERIPHERY); (FSOTHER, CENTER)
208
Graciela L. Kaminsky and Carmen Reinhart
Appendix Table (continued) DAY
ALL
ASIA
26 Aug 98
27 Aug 98
EUR
G7
74
86
57
2 Sep 98
71
3 Sep 98
57
100
100
86
EUR, G7, LA, TRA: Russia’s government unable to sell its newly restructured GKO bills, spreading fear that global crisis will continue. (FS-OTHER, PERIPHERY) LA: Stocks end sharply higher mirroring the DJIA’s rebound. (FS-OTHER, CENTER) EUR: Stocks up on optimism about Europe’s prospects. (FS-OTHER, CENTER)
57
TRA: Markets rebound as investors went for bargains. (FS-OTHER, PERIPHERY) EUR: Stocks follow rebound in the US stock market. (FS-OTHER, CENTER)
57
57
57
LA: Moody’s downgraded Brazil’s and Venezuela’s foreign debt and put Argentina’s foreign currency debt and 11 banks on review for a possible downgrade. (E&PN, PERIPHERY), (FSBANKING, PERIPHERY) EUR, G7: European stock markets were hurt by a dollar plunge and worries that financial troubles are spreading from Russia and Asia to Latin America. (FSOTHER, PERIPHERY) 57
57
NEWS EUR: Stocks fall as Russia announces its debt restructuring plan. (FS-OTHER, PERIPHERY)
4 Sep 98
8 Sep 98
TRA
71
1 Sep 98
7 Sep 98
LA
57
TRA: Russia’s parliament delays a vote on Chernomyrdin’s appointment as Prime Minister at Yeltsin’s request. (E&PN, PERIPHERY) ASIA: Stronger yen and a higher stock market helps Japanese banks but fund managers stay skeptical. (FS-BANKING, CENTER) EUR: Greenspan hints he would favor cutting interest rates. (MP, CENTER)
57
57
G7, TRA: Renewed confidence was felt thanks to market-supportive comments from Fed Chairman Alan Greenspan. (MP, CENTER)
The Globalization of Financial Turmoil
209
Appendix Table (continued) DAY
ALL
10 Sep 98
60
ASIA
EUR 71
G7
LA
71
86
11 Sep 98
14 Sep 98
71
15 Sep 98
17 Sep 98
86
NEWS
57
EUR, G, TRA, LA: Worries about banks’ exposures as S&P downgrades Spain’s second-largest bank. Credit ratings for Argentina’s two largest banks were also reduced. (FS-BANKING, CENTER); (FSBANKING, PERIPHERY)
57
LA: Brazilian Government boosted overnight interest rates by 20 percentage points to try to stem capital flight, which reached 2.2 billion dollars the day before. (FS-OTHER, PERIPHERY)
71
G7, TRA: Russia’s new PM pledges to revive the economy. (E&PN, PERIPHERY)
57
LA, TRA: G7 meeting hints at financial aid for Latin America. Argentina may borrow 5.7 billion dollars from the World Bank and other international institutions. (IA, PERIPHERY)
86
86
EUR, G7: Greenspan states that there is no move to coordinate interest rates (MP, CENTER)
21 Sep 98
86
57
G7, EUR: Concern about Japan’s recession and low growth potential for OECD countries due to emerging markets collapse and deepening financial collapse. Political parties in Japan remains at odds on how to use taxpayer money to prop up LTCB of Japan. (E&PN, CENTER); (FS-OTHER, EMERGING); (FS-BANKING, CENTER)
22 Sep 98
71
23 Sep 98
57
24 Sep 98
54
TRA
86
EUR: US markets rebound day after the Clinton grand jury testimony. (E&PN, CENTER) 57
100
LA: President of IDB says Brazil could receive up to 50 billion dollars in aid from international institutions. IMF and US also gave statements of support for Brazil aid. (IA, PERIPHERY) G7, EUR: Investors hope that Greenspan will hint at a possible rate cut when he testifies before the Senate banking committee. (MP, CENTER)
57
ASIA: Stocks up as Greenspan suggests he may lower interest rates. (MP, CENTER)
210
Graciela L. Kaminsky and Carmen Reinhart
Appendix Table (continued) DAY
ALL
ASIA
EUR
G7
LA
TRA
NEWS LA: Stocks down as concern over banks is felt after some of the largest banks put together a 4 billion dollar bailout of LTCM, raising concern about credit. Brazil announces fiscal austerity measures. (E&PN, PERIPHERY); (FSBANKING, CENTER)
25 Sep 98
71
30 Sep 98 1 Oct 98
57 66
86
100
2 Oct 98
8 Oct 98
9 Oct 98
G7: US cut interest rates and asked other countries to follow suit. (MP, CENTER) 71
57
6 Oct 98
51
86
57
TRA: Russian tax collection continued to plummet in September, due to the crash on Russian financial markets and the country’s ensuing banking crisis (statement by tax official). (E&PN, PERIPHERY)
G7, EUR, LA: Concerns about global economic slump. Report US manufacturing production weakened for fourth straight month as exports slumped. (E&PN, CENTER) 57
LA: Stock markets soared on hopes of a financial package for troubled Brazil. (IA, PERIPHERY) TRA: Stocks still falling following global declines of October first. (FS-OTHER, CENTER)
57
G7: Disappointment that the G7 meeting in Washington failed to adopt a clear strategy to address global economic issues drove share prices sharply lower in world markets. (IA, CENTER)
71
EUR, G7: Speculation the Fed would cut interest rates. Japan moves to repair its economy. (MP, CENTER) (E&PN, CENTER)
57
57
ASIA, G7: Interest cuts in UK and other European countries in the preceding week generated rallies in several markets. (MP, CENTER) LA: Brazilian authorities and the International Monetary Fund issued a joint statement on the availability of a rescue package to help cushion the region from market turmoil. (IA, PERIPHERY)
The Globalization of Financial Turmoil
211
Appendix Table (continued) DAY
ALL
12 Oct 98
54
16 Oct 98
ASIA
EUR
57
71
71
71
20 Oct 98
57
27 Oct 98
71
G7
LA
TRA
71
EUR, G7, ASIA: Japan will substantially increase the amount of money it will spend on shoring up its fragile banking system. (FS-BANKING, CENTER) EUR, G7: Fed Funds rate cut by a quarter percentage point on Oct. 15. (MP, CENTER)
57
EUR, G7: Suggestions that France and Germany would lower their interest rates boosted investor sentiment in Europe as well as continued gains in the USA and a rally in Asian markets. (MP, CENTER); (FS-OTHER, CENTER); (FSOTHER, PERIPHERY) EUR: Italy makes a surprise cut in interest rate by a full percentage point. (MP, CENTER)
30 Oct 98
71
2 Nov 98
71
4 Nov 98
NEWS
LA: G7 countries said they would back a new IMF credit line to Brazil, speeding aid to Brazil. (IA, PERIPHERY) 57
71
EUR, TRA: Stocks rallied after the October 30 US commerce department report announcing better than expected third quarter growth rates. (E&PN, CENTER) G7: Democrats increased seats in the US Congressional elections, the first party with an incumbent resident to do this since 1934. Stocks rally after interest rate cuts in Italy and Sweden in the past week. (E&PN, CENTER); (MP, CENTER)
10 Nov 98
57
ASIA: Investors await the release of the Japanese government’s stimulus package. (E&PN, CENTER)
11 Nov 98
57
ASIA: Japan’s newest economic stimulus package is expected to be the largest ever. (E&PN, CENTER) EUR: European stocks finished with strong gains as bourses benefited from hopes of further European rate cuts. (MP, CENTER) G7: Global markets were given a boost after the DJIA marked a record high. (FS-OTHER, CENTER)
20 Nov 98
57
30 Nov 98
57
57
212
Graciela L. Kaminsky and Carmen Reinhart
Appendix Table (continued) DAY
ALL
ASIA
1 Dec 98
EUR
57
G7
TRA
57
21 Dec 98 4 Jan 99
86
6 Jan 99
57
LA: The US dollar weakened as investors were discouraged by the continuing decline in U.S. stocks and Wednesday’s defeat in the Brazilian Congress of an important government austerity measure. (FS-OTHER, CENTER); (E&PN, PERIPHERY)
57
G7: High expectations on the euro boost stocks. (FS-OTHER, CENTER)
57
EUR, G7: Stock prices ended up higher lured by a weak dollar and start of euro trading. (FS-OTHER, CENTER)
71
G7: US rallied on the back of technology stocks. (FS-OTHER, CENTER) ASIA: Japanese market followed an overnight jump in New York stocks lead by strength in the high-technology sector. (FS-OTHER, CENTER)
12 Jan 99
57
66
100
NEWS LA: Latin American investors were influenced by heavy profit taking on Wall Street and Brazil. (FS-OTHER, PERIPHERY) G7, EUR: Stocks down on weak dollar. (FS, CENTER, OTHER)
71
3 Dec 98
13 Jan 99
LA
57
86
LA: Markets closed sharply lower due to rumors of an interest rate hike in Brazil and a near $200 million outflow. (FSOTHER, PERIPHERY) 57
EUR, LA, TRA: Brazil’s Central Bank chairman resigns. Brazil devalues its currency. (E&PN, PERIPHERY); (MP, PERIPHERY)
14 Jan 99
57
LA: Standard & Poor’s downgraded certain Latin American banks and some of Brazil’s foreign currency debt. (FSBANKING, PERIPHERY)
15 Jan 99
71
LA: Brazil lets its currency float against the dollar. (FS-OTHER, PERIPHERY)
18 Jan 99
9 Feb 99
57
EUR: Bank mergers in France, Spain and calmer financial markets in Brazil pushed stocks higher. (FS-BANKING, CENTER); (FS-OTHER, PERIPHERY) 57
G7: There were growing concerns in Europe about a slowdown in the economy. European markets fell following financial turmoil in emerging markets.
The Globalization of Financial Turmoil
213
Appendix Table (continued) DAY
ALL
ASIA
EUR
G7
LA
TRA
NEWS Japanese investors were waiting for measures, if any, from the BOJ to curb the recent sharp rise in bond yields, which would increase borrowing costs for companies and could stall Japan’s efforts to revive its battered economy. (E&PN, CENTER); (FS-OTHER, PERIPHERY)
5 Mar 99
86
16 Apr 99
57
19 Apr 99
57
26 May 99
29 July 99
G7: Labor department reported hourly wages rose 0.1 percent in February, less than the 0.3 percent forecasted. Unemployment went up 0.1 percent point. (E&PN, CENTER) ASIA: Influx of European funds brought up Asian stocks posting sharp gains throughout the region. (FS-OTHER, CENTER) ASIA: Investors confident that the global financial crisis is largely over. (FSOTHER, PERIPHERY) 71
57
LA: Markets rebound as fears concerning Argentina’s ability to maintain its currency board (as well as fears about a potential political scandal involving Brazilian President Cardoso) subside. (FS-OTHER, PERIPHERY); (E&PN, PERIPHERY) G7: Investors were relieved when Alan Greenspan offered nothing new to upset global markets in a testimony to US lawmakers. (MP, CENTER)
Notes: FS: News from the financial sector. They could either originate in the banking sector (BANKING) or not (OTHER). MP: News about monetary policy. E&PN: News about the economy (excluding the financial sector) and political news. IA: Refers to international agreements or policy coordination actions. ASIA: Includes Hong Kong, Indonesia, Malaysia, the Philippines, Singapore, South Korea, and Thailand. EUR: Includes Finland, Greece, Holland, Norway, Spain, Sweden, and Turkey. G7: Includes Canada, France, Germany, Italy, Japan, United Kingdom, and the United States. LA: Includes Argentina, Brazil, Chile, Colombia, Mexico, Peru, and Venezuela. TRA: Includes Czech Republic, Estonia, Hungary, Poland, Russia, Slovakia, and Ukraine. Numbers in cells represent the percentage of countries in their respective region (or world) experiencing turmoil on that day. The parenthetical statements after each news event explain the region from which news originated and our classification of news. For example, on July 29, 1999, 57 percent of the G7 countries were affected by Alan Greenspan’s testimony. His testimony was classified as Monetary Policy News originating in the Center (MP, CENTER).
214
Graciela L. Kaminsky and Carmen Reinhart
Notes 1. Extreme returns are those returns in the 5th and 95th percentile of the distribution. 2. Also see Calvo and Mendoza (2000) for evidence suggesting that this mechanism can be important. 3. See, for example, Kaminsky and Reinhart (1999). 4. See, for example, Eichengreen, Rose, and Wyplosz (1996), Glick and Rose (1998), and Kaminsky and Reinhart (2000). 5. See also Danielsson and de Vries (1997), De Bandt and Hartmann (2000), Hartman, Straetmans, and Devries (2004), Longin (1996), and Mati (2001) for studies of extreme returns in stock and bond markets. 6. See, for example, Gelos and Sahay (2000), Glick and Rose (1998), and Kaminsky and Reinhart (2000). 7. In Kaminsky and Reinhart (2000) we constructed a similar index. In that paper, the index was the proportion of countries with currency crises, which was used to predict currency crises in other countries. Bae, Karolyi, and Stulz (2000) also look at simultaneous financial strains in Asia and Latin America and construct a similar index, finding that contagion is predictable using a small set of macroeconomic variables.
References Bae, Kee-Hong, Andrew G. Karoyli, and Rene M. Stulz. 2003. ‘‘A New Approach to Measuring Financial Contagion.’’ Review of Financial Studies 16, no. 3: 717–763. Calvo, Guillermo. 1998. ‘‘Capital Market Contagion and Recession: An Explanation of the Russian Virus.’’ Mimeo., University of Maryland. Calvo, Guillermo A., Leonardo Leiderman, and Carmen M. Reinhart. 1993. ‘‘Capital Flows and Real Exchange Rate Appreciation in Latin America: The Role of External Factors.’’ IMF Staff Papers 40, no. 1. ———. 1996. ‘‘Capital Flows to Developing Countries in the 1990s: Causes and Effects.’’ Journal of Economic Perspectives 10: 123–139. Calvo, Guillermo, and Enrique Mendoza. 2000. ‘‘Rational Contagion and the Globalization of Securities.’’ Journal of International Economics 51, no. 1: 79–113. Calvo, Sara, and Carmen M. Reinhart. 1996. ‘‘Capital Flows to Latin America: Is There Evidence of Contagion Effects?’’ In Private Capital Flows to Emerging Markets After the Mexican Crisis, eds. Guillermo A. Calvo, Morris Goldstein, and Eduard Hochreitter, 151–171. Washington, D.C.: Institute for International Economics. Corsetti, Giancarlo, Paolo Pesenti, Nouriel Roubini, and Cedric Tille. 1998. ‘‘Structural Links and Contagion Effects in the Asian Crisis: A Welfare-Based Approach.’’ Mimeo., Yale University, New Haven, CT. Danielsson, J., and C. de Vries. 1998. ‘‘Value-at-Risk and Extreme Returns.’’ Working Paper No. 98017/2, Tinbergen Institute, Amsterdam. De Bandt, Olivier, and Philipp Hartmann. 2000. ‘‘Systemic Risk: A Survey.’’ Discussion Paper No. 2634, CEPR, London. Eichengreen, Barry, Andrew Rose, and Charles Wyplosz. 1996. ‘‘Contagious Currency Crises.’’ Working Paper No. 5681, NBER, Cambridge, MA.
The Globalization of Financial Turmoil
215
Frankel, Jeffrey, and Sergio Schmukler. 1998. ‘‘Crises, Contagion, and Country Funds: Effects on East Asia and Latin America.’’ In Managing Capital Flows and Exchange Rates: Perspectives from the Pacific Basin, ed. Reuven Glick, 232–266. New York: Cambridge University Press. Gelos, Gasto´n, and Ratna Sahay. 2000. ‘‘Financial Market Spillovers in Transition Economies.’’ Working Paper WP/00/71, IMF, Washington, D.C. Glick, Reuven, and Andrew Rose. 1998. ‘‘Contagion and Trade: Why Are Currency Crises Regional?’’ Working Paper No. 6806, NBER, Cambridge, MA. Kaminsky, Graciela L., and Carmen M. Reinhart. 1999. ‘‘The Twin Crises: The Causes of Banking and Balance-of-Payments Problems.’’ American Economic Review 89, no. 3: 473–500. ———. 2000. ‘‘On Crises, Contagion, and Confusion.’’ Journal of International Economics 51, no. 1: 145– 168. ———. 2001. ‘‘Bank Lending and Contagion: Evidence from the Asian Crisis.’’ In Regional and Global Capital Flows: Macroeconomic Causes and Consequences, eds. T. Ito and A. Krueger, 73–99. Chicago: University of Chicago Press. Kaminsky, Graciela, Richard Lyons, and Sergio Schmukler. 2001. ‘‘Economic Fragility, Liquidity, and Risk: The Behavior of Mutual Funds during Crises.’’ Working Paper, World Bank, Washington, D.C. Kaminsky, Graciela, Richard Lyons, and Sergio Schmukler. 2004. ‘‘Managers, Investors, and Crises: Mutual Fund Strategies in Emerging Markets.’’ Journal of International Economics 64, no. 1: 113–134. Lewis, Arthur. 1977. The Evolution of the International Economic Order. Princeton: Princeton University Press. Longin, F. M. 1996. ‘‘The Asymptotic Distribution of Extreme Stock Market Returns.’’ Journal of Business 69: 383–408. Mati, Amine. 2001. ‘‘Extreme Returns: News and Patterns of Contagion.’’ Mimeo., George Washington University, Washington, D.C. Van Rijckeghem, Caroline, and Beatrice Weder. 2000. ‘‘Financial Contagion: Spillovers through Banking Centers.’’ Mimeo., IMF, Washington, D.C.
9
Why Do Some Countries Recover More Readily from Financial Crises? Padma Desai and Pritha Mitra
9.1
Introduction
The origins of the financial crises of the 1990s around the globe have been studied extensively. Calvo (1998); Desai (2003a, b); Krugman (1999); Corsetti, Pesenti, and Roubini (2001); Cespedes, Chang, and Velasco (2000); Rodrik and Velasco (1999); and Kaminsky and Reinhart (1999) trace them to an environment of stable exchange rates, high interest rates, and free, cross-border capital flows in emerging markets, which prompted foreign lending to their businesses and governments.1 The poorly regulated financial markets of these economies not only led to excessive borrowing from abroad but also to a double mismatch in their borrowing pattern. The double mismatch consisted of short-term borrowing by banks from abroad and their long-term lending at home, coupled with their accumulated debt liabilities (largely unhedged) in foreign currencies and dubious asset acquisition in domestic currencies. Latin American sovereign borrowers, among them the governments of Argentina and Brazil, also accumulated a significant foreign debt burden without concern for their ability to meet their repayment obligations via export earnings. The East Asian economies were swept in a capital-outflow-led currency and financial crisis that began in Thailand in mid-1997 and spread to neighboring Malaysia, Indonesia, and South Korea. Currencies tumbled at varying rates in Russia in August 1998 and Brazil in January 1999. In unrelated developments, Argentina faced similar turmoil leading to its sovereign debt default in December 2001. Soaring interest rates, which were calculated to arrest capital flight and stabilize exchange rates, plunged these economies into varying levels of recession. Following the crisis, the affected country policy makers adopted a triple framework of free capital mobility, floating exchange rates, and high interest rates aimed at restoring investor confidence. Despite similar policies, their economies recovered at varying speed. The rapid recovery of the East Asian group
218
Padma Desai and Pritha Mitra
contrasted with the slow pace in the Latin American set. Why is it that in East Asia, the postcrisis interest rates almost never surpassed 25 percent, whereas in Latin America they were as high as 91 percent? The dynamics underlying the relatively fast recovery of some countries can provide new insights into the design of appropriate post-crisis policy agendas. However, postcrisis recovery has not been subjected to a rigorous analysis with a view to drawing policy lessons from the exercise. The relevant literature is predominantly empirical. For example, Charoenseang and Manakit (2002); Claessens, Klingebiel, and Laeven (2001); and Koo and Kiser (2001) analyze the role of the private sector, the reform of corporate governance, and the importance of policy changes that can improve interaction between banks and corporations in crisisprone economies. Park and Lee (2001) come up with a comparative perspective by affirming that the post-1999 revival of East Asian countries was faster than could be predicted from previous episodes of crisis elsewhere.2 By contrast, Calvo (2003) initiates a distinct theoretical departure in financial crisis analysis by modeling the importance of fiscal status and institutions in the growth of emerging markets, and by highlighting the role of dysfunctional domestic policies and the resulting financial vulnerabilities in amplifying minor shocks into major turmoil. Christiano, Gust, and Roldos (2003) also provide a theoretical analysis of postcrisis recovery. They examine the effects of an interest rate cut in a postcrisis economy with collateral constraints.3 Our purpose is to explore the largely unexamined topic of the varying recovery rates of crisis-affected economies by designing a simple model that can be subjected to simulations. Evidently the precrisis health of countries’ macroeconomic fundamentals—among them balanced government budgets reflecting their fiscal status, high national saving rates, and strong export performance—can facilitate a quick recovery of investment potential and output growth in the postcrisis phase. We will compare the impact of varying these three precrisis macroeconomic fundamentals, one at a time, on the recovery paths of two contrasting performers, Thailand in East Asia and Argentina in Latin America. We develop an open-economy macroeconomic model that we then use for simulating the quarterly interest and exchange rates of Argentina for the period from the third quarter of 2000 to the third quarter of 2003. We refer to this approximation of the actual Argentine postcrisis profile as the base case. We then simulate another Argentine scenario, called case 1, by altering a single macroeconomic fundamental from the Argentine to the Thai value in the precrisis years, holding all other parameters and variables constant. The impact of changing a given macroeconomic fundamental for Argentina—say, the saving rate—is then assessed by comparing the simulated results of each case 1 with the actual values of four indicators for Thailand. More specifically, if Argentina were to be endowed with the
Why Do Some Countries Recover More Readily from Financial Crises?
219
high saving rate of Thailand in the precrisis years, how will its exchange rate decline, interest rate hike, inflation, and GDP growth rates in the recovery phase compare with the actual magnitudes in Thailand? We then undertake similar simulations by imposing the Thai fiscal status and export growth performance (precisely defined later) on precrisis Argentina for assessing its postcrisis outcomes with respect to the four indicators. We present, in section 2, the contrasting patterns of the three macroeconomic fundamentals of relevance to our analysis for Indonesia, Malaysia, South Korea, and Thailand in East Asia and Argentina and Brazil in Latin America for their respective precrisis years from 1985 to 2003. We then set out our theoretical model in section 3, discuss our simulation procedure in section 4, and provide our underlying data in section 5. Our simulation results are presented in section 6. We draw policy conclusions and suggest ideas for further research in section 7. 9.2
Contrasting Macroeconomic Fundamentals: East Asia and Latin America
The speed of recovery in our model will depend on the status of three macroeconomic fundamentals at the onset of financial turmoil. These are balanced or surplus government budgets, high overall saving in the economy, and solid export performance.4 The relatively healthy fiscal condition of the East Asian Four created the potential for them to pay back their external debt despite significant decline of their currencies, and to extend funds to their corporate sector in the midst of a financial crunch. Their governments were also better poised to extend unemployment benefits to the jobless during the downturn. By contrast, the shaky fiscal health of the Latin American Two had an opposite effect. They lacked the resources to repay their external debt, assist their private sector recovery, or extend unemployment benefits to the unemployed. The contrasting fiscal status of the two groups in the years prior to the emergence of the crisis—from 1991–1996 for the East Asian Four, and from 1991–2000 for the Latin American Two (1999 for Brazil)—is brought out in figures 9.1 and 9.2. Government budgets in the East Asian Four during 1991–1996 in figure 9.1 were in surplus in most years, whereas they were negative throughout the period 1991–2000 for Argentina, and hit 10 percent of GDP in 1999 for Brazil. Next, high saving rates have a mediating effect on interest rates during a crisis. High overall saving in the economy implies that more domestic funds may be available for lending, and thus for supporting investment even during the crisis phase. Consequently interest rates may shoot up less, and the contraction of investment and GDP in the postcrisis period may be moderated. Thus saving rates in the East Asian Four ranged from 25 to 35 percent or higher of GDP (figure 9.3),
220
Padma Desai and Pritha Mitra
Figure 9.1 Government Budget Balance of East Asian Countries. Source: Economist Intelligence Unit.
Figure 9.2 Government Budget Balance of Argentina and Brazil. Source: Economist Intelligence Unit. Data prior to 1994 for Argentina and prior to 1995 for Brazil are not available.
Why Do Some Countries Recover More Readily from Financial Crises?
221
Figure 9.3 Saving Rates of East Asia Countries. Source: Economist Intelligence Unit. Aggregate national saving by the public and private sectors as a percent of nominal GDP.
whereas the saving rate was stagnant at 15 percent for Argentina and declined to that level in 1999 for Brazil from a high of 20 percent in 1991 (figure 9.4). Finally, a strong export sector would provide the East Asian group with the capability to generate the much-needed foreign exchange in the postcrisis period. Although a few companies went bankrupt because of their inability to repay foreign debt, the survivors benefited from the increased foreign demand resulting from the lower exchange rate. Moreover, companies that were unable to meet their external financial obligations had highly developed manufacturing infrastructure. The depreciated exchange rates made them attractive pickings for foreign investors. These factors contributed to a rapid revival of foreign investor confidence in the East Asian Four. By contrast, the uneven and generally feeble performance of Argentine and Brazilian export sectors was inadequate to attract foreign creditors, in the process requiring significantly higher interest rates for the purpose. According to the contrasting export performance of the two sets of countries in figure 9.5, the precrisis average annual growth of Thai exports in dollars was three times that of Argentina in its precrisis years. The model, which we present in the next section, is designed to assess the impact of these three macroeconomic features on the interest and exchange rates, and therefore on the inflation and GDP growth rates, in the postcrisis recovery phase of Argentina in the context of our simulation design.
222
Padma Desai and Pritha Mitra
Figure 9.4 Saving Rates of Argentina and Brazil. Source: Economist Intelligence Unit. Aggregate national saving by the public and private sectors as a percent of nominal GDP.
Figure 9.5 Total Annual Exports (f.o.b.) in U.S. Dollars. Source: Economist Intelligence Unit.
Why Do Some Countries Recover More Readily from Financial Crises?
9.3
223
The Model
We begin with the classic Dornbusch (1976) exchange rate over-shooting framework with its three components of the asset market, the money market, and the goods market. The Dornbusch model, however, does not account for investor expectations, which are crucial in our framework. The Dornbusch model simply represents the expected depreciation in the exchange rate as a function of its deviation from its equilibrium value. The only shock in the model is a monetary one. Our innovation in the familiar Dornbusch model is precisely in the asset market component. Rather than relying on a monetary shock, we introduce exogenous shocks in the form of investor expectations that affect the economy as it revives from crisis. The expected depreciation of the exchange rate is reconstructed to reflect investor expectations, which, in turn, are a function of the three macroeconomic fundamentals listed earlier. We begin with the asset market. 9.3.1
The Asset Market
r t ¼ r þ xt
ð9:1Þ
xt ¼ yðet et Þ þ ðet þ eÞð1 þ FÞ
ð9:2Þ
where F represents the macroeconomic fundamental that affects investor expectations. It is defined as F ¼ expfwððtax gÞ þ z þ ð1 gÞÞg: The domestic interest rate, rt , in equation 9.1 is determined by the world interest rate, r , and the expected rate of exchange rate depreciation, xt . This expected rate of exchange rate depreciation has two components in equation 9.2. The first component is the difference between the long-run equilibrium exchange rate and the current exchange rate, where et is the logarithm of the long-run equilibrium exchange rate, et is the logarithm of the current exchange rate, and y is an adjustment factor. The second component of the expected exchange rate depreciation, our innovation in the Dornbusch model, is investor expectations. Investor expectations about the depreciation of the exchange rate are based on numerous factors such as the current political and economic climate of a country. For example, if some companies in an emerging market default on their external debt, investors will expect other investors to withdraw their funds from this market. This results in a selffulfilling speculation against the emerging market currency.
224
Padma Desai and Pritha Mitra
In equation 9.2 of our model, investor speculation against the currency is represented by et þ e, where et is a transitory component and e is a permanent component.5 The more confidence investors have in the macroeconomic fundamentals of the economy, the lower the effect of speculation on the expected depreciation. F represents the macroeconomic fundamentals that affect investor expectations in our model. F has three components, ðtax gÞ, z, and ð1 gÞ. •
ðtax gÞ represents the precrisis government budget balance.6
z represents the precrisis export sector strength. This value is represented by the annual percentage growth of real exports of goods and services relative to the annual percentage growth of real GDP. The growth rates are averaged over three years prior to the onset of the crisis.7,8
•
•
ð1 gÞ represents the precrisis national saving rates.9
9.3.2
The Money Market
m pt ¼ lrt þ fyt
ð9:3Þ
The money market equilibrium sets money demand—the right-hand side of equation 9.3—equal to money supply—the left-hand side of equation 9.3. Here m, pt , and yt are the natural logarithms of nominal money, the price level, and real GDP, respectively. 9.3.3
The Goods Market
yt ¼ u þ dðet pt Þ þ gyt þ ag btax srt
ð9:4Þ
p_ ¼ pðyt yt Þ
ð9:5Þ
The goods market equilibrium sets aggregate demand (the right-hand side of equation 9.4) equal to aggregate supply (the left-hand side of equation 9.4). Here aggregate demand is represented by a shift factor, u (which includes the exogenous and constant demand for exports10); the relative price of domestic goods and services, et pt ; income, yt ; government expenditures relative to revenue, ag btax; and interest rates, rt . The consumers’ marginal propensity to consume in the aggregate demand equation, g, determines the saving rate, 1 g. In equation 9.5, the rate of increase in prices, p_ , is proportional to the deviation of GDP, yt , from potential GDP, yt .
Why Do Some Countries Recover More Readily from Financial Crises?
9.3.4
225
Equilibrium
In equilibrium, the interest rate, GDP, and exchange rate must adjust such that the asset, money, and goods markets are simultaneously in equilibrium. The resulting equilibrium conditions are: yt yt ¼ oð pt pt Þ
ð9:6Þ
et et ¼ ½ð1 fmdÞ=Dð pt pt Þ
ð9:7Þ
p_ ¼ poð pt pt Þ
ð9:8Þ
where o¼
½mðd þ ysÞ þ mdyl D
m¼
1 ð1 gÞ
D ¼ fmðd þ ysÞ þ yl Following rational expectations, the rate at which the current exchange rate adjusts to the long-run equilibrium exchange rate is equal to the rate at which the exchange rates actually adjust. That is, y ¼ po. 9.4
Model Simulations
Initially, we assume that the economy in our model is in equilibrium and investor speculation against the currency is absent. That is, e0 ¼ e0 ¼ 0. Then the economy is hit with an external shock such that investors now speculate against the currency. That is, et > 0, e > 0. The speculation of investors initially affects the expected exchange rate depreciation, which in turn affects the interest rate. The goods and money markets are affected through the interest rate. The exchange rate, GDP, and prices then adjust accordingly. As time passes, investors speculate less against the currency as the currency adjusts to its new long-run value. Consequently, the speculation of investors against the currency is modeled as an exponentially decreasing shock. et ¼ expfktg, k < 0, where t represents time.11 The precrisis values of the macroeconomic fundamentals affecting investor expectations, F, either moderate or amplify the effects of investor speculation. When the macroeconomic fundamental variable is strong, investors have more
226
Padma Desai and Pritha Mitra
confidence in the economy, and thus F acts to reduce the effects of speculation. The opposite occurs when the macroeconomic fundamental is weak. As already noted, the purpose of our exercise is to assess the impact of precrisis macroeconomic fundamentals on an economy experiencing an exogenous shock that induces investors to speculate against the currency. Having selected a specific macroeconomic fundamental—for example, the government’s precrisis fiscal status—we use our model to simulate the Argentine economy in its recovery phase. In the first step, the model simulation is calibrated to approximately match the actual interest rate and exchange rate profiles of Argentina. We refer to this simulation as the base case.12 Next, the Argentine government’s precrisis fiscal status is changed to that of Thailand. All other variables and parameters are kept at their base case values. We apply the same shock as the one in the base case. In other words, referring to equation 9.2, et and e remain the same as in the base case. This simulated model is referred to as case 1. Finally, the impact of the precrisis government fiscal status is assessed by comparing the case 1 simulation results of the interest rate, the exchange rate, the GDP growth rate, and inflation with the actual values for Thailand. If the case 1 simulated Argentine values are a reasonable match to the actual values for Thailand, then we can suggest that had the precrisis fiscal status of Argentina been more like that of Thailand, Argentina’s postcrisis interest rate, exchange rate, GDP growth rate, and inflation would have been closer to those of Thailand in the postcrisis period. We repeat these steps in two additional simulations by bringing in export sector strength and saving rate in place of government fiscal status. Thus, in the second simulation, we replace the Argentine precrisis export sector strength with that of Thailand. In the third and final simulation, the Argentine precrisis saving rate is replaced with that of Thailand. 9.5
The Data
We apply three sets of data in our model: the actual data series for Argentina and Thailand, parameter values, and values of macroeconomic fundamentals. The actual data series includes quarterly data of interest rates, exchange rates, GDP, and price levels for each country. Details of these data are presented in table 9.1. We adopt previously estimated parametric values for our money-market and goods-market parameters. These estimates and their sources are reported in table 9.2. Some of the parameters are calibrated such that the simulated time series of interest rates and exchange rates matches the actual data for Argentina and Thailand. For example, ut , which represents a shift factor in the aggregate demand
Why Do Some Countries Recover More Readily from Financial Crises?
227
Table 9.1 Actual Series Series
Symbol
Year(s)
Y p_
1997,QI–2003,QI
Lending interest rate (%) Consumer prices (% average change per annum)
r
1997,QI–2003,QI
Exchange rate baht:US$ (period average)
E
1997,QI–2003,QI
Y p_
2000,QIII–2003,QIII
Lending interest rate (%) Consumer prices (% average change per annum)
r
2000,QIII–2003,QIII
Exchange rate peso:US$ (period average)
E
2000,QIII–2003,QIII
Thailand Real GDP
1997,QI–2003,QI
Argentina Real GDP
2000,QIII–2003,QIII
Note: Source for all is economist intelligence, unit country data. Frequency for all is quarterly.
equation, is adjusted such that the simulated base case time series matches the actual time series for Argentina. Finally, the values of the precrisis macroeconomic fundamentals of government expenditures and revenues, export sector strength, and saving rates, and their sources for each country, are presented in table 9.3. The data in table 9.4 bring out the striking contrast between the crisis impact on the economies of Argentina and of Thailand. At its peak, the Argentine interest rate, at 90.61 percent in 2002, QIII (table 9.4, row 3, column 2), was almost six times its level of 15.25 percent in Thailand in 1998, QI and QII (table 9.4, row 3, column 3). The high Argentine interest rate reflects the abysmally low investor confidence in its recovery prospects. The difference in crisis severity is also reflected in the exchange rate depreciation. Prior to the crisis, Argentine and Thai exchange rates were fixed to the dollar, the former more strictly than the latter. When they were allowed to float in the postcrisis phase, the peso’s maximum plunge was 258.47 percent in 2002, QIII (table 9.4, row 6, column 2) in contrast to the baht’s maximum decline of 82.10 percent in 1998, QI (table 9.4, row 6, column 3), almost three times less. The postcrisis GDP growth rates of Argentina and Thailand are also reported in table 9.4. The two growth rates are similar in their pattern and magnitude although, at its lowest, Argentina’s GDP growth rate in 2002, QI is slightly more negative at 15.23 percent (table 9.4, row 9, column 2) than Thailand’s in 1998, QIII at 13.92 percent (table 9.4, row 9, column 3). Argentina’s GDP growth rate, however, continues to be negative for a longer stretch after the crisis, representing a more painful recovery.
228
Padma Desai and Pritha Mitra
Table 9.2 Parameter Estimates Parameter
Value
Goods market d 0.11
Interpretation
Source of Estimate
Sensitivity of aggregate demand to relative prices of domestic goods and services Proportion of income spent on purchases of goods and services (approximately the inverse of the saving rate) Proportion of income spent on purchases of goods and services (approximately the inverse of the saving rate)
Ghosh and Masson (1991) World Development Indicators
gArgentina
0.85
gThailand
0.7
s
0.15
Sensitivity of aggregate demand to interest rates
a
0.3
Sensitivity of aggregate demand to government expenditures
Calibrated*
b
0.2
Sensitivity of aggregate demand to taxes
Calibrated*
p
0.2
Proportion representing the constant relationship between changes in output and changes in prices
Calibrated*
u yArgentina
1.472 69 billion pesos
Shift parameter Initial long-run equilibrium output
Calibrated* Economist Intelligence Unit Country Data, Quarterly Real GDP 2000 average
yThailand
779 billion bahts
Initial long-run equilibrium output
Economist Intelligence Unit Country Data, Quarterly Real GDP 1996 average Chowdhury (1997), Dekle and Pradhan (1999) Dekle and Pradhan (1999) Economist Intelligence Unit Country Data, M2 Money Supply, 1996,QIV
World Development Indicators, Baharumshah, Thanoon, and Rashid (2003) Ghosh and Masson (1991)
Money market j
1
Sensitivity of real money demand to real income
l
0.04
MThailand
3726 billion bahts
Sensitivity of real money demand to interest rates Money supply
MArgentina
91 billion pesos
Money supply
Economist Intelligence Unit Country Data, M2 Money Supply, 2000,QIV
8%
World interest rate
World Development Indicators, approximated by the average of 1995– 2000 US prime rate
Asset market r*
Why Do Some Countries Recover More Readily from Financial Crises?
229
Table 9.2 (continued) Parameter
Value
Interpretation
Source of Estimate
y
0.2447
Factor for adjustment of current exchange rate to long-run exchange rate value
Endogenously Determined**
w
0.033
Sensitivity of investor speculation to precrisis macroeconomic fundamental
Calibrated*
Shock propagation factor Permanent component of shock Initial value of transitory component of shock
Calibrated* Calibrated* Calibrated*
Shock equation k e e0
0.07 0.01 0.178
* These parameter values are adjusted so that the simulated Argentine time series match the actual Argentine time series. ** This parameter value is determined endogenously within the model (refer to section 3).
Finally, inflation soared higher in Argentina than in Thailand in the postcrisis phase. At its height in 2002, QIV, the rate was 40.31 percent (table 9.4, row 13, column 2), in contrast to Thailand’s 10.35 percent in 1998, QII (row 13, column 3). The combination of high inflation, exchange rate depreciation, and soaring interest rate, along with the persistence of a negative GDP growth rate, made Argentina’s postcrisis recovery more arduous than Thailand’s. 9.6
Simulation Results
In our simulations, the adoption of Thailand’s precrisis macroeconomic fundamentals for Argentina should result in an easier recovery path for Argentina, resembling that of Thailand in terms of interest rate, exchange rate depreciation, GDP growth, and inflation. Recall that our base case simulates the Argentine economy from 2000, QIII to 2003, QIII. Each version of the model alters the base case simulation by changing the value of a precrisis macroeconomic fundamental from its Argentine value to its Thai value. The resulting simulation is compared to the actual performance of the recovering Thai economy. A closer match brings out the effectiveness of the particular precrisis macroeconomic fundamental. The simulation results are summarized in tables 9.5–9.8, one each for interest rate, exchange rate depreciation, GDP growth rate, and inflation rate, and presented in figures 9.6–9.9. We notice similar patterns in these variables across the simulations: interest rates shoot up, exchange rates decline, GDP growth rates tumble, and inflation rates go up after the shock and subsequently adjust to take their equilibrium values. However, the simulation version with the precrisis export sector strength gives significantly lower values of interest rate, exchange rate
Table 9.3 Macroeconomic Fundamentals Variable
Value
Interpretation
Source of estimate
GArgentina
15 billion pesos
Government expenditures, quarterly data
Economist Intelligence Unit Country Data, Quarterly Real Government Expenditures, 1999 average
GThailand
137 billion bahts
Government expenditures, quarterly data
Economist Intelligence Unit Country Data, Quarterly Real Government Expenditures, 1996 average
TArgentina
14 billion pesos
Government revenues, quarterly data
Economist Intelligence Unit Country Data, Quarterly Real Government Expenditures, 1999 average
TThailand
142 billion bahts
Government revenues, quarterly data
Economist Intelligence Unit Country Data, Quarterly Real Government Expenditures, 1996 average
Version One: ZArgentina
36.55
World Development Indicators
Version One: ZThailand
1.00
4.02/0.11—Annual percentage growth of real exports of goods and services/ annual percentage growth of real GDP. Averaged over 3 years prior to the crisis (1998, 1999, 2000)*. 8.06/8.04—Annual percentage growth of real exports of goods and services/annual percentage growth of real GDP. Averaged over 3 years prior to the crisis (1994, 1995, 1996).
Version Two: ZArgentina
1.50
1/Total debt service—Exports of goods and services/Total debt payment. Averaged over 3 years prior to the crisis (1998, 1999, 2000)*.
World Development Indicators
Version Two: ZThailand
8.00
1/Total debt service—Exports of goods and services/Total debt payment. Averaged over 3 years prior to the crisis (1994, 1995, 1996).
World Development Indicators
1 gArgentina
0.15
Saving rate ¼ 1 Proportion of income spent on purchases of goods and services (1999)
World Development Indicators
1 gThailand
0.30
Saving rate ¼ 1 Proportion of income spent on purchases of goods and services (1996)
World Development Indicators, Baharumshah, Thanoon, and Rashid (2003)
World Development Indicators
* The year 2000 is considered a year prior to the crisis since the crisis began at the end of the year 2000.
Why Do Some Countries Recover More Readily from Financial Crises?
231
Table 9.4 Actual Rates of Interest, Exchange Rate Depreciation, GDP Growth Rate, and Inflation Rates for Argentina and Thailand 1
2 Actual value for Argentina %
3 Actual value for Thailand %
4 Difference %
1
Variable*
2 3
Precrisis interest rate Maximum postcrisis interest rate
9.66 90.61
13.00 15.25
3.34 75.36
4
Equilibrium postcrisis interest rate
13.90
6.50
7.40
5
Precrisis exchange rate depreciation
0.00
2.40
2.40
6
Maximum postcrisis depreciation
258.47
82.10
176.37
7
Equilibrium postcrisis depreciation
20.91
2.16
18.75
8
Precrisis GDP growth rate
9
Maximum negative postcrisis GDP growth rate
0.33
1.00
1.33
15.23
13.92
1.31
10 Maximum positive postcrisis GDP growth rate
8.62
8.41
0.21
11 Equilibrium postcrisis GDP growth rate
8.62
6.73
1.89
12 Precrisis inflation rate
0.78
4.37
5.15
13 Maximum postcrisis inflation rate
40.31
10.35
29.96
5.20
1.97
3.24
14 Equilibrium postcrisis inflation rate
* All variables are measured on a quarterly basis. 1. The precrisis rates of all variables refer to their values in the quarter preceding 2000,QIV for Argentina and 1997,QII for Thailand. 2. The maximum postcrisis rates represent the highest values of the variables in the postcrisis period. 3. The equilibrium postcrisis rates represent their final and stable values in the postcrisis period after all factors have fully adjusted in the simulation. The actual equilibrium values refer to 2003,QIII for Argentina (this was the most recent value available at the time this analysis was done) and 2003,QI for Thailand. 4. The Argentine peso was allowed to float in 2002,QI. The baht was allowed to float in 1997,QIII.
depreciation, GDP growth rate decline, and inflation rate than the alternatives. This version outperforms the simulations in which the precrisis macroeconomic fundamentals represent the government’s fiscal status and the economy’s saving rate. We now discuss the details of our simulations, beginning with interest rates. 9.6.1
Interest Rates
In figure 9.6, the base case simulation tracks the actual Argentine interest rate relatively well. The maximum postcrisis interest rate for Argentina was 90.61 percent (table 9.5, row 2, column 3). The base case simulation, emulating the Argentine economy, attains a similar maximum postcrisis interest rate of 90.37 percent (table 9.5, row 5, column 3). In contrast, the actual Thai maximum postcrisis interest rate, 15.25 percent (table 9.5, row 4, column 3), was 75.36 basis points lower than the corresponding Argentine rate.
Figure 9.6 Interest Rate Response to Shock. Source: The actual interest rates for Thailand and Argentina are lending rates put together from the Economist Intelligence Unit (EIU). EIU Source: IMP, International Financial Statistics. Notes: 1. The financial crisis began in Argentina in the fourth quarter of 2000 (2000, QIV) and hit Thailand in the second quarter of 1997 (1997, QII). In the figure, the quarter preceding the shock (Q0) corresponds to 2000, QIII for Argentina, and 1997, QI for Thailand. 2. The interest rate paths generated by the three simulations of the base case, of Argentina with Thai fiscal status, and of Argentina with Thai saving rate almost perfectly overlap, creating only one visible solid line in the figure. Recall that the base case represents Argentine values generated by our model. This overlap reflects the weakness of the precrisis fiscal status and saving rate in affecting the postcrisis interest rate. In other words, even if Argentina had the precrisis fiscal status or saving rate of Thailand, the postcrisis interest rate of Argentina would match that in the base case in which Argentina retains its own precrisis fiscal status and saving rate. 3. The interest rate path generated by the simulation of Argentina with Thai export strength produces postcrisis interest rates that are significantly lower than those resulting from the base case simulation. In the context of our model, this result implies that the postcrisis interest rates in Argentina would have been much lower, closer to the postcrisis interest rates of Thailand, if Argentina had the precrisis export sector strength of Thailand. 4. We undertake the following steps for deriving the interest rates in the figure. Solving the differential equation for prices (equation 9.8), we obtain the path of prices in the quarters after the shock. Substituting prices in equations 9.6 and 9.7, we solve for GDP and the exchange rate, respectively. Substituting the exchange rate in equations 9.1 and 9.2, we obtain the interest rates. Details of the calculations for solving each series are in the appendix. The solutions are generated in Matlab.
Why Do Some Countries Recover More Readily from Financial Crises?
233
Table 9.5 Interest Rate Analysis 1
2
Variable*
Precrisis interest rate %
3 Maximum postcrisis interest rate %
4 Equilibrium postcrisis interest rate %
1 2 3
Actual value for Argentina Actual value for Thailand
9.66 13.00
90.61 15.25
13.90 6.50
3.34
75.36
7.40
8.00
90.37
12.38
4
Difference
5
Base case**
6
Scenario 1—Replace Argentine fiscal status with that of Thailand
7
Case 1***
8.00
90.15
12.37
8
Difference ¼ base case case 1
0.00
0.22
0.01
9
Scenario 2—Replace Argentine export sector strength with that of Thailand
10 Case 1***
8.00
44.94
9.97
11 Difference ¼ base case case 1
0.00
45.43
2.41
12 Scenario 3—Replace Argentine saving rate with that of Thailand 13 Case 1***
8.00
90.06
12.37
14 Difference ¼ base case case 1
0.00
0.31
0.01
* All variables are measured on a quarterly basis. **B ase case values refer to simulated Argentine values. *** Case1 values are simulated Argentine values which result when its precrisis macroeconomic fundamental is replaced with a precrisis Thai value. 1. The precrisis rates prevailed in the quarter preceding 2000,QIV for Argentina and 1997,QII for Thailand. 2. The maximum postcrisis rates represent their highest values in the postcrisis period. 3. The equilibrium postcrisis rates represent their final and stable values in the postcrisis period after all factors have fully adjusted in the simulation. These actual values are 2003,QIII value for Argentina (this was the most recent value available at the time this analysis was done) and 2003,QI value for Thailand. 4. GDP growth simulations assume a relatively stable GDP level with a zero GDP growth rate for the quarter prior to crisis onset. 5. The Argentine peso was allowed to float in 2002,QI, the baht in 1997,QIII.
When Argentina’s precrisis fiscal status is changed to that of Thailand, the maximum Argentine postcrisis interest rate drops by only 0.22 basis points (table 9.5, row 8, column 3). By contrast, if the precrisis government fiscal status played a critical role in postcrisis recovery, the maximum postcrisis interest rate in the simulation would have dropped by a number closer to 75.36 basis points (the difference between the actual Argentine and Thai interest rates, table 9.5, row 4, column 3). Despite a robust precrisis fiscal status closer to Thailand’s, the crisis would have landed Argentina in exorbitant interest rates. However, with the imposition of the Thai export sector strength on precrisis Argentina, the maximum postcrisis interest rate in Argentina is reduced by 45.43 basis points (table 9.5, row 11, column 3). In figure 9.6, this version brings the
234
Padma Desai and Pritha Mitra
Table 9.6 Exchange Rate Depreciation Analysis
1
1
2
Variable*
Precrisis depreciation %
3 Maximum postcrisis depreciation of exchange rate %
4 Equilibrium postcrisis depreciation of exchange rate % 20.91
2
Actual value for Argentina
0.00
258.47
3
Actual value for Thailand
2.40
82.10
2.16
4
Difference
2.40
176.37
18.75
5
Base case**
0.00
216.71
0.00
6
Scenario 1—Replace Argentine fiscal status with that of Thailand
7
Case 1***
0.00
215.73
0.00
8
Difference ¼ base case case 1
0.00
0.98
0.00
9
Scenario 2—Replace Argentine export sector strength with that of Thailand
10 Case 1***
0.00
67.83
0.00
11 Difference ¼ base case case 1
0.00
148.88
0.00
12 Scenario 3—Replace Argentine saving rate with that of Thailand 13 Case 1***
0.00
211.32
0.00
14 Difference ¼ base case case 1
0.00
5.39
0.00
Notes: See notes for table 9.5.
simulated value of the Argentine interest rate remarkably close to the actual Thai value, in contrast to the simulation versions employing Thailand’s precrisis fiscal status and saving rate (to be analyzed later). If Argentina had the precrisis export sector strength of Thailand, its interest rate would not have soared to 90 percent. It would have capped around a more manageable 45 percent. Finally, the augmentation of the precrisis saving rate of Argentina to that of Thailand pulls down the maximum postcrisis interest rate of the simulated Argentine economy by a minuscule 0.31 basis points (table 9.5, row 14, column 3). The postcrisis interest rate in Argentina turns out to be insensitive to the precrisis saving rate. 9.6.2
Exchange Rate Depreciation
The maximum depreciation of the actual Argentine exchange rate was 258.47 percent in 2002, QIII (table 9.6, row 2, column 3). The exchange rate in our base case simulation attains a maximum depreciation of 216.71 percent (table 9.6, row 5, column 3). By contrast, the maximum postcrisis depreciation of the Thai baht was
Why Do Some Countries Recover More Readily from Financial Crises?
235
Figure 9.7 Depreciation of Exchange Rates in Response to Shock. Source: Actual baht and peso exchange rates are from Economist Intelligence Unity (EIU). EIU Source: IMF, International Financial Statistics. Notes: 1. In the years preceding the onset of the crisis and for some time after that, the Argentine peso and the Thai baht were pegged to the U.S. dollar, the former under a regime resembling a currency board, and the latter under a managed exchange rate arrangement. The peso was allowed to float in 2002, QI, and the baht in 1997, QIII. In the figure, Q0 corresponds to the quarter preceding the currency float 2001, QIV for Argentina and 1997, QII for Thailand. 2. The depreciation path generated by the simulations of the base case and of Argentina with Thai fiscal status almost perfectly overlap, creating only one visible solid line in the figure. This result reflects the weakness of the fiscal status in affecting postcrisis exchange rate depreciation. Even if Argentina had the precrisis robust fiscal status of Thailand, the postcrisis exchange rate depreciation of Argentina would remain the same as in the base case in which Argentina retains its own precrisis weak fiscal status. 3. The exchange rate depreciation path generated by the simulation of Argentina with Thai saving rate is only slightly lower than the base case simulation path. In other words, even if Argentina had the precrisis high saving rate of Thailand, the postcrisis exchange rate depreciation of the peso would be minimally reduced. 4. The depreciation path generated by the simulation of Argentina with Thai export sector strength is significantly lower than the base case simulation path and the paths generated by the simulations of Argentina with Thai fiscal status and Argentina with Thai saving rate. This result reflects the strength of the precrisis export sector in positively affecting postcrisis depreciation. 5. The simulation steps are stated in note 4 of figure 9.6.
236
Padma Desai and Pritha Mitra
relatively modest at 82.10 percent (table 9.6, row 3, column 3), 176.3 percent lower than that of the peso. Figure 9.7 brings out this striking contrast. As with the interest rate, the peso exchange rate is relatively insensitive to the Argentine government’s precrisis fiscal status. When it is replaced by the precrisis fiscal status of Thailand, the maximum depreciation of the peso is reduced by only 0.98 percent (table 9.6, row 8, column 3). If the government’s precrisis fiscal status were a critical factor in affecting its level, the massive actual decline of the peso, postshock, would have been reduced substantially by 176.37 percent, reaching the baht’s postcrisis maximum decline of 82.10 percent (table 9.6, row 4, column 3). By contrast, a change in the precrisis export sector strength of Argentina to that of Thailand significantly reduces the peso’s depreciation in the postcrisis period in figure 9.7. The maximum depreciation is reduced by as much as 148.88 percent (table 9.6, row 11, column 3), closer to the 176.37 percent difference between the actual peso–baht maximum tumble (table 9.6, row 4, column 3). Finally, the impact of the precrisis saving rate on the postcrisis peso exchange rate decline is similar to the impact of the precrisis fiscal status of the government: the peso’s depreciation is insensitive to the precrisis saving rate in the simulation. The maximum postcrisis depreciation of the peso is contained by only 5.39 percent (table 9.6, row 14, column 3). The simulation results suggest substantially moderate postshock interest and exchange rate movements, associated with a strong precrisis export sector strength. As a result, Argentina’s recovery path could possibly have been less prolonged and painful. 9.6.3
GDP Growth Rates
As noted earlier, our base case simulations understate the magnitudes of Argentina’s GDP growth and inflation rates while tracking their trends. In figure 9.8, during the postcrisis period, actual GDP in Argentina and Thailand experiences a significant decline, with negative growth rates, in the initial postcrisis quarters. As the economies recover, both GDPs move up. Thus the actual GDP growth rates of the two economies follow a similar path. The maximum negative postcrisis actual GDP growth of Argentina is only 1.31 percent lower than that of Thailand (table 9.7, row 4, column 3). Our simulations suggest that if Argentina had the precrisis fiscal status of Thailand, its maximum negative GDP growth would have matched it in the base case simulation in which Argentina retains its own precrisis fiscal status (table 9.7, row 8, column 3). Essentially, the precrisis fiscal status has no effect on Argentine GDP growth as it recovers. Similarly, the postcrisis Argentine economy is insensitive to
Why Do Some Countries Recover More Readily from Financial Crises?
237
Table 9.7 GDP Growth Rate Analysis
1
1
2
3 Maximum negative postcrisis GDP growth %
4 Maximum positive postcrisis GDP growth %
Variable*
Precrisis GDP growth %
Equilibrium postcrisis GDP growth %
0.33
15.23
8.62
8.62
1.00
13.92
8.41
6.73
1.33
1.31
0.21
1.89
0.00
1.25
0.47
0.00
2
Actual value for Argentina
3
Actual value for Thailand
4
Difference
5
Base case**
6
Scenario 1—Replace Argentine fiscal status with that of Thailand
5
7
Case 1***
0.00
1.25
0.47
0.00
8
Difference ¼ Base case case 1
0.00
0.00
0.00
0.00
9
Scenario 2—Replace Argentine export sector strength with that of Thailand
10 Case 1***
0.00
1.06
0.43
0.00
11 Difference ¼ Base case case 1
0.00
0.19
0.04
0.00
12 Scenario 3—Replace Argentine saving rate with that of Thailand 13 Case 1***
0.00
1.25
0.47
0.00
14 Difference ¼ Base case case 1
0.00
0.00
0.00
0.00
Notes: See notes for table 9.5.
the precrisis saving rate. Changing the precrisis saving rate of Argentina to that of Thailand results in zero change in the postcrisis GDP growth path (table 9.7, row 14, column 3, also figure 9.8). However, when we replace the precrisis export sector strength of Argentina with that of Thailand in our simulated Argentine economy, Argentina’s maximum negative GDP growth rate is pulled up by 0.19 percent (table 9.7, row 11, column 3). The actual maximum negative postcrisis GDP growth of Argentina is 1.31 lower than in Thailand (table 9.7, row 4, column 3). In figure 9.8, the attribution of the precrisis export sector strength of Thailand to Argentina is most effective in reducing fluctuations in GDP growth rate in recovering Argentina. 9.6.4
Inflation
The maximum postcrisis inflation rate in Thailand is 29.96 percent lower than in Argentina (table 9.8, row 4, column 3). As before, the potentially large impact of a precrisis macroeconomic fundamental in mediating investor expectations, and thus reducing the severity of the crisis, would result in a significant reduction of
238
Padma Desai and Pritha Mitra
Figure 9.8 GDP Growth Rates in Response to Shock. Source: Actual GDP growth rates for Argentina from 2000, QIII to 2003, QIII are from Economist Intelligence Unity (EIU). EIU data are taken from Ministerio de Economia y Obras Y Servicios Publico. Actual GDP Growth Rates for Thailand from 1997, QI to 2003, QI are from Economist Intelligence Unity (EIU). EIU data are taken from the National Economic and Social Development Board of Thailand. Notes: 1. The financial crisis hit Argentina in 2000, QIV and Thailand in 1997, QII. In the figure, the quarter preceding the shock, Q0 corresponds to 2000, QIII for Argentina, and 1997, QI for Thailand. 2. The GDP growth rate paths generated by the simulations of the base case, of Argentina with Thai fiscal status, and of Argentina with Thai saving rate almost perfectly overlap, creating only one visible solid line in the figure. This result reflects the weakness of the fiscal status and saving rate in affecting the postcrisis GDP growth rate. Even if Argentina had the robust precrisis fiscal status or saving rate of Thailand, the postcrisis growth rate of Argentina would remain the same as in the base case, in which Argentina retains its own precrisis fiscal status and saving rate. 3. The GDP growth path generated by the simulation of Argentina with Thai export sector strength fluctuates less than the GDP growth path simulated by the base case. In the context of our model, this may be interpreted to mean that Argentina’s postcrisis GDP growth rate would have fluctuated less if its precrisis export sector were as strong as that of Thailand. 4. The simulation steps are stated in note 4 of figure 9.6.
Why Do Some Countries Recover More Readily from Financial Crises?
239
Table 9.8 Inflation Rate Analysis 1
2
Variable*
Precrisis inflation rate %
3 Maximum postcrisis inflation rate %
4 Equilibrium postcrisis inflation rate %
1 2 3
Actual value for Argentina Actual value for Thailand
0.78 4.37
40.31 10.35
5.20 1.97
5.15
29.96
3.24
0.00
4.32
0.00
4
Difference
5
Base case**
6
Scenario 1—Replace Argentine fiscal status with that of Thailand
7
Case 1***
0.00
4.32
0.00
8
Difference ¼ Base case case 1
0.00
0.01
0.00
9
Scenario 2—Replace Argentine export sector strength with that of Thailand
10 Case 1***
0.00
2.44
0.00
11 Difference ¼ Base case case 1
0.00
1.88
0.00
12 Scenario 3—Replace Argentine saving rate with that of Thailand 13 Case 1***
0.00
4.31
0.00
14 Difference ¼ Base case case 1
0.00
0.01
0.00
Notes: See notes for table 9.5.
the inflation rate in recovering Argentina if its precrisis macroeconomic fundamentals were to assume Thai values. In our simulations, Argentina’s assumption of Thai fiscal status and saving rate lowers the maximum postcrisis inflation rate of Argentina by a mere 0.01 percent in both cases (table 9.8, row 8, column 3 and row 14, column 3). The Argentine inflation rate is insensitive to its precrisis fiscal status and saving rate. By contrast, Argentina’s maximum inflation rate in the recovery phase drops by 1.88 percent when it takes the precrisis export sector strength of Thailand (table 9.8, row 11, column 3, also figure 9.9). 9.7
Conclusions
We analyze the postcrisis performance of two crisis-affected economies, Argentina and Thailand, with vastly different precrisis macroeconomic fundamentals, by applying a simple macroeconomic model. Their precrisis macroeconomic fundamentals, among them their governments’ fiscal status, their national saving rates, and their export sector strength, can be expected to affect postcrisis recovery differently. In fact, our model simulations suggest that the precrisis difference in
240
Padma Desai and Pritha Mitra
Figure 9.9 Inflation Rates in Response to Shock. Source: Actual GDP growth rates for Argentina from 2000, QIII to 2003, QIII are from Economist Intelligence Unity (EIU). EIU data are taken from Instituto Nacional de Estadistica y Censo. Actual GDP growth rates for Thailand from 1997, QI to 2003, QI are from Economist Intelligence Unity (EIU). EIU data are taken from the IMF: International Financial Statistics. Notes: 1. The financial crisis hit Argentina in 2000, QIV and Thailand in 1997, QII. In the figure, the quarter preceding the shock, Q0, corresponds to 2000, QIII for Argentina and 1997, QI for Thailand. 2. The inflation paths generated by the simulations of the base case, of Argentina with Thai fiscal status, and of Argentina with Thai saving rate almost perfectly overlap, creating only one visible solid line in the figure. This result reflects the weaknesses of the fiscal status and the saving rate in affecting the postcrisis inflation rate. Even if Argentina had the robust precrisis fiscal status or saving rate of Thailand, the postcrisis inflation rate of Argentina would remain the same as in the base case, in which Argentina retains its own weak, precrisis fiscal status and saving rate. 3. The inflation path generated by the simulation of Argentina with Thai export strength is notably lower than the inflation path generated by the base case simulation. In other words, if Argentina had the precrisis export strength of Thailand, its postcrisis inflation rates would have been significantly lower. 4. The simulation steps are stated in note 4 of figure 9.6.
Why Do Some Countries Recover More Readily from Financial Crises?
241
export sector strength between Argentina and Thailand (acting through investor expectations) is large enough to explain most of the difference in postcrisis interest rate and exchange rate movements between the two countries. The GDP growth and inflation rate results are quantitatively less strong; however, the precrisis difference in export sector strength is able to explain some of the postcrisis difference in these two variables as well. In other words, if Argentina had the precrisis export sector strength of Thailand, Argentina’s postcrisis recovery path could have been closer to that of Thailand. On the other hand, a strong fiscal status and a high national saving rate in the precrisis years could not have spared the Argentine economy from high interest rates and substantial peso depreciation, leading in turn to substantial GDP decline and high inflation. The export earning capacity of an economy reflects its debt repayment potential by generating foreign exchange earnings and restoring investor confidence. The contrasting results suggest that investors focus essentially on an economy’s ability to generate foreign exchange and repay its external debts. However, our model lacks microeconomic foundations that could adequately capture the interaction among consumers, producers, foreign investors, and the government. We plan to design and empirically test such a model with appropriate microeconomic underpinnings. Despite this shortcoming, we believe that our simulation results linking superior postcrisis recovery to precrisis export strength are eminently credible. Our analysis examines the impact of actual, ex post, precrisis macroeconomic fundamentals on postcrisis recovery. We are aware that a variety of other factors played a role in the Thai and Argentine recoveries. One factor that future analysts may want to introduce in the model is the expectation of postcrisis macroeconomic and financial fundamentals both during and just prior to the crisis. Note that these expectations tend to be related to the precrisis macroeconomic fundamentals. In this sense, our analysis indirectly accounts for them by explicitly including the macroeconomic impact of the precrisis fundamentals in our model. On the other hand, expectations of future macroeconomic and financial fundamentals will influence forward-looking expectations of servicing sovereign debt, which in turn will affect the country risk premium, leading to higher domestic interest rates. From this perspective, a more detailed model will require the introduction of the country risk premium. 9.8
Appendix: Derivation of Model
The model is defined by the following five equations: Asset Market r t ¼ r þ xt
ð9:A1Þ
242
Padma Desai and Pritha Mitra
xt ¼ yðet et Þ þ ðet þ eÞð1 þ FÞ
ð9:A2Þ
f ¼ expfwððtax gÞ þ z þ ð1 gÞÞg Money Market m pt ¼ lrt þ fyt
ð9:A3Þ
The Goods Market yt ¼ u þ dðet pt Þ þ gyt þ ag btax srt
ð9:A4Þ
p_ ¼ pðyt yt Þ
ð9:A5Þ
We take as given (that is, exogenous) values for all parameters and m, r , g, tax, ðet þ eÞ, and yt . Step 1: Goods and Asset Market Equilibrium For a given shock, ðet þ eÞ, the economy converges to equilibrium where y t ¼ yt ;
et ¼ et ;
pt ¼ pt ) r ¼ r þ ðe þ eÞð1 þ FÞ:
Substituting equations 9.A1 and 9.A2 into equation 9.A4, we get the equilibrium value for yt : yt ¼ m½u þ dðet pt Þ þ ag btax sr sðet þ eÞð1 þ FÞ m¼
ð9:A6Þ
1 : 1g
Subtracting equation 9.A6 from 9.A4, we get yt yt ¼ m½ðd þ syÞðet et Þ mdð pt pt Þ:
ð9:A7Þ
Step 2: Money and Asset Market Equilibrium For a given shock, ðet þ eÞ, the economy converges to equilibrium where y t ¼ yt ;
et ¼ et ;
pt ¼ pt ) r ¼ r þ ðe þ eÞð1 þ FÞ:
Substituting equations 9.A1 and 9.A2 into equation 9.A3, we get pt ¼ m þ lr þ lyðet et Þ þ lðet þ eÞð1 þ FÞ fyt :
ð9:A8Þ
The equilibrium value for pt in 9.A8 is given by pt ¼ m þ lr þ lðet þ eÞð1 þ FÞ fyt :
ð9:A9Þ
Subtracting equation 9.A9 from 9.A8, we get pt pt ¼ lyðet et Þ fð yt yt Þ:
ð9:A10Þ
Why Do Some Countries Recover More Readily from Financial Crises?
243
Step 3: Combining Goods, Asset, and Money Market Equilibrium Substituting ðet et Þ from equation 9.A10 into equation 9.A7, we derive ð yt yt Þ ¼ oðpt pt Þ
ð9:A11Þ
o ¼ ½mðd þ ysÞ þ mdyl=D D ¼ fmðd þ ysÞ þ yl: Rational expectations dictate that y ¼ po: Substituting equation 9.A11 into equation 9.A10, we derive et et ¼
ð1 fmdÞ ð pt pt Þ: fmðd þ syÞ þ ly
ð9:A12Þ
Substituting 9.A11 into 9.A5, we get p_ ¼ poð pt pt Þ:
ð9:A13Þ
Step 4: Defining Equilibrium Exchange Rate and Prices From equation 9.A9, we have pt ¼ m þ lr þ lðet þ eÞð1 þ FÞ fyt :
ð9:A14Þ
Substituting equation 9.A6 into 9.9, we get et ¼
½ð1 mdfÞ yt mu þ mdm þ mðdl þ sÞr md þ
½mðdl þ sÞðet þ eÞð1 þ FÞ mag þ mbtax : md
ð9:A15Þ
Step 5: Solving the Variables We take as given (that is, exogenous) values for m, r , g, tax, ðet þ eÞ, and yt . Substituting these values into equations 9.A14 and 9.A15, we have m; r ; g; tax; ðet þ eÞ; yt ) pt ; and et : We can then solve for the differential equation from 9.A13: pt ; p_t ) pt : Using equation 9.A11, we can solve for yt : pt ; pt ) y t : Using equation 9.A12, we can solve for et : p t ; e t ; pt ) e t : Using equations 9.A1 and 9.A2, we can solve for rt : r ; et ; et ; ðet þ eÞ ) rt :
244
Padma Desai and Pritha Mitra
Step 6: Simulation 1. Initially we assume e0 ¼ e0 ¼ 0 and we have m, r , y0 . Applying step 5, we solve for the initial values of all the variables. 2. A shock hits the economy: et ¼ expfktg, k < 0, et b 0. 3. We define yt ¼ y0 f ðet þ eÞ. Therefore, for each time, t, et ) yt . 4. Next, for each time, t, we have m, r , ðet þ eÞ, and yt . Applying step 5, we solve for the time, t, values of all the variables. 5. We repeat parts 3 and 4 of the simulation process until et ¼ 0, thereby reaching the long-run equilibrium. The permanent component of the shock has permanently reduced the equilibrium value of yt , permanently depreciated et , and permanently increased the price level pt . 6. In our simulations, we define the base case as the scenario in which the values of F, the macroeconomic fundamental that affects investor expectations, reflect the Argentine values of fiscal status, export sector strength, and saving rate13 ½ðtax gÞ; z; ð1 gÞ. In the alternative cases, the simulation described above is performed by changing the value of one macroeconomic fundamental: for example, z is changed to its Thai value.14 Notes 1. The literature on financial-crisis related issues such as capital account controls, current account deficits, contagion transmission, currency boards, exchange rate regimes, and interest rate policies for emerging markets, is vast and varied. Caballero and Krishnamurthy (2001), Calvo and Reinhart (2000), Lahiri and Vegh (2001), Calvo and Mishkin (2003), Kaminsky, Reinhart, and Vegh (2003) are among the noteworthy references. 2. Park and Lee differ from previous studies by explicitly contrasting the recovery paths of the East Asian economies from numerous past crises (of as many as 95 previous episodes during the period from 1970 to 1995). They adopt cross-country regression analysis for the purpose. 3. In the authors’ modeling, foreign borrowing must be collateralized by physical assets. In other words, the value of physical capital in foreign currency in each period must be greater than or equal to the value of foreign debt, short-term and long-term as well. 4. Details are in Desai (2003b). 5. The speculation term has transitory and permanent components to account for the fact that a financial crisis occasionally results in permanent GDP decline. Cerra and Saxena (2003) find evidence of permanent losses in GDP levels in the postcrisis phase of East Asian economies. In our model, this implies that long-term GDP is defined as yt ¼ y0 f ðet þ eÞ. Setting e ¼ 0, and defining y ¼ yt ¼ y0 , we can also simulate a model without permanent GDP losses. 6. The fiscal data employed in our simulations and their sources are reported in table 9.3. For Thailand, the precrisis values for government budget balance correspond to 1996 average quarterly real government expenditures less revenues. For Argentina, the precrisis values for government budget balance correspond to 1999 average quarterly real government expenditures less revenues. 7. The export and GDP data employed in our simulations and their sources are reported in table 9.3. For Thailand, the precrisis values for the export sector strength correspond to the annual percentage
Why Do Some Countries Recover More Readily from Financial Crises?
245
growth of real exports of goods and services relative to the annual percentage growth of real GDP, averaged over 1994, 1995, and 1996. For Argentina, the precrisis values for the export sector strength correspond to the annual percentage growth of real exports of goods and services relative to the annual percentage growth of real GDP, averaged over 1998, 1999, and 2000. 8. A version of the model where export sector strength is modeled as export earnings relative to external debt is also analyzed. Our preferred model in which we represent export sector strength via export growth relative to GDP growth gives stronger results than the alternative in which export sector strength is modeled as export earnings relative to external debt. We therefore analyze in depth the results of our preferred model. 9. The saving rate data and their sources are reported in table 9.3. For Thailand, the precrisis saving rate corresponds to the 1996 annual saving rate. For Argentina, the precrisis saving rate corresponds to the 1999 annual rate. 10. We assume for simplicity that demand for exports is exogenous. The country being modeled is assumed to be a small open economy. We assume that there is always a foreign country that demands its exports. So if the economy produces export goods, these will immediately be purchased by a foreign country. 11. After the economy is hit with an external shock, the permanent component of the shock, e b 0, remains constant. 12. Our base case simulations of Argentine GDP and inflation track their trends rather than their actual magnitudes. 13. Argentina: The precrisis fiscal status is measured as the 1999 average quarterly government expenditures less revenue. The precrisis export sector strength is measured as the average 1997–1999 annual percentage growth of real exports divided by the annual percentage growth of real GDP. The precrisis saving rate is measured as the 1999 annual saving rate. Details are given in table 9.2. 14. Thailand: The precrisis fiscal status is measured as the 1996 average quarterly government expenditures less revenue. The precrisis export sector strength is measured as the average 1994–1996 annual percentage growth of real exports divided by the annual percentage growth of real GDP. The precrisis saving rate is measured as the 1996 annual saving rate. Details are given in table 9.2.
References Baharumshah, Ahmad Z., Marwan A. Thanoon, and Salim Rashid. 2003. ‘‘Savings Dynamics in the Asian Countries.’’ Journal of Asian Economics 13: 827–845. Caballero, Ricardo J., and Arvind Krishnamurthy. 2001. ‘‘A ‘Vertical’ Analysis of Crises and Intervention: Fear of Floating and Ex-ante Problems.’’ Working Paper No. 8428, NBER, Cambridge, MA. Calvo, Guillermo A. 1998. ‘‘Capital Flows and Capital-Market Crises: The Simple Economics of Sudden Stops.’’ Journal of Applied Economics 1, no. 1: 35–54. Calvo, Guillermo A. 2003. ‘‘Explaining Sudden Stops, Growth Collapse and BOP Crises: The Case of Distortionary Output Taxes.’’ Working Paper No. 9864, NBER, Cambridge, MA. Calvo, Guillermo A., and Frederic S. Mishkin. 2003. ‘‘The Mirage of Exchange Rate Regimes for Emerging Market Countries.’’ Working Paper No. 9808, NBER, Cambridge, MA. Calvo, Guillermo A., and Carmen M. Reinhart. 2000. ‘‘When Capital Inflows Come to a Sudden Stop: Consequences and Policy Options.’’ In Reforming of the International Monetary and Financial System, eds. Peter Kenen and Alexandre Swoboda, 175–201. Washington, D.C.: International Monetary Fund.
246
Padma Desai and Pritha Mitra
Cerra, Valerie, and Sweta C. Saxena. 2003. ‘‘Did Output Recover from the Asian Crisis?’’ Working Paper No. 03/48, IMF, Washington, D.C. Cespedes, Luis F., Roberto Chang, and Andres Velasco. 2000. ‘‘Balance Sheets and Exchange Rate Policy.’’ Working Paper No. 7840, NBER, Cambridge, MA. Charoenseang, June, and Pornkamol Manakit. 2002. ‘‘Financial Crisis and Restructuring in Thailand.’’ Journal of Asian Economics 13: 597–613. Chowdhury, Abdur R. 1997. ‘‘The Financial Structure and the Demand for Money in Thailand.’’ Applied Economics 29: 401–409. Christiano, Lawrence J., Christopher Gust, and Jorge Roldos. 2003. ‘‘Monetary Policy in a Financial Crisis.’’ Journal of Economic Theory 119, no. 1: 64–103. Claessens, Stijn, Daniela Klingebiel, and Luc Laeven. 2001. ‘‘Financial Restructuring in Systemic Crises: What Policies to Pursue?’’ Paper presented at NBER Conference on Management of Currency Crises, Monterey, March 28–31. Corsetti, Giancarlo, Paolo Pesenti, and Nouriel Roubini. 2001. ‘‘Fundamental Determinants of the Asian Crisis: The Role of Financial Fragility and External Imbalances.’’ In Regional and Global Capital Flows: Macroeconomic Causes and Consequences, eds. Takatoshi Ito and Anne O. Krueger, 11–41. Chicago: University of Chicago Press. Dekle, Robert, and Mahmood Pradhan. 1999. ‘‘Financial Liberalization and Money Demand in the ASEAN Countries.’’ International Journal of Finance and Economics 4: 205–215. Desai, Padma. 2003a. ‘‘Explorations in Light of Financial Turbulence from Asia to Argentina.’’ Paper presented at Conference on the Future of Globalization: Explorations in Light of Recent Turbulence, Yale Center for the Study of Globalization, Yale University, October 10–11. Desai, Padma. 2003b. Financial Crisis, Contagion, and Containment: From Asia to Argentina. New Jersey: Princeton University Press. Dornbusch, Rudiger. 1976. ‘‘Expectations and Exchange Rate Dynamics.’’ Journal of Political Economy 84, no. 6: 1161–1176. Ghosh, Atish, and Paul R. Masson. 1991. ‘‘Model Uncertainty, Learning and the Gains from Coordination.’’ American Economic Review 81, no. 3: 465–479. Kaminsky, Graciela L., and Carmen M. Reinhart. 1999. ‘‘The Twin Crises: The Causes of Banking and Balance-of-Payments Problems.’’ American Economic Review 89, no. 3: 473–500. Kaminsky, Graciela L., Carmen M. Reinhart, and Carlos A. Vegh. 2003. ‘‘The Unholy Trinity of Financial Contagion.’’ Working Paper No. 10061, NBER, Cambridge, MA. Koo, Jahyeong, and Sherry L. Kiser. 2001. ‘‘Recovery from a Financial Crisis: The Case of South Korea.’’ Economic and Financial Policy Review, Federal Reserve Bank of Dallas 4: 24–36. Krugman, Paul. 1999. The Return of Depression Economics. New York: W.W. Norton & Company. Lahiri, Amartya, and Carlos A. Vegh. 2001. ‘‘Living with the Fear of Floating: An Optimal Policy Perspective.’’ Working Paper No. 8391, NBER, Cambridge, MA. Park, Yung C., and Jong W. Lee. 2001. ‘‘Recovery and Sustainability in East Asia.’’ Paper presented at Conference on Management of Currency Crises, NBER, Cambridge, MA, March 28–31. Rodrik, Dani, and Andres Velasco. 1999. ‘‘Short-Term Capital Flows.’’ Working Paper No. 7364, NBER, Cambridge, MA.
IV
Debt, Taxation, and Reforms
10
Government Debt: A Key Role in Financial Intermediation Michael Kumhof and Evan Tanner
10.1
Introduction
Optimally, governments should finance their expenditures such that losses from distortionary taxation are minimized. Many authors have emphasized that such losses might be substantially reduced through the use of state-contingent capital levies on government debt—in bad states of nature, the government defaults outright and/or engineers a debt devaluation through a price level increase.1 However, real-world policy debates are not typically cast in such terms. First, bonds with explicit state-contingent returns are rarely available to governments. Second, default or debt devaluation is typically considered to be a policy of last resort, and certainly not a desirable way to balance the budget.2 For example, according to the International Monetary Fund’s (2003) recent ‘‘stress test’’ approach to fiscal policy, a fiscal adjustment should be large enough to preempt inflation, default, or additional adjustment in the future. Why do policy makers think of debt devaluation so differently from the academic literature? As economists, we typically address such questions by considering various market imperfections. The most popular approach in the literature is the sticky price friction, as in Siu (2004), Schmitt-Grohe and Uribe (2004), and Angeletos (2003). Another is a cash-in-advance constraint on consumption, as in Nicolini (1998) and Dı´az-Gime´nez, Giovanetti, Marimon, and Teles (2003). But among policy makers, the reason that is often advanced is the damage that a debt devaluation would do to the domestic financial system. We therefore hypothesize that the difference between academic theory and policy reality is due to several features of financial markets that are absent from existing models, and that are particularly pronounced in developing countries. Most important, financial markets suffer from problems associated with asymmetric information,3 and legal or institutional imperfections can make it prohibitively costly and time-consuming to take security interests in real estate or movable property that could help to overcome these asymmetries. In this case
250
Michael Kumhof and Evan Tanner
fiscal policy, and especially predictable debt management that provides a safe asset, can be crucial in supporting at least some intermediation, thereby helping overcome barriers between borrowers and lenders. In essence, safe government debt facilitates financial intermediation by serving a collateral-like function. In developing countries, as we will show, this is reflected in the presence of large amounts of government debt on bank balance sheets. In some cases, government debt is also used as explicit collateral in repurchase agreements—a transaction requiring that government debt be safe. And finally, as is stressed in a large literature on the development of emerging bond markets, government debt plays an important role as a benchmark for private-sector bond markets, which are key to successful overall financial development. Consider how all this could be incorporated into a simple theoretical model.4 Assume that the government issues nominal debt and levies a proportional tax on labor income to finance a given expenditure flow. Producers need to borrow capital from an intermediary, who requires that a certain minimum fraction of the loan must be covered by government debt as collateral. Then fiscal policy creates two distortions. First, the positive labor tax rate implies a suboptimal work effort. Second, the collateral requirement implies a suboptimal level of capital. The first distortion usually makes a highly volatile inflation rate desirable in order to engineer state-contingent real returns on debt. But the second distortion does the opposite, because an inflation-induced erosion of the public debt stock would affect the economy’s ability to intermediate capital. A welfare-maximizing government therefore faces a trade-off. And that trade-off may suggest why defaults and debt devaluations are so rarely used in practice.5 International Monetary Fund (2004) illustrates the practical importance of this reasoning. This study finds that in each of the four recent government debt restructurings it examines, the key consideration in delaying a restructuring for as long as possible has been fear of the resulting damage to domestic banks, especially given their potential to spread and amplify the negative effects of a restructuring throughout the economy. This line of thinking can draw on several pieces in Guillermo Calvo’s inspiring body of work. A key element of Calvo (1988) is that surprise inflation is not costless, thereby giving rise to a trade-off between taxation and inflation. The paper implements this by way of an exogenous cost function for inflation. And Calvo and Vegh (1990, 1995) assume that bonds enter the utility function. While the application of that technology in those papers was different, this holds the key to a more formal modeling of costly unanticipated inflation, such as the approach suggested in the previous paragraph. In this paper, we provide evidence relevant to the outlined discussion. We show that the type of debt that is important for the effect of default and debt de-
Government Debt
251
valuation on domestic financial systems, domestically held debt, is of greater and growing importance than external debt for almost every single emerging market. This in itself is interesting, given that most of the interest in the context of emerging-market debt has focused on external debt, to which very different economic criteria apply in a default scenario. Next, to address the question of government debt in financial intermediation, we show that in most of these countries banks hold an extremely high proportion of their assets in government debt. To relate this finding to financial market imperfections, we show that there is a positive relationship between such holdings and a comprehensive index of legal and institutional imperfections, containing measures of difficulties in contract enforcement, in registering property, and in obtaining credit. Finally, we provide detailed evidence that suggests that the high bank holdings of government debt are no longer due to financial repression in all but a handful of the larger developing countries—they are indeed the result of a choice to hold the safe asset. 10.2 10.2.1
Government Debt in Developing Countries Domestic versus External Debt
The optimal fiscal policy literature has mainly used closed-economy models, and the analysis has mainly been applied to large industrialized countries such as the United States. On the other hand, when it comes to government debt in developing countries, the literature has used open-economy models, and has been almost exclusively concerned with external or sovereign debt. External debt default is clearly considered costly in practice,6 but in the literature that cost is not generally assumed to include direct negative effects on the domestic financial system. The question arises as to whether this exclusive preoccupation with external debt is still justified in today’s developing countries. As we will now show, with very few exceptions—including, most notably, Argentina—the answer to that question is no. Our main data source for this exercise is the Bank for International Settlements (BIS) (2003), which aggregates comprehensive data on individual securities issues obtained from financial markets sources.7 As explained in the data appendix, the split of the BIS data into domestic and international securities is very conservative in what it classifies as a domestic security. The only securities classified as domestic are issues by residents, targeted at resident investors, in domestic currency. As the BIS data set excludes Brady bonds, we merge it with a JP Morgan data set on outstanding Bradies.
252
Michael Kumhof and Evan Tanner
The results are presented in figures 10.1–10.6, which plot quarterly data from 1994, Q1 through 2003, Q4. For each group of countries we first present the domestic share of marketable government debt and then the ratio of marketable government debt to GDP. We first show various groups of developing countries, and then industrialized countries for comparison. The broad trends for the marketable debt-to-GDP ratios for industrialized and developing countries are quite different. Most industrialized countries started with relatively high ratios (around 50 percent but with wide deviations in either direction) that have not exhibited a long-term growth trend, a recent slight increase having been preceded by a decline in the 1990s. The well-known exception is Japan. Developing countries typically had much lower marketable debt-to-GDP ratios of around 20 percent in the mid-1990s, but with few exceptions they have since exhibited a positive growth trend. More importantly, the growth trend in the domestic part of that debt has been even stronger, as figures 10.1a, 10.2a, and 10.3a illustrate. In almost all major developing countries the share of domestic debt has risen and now represents in excess of 70 percent of overall debt.8 This is remarkable in view of the conservative criteria used for classifying debt as domestic. There is no clear trend in industrialized countries (see figures 10.4–10.6). The only major exceptions to this trend are countries that have experienced severe financial crises accompanied by large devaluations and consequent reductions in the real value of domestic debt. But the only cases where this has kept the domestic debt share well below 50 percent until the present day are Russia and Argentina.9 Mexico (in 1995) and Thailand (in 1997) also experienced such episodes but have since returned to domestic debt shares of over 70 percent. In these four cases the large devaluations that eroded nominal domestic debt were accompanied by severe banking crises and output contractions, with recoveries taking several years (see Kaminsky and Reinhart 1999). This provides ample evidence to justify our interest in domestically held debt in developing countries.10 As a natural extension of the above, we also briefly examine the cyclical properties of domestic debt. Does domestic debt act as a countercyclical ‘‘shock absorber,’’ a consequence of the positive relationship between tax revenue and GDP, or of countercyclical government expenditures? Or are governments able to borrow more and therefore run larger deficits during cyclical upswings, consistent with findings for developing countries by Kaminsky, Reinhart, and Ve´gh (2004)? Table 10.1 presents results from a simple bivariate regression of the cyclical components of the domestic debt-to-GDP ratio on real economic activity. Domestic debt moves procyclically (countercyclically) as the slope coefficient ðb1 Þ is greater than (less than) zero.
Government Debt
253
Figure 10.1 Latin America (a) Domestic share of marketable government debt, (b) Marketable government debt to GDP ratios
254
Michael Kumhof and Evan Tanner
Figure 10.2 Asia (a) Domestic share of marketable government debt, (b) Marketable government debt to GDP ratios
Government Debt
255
Figure 10.3 Other developing countries (a) Domestic share of marketable government debt, (b) Marketable government debt to GDP ratios
256
Michael Kumhof and Evan Tanner
Figure 10.4 Small industrialized countries (a) Domestic share of marketable government debt, (b) Marketable government debt to GDP ratios
Government Debt
257
Figure 10.5 Scandinavia (a) Domestic share of marketable government debt, (b) Marketable government debt to GDP ratios
258
Michael Kumhof and Evan Tanner
Figure 10.6 Large industrialized countries (a) Domestic share of marketable government debt, (b) Marketable government debt to GDP ratios
Government Debt
259
Table 10.1 Domestic Government Debt and the Business Cycle Estimated Regression: [Domestic Debt/GDP]tCYC ¼ b0 þ b1 Econ ActivtytCYC þ error Quarterly Data, 1994:3–2004:2 Industrialized Countries Country
Emerging/Developing Countries b1
t-stat
Country
b1
t-stat
Australia
0.21
0.56
Argentina
0.03
0.38
Austria
0.19
1.47
Brazil
0.05
0.21
Belgium
0.06
0.15
Chile
0.26
2.56
Canada
1.14
3.28
Czech Rep.
0.03
0.07
Denmark
0.18
0.30
Hungary
0.11
1.01
Finland France
0.34 0.12
1.73 0.23
Malaysia Mexico
0.03 0.03
0.54 0.36
Greece
0.27
0.72
Peru
0.01
0.35
Iceland
0.13
0.87
Philippines
0.34
1.56
Ireland
0.09
0.45
Poland
0.10
1.51
0.47
0.68
0.72
1.37
0.60
2.91
0.24
0.62
0.52 0.05
2.07 0.44
0.52
1.92
Italy Japan Netherlands New Zealand Norway Portugal South Korea Spain Sweeden
0.12
3.92
0.14
0.42
0.34
0.63
0.40
1.65
United Kingdom
0.06
0.13
United States
0.05
0.20
Switzerland
South Africa Thailand Turkey
0.00
0.02
0.41
2.92
Notes: 1. Data sources: Domestic Debt/GDP, Bank for International Settlements; Economic Activity (Industrial Production, Real GDP), International Monetary Fund—International Financial Statistics. 2. Trend/Cycle decomposition uses Hodrick–Prescott filter. 3. For the following countries, the industrial production index is used as an indicator for quarterly economic activity: Austria, Belgium, Brazil, Finland, Greece, Hungary, Japan, Korea, Malaysia, Mexico, Netherlands, Norway, Poland. In all other cases, real GDP is used.
260
Michael Kumhof and Evan Tanner
In most cases, the slope coefficient is statistically indistinguishable from zero. However, in four industrialized countries (Canada, Finland, Portugual, Switzerland) and one emerging country (Turkey) domestic debt appears to be countercyclical (b 1 < 1, jt-statj > 1:5). By contrast, in three industrialized countries ( Japan, New Zealand, South Korea) and three emerging/developing countries (Chile, Phillipines, Poland) domestic debt appears to move procyclically (b 1 > 1, jt-statj > 1:5). In the following subsections, we discuss several economic roles played by government debt in a developing economy. We emphasize throughout how such roles would be diminished if the government regularly employed state-contingent capital levies on debt. One key idea is that stable government debt management may facilitate bank-based financial intermediation, especially in an environment with a weak legal and institutional infrastructure. In addition to that, practitioners emphasize that stable government debt markets are the backbone for further development of financial markets beyond a bank-based system. Given that welldeveloped financial intermediation may have positive effects on capital accumulation and growth, this implies important real effects of fiscal and debt policy. 10.2.2 Debt
Financial Institutions Hold a Large Part of Their Assets in Government
Holders of government debt in developing countries include banks and pension funds. The role of the latter is still relatively minor, with a few notable exceptions such as Chile, as is the role of private securities markets. Financial intermediation is therefore highly dependent on banks. As we will now show, banks hold a very large proportion of their assets in government debt, which means that stable debt management becomes critical for financial stability. Figures 10.7–10.12 show the exposures of national banking systems to government debt in developing and developed economies. We define that exposure as the average ratio (1998–2002) of financial institutions’ net credit to the government11 to their total assets. (See the data appendix for further detail.) In industrialized countries (figures 10.11–10.12), where the fiscal situation is mostly very robust and government debt therefore extremely safe, that ratio is typically around 10 percent, whereas in developing countries (figures 10.7–10.10) ratios of 20 percent–40 percent are very common, and even ratios above 50 percent are observed in some of the largest developing countries. In such an economy, a government that contemplates a debt devaluation is faced with the insolvency of its banking system, because the typical bank capitalization ratio is less than 10 percent of assets. This is indeed exactly what happened in Argentina, where Perry and Serven (2002) conclude that ‘‘the roots of the [Argentinean] crisis lie in
Government Debt
Figure 10.7 Latin American financial institutions—credit to public sector/total assets
Figure 10.8 Asian financial institutions—credit to public sector/total assets
261
262
Michael Kumhof and Evan Tanner
Figure 10.9 Eastern European financial institutions—credit to public sector/total assets
Figure 10.10 Middle Eastern/African financial institutions—credit to public sector/total assets
Government Debt
Figure 10.11 Industrialized countries’ financial institutions—credit to public sector/total assets
Figure 10.12 Small European countries’ financial institutions—credit to public sector/total assets
263
264
Michael Kumhof and Evan Tanner
the . . . rigid exchange rate regime, the fragile fiscal position, and the hidden vulnerability of the banking system behind its strong facade.’’ During the protracted negotiations that eventually led to Argentina’s massive default, it was a key consideration that a large government default would seriously harm the banking system. 10.2.3 The Reason for Large Government Debt Holdings: Legal and Institutional Weaknesses Large holdings of government debt by banks in developing countries have in the past been the result of financial repression. However, as first stressed by Reinhart, Rogoff, and Savastano (2003), forced holdings of government debt by captive financial institutions are gradually disappearing and being replaced by government debt securities issued at market interest rates.12 The evidence presented in table 10.2 confirms that view.13 Reserve requirements no longer exist in several key developing countries, and have been falling in others, to the point that they are actually no longer binding. Binding reserve requirements seem to mostly affect pension funds in some countries, because they are prevented from investing abroad and have only limited domestic investment possibilities (see also World Pension Association 2002). The large holdings of government debt by banks must therefore be largely voluntary. On the other hand, as we have seen, government debt in developing countries is in fact mostly lower than in industrialized countries. The main problem is, therefore, that banks do not lend very much, or, in other words, that important segments of an economy are unable to borrow, or can only obtain credit with great difficulty. De Soto (2000) suggests that the culprit is poorly developed laws and institutions.14 Potential borrowers cannot collateralize their loans with physical assets, while lenders cannot enforce lending contracts and security provisions in the courts. Fleisig (1996, 1998) quantifies some of the resulting differences in lending behavior between industrialized and developing countries, and emphasizes that the major problem is the inability to use movable capital as collateral, with mortgages being subject to fewer problems. He mentions that in the U.S. 70 percent of all loans are secured by collateral, and almost 50 percent of all credit is secured by movable collateral. Movable capital represents 67 percent of the U.S. industrial capital stock and 75 percent of gross business investment (Asian Development Bank 2000). It is obviously critical for businesses to be able to use such capital as collateral. But in Argentina, for example, they are unable to do so. Only around 10 percent of all bank credit is secured at all, and practically none of it is secured by movable collateral. This is because, as described by Fleisig (1996,
Government Debt
265
Table 10.2 Reserve Requirements Country
Year
Description (RR ¼ reserve requirement)
Brazil
1996 1999
RR remains over 80% on demand deposits. April: RR on demand deposits up to 100%, on other acc. claims up to 60%. October: RR on bank deposits reduced from 65% to 45%. RR on demand deposits 45%, RR on time deposits 10%. No de jure RR (source: IMF Staff Economists). No de jure RR (source: IMF Staff Economists).
China
2001 2004 2004
Czech Republic
2004
No de jure RR (source: IMF Staff Economists).
India
1990 1998 1999 2000 2001 2004
Indonesia
1999 2004
Cash reserve ratio (CRR) 10%. CRR raised to 10.5%, lowered to 10%, raised to 11%. CRR lowered to 10.5%, then to 10%, and to 9% in November. CRR lowered to 8%. CRR lowered to 7.5%. Statutory Liquidity Ratio (SLR): RR of 25% in the form of government bond holdings. No RR. No de jure RR (source: IMF Staff Economists).
Jamaica
2004
Actual holdings of government bonds far exceed RR (source: IMF Staff Economists).
Korea Lebanon
2001 2004
RR ¼ 2.9%. No de jure RR (source: IMF Staff Economists).
Mexico
1991
1999
Liquidity coefficients on foreign currency deposits up to 50% dep. on maturities. Change from a daily-zero RR system to cumulative-zero RR over 28-day period. RR ¼ 0%.
Philippines
1991 1993 1994 1995 1997 1998 1999 2002
RR rationalized. RR for bank deposits lowered from 25% to 22%. RR lowered to 17%. RR lowered to 15%. RR lowered to 13%. RR lowered to 10%. RR ¼ 11% for commercial banks, RR ¼ 0% for rural and cooperative banks. RR ¼ 9%.
Poland
2004
No RR on banks (source: IMF Staff Economists).
Russia
2004
RR ¼ 3.5%.
Turkey
1994 1995 1996 2004
Comprehensive reform of RR. RR ¼ 10.8%. RR ¼ 9.3%. RR ¼ 8% for domestic currency deposits, RR ¼ 11% for foreign currency. Actual holdings of government bonds far exceed RR (source: IMF Staff Economists).
1995
266
Michael Kumhof and Evan Tanner
1998), creation of security interests, especially over movable property, is difficult, costly, and uncertain, perfection and registration is not effectively possible, and enforcement is slow and expensive. As a result, U.S. banks relying predominantly on movable collateral are seen as safe by both bank examiners and the public, while Argentinean banks that do so are not. They will therefore have to curtail the amount of such lending and seek to bolster their balance sheet through safer investments, including holdings of government debt. Asian Development Bank (2000) reports that when debtors can offer acceptable collateral, private creditors offer six to eight times more credit, two to ten times longer maturity, and 30 to 50 percent lower interest rates. The effect on interest rates is also described by Fleisig (1998), who compares loans against movable property in the United States and Argentina. Because in Argentina movable security is considered almost irrelevant to the recoverability of a loan, a loan for agricultural machinery (for example) therefore has lending rates of around 60 percent. Of the enormous difference of that rate to a corresponding U.S. lending rate, only about one-sixth is explained by macroeconomic risk and one-tenth by the general difficulty of collecting against security (including mortgages). A full three-quarters of the difference is due to problems with the legal framework fr secured transactions against movable property. This is a more general problem for both Latin America, as discussed in Garro (1998), and for Asia, as discussed in Asian Development Bank (2000) and Cranston (1998). These authors find that banks generally prefer real estate (and personal guarantees). The macroeconomic consequences of such problems include an inefficient savings allocation, an inefficient allocation of capital (with a preference for assets that can be used as collateral), fragile banks highly exposed to credit risks, an inability to use securitization for risk diversification (thereby increasing the dependence on fragile banks), and negative effects on equity because small and start-up businesses, who do not own land, find it almost impossible to obtain credit. All of this has negative effects on real economic activity. Fleisig (1998) cites country studies that put the cost of defective systems of lending security at 5 percent–10 percent of annual GNP, with the cost of reforming such systems a small fraction of that. The major international financial institutions have therefore in recent years paid increasing attention to these issues, as exemplified by the European Bank for Reconstruction and Development (Simpson and Menze, 2000),15 Asian Development Bank (2000), and World Bank (2004). It should be added that, in developing countries, even real estate security is often subject to numerous problems, including cost, time delays, difficulties in establishing priority in the absence of efficient registries, and other issues. Cranston (1998) explains that in several Asian countries much of the land remains
Government Debt
267
communally held according to customary law, ruling out its use as collateral. Fleisig (2002) mentions that in Latin America most of the land is not titled, and existing registries are rudimentary. Legal and institutional imperfections are therefore of critical importance for business, but are at the same time hard to quantify. The work of La Porta, Lopez de Silanes, Shleifer, and Vishny (1998) has therefore been extremely valuable in creating cross-sectional country indices for the quality of law and institutions affecting businesses. It is now available through World Bank (2004) in the form of updated indicators,16 on the basis of, among others, the papers by Djankov, McLiesh, and Shleifer (2004) and Djankov, La Porta, Lopez de Silanes, and Shleifer (2003). For the purpose of this study we are interested specifically in the factors responsible for segmented credit markets, which are represented by three factors. The first is ‘‘getting credit,’’ describing the quality and accessibility of credit information and the quality of collateral and bankruptcy laws. The second is ‘‘enforcing contracts,’’ measuring the cost and time delays in the collection of an overdue debt. And the third is ‘‘registering property,’’ measuring the cost and time delays of transferring title to real estate. We compute an overall raw index for all countries in our sample by attaching equal weights to all three categories.17 Our final index ranges between 0 and 100, with 0/100 for the countries with the worst/best raw index. We create another index for the share of debt on financial institutions’ balance sheets, as discussed earlier, with 0/100 for the lowest/highest government debt-to-total assets ratio among all countries in the sample. Figure 10.13 plots these two indices against each other. The results are striking. There is a very strong negative relationship between the quality of law and institutions and the amount of government debt that financial institutions choose to hold on their balance sheets. Almost all the countries in the top left corner of figure 10.13 are developing countries, and almost all in the bottom right are industrialized (Chile does the best among developing countries). Financial institutions with highly risky and largely unsecured lending books are very vulnerable to macroeconomic shocks. Their high holdings of government debt are generally an attempt to offset the credit risk inherent in private credit with government debt, which ideally bears no credit risk. This is recognized by the Basle rules for capital adequacy, which give a much lower risk weighting (zero) to such debt, even if the debtor is an emerging-market government. The safety of banks’ government debt portfolio is a precondition for even the limited amount of private sector lending that does take place. In such countries prudent management of public debt is critical for the health of the banking system, and using state-contingent returns on government debt would be highly damaging.
268
Michael Kumhof and Evan Tanner
Figure 10.13 Government debt on balance sheets and legal/institutional quality
10.2.4 Government Debt and Further Financial Development: Private Bond Markets As we have argued, prudent management of government debt is important to safeguard fragile domestic banking systems. But financial development in the long run benefits greatly from moving beyond a purely bank-based system to include not only a stock market, but also a long-term private bond market. And here again, as emphasized by World Bank-IMF (2001),18 the government bond market plays a critical role because it is the backbone of most fixed income securities markets. Herring and Chatusripitak (2000) present the key arguments for why private bond markets matter for financial development, stability and growth,19 and show how legal and institutional factors have impeded the development of private bond markets in Asia. The lack of developed bond markets in developing coun-
Government Debt
269
tries has in recent years also been a major theme for the international financial institutions, including the Inter-American Development Bank (Castellanos 1998, Reinstein 2002, and del Valle 2002), Asian Development Bank (2002), World Bank-IMF (2001), and policy makers in individual countries.20 The Asian Bond Market Initiative (Rhee 2004) is a concerted effort to develop bond markets across that region. Underdeveloped bond markets make the pricing of credit risks and equities harder because of the absence of a benchmark yield curve. Derivatives markets cannot develop, making the diversification of risk exposures harder. Fewer savings are utilized, and borrowers face higher borrowing costs and shorter maturities (potentially leading to foreign borrowing and thereby resulting in foreign exchange risks), and banks become too large relative to the overall financial sector, thereby making the economy more vulnerable to crises because of its dependence on a small number of institutions. Herring and Chatusripitak (2000) identify two major prerequisites for the development of bond markets. The first is the legal and institutional infrastructure mentioned in section 2.3. The second, already mentioned as a key concern in World Bank-IMF (2001), is a deep, liquid, government bond market. As we have seen, legal problems restrict even bank-based private sector lending, and clearly they do the same to bonds-based lending. But while it is important to remove those weaknesses, this is a time-consuming process, and in the meantime a stable government bond market is all the more important, and can help overcome at least some of the problems caused by those weaknesses. In several developing countries that have still not tackled their legal frameworks effectively, development of the government bond market has nevertheless started to support a private bond market, at least to strong borrowers that are relatively less dependent on collateral. A government debt market does this first by putting in place a basic financial infrastructure including laws, institutions, products, services, repo, and derivatives markets, and second by playing a role as an informational benchmark. A single private issuer of securities would never be of sufficient size to generate a complete yield curve, and his securities would not be nominally riskless because only the government has the power to print domestic currency. The government, through the government debt market, can therefore provide a public good to financial markets, but only under two further conditions. First, macroeconomic volatility, especially inflation volatility, must be low so that a nominal yield curve is informative about the real cost of borrowing. State-contingent inflation would interfere with this goal, and experience confirms this—the volatility in Brazil in 1999 and in Argentina in 2001–2002 were major impediments to the further development of their local financial markets (del Valle 2002). Second, the government
270
Michael Kumhof and Evan Tanner
must issue a sufficient volume of debt. For this latter reason, during the era of shrinking U.S. public debt (the late 1990s and early 2000s), some observers expressed concern that it would be more difficult to conduct monetary policy in a smaller government debt market (Reinhart, Sack, and Heaton 2000 and Fleming 2000).21 Similarly, for developing countries Herring and Chatusripitak (2000) conclude that the goal of developing a robust bond market may conflict with the goal of minimizing the cost of government borrowing if a government with fiscal surpluses decides to issue government debt and invest the proceeds in foreign securities in order to provide liquidity, as Hong Kong has done in recent years. We have seen that government debt plays two key roles. It provides infrastructure and acts as an informational benchmark in securities markets. And on the balance sheets of financial institutions it is, to depositors, a form of security that increases their willingness to have their funds intermediated in a generally risky environment. As such, it informally acts as collateral. However, government debt can also play a more direct role as collateral in wholesale securities markets. In particular, it plays a critical role in managing risks in derivatives markets, payment and settlement systems, and in the market for repurchase agreements. Bank for International Settlements (2001) shows that there has been an enormous increase in such collateralized transactions in recent years. Under a repurchase agreement, a market participant sells a security while simultaneously agreeing to repurchase it in the future. Such a transaction functions as a secured loan. The party purchasing the security makes funds available to the seller while holding that security as collateral. In virtually every financial market worldwide, the only form of security acceptable in repo transactions is government debt, due to its low or zero risk and high liquidity. Because repo markets require quite well-developed financial markets, they generally appear at a later stage of development. We nevertheless found sizeable repo markets in Brazil and Mexico, using as a criterion Bankscope balance-sheet data to express the stock of outstanding repo loans as a fraction of total assets. In Brazil that ratio is around 7 percent, and in Mexico around 5 percent in 2002–2003. Other sizeable markets exist in the Philippines, Poland, and India. 10.3
Summary and Conclusions
Is state-contingent inflation (or default) on government debt an optimal way to conduct fiscal policy in response to shocks? This paper makes an attempt to understand the very different replies to this question given by the theoretical literature on optimal fiscal policy on the one hand, and by experienced practitioners on the other. Many contributions to the theoretical literature advocate what we have referred to as debt devaluations (in response to negative shocks), while prac-
Government Debt
271
titioners only think of it as a last resort, to be avoided at almost all costs. The emphasis in this paper is on understanding the thinking of practitioners, particularly decision makers in developing countries, by presenting some pertinent data. We show that debt devaluations are likely to have very negative effects on financial intermediation in developing countries, as their banks are highly exposed to government debt and at the same time face much higher risks in private sector lending because of weak legal and institutional infrastructures. Keeping government debt a safe investment for banks is critical to support the already low financial intermediation that does exist. It is also critical to support further improvements in financial intermediation, away from banks and toward securities markets. Many countries that have not followed, or have been unable to follow, this advice have experienced serious lending crunches and recessions, and have set back the development of their financial markets by years. It would be beneficial to further develop the theory of optimal fiscal policy against the background of these arguments, as this would greatly enhance its usefulness to applied policy analysis. In doing so, great inspiration can be drawn from the work of Guillermo Calvo, who has made several key contributions to this literature. 10.4
Data Appendix
Domestic versus External Debt The data were obtained from Bank for International Settlements (2004), which follows the following classification system for domestic versus external/international securities: Issues by Residents
Issues by Nonresidents
In domestic currency, targeted at resident investors
Domestic
International
In domestic currency, targeted at nonresident investors
International
International
In foreign currency
International
International
Note that the classification covers the borrower side of securities issues, not the actual ownership. This system is very conservative in what it classifies as a domestic security.22 Issues by nonresidents targeted at domestic investors, whether in domestic or foreign currency, are invariably classified as international,
272
Michael Kumhof and Evan Tanner
although many if not most holders will be domestic. The same is true for issues in domestic currency targeted at foreign investors, because it is well known that the ultimate holders of developing countries’ domestic currency debt are in many cases domestic residents.23 Bank for International Settlements (2004) also states that notes and money market instruments issued by nonresidents in domestic currency and in domestic markets are not included due to lack of data, which again biases results against domestically held debt. The opposite bias results from another missing debt class, Brady bonds, but we remedy this by merging the BIS data set with a JP Morgan data set on outstanding Brady bonds, using the Merrill Lynch (1994) guide for the classification of securities. Bank Holdings of Government Debt We define the exposure of banks to government debt as the ratio of (Financial Institutions’ Net Credit to Government) / (Financial Institutions’ Net Total Assets). The percentage figures are computed from International Financial Statistics, and are averages for 1998–2002, except for countries in the euro area, where 1999–2003 was used. The numerator is the sum of all entries representing net credit to the public sector by deposit money banks, other banking institutions, and nonbank financial institutions (the latter series do not exist for all countries). The denominator is the sum of the net total assets of these three groups, after canceling out credit items between them. To obtain the net figures we deduct from both numerator and denominator the sum of all entries representing credit by the public sector to these institutions. Crisis years are excluded from the computation of averages as follows: Argentina 2002, Indonesia 1998. Acknowledgements This paper was prepared for the festschrift in honor of Guillermo Calvo, held at the International Monetary Fund on April 16, 2004. The authors would like to thank the editor for his patience. We are grateful to Abdul Abiad and Ashoka Mody, who generously shared their data set on financial repression with us. Excellent research assistance was provided by Nicolas Amoroso and Wolfgang Harten. All errors are our own. Notes 1. See Lucas and Stokey (1983), Chari, Christiano, and Kehoe (1994), and a large subsequent literature surveyed, for example, in Chari and Kehoe (1999). More recent contributions include Angeletos (2002) and Buera and Nicolini (2004).
Government Debt
273
2. Such a policy must be unanticipated and credibly on a one-time basis. If it does become unavoidable to resolve a fiscal problem in this way, then achieving that credibility is a key objective of a welldesigned IMF fiscal adjustment program. The time-inconsistency problem associated with such a policy discussed by Calvo (1978) and Calvo and Guidotti (1993) is not considered in this paper. 3. A related literature employs models of segmented asset markets; for example, Alvarez, Lucas, and Weber (2001). In such models a subset of agents is unable to trade assets and therefore to smooth consumption over time and states of nature. Instead, state-dependent taxes and transfers (including the inflation tax, in that paper) help them do so. In a similar vein, Shin (2003) develops a model of fiscal policy with heterogeneous agents. To us it seems that the key issue is the ability of firms to continue to obtain financing. Studies such as Kaminsky and Reinhart (1999) have shown that (debt) devaluations in developing countries are almost invariably accompanied by a banking collapse that causes a severe output contraction. 4. See Kumhof (2004) for more details. 5. Depending on the parameterization of the model, it may also address another problem of the optimal fiscal policy literature, which is that in many of those models only very small stocks of government debt can be sustained in equilibrium. 6. It also involves an effect that does not arise under a domestic debt default, and that makes default beneficial: a wealth transfer from foreigners. 7. This data set excludes nonmarketable debt. For more comprehensive measures of government debt than the one used in this study, a disaggregation into domestic and international debt is not available at the frequency and country coverage required. 8. The overall size of emerging local bond markets at the end of 2002 was four times the size of those countries’ foreign currency external debt (see Mathieson et al. 2004, IMF 2004). 9. The latter is a special case because its commitment to a currency board left little incentive to develop a domestic currency bond market. 10. Similar evidence for many smaller developing countries is presented in Cabbar and Jonasson (2004) for South Asia, del Valle (2002) and Batlay and del Valle (2002a) for the Middle East and North Africa, and Batlay and del Valle (2002b) for Eastern Europe and Central Asia. 11. Unlike the BIS measure used above, this includes nonmarketable government debt. 12. Caprio (1999) also discusses evidence to this effect. 13. We are extremely grateful to Abdul Abiad and Ashoka Mody for sharing with us the historic information contained in table 10.2, which can also be found in Abiad (2004). The 2004 entries were collected by contacting IMF staff economists. 14. Pagano (2001) contains similar arguments. 15. The EBRD’s Model Law on Secured Transactions of 1994 has been used to help several Eastern European transition economies in the redesign of their legal systems. 16. The data set is available at www.doingbusiness.org. 17. Our qualitative results are not dependent on the details of the weighting scheme. 18. See chapter 1, page 3, table 1.1 in World Bank-IMF (2001). 19. Similar arguments are presented by Fry (1997). 20. For example, Kang, Kim and Rhee (2004) discuss Korea, and Kim (2001) discusses a number of Asian countries in great detail.
274
Michael Kumhof and Evan Tanner
21. Not everyone shared this concern for the U.S., where financial markets are so deep that some substitutes may be found, especially the abundant issues by government-sponsored enterprises. But Herring and Chatusripitak (2000) stress that the same is not true for developing countries. This is precisely because there the historic lack of developed public debt markets has been an impediment to the growth of private financial markets. 22. Notice that limiting the attention to securities is necessary for practical purposes, but a more complete picture would also consider nonmarketable government obligations. These include on the domestic side a variety of spending commitments and outstanding debt balances, and on the international side concessional lending from governments and international financial institutions, as well as syndicated bank loans. 23. Another source of bias is that local issues in foreign currency are generally classified as international. But here exceptions are made by BIS for Argentina, Peru, and the Philippines.
References Abiad, A. 2004. ‘‘A New Database of Financial Reforms.’’ Working Paper, IMF, Washington, D.C. Alvarez, F., R. E. Lucas, Jr., and W. E. Weber. 2001. ‘‘Interest Rates and Inflation.’’ American Economic Review 91, no. 2: 219–225. Angeletos, G.-M. 2002. ‘‘Fiscal Policy with Noncontingent Debt and the Optimal Maturity Structure.’’ Quarterly Journal of Economics 117, no. 3: 1105–1131. Angeletos, G.-M. 2003. ‘‘Optimal Fiscal and Monetary Policy.’’ NBER Macroeconomics Annual 2003, Cambridge, MA: MIT Press. Asian Development Bank. 2000. ‘‘Secured Transactions Law Reform in Asia: Unleashing the Power of Collateral.’’ In Law and Policy Reform at the Asian Development Bank, Vol. II. Manila, Philippines: Asian Development Bank. Bank for International Settlements. 2001. ‘‘Collateral in Wholesale Financial Markets: Recent Trends, Risk Management and Market Dynamics.’’ Publication No. 17, Committee on the Global Financial System, BIS, Basel, Switzerland. Bank for International Settlements. 2003. ‘‘Guide to the International Financial Statistics.’’ Paper No. 14, BIS, Basel, Switzerland. Available at http://www.bis.org/publ/bispap14.pdf. Batlay, M., and C. del Valle. 2002a. ‘‘Regional Snapshot of Government Bond Markets in the Middle East and North Africa.’’ Financial Sector Development, The World Bank. Powerpoint presentation. Batlay, M., and C. del Valle. 2002b. ‘‘Regional Snapshot of Government Bond Markets of Eastern Europe and Central Asia.’’ Financial Sector Development, The World Bank. Powerpoint presentation. Buera, F., and J. P. Nicolini. 2004. ‘‘Optimal Maturity of Government Debt without State Contingent Bonds.’’ Journal of Monetary Economics 51, no. 3: 531–554. Cabbar, D. and T. Jonasson. 2004. ‘‘Regional Snapshot of Government Bond Market Development in South Asia Economies.’’ Paper presented at Seacen-World Bank-IMF Seminar on Strengthening the Development of Domestic Debt Securities Markets, Colombo, Srilanka, June 8–11. Calvo, G. A. 1978. ‘‘On the Time Consistency of Optimal Policy in a Monetary Economy.’’ Econometrica 46, no. 6: 1411–1128. ———. 1988. ‘‘Servicing the Public Debt: The Role of Expectations.’’ American Economic Review 78: 647– 671.
Government Debt
275
Calvo, G. A., and P. E. Guidotti. 1993. ‘‘Optimal Maturity of Nominal Government Debt: An Infinite Horizon Model.’’ International Economic Review 33, no. 4: 895–919. Calvo, G. A. and C. A. Vegh. 1990. ‘‘Interest Rate Policy in a Small Open Economy: The Predetermined Exchange Rates Case.’’ IMF Staff Papers 37, no. 4: 753–776. ———. 1995. ‘‘Fighting Inflation with High Interest Rates: The Small Open Economy Case under Flexible Prices.’’ Journal of Money, Credit and Banking 27: 49–66. Caprio, G. Jr. 1999. ‘‘Review of S. Fry: Emancipating the Banking System and Developing Markets for Government Debt.’’ Journal of Economic Literature 37, no. 4: 1723–1725. Castellanos, J. 1998. ‘‘Developing Government Bond Markets.’’ Working Paper IFM-111, InterAmerican Development Bank, Washington, D.C. Chari, V. V., L. Christiano, and P. J. Kehoe. 1994. ‘‘Optimal Fiscal Policy in a Business Cycle Model.’’ Journal of Political Economy 102, no. 4: 617–652. Chari, V. V., and P. J. Kehoe. 1999. ‘‘Optimal Fiscal and Monetary Policy.’’ In Handbook of Macroeconomics, Vol. 1C, eds. J. B. Taylor and M. Woodford, 1671–1745. Amsterdam: Elsevier. Cranston, R. 1998. ‘‘Credit Security and Debt Recovery: Law’s Role in Reform in Asia and the Pacific.’’ In Emerging Financial Markets and Secured Transactions, International Economic Development Law Volume 6, eds. J. J. Norton and M. Andenas. Boston: Kluwer Law International. del Valle, C. 2002. ‘‘Developing Government Securities Markets—A Framework.’’ Mimeo. Financial Sector Development, The World Bank. De Soto, H. 2000. The Mystery of Capital: Why Capitalism Triumphs in the West and Fails Everywhere Else. New York: Basic Books. Dı´az-Gime´nez, J., G. Giovanetti, R. Marimon, and P. Teles. 2003. ‘‘Nominal Debt as a Burden on Monetary Policy.’’ Working Paper No. 8. Centre de Refere`ncia en Economia Analitica, Barcelona. Djankov, S., C. McLiesh, and A. Shleifer. 2004. ‘‘Private Credit Around the World.’’ Working Paper, Department of Economics, Harvard University. Djankov, S., R. La Porta, R. Lopez-de-Silanes, and A. Shleifer. 2003. ‘‘Courts.’’ Quarterly Journal of Economics 118: 453–517. Fleisig, H. W. 1996. ‘‘Secured Transactions: The Power of Collateral.’’ Finance and Development, The World Bank. Available at http://worldbank.orgfandd/english/0696/articles/0150696.htm. ———. 1998. ‘‘Economic Functions of Security in a Market Economy.’’ In Emerging Financial Markets and Secured Transactions, International Economic Development Law Volume 6, eds. J. J. Norton and M. Andenas. Boston: Kluwer Law International. ———. 2002. ‘‘Microenterprises and Collateral.’’ Issues Brief No. 2, Center for the Economic Analysis of Law, Washington, D.C. Fleming, M. J. 2000. ‘‘Financial Market Implications of the Federal Debt Paydown.’’ In Brookings Papers on Economic Activity 2: 221–251. Fry, M. J. 1997. Emancipating the Banking System and Developing Markets for Government Debt. New York: Routledge. Garro, A. M. 1998. ‘‘Difficulties in Obtaining Secured Lending in Latin America: Why Law Reform Really Matters.’’ In Emerging Financial Markets and Secured Transactions, International Economic Development Law Volume 6, eds. J. J. Norton and M. Andenas. Boston: Kluwer Law International.
276
Michael Kumhof and Evan Tanner
Herring, R. J. and N. Chatusripitak. 2000. ‘‘The Case of the Missing Market: The Bond Market and Why It Matters for Financial Development.’’ Working Paper 11, ADB Institute, Tokyo, Japan. International Monetary Fund. 2003. ‘‘Sustainability Assessments—Review of Application and Methodological Refinements.’’ Policy Development and Review Department. Available at http://www.imf.org /external/np/pdr/sustain/2003/061003.pdf. International Monetary Fund. 2004. ‘‘Sovereign Debt Restructurings and the Domestic Economy— Experience in Four Recent Cases.’’ Policy Development and Review Department. Available at http:// www.imf.org/external/NP/pdrsdrm/2002/022102.pdf. Kaminsky, G., and C. M. Reinhart. 1999. ‘‘The Twin Crises: The Causes of Banking and Balance of Payments Crises.’’ American Economic Review 89, no. 4: 473–500. Kaminsky, G., C. M. Reinhart, and C. Ve´gh. 2004. ‘‘When It Rains It Pours: Procyclical Capital Flows and Macroeconomic Policies.’’ In NBER Macroeconomics Annual 2004, eds. M. Gertler and K. Rogoff, 11–53. Cambridge, MA: MIT Press. Kang, K., G. Kim, and C. Rhee. 2004. ‘‘Developing the Government Bond Market in Korea: History, Challenges, and Implications for Asian Countries.’’ Paper presented at Asia Economic Panel (AEP), Columbia University, October 7–8. Kim, Y.-H., ed. 2001. Government Bond Market Development in Asia. Manila, Philippines: Asian Development Bank. Kumhof, M. 2004. ‘‘Fiscal Crisis Resolution: Taxation versus Inflation.’’ Working Paper, Stanford University. La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. Vishny. 1998. ‘‘Law and Finance.’’ Journal of Political Economy 106, no. 6: 1113–1155. Lucas, R. E. Jr., and N. Stokey. 1983. ‘‘Optimal Fiscal and Monetary Policy in an Economy Without Capital.’’ Journal of Monetary Economics 12: 55–93. Mathieson, D. J., J. E. Roldos, R. Ramaswamy, and A. Ilyina. 2004. ‘‘Emerging Local Securities and Derivatives Markets.’’ Selected Topics Paper, IMF, Washington, D.C. Merrill Lynch. 1994. ‘‘The 1995 Guide to Brady Bonds.’’ Nicolini, J. P. 1998. ‘‘More on the Time Inconsistency of Optimal Monetary Policy.’’ Journal of Monetary Economics 41: 333–350. Pagano, M., ed. 2001. Defusing Default-Incentives and Institutions. Washington, D.C.: Johns Hopkins University Press. Perry, G., and L. Serven. 2002. ‘‘La Anatomia de Una Crisis Multiple: Que Tenia Argentina de Especial y Que Podemos Aprender de Ella (with English summary).’’ Desarrollo Economico 42, no. 167: 323–375. Reinhart, C. M., K. S. Rogoff, and M. A. Savastano. 2003. ‘‘Debt Intolerance.’’ Brookings Papers on Economic Activity 2003, no. 1: 1–74. Reinhart, V., B. Sack, and J. Heaton. 2000. ‘‘The Economic Consequences of Disappearing Government Debt.’’ Brookings Papers on Economic Activity 2: 163–220. Reinstein, A. 2002. ‘‘Issues in Building Corporate Money and Bond Markets in Developing Market Economies.’’ Working Paper, IFM Division, Inter-American Development Bank, Washington, D.C. Rhee, C. 2004. ‘‘Asia Bond Market Initiative: Progress and Future Directions.’’ Presentation at the International Monetary Fund.
Government Debt
277
Schmitt-Grohe, S., and M. Uribe. 2004. ‘‘Optimal Fiscal and Monetary Policy under Sticky Prices.’’ Journal of Economic Theory 114, no. 2: 198–230. Shin, Y. 2003. ‘‘Optimal Fiscal Policy with Incomplete Markets.’’ Working Paper, Department of Economics, University of Wisconsin, Madison. Simpson, J., and J. Menze. 2000. ‘‘Secured Transactions: Ten Years of Legal Reform.’’ Law in Transition (European Bank for Reconstruction and Development) Autumn: 20–27. Siu, H. E. 2004. ‘‘Optimal Fiscal and Monetary Policy with Sticky Prices.’’ Journal of Monetary Economics 51, no. 3: 575–607. World Bank-IMF. 2001. Developing Government Bond Markets: A Handbook. Washington, D.C.: World Bank. World Bank. 2004. Doing Business in 2004: Understanding Regulation. Washington, D.C.: World Bank. World Pension Association. 2002. ‘‘Investment Restrictions to Pension Funds.’’ Available at http:// www.world-pensions.org/docus/tres16.doc.
11
Capital Income Taxation in the Globalized World Assaf Razin and Efraim Sadka
11.1
Introduction
These days globalization across various economies is a universal force to reckon with. Guillermo Calvo’s framework for his prolific research was, indeed, the constraints on economic policy imposed by the integrated, and fluctuating, world capital market. Maurice Obstfeld and Alan M. Taylor (2003) examine the historical development of globalization (in particular, international capital mobility) by political-economy forces. After World War I, ‘‘newly or better-enfranchised groups such as the working classes’’ contributed to severely impede capital mobility. The peace and prosperity that emerged following World War II, and that intensified after the end of the Cold War, unleashed political forces for freer capital mobility. Nevetheless, the aging population (through falling birth rates and increased longevity) raises the need for tax revenues by the welfare state. This chapter focuses on capital income taxation. Can high domestic capital taxes survive international tax competition brought about by recent widespread globalization? Evidently, the answer is negative. As put succinctly by The Economist (1997): Globalization is a tax problem for three reasons. First, firms have more freedom over where to locate. . . . This will make it harder for a country to tax [a business] much more heavily than its competitors. . . . Second, globalization makes it hard to decide where a company should pay tax, regardless of where it is based. . . . This gives them [the companies] plenty of scope to reduce tax bills by shifting operations around or by crafting transferpricing. . . . [Third], globalization . . . nibbles away at the edges of taxes on individuals. It is harder to tax personal income because skilled professional workers are more mobile than they were two decades ago.
In fact, Oates (1972, 192), in his classic treatise on the related issue of fiscal federalism, states:
280
Assaf Razin and Efraim Sadka
The result of tax competition may well be tendency towards less than efficient levels of outputs of local public services. In an attempt to keep tax rates low to attract business investments, local officials may hold spending below those levels for which marginal benefits equal marginal costs.
The literature on tax competition has developed a great deal since this early ‘‘race-to-the-bottom’’ prediction. Specifically, Razin and Sadke (1991) point out that this race is halted when residence-based taxation can be effectively enforced. Baldwin and Krugman (2000) invoke agglomeration effects in order to explain why capital may be a quasifixed factor on which the tax does not have to race to the bottom. Indeed, Krogstrup (2002) indicates that capital taxes did not fall in the 1960s and 1990s in the European Union (EU), and the average tax revenues from the corporate sector even increased, both as percentages of GDP and of total tax revenues. Nevertheless, residence-based taxation is rarely enforced effectively and uniformly. For instance, a member country of the EU-15 can more effectively tax its resident on income originating elsewhere in the EU-15, but less effectively on income originating elsewhere in the EU-27, in this transition stage, and even less effectively on income originating in tax havens elsewhere in the world. We illustrate in this chapter how strong international tax competition in the era of globalization imposes severe constraints on capital income taxation, and thereby put into question its standing in the public finance of the welfare state. We develop a political economy model to assess how the forces of globalization affect the taxation of capital income. The chapter is organized as follows: section 2 provides a simple analytical framework for the study of capital taxation in the presence of international capital mobility. In particular, we analyze the tax structure in the political-economy equilibrium. In section 3 we apply the model for the analysis of international tax competition. Section 4 concludes. 11.2
International Capital Mobility: A Stylized Political-Economy Tax Model
We present a stripped-down model of international capital mobility, which enables us to explore key issues of international taxation without being sidetracked by irrelevant complications. We consider an economy that lives for two periods, indexed by t ¼ 1; 2. There is one aggregate, all-purpose good in each period, serving for both consumption and investment. See Ruzin, Sadka, and Swagel (2002, 2004) for a related analysis. 11.2.1
Households
There are two types of workers: skilled workers, who have high productivity and provide one efficiency unit of labor per unit of labor time, and unskilled workers,
Capital Income Taxation in the Globalized World
281
who provide q < 1 efficiency units of labor per unit of time. Workers have one unit of labor time during each one of the two periods of their life. They are born without skills and thus with low productivity. In the first period, each worker chooses whether to get an education and become a skilled worker, or instead remain unskilled. There is a continuum of individuals, characterized by an innate ability parameter, e, which is the time needed to acquire a skill. By investing eb units of labor time in education during the first period, a worker becomes skilled, after which the remaining ð1 eÞ units of labor time in the first period provide an equal amount of efficiency units of labor in the balance of the first period. We assume that the individual also provides one efficiency unit of labor in the second period. We also assume a positive pecuniary cost of acquiring skills, g, which is not tax-deductible. Given these assumptions, there exists a cutoff level, e , such that those with education cost parameters below e will invest in education and become skilled, whereas everyone else will remain unskilled. The cutoff level is determined by the equality between the present value of the payoff to education and the cost of education (including foregone income): w2 ð1 tL Þð1 qÞ w1 þ ¼ ð1 tL Þw1 e þ g; ð11:1Þ 1 þ ð1 tD Þr where wt is the wage rate per efficiency unit of labor in period t ¼ 1; 2; r is the domestic rate of interest; tL is the tax rate on labor income (constant over time); and tD is the tax rate on capital income of residents from domestic sources (see next paragraph). Rearranging terms, equation 11.1 yields w2 =w1 g e ¼ ð1 qÞ 1 þ : ð11:2Þ 1 þ ð1 tD Þr ð1 tL Þw1 Note that the two taxes, the tax on labor income and the tax on capital income, have opposite effects on the decision to acquire skill. The tax on labor income reduces the foregone (net of tax) income component of the cost of education. It also reduces the payoff to education by the same proportion.1 Were the pecuniary cost g equal to zero (or else tax-deductible), the labor income tax would have no effect on the decision to acquire skill. However, with a positive pecuniary cost of education, the labor income tax has a negative effect on acquiring skills: it reduces e and, consequently, the proportion of the population who becomes skilled (namely, Gðe Þ). On the other hand, the tax on capital income has a positive effect on education because it reduces the (net-of-tax) discount rate, thereby raising the present value of the future payoff to education.
282
Assaf Razin and Efraim Sadka
We assume for the sake of simplicity that the individual’s leisure time is exogenously given. Nevertheless, total labor supply is distorted by the taxes, as can be seen from equation 11.2. Note that there are Gðe Þ skilled individuals and 1 Gðe Þ unskilled individuals in each period. The labor supply of each one of the unskilled individuals, in efficiency units, is q in each period. Therefore, total labor supply in efficiency units of the unskilled individuals is q½1 Gðe Þ in each period. However, a skilled individual devotes e units of her time in the first period to acquiring education, and hence works only 1 e units of time in the first period. Thus, the individual labor supply in the first period varies over e. The labor supply of skilled individuals is equal to ð e
ð1 eÞ dG:
0
Any skilled individual supplies as labor all of her unit time in the second period. Thus, total labor supply ðLt Þ in efficiency units in period t ¼ 1; 2, is given by L1 ¼
ð e
ð1 eÞ dG þ q½1 Gðe Þ
ð11:3Þ
0
and L2 ¼ Gðe Þ þ q½1 Gðe Þ:
ð11:4Þ
For the sake of simplicity, assume that all individuals have identical preferences over first- and second-period consumption [c1 ðeÞ and c2 ðeÞ, respectively], represented by a common, concave utility function u½c1 ðeÞ; c2 ðeÞ. Each individual has initial income (endowment) in the first period of I1 units of the consumptioncapital good. The total amount of the initial endowment (I1 , because the size of the population is normalized to one) serves as the stock of capital employed in the first period. (This initial endowment is generated by past savings or is inherited.) Because taxation of the fixed initial endowment is not distortionary, we may assume that the government could efficiently tax away the entire value of the initial endowments. Thus, an individual of type e faces the following budget constraints in periods one and two, respectively: c1 ðeÞ þ sD ðeÞ þ sF ðeÞ ¼ E1 ðeÞ þ T1
ð11:5Þ
and C2 ðeÞ ¼ T2 þ E2 ðeÞ þ SD ðeÞ½1 þ ð1 tD Þr þ SF ðeÞ½1 þ ð1 tF tN Þr ;
ð11:6Þ
Capital Income Taxation in the Globalized World
283
where Et ðeÞ is after-tax labor income (net of the cost of education), t ¼ 1; 2, and where Tt is a uniform lump-sum transfer (demogrant) in period t ¼ 1; 2. That is, E1 ðeÞ ¼
ð1 tL Þð1 eÞw1 g for e a e ð1 tL Þqw1 for e b e
ð11:7Þ
and ð1 tL Þw2 E2 ðeÞ ¼ ð1 tL Þqw2
for e a e for e b e :
ð11:8Þ
An individual can channel savings to either the domestic or foreign capital market, because the economy is open to international capital flows. We denote by SD ðeÞ and SF ðeÞ savings channeled by an e-individual to the domestic and foreign capital markets, respectively. We denote by r and r the real rate of return in these markets, respectively.2 The government levies a tax at the rate tD on capital (interest) income from domestic sources. Capital (interest) income from foreign sources is subject to a nonresident tax at the rate of tN , levied by the foreign government. The domestic government may levy an additional tax on its domestic residents, on their foreign-source income, at an effective rate of tF . Note that tF þ tN is the effective tax rate on foreign-source income of residents. For the sake of brevity, we consider only the case of a capital-exporting country—that is, its national savings exceed domestic investment, with the difference (defined as the current account surplus) invested abroad.3 (The analogous case of a capital-importing country can be worked out similarly.) By arbitrage possibilities, the net-of-tax rates of interest, earned at home and abroad, are equalized; that is, ð1 tD Þr ¼ ð1 tF tN Þr :
ð11:9Þ
Employing 11.9, one can consolidate the two one-period budget constraints 11.5 and 11.6 into one lifetime budget constraint: c1 ðeÞ þ Rc2 ðeÞ ¼ E1 ðeÞ þ RE2 ðeÞ þ T
ð11:10Þ
where R ¼ ½1 þ ð1 tD Þr1
ð11:11Þ
is the net-of-tax discount factor (which is also the relative after-tax price of second-price consumption), and T 1 T1 þ RT 2
ð11:12Þ
284
Assaf Razin and Efraim Sadka
is the discounted sum of the two transfers (T1 and T2 ).4 As usual, the consumer maximizes her utility function, subject to her lifetime budget constraint. A familiar first-order condition for this optimization is that the intertemporal marginal rate of substitution is equated to the tax-adjusted interest factor MRSðeÞ 1 u1 ½c1 ðeÞ; c2 ðeÞ=u2 ½c1 ðeÞ; c2 ðeÞ ¼ 1 þ ð1 tD Þr ¼ R1 ;
ð11:13Þ
where ui denotes the partial derivative of u with respect to its ith argument, i ¼ 1; 2. Equations 11.13 and 11.10 yield the consumption-demand functions c1 ½R; E1 ðeÞ þ RE2 ðeÞ þ T and c2 ½R; E1 ðeÞ þ RE2 ðeÞ þ T of an e-individual. The maximized value of the utility function of an e-individual, v½R; E1 ðeÞ þ RE2 ðeÞ þ T, is the familiar indirect utility function. Denote the aggregate consumption demand in period t ¼ 1; 2 by Ct ½R; ð1 tL Þw1 ; ð1 tL Þw2 ; T ð1 1
ct ½R; E1 ðeÞ þ RE2 ðeÞ þ T dG
0
¼
ð e
ct ½R; ð1 tL Þð1 eÞw1 þ Rð1 tL Þw2 þ T g dG
0
þ ½1 Gðe Þct ½R; ð1 tL Þqw1 þ Rð1 tL Þqw2 þ T;
ð11:14Þ
where use is made of equations 11.7 and 11.8. Note that e is a function of ð1 tL Þw1 and of Rw1 =w2 (see equation 11.2). 11.2.2
Producers
All firms are identical and possess constant-returns-to-scale technologies, so that with no further loss of generality we assume that there is only one firm, which behaves competitively. Its objective, dictated by the firm’s shareholders, is to maximize the discounted sum of the cash flows accruing to the firm. We assume that the firm finances its investment by issuing debt. In the first period, it has a cash flow of ð1 tD Þ½FðK1 ; L1 Þ w1 L1 ½K2 ð1 dÞK1 þ tD dK1 ; where f ðÞ is a neoclassical, constant-returns-to-scale production function. In the second period, the firm has an operating cash flow of ð1 tD Þ½FðK2 ; L2 Þ w2 L2 þ ð1 dÞK2 þ tD dK2 :
Capital Income Taxation in the Globalized World
285
We denote by d both the physical and the economic rate of depreciation (assumed for the sake of simplicity to be equal to each other). This depreciation rate is also assumed to apply for tax purposes. We essentially assume that the corporate income tax is fully integrated into the individual income tax. With such integration of the individual income tax and the corporate tax, there is no difference between debt and equity finance. Specifically, we assume that the individual is assessed a tax (at the rate tD ) on the profits of the firm, whether or not they are distributed, and that there is no tax at the firm level. The firm’s discounted sum of its after-tax cash flow is therefore p ¼ ð1 tD Þ½FðK1 ; L1 Þ w1 L1 ½K2 ð1 dÞK1 þ tD dK1 þ fð1 tD Þ½FðK2 ; L2 Þ w2 L2 þ tD dK2 þ ð1 dÞK2 g=½1 þ ð1 tD Þr:
ð11:15Þ
Note that K1 is the preexisting stock of capital at the firm, carried over from period zero. Maximizing 11.15 with respect to K2 , L1 , and L2 yields the standard marginal productivity conditions FL ðK1 ; L1 Þ ¼ w1 ;
ð11:16Þ
FL ðK2 ; L2 Þ ¼ w2 ;
ð11:17Þ
and FK ðK2 ; L2 Þ d ¼ r:
ð11:18Þ
Note that although taxes do not affect the investment rule of the firm, the taxes are nevertheless distortionary. To see this distortion, consider the intertemporal marginal rate of transformation ðMRTÞ of second-period consumption (namely, c2 ) for first-period consumption (namely, c1 ). It is equal to ð1 dÞ þ FK ðK2 ; L2 Þ. When the economy gives up one unit of first-period consumption in order to invest it, then it receives in the second period the depreciated value of this unit (namely, ð1 dÞ), plus the marginal product of capital (namely, FK ). From equation 11.18, we can see that MRT ¼ 1 þ r: However, from equation 11.13 we can see that the common intertemporal marginal rate of substitution of all individuals is equal to MRS ¼ 1 þ ð1 tD Þr: Hence, the MRT need not equal the MRS; in fact, the MRT is larger than the MRS when the tax rate on capital income from domestic sources ðtD Þ is positive. This violates one of the Pareto efficiency conditions.
286
Assaf Razin and Efraim Sadka
Note that the firm has pure profits (or surpluses) stemming from the preexisting stock of capital, K1 . We denote this surplus by p1 which is equal to p1 ¼ ð1 tD Þ½FðK1 ; L1 Þ dK1 w1 L1 þ K1 :
ð11:19Þ
The surplus consists of the after-tax profit of the first period, plus the level of the preexisting stock of capital. Given the constant-returns-to-scale technology, the firm’s after-tax cash flow consists entirely of this surplus; that is, p ¼ p1 . This equality follows by substituting the Euler’s equation FðK2 ; L2 Þ ¼ FK ðK2 ; L2 ÞK2 þ FL ðK2 ; L2 ÞL2 and the marginal productivity conditions—equations 11.17 and 11.18—into equation 11.15. Naturally, the government fully taxes away the surplus p1 before resorting to distortionary taxation (via the various tS0 ) 11.2.3
Policy Tools: Taxes, Transfers, and Debt
The government has a consumption demand of CtG in period t ¼ 1; 2. We assume that the government can lend or borrow at market rates. With no loss of generality, we assume that the government operates only in the foreign capital market— that is, its first-period budget surplus is invested abroad; for concreteness, suppose that this is positive. Therefore, the government does not have to balance its budget period by period, but only over the two-period horizon C1G þ R C2G þ T1 þ R T2 ¼ tL w1 L1 þ tL R w2 L2 þ tD R rSD þ tF R r SF þ p1 þ tD ½FðK1 ; L1 Þ dK1 w1 L1 ;
ð11:20Þ
where SD ¼
ð1
sD ðeÞ dG
ð11:21Þ
0
is the aggregate private savings, channeled into the domestic capital market; SF ¼
ð1
sF ðeÞ dG
ð11:22Þ
0
is the foreign aggregate private savings, channeled into the foreign capital market; and R ¼ ½1 þ ð1 tN Þr 1
ð11:23Þ
Capital Income Taxation in the Globalized World
287
is the foreign discount rate faced by the domestic economy. Note that the foreign government levies a tax at the rate tN on interest income from the home government budget surplus invested abroad. The left-hand side of equation 11.20 represents the present value of the government expenditures on public consumption and transfers, discounted by the factor R , which is the interest factor at which the domestic economy can lend. The right-hand side of equation 11.20 represents the present value of the revenues from the labor income taxes, the interest income taxes, and the pure surplus of the firm. Market clearance in the first period requires that CA þ C1 þ C1G þ K2 ð1 dÞK1 þ Gðe Þg ¼ FðK1 ; L1 Þ;
ð11:24Þ
where CA is the current account surplus.5 Market clearance in the second period requires that C2 þ C2G ¼ FðK2 ; L2 Þ þ ð1 dÞK2 þ CA½1 þ ð1 tN Þr :
ð11:25Þ
Note that the tax at the rate tN is levied by the foreign country on the interest income of the residents of the home country, and must therefore be subtracted from the resources available to the home country. In order to get one present-value resource constraint, we can substitute the current account surplus, CA, from equation 11.24 into equation 11.25 C1 þ R C2 þ C1G þ R C2G þ K2 ð1 dÞK1 þ Gðe Þg ¼ FðK1 ; L1 Þ þ R FðK2 ; L2 Þ þ R ð1 dÞK2 : 11.2.4
ð11:26Þ
Median Voter Equilibrium
In this model, e b is the only characteristic that distinguishes one individual from another. Recall that the lower is e, the more able is the individual and more objectionable she is to tax hikes. We therefore take the median voter to be the decisive voter. Thus, the political-economy equilibrium tax rates maximize the (indirect) utility of the median voter. Policy tools at the government’s disposal are, inter alia, labor income taxes and capital income taxes. The derivation of the equilibrium is relegated to the appendix. The equilibrium tax on capital income is implicitly given by the following condition: 1 d þ FK ðK2 ; L2 Þ ¼ 1 þ ð1 tN Þr :
ð11:27Þ
288
Assaf Razin and Efraim Sadka
Figure 11.1 The Political-Equilibrium Stock of Capital ðK2 Þ
The political-economy equilibrium stock of capital (implicitly determined from equation 11.27, ascertains the aggregate production efficiency theorem (Diamond and Mirrlees 1971): the intertemporal marginal rate of transformation (which is 1 d þ FK ðK2 ; L2 Þ) must be equated to the world intertemporal marginal rate of transformation faced by the domestic economy (which is equal to 1 þ ð1 tN Þr ). This rule can be illustrated in figure 11.1, where first-period total (private and public) consumption ðC1 þ C1G Þ is plotted on the horizontal axis and secondperiod total consumption ðC2 þ C2G Þ on the vertical axis. Suppose that L1 , L2 , and e were already set at their political-economy equilibrium levels. The production possibility frontier is described by the curve ABD, whose slope is equal (in absolute value) to 1 d þ FK ðK2 ; L2 Þ. The political equilibrium stock of K2 is HD, which gives rise to the consumption possibility frontier given by MBN. Any other level of K2 , say H 0 D, must generate a lower consumption possibility frontier—the curve M 0 B 0 N 0 . Employing the firm’s investment rule (the marginal productivity condition 11.18) and the arbitrage condition (equation 11.9), we can conclude from equation 11.27 that: r ¼ ð1 tN Þr :
ð11:28Þ
Capital Income Taxation in the Globalized World
289
That is, the pretax domestic rate of interest ðrÞ must be equated to the world rate of interest faced by the domestic economy, which is the world rate of interest, net of the source taxes. Equations 11.9 and 11.28 yield the political-economy equilibrium tax on foreign-source income: tF ¼ tD ð1 tN Þ:
ð11:29Þ
Thus, in the political-economy equilibrium, the home country imposes the same tax rate ðtD Þ on foreign-source income from capital as on domestic-source income from capital, except that a deduction is allowed for foreign taxes paid (and levied at source): one dollar earned abroad is subject to a tax at source at the rate tN ; the after-foreign-tax income, which is 1 tN , is then taxed by the home country at the rate tD . The total effective tax rate paid on foreign-source income is therefore tF þ tN ¼ tD þ tN tN tD : 11.3
International Tax Competition and Capital Taxation
A critical issue of taxation in the era of the globalization of the capital markets is the ability, or the inability, of national governments to tax their residents on foreign-source capital income. An editorial in the New York Times (May 26, 2001) underscores the severity of this issue: From Antigua in the Caribbean to Nauru in the South Pacific, offshore tax havens leach billions of dollars every year in tax revenues from countries around the world. . . . The Internal Revenue Service estimates that Caribbean tax havens alone drain away at least $70 billion per annum in personal income tax revenue. The OECD suspects the total worldwide to be in the hundreds of billions of dollars . . . the most notorious tax havens do not even extend their minimal tax rates to their own citizens or domestic enterprises. Their primary aim is to encourage and profit from individuals and businesses seeking to evade taxes in their own countries.
It is fairly safe to argue that tax havens, and the inadequacy of cooperation among OECD national tax authorities in information exchanges, put binding ceilings on how much foreign-source capital income can be taxed. What then are the implications for the taxes on domestic-source capital income? Consider the extreme situation where the home country cannot effectively enforce any tax on foreign-source capital income of its residents. That is, suppose that tF ¼ 0. Then we can see from the political-equilibrium tax rule applying to foreign-source capital income, equation 11.29, that the tax rate on domestic-source capital income, tD , would be set to zero too. Thus the capital income tax vanishes
290
Assaf Razin and Efraim Sadka
altogether. And even if some enforcement of taxation on foreign-source capital income is feasible so that tF does not vanish altogether, it still follows from equation 11.29 that tD ¼ tF =ð1 tN Þ, so that a low tF generates a low tD . Indeed, a poor enforcement of international taxes would generate political processes that would reduce significantly the domestic-source capital income taxation. The unwillingness of foreign tax authorities to cooperate with the home tax authority in helping to enforce capital taxation on home country residents’ capital income originating abroad usually stems from their desire to lure capital to their countries. This is what is meant by tax competition. They further compete with the home country by lowering the source tax ðtN Þ they levy on the capital income of the home country residents. We thus capture formally the effect of tax competition by assuming that tN falls as foreign governments lure capital to their countries. Then we can see from equation 11.28 that r, the net (of depreciation d) marginal product of domestic capital, must rise. With diminishing marginal product, this must happen when the stock of domestic capital falls and more capital flows abroad. Hence, the tax base for the domestic-source capital income shrinks, thereby turning the enforcement of foreign-source capital income all the more acute. 11.4
Conclusion
The behavior of taxes on capital income in the recent decades points to the notion that international tax competition that follows globalization of capital markets puts strong downward pressures on the taxation of capital income—a race to the bottom. This behavior has been perhaps most pronounced in the EU-15 following the single market act of 1992. (See, for example, Razin and Sadka 2005.) The 2004 enlargement of the EU with ten new entrants put a strong downward pressure on capital income taxation for the EU-15 countries. Table 11.1 describes the corporate tax rates in the twenty-five EU countries in 2004. It reveals a marked gap between the original EU-15 countries and the ten accession countries. The latter have significantly lower rates. Estonia, for instance, has no corporate tax, the rates in Cyprus and Lithuania are 15 percent, and Latvia, Poland, and Slovakia are at 19 percent. In sharp contrast, the rates in Belgium, France, Germany, Greece, Italy, and the Netherlands range from 33 percent to 40 percent. Europe’s new constitution, adopted by the European Summit in Brussels, June 19, 2004 (the text, however, has yet to be ratified in all twenty-seven member states), cannot stop the tax competition process because the new constitution retains the national veto in tax cooperation and harmonization. But, there is flexibility for some countries that want to push ahead tax harmonization to do so. In-
Capital Income Taxation in the Globalized World
291
Table 11.1 Statutory Corporate Tax Rates in the Enlarged EU, 2003 Country
Tax Rates (%)
Austria Belgium
34 34
Cyprus*
15
Czech Republic*
31
Denmark
30
Estonia*
0
Finland
29
France
33.3
Germany Greece
40 35
Hungary*
18
Ireland
12.5
Italy
34
Latvia*
19
Lithuania*
15
Luxembourg
22
Malta* Netherlands
35 34.5
Poland*
27
Portugal
30
Slovakia*
25
Slovenia*
25
Spain
35
Sweden
28
UK
30
* New entrants.
deed, Germany and France are currently pushing some of the new entrants to raise their corporate tax rates. Tax competition within the EU is in sharp contrast to the U.S. federal fiscal system, where the capital income tax (on individuals and corporations) is federal and not state specific. But both the EU and the U.S. are subject to severe tax competition from the rest of the world. 11.5
Appendix: Derivation of the Political-Economy Equilibrium
The political-economy equilibrium tax rates maximize the (indirect) utility of the median voter. Denoting the indirect utility function of the median voter by V, it is given by:
292
Assaf Razin and Efraim Sadka
VðeM ; R; w1N ; w2N ; TÞ ¼
v½R; ð1 eM Þw1N þ Rw2N þ T g v½R; qðw1N þ Rw2N Þ þ T
if eM < e if eM > e ;
where wtN ¼ ð1 tL Þwt is the after-tax wage per efficiency unit of labor in period t ¼ 1; 2, and eM is the innate ability parameter of the median voter. Policy tools at the government’s disposal are, inter alia, labor income taxes and capital income taxes. We therefore assume that the government can effectively choose the aftertax wage rates (w1N and w2N ) and the after-tax discount factor ðRÞ. The government can choose also T, the discounted sum of the lump-sum transfers (T1 and T2 ). Once w1N , w2N , R, and T are chosen, then private consumption demands [C1 ðR; w1N ; w2N ; TÞ and C2 ðR; w1N ; w2N ; TÞ] are determined. The cutoff level, e , and labor supplies, L1 and L2 , are also determined as follows: e ðR; w1N ; w2N Þ ¼ ð1 qÞ½1 þ Rw2N =w1N g=w1N L1 ðR; w1N ; w2N Þ ¼
ð e ðR; w N ; w N Þ 1
0
2
ð1 eÞ dG þ qf1 G½e ðq; w1N ; w2N Þg
L2 ðR; w1N ; w2N Þ ¼ G½e ðR; w1N ; w2N Þ þ qf1 G½e ðR; w1N ; w2N Þg:
ð11:2 0 Þ ð11:3 0 Þ ð11:4 0 Þ
In choosing its policy tools (R, w1N , w2N , and T) and its public-consumption demands (C1G and C2G ), the government is constrained by the economy-wide ‘‘budget’’ constraint 11.26, where C1 , C2 , L1 , L2 , and e are replaced by the functions C1 ðÞ, C2 ðÞ, L1 ðÞ, L2 ðÞ, and e ðÞ, given by equations 11.14 and 11.2 0 –11.4 0 , respectively. Note that the capital stock in the first period ðK1 Þ is exogenously given. The capital stock in the second period ðK2 Þ must satisfy the investment rule of the firm (equation 11.18). Note that because the economy is financially open, the individuals, by the arbitrage condition (equation 11.9), are indifferent between channeling their savings domestically or abroad. This means that the government can choose K2 , and then r and the pretax wages (w1 and w2 ) are determined so as to clear the capital market and labor market in each period through equations 11.18, 11.16, and 11.17, respectively. This does not mean that the government actually chooses the stock of capital ðK2 Þ for the firm, or the pretax wage rates (w1 and w2 ), or the domestic interest rate ðrÞ. Rather w1 , w2 , and r are determined by market clearance, and the firm chooses K2 so as to maximize its value. What we did is to determine K2 , w1 , w2 , and r at levels that are compatible with firm-value maximization and market clearance in the presence of taxes. To sum up, the government in a political-economy equilibrium chooses C1G , C2G , R, w1N , w2N , T, and K2 so as to maximize the utility of the median voter (as given by equation 11.29), subject to the economy-wide ‘‘budget’’ constraint, equation
Capital Income Taxation in the Globalized World
293
11.26. Note that C1 , C2 , L1 , L2 , and e in the latter constraint are replaced by the functions C1 ðÞ, C2 ðÞ, L1 ðÞ, L2 ðÞ, and e ðÞ, respectively. We can ignore the government budget constraint 11.20 by Walras Law. Note that in this maximization, K2 appears only in the economy-wide ‘‘budget’’ constraint, equation 11.26. Thus the first-order condition for the political-economy equilibrium level of K2 is given by 1 R FK ðK2 ; L2 Þ R ð1 dÞ ¼ 0: Note that this choice does not depend on whether the median voter is skilled or unskilled. Substituting the firm’s investment rule, equation 11.18, and rearranging terms yields 1 d þ FK ðK2 ; L2 Þ ¼ 1 þ ð1 tN Þr : Notes This chapter is based on a keynote address to the annual congress of the International Institute of Public Finance (IIPF), Prague, August 25–28, 2003. 1. Evidently, if the tax is progressive, the payoff would be reduced proportionally more than the foregone-income cost. 2. These rates (r and r ) hold in essence between periods one and two and we therefore assign no time subscript (one or two) to these rates. 3. Evidently in a nonstochastic set-up like ours, the country is either capital exporter or capital importer. 4. Note that even though T may seem at first glance to be dependent on tD (through the discount factor R), we may nevertheless assume that these are two independent policy tools because the government can always change either T1 and T2 in order to keep T constant when it changes tD . 5. For notational simplicity, we assume that the net external assets are initially equal to zero, so that there is no initial external debt payment term in the current account.
References Baldwin, Richard and Paul Krugman. 2000. ‘‘Integration and Tax Harmonization.’’ Discussion Paper No. 2630, CEPR, London. Diamond, Peter A., and James A. Mirrlees. 1971. ‘‘Optimal Taxation and Public Production.’’ American Economic Review 61, no. 1: 8–27 and no. 3: 261–278. Krogstrup, Signe. 2002. ‘‘What Do Theories of Tax Competition Predict for Capital Taxes in EU Countries? A Review of the Tax Competition Literature.’’ Working Paper No. 05/202, Graduate Institute of International Studies, Geneva, Switzerland.
294
Assaf Razin and Efraim Sadka
Oates, William E. 1972. Fiscal Federalism. New York: Harcourt Brace Jovanovich. Obstfeld, Maurice, and Alan M. Taylor. 2003. ‘‘Globalization and Capital Markets.’’ In Globalization in Historical Perspective, eds. Michael D. Bordo, Alan M. Taylor, and Jeffrey G. Williamson, 121–124. Chicago: University of Chicago Press. Razin, Assaf, and Efraim Sadka. 1991. ‘‘International Tax Competition and Gains from Tax Harmonization.’’ Economics Letters 37, no. 1: 69–76. Razin, Assaf, and Efraim Sadka, with the cooperation of Chang Woon Nam. 2005. The Decline of the Welfare State: Demography and Globalization. Cambridge, MA: MIT Press.
12
Can Public Discussion Enhance Program ‘‘Ownership’’? Allan Drazen and Peter Isard
12.1
Introduction
IMF programs have a mixed record of success. A significant share of programs fail to reach their originally scheduled completion dates, many others require substantial modifications along the way,1 and many countries remain under IMF programs for prolonged periods of time.2 An oft-cited reason that Fund programs get off track is a lack of ownership on the part of the program country, where ‘‘ownership’’ refers loosely to the extent to which a country is committed to the general reform process and the conditions specified by the program independently of the incentives provided by multilateral institutions. Genuine ownership on the part of a country, combined with sufficient lending from the Fund and good program design, is believed to yield program success (or at least a high probability of success). And indeed, when the Fund refers to ownership it defines the concept in a circular way, based implicitly on the vision of reaching program completion dates: ‘‘National ownership refers to a willing assumption of responsibility for a program of policies, by country officials who have the responsibility to formulate and carry out those policies, based on an understanding that the program is achievable and is in the country’s best interest.’’3 The main challenges that arise when wrestling with the concept of ownership do not stem primarily from disagreement over the most appropriate definition, but rather reflect the difficulties in making the concept of ownership operational.4 How can the Fund determine whether a country is likely to remain committed to the reform effort and comply with the conditions of its program through the completion date? And how can the Fund design programs to enhance the prospect that countries will remain committed and completion dates will be reached? These questions identify and distinguish between two central issues: demonstrating that ownership is present and building or creating ownership when it is thought to be weak. The former issue is perhaps most relevant for countries that
296
Allan Drazen and Peter Isard
have failed to comply with program conditions in the past and must demonstrate that, contrary to past performance, they are now truly interested in and committed to reform.5 The latter issue recognizes that ownership is endogenous, not an innate or exogenous characteristic. Ownership can be improved and fostered through various devices. That is why we argue that a key challenge faced by a reform-minded government is building ownership—that is, creating support for the process of reform and for the specific measures embodied in a Fund program. Among the general public and especially critics of the IMF (and the World Bank), the lack of ownership and the failures of Fund programs are often attributed to a lack of public discussion of the programs during their formulation stages. Our use of the term public discussion in this paper refers broadly to mechanisms through which policy makers disseminate and exchange information with the public, including discussions both to formulate programs and to explain program design. It appears to be similar what others have referred to as ‘‘participation’’ in the recent literature on economic development and poverty and social impact analysis. Critics of the IMF often characterize the process of formulating a program as one of negotiations between the IMF (seen as having its own agenda) and a thin layer of high-level government officials, with little or no input from the public, from nongovernmental organizations (NGOs), or from other interested parties. The failure of these programs to lead to a significant bettering of the general population’s economic condition is often seen as reflecting this lack of public input. Demonstrations (or riots) against the IMF and its programs are taken as evidence of the seriousness of the problem. Public discussion is considered an important vehicle for establishing ownership, and lack of public discussion is perceived as part of the reason that some programs are never adopted or fail to reach completion. The need for public discussion is recognized within the IMF and World Bank.6 To date, however, the case for public discussion is not much more developed than the cases for other terms that sound unambiguously positive. In particular, the discussion of ‘‘public discussion’’ seems quite unfocused and lacking in coherent analysis of what form it should take and how it can contribute to strengthening Fund programs or demonstrating and building ownership. Such lack of careful, analytic argument about the role of public discussion can leave the unfavorable impression that the IMF and World Bank simply view public discussion as politically correct rather than as a vehicle that can significantly improve program design—or that they engage in public discussion primarily to make the public feel more comfortable, or warm and fuzzy, about their programs. A more serious concern is that the lack of a clear understanding of how public discussion works can leave the impression that public discussion strengthens ownership only
Can Public Discussion Enhance Program ‘‘Ownership’’?
297
by leading counterproductively to the adoption of weaker programs. At worst, public discussion is seen as making strong program design very difficult or impossible by giving too much influence to groups that are either against real reform or hold views strongly at odds with generally accepted economic wisdom. Thus, it is not surprising that public discussion has a bad name in some circles (at least when it is not being discussed publicly). Both the association of public discussion with fuzzy thinking and the view that it detracts from successful program design are especially unfortunate. In our view, careful public discussion of programs can increase necessary public support and in any case is often unavoidable. The purpose of this paper is to distinguish a number of different functions that public discussion can serve and to illustrate its potential for improving collective choice. This paper does not break new theoretical ground, but it uses the tools of economic theory and political science to develop a better understanding of how public discussion can contribute to building and demonstrating ownership. The first task is to develop a better understanding of both ownership and public discussion. In section 12.2, we argue that ownership is more complex than many discussions of it would suggest. Ownership must include not only willingness, but also the technical capacity and political ability to carry out a program (especially the political ability.) In section 12.3, we argue that public discussion can serve a number of purposes—including educating the public, revealing public preferences and constraints (or educating policy makers), demonstrating the unbiasedness of policy makers, and finding common ground among heterogeneous interests— each of which can be better understood by moving to a more formal treatment. What public discussion can or cannot achieve depends on what the ownership problem is, and the different purposes of public discussion may interact with one another. Here too, understanding the political constraints affecting a government may help us understand what to expect from public discussion. In section 12.4, we illustrate our arguments using simple examples relevant to actual program design. In section 12.5 we consider some of the drawbacks of public discussion, especially as applied to Fund programs. After having developed a clearer sense of ownership and a more formal understanding of what public discussion can achieve in different types of circumstances, we focus in section 12.6 on the design of IMF programs and the role that public discussion can play in enhancing program ownership. We argue that the difficulties in attaining program ownership depend on the nature of the conditions involved, that this should be taken into account in designing programs and seeking to build overall program ownership, and that the recent emphasis on streamlining Fund conditionality can be counterproductive if it results in the elimination of conditions for which ownership is potentially very high. We also suggest that
298
Allan Drazen and Peter Isard
the effectiveness of both Fund programs and the Fund’s surveillance of nonprogram countries might be enhanced through more explicit focus ex ante on the obstacles to building ownership for policies and programs, and on how those obstacles can be overcome or mitigated through public discussion. Section 12.7 provides a summary of our arguments and conclusions. 12.2
Defining Ownership
Ownership is a more complex concept than it may seem. As noted above, the degree to which a country ‘‘owns’’ a Fund program can be broadly defined as the extent to which the country is oriented toward meeting the program conditions, independent of any incentives provided by multilateral lenders. Alternatively, and from the perspective of seeking to draw meaningful implications for Fund program design and the role of public discussion, the degree of ownership can be defined as the probability that a Fund program will be adopted and reach completion without a significant weakening of the reform effort. The latter definition points to a number of factors on which ownership depends. One is the willingness of the government to meet the conditions of the program. Without such willingness, the probability of reaching completion is unlikely to be high. Suppose, however, that there is willingness, but a lack of technical capacity to collect taxes or carry out other measures on which the program crucially depends. (‘‘The spirit is willing, but the infrastructure is weak.’’) It would not seem meaningful to suggest that a country owns a program that it has no hope of carrying out: ownership also depends on technical capacity. Conversely, ownership for its own sake is not a desirable objective. Just as it is not very meaningful to design diets that individuals are willing and able to undertake because they require only minimal changes in eating habits, it would not be very desirable to design programs that countries are willing and technically able to carry out simply because they require only minimal changes in policy. A third crucial factor for ownership is the political ability of the government to carry out a meaningful program. To illustrate the point, suppose policy makers in a country with continuing large fiscal deficits want to implement a program of fiscal austerity, but face powerful special interest groups (SIGs), either inside or outside government, who can block fiscal reform. One could not argue that there is country ownership if the SIGs do not agree to the fiscal austerity program, even if the government is willing and technically able to carry the program out.7 The same argument suggests that ownership can be strengthened by weakening either the power of or the incentives for SIGs to block the program. The nature of the political system can very much determine the power of SIGs, while both the design
Can Public Discussion Enhance Program ‘‘Ownership’’?
299
and public discussion of programs can affect their incentives. We return to the latter point in our analysis of how public discussion can affect ownership. Willingness, technical capacity, and political ability are not independent of each other as determinants of program ownership. The distinction between the willingness of the government to carry out reforms and its political ability is often unclear, as is the distinction between its technical capacity and its political ability to collect taxes or discipline public spending. In some cases, distinguishing between technical and political capacity may be unimportant or irrelevant. One may want to think simply of ‘‘institutional capacity,’’ broadly defined to reflect the technical characteristics of fiscal processes as well as the nature of the political institutions that influence what comes out of fiscal processes. The Fund is unlikely to find itself involved in a program in which country ownership is ex ante complete. Were ownership complete—that is, if a country had the willingness, technical capacity, and political ability to pursue serious reforms—it would not have to approach the Fund for support.8 In this sense, the challenge faced by the Fund in negotiating and designing programs and in considering how to make effective use of public discussion is not simply to try to determine the degree of ownership, but also to try to increase—that is, to build and strengthen—ownership. This leads to the question of the role of public discussion in designing effective programs and especially in enhancing ownership. 12.3
A Framework for Understanding the Role of Public Discussion
Public discussion has an image problem: calls for public discussion are often perceived to be based primarily on the desire to increase ownership by making the public feel more comfortable about a program in some general way. Our aim in this paper is to move away from the warm and fuzzy and to try to analyze public discussion more rigorously. From a more analytical perspective, public discussion of a program may be seen as having a number of functions, which may be grouped into four categories: To educate the public about the nature of the program and, more generally, about macroeconomic realities, including the trade-offs between short-run costs and medium-run benefits and the implications of the program for relevant interest groups. It can also provide a vehicle for the government (and/or the Fund) to make the public aware of information on which the program is based;
•
To learn public preferences and constraints, and to convince the public that program design takes this information into account, which (along with educating the public) may enhance compliance and cooperation with the program;
•
300
Allan Drazen and Peter Isard
To demonstrate unbiasedness—that is, to convince the public that the program is designed for the general good, rather than to serve the interests of the authorities or the Fund;
•
•
To find common ground among heterogeneous interests.
‘‘Public discussion’’ might appear to be a topic not easily formalized or discussed with any rigor. However, there are two highly relevant strands of literature—one in economics, the other in political science—that are rigorous and, we believe, potentially quite relevant for developing a more concrete awareness of the different ways that public discussion can affect policy outcomes. As we discus in section 3.1, the economics literature on cheap talk is useful for illustrating how discussion can help country governments and/or the Fund learn about preferences and constraints, demonstrate their unbiasedness, or find common ground among different interest groups. And the political theory literature on deliberative democracy is helpful in thinking about the educational role of public discussion. 12.3.1
Cheap Talk
Economic theory distinguishes between costly signaling mechanisms and ‘‘cheap talk,’’ where the latter is defined as a signal an agent can send that is neither costly nor binding and that does not directly affect the payoffs associated with any given outcome.9 As an example of costly signaling, suppose that two jobs require different levels of ability, but that individual ability is unobserved. If innately high-ability individuals find it less costly to become educated than do low-ability individuals, observable educational attainment may serve as a signal of unobserved ability. The signal works because it is differentially costly for different types of individuals to send it. Suppose, however, that individuals simply announce to employers what their ability is, it being costless for an individual to make any announcement he wants (hence the term ‘‘cheap talk’’). One might think that such costless signals are completely uninformative. However, under certain conditions, cheap talk can in fact convey information about unobserved characteristics. Cheap public discussion may enable the government to convey the information that its objective is maximizing social welfare (rather than having its own agenda), and cheap talk may enable the public to convey information about its preferences to those who design programs. We return later to this application of public discussion. Cheap talk may also be a useful coordination mechanism; that is, a device for finding common ground. The challenge of program design presents the govern-
Can Public Discussion Enhance Program ‘‘Ownership’’?
301
ment not only with the difficulties of convincing the populace of its good intentions, but also with the need to forge an agreement among different special interest groups. We also consider how public discussion applies to the latter task. Note that in considering the importance of talking we have not made any reference to bargaining or negotiations per se. The issue of how bargaining affects outcomes can, of course, be analyzed formally, but that is not our intention in studying public discussion. 12.3.2
Deliberative Democracy
Political science has generated quite a bit of literature on communication between political leaders and the public, and on the role of such communication in shaping policy decisions and generating support for policies and political leaders. Much of this literature is empirical or illustrative, arguing the importance of communication but without providing a theoretical or formal way to understand why public discussion may be important. There is, however, a rigorous strand of literature focusing on why public discussion may be important in a democracy.10 This is the analysis of ‘‘deliberative democracy,’’ which can be roughly defined as ‘‘decision making by discussion among free and equal citizens’’ (Elster 1998, 1), with the stress on ‘‘discussion.’’ According to this literature, a crucial part of the democratic process is discussion and deliberation before a collective choice is made. The central question in assessing discussion is put clearly by Fearon (1998): What good reasons might a group of people have for discussing matters before making some collective decision, rather than simply voting on the issue or using some other decision rule that does not involve discussion? In other words, what is the point or value of discussing things before making political decisions? (44, italics in original)
Advocates of deliberative democracy argue that decision-making in a democracy is not simply the aggregation of (existing) policy preferences among heterogeneous agents in a society. That is, it is not simply a collective choice rule (such as majority voting) for reaching an aggregate policy decision when individual citizens do not agree on their most preferred policy. Instead, and perhaps more importantly, democratic decision-making involves the transformation of individual preferences through deliberation and discussion (Habermas 1987).11 In common parlance, if two individuals have different opinions, the free and open discussion of issues that is part of the democratic process enables each side to attempt to convince the other of the correctness of its own position. Thus, discussion provides a mechanism for trying to narrow differences of opinion. In the limit, deliberation
302
Allan Drazen and Peter Isard
and discussion would bring about a consensus of opinion so that there would be no need to use voting or some other mechanism to aggregate preferences. In applying this perspective to Fund programs, the relevant question is, how can public discussion lead to a consensus, or at least a narrowing of differences of opinion, on what policy should be followed? It is crucial to note that the focus here is on policies and not on outcomes. Two constituencies may differ in their desires for narrowing income inequality, for example, as an outcome, or, they may agree on the desired degree of income inequality, but differ over the best way to achieve it. Hence, their disagreement is not about goals, but about the best way to achieve them.12 They disagree about the connection between policies and outcomes, or more generally, about how the world works. We term this an issue of causal policy-to-outcome relations, or simply ‘‘causal relations.’’ We note this point since—in the context of Fund programs—the possibility of educating the public about the effects of policies is a key consideration. That is, public discussion can make the public aware of the payoff matrix they face, and thus perhaps lead them to choose to comply with the proposed program (to choose ownership) as leading to better outcomes for all parties than the outcomes associated with noncompliance. In addition to the positive question of whether deliberation can lead to ownership, there is a normative question of whether deliberation leads to welfareimproving outcomes. As already mentioned in the introduction, some observers stress the possible negative implications, a point we consider in detail in section 5. 12.3.3
Voting versus Program Design
Many discussions of deliberative democracy focus on the role of public discussion prior to voting on policy. By contrast, we are interested in the value of public discussion of policy proposals before a government commits to a policy program with the IMF. The fact that public discussion is followed by the government adopting a program, rather than voting, is a key distinction between our framework and most applications in the political theory literature. This difference may have implications for the functions of public discussion, as well as for the pathologies (the circumstances in which public discussion may be counterproductive). A second factor distinguishing our framework, which may influence the relative importance of various arguments, is that IMF programs—especially difficult ones that need the most public discussion—are typically designed when the economic outlook is particularly bad. As we shall see, these features of our framework introduce some considerations not present in a voting setup.
Can Public Discussion Enhance Program ‘‘Ownership’’?
12.4
303
Illustrating the Functions of Public Discussion
Public discussion is a vehicle for transmitting information. This includes both economic information about the state of the economy and the links between policies and economic outcomes and sociopolitical information about the preferences and agendas of the government, relevant national constituencies, or the Fund itself. This section focuses on how the design and ownership of programs can be strengthened by the transmission of such information. Our purpose is not to break new theoretical ground, but to illustrate these points by means of simple examples as applied to actual Fund programs. In analyzing the various functions that public discussion may serve, it is useful to address three separate questions sequentially. First, as a reference case, what are the functions of public discussion when the government is known to maximize social welfare (rather than being thought to have its own agenda) and the public is homogeneous (rather than having heterogeneous interests)? Second, how can public discussion help convince the public that the government’s objective is in fact the maximization of social welfare (thus presumably increasing public support and compliance)? Third, what additional functions might public discussion have in addressing the conflicting interests of groups in society? Throughout this section we restrict attention to communication between two parties, focusing on communication either between the government and the public or between two interest groups. We assume that the Fund’s objectives coincide with those of the government.13 This assumption implies that the argument that the Fund’s interests do not coincide with those of countries entering programs can be captured by case 2 (section 4.2) on the role of public discussion when the authorities in a country may not be acting in the interests of social welfare. 12.4.1 Case 1: Government Known to Maximize the Social Welfare of a Homogeneous Public Consider first the case where it is common knowledge that the government’s objective is the maximization of social welfare and the public is homogeneous in its preferences and beliefs.14 These two assumptions are admittedly extreme and unrealistic, but this case is useful in understanding the importance of public uncertainty about the government’s objectives and of heterogeneity in agents’ objectives. In this reference case we ask, since the public knows that the government is maximizing its welfare, why isn’t it effective for the government simply to announce its program without any prior discussion? Our answer is that achieving
304
Allan Drazen and Peter Isard
the optimal outcome may depend on a number of functions that public discussion can serve: (1) revealing the public’s preferences; (2) convincing the public that the government is aware of what it values; (3) educating the public; (4) revealing information about available resources and/or efficient choices; and (5) making the government’s commitment to a program more credible. When the government has imperfect information about what the public wants, public discussion can reveal information. This can be a serious problem when program negotiation involves only the Fund and a thin layer of high-level government officials. Since preferences are necessarily complex, it can be argued that the government and the Fund cannot possibly design a program that is socially optimal (or perceived as such) without consulting the public. Compared to other ways of forming judgments about ownership, public discussion has the attraction of being a relatively quick and resource-efficient way to seek feedback.15 We may illustrate our points in simple matrices of payoffs to the government and a homogeneous public. Consider matrix 12.1. (In this case the use of a formal model is perhaps trivial, but it sets the stage for later examples.) The public’s type is left, but the government does not know this. (Formally, the game is either left or right, which public knows but government does not.) For example, left means relatively large tax-financed social welfare programs and right means relatively small tax-financed social welfare programs. A policy of high (government spending) is optimal when the public prefers left and a policy of low (government spending) is optimal when the public prefers right. (Since the government is known to maximize social welfare, the payoffs are identical to both public and government in each cell.) Matrix 12.1 Public Government Program
Left
Right
High
3,3
1,1
Low
1,1
3,3
The government must choose the optimal policy when it does not know what the public wants. In the absence of communication, the government must choose on the basis of its prior beliefs about the public’s preferences (where we assume that once government chooses a policy it cannot change). Suppose government
Can Public Discussion Enhance Program ‘‘Ownership’’?
305
assigns probability of 1/2 to each possibility (actually, any interior probabilities will suffice for this example), so that the expected payoff is ð2; 2Þ. Here, the problem is solved if the public simply announces that it is left, which it is optimal for it to do. Public discussion increases welfare since it allows the public to credibly signal its preferences even though such discussion is cheap talk. (Formally, the public’s preferences are self-signaling.)16 A related role for public discussion is to convince the public that the government is really aware of what the public values. This is relevant when the public knows that the government’s objective is maximization of social welfare, but feels the government may not be aware of society’s preferences. This is clearly related to the previous argument, but it is not identical: the previous argument applies when the government does not have information on the public’s preferences; the second role recognizes that even a perfectly informed government must sometimes convince people that it is informed. Put another way, in the first case public discussion is used to shape a program; in the second, to help sell a program. Formally, the public’s preferences must not simply be known, but must also be common knowledge. This possibility suggests one interpretation of the notion that people want to be included in program formulation—in this case, enough so to be convinced that the decision maker really does know their preferences. In other words, public discussion may serve to increase the legitimacy of policy. Another function of public discussion is to address the possibility that the general public knows the government’s preferences (or at least knows that the government’s objective is the maximization of its welfare) but is uninformed about how the world works; that is, about what we earlier termed causal relations.17 In other words, the function of public discussion is to educate the public about economic realities. Przeworski (1998), for example, argues that transmitting knowledge about such causal relations is central to what deliberation is supposed to do: The preferences on which people act are contingent on their beliefs about the consequences of their actions. And, indeed, parties and candidates competing for office do not offer merely policies: they explain to the electorate how these policies will affect the outcomes, trying to persuade citizens that their policies, as distinct from those of their opponents, will lead to the outcomes they want. (144)
We may represent this case by making a number of changes to matrix 12.1. First, left and right now refer not to the public’s type, but to the state of the world. Second, it is now the government that has superior information—it knows the state of the world and the associated optimal policy, while the public does not. For purposes of the example, let the state be right, calling for a policy of low government spending. (That is, in matrix 12.1, the true ‘‘game’’ is only the right-hand column.) Third, a government program can be enacted only if the public favors it.
306
Allan Drazen and Peter Isard
(Hence, it is the public that chooses high or low.) Finally, the public does not know the true payoffs connected with different policies. As a concrete example, one may think of the state of the world as overheated, with the public not fully understanding the longer-run effects of countercyclical and procyclical policies. Though the true payoffs are as in matrix 12.1, the public believes that the payoffs associated with different policies are as in matrix 12.2. Matrix 12.2 State of the World Government Program
Left
Right
High
2,2
1,1
Low
1,1
2,2
In this case, the public would favor high government spending even if the government announced that the state was right and that announcement was believed. Hence, though the government has superior information, it must convince the public not only about the state of the economy, but also about what works. Once it does that (that is, once it convinces the public that the payoffs are those in matrix 12.1 rather than matrix 12.2), then it need simply announce the state of the world and a policy of low spending would be supported. The educational function of public discussion may be especially important in countries where IMF programs have historically been viewed as ‘‘bitter medicine prescribed by an unsympathetic international financial community,’’ as represented by the public belief in payoffs described by matrix 12.2. In these cases, public discussion has the potential to help identify the truth by sifting through whatever valid or invalid inferences have been drawn from historical experience. Educating the public about causal relations may also be important to ensure compliance with a program. A program may only be successful if the public complies with it, which depends on the public understanding why it is beneficial. To illustrate, suppose left and right in the previous two matrices now refer to actions the public may take, high and low to actions the government takes. The government moves first, but knows how the public will react (that is, whether the public will comply with a government program). Both the public’s and the government’s actions are fully observable. Suppose the true payoffs are as in matrix 12.3, but the public believes that matrix 12.2 represents the true payoffs.
Can Public Discussion Enhance Program ‘‘Ownership’’?
307
Matrix 12.3 Public Actions Government Program
Left
Right
High
2,2
1,1
Low
1,1
4,4
The public would then choose left, no matter what the government chose. The government would be induced to choose high. If the public knew the true payoffs to their actions, they would choose right, and comply with a program of low government spending. Consider, for example, a program of fiscal restraint that requires replacing government supply of goods (high government spending) with reliance on the private sector (low government spending) for supply. Success of the market solution requires public compliance. Public willingness to rely on the market rather than government is represented by right rather than left. Matrix 12.2 thus might represent public distrust of the market’s ability to supply services, combined with an unwillingness to use the market. Public education thus means educating the public on the value of greater reliance on the market when there is a need to reduce the size of government, with matrix 12.3 showing the value of the public acting in accordance with this view. The functions of public discussion in revealing information to the government and in eliciting optimal public responses interact when it enables the government to learn about the resources the public has. That is, public discussion can also play an important role in educating the government and identifying policy choices that are likely to be most efficient, or elicit the greatest ownership, when there are a number of options. (We leave it to the reader to combine the previous cases to demonstrate this formally.) Governments are rarely (if ever) fully informed of the effects of their policies on society, and there are endless examples of situations in which public discussion has made governments aware of adverse policy effects on certain groups of society and of the possible desirability of policy adjustments or compensating policy actions. This may be seen in the broad consultation that takes place among stakeholders and development partners during the process of preparing country strategy papers under the IMF and World Bank’s poverty reduction initiative. For example, public discussion made governments aware of the regressive nature of local graduated taxes in Uganda and of commune-level
308
Allan Drazen and Peter Isard
revenue-raising measures in Vietnam, of the effects of energy price liberalization on the livelihoods of fishing communities and of low-paid casual laborers who were heavily affected by transportation costs in Mozambique, and of the scope for building ownership by addressing these public concerns.18 In addition to making the government more aware of the effects of its policies, public discussion and participation sometimes identify resources that can be mobilized to pursue policy objectives more efficiently.19 Yet another function of public discussion is to induce the government to make a greater commitment to a program, and hence make it more credible that it will be carried through. One way to think of this is in terms of solving a time consistency problem. By taking publicly announced positions, the government changes the payoff structure associated with following a given policy in order to commit itself to a specific course of action. For example, following a severe banking and currency crisis that led to near hyperinflation in early 1997, and in the context of negotiations with the IMF, the newly elected president of Bulgaria engaged in discussions with major political parties to forge a commitment to an economic program that included, inter alia, the introduction of a currency board. The commitment resulting from the discussions was formalized in a declaration proclaiming the readiness of the major political parties to cooperate in fulfilling the agreement with the Fund. Elections in April 1997 shifted political power to a new majority, but the negotiated program continued to enjoy broad national ownership and was endorsed by the new parliament.20 12.4.2
Case 2: Public Uncertainty about Government Objectives
We now focus on the issue of uncertainty about the government’s objectives in the design of a program. That is, the government faces the problem of convincing the public it is acting in the public’s best interests, rather than having its own agenda. Though the imperfect information issues discussed in the previous case may still be present (that is, where the government does not know the public’s preferences), we assume these away in order to focus on the problem of the government making it credible to the public that it is acting in its best interests. We also continue to assume that the public is homogeneous (though not necessarily informed about what policies would maximize their welfare). If the public believes that the government’s preferences are not correlated with its own, then cheap talk may fail to convey information. Specifically, suppose that only the government knows the state of the economy, high or low, which the public cannot observe, though it knows how the world works (that is, the true payoff matrix). Let us interpret more and less as the public’s action (say, the level of
Can Public Discussion Enhance Program ‘‘Ownership’’?
309
wage demands). Public discussion (cheap talk), where the government announces the state of the economy, precedes any actions. We now assume that the government’s and the public’s objectives can differ, which might reflect, for example, the weight government puts on the reaction of global capital markets. Hence, suppose that instead of a payoff matrix such as matrix 12.1, payoffs were described by matrix 12.4 (where the government’s payoff is listed first). Matrix 12.4 Public Behavior State of Economy
More
Less
High
1,3
5,1
Low
6,3
3,2
In this case, the government would like the public to play less, no matter what the economic situation is. Hence, government would like the public to believe that the state of the economy is low. But the public, knowing the payoff matrix, knows government has this incentive and so does not believe any government announcement. Talking doesn’t reveal the government’s information because it is believed to have an objective other than the maximization of social welfare—that is, to have different preferences than the public. However, even if the preferences of government and public are not perfectly correlated, information can be conveyed, but the type of message sent will be crucial in determining whether it is successfully conveyed. Suppose, more realistically, that the state of the economy is not discrete as in matrix 12.4, but is continuous. Suppose the payoffs are such that government wants the public to believe that the state of the economy is somewhat worse than it really is (perhaps to give them an incentive to accept lower wages as part of a tough program). For simplicity, suppose that if the state of the economy was S, the government would want the public to believe that the state was S x, where this incentive is known to the public. One might think that cheap talk can convey no information here, since if the public discounts the government’s announcement by some amount y (that is, in response to an announcement of S x, public believes the state is S x þ y), government will simply announce S x y. A key result in the cheap talk literature (Crawford and Sobel 1982) is that imprecise messages can convey information as long as the incentive to distort
310
Allan Drazen and Peter Isard
information (in the example above, the amount x by which the government wants to understate S) is not too large. That is, while a supposedly precise announcement of the state of the economy will not be believed, a crude announcement of the state of the economy, such as simply high or low, even when the state is continuous, will convey information. This result has applications not only to cases in which the public is uncertain about the government’s preferences, but also to cases in which the public knows that the government has different preferences, as long as the difference in preferences is not too large. In such situations, trying to get the public to comply with government-recommended policies may be difficult when the choice set includes a continuum of policies (for example, different degrees of adjustment). However, presenting a choice between two discretely different alternatives may allow the government to credibly convey its superior information to a skeptical public and thus gain acceptance of a program (roughly) tailored to the economic situation. Such a situation is often encountered in Fund programs—say, in fiscal tightening in response to the state of the economy. Let the state of the economy be continuous, with high and low being very good or very bad states. The degree of possible fiscal tightening is also continuous, but with the government and public differing on what is optimal in each state of the economy, due to different discount rates for evaluating the trade-offs between current costs and future benefits. Both parties know that the public discounts the future more than does the government. The government has more information than the public about the economic situation, both ex ante or ex post, and the public’s support is required for fiscal tightening. The differences in discount rates give the government an incentive to overstate the seriousness of the economic situation by a discrete amount, and the public in turn is fully aware of government’s incentive. Suppose however that the public is presented with a discrete choice—either government enacts no policy change or a specific amount of fiscal tightening. (One may think of the choice of either accepting or rejecting a specific Fund program). Accordingly, if the difference between the discount rates of government and public is not too large, government’s characterization of the economic situation (cheap talk) in combination with a choice restricted to two options can be successful in inducing the public to accept the proposed amount of fiscal tightening. On the other hand, if the difference between government and public preferences is too large relative to the outcomes associated with the two choices, or—to return to the case where the public is uncertain about the government’s preferences—if the public believes, rightly or wrongly, that the difference between the government’s objectives and its own is sufficiently large (that is, that government has
Can Public Discussion Enhance Program ‘‘Ownership’’?
311
too strong an incentive to mislead the public), then cheap talk cannot convey information. When cheap talk provides no information, a government may need to use costly signals to convey information about its preferences. For example, a government whose objective is social-welfare maximization may reveal this by taking actions that would not be mimicked by a government with other objectives, because the latter would find them too costly. In formal terms, this is the issue of unobserved type, where one can think of two types of governments: a ‘‘good,’’ or social-welfare-maximizing government, and a ‘‘bad,’’ or non-social-welfaremaximizing government. How can a government whose objective is socialwelfare maximization separate itself from one that has other objectives? For the good government to care (and for the bad government to have an incentive to try to masquerade as a good government), there must be a favorable response to the government, or its program, being perceived as in the public’s interest. Such a response is present when public support is required for programs to succeed and when the public only complies with programs that it either perceives to be associated with a good government (that is, accepts because it is associated with a government it trusts) or is otherwise convinced to support. Even when costly signals are used to reveal type, public discussion may play a role. Jamaica provides an interesting example in which the government established good credentials by taking the costly action of relinquishing IMF financial support and by employing transparency (public discussion) to win support for a homegrown macroeconomic program. Following three decades of heavy reliance on Fund support from the 1960s through 1996, and while still facing very large adjustment problems, the government declared its independence from the IMF, essentially rejecting the Fund’s push for a devaluation to restore competitiveness along with work-out measures for the financial sector. It opted instead for its own strategy of tight fiscal and monetary policies, guarantees for bank deposits and other liabilities, and a gradual approach to dealing with problem financial institutions. The government also made a strong commitment to transparency by agreeing to the publication of the Fund’s fairly critical Public Information Notice following the 1997 Article IV consultation, along with subsequent Fund reports, and by issuing its own commentaries in which it emphasized both points of agreement and disagreement, thereby fostering public debate. Although it is difficult to judge whether the Jamaican government’s homegrown approach was more appropriate than the Fund’s approach would have been in the absence of political commitment, it is clear that the government succeeded in building a strong degree of ownership for its homegrown program by putting its credibility on the line, incurring the cost of foregoing Fund credit, and fostering public discussion through its transparency policy.
312
Allan Drazen and Peter Isard
When the government demonstrates its commitment to social welfare by choosing policies favored by the public, another problem may arise. What if the authorities believe that the public does not necessarily know what the best policy is? (The notion that a key function of public discussion is to educate the public about economic realities presumes that this is often the case.) A government that cares about social welfare faces a problem. Choosing the policy that the public says it prefers enables the government to demonstrate that it has ostensibly the same objectives as the public, but often at the cost of not actually maximizing the public’s welfare! (This distinction is sometimes referred to as the difference between a government being responsive to citizens’ expressed preferences and representative of public preferences. See Manin, Przeworski, and Stokes 1999.) This suggests that the only effective strategy that a ‘‘good’’ government can use to convince an uncertain public of its type is to educate the public that it is indeed advocating a good program—that is, a set of policies likely to generate attractive payoffs for the public relative to the payoffs associated with other courses of action (or inaction). Genuine public discussion may serve a purpose here that cannot be achieved through a simple announcement of policy or assertions about the economic realities on which policy is based.21 The reason is that the requirement that positions be argued and justified to the public (either directly or by convincing experts who have credibility with the public) will eliminate some arguments as convincing rationales for policy. Exposing the (supposed) rationale for policy to the cold light of open public debate and scrutiny can often reveal whether or not the rationale is really social-welfare maximization. (Fearon 1998 provides a nice discussion of this point.) Public debate can thus serve the purpose of demonstrating unbiasedness or revealing government type. The willingness to engage in public discussion may itself signal that the government’s objective is the maximization of social welfare, independent of any information that public discussion reveals. Good programs may face a lower cost of exposure to public discussion as they are more likely to survive public scrutiny. Public discussion may thus have different costs for different types of governments, or different types of programs.22 12.4.3
Case 3: The Public is Heterogeneous
In the third case, we focus on the role of public discussion when groups in society have conflicting interests. Conflict among members of a society is central to political economy in general (Drazen 2000), and is a key problem in formulating reform and stabilization programs. Though the problem of the public having imperfect information about the government’s preferences may be important here as well, we abstract from it to focus more exclusively on how public discussion might
Can Public Discussion Enhance Program ‘‘Ownership’’?
313
help hammer out an acceptable program when the public is heterogeneous. For the same reason, we also assume that the problem is not that the government must make credible that its objective is social-welfare maximization. (Of course, there are interesting questions here—for example, the government must make credible it is not too closely aligned with one interest group, such as the financial sector.) Suppose the public is divided into three groups: the general public, and two special interest groups (SIGs) denoted ‘‘Green’’ and ‘‘Blue,’’ who must consent for a program to be adopted. The general public is unorganized and lacks the power that the SIGs have to block a Fund program. The government may be seen as maximizing the welfare of the general public. Here we focus on public discussion as revealing information about SIGs or coordinating their actions. It might also weaken their ability to achieve selfish aims by making it necessary for them to justify their demands, analogous to the discussion at the end of section 4.2. Getting the approval of SIGs may be seen as a coordination problem when it is not known exactly what they might be willing to concede (though their general preferences are known) or when there is initial disagreement about how the world works. Let’s start with a simple reference case. Each group must simultaneously choose an action, and each perceives that the payoffs depend on the different combinations of actions. As in matrix 12.4, public discussion—that is, an announcement by both Blue and Green—precedes any actions. Matrix 12.5 Green Blue
Left
Right
Up
3,3
1,1
Down
1,1
3,3
Here, cheap talk works as a coordination device because there is no real conflict of interest. If Blue announces that his interest is down, then Green will clearly want to say right and we get to a better solution than we could expect without communication. In formal terms, Blue’s message is both self-signaling and selfcommitting. It is self-signaling in that Blue wants to announce down if and only if it is best. It is self-committing in that if Green believes Blue (which we just argued it should), the announcement creates the incentive for Blue to actually carry it out. Public discussion has the function of simply informing groups of
314
Allan Drazen and Peter Isard
where there may be common ground. If there is, then the program adopted will be superior to what would be adopted in its absence. Of course, this example is too simple, because there is no conflict of interest, only a problem of coordination. Farrell and Rabin (1996) point out that when the interests of SIGs are not so potentially well aligned, cheap talk messages are less likely to be self-signaling or self-committing. A simple example is the well-known ‘‘prisoner’s dilemma,’’ where the payoffs are as in matrix 12.6. Cheap talk precedes simultaneous action by the two groups. Matrix 12.6 Green Blue
Left
Right
Up
7,7
4,8
Down
8,4
5,5
Here, for Blue it is optimal to choose down for any choice Green makes, while for Green it is optimal to choose right for any choice Blue will make. Cheap talk will not allow them to coordinate. Messages are not self-committing. Though one may argue that repeated play of the game by the same players may lead to a cooperative solution, and that is what is seen in experimental studies,23 the basic message is clear: if SIGs see their interests as sufficiently in conflict in relation to a stabilization program, public discussion in the sense of simply exchanging information may have limited value. An argument put forward in discussing case 2 is relevant here as well. Public discussion that requires interest groups to justify their demands may eliminate some claims because they are unjustifiable in a setting of genuine, open discussion. Forums to elicit public input and reaction to proposed programs may be effective in revealing useful information only to the extent that they are not too warm and fuzzy. More exactly, eliciting credible information about preferences of groups is not easy once the groups know they have strongly conflicting objectives. The situation is somewhat different when government and IMF actions (including actions other than public discussion) can influence the payoffs that SIGs associate with different outcomes. When the interaction is among SIGs, but the authorities can change the payoff matrix, they could, for example, transform a prisoner’s dilemma as in matrix 12.6 to a coordination game as in matrix 12.5. More concretely, the ability to modify proposed programs and sweeten the deal
Can Public Discussion Enhance Program ‘‘Ownership’’?
315
for interest groups can be seen in this light. In the process of drafting legislation that can get through national parliaments, it is common to add amendments to win the support of specific legislators or groups. Fund programs are often structured with similar political constraints in mind. When the ownership problem is one of getting the political consent of powerful groups, building ownership may in fact be identified with cases in which the cooperative outcome is made credible, as distinct from cases in which it is not. We return to this issue in section 6. 12.4.4
Public Discussion and Democratic Values
A very different, but important point is that public discussion can be instrumental in strengthening democratic values, independent of the outcome of specific discussions. Encouraging public discussion on one type of issue may increase the likelihood that public discussion is used in other contexts as well. If the political culture is such that government decisions are not generally subjected to public discussion—if the paradigm of deliberative democracy is not well developed— then public discussion of economic policy can have a valuable learning effect. This may well be the case in new democracies.24 An outside body, such as the Fund, may be effective in fostering such a discussion in countries where this tradition is weak or nonexistent. 12.5
Drawbacks of Public Discussion
Public discussion may have significant drawbacks as well. Przeworski (1998) and Stokes (1998) have argued, for example, that when public discussion is followed by voting, lobbyists and interest groups may have incentives to mislead voters during public discussion, such that voting actually becomes less informed rather than more informed. The possibility that public discussion could cause people to be misled by SIGs is also present in the context of program design, but we don’t think that this is the central problem of public discussion of Fund programs. In fact, the key drawbacks of public discussion in the context of adopting a Fund program may be quite different than the drawbacks of discussion as a prelude to voting. In the case of public acceptance of a Fund program, one problem is that public discussion may essentially give interest groups more veto power over a proposed program, or a heightened awareness of adverse implications and a stronger incentive to use their veto power. To the extent that public discussion simply informs governments about the public’s true preferences, as in case 1 (where the public as a whole is seen as homogeneous), discussion improves program design. But when the public includes strong SIGs whose interests are not those of the general public
316
Allan Drazen and Peter Isard
(case 3), there is a danger that a policy-making process that gives them too large a voice will result in a program that does not serve the public good. Social welfare may thus be higher when the authorities make take-it-or-leave-it offers rather than allowing public discussion to give SIGs too much influence. This has been a criticism, for example, of giving NGOs a large say in program design, where the implicit assumption is that their interests do not necessarily reflect the broader public interest. Second, public discussion can be time-consuming and may slow down the process of shaping a program. This is potentially quite a serious problem when the economy is in crisis and a program needs to be put in place quickly. The problem of having exactly the right amount of discussion is obviously a very difficult one that probably is not conducive to any general rules. The right amount of public discussion will be very much situation-specific. Third, since programs may go through a lot of changes, making discussion public will involve groups (and the government) in espousing positions that may subsequently be rejected and even become seen as incorrect. The costs of being put in such a situation may induce groups to refrain from putting forward public positions at all, or to take positions in public that differ from the positions they would put forward in private. Hence, public discussion could actually greatly hinder information transmission relative to more confidential means of discussion. Fourth, since the discussion of Fund programs tends to focus attention on a country’s bad economic situation, making information public may in itself make the situation even worse. Some types of information may spook financial markets and lead to major capital outflows, making it even harder to design a successful program. This appears to be the main concern expressed in relevant IMF Board meetings by those executive directors who oppose full transparency—that is, full revelation of the Fund’s information and/or concerns about countries (IMF 2003c). None of these concerns is easy to address, since they are all both genuine and situation-specific (and hence hard to address in any generality). They should not be taken as arguments against public discussion per se, but only as cautions in considering how public discussion of a program should be structured. 12.6
The Design of IMF Programs
The importance of national ownership is clearly recognized in the principles governing the design of Fund programs. The Fund’s Guidelines on Conditionality (IMF 2002), as revised in September 2002, indicate that national authorities have the lead role in shaping program documents25 and that the Fund should encourage countries to build broad support for sound policies. Moreover, in association
Can Public Discussion Enhance Program ‘‘Ownership’’?
317
with the Guidelines, the set of general operational instructions provided to Fund staff engaged in program design is based on five key principles: ‘‘national ownership of reform programs; (ii) parsimony in program conditions; (iii) tailoring of programs to a member’s circumstances; (iv) effective coordination with other multilateral institutions; and (v) clarity in the specification of conditions’’ (IMF 2003a, paragraph 2). In moving from principles to procedures, the operational instructions emphasize that Fund staff ‘‘should seek proposals from national authorities at an early stage in the policy dialogue.’’ They also stress that Fund staff ‘‘should encourage the authorities to engage in a transparent participatory process in developing a policy framework, and should continue to be prepared to assist the authorities in this process by giving seminars, meeting with various interest or political groups (parliamentary committees, trade unions, business groups, etc.) and by being available to the media. . . . [while being mindful] of the authorities’ views on staff contact with domestic groups. . . . ’’ In addition, the operational guidance note stresses that documents prepared by Fund staff in the course of briefing the Fund’s management and reporting formally to the Executive Board should assess the challenges to broad ownership, including key capacity weaknesses and issues relating to political structures. It also clarifies the principle of parsimony in program conditions: performance criteria, prior actions, and other program conditions must be limited to those that, if excluded, ‘‘would seriously threaten the achievement of program goals or the Fund’s ability to monitor implementation’’ (IMF 2003a, paragraphs 6 and 7). The emphasis on ownership and transparent participatory processes in the recently revised Fund Guidelines and guidance notes is consistent with the position of the Fund’s Independent Evaluation Office (IEO). In its evaluation of cases in which Fund programs have not succeeded, the IEO concluded that the extent and structure of program conditions was much less important than securing an underlying commitment to core policy adjustments. It consistently suggested that the aim in program design should be to move as quickly as possible to a situation in which the core elements of a program are subject to a policy debate within the country’s own policy-making institutions, and that the Fund staff should actively seek to present policy options, analyze the tradeoffs between them, and encourage open debate on the alternatives (Independent Evaluation Office 2002). These recommendations are supported by case studies of Pakistan, the Philippines, and Senegal, where lack of political commitment was a major factor in program failures, and of Morocco and Jamaica, where a ‘‘real difference seems to have been made by strong domestic ownership’’ (Independent Evaluation Office 2002, 16–18).
318
Allan Drazen and Peter Isard
In the case of Senegal, despite significant progress during the 1980s in moving the economy away from excessive state intervention, roughly one-third of the measures envisioned under World Bank structural adjustment loans were not implemented as scheduled. These measures were concentrated in areas such as labor regulations, where there was strong opposition from vested interests and where the government apparently made little prior effort to generate public discussion and reach consensus on key issues. Senegal’s failure to pursue adequate public discussion also contributed to a lack of clarity and progress in its efforts to restructure the groundnut sector during the 1980s, which was a source of income for the majority of the rural population and a key sector in the effort to reduce poverty.26 By contrast, in Morocco, where evaluation reports have found no major differences in the approach to economic program design from that followed in other countries with IMF programs, ‘‘[i]ncreasing transparency in putting information and policies out for public discussion . . . appears to have helped develop a broader consensus [for reform]’’ and was a critical factor in weaning the country from prolonged reliance on IMF loans (Independent Evaluation Office 2002, 196). Despite the heightened awareness of the importance of country ownership and the relevance of public discussion (participation), most references to public discussion or participation in Fund documents are cast in broad terms, with little or no explicit recognition of the different functions that public discussion can serve, and little or no explicit focus on which specific functions it can usefully serve in specific circumstances. Of course, Fund staff rarely engage in public discussion without advance brainstorming or other forms of preparation, which often is not reflected in any written documents. And it would be misleading to suggest that Fund staff engage in public discussion without a fairly clear implicit sense of what they are trying to achieve. Nevertheless, the effectiveness of public discussion in building and assessing ownership of Fund programs might be significantly enhanced by greater awareness of the range of functions that public discussion can serve and by a more systematic focus on strategies for engaging in public discussion in specific circumstances. Although public discussion can enhance ownership in various ways, it would be misleading to suggest that discussion alone can induce complete ownership of Fund programs. Many of the countries that seek to negotiate Fund programs require fiscal adjustment to restore and maintain macroeconomic stability. Because such adjustment necessarily leaves some interest groups worse off in the short run, complete ownership of the fiscal components of Fund programs is generally difficult to achieve. Nevertheless, public discussion in this context—with sufficient efforts to keep it well focused—can be particularly important for building ownership by educating the public about the overall economy-wide benefits of fiscal ad-
Can Public Discussion Enhance Program ‘‘Ownership’’?
319
justment in the short run, about prospects that the benefits will be widely shared over the medium run, and about the gains from a cooperative solution that elicits financial support from the Fund and other sources. It can also contribute to building ownership by helping governments find the type of cooperative solution or policy mix that stands the best chance of sustaining political support— particularly so when the Fund gives policy makers broad freedom to design the specific details of the fiscal adjustment effort. In this context, it may be very important for the Fund to refrain from pressuring governments to adopt policies that are appealing on efficiency grounds but typically provoke strong public resentment, such as raising or eliminating ceilings on the price of necessities. Despite these various ways that public discussion can enhance ownership of fiscal adjustment policies, it cannot overcome the fact that various interest groups may continue to seek to benefit in the short run by undermining the fiscal adjustment effort. This is why the effectiveness of many Fund programs depends critically on appropriate prior conditions, fiscal performance criteria, and the phased provision of Fund credit.27 Although public discussion on its own cannot achieve complete ownership of programs for countries in which the restoration of macroeconomic stability requires fiscal adjustment, Fund programs need not be (and generally are not) limited to policy actions that impose short-run costs. Indeed, one approach to enhancing the overall ownership of Fund programs is to try to counteract conditions that impose short-run costs with conditions that the public correctly perceives to convey major benefits. In this context, public discussion, through many of the functions it can serve (eliciting information about what the public wants, educating the public, educating the government and the Fund, revealing that the government is indeed oriented toward maximizing social welfare, and leading heterogeneous interests to a cooperative solution), may well be able to strengthen the overall ownership of programs by building virtually complete support for various types of structural measures (such as clarification of property rights or strengthening of tax-collection mechanisms and accounting standards) that do not impose costs on any powerful interest groups, at least in the context of a good government where there is little or no inherent incentive to oppose important growth-enhancing structural reforms. 12.7
Conclusions
There has been growing recognition in recent years that the effectiveness of IMFsupported programs depends heavily on the degree of country ownership, and that ownership can be promoted by seeking to broaden and deepen the base of support for sound policies among a country’s domestic interest groups. These
320
Allan Drazen and Peter Isard
perceptions are reflected both in the September 2002 revision of the Fund’s Guidelines for Conditionality and in the international community’s decision to link debt relief for highly indebted poor countries (HIPCs) to the formulation of poverty reduction strategies through processes that involve the broad participation of stakeholders. This paper has focused on public discussion as one potentially important vehicle for enhancing program ownership (the functions of public discussion are summarized at the beginning of section 3). The motivation comes from our sense that economists do not yet have a very clear understanding of the various channels through which public discussion can work, or of the circumstances in which public discussion can be effective. This lack of understanding has contributed to the view—held in some quarters—that public discussion strengthens ownership primarily by leading counterproductively to the adoption of weaker programs. More seriously, failure to appreciate and distinguish between the range of functions that public discussion can serve and the circumstances that determine its effectiveness or ineffectiveness implies that country governments and the Fund may not be exploiting the potential of public discussion in the most effective ways. A better understanding of public discussion can contribute to the effort to design strong programs that can command broad country ownership. We have argued, primarily by way of example in illustrating the functions that public discussion may serve, that it can in fact be an important tool in raising the probability of program success. It would be wishful thinking, however, to suggest that public discussion alone can lead to complete ownership of a program that imposes significant costs in the short run. Those interest groups that stand to incur the short-run costs pose significant risks to the adjustment effort in the absence of appropriately structured conditionality, which provides financial incentives for countries to remain in compliance with their policy commitments. Moreover, as we argued in section 5, there are drawbacks to public discussion, especially in the presence of strong interest groups. By contrast, public discussion may be able to achieve virtually complete ownership of growth-enhancing structural reforms that impose no significant costs in either the short run or the longer run. Accordingly, public discussion has the potential to contribute importantly to growth by inducing the public to want to pursue more growth-enhancing structural reforms as part of their policy programs. In principle, program conditionality is not necessary for achieving compliance when there is complete ownership of reforms, although in practice it may be useful for spurring progress. Ironically, the efforts that the Fund has made to build program ownership by streamlining conditionality in recent years appear to have shifted the focus of
Can Public Discussion Enhance Program ‘‘Ownership’’?
321
Fund programs away from the types of growth-enhancing structural policies on which ownership would seem relatively easy to build. We attribute this in part to the limited attention that economists have paid to the concept of ownership and to the process of building ownership through public discussion. Acknowledgments This paper was prepared for the IMF Conference in honor of Guillermo Calvo, held in Washington, D.C., April 15–16, 2004. We wish to thank our discussant Raquel Fernandez, as well as Jim Boughton, Eddie Dekel, Thomas Dorsey, Marcela Eslava, Judith Gold, Caroline Kende-Robb, and Nuno Lima˜o for helpful comments and discussions. Parts of this paper were written while the first author was visiting the IMF, which he thanks for its hospitality. The views expressed in the paper are those of the authors and should not be attributed to the IMF. Notes 1. For readers unfamiliar with Fund jargon, the term completion date refers to the date at which the borrowing country becomes eligible for the last installment of the IMF loan, based on demonstrated compliance with, or agreed waivers or modifications of, the performance criteria and other conditions of the program. Out of the 615 Fund programs approved between 1973 and 1997, only 70 percent achieved their originally agreed completion dates (in many cases with modifications along the way). An additional 12 percent were extended beyond their original durations, another 11 percent were cancelled early but followed promptly by successor arrangements, and 7 percent were effectively suspended; see Mussa and Savastano (2000). 2. Under a definition that treats a country as a prolonged user of IMF resources if it has been under IMF programs for seven or more years within a ten-year period, forty-four countries were prolonged users during some part of the 1971–2000 period; see Independent Evaluation Office (2002). 3. IMF (2002). 4. See Boughton and Mourmouras (2002). 5. It is not only to the IMF that ownership must be demonstrated. Demonstration of ownership may also have a catalytic effect in inducing both official bilateral lending and private capital flows, which exceed or add substantially to official multilateral loans. 6. See, for example, IMF (2003a), paragraph 7. 7. This distinction between country and government (or authority) ownership is central to the arguments on conditionality presented in Drazen (2002) and is explored in detail there. 8. Fund involvement might be crucial in inducing complete country ownership (as in the use of conditionality in Drazen 2002), so that one could observe Fund programs where country ownership appears to be complete (ex post). 9. Farrell and Rabin (1996) present an excellent, easily accessible introduction to the subject. 10. From a formal point of view, political scientists have also used cheap talk models to study, for example, communication in legislatures.
322
Allan Drazen and Peter Isard
11. Some critics argue that this transformation may actually be a distortion of individual preferences. See section 12.5 below. 12. The reader will probably quickly point out that individuals may view income inequality not as itself the outcome, but rather in terms of its effect on other variables, such as growth. That is, individuals may both favor high growth, but differ on the connection between income inequality and growth. For illustration, see the discussion in chapter 11 of Drazen (2000). 13. Financial constraints on the IMF, for example, may lead its objectives to differ from those of a country’s authorities. See Drazen (2002) for a discussion of some of the issues involved in whether the government and the IMF have the same or different objectives. 14. ‘‘Common knowledge’’ means not only that something is known by all parties, but that everyone knows that everyone knows, everyone knows that everyone knows that everyone knows, and so on. 15. By contrast, the methods that Wimmer, de Soysa, and Wagner (2002) suggest for assessing the feasibility and sustainability of reforms would appear to require considerable resources and time to implement. 16. When the public consists of groups with conflicting interests, then public discussion meant to elicit information on preferred policy may not serve the same function, since groups have the incentive to reveal information strategically. We return to this problem in section 12.4.3. 17. Since programs are economically complex, the public will never fully understand all the details. Hence, there must be a degree of trust in the government. This can be attained by generating a feeling of inclusion in the process. 18. See Robb (2003), box 3. 19. See IMF (2003b). 20. See IMF (2001). 21. In the cheap talk examples, there is no real discussion or deliberation—the informed party sends a single message to the uninformed party. This is not unique to these examples. As Aumann and Hart (2003, 1619) point out, ‘‘Most formal models in this area [strategic information transmission] allow for at most one message from each player.’’ They consider multistage cheap talk (‘‘long cheap talk’’), but the application of their results to public discussion is an issue yet to be explored. 22. However, the signal is less than perfect, both because some regimes seem adverse to public discussion per se, and because public discussion may have adverse effects, as discussed in the next section. 23. It is argued that players in situations like this often play cooperatively, and that communication may have a coordinating effect. See footnote 14 in Farrell and Rabin (1996). 24. Brender and Drazen (2005) find that political deficit cycles are present only in new democracies and are not statistically significant in established democracies as a group. They suggest that this observed difference may reflect in part the lack of availability and analysis of information about the fiscal process. They also find that as a country gains experience with democracy, the political cycle disappears. See also Akhmedov and Zhuravskaya (2004) for evidence of this effect in Russia after the transition to democracy. 25. There is no requirement that country authorities draft program documents, but rather the directive that the Fund staff be responsive when the authorities desire a greater role in the drafting process. 26. Independent Evaluation Office (2002), chapter 11, which cites reports of the World Bank’s Operational Evaluation Department.
Can Public Discussion Enhance Program ‘‘Ownership’’?
323
27. The time dimension of Fund programs provides an interesting direction for further thought on the role of public discussion. In particular, given that the Fund is often criticized for being too willing to waive or modify performance criteria (not being a sufficiently ‘‘tough cop’’), it is interesting to consider whether public discussion that starts at the time of the initial request for financing can strengthen compliance with program conditions by making it more credible that the Fund will say ‘‘no’’ on test dates for subsequent installments of financing when governments breach the conditions established at the outset of their programs.
References Akhmedov, A., and Zhuravskaya, E. 2004. ‘‘Opportunistic Political Cycles: Test in a Young Democracy Setting.’’ Quarterly Journal of Economics 119: 1301–1338. Aumann, R., and S. Hart. 2003. ‘‘Long Cheap Talk.’’ Econometrica 71: 1619–1660. Boughton, J., and A. Mourmouras. 2002. ‘‘Is Policy Ownership An Operational Concept?’’ Working Paper WP/02/72, IMF, Washington, D.C. Brender, A., and A. Drazen. 2005. ‘‘Political Budget Cycles in New versus Established Democracies.’’ Journal of Monetary Economics 52: 1271–1295. Crawford, V., and J. Sobel. 1982. ‘‘Strategic Information Transmission.’’ Econometrica 50: 1431–1451. Drazen, A. 2000. Political Economy in Macroeconomics. Princeton, NJ: Princeton University Press. ———. 2002. ‘‘Conditionality and Ownership in IMF Lending: A Political Economy Approach.’’ IMF Staff Papers 49 (Special Issue). Elster, J., ed. 1998. Deliberative Democracy. Cambridge, UK: Cambridge University Press. ———. 1998. ‘‘Introduction.’’ In Deliberative Democracy, 1–18. Cambridge, UK: Cambridge University Press. Farrell, J., and M. Rabin. 1996. ‘‘Cheap Talk.’’ Journal of Economic Perspectives 10: 103–118. Fearon, J. 1998. ‘‘Deliberation as Discussion.’’ In Deliberative Democracy, J. Elster, ed., 44–68. Cambridge, UK: Cambridge University Press. Habermas, J. 1987. Theory of Communicative Action. Boston: Beacon Press. Independent Evaluation Office. 2002. Evaluation of the Prolonged Use of Fund Resources. Washington: International Monetary Fund. Available at www.imf.org. International Monetary Fund. 2001. ‘‘Strengthening Country Ownership of Fund-Supported Programs.’’ Policy Development and Review Department. Available at http://www.imf.org/External/ np/pdr/cond/2001/eng/strength/120501.pdf. International Monetary Fund. 2002. ‘‘Guidelines on Conditionality.’’ Legal and Policy Development and Review Departments. Available at http://www.imf.org/External/np/pdr/cond/2002/eng/ guid092302.pdf. ———. 2003a. ‘‘Operational Guidance on the New Conditionality Guidelines.’’ Available at http:// www.imf.org/External/np/pdr/cond/2003/eng/050803.htm. ———. 2003b. ‘‘Chad: Poverty Reduction Strategy Paper.’’ Country Report No. 03/209, IMF, Washington, D.C.
324
Allan Drazen and Peter Isard
———. 2003c. ‘‘IMF Reviews the Fund’s Transparency Policy—Issues and Next Steps.’’ Public Information Notice (PIN) No. 03/122. Available at http://www.imf.org/external/mp/sec/pn/2003/ pn03122.htm. Manin, B., A. Przeworski, and S. C. Stokes. 1999. ‘‘Elections and Representation.’’ In Democracy, Accountability, and Representation, eds. A. Przeworski, S. C. Stokes, and B. Manin, 29–54. Cambridge, UK: Cambridge University Press. Mussa, M., and M. Savastano. 2000. ‘‘The IMF Approach to Economic Stabilization.’’ NBER Macroeconomics Annual 1999. Cambridge: MIT Press: 79–122. Przeworski, A. 1998. ‘‘Deliberation and Ideological Domination.’’ In Deliberative Democracy, ed. J. Elster, 140–160. Cambridge, UK: Cambridge University Press. Robb, C. M. 2003. ‘‘Poverty and Social Impact Analysis—Linking Macroeconomic Policies to Poverty Outcomes: Summary of Early Experiences.’’ Working Paper WP/03/43, IMF, Washington, D.C. Stokes, S. C. 1998. ‘‘Pathologies of Deliberation.’’ In Deliberative Democracy, ed. J. Elster, 123–139. Cambridge, UK: Cambridge University Press. Wimmer, A., I. de Soysa, and C. Wagner. 2002. ‘‘Political Science Tools for Assessing Feasibility and Sustainability of Reforms.’’ Paper prepared for the Independent Evaluation Office, International Monetary Fund, Washington, D.C.
V
Transition and Growth
13
Sources and Obstacles for Growth in Transition Countries: The Role of Credit Fabrizio Coricelli, Bostjan Jazbecˇ, and Igor Masten
13.1
Introduction
One of the distinguishing features of planned economies was the total irrelevance of credit and finance for the determination of output. Market reforms gave credit and finance a central role in economic activity. Guillermo Calvo was among the first economists to emphasize the role of credit and financial markets in the transition process (Calvo and Coricelli 1992, 1993), arguing that the initial collapse in output could be interpreted as a trade implosion due to the sudden dry-up of financing for firms, induced by the initial liberalization and stabilization programs. Following such initial fall in output, transition economies moved along a path of relatively fast growth, although with significant differences across countries. This phase of growth has been accompanied by progress in financial development. However, the depth of credit markets remained low, even though during the period 2000–2005 credit has grown rapidly in several transition countries, leading in some cases to credit booms. In this paper we identify some causes for the continuing underdevelopment of financial markets and explore the implications of such underdevelopment for growth in transition countries. Through an econometric analysis of microdata for a relatively large set of countries, including both advanced European and transition countries, we find that financial sector development, through an increase in the depth of both bank credit and stock markets, would induce large growth effects. We also find that the impact of financial sector development on growth seems to be larger in transition rather than in mature industrial countries. This suggests that growth effects are larger at lower levels of development of a financial sector, a condition characterizing most transition countries. We interpret this evidence as suggesting that financial sector development affects growth mainly through a softening of liquidity constraints on enterprises. This belief is strengthened by the additional finding that trade credit plays a much more relevant role in substituting for official lines of financing in transition countries.
328
Fabrizio Coricelli, Bostjan Jazbecˇ, and Igor Masten
The paper is structured as follows: section 13.2 provides some evidence on the level of financial development of transition countries and emphasizes the role of institutional variables in explaining underdevelopment of financial markets in transition countries. Section 13.3 reviews the channels through which credit markets could affect output growth, discussing the role of bank credit, trade credit, and self-financing. Section 13.4 contains the empirical results of econometric analysis on industry-level data for a sample of countries including both advanced European market economies and transition countries, for a ten-year period starting in 1995. We show that progress in credit market development, through the convergence of the depth and efficiency of credit markets toward the level prevailing in advanced market economies, could have a major impact on growth in transition countries. The impact of financial market development on transition countries turns out to be much larger than for more advanced market economies. We show that this result may be due to the presence of threshold effects, related to the initial level of financial sector development. Indeed, financial sector development has a stronger effect in countries starting from underdeveloped financial markets. Furthermore, trade credit is found to be a relevant substitute for bank credit in countries at the lower end of the financial sector development Spectrum. Section 5 concludes. 13.2
Financial Sector Development in Transition Countries
In spite of rapid growth, especially in the Baltic countries and in Southern-Eastern European countries (SEE), financial markets remain shallow in transition countries (table 13.1). However, there are significant differences across countries, with CEB displaying much higher levels of financial market development with respect to CIS countries, with SEE in the middle.1 If we measure financial development against the level of economic development, summarized by the level of GDP per capita at purchasing power parity, we can identify the presence of a transition gap. Indeed, most transition countries are placed below the regression line that predicts the levels of financial development as a linear function of per capita incomes (figure 13.1). Similar results have been obtained in studies that take into account other factors, in addition to incomes per capita, that may affect financial development (Cottarelli, Dell’Ariccia, and Vladkova-Hollar 2003). The reasons for the underdevelopment of financial markets probably have to do with the initial design of liberalization and reform policies and with objective difficulties in developing financial markets in the midst of enormous structural change and transformation of the economy. Partly stimulated by the literature on transition, it is now acknowledged that institutional development plays a key role in macroeconomic performance, both
Sources and Obstacles for Growth in Transition Countries
329
Table 13.1 Domestic Credit to Private Sector and Stock Market Capitalization, 2005 (in percent of GDP) Domestic credit to private sector
Stock market capitalization
2000
2005
2000
2005
Albania
3.0
10.3
na
na
Armenia
7.1
8.0
1.3
0.9
Azerbaijan
5.9
9.5
0.1
na
Belarus
8.6
1.0
4.1
na
Bosnia and Herzegovina
5.6
22.6
na
na
Bulgaria
11.6
26.0
4.8
20.1
Croatia
36.0
55.6
14.5
35.2
Czech Republic Estonia
35.7 24.0
16.8 61.8
19.3 32.4
31.8 26.5
FYR Macedonia
11.4
10.5
18.6
0.2
Georgia
6.4
9.5
0.8
5.5
Hungary
30.1
51.1
25.8
31.9
Kazakhstan
11.2
26.7
7.5
21.6
Kyrgyz Republic
11.2
8.0
0.3
1.8
Latvia
18.1
67.4
7.4
17.4
Lithuania Moldova
10.0 12.6
34.0 21.2
13.9 30.3
31.8 na
Poland
20.8
23.4
17.9
31.6
7.2
10.2
3.4
22.3
Russian Federation
13.3
25.7
15.3
71.9
Slovak Republic
33.6
26.8
3.5
9.5
Slovenia
38.5
56.9
13.6
23.8
Tajikistan
19.2
17.1
na
na
Turkmenistan Ukraine
2.1 11.2
1.4 31.2
na 6.0
na 31.3
Uzbekistan
27.9
20.4
1.0
0.3
CEB SEE
26.4 12.3
42.3 23.9
16.7 5.8
26.0 23.0
Romania
CIS
8.4
9.4
6.7
17.0
World
45.7
55.8
53.2
57.7
European Union
74.4
85.8
78.7
67.0
Brazil
34.9
33.7
37.6
54.6
Philippines
40.6
32.7
68.0
33.5
Source: World Bank, EBRD banking survey, capital markets survey.
330
Fabrizio Coricelli, Bostjan Jazbecˇ, and Igor Masten
Figure 13.1 Development of Credit Markets and Incomes Per Capita
on growth rates and on their volatility (Roland 2000 and Acemoglu et al. 2002, among others). One of the main channels through which institutions affect growth is that of financial markets. The latter are indeed extremely sensitive to institutional design.2 How can we explain such a transition gap? A recent study by Djankov, McLiesh, and Shleifer (2007) provides a useful reference for computing the causes of such a gap, as it analyzes the determinants of credit-to-GDP ratios for the largest sample of countries considered so far in the literature (129 countries). The study considers macroeconomic factors (such as inflation and GDP growth), fixed costs in setting up financial markets (which give an advantage to large economies in terms of the absolute size of their GDP), and institutional variables, related to information available, through credit registries, the protection of creditor rights, and inefficiency in the legal procedures for recovering credit in cases of default. These institutional variables, especially information and creditor rights, explain much of the cross-country differences in credit ratios. Therefore, further progress in institution building in these areas appears to be the main avenue to reducing the gap in financial sector development in transition countries.
Sources and Obstacles for Growth in Transition Countries
331
Transition countries offer a unique vantage point for analyzing the role of institutions as they went through a process of radical institutional change, as most institutions necessary for the functioning of a market economy were absent in the previous regime. Liberalization policies left the determination of output to decentralized decisions in a market with free prices. At the same time, firms had to obtain in the market the required financing for their operations. Financial markets, however, were completely absent in the previous regime and could not be developed instantaneously. Transition countries went through a process that reversed the developments typical for market economies. In the history of capitalism, the development of finance preceded the development of industry. Most financial contracts were developed to support the activity of merchants. When the industrial revolution took place, there was financial capital available to sustain the growth of industry. By contrast, planned economies were characterized by very developed industrial systems. Indeed, most planned economies had an excessive weight of industry in the economy. In contrast, finance and money were a ‘‘veil’’ for the enterprise sector. They did not play any active role. The real equilibrium was determined by planning on real variables, and banks served as pure accounting institutions. There was no accumulation of information and skills to assess credit-worthiness. The same concept of credit-worthiness was irrelevant, as there was no possibility of bankruptcy. This was the financial sector that most transition countries had when they started their reforms. Whatever the final verdict on what caused the initial collapse in output, it is hard to dispute that there was a fundamental friction in the economy between the possibility of determining a decentralized equilibrium and the total absence of financial markets. Transition countries had to simultaneously develop financial markets and effect an enormous reallocation of resources in the economy. Following the initial shock, financial markets had to be created through a combination of policy intervention and spontaneous developments in the economy. Several features of the adjustment process after the initial shock were linked to the essence of finance. In some countries private credit markets developed through trade credit, in others barter trade became predominant, and in others generalized payment default through interenterprise arrears emerged. During this second phase, the experience of transition countries was highly heterogeneous. To simplify, CEB were able to drive through the initial crisis and establish a relatively well-functioning market economy. In the countries of the former Soviet Union, the output collapse was more persistent, resembling a bad equilibrium, a distant relative of a market economy. However, in recent years several CIS countries, including Russia, have begun to grow very rapidly.
Fabrizio Coricelli, Bostjan Jazbecˇ, and Igor Masten
332
Whether these two different paths were caused mainly by different macroeconomic policies or by different paths of institutional change and different initial conditions has been debated. In the review of the debate in Campos and Coricelli (2002), different initial conditions in terms of a minimal structure of market institutions and the subsequent impetus given by the attraction of Europe played a crucial role. Before turning to the empirical analysis of the effect of finance on growth in transition countries, we next briefly sketch a simple framework that may help to highlight the main channels through which finance may affect growth in transition. 13.3
Credit and Growth During Transition: A Simple Framework
We sketch a simple framework to assess the relationship between output and credit during transition. The model may help to organize our analysis of the different stages of transition. Assume there is a cash-in-advance constraint for the purchase of inputs by firms. The cash-in-advance constraint for the purchase of nonlabor inputs (N) can be expressed as follows, with C denoting cash flow and A an unanticipated shock: N a gðCðAÞÞ1
with g b 1:
ð13:1Þ
When g ¼ 1, the firm has no access to external financing, a condition of extreme credit constraint that was not uncommon in several countries in the initial phases of transition (Coricelli 1998). Even in 2005, in several CIS countries, most enterprises were completely excluded from official credit markets (table 13.2). In general, the share of both working capital and investment that was financed through internal funds in transition countries was much larger than in advanced market economies (table 13.3). Assume a simple production function with intermediate goods (N) as the only variable input, and labor (L) as the fixed factor, Q ¼ f ðL; sNÞ;
with f 0 ðNÞ > 0 and f 00 ðNÞ < 0:
ð13:2Þ
The parameter s denotes the importance of intermediate inputs in production. We also assume that payments to factor L are made at the end of the period and thus do not affect the cash-in-advance constraint in the current period, although they are important for the accumulation of liquidity by firms over time.3 We can thus continue our discussion without considering the labor costs in following equations. In the pre-reform regime, bank credit covered all purchases of inputs. After reforms, input prices change, interest rates are imposed on loans, and taxes are
Sources and Obstacles for Growth in Transition Countries
333
Table 13.2 Firms without Bank Loans and Financially Constrained (in percent of the sample) Small firms
Medium firms
Without bank loans
Without bank loans and financially constrained
CEB
60.8
SEE CIS
59.7 67.5
Germany
37.2
Large firms
Without bank loans
Without bank loans and financially constrained
Without bank loans
27.3
41.9
13.1
29.2
7.4
29.1 34.9
39.8 51.5
15.9 24.4
32.1 45.8
11.0 13.4
14.6
24.6
9.8
15.3
4.8
Without bank loans and financially constrained
Source: BEEPS 2005. Notes: A firm is classified as ‘‘small’’ if it has less than 49 employees, ‘‘medium’’ if it has between 50 and 249 employess, and ‘‘large’’ if it has 250 employees or more.
Table 13.3 Financing Structure of Firms, 2005 (in percent of total financing) Working capital financing Internal finance
Borrowing from banks
Equity
Trade credit
CEB
68.0
10.1
6.9
6.2
6.6
SEE
73.2
12.9
1.0
5.6
5.8
CIS
77.3
10.1
2.0
4.0
6.0
Other
Fixed investment financing CEB
62.4
14.3
6.5
1.9
12.0
SEE
70.8
17.7
0.9
2.4
6.8
CIS
77.2
11.6
1.9
1.8
6.9
Source: BEEPS.
raised on enterprise profits. Let us assume that banks keep their loans to enterprises constant in nominal terms. Define Q as output, P the price of output, B bank debt, t enterprise tax rate, and P after-tax profits. We define firm liquidity in real terms (nominal variables deflated by input prices); that is, the real liquidity necessary to buy inputs. Furthermore, let us assume that in the pre-reform regime the zero-profit condition holds: PQðNÞ ¼ Pi N ¼ B, or, at relative prices equal to 1, Q ¼ B. Thus, enterprise liquidity in the new regime is ðB þ PÞ=Pi ¼ ðB þ ðPQðNÞ BÞð1 tÞÞ=Pi :
ð13:3Þ
334
Fabrizio Coricelli, Bostjan Jazbecˇ, and Igor Masten
We also assume that the firm is viable in the new regime—specifically, its revenues at full capacity utilization (pre-reform level of output) are sufficient to cover its costs, inclusive of interest payments. It is clear from 13.3 that at unchanged nominal bank credit, the purchase of inputs necessary to produce full-capacity output is affordable only if t ¼ 0 and relative prices P=Pi remain equal to 1. In such a case 13.3 implies Q ¼ B. However, when t > 0 and there is an adverse shift in relative prices, with input prices increasing more than output prices (adverse supply shock), real liquidity declines and with it input purchases and thus output. Thus, such adverse changes can be thought of as bad realization of the shock A. In such a case constraint 13.2 is binding and firm’s output is liquidity constrained. An implication of the above model is that firms with higher s—that is, firms that are more liquidity-dependent for their production—will display a larger decline in output following a liquidity squeeze. This is indeed what Calvo and Coricelli (1993) found for the case of Poland at the outset of the reforms of the 1990s. In addition, given the assumption of decreasing marginal product from intermediate inputs, the impact of credit expansion on output would be larger for more constrained activities. From 13.3 we can also note that higher profit rates allow firms to decouple their production from external finance, as they can use their cash flow for purchases of inputs. The high profit rates in transition countries may thus explain the coexistence of high growth rates of output and low external finance. Let us now reinterpret the analysis in terms of the constraint 13.2. The parameter g (greater or equal to 1) identifies the development of financial markets: the higher is g, the less dependent firms are on their cash flow to finance purchase of inputs. It should be noted that in addition to bank credit, other forms of financing such as trade credit would reduce the dependence of output on cash-flow fluctuations. As discussed later, trade credit could be a substitute for bank credit. However, a large use of trade credit in a system with an underdeveloped banking system increases sharply the risk of contagion and the magnification of local shocks to the entire system, and thus is likely to increase output volatility. What is special in the experience of transition countries? There were two special features in transition countries: first, the degree of underdevelopment of financial markets was extreme, as credit did not play any role in the planned economy. Second, transition countries are perhaps the only example in which one could identify liquidity needs at the start of market reforms, without risking confusing supply and demand for credit.4 The paradox is that the equilibrium in the pre-reform period was observationally equivalent to having perfect financial markets. Indeed, output decisions and the distribution of inputs among firms were decided by the planner. Credit played just an accounting role, accommodating real decisions. There was no connection between financial markets and liquidity needs by firms. Decisions at t 1
Sources and Obstacles for Growth in Transition Countries
335
were not constrained by availability of liquidity. These constraints emerged at t when a decentralized market system was introduced. With the benefit of hindsight, the downplay of the role of financial markets was one of the major drawbacks of stabilization and reform programs in formerly planned economies. We stress this point not for pure intellectual reasons, but because we believe this initial sin conditioned the experience of transition economies and explains some of the persistent problems characterizing even the more advanced among the transition countries—namely, the new members of the European Union and those that are candidates to entry. Of course, we do not want to underestimate the importance of imposing hard budget constraints on enterprises. In the planned economy, incentives were heavily distorted by lack of discipline in the use of resources that can derive by having to repay bank loans. Eliminating automatic accommodation of real decisions was thus a key component of market reforms. However, the difficult task is to create the right incentives but at the same time ensure that liquidity needs for viable and growing firms are available. Combining imposition of right incentives and provision of liquidity is a feature of well-developed financial markets and of effective institutional mechanisms for contract enforcement (Coricelli and Djankov 2001). Indeed, such institutional preconditions were a fundamental aspect for the creation of private markets, especially trade credit, that could compensate the slow progress in developing an efficient banking sector. 13.3.1
Trade Credit and Inter-enterprise Arrears
A distinctive feature of successful reformers—namely the CEB group—in contrast to CIS countries is that in the former private credit markets developed rather quickly, while in the latter market liberalization was accompanied by a demonetization of the economy or the explosion of payment default. Calvo and Coricelli (1996) developed a simple model of credit chains within a circular system in which each firm is both buyer and seller of inputs, thus potentially both creditor and debtor in the inter-enterprise trade credit market.5 Here we illustrate the main mechanism through a simple example. There are n firms producing inputs, using labor and inputs from other firms. Assume that trade credit involves a proportion y of input transactions. Each firm exchanges one unit of good. The proportion ð1 yÞ is the share paid in cash. This cash can be obtained from banks or simply be liquidity accumulated by firms. If the same proportion of trade credit and cash payments is used by every firm, ð1 yÞ rather than ð1 yÞn is the amount of cash needed to support transactions for an amount n of goods. In the simplest case of identical value of purchases and sales of inputs, each firm has a balanced net position.6
336
Fabrizio Coricelli, Bostjan Jazbecˇ, and Igor Masten
Cash is needed only at one point of the chain. However, if we assume that the cash constraint is binding, a contraction of bank credit by ð1 yÞ induces a decline of sales by ð1 yÞn. The same effect would occur if there is a negative shock to the cash flow of an individual firm that cannot pay its suppliers any more, inducing a chain reaction of reductions in sales. Consider again the cash-in-advance constraint 13.1 N a gðCðAÞÞ1 :
ð13:4Þ
This now becomes ð1 yÞN a gðC 0 ðAÞÞ1 ;
ð13:5Þ
where C 0 ðAÞ now takes into account that the cash flow of a firm is affected by its credit provided to other firms for its sales. In our example of zero net trade credit, CðAÞ C 0 ðAÞ ¼ ð1 yÞN. Thus, the cash-in-advance constraint remains the same, but it affects a smaller amount of input transactions. The main difference between a situation with trade credit and one without it is that the amount of cash needed to effect transactions would be higher without trade credit. In our scheme, it would be 1 rather than y. Alternatively, trade credit could replace bank credit, playing ex post the same role of the netting-out operation carried out by banks in the pre-reform regime. Even though it could be argued that trade credit is a way to overcome informational problems associated with bank credit, it is nevertheless reasonable to assume that trade credit requires a minimum set of institutions ensuring enforcement of private contracts. Lacking these institutions, liquidity shocks induce an equilibrium of generalized default, in the form of inter-enterprise arrears (or involuntary credit). While in the very short run arrears can reduce output losses, as firms can acquire inputs, such an equilibrium is bound to create large costs. As a result of payment arrears, credit risk becomes so large as to impede any transactions based on credit. Indeed, it is not an accident that following a burst in inter-enterprise arrears one observes a boom in barter trade, as was the case in several CIS countries. In Calvo and Coricelli (1996) it is shown that institutional factors affecting the penalty imposed to bad behavior are key in determining whether the bad equilibrium with generalized default takes place. In the case of the good equilibrium that applies to CEB, one can note that trade credit was an important factor for financing interenterprise transactions.7 As noted earlier, there is a view that explains trade credit on the basis of informational advantages and of better enforcement of contracts than bank credit. However, this purely microeconomic view downplays a major drawback of trade credit, the bilateral nature of the contracts. Credit positions are
Sources and Obstacles for Growth in Transition Countries
337
not diversified and thus local shocks are transmitted to the whole system. In addition, while direct information on the customer is valuable, the presence of trade credit complicates the analysis of financial positions of firms, and thus of creditworthiness. A perfectly viable firm can run into liquidity problems if its customer does not pay. As a result, the firm cannot pay its supplier. In the bilateral contract the assessment of the customer’s capacity to repay depends on the whole chain. Thus, the information necessary to establish the capability of an individual firm to repay its debt is much higher. All these factors may imply that a predominance of trade credit with respect to bank credit increases output volatility by magnifying the aggregate impact of idiosyncratic shocks. However, the Kiyotaki and Moore (1997) story and its general equilibrium extension used by Cardoso-Lecourtois (2003) to explain higher volatility of output in emerging markets fail to recognize that the credit chain reaction is relevant only when there are large net positions in trade credit. In such an event, problems of cash-poor firms are transferred to cash-rich firms, inducing an inefficient magnifying effect to the whole system of liquidity shocks to cash-poor firms. The more circular the system is, the less relevant is this channel. Circularity means that net positions tend to be close to zero. This is indeed the case in most countries, including transition countries. For instance, in Romania, Calvo and Coricelli (1996) found that trade credit was almost perfectly circular, with net debt positions accounting for a marginal fraction of total input transactions. Furthermore, trade credit tends to be a major source of enterprise transaction financing in advanced market economies as well. We argue that the overall development of the financial sector is a fundamental factor in determining whether trade credit is a source of greater volatility. Consider again the cash-in-advance constraint 13.1 that would still apply to a model with circular trade credit, but let us think of an economy evolving over time. The constraint implies that the higher the level of financial sector development is, the higher is g and the lower are the effects on output of shocks to cash flow. In contrast, if g is close to 1, cash-flow shocks affect the capability of firms of effecting their cash payments and carrying out production at its full capacity. In such a situation, it is likely that firms will try to raise their liquidity by reducing the amount of trade credit provided to their customers. In other words, firms will try to get more cash from their sales. If all firms do the same as the overall amount of cash available in the enterprise sector remains the same, a chain reaction with lower amounts of inputs purchased will emerge, as in the credit chain foreseen by Kiyotaki and Moore (1997). The ensuing equilibrium is analogous to one of a chain reaction of default. This mechanism is particularly relevant if one takes into account that there are forms of payments, like those to workers or the government, that are rigid.
Fabrizio Coricelli, Bostjan Jazbecˇ, and Igor Masten
338
The discussion so far has emphasized the liquidity channel through which financial markets may affect output in transition countries. Development of financial markets affects output mainly by softening the liquidity constraints on firms. Under the maintained assumption of decreasing marginal product for inputs in production, it turns out that the effect of financial development should be larger the more severe are the liquidity constraints, and thus the underdevelopment of financial markets. In our view, this channel is the main one in transition countries, as opposed to the channel associated with financing innovation that has been identified in the literature (see Levine 2005). In the next section we analyze the impact of financial development on growth by looking at microdata for a relatively large sample of countries, including several transition economies. 13.4
Finance and Growth during Transition: An Empirical Analysis
Having established that transition countries are still lagging behind in their development of financial markets, the issue is to give a measure of how relevant such a gap is for the growth rate of transition economies. We analyze this issue on a large enterprise data set. Firm- or sector-level observations seem necessary to assess the impact of financial development on growth. Analyses at the country level of aggregation are affected by serious drawbacks in terms of identification of the effect of finance on growth. 13.4.1
The Data Set
Our sample of data covers thirty European countries and twenty-six three-digit International Standard Industrial Classification Revision 2 (ISIC R2) manufacturing industries for the period 1996–2004, which are obtained by aggregating firmlevel data from the Amadeus database. The countries in the sample are EU-25 countries to which we added data for Iceland, Norway, Croatia, Bulgaria, Romania, the Russian Federation and Ukraine. Since the time span of data is not uniform across countries, we are dealing with an unbalanced panel of data. Data on external finance dependence at industry level (three- and four-digit ISIC R2 level) are taken from Rajan and Zingales (1998) (hereafter RZ). They define external finance dependence as the share of capital expenditure that a given industry cannot finance through internal cash flow. Data on financial market development (market capitalization of listed firms, domestic credit, and bank credit to the private sector, all expressed as share in GDP) are taken from the World Development Indicators database (WDI) of the World Bank. These variables are then interacted with the RZ measure of external finance dependence to obtain the vari-
Sources and Obstacles for Growth in Transition Countries
339
able that measures the effect on growth of external financial funds provided through financial markets. Similarly, Fisman and Love (2003) (hereafter FL) construct a variable that measures the dependence on trade credit in the benchmark U.S. case as another source of external finance. We use their indicator in addition to the one by Rajan and Zingales. Not just trade credit is likely to play a major role in transition countries, but generally it seems more appropriate to consider external finance not only for capital expenditure, but also for working capital— that is, the main determinant of enterprise debt. Growth of industry output is calculated from firm-level data on sales drawn from the Amadeus database of the Bureau Van Dijck, which includes also small and medium-sized firms. Sales are deflated with the producer price index obtained from the IMF IFS database.8 The final dataset consists of 4,449 observations, comprising 638 country–industry units with seven years of time observations on average. 13.4.2
The Methodology
One of the major challenges in the literature on financial development and growth has been how to address the potential endogeneity problem between the growth rate of firm-level output and the degree of financial development. Using industrylevel data, Rajan and Zingales (1998) proposed a solution to the problem by using the dependence on external finance by different sectors in the United States as the benchmark. The idea is that the financial market in the United States can be assumed to be close to perfect and thus the financial structure of firms is determined by an optimal choice that is not constrained by supply factors. In addition, Rajan and Zingales argue that differences across firms of the same sectors are minor, and thus sectoral indicators are a good proxy for firm-level dependence on external finance. The U.S. indicators can be considered exogenous indicators of financing needs. Cross-country analysis of growth of real sales of firms, excluding the United States, can then be used to determine the role of financial development on growth. The sectoral U.S. financial dependence indicator is multiplied by the level of financial sector development in different countries to construct what is by now a familiar indicator in the literature, the Rajan-Zingales indicator. In our estimations we interact the RZ measure of external finance dependence with the share of total finance (market capitalization of listed firms and credit to the private sector) in GDP. We concentrate on this measure of financial development because it is the most general among the measures used in the literature. If the coefficient on the RZ indicator in a cross-country regression with the growth of real sales as a dependent variable turns out to be positive, this indicates
340
Fabrizio Coricelli, Bostjan Jazbecˇ, and Igor Masten
that financial sector development affects the growth rates of firms. A positive coefficient on the RZ indicator implies that firms that need more external finance grow faster in countries with a more-developed financial sector. As Fisman and Love (2003) pointed out, this raises the issue of sufficient financing for the firms with high returns in the countries with less-developed financial markets. From RZ it follows that the additional financing needed could be collected from internal financing. Petersen and Rajan (1997) argue that alternative funds could be raised by borrowing from suppliers. FL made a natural extension of Petersen and Rajan’s approach by constructing a measure of trade credit using a methodology similar to RZ. In order to obtain an industry-level measure of trade credit usage, they employ the ratio of accounts payable to total assets, calculated for the U.S. firms for different sectors. Similarly to RZ, this measure is also multiplied by the level of financial sector development in different countries. A negative sign of the coefficient is consistent with the hypothesis that firms that are more dependent on trade credit have a relative advantage in countries with lessdeveloped financial intermediaries, which implies a substitutability between trade credit and bank credit. In contrast, if the coefficient is positive, there is a complementarity between the two forms of financing. Even if RZ solve the problem of endogeneity of the financial indicator, there is still a problem of possible reverse causality from growth of output to the level of financial development. As emphasized by Guiso et al. (2004), a potential problem of RZ is that financial development may affect both the growth rate of firms and industries and the pattern of industry specialization. As a consequence, firms in financially less-developed markets may adopt technologies that make them less dependent on external finance. When estimating the effect of financial development on growth using industry-level data, RZ tackle this endogeneity problem by including in the estimated equations the beginning-of-period industry share in value added. This has been used by other authors as well, including Guiso et al. (2004). In our panel data set with time-varying data, initial period industry shares in total value added are simple fixed effects. This means that a simple within estimator corrects for the potential bias induced by the correlation between industry specialization pattern and financial development. However, it must be noted that by allowing for time variation in the panel, we may induce an additional potential bias simply because financial development may be demand- and not only supplydetermined. Indeed, we use the contemporaneous relation between financial development and growth, in contrast to RZ who use the initial-period level of financial development. In our approach this problem is unlikely to be too damaging, as our original units of observation on sales are firms who may have only a very limited effect on aggregate supply of finance. Furthermore, the aggregate effect of all firms together may be limited since we concentrate on manufacturing
Sources and Obstacles for Growth in Transition Countries
341
that normally accounts for less than half of aggregate value added. In addition, a significant share of credit may be supplied to households that even in most transition countries account for a large share of total credit. These are all indications that the endogeneity problem in our estimations may be very limited, if present at all. We formally tested this hypothesis with the Hausman test, which revealed that our within-panel estimates do not suffer from an endogeneity problem.9 This allowed us to treat the RZ and FL indicators as exogenous. Our baseline empirical model is Dyict ¼ aic þ bðRZi FDct Þ þ gðFLi FDct Þ þ dt þ uict
ð13:6Þ
where Dyict denotes growth of real sales in industry i, country c, and year t. RZi represents the Rajan and Zingales (1998) measure of external finance dependence, while FLi stands for the corresponding measure of the use of trade credit assembled by Fisman and Love (2003). FDct is a measure of financial development (the sum of stock market capitalization and private credit as a percent of GDP). aic is a full set of industry–country fixed effects, while dt denote common time effects.10 The baseline model was extended in two directions in order to capture potentially nonlinear effects of financial development on growth. First, explanatory variables were interacted with a dummy variable for transition countries, DTr , and its complement, DNTr . In other words, Dyict ¼ aic þ bTr ðRZi FDct Þ DTr þ gTr ðFLi FDct Þ DTr þ bNTr ðRZi FDct Þ DNTr þ gNTr ðFLi FDct Þ DNTr þ dt þ uict :
ð13:7Þ
The transition dummy de facto divides our sample between countries with low and countries with high financial development. Significant differences in coefficients between the two groups of countries may thus reflect nonlinearities in the effects of financial development on growth. The second extension adopts a more systematic approach to modeling nonlinearities in the effect of financial development conditional on the level of financial development itself, without resorting to the use of country dummies. This was achieved by allowing for explicit threshold effects. Following Hansen (1999), we allow for a multiple threshold model. In particular, the model can be compactly written as Dyict ¼ aic þ dt þ
4 X ðbj ðRZi FDct Þ j¼1
þ gj ðFLi FDct ÞÞIðtj1 < FD < tj Þ þ uict :
ð13:8Þ
Fabrizio Coricelli, Bostjan Jazbecˇ, and Igor Masten
342
This corresponds to a triple threshold model with t0 and t4 unspecified. For comparison, we also report the estimates of the double-threshold model. The threshold variable is the measure of financial development.11 13.4.3
Empirical Results
Empirical results are reported in tables 13.3–13.6. Table 13.3 replicates the original RZ and FL approaches on our data; that is, with no time variation in growth of industry output and financial development, but with the full set of country and sector dummies. Growth is averaged over the period under investigation, while the initial level of financial development is taken in 1995. The RZ variable is correctly signed, but insignificant. Including the FL variable also shows insignificant effects of trade credit, while leaving the estimate of the RZ coefficient virtually unchanged. Similar conclusions also emerge if we allow the coefficient to differ between transition and developed European countries. Results significantly change when we allow for time variation in the panel. Table 13.4 shows that, when inserted individually, the RZ variable is again insignificant. Inclusion of the FL variable and allowing for the effects to differ between transition and other European countries, however, produce meaningful results. The RZ turns out to be positive and significant in both groups of countries, and considerably higher in transition countries. What is more important, the size of coefficients is much higher in the panel analysis relative to the original RZ approach. Moreover, trade credit appears to be a substitute to official financing, with stronger effects in transition countries. Table 13.4 Effect of Financial Development on Growth: Original Rajan and Zingales Specification (1995 Financial Development) RZ FL
0.0012 (0.0021)
0.0011 (0.0022) 0.0092 (0.168)
RZ transition RZ nontransition
0.0017 (0.0066)
0.0024 (0.0066)
0.0007 (0.0023)
0.0006 (0.0024)
FL transition
0.642 (0.489)
FL nontransition
0.107 (0.183)
Note: Standard errors in parentheses. Constant not reported.
Sources and Obstacles for Growth in Transition Countries
343
These results confirm the original RZ result that financial development positively affects growth. However, it is worth noting that this result obtains only after we control for trade credit as an alternative source of external finance and, more importantly, after we allow for time variation in the data and nonlinearity of the effect. In this respect, we find that financially less-developed countries benefit considerably more from financial development, in line with our discussion in section 3. Evidence of nonlinearity in the effect of financial development on growth is more refined in the estimated threshold models. Formal testing supports the triple-threshold model as more appropriate than a double-threshold model. As can be seen by comparing the upper panels of tables 13.5 and 13.6, allowing for Table 13.5 Effect of Financial Development on Growth—Within Estimates RZ
0.0215 (0.0149)
0.0430 (0.0373) 0.461* (0.190)
FL RZ transition RZ nontransition
0.1442* (0.0054)
0.2447* (0.0757)
0.0352* (0.0155)
0.0140* (0.0017)
FL transition
0.688* (0.375)
FL nontransition
0.395* (0.103)
Note: Standard errors in parentheses. Constant not reported. Tests reveal significant presence of fixed effects in all specifications. Time dummies included in the model. * denotes statistical significance.
Table 13.6 Threshold Effects of Financial Development on Growth—Double Threshold Model Threshold 95% conf. int.
52.93 [40.94, 52.99]
F2 (p-value)
25.84 (0.03)
Parameters
FD < t1
RZ FL
0.308* (0.113) 1.53* (0.56)
70.29 [64.69, 256.10] t1 < FD < t2 0.052 (0.161) 2.78* (0.78)
FD > t2 0.020 (0.018) 0.43* (0.09)
F2 is the test statistic for presence of thresholds using 300 bootstrap replications. Standard errors of coefficients in parentheses. Time dummies included in the model. * denotes statistical significance.
Fabrizio Coricelli, Bostjan Jazbecˇ, and Igor Masten
344
Table 13.7 Threshold Effects of Financial Development on Growth—Triple-Threshold Model Threshold
52.93
62.92
70.29
95% conf. int.
[40.94, 52.99]
[62.11, 92.97]
[70.29, 70.34.10]
F2 (p-value)
55.12 (0.00)
Parameters RZ FL
FD < t1 0.308? (0.112) 1.56* (0.56)
t1 < FD < t2
t2 < FD < t3
0.133 (0.187)
0.154 (0.316)
0.83 (0.90)
8.02* (1.50)
FD > t3 0.020 (0.018) 0.44* (0.09)
F2 is the test statistic for presence of thresholds using 300 bootstrap replications. Standard errors of coefficients in parentheses. Time dummies included in the model. * denotes statistical significance.
the third threshold, found to lie between the first two, considerably reduces the estimated confidence interval of the second-threshold model (Table 13.7). We find thresholds at levels of financial development at 53, 63, and about 70 percent of GDP. The lower and upper thresholds are very precisely estimated, while the confidence intervals for the middle threshold prove to be much wider. In line with our priors, the largest effect of financial development on growth is found for countries that have a financial development ratio below the first threshold. Indeed, the coefficient is even larger than found for the transition countries in the previous panel analysis. Above that threshold the coefficient declines and becomes insignificant and close to zero for levels of financial development above the upper threshold.12 All developed Western European countries in the sample were above that threshold throughout the period under investigation. It is important to note that the most advanced transition countries passed the threshold in recent years as well, implying that considerably smaller growth dividends than those observed in the past can be expected from further financial development.13 Regarding the effects of trade credit, we found that at lower levels of financial development trade credit acts as a strong substitute for official finance. Going through insignificance in the second-threshold region, trade credit becomes a strong and significant complement in the third-threshold region. Above the third threshold, trade credit becomes a substitute for official finance, even though the effect is much smaller than the one prevailing below the first threshold. In summary, the effect of financial development appears to be much stronger for transition countries. Trade credit serves as a relevant substitute for official financing, softening the adverse impact on the growth of financial underdevelopment.
Sources and Obstacles for Growth in Transition Countries
13.5
345
Conclusions
Credit markets play a crucial role in the transformation from centrally planned to market economies. Our empirical analysis suggests that financial sector development has a significant independent effect on output growth. For transition countries, the growth dividend from financial sector development appears to be much larger than for more advanced market economies, indicating the potentially larger effects of financial development on growth in countries with low levels of development. Trade credit plays a relevant role in substituting for lack of official finance, especially in transition countries. The policy challenge is how to foster the efficient development of financial markets, which continue to be shallow in many transition countries. The process of financial sector development crucially hinges on institution building. CEB had a great opportunity for strengthening their institutions by joining the European Union. Further progress is likely to come from joining the Eurozone. Contrary to the traditional optimal currency area theory, countries with large structural asymmetries with respect to the Eurozone countries (because of much lower GDP per capita) and with still-underdeveloped financial markets can benefit significantly from joining the Eurozone. Joining a currency union may lead to large benefits in terms of diversification of risk and access to smoothing instruments. Often empirical analyses measure the readiness to join a monetary union in terms of structural asymmetries, measuring the overall volatility of output and the correlation between individual country shocks and those of the currency area to be joined. These analyses neglect the importance of transmission of shocks through the financial sector. If the financial sector of a union is more efficient than the national one, joining the union reduces the transmission of nominal shocks to output, or, for that matter, the transmission of temporary shocks to output, whatever their origin. Notes 1. CEB denotes Central-Eastern European countries and the Baltic States: Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia and Slovenia. SEE stands for Southern-Eastern European countries: Albania, Bosnia-Herzegovina, Bulgaria, Croatia, Macedonia, Romania, and Serbia. CIS stands for Commonwealth of Independent States: Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyz Republic, Moldova, Russia, Taijikistan, Turkmenistan, Ukraine, and Uzbekistan. 2. See, for instance, EBRD (2003). 3. See Calvo and Coricelli (1992) for a discussion of the possibility of ‘‘borrowing from workers.’’ 4. Rajan and Zingales (1998) discuss the issue and, to avoid these problems, assume that the liquidity needs of a sector are those observed in the United States, defined as a benchmark for perfect financial markets (see discussion in section 4).
346
Fabrizio Coricelli, Bostjan Jazbecˇ, and Igor Masten
5. See Kiyotaki and Moore (1997) for a model of credit chains. 6. If we had a model with nonzero net positions, the amount of liquidity needed would be equal to the sum of these net positions plus the amount ð1 yÞ. 7. In microdata for Hungary for the mid-1990s, for instance, it was found that many firms were completely cut off from access to bank loans and financed their purchases of inputs through trade credit (Coricelli 1998). 8. All observations with growth of real sales that exceeded 100 percent were treated as outliers and thus excluded from the database. Industry-level growth of output was calculated as a simple average. 9. In the construction of the Hausman test, we compared within estimates with the instrumental variables within estimates using lagged values of growth of real sales, the RZ indicator, and FL indicators. The Hausman test statistic is w 2 ð8Þ ¼ 10:61 with a corresponding p-value of 0.22. The test thus suggests the use of the normal within estimator as the most appropriate since it is more efficient under the null hypothesis that cannot be rejected. 10. Note that the Rajan and Zingales (1998) in their original specification estimated a different model. Growth of output was measured as the average over a period, while financial development was taken from the initial period. 11. In line with the results of the Hausman test in the basic panel analysis, we treat it as exogenous, which means that the model setting fits the assumption in Hansen (1999). Estimation of threshold levels and their confidence regions follows the multistep procedure described in Hansen (1999). Hansen’s method is designed for balanced panels, while we operate with an unbalanced panel. In such a case it must be noted that it is unknown whether all Hansen’s results regarding inference carry completely through. 12. Virtually the same findings as regards the RZ variable emerge from both the double- and triplethreshold model. 13. This statement is confined to measures of development that we use in the sample; that is, the depth of financial markets. The effects of other, mainly institutional, characteristics related to the efficiency of financial markets (that may or may not be correlated with the depth of financial markets) were not accounted for in present empirical analysis.
References Acemoglu, D., S. Johnson, J. Robinson, and Y. Thaicharoen. 2002. ‘‘Institutional Causes, Macroeconomic Symptoms, Crises and Growth.’’ Mimeo., Massachusetts Institute of Technology, Cambridge, MA. Calvo, G., and F. Coricelli. 1992. ‘‘Stagflationary Effects of Stabilization Programs in Reforming Socialist Countries: Enterprise-side and Household-side Factors.’’ World Bank Economic Review 6, no. 1: 71–90. Calvo, G., and F. Coricelli. 1993. ‘‘Output Collapse in Eastern Europe: The Role of Credit.’’ International Monetary Fund Staff Papers 40, no. 1: 32–52. Calvo, G., and F. Coricelli. 1996. ‘‘Credit Market Imperfections and Low-output Equilibria in Economies in Transition.’’ In Financial Factors in Economic Stabilization and Growth, eds. M. Blejer, Z. Eckstein, Z. Hercowitz, and L. Leiderman, 75–102. New York: Cambridge University Press. Campos, N. F., and F. Coricelli. 2002. ‘‘Growth in Transition: What We Know, What We Don’t, and What We Should.’’ Journal of Economic Literature 40, no. 3: 793–836.
Sources and Obstacles for Growth in Transition Countries
347
Cardoso-Lecourtois, M. 2003. ‘‘Chain Reactions, Trade Credit and the Business Cycle.’’ Mimeo., Universidad Carlos III de Madrid. Coricelli, F. 1998. Macroeconomic Policies and Development of Markets in Economies in Transition. Budapest: Central European University Press. Coricelli, F., and S. Djankov. 2001. ‘‘Hardened Budgets and Enterprise Restructuring: Theory and the Application to Romania.’’ Journal of Comparative Economics 29: 749–763. Cottarelli, C., G. Dell’Ariccia, and I. Vladkova-Hollar. 2003. ‘‘Early Birds, Late Risers, and Sleeping Beauties: Bank Credit Growth to the Private Sector in Central and Eastern Europe and the Balkans.’’ Working Paper No. WP/03/213, IMF, Washington, D.C. Djankov, S., C. McLiesh, and A. Shleifer. 2007. ‘‘Private Credit in 129 Countries.’’ Journal of Financial Economics, 84(2), 299–329. European Bank for Reconstruction and Development (EBRD). 2003. Transition Report 2003: Integration and Regional Cooperation. London: EBRD. Fisman, R., and Love, I. 2003. ‘‘Trade Credit, Financial Intermediary Development and Industry Growth.’’ Journal of Finance 58: 353–374. Guiso, L., T. Jappelli, M. Padula, and M. Pagano. 2004. ‘‘Financial Market Integration and Economic Growth in the EU.’’ Economic Policy 19, no. 40: 523–577. Hansen, B. E. 1999. ‘‘Threshold Effects in Non-dynamic Panels: Estimation, Testing and Inference.’’ Journal of Econometrics 93: 345–368. International Monetary Fund. International Financial Statistics. Various issues, Washington, D.C. Kiyotaki, N., and J. Moore. 1997. ‘‘Credit Chains.’’ Mimeo., London School of Economics. Levine, R. 2005. Finance and Growth: Theory and Evidence. In P. Aghion and S. Durlauf, eds. Handbook of Economic Growth. Amsterdam: Elsevier Science. Petersen, M., and R. Rajan. 1997. ‘‘Trade Credit: Theories and Evidence.’’ The Review of Financial Studies 10(3): 661–691. Rajan, R., and L. Zingales. 1998. ‘‘Financial Dependence and Growth.’’ American Economic Review 88: 559–586. Roland, G. 2000. Transition and Economics: Politics, Markets, and Firms. Cambridge, MA: MIT Press. World Bank. World Development Indicators. Various issues. Washington D.C.
14
Growth in Transition Economies: Domestic Policies, External Assistance, and Institution Building Stanley Fischer and Ratna Sahay
14.1
Introduction
The transition process in the economies of the former Soviet bloc is essentially at an end, as the twenty-five transition economies go their own ways and their policy problems become less distinguishable from those confronting other economies. After a little more than a decade, some countries are headed for Europe, with different expected dates of arrival; some in the CIS that, for a while, looked as if they would never progress, most notably Russia, are growing rapidly; and a few remain in severe difficulties. It is unlikely there will ever be another natural economic experiment on a scale as large as that of the transition process, with twenty-five economies changing policies radically at almost the same moment.1 What have we learned from the experience?2 That depends on what was expected. The consensus was that there would be an initial decline in output due to disorganization as the price and institutional structure of the economies changed and macroeconomic stability was established, and that after that period, countries would begin to grow more rapidly than the advanced countries, toward whose levels of income they would eventually converge. The initial decline was probably expected to last one to two years, but as we know now the decline—in magnitude and length of time—was grossly underestimated, especially in the CIS countries. Although the transition economies are now all growing, and although our reading of the history of the period, and the statistical representation of it, convinces us that the basic strategy recommended by international financial institutions (IFIs) was right, that is a controversial view. The accusation that the IFIs lost Russia and the charge that shock treatment and too-rapid privatization produced unnecessary output losses, disorganization, corruption, and misery have been familiar parts of the indictment of the approach recommended by Western officials and other advisers.3
350
Stanley Fischer and Ratna Sahay
The paper is organized as follows: in section 14.2 we take stock of the evolution of output in the twenty-five transition economies since the dismantling of the central planning system. In section 14.3, we extend the data set used in our previous work and confirm our earlier results that both macroeconomic policies and structural reforms help growth to revive. The role of initial conditions and policy performance are explored in greater detail in sections 14.4 and 14.5. We then focus on institutional development in the next three sections. In particular, we focus on an additional element in the indictment brought by critics of the advice offered to the transition economies by the IFIs and other mainstream economists; namely, that their advice ignored the need to build the institutional framework for a market economy. We argue that this charge that the IFIs did not take into account the importance of institutional development, especially the rule of law, is without merit. In section 14.9, we present empirical analyses that highlight the role of institutions, broadly defined. We present concluding comments in section 14.10. 14.2
Evolution of Output During Transition
This study is based on data through the end of 2003.4 Figure 14.1a shows the evolution of GDP since transition began in four groups of countries: those of Central and Eastern Europe (CEE); the Baltics; the CIS-5,5 which includes Russia; and the CIS-7,6 which consists of the poorest countries of the CIS. Transition year, T, is defined as the year in which central planning was abandoned in the former communist countries (table 14.1). Transition began in 1992 in the former Soviet Union (Baltics, the CIS-5, and the CIS-7), and somewhat earlier in the CEE countries. The chart shows unweighted averages of output for the countries in each of the groups. Output began to fall toward the end of the 1980s, even before transition began in most countries. As is well known, the output fall was much larger in the former Soviet Union countries, including the Baltics, than in the central and eastern European countries. It is also true that output took longer to recover in the CIS countries than in eastern and central Europe and the Baltics. As is evident from table 14.1 and figure 14.1b, once output began to grow, the average growth rate in the CIS-5 and CIS-7 was higher by nearly four and more than two percentage points, respectively, than in the CEE countries, which grew at about 3.5 percent. However, because of their late start and much larger output decline, measured output in most CIS countries still had not significantly surpassed its pretransition levels by 2003.7 In the event, the initial output declines looked more like collapses—which began before the end of the old regime—than the more measured declines that were expected. Growth began to revive at different times in each transition economy. The rate of growth in the initial recovery phase was probably below expectations.
Figure 14.1 Transition Economies (a) Real GDP Index, (b) Since Lowest Level. Source: International Monetary Fund. Authors’ calculations.
Table 14.1 Transition Economies: Output Performance, 1989–2003
Cumulative output growth, lowest to 2003 (5)
Average output growth since lowest level until 2003a (6)
Transition year (T) (1)
GNP (PPP) per capita in 1989 (2)
Year in which output was lowest (3)
Maximum output decline since T-1 (4)
Baltics
1992
7973
1993/94
38.1
52.3
4.87
Estonia
1992
8900
1994
29.4
54.1
4.80
Latvia Lithuania
1992 1992
8590 6430
1993 1994
44.2 40.6
53.0 49.8
5.32 4.49
CEE Albania
1990/91 1991
5760 1400
1991/97 1992
24.7 33.2
44.5 95.5
3.54 6.09
Bulgaria
1991
5000
1997
39.3
27.5
4.05
Croatia
1990
6171
1993
37.6
54.8
4.37
Czech Republic
1991
9000
1992
12.1
24.4
1.98
Hungary
1990
6810
1993
18.1
40.4
3.39
Macedonia, FYR
1990
3394
1995
27.2
15.1
1.76
Poland
1990
5150
1991
13.7
62.9
4.07
Romania Slovak Republic
1991 1991
3470 8000
1992 1993
20.6 24.4
22.3 51.0
1.83 4.12
Slovenia
1990
9200
1992
20.4
51.3
3.76
CIS-7 Armenia
1992 1992
4191 5530
1993/99 1993
46.0 14.1
58.2 100.7
5.72 6.97
Azerbaijan
1992
4620
1995
57.9
88.5
7.93
Georgia
1992
5590
1994
65.4
62.1
5.37
Kyrgyz Republic
1992
3180
1995
44.8
45.4
4.68
Moldova
1992
4670
1999
62.2
24.1
5.40
Tajikistan
1992
3010
1996
58.8
59.3
6.66
Uzbekistan
1992
2740
1995
17.5
27.2
3.01
CIS-5
1992
5954
1995/99
42.0
54.4
7.00
Belarus
1992
7010
1995
31.5
58.4
5.75
Kazakhstan Russia
1992 1992
5130 7720
1995 1998
31.1 45.6
53.3 35.5
5.34 6.08
Turkmenistan
1992
4230
1997
45.9
93.5
11.00
Ukraine
1992
5680
1999
55.2
31.3
6.81
Sources: GNP per capita in 1989 (DDGT 1998, World Bank), real GDP (World Economic Outlook 2004). Note: Transition year is defined as the year in which central planning was dismantled. T-1 refers to the year before transition began. Through columns 4 to 6, real GDP index (T-1 ¼ 100) is used. The calculations for Armenia based on time T, as there are no data for T-1 and 1989. The transition year and the year in which output is lowest for each group corresponds to the modal average. a Geometric averages.
Growth in Transition Economies
353
However, in more recent years, the poorer countries of the CIS-7 as well as the CIS-5 have been growing faster than CEE and the Baltics, and if high rates of growth continue, there is hope for convergence—though that could take a very long time. 14.3
Revisiting the Earlier Empirical Studies
In our earlier work (Fischer, Sahay, and Ve´gh 1996a, 1996b, and 1998), we concluded that the transition experience confirmed the view that both macroeconomic stabilization and structural reforms contribute to growth, and that the more structural reform that took place, the more rapidly the economy grew. Table 14.2 presents results obtained from running the panel data specifications used in our previous work on a data set that has been extended to 2001.8 Note that the number of observations nearly tripled with this extension of the data set. Table 14.2 Transition Economies: Growth Regressions—Fischer-Sahay-Vegh Specifications with Expanded Data Sets (t-statistics in parentheses) Fixed effects regressions: Dependent variable: GDP growth (1)
(3)
FSV 96
(2) Extended data
Fixed exchange rate
11.35 (2.00)*
5.50 (2.95)*
—
Fiscal balance
0.30 (1.42)
0.49 (7.03)*
CLI Inflation
7.42 (3.54)* —
2.78 (9.68)* —
Official assistance
—
—
LIP
—
—
LI
—
—
R-Squared Total Observations
0.72 75
0.48 216
FSV 97
0.24 (1.90)
(4) Extended data
0.42 (5.32)*
—
(5) FSV 98
(6) Extended data
7.08 (2.15)*
4.53 (2.27)*
0.29 (2.60)*
0.41 (5.28)*
—
—
2.50 (10.03)*
—
—
0.26 (2.32)*
0.02 (0.14)
— —
— —
12.97 (2.86)*
16.26 (4.17)*
—
—
2.73 (5.27)*
— — —
—
30.72 (2.70)* 0.57
208
0.66 100
38.44 (7.69)* 0.40 216
Source: Cumulative liberalization index (CLI), liberalization index of privatization (LIP), and liberalization index (LI) from DDGT (1998)/World Bank, inflation and fiscal balance from WEO 2002, aid share of gross national income from WDI 2002. Note: Fixed exchange rate regime is a dummy, it takes 0 for floating exchange rate regime and 1 for fixed exchange rate regime. Fiscal balance is general government balance as share of GDP.
354
Stanley Fischer and Ratna Sahay
The earlier results showed that lower inflation or fixed exchange-rate regimes (possibly because they brought inflation down much faster), more foreign assistance, and faster and deeper structural reforms helped raise growth rates. The results on fiscal balance (the tighter the fiscal policy, the faster the growth rates) were weak. The new results (with the extended data set) are consistent with our earlier results. In fact, confirming the IFI view, the fiscal balance variable that was not significant in the earlier data set is now significant across all three regressions.9 Moreover, the coefficient on the fiscal policy variable (the budget balance) increases with the longer-period data set. The coefficients on the reform indices— whether the cumulative liberalization index (CLI), or simply the liberalization index (LI), or the liberalization index of privitization (LIP), which captures privatization and enterprise reforms—are significant throughout the period, irrespective of the time period considered. An interesting result relates to the foreign assistance variable: with the extended data set, the coefficient on foreign assistance is no longer significant, indicating possibly that aid mattered more during the initial years of transition than later, or reverse causation, with the slow growers receiving more aid in later years. In summary, the extension of the data set strengthens our confidence in the empirical results obtained a few years ago by us and other researchers, as well as in the policy implications of those results.10 We elaborate more on these aspects in later sections. 14.4
Initial Conditions, History, and Geography
Our earlier work (Fischer, Sahay, and Ve´gh 1996a, 1996b, and 1998) found a significant effect of initial conditions on growth in the early years of the transition. For example, growth was generally worse, ceteris paribus, for those countries that were farther to the east (as measured by distance from Brussels), or that had spent a longer period under communism. Table 14.3 presents different measures of initial conditions, history, and geography of the transition economies. At the start of the transition, per capita income levels in the CIS countries were generally lower than those in the Baltics and the CEE countries. Compared to the Baltics and the CEE countries, the CIS countries had also endured a longer period of communism and are farther away from the more advanced countries of Europe (as measured by the distance from Du¨sseldorf). Of the four groups of countries, the CIS-5 are better endowed with natural resources than the other regions. While on average, the Baltics and the CIS countries are much more fragmented ethnically than the CEE countries, there is large diversity on this score among the countries within the groups.11
Growth in Transition Economies
355
Table 14.3 also includes an index of initial conditions computed by the European Bank for Reconstruction and Development (EBRD). A higher value of the initial conditions index indicates a more favorable starting position. This index is derived by the EBRD from factor analysis on a set of measures for the level of development, trade dependence on the Council for Mutual Economic Assistance (CMEA), macroeconomic disequilibria (such as repressed inflation and black market premium), distance from the EU, natural resource endowment, market memory, and state capacity. As expected, economic distortions as measured by the EBRD index were much higher in the CIS countries than in the CEE. We use this index in the regression exercises to examine whether and for how long initial conditions mattered. The role of initial conditions was examined by De Melo, Denizer, Gelb, and Tenev (1998); Berg et al. (1999); Havrylyshyn and van Rooden (2001); and others. The general conclusion was that the effect of initial conditions, while strong at the start of transition, wears off over time. The diminishing role of initial conditions is confirmed by the results presented in table 14.4. The coefficients are insignificant in two of three regressions. In the regression where it is significant (equation (7) in table 14.4), the initial condition index has the right sign, indicating that those with more favorable initial conditions grew faster. When the index is interacted with the time-in-transition dummy, the coefficient remains significant but turns negative, indicating that the longer the period, the less the significance of the initial conditions. These results are reassuring for countries aiming to overcome their adverse starting conditions by implementing good policies. 14.5
Policy Performance
As reported in section 14.3, in our earlier papers (Fischer, Sahay, and Ve´gh 1996a, 1996b, and 1998) the results on the macroeconomic policy variables were not robust across specifications. But before we report in greater detail the results with the expanded data set, it is worthwhile to look at the policy performance that accompanied the revival of growth in the later years of the transition. Table 14.5 reports on policy performance. We look at two key variables—the inflation rate and budget balance. The inflation performance across each of the subgroups—the Baltics, CEE, CIS-7, and CIS-5—is impressive. During the first three years of transition virtually all the countries averaged three-digit inflation rates. By 2003, inflation rates were reduced to single- or low double-digit levels. In fact, in the Baltics and the CEE, the inflation performance by 2003 was similar to that in the OECD countries. Pegging the exchange rate to a strong currency in the initial years helped most countries to bring inflation down very rapidly. This performance was sustained by the expectation of EU accession in later years in
5150 3470
1990 1990
Macedonia, FYR
Poland
1992 1992 1992
Moldova
Tajikistan
Uzbekistan
5590
1992 1992
5530 4620
1992 1992
Georgia
4191
1992
CIS-7
Armenia Azerbaijan
Kyrgyz Republic
9200
1990
Slovenia
2740
3010
4670
3180
8000
1991 1991
Romania
Slovak Republic
3394
6810
9000
1991
5000 6171
1990
1991 1990
Bulgaria Croatia
1400
5760
Czech Republic
1991
Albania
Hungary
1990/91
CEE
6430
8590
1992 1992
1992 1992
Baltics
Estonia
Latvia
7973 8900
Transition year (T)
Lithuania
Per capita income (PPP) in 1989
Table 14.3 Transition Economies: History and Geography
2195
1866
3311
2978
4650
2160 3046
2887
11345
7938
...
5684
5011
9447
10801
5632 7133
2186
7242
6703
5335
6320
6119
Per capita GDP (PPP) in T
1921
1921
1940
1921
1921
1920 1921
1924
1945
1948
1948
1948
1945
1948
1948
1947 1945
1945
1947
1940
1940
1940
1940
First year of communism
2.6
1134
3069 5047 1673 4938 4788
2.2 2.3 1.1 2.9 2.8
3704 3143 3270
1.1 3.2
815
824
1637
995
1522
1002
559
1574 913
1494
C0.6
3.2
2.9
1.7
1.9
2.5
3.3
3.5
2.1 2.5
2.1
1299
1293
0.2 0
1347 1449
0.4
Distance from Du¨sseldorf
C0.2
Initial condition index
0.48
0.58
0.55
0.66
0.55
0.12 0.31
0.46
0.17
0.25
0.20
0.04
0.57
0.17
0.11
0.26 0.37
0.46
0.26
0.35
0.61
0.53
0.50
Ethnic fractionalizationa
Moderate
Poor
Poor
Poor
Moderate
Poor Rich
Poor
Poor
Poor
Moderate
Moderate
Poor
Poor
Poor
Poor Poor
Poor
Poor
Poor
Poor
Poor
Poor
Natural resources
356 Stanley Fischer and Ratna Sahay
1992 1992 1992 1992
Kazakhstan Russia
Turkmenistan
Ukraine
5680
4230
5130 7720
7010
5954
6501
5998
5154
5615 9077
6660
1920
1918
1921
1921 1917
1918
2924 1435 5180 2088 4254 1664
C1.9 1.1 2.5 1.1 3.4 1.4
0.45
0.42
0.46
0.68 0.31
0.37
Rich
Moderate
Rich
Rich Rich
Poor
Source: Per capita income (World Bank), per capita GDP (World Economic Outlook, 2002), initial condition index (EBRD reports), ethnic fractionalization (Kok Kheng, 2001). Note: GDP data in transition year were not available for Uzbekistan, Slovenia, and Czech Republic. For these countries GDP data in the first year after transition are reported. a Range 0–1; 0 represents no ethnic fractionalization.
1992 1992
CIS-5
Belarus
Growth in Transition Economies 357
358
Stanley Fischer and Ratna Sahay
Table 14.4 Transition Economies: Output Performance, Initial Conditions, Policies, and Institutional Development, 1991–2001 (t-statistics in parentheses) 2SLS Panel regressions1 dependent variable: GDP growth (7)
(8)
(9)
1.439 (2.08)*
C0.013 (0.03)
C0.441 (0.94)
Initial conditions index *year
C0.213 (3.45)**
C0.039 (0.99)
0.027 (0.67)
Exchange rate regimeb
C4.293 (3.05)** C0.017 (7.30)**
C0.017 (7.33)**
0.387 (4.55)**
0.583 (7.59)**
0.685 (8.70)**
15.382 (10.58)**
4.77 (7.19)**
3.608 (5.91)**
a
Initial conditions index
First lag of inflation Change in fiscal balance Reform indexc State capture indexd Intercept
C0.126 (2.85)** C36.963 (11.13)**
C9.319 (6.11)**
C4.797 (2.79)**
R-Squared
0.41
0.4
0.47
Probability value
0.000
0.000
0.000
Total observations
265
296
252
Source: Initial conditions index, reform index, and state capture index (Transition Report); exchange rate regime (EAER); fiscal balance and inflation (WEO 2002). 1 One star (*) indicates significance at the five-percent level, while two stars (**) indicate significance at the one-percent level. a Initial conditions index is derived from factor analysis and represents a weighted average of measures for the level of development, trade dependence on CMEA, macroeconomic disequilibria, distance to the EU, natural resource endowments, market memory, and state capacity. The higher values of the index relate to more favorable starting positions. b Exchange rate regime dummy. Takes 0 for floating exchange rate system and 1 for fixed exchange rate system. c Reform index is average of liberalization index, market reform index, financial reform index, and privatization. The reform index is instrumented by the sum of reform indices in other 24 transition countries excluding the country for which it is instrumented. d State capture index measures the extent to which businesses have been affected by the sale of government decisions and policies to private interests. A higher value indicates more ‘‘capture.’’ It is based on the business environment and enterprise performance survey implemented in 1999 by the EBRD in collaboration with the World Bank.
Growth in Transition Economies
359
Table 14.5 Transition Economies: Policy Performance
Transition year (T)
Average inflation rate, first three years since price liberalization
Baltics
1992
298
1.0
C1.5
Estonia
1992
150
1.3
0.1
0.6
Latvia
1992
395
2.9
1.5
1.8
Lithuania
1992
350
1.2
3.2
1.4
CEE
1990/91
181
4.4
C5.2
C4.0
Albania
1991
116
2.3
20.1
5.1
Bulgaria
1991
163
2.3
10.3
0.0
Croatia
1990
455
1.5
3.0
4.5
Czech Republic Hungary
1991 1990
30 27
0.2 4.7
1.8 3.5
7.0 5.9
Inflation rate in 2003
Average budget balance (% GDP), first three years since T
Budget balance (% GDP) in 2003 C0.8
Macedonia, FYR
1990
113
2.5
3.9
2.0
Poland
1990
302
0.8
3.6
6.6
Romania
1991
209
15.3
0.6
2.5
Slovak Republic
1991
31
8.5
9.0
4.7
Slovenia
1990
363
5.6
0.5
1.4
CIS-7
1992
434
8.2
C19.7
C1.3
Armenia
1992
462
4.8
36.7
2.0
Azerbaijan
1992
509
2.2
8.2
1.5
Georgia Kyrgyz Republic
1992 1992
483 570
4.8 2.7
34.0 14.2
2.3 4.7
Moldova
1992
475
11.7
13.9
1.6
Tajikistan
1992
112
16.4
19.8
1.8
Uzbekistan
1992
430
14.8
11.3
1.6
CIS-5
1992
298
12.9
C5.6
0.5
Belarus
1992
526
28.4
2.4
1.2
Kazakhstan
1992
91
6.4
5.3
3.2
Russia
1992
485
13.7
12.1
1.0
Turkmenistan
1992
298
11.0
6.7
0.0
Ukraine
1992
91
5.2
14.9
0.7
Source: Inflation and general government budget balance (World Economic Outlook 2004). Authors’ calculations.
360
Stanley Fischer and Ratna Sahay
many of those countries. The development of domestic financial markets and access to international capital markets over time also helped in financing budgets without resorting to central bank money creation. There has been a substantial fiscal consolidation across all groups since the start of the transition. The CIS-7 countries stand out for their impressive feat in reducing the budget deficit from nearly 20 percent of GDP in the first three years to just over 1 percent in 2003. Due to financing constraints and lack of access to capital markets, these countries faced severe expenditure compression throughout the period. Helbling, Mody, and Sahay (2004) show that foreign aid, especially grants, came on too small a scale to provide significant support to their budgets; this appeared to have both hindered their recovery process and led to a rapid buildup of external debt-to-GDP ratios. The CIS-5 and the Baltics also managed to nearly balance their budgets by 2003, although the starting points in the Baltics were already quite favorable. Interestingly, the consolidation in the CEE countries was much smaller, but less was needed, given their more favorable starting points. Moreover, most of these countries invested in public sector reforms that initially contributed little to deficit reduction. So the quality of spending improved over time even though the trend fall in the level of the deficit was small. The CEE countries also had greater market access that enabled them to finance their budgets by borrowing from abroad. The empirical results with the extended data in table 14.2 reinforce the message that good output performance was accompanied by good macroeconomic policy performance. While policy measures of inflation stabilization (exchange-rate dummy or inflation itself) remained significant irrespective of the size of the data set, the fiscal variable gains significance consistently in the extended data set. Hence, the focus on macroeconomic stabilization by domestic policy makers, reinforced by advice from multilateral agencies, seemed to have helped the transition countries to grow—although the possibility of reverse causation has to be taken into account. 14.6
The Role of Institutions
We now examine the role of institution-building in the transformation of the centrally planned economies to market economies. In its 1998 volume Institutions Matter,12 the World Bank (p. 11) defines institutions as ‘‘formal and informal rules and their enforcement mechanisms that shape the behavior of individuals and organizations in society.’’ They are distinguished from organizations, which are ‘‘entities composed of people who act collectively in pursuit of shared objectives.’’13 Examples of formal institutions presented by the authors are laws, regulations, and contracts; among informal institutions are trust, ethics, and political
Growth in Transition Economies
361
norms. Organizations include political (legislatures, political parties, government agencies, the judiciary), economic (private firms, trade unions, business associations), and social (NGOs, schools, PTAs). A similar definition is used in the new institutional growth literature, which claims that institutions dominate policies. For instance, Acemoglu, Johnson, Robinson, and Thaicharoen (2003, 52) define institutions as ‘‘a cluster of social arrangements that include constitutional and social limits on politicians’ and elites’ power, the rule of law, provisions for mediating social cleavages, strong property rights enforcement, a minimum amount of equal opportunity, and relatively broadbased access to education, etc.’’ (italics in original). This seems to be a description of a well-run country, or a measure of governance. Similarly, but less normatively, Rodrik, Subramanian, and Trebbi (2002, 132) specify as institutions ‘‘the rules of the game in a society and their conduciveness to desirable economic behavior,’’ in particular ‘‘the role of property rights and the rule of law.’’ The institutional variable in Dollar and Kraay (2003) is a measure of the rule of law. Hall and Jones (1999, 114) attribute national differences in economic performance to differences in social infrastructure, defined as ‘‘the institutions and government policies that make up the economic environment within which individuals and firms make investments, create and transfer ideas, and produce goods and services.’’ Most of these definitions are consistent with those set out by the World Bank (2002): institutions are rules and modes of behavior, including the rule of law; organizations are ‘‘entities composed of people who act collectively in pursuit of shared objectives.’’ The new institutional literature is sometimes taken as saying that not much matters for the development of a country beyond its institutions, which are themselves determined by a more or less immutable history, such as those of European settlement. The italicized component mistakes econometric convenience in choosing instruments for a substantive argument: for instance, no one observing modern Zimbabwe would take the institutional structure—in the sense of rule of law—as immutable. We also doubt the first statement, having seen too many policy changes that produced changes in real behavior—for example, successful economic stabilizations. But that is not the subject of this paper. Those who argued that the IFIs and Western advisers were missing the point in their efforts to help the transition economies referred mainly to the need to promote the rule of law, property rights, and the investment climate (which are closely linked). They were thus talking about institutions as we have just defined them. However, we shall be examining both institutions and organizations, for we find the distinction a difficult one to draw, and in particular believe that institutions in the World Bank sense require organizations to be effective, and that in seeking to develop institutions, it is also necessary to develop organizations. It may well be, though, that it is easier to set up organizations than to develop
362
Stanley Fischer and Ratna Sahay
institutions such as the rule of law, which may require changes in modes of social interaction that take a long time to be implanted in social behavior. Apart from the complication that arises in defining institutions clearly, it is also difficult to measure them meaningfully and without subjectivity. Moreover, many measures (such as those based on surveys conducted in a particular year) are not available on a time series basis, which complicates our ability to run regressions that explain short-run growth (the latter because it would be more meaningful to see the impact of institutional development, rather than the institutions themselves, on how output evolves in the short run). Hence, in our approach, we use qualitative information and simple statistics to make our assessment on institutionbuilding, and interpret institution-building in as broad a sense as possible. 14.7
Institution Building During the Transition
Table 14.6 and figures 14.2 and 14.3, based primarily on indices constructed by the EBRD, present several measures of the extent and success of the institution building that took place in the last decade.14 The reform index is a composite measure of EBRD subindices: the financial reform index, market reform index, the liberalization index, and a privatization index. The financial reform index, in turn, has two subindices: banking and nonbanking reforms. The market reform index comprises the enterprise reform index and competition policy. The liberalization index is made up of price, trade, and exchange-regime liberalization indices. Finally, the privatization index measures small- and large-scale privatization. Liberalization indices increased very rapidly, as they could be adopted and implemented quickly. But progress is evident even in the other categories; for instance, commercial law and the legal environment, where changes take longer to implement. Table 14.6 and figure 14.2 show that while all three regions (the CEE and Baltics, CIS-5, and CIS-7) have progressed, the CEE and the Baltics are more advanced, and have advanced faster, than the others. Regarding laws, indices on the legal framework have only recently (since 1997) begun to be constructed. Available data indicates that as expected, the CEE and the Baltics are more advanced than the others. However, as seen in table 14.6 and figure 14.3c, the legal environment measures in the CIS-5 advanced rapidly, reaching levels closer to those of the CEE by 2000, even though they were behind both the CIS-7 and the CEE in 1997. Financial regulations (figure 14.3b), by contrast, appear to have progressed faster in the CIS-7 than in the CIS-5, even though they were lagging in 1997. In the three categories of commercial law, financial regulations, and legal environment (company law), all three groups of countries are more advanced in the legal environment than in commercial law or financial regulations. However, while the charts show improvements, the absolute levels for the variables shown in figure 14.3 are low in the CIS countries. The regulatory
Growth in Transition Economies
363
quality (figure 14.3d) is good in the CEE and the Baltics but poor in the other countries. Table 14.7 shows high—though not perfect—correlations among all measures of reform. In particular, if we take the legal variables (the last three rows in the matrix) as representing institutional reform, their average correlation with the overall reform index is about 0.7. A surprising feature of the table is that the correlation between the commercial law index and measures of price liberalization, competition policy, and privatization is relatively low—but recall that the legal indices are available for only the later part of the period. Table 14.8 shows that the legal variables are negatively correlated with the time a country spent under communism—another way of saying that these reforms are more advanced in the CEE and the Baltics. We should note also the strong correlations among the many reform measures in tables 14.7 and 14.8: once a country is reforming, it typically advances on many fronts. In sum, there are no major surprises here, except perhaps that there has been progress in all areas of institutional reform, narrowly and broadly interpreted. Nor is it a surprise that countries that have been more successful in implementing policy reforms have also been more successful in implementing structural or institutional reforms. 14.8
Role of IFIs in Institution Building
The charge that the IFIs were unaware of the importance of institutional development, especially of the rule of law, cannot be sustained. An early awareness of the importance of institutions is evident in Fischer and Gelb (1991). The conditionalities in IFI programs generally included elements of institutional and organizational development. Moreover, much of the technical assistance was focused on the building-up of institutions; for instance, in the case of the IMF, the development of central banks, treasuries, tax systems, financial systems, and statistical systems. Further, there were many bilateral non-IFI efforts at providing technical assistance, including, for example, technical assistance by the American Bar Association to develop legal systems.15 Table 14.9 indicates the extent to which restrictions on foreign direct investment and portfolio flows were lifted in the transition countries, which often reflected key recommendations by the IMF-supported programs in these countries. In general, the countries liberalized their capital accounts substantially. As expected, there were fewer restrictions in the Baltics and the CEE countries than in the CIS7 or the CIS-5. The table also shows the extent to which these countries implemented IMF-supported programs. Here the record is the best in the Baltics, where almost all conditions were met (95 percent), followed by CEE and the CIS-7 with similar records at around 85 percent. The CIS-5 has a lower implementation rate
2.6 2.3
1.1
1.80
1.1
1.6
1.8
2.1
2.3
1.8
2.1 1.2
2.1
1.9
1.00
1
1
1
1
1
1
1
Lithuania
CEE
Albania
Bulgaria
Croatia
Czech Republic
Hungary
Macedonia, FYR
Poland Romania
Slovak Republic
Slovenia
CIS-7
Armenia
Azerbaijan
Georgia
Kyrgyz Republic
Moldova
Tajikistan
Uzbekistan
2.3
2.8
2.8
2.4
2.8
2.50
3.3
3.3
3 2.9
2.9
3.6
3.5
3
3
2.6
3.11
3.3
3.5 3
1.3 1.1
Estonia Latvia
3.27
1.17
Baltics
1
1
1
1
1
1
1
1.00
1
2
2 1
1
2
2
1
1
1
1.40
1
1 1
1.00
1991
1991
2001
Banking reform
Reform index
Table 14.6 Transition Economies: Institutional Development
2
1
2
2
2
2
2
1.86
3
3
3 3
3
4
4
3
3
2
3.10
3
4 3
3.33
2001
1
1
1
1
1
1
1
1.00
2
1
2 1
1
2
2
1
1
1
1.30
1
1 1
1.00
1991
2
1
2
2
2
2
2
1.86
3
2
4 2
2
4
3
2
2
2
2.60
3
3 2
2.67
2001
Nonbanking reform
1
1
1
1
1
1
1
1.00
1
2
2 1
1
2
2
1
1
1
1.40
1
1 1
1.00
1991
2
2
2
2
2
2
2
2.00
3
3
3 2
2
3
3
3
2
2
2.60
3
3 3
3.00
2001
Enterprise reform
2
..
2
2
2
1
3
2.00
3
3
4 3
2
4
4
4
3
2
3.20
3
4 3
3.33
1997
3
2
4
3
2
2
2.67
4
3
3 4
4
4
3
4
4
2
3.50
4
4 4
4.00
2001
Commercial law
..
1
2
2
1
2
2
1.67
3
3
4 3
2
4
3
3
3
2
3.00
3
3 3
3.00
1998
3
4
2
3
3
2
2
2.71
3
3
.. 4
2
4
3
3
3
2
3.00
4
3 3
3.33
2000
Financial regulations
2
..
3
3
3
2
3
2.67
4
3
4 3
2
4
4
4
3
2
3.30
4
4 3
3.67
1997
3
2
3
3
3
3
3
2.86
4
3
4 4
3
4
3
4
4
3
3.60
4
4 4
4.00
2000
Legal environment
364 Stanley Fischer and Ratna Sahay
1
1 1.1
1
1
Belarus
Kazakhstan Russia
Turkmenistan
Ukraine
2.5
1.3
2.8 2.6
1.3
2.10
1.00
1
1
1 1
1
1.80
2
1
3 2
1
1.00
1
1
1 1
1
1.80
2
1
2 2
2
1.00
1
1
1 1
1
Source: Transition Report (various issues), EBRD, London. Note: The indices are ranked from 1 to 4, where 4 indicates the highest level of reform.
1.00
CIS-5
1.75
2
1
2 2
..
2.25
2
..
2 3
2
3.00
3
2
4 3
3
1.80
2
1
2 3
1
2.50
..
2
3 3
2
2.25
2
..
2 3
2
3.25
3
..
4 4
2
Growth in Transition Economies 365
366
Stanley Fischer and Ratna Sahay
Figure 14.2 Transition Economies: Institutional Development—Financial and Private Sectors. Source: EBRD, various issues.
at nearly 75 percent—it is possible that the institutional capacity of these countries to implement reforms may have been overestimated. We examine IMF conditionality more closely in table 14.10. This table shows the number of conditions in IMF-supported programs with transition countries during 1993–1997.16 Performance criteria, usually imposed on macroeconomic variables, are more stringent conditions than structural benchmarks. In addition to these two types of conditions, prior actions (not reported here) are also often needed before a country can enter an IMF-supported program.17 Many of the structural benchmarks cover a broad definition of institutionbuilding, and include conditions that relate to the building of organizations; for instance, in financial sector reform. Table 14.10 shows that countries that were less advanced institutionally had, on average, almost twice as many structural benchmarks as compared to the more advanced ones. These benchmarks were fairly comprehensive, covering several areas in trade and exchange systems, tax policy and systems, expenditure management systems, public enterprise reforms,
Growth in Transition Economies
367
Figure 14.3 Transition Economies: Legal Institutional Development. Source: EBRD, various issues.
privatization, and financial sector reforms. There were fewer conditions relating to macroeconomic variables in the CIS countries than those relating to structural changes or institution-building; in the CEE countries, the number of macroeconomic conditions was roughly equal to the nonmacroeconomic conditions. During the early years, there were many benchmarks to prevent the accumulation of arrears by the state budget, enterprises, and the central bank—these benchmarks were imposed to curb the culture of soft budget constraints that was common in the communist regime. Budget constraints generally hardened over time. In Table 14.11 we present data on IMF technical assistance to the twenty-five transition economies over the period 1989–2003. The volume of technical assistance to each of the three groupings of countries in the table peaked in the period 1992–1995, but was maintained at a high level through the end of 2003, especially to the twelve CIS countries. This assistance was aimed at building the institutional infrastructure for the macroeconomic management of the economy. We conclude that if there was a problem in the slow development of institutions, it was not for lack of effort on the part of outside advisers. Rather, it was
0.94*
0.97*
0.92*
0.80*
0.92*
0.77*
0.74*
0.90*
0.91*
0.90*
0.63*
0.78*
0.71*
Liberalization index
Privatization
Banking reform
Nonbanking reform
Enterprise reform
Competition policy
Price liberalization
Trade liberalization
Small privatization
Large privatization
Commercial law
Financial law
Legal environment
0.73*
0.81*
0.67*
0.81*
0.78*
0.76*
0.56*
0.77*
0.90*
0.92*
0.95*
0.90*
0.82*
0.91*
1.00
0.62*
0.77*
0.54*
0.81*
0.75*
0.72*
0.56*
0.93*
0.94*
0.82*
0.85*
0.86*
0.80*
1.00
0.53*
0.54*
0.42*
0.73*
0.84*
0.97*
0.88*
0.55*
0.76*
0.57*
0.79*
0.89*
1.00
0.64*
0.68*
0.54*
0.94*
0.96*
0.83*
0.68*
0.68*
0.84*
0.70*
0.84*
1.00
Market Lib. reform index Priv. (3) (4) (5)
0.71*
0.73*
0.66*
0.81*
0.80*
0.81*
0.61*
0.67*
0.90*
0.74*
1.00
0.63*
0.79*
0.56*
0.70
0.64*
0.59*
0.42*
0.77*
0.77*
1.00
NonBanking banking reform reform (6) (7)
0.68*
0.74*
0.60*
0.82*
0.79*
0.77*
0.58*
0.75*
1.00
Enterpr. reform (8)
0.43*
0.67*
0.39*
0.68*
0.61*
0.55*
0.45*
1.00
Competition policy (9)
0.42*
0.54*
0.32*
0.59*
0.70*
0.74*
1.00
Price liberalization (10)
0.53*
0.51*
0.43*
0.74*
0.83
1.00
0.54*
0.62*
0.43*
0.81*
1.00
Trade liberal- Small ization priv. (11) (12)
0.64*
0.65*
0.56*
1.00
Large priv. (13)
0.92*
0.65*
1.00 0.69*
1.00 1.00
Comm. Finan. Legal law law envi. (14) (15) (16)
Note: Column (1) ¼ Column (2) þ Column (3) þ Column (4) þ Column (5); Column (2) ¼ Column (6) þ Column (7); Column (3) ¼ Column (8) þ Column (9); Column (4) ¼ Column (10) þ Column (11); Column (5) ¼ Column (12) þ Column (13).
0.95*
0.92*
Market reform
1.00
Financial reform
Reform index
FinanReform cial reform index (1) (2)
Table 14.7 Institutions and Reforms
368 Stanley Fischer and Ratna Sahay
0.39* 0.35* 0.53* 0.29*
0.78* 0.71* 0.40* 0.46* 0.46* 0.20*
Initial conditions index Distance from Du¨sseldorf
Time under communism
Ethnic fractionalization
0.65* 0.92*
0.63*
Financial law Legal environment
1.00
1.00
Commercial law
Commercial law
Reform index
Reform index
Table 14.8 Initial Conditions, Institutions, and Endowments
0.45*
0.56*
0.45*
0.46*
1.00 0.69*
Financial law
0.40*
0.60*
0.42*
0.45*
1.00
Legal environment
0.62*
0.84*
0.75*
1.00
Initial condition index
0.49*
0.72*
1.00
Distance from Du¨sseldorf
0.44*
1.00
Time under communism
1.00
Ethnic fractionalization
Growth in Transition Economies 369
370
Stanley Fischer and Ratna Sahay
Table 14.9 Transition Economies: Foreign Investment Indices Foreign direct investment restrictions indexa 1993–1999
Portfolio investment restrictions indexb 1996–1999
IFI indexc 1993–1997
Baltics
1.4
0.0
Estonia
0
0.0
95.2
Latvia Lithuania
1.4 2.8
0.0 0.0
93.1 92.6
CEE Albania
1.3 1.8
0.6 1
84.0 90.9
Bulgaria
1.3
0.4
50
Croatia
0.9
0.6
80
Czech Republic
0.3
0.1
96
Hungary
1.1
0.4
98.1
Macedonia, FYR
0.8
0.9
83.5
Poland
1.6
0.5
96.4
Romania Slovak Republic
2.8 0.8
1 0.6
76.7 ...
Slovenia
1.8
0.7
...
100
CIS-7
1.6
0.7
86.8
Armenia
0.4
0
94.4
Azerbaijan
0.8
0.6
93.4
Georgia
0.8
0.5
94.6
Kyrgyz Republic
1.4
1
78
Moldova
3.1
0.6
Tajikistan
1.8
1
Uzbekistan
2.8
1
70.6
89.5 ...
CIS-5
2.6
0.9
73.8
Belarus
3.4
1
62.2
Kazakhstan Russia
2.6 2.6
1 0.6
75.4 88.5
Turkmenistan
2.8
1
Ukraine
1.8
1
... 69.1
Source: Foreign direct investment restriction index and portfolio investment restriction index (Garibaldi, Mora, Sahay, and Zettelmeyer (GMSZ), 2002), IFI index (MONA database, IMF). a Calculated by GMSZ 2002 based on Annual Report on Exchange Arrangements and Restrictions (IMF). The index ranges from 0.2 to 6, where 6 reflects most restrictions. b Calculated by GMSZ 2002 based on Annual Report on Exchange Arrangements and Restrictions. Ranges from 0 to 2, where 2 indicates outright prohibition of portfolio flows. c Measures percentage of IMF performance criteria met (100 equals all performance criteria met).
16
SBA
EFF, SBA
Latvia
Lithuania
18
EFF, SBA
SBA
SBA
SBA, ESAF
SBA
SBA
Bulgaria
Croatia
Czech Republic
Hungary
Macedonia, FYR
Poland
Romania
48
SBA, EFF, ESA
SBA, ESAF
SBA, ESAF
EFF, SBA
SBA
Armenia
Azerbaijan
Georgia
Kyrgyz Rep.
Moldova
Uzbekistan
11
EFF, SBA
EFF, SBA
SBA
Belarus
Kazakhstan
Russia
Ukraine
—
13
—
—
—
—
4
—
4
—
—
—
—
—
—
—
—
—
—
—
8
42
—
Source: MONA Database, International Monetary Fund.
27
80
18
34
SBA
CIS-5
19
20
9
11
21
SBA, ESAF
CIS-7
52
11
5
33
4
35
23
SBA
CEE
21
42
26
SBA
Estonia
MacroLaws, economic others
Baltics
Arrangement type
Performance Criteria (PC)
SBA
EFF, SBA
SBA, EFF
SBA
SBA
SBA, EFF
ESAF, SBA
SBA, ESAF
EFF, ESAF, SBA
SBA, ESAF
SBA
SBA
SBA, ESAF
SBA
—
SBA, EFF
SBA
EFF
SBA
SBA
Arrangement type
3
5
2
1
3
4
12
1
1
8
3
5
1
3
0
1
—
1
3
2
2
1
2
2
2
1
4
1
2
3
2
4
2
9
4
4
1
0
1
0
—
0
3
1
1
0
3
1
2
1
3
5
3
2
2
7
4
4
4
4
8
1
0
0
—
0
2
2
0
0
1
0
2
19
9
5
9
2
5
10
13
24
19
12
7
5
13
12
—
4
3
7
0
1
6
2
3
7
11
2
6
1
3
8
7
13
13
8
11
4
10
0
—
8
7
7
6
0
2
3
6
6
15
3
8
0
9
5
7
21
2
7
6
3
4
2
—
6
6
5
1
0
3
1
PrivaTax/ Trade/ expendi- Financial tization exchange Pricing Public reform sector systems marketing enterprise ture
Structural Benchmark (SB)
Table 14.10 Transition Economies: IMF Conditionality on Macroeconomic Policies, Reforms, and Institution Building, 1993–1997
5
4
3
4
4
0
1
6
5
19
8
7
0
1
8
0
—
1
4
2
1
1
1
1
Other
23
43
47
21
34
12
34
41
39
98
53
46
34
17
36
15
—
20
28
25
11
3
18
11
Total SB
Growth in Transition Economies 371
372
Stanley Fischer and Ratna Sahay
Table 14.11 Transition Economies: IMF Assistance in Institution Building, 1989–2003 (Person-years) Fiscal area Central Europe (10) 1989–1991
Financial sector
Statistics
7.1
11.6
1.2
1992–1995
20.7
25.2
4.9
1996–1999
13.1
20.1
4.5
2000–2003
19.2
19.7
8.4
Baltics (3) 1989–1991
—
0.3
1992–1995
7.6
9.6
1996–1999
2.8
7.6
2000–2003
1.4
1.7
— 5.7 — 0.0
CIS Countries (12) 1989–1991
2.9
0.6
0.8
1992–1995
57.0
66.3
22.1
1996–1999 2000–2003
53.9 42.3
51.5 31.7
21.6 10.5
Source: International Monetary Fund. Note: Number of countries in parentheses.
the sheer difficulty of developing some institutions (in the new institutional literature sense) such as the rule of law, which require changes in societal norms and ways of doing business. Those changes inherently take a long time. Other institutions, or organizations, such as a central bank, can be created more easily and rapidly, and a great deal of technical assistance was directed to building central banks and fiscal systems. It could also be argued that there was no point in trying to move ahead with economic policy changes and organization-building until the right institutions had been created. This is not a point we agree with, for we believe that changes in institutions, in the sense of modes of behavior, are strengthened and sustained by an appropriate organizational framework. 14.9
Regression Analysis: Including Institutions
In table 14.4, we add measures of organizational and institutional development to the basic regressions reported in table 14.2. Also included in the regressions is an index of ‘‘state capture’’—or corruption—based on work by the EBRD and World Bank, which attempts to measure the extent to which businesses have been affected by the sale of government decisions and policies to private interests. In a
Growth in Transition Economies
373
broader sense, this index is likely to capture the extent to which laws are in practice respected. There are some changes in the regressions between tables 14.2 and 14.4. In particular, instead of using the CLI, which runs only through 1999, we replace it with the reform index (described earlier), which runs through 2001, and which is a more comprehensive measure of institution-building than the CLI. In addition, the regressions are two-stage least-squares panel regressions, with the reform index in each case being instrumented by the sum of the reform indices for the other countries in the regression. Regressions (7) and (8) in table 14.4 in essence repeat the results of table 14.2, and show strong statistical significance of both macroeconomic variables and the reform index. Inflation and the exchange-rate-regime variable are strongly correlated, and they are not separately significant if both are entered into the regression. Initial conditions lose their statistical significance once the exchangerate-regime variable is replaced by a measure of inflation. Regression (9) in table 14.4 adds the state capture index to the regression. The coefficient on that variable is statistically significant, indicating that even after accounting for the extent of structural reforms and institution-building reflected in the reform index, state capture has an independent negative impact on growth. The reform index varies within the panel of data from close to 0 to 4. This means that the quantitative impact of the reform index on growth implied by the regression equations is very high, possibly too high to be plausible. The statecapture index has a range within the sample of about 25, suggesting an impact on growth between the most and the least state-captured countries of about 3 percentage points—a number that we find plausible. 14.10
Conclusions
In a few years, history will have recorded the rapid transformation of centrally planned economies into market economies, indistinguishable from other market economies, irrespective of their level of development. It will also have recorded the speed and magnitude of change, which were remarkable by any standards. By and large, countries that are closer to industrialized Europe are at a more advanced stage, but the pace of reforms and the catching up in the lesserdeveloped countries of the former Soviet Union has been swift since their economic and political liberalization. We have demonstrated in this paper that after the large fall in output at the start of the transition process, the resumption of growth occurred at different times and paces, accompanied by a tightening of macroeconomic policies, rapid structural reforms, and institutional development. We also documented that the
374
Stanley Fischer and Ratna Sahay
IFIs were well aware of the need for institutional and organizational development in the transition process, and that a great deal of effort was devoted to helping build institutions in the transition economies. One final word: we noted earlier that it is sometimes assumed that a country’s institutions are determined by a more or less immutable history, and thus must be slow to change. Institutions can change, particularly at a time of crisis. And our ability to predict which institutions are immutable is low. So, at the same time as we emphasize the role of institutions in growth and development, we should also recognize that institutions can change. And they have changed during the transition process in the former Soviet bloc. Notes Stanley Fischer is Governor of the Bank of Israel, and Ratna Sahay is Senior Advisor in the Finance Department at the International Monetary Fund. This paper was presented at a conference in honor of Guillermo Calvo at the IMF on April 14–15, 2004. An earlier version was presented at the AEA Meetings in Washington, D.C., in 2003. We would like to thank Andrew Feltenstein, Rodney Ramcharan, David O. Robinson, Andrei Shleifer, and Antonio Spilimbergo for their insights and helpful comments. The authors are grateful to Hulya Ulku, Manzoor Gill, and Ping Wang for their excellent research assistance. The views expressed in this paper are those of the authors and do not necessarily reflect the views or policies of their institutions. 1. In terms of what can be drawn from the data, radical changes in direction in an economy like China’s or India’s, where more than twenty-five economic units (states or provinces) collect data and differ in aspects of economic policy, geography, and other conditions, also provide a great deal of information. But in those cases, there is only one national policy at a given time. 2. See, for instance, Aslund (2001), EBRD Transition Reports, Stiglitz (2002), and World Bank (1999/ 2000). 3. Many different economists and other advisers were active in the transition economies, and their advice was by no means uniform. However the mainstream advice was generally for rapid change where possible (for example, price and trade reform) and as rapid change as possible in other areas, such as privatization. See Aslund (2001). 4. In analyzing some questions, we used the latest available data that end earlier than 2003. 5. Belarus, Kazakhstan, Russia, Turkmenistan, and Ukraine. 6. Armenia, Azerbaijan, Georgia, Kyrgyzstan, Moldova, Tajikistan, and Uzbekistan. 7. Of course, the pretransition data have to be interpreted with some caution in that they were measured at distorted prices and the quality of products was most likely inferior. GDP is likely to be underestimated during the initial transition years, more so in the countries of the former Soviet Union. 8. Given data constraints, regressions with the liberalization indices (Cumulative Liberalization Index [CLI], Liberalization Index of Privatization [LIP], Liberalization Index [LI]) have been run with data updated only until 1999, while those with aid as a share of GDP were updated until 2001. 9. Aside from the currency boards in the Baltics and Bulgaria, most of the transition economies are now operating with flexible exchange-rate regimes (albeit ‘‘managed’’ in most cases). Early in the tran-
Growth in Transition Economies
375
sition period, countries that pegged the exchange rate generally were able to stabilize more rapidly, but as the Russian case vividly illustrates, adoption of the peg sometimes led to setbacks later. In other cases (central and eastern Europe) there was a gradual transition to exchange-rate flexibility as capital accounts were opened. 10. See, for instance De Melo, Denizer, and Gelb (1996), Berg, Borensztein, Sahay, and Zettelmeyer (1999), and Garibaldi et al. (2002). 11. While the results are not reported in this paper, we examined empirically whether these additional factors mattered for growth. We found that, controlling for other factors, resource-rich countries performed worse than others and, as expected, the more ethnically diverse countries grew more slowly. 12. Burki and Perry (1998). 13. The distinction is attributed to the new institutional economics. 14. The top grade for any index is 4. 15. We believe it would be useful to construct a full listing of all the technical assistance provided to any given transition country. One such listing, for the CIS-7, can be found in table 6 in Hare (2003). Hare estimates that on average about 20 percent of the financial aid provided to the CIS-7 was for purposes of institution-building. We are not aware of such information having been collected for other transition economies. 16. The numbers are based on data collected by IMF economists Valerie Mercer-Blackman and Anna Unigovskaya. 17. The number of conditions are counted by the test dates that were set for each of the measures. Hence, if there were four test dates for ceilings on reserve money, the total would include these four numbers even though they were on the same variable.
References Aslund, Anders. 2001. Building Capitalism: The Transformation of the Former Soviet Bloc. New York: Cambridge University Press. Acemoglu, Daron, Simon Johnson, James Robinson, and Yunyong Thaicharoen. 2003. ‘‘Institutional Causes, Macroeconomic Symptoms: Volatility, Crises and Growth.’’ Journal of Monetary Economics 50, no. 1: 49–123. Berg, Andrew, Eduardo Borensztein, Ratna Sahay, and Jeromino Zettelmeyer. 1999. ‘‘The Evolution of Output in Transition Economies: Explaining the Differences.’’ Working Paper No. WP/99/73, International Monetary Fund, Washington, D.C. Burki, Shahid Javed, and Guillermo Perry, eds. 1998. Beyond the Washington Consensus: Institutions Matter. Washington, D.C.: World Bank. De Melo, Martha, Cevdet Denizer, and Alan Gelb. 1996. ‘‘From Plan to the Market: Patterns of Transition.’’ World Bank Economic Review 10: 397–424. De Melo, Martha, Cevdet Denizer, Alan Gelb, and S. Tenev. 1998. ‘‘Circumstances and Choice: the Role of Initial Conditions and Policies in Transition Economies.’’ Working Paper No. 1866, World Bank, Washington, D.C. Dollar, David, and Aart Kraay. 2003. ‘‘Institutions, Trade, and Growth.’’ Journal of Monetary Economics 50, no. 1: 133–165.
376
Stanley Fischer and Ratna Sahay
EBRD Transition Report, several issues. London. Fischer, Stanley, and Alan Gelb. 1991. ‘‘The Process of Socialist Economic Transformation.’’ Journal of Economic Perspectives 5, no. 4: 91–105. Fischer, Stanley, Ratna Sahay, and Carlos Ve´gh. 1996a. ‘‘Stabilization and Growth in Transition Economies: The Early Experience.’’ Journal of Economic Perspectives 10, no. 2: 45–66. Fischer, Stanley, Ratna Sahay, and Carlos Ve´gh. 1996b. ‘‘Economies in Transition: The Beginnings of Growth.’’ American Economic Review 86, no. 2: 229–233. Fischer, Stanley, Ratna Sahay, and Carlos Ve´gh. 1998. ‘‘From Transition to Market: Evidence and Growth Prospects.’’ In Lessons from the Economic Transition: Central and Eastern Europe in the 1990s, ed. Salvatore Zecchini, 79–101. Norwell, MA: Kluwer Academic Publishers. Garibaldi, Pietro, Nada Mora, Ratna Sahay, and Jeromino Zettelmeyer. 2002. ‘‘What Moves Capital to Transition Economies?’’ IMF Staff Papers 48, no. 4: 109–145. Hall, Robert E., and Charles I. Jones. 1999. ‘‘Why Do Some Countries Produce So Much More Output Per Worker than Others?’’ The Quarterly Journal of Economics 114, no. 1: 83–116. Hare, Paul. 2003. ‘‘Institution Building in the CIS-7: Role of the International Community.’’ Paper presented at the Lucerne Conference of the CIS-7, January. Available at http://www.cis7.org. Havrylyshyn, Oleh, and Ron van Rooden. 2003. ‘‘Institutions Matter in Transition, But So Do Policies.’’ Comparative Economic Studies 45, no. 1: 2–24. Helbling, Thomas F., Ashoka Mody, and Ratna Sahay. 2004. ‘‘Debt Accumulation in the CIS-7 Countries: Bad Luck, Bad Policies, or Bad Advice?’’ In The Low-Income Countries of the Commonwealth of Independent States, eds. C. Shiells and S. Sattar, 15–50. Washington, D.C.: International Monetary Fund. Longer version issued as Working Paper No. WP/04/93, IMF, Washington, D.C. Rodrik, Dani, Arvind Subramanian, and Francesco Trebbi. 2004. ‘‘Institutions Rule: The Primacy of Institutions Over Geography and Integration in Economic Development.’’ Journal of Economic Growth 9: 131–165. Stiglitz, Joseph. 2002. Globalization and Its Discontents. New York: W. W. Norton. World Bank. 1999/2000. ‘‘Entering the 21st Century: World Development Report 1999/2000.’’ Available at http://www.worldbank.org/wdr/2000/fullreport.html. World Bank. 2002. ‘‘Building Institutions for Markets.’’ World Development Report. 2002. The World Bank, Washington, D.C.
15
Growth Collapses Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
15.1
Introduction
According to the 2006 World Development Indicators, no developed economy attained its peak per capita GDP before 2000. By contrast, 53 percent (61 out of 116) of developing economies saw their best times before that year. Of these, more than 40 percent (26) saw their best peak GDP before 1980. This number is even higher in particular regions of the developing world: in Latin America, 56 percent (15/27) of economies saw peak output before 2000, while in sub-Saharan Africa the corresponding figure is 74 percent (25/34). Recessions in the developing world are much deeper and longer than in the developed world. Figure 15.1 presents histograms of peak-trough ratios and duration of output contractions for developing and developed countries.1 The comparisons are striking. Of recessions in the developed world, 87.8 percent have peak-trough ratios less than 5 percent of peak GDP. The remaining 12.2 percent have peak-trough ratios between 5 and 15 percent of peak GDP. For developing countries, the corresponding figures are 48.3 percent and 25.4 percent, with the remaining 26.3 percent having contractions where the peak-trough ratio exceeds 15 percent of peak GDP. In terms of duration, 88.9 percent of recessions in developed economies last less than four years. In the developing world, the corresponding figure is 63.6 percent. The phenomenon of deep and prolonged recessions constitutes a challenge to macroeconomic theory. At the very least, it suggests that a vision of economic fluctuations as trends around a stable and growing level of potential GDP is problematic for understanding growth in the developing world. It also suggests that understanding the causes of low-frequency fluctuations may be key in accounting for differences in development performance. Guillermo Calvo has been one of the pioneers of the study of deep recessions in developing economies. In a series of seminal papers,2 Calvo and his coauthors have strived to achieve a complete characterization of a specific type of economic
378
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
Figure 15.1 Comparison of Crisis Characteristics
contraction: output collapses that occur in the context of sudden stops in capital flows in developing economies that are highly integrated into world financial markets. Among their most important conclusions is that these output collapses tend to be followed by rapid recoveries of output despite the lack of recoveries of either domestic or foreign credit. Our paper goes one step further and attempts to tackle a broader question. Instead of focusing on particular types of output collapses, we ask what can be learned from studying the distribution of economic contractions across countries and over time. In other words, we investigate whether it is possible to go beyond the aggregate characterization shown in figure 15.1 to a deeper understanding of the causes behind prolonged recessions. In this paper we will study both the events that coincide with the onset of crises and the determinants of the duration of crises. Among our main results, we find that a number of events—wars, export collapses, sudden stops in capital flows, and high levels of inflation—coincide with the onset of crises. However, we find that the duration of crises is particularly difficult to predict. Aside from region- and time-specific effects, we find that
Growth Collapses
379
a measure of export flexibility—given by the distance-weighted density of unexploited export products—which may capture the flexibility of the economy to adapt to external shocks, is an important predictor of crisis recovery. We also find that both conditional and unconditional hazard rates are declining in time, suggesting that countries have a harder time exiting crises the longer that they spend in them. Francis Galton once criticized his statistician colleagues because they ‘‘limited their inquiries to averages, and do not seem to revel in more comprehensive views’’ (1889, 62). Sir Galton could have just as well been referring to modern growth empirics. Ever since Barro’s (1989) seminal contribution, empirical work on economic growth has to a great extent been concerned with explaining differences in average growth among countries. Despite Pritchett’s (2000) call to think closely about the ‘‘hills, plateaus, mountains, and plains’’ characterizing the growth data, until recently very little work had been carried out attempting to explain the substantial differences in patterns taken by growth series with similar first and second moments.3 Among the notable exceptions are the work of Calvo, Izquierdo, and Mejı´a (2004); Calvo, Izquierdo, and Loo-Kung (2005); and Calvo, Izquierdo, and Talvi (2006). These authors have concentrated on studying the characteristics of output collapses that are associated with sudden stops in capital flows in emergingmarket economies. They have found that some of these sudden stops—particularly those associated with systemic international capital market turmoil—are followed by rapid output recoveries arising in spite of the limited reconstruction of credit markets. By concentrating on a very specific form of output collapses, Calvo and his coauthors have been able to take advantages of the similarities between comparable episodes. Other works in this literature have taken a more general approach, attempting to exploit the differences arising in broader samples. Ben-David and Papell (1998), for example, study episodes of growth slowdowns, defined by statistically significant breaks in the time-series trends, and find differences between the magnitude and timing of these slowdowns among developed and developing countries. Pritchett (2000), in contrast, studied declines of at least two percentage points in trend growth rate—without distinguishing between significant or insignificant trends. Neither of these papers makes a serious attempt to understand the causes behind the onset or magnitude of growth slowdowns. Two recent papers have made a more direct attempt to understand the dynamics of growth decelerations. Cerra and Sexena (2005) study the long-run implications of economic recoveries. In essence, they show that after economic contractions, GDP growth does not typically return to trend growth. Additionally, they show that the incidence of crises is an important reason for unconditional
380
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
divergence in the postwar growth data.4 Reddy and Miniou (2006) study ‘‘real income stagnations,’’ which they define as long and sustained periods of negative growth. Their definition is perhaps the one that comes closest to ours in the literature.5 They find that countries that suffered spells of real income stagnation were more likely to be poor, located in Latin America or Africa, undergoing armed conflict, and highly dependent on primary exports.6 The main characteristics of these and other studies dealing with long-run economic contractions are summarized in table 15.1. One common feature of all these studies is the lack of any explicit method to deal with unfinished episodes of contraction (meaning those that have not ended by the last year of the dataset). Of the ten papers listed in table 15.1, five take an explicit decision to omit or truncate the unfinished contraction episodes; two additional ones (Ben-David and Pappell 1998 and Pritchett 2000) adopt a methodology that is incapable of handling these breaks, thus also dropping them in practice. An eighth paper (Calvo, Izquierdo, and Talvi 2006) adopts a very restrictive definition of crises—as drops in aggregate output that occur in the context of systemic capital market turmoil—with the result that recoveries are very rapid and in their main sample there are no episodes of unfinished crises. This decision comes at a cost: there is no obvious reason from a growth theory perspective why one should concentrate on aggregate instead of per capita or per worker output. The two remaining papers analyze the change in growth rates among two predetermined periods and thus do not have to deal directly with this issue. In those papers, the issue of unfinished crises is avoided by choosing an arbitrary cut-off date to calculate changes in growth rates. The issue of dealing with unfinished episodes is important because a substantial number of crisis episodes have generally not finished by the end of the period for which continuous data is available. Using the definition of crisis that we will adopt in this paper (periods of continuous negative average growth), we find that 16.07 percent of the events defined as crises are censored. While this number may not seem large, what is problematic is that it is asymmetrically composed of long crises episodes. Only 4.95 percent (18/363) of crisis episodes lasting less than 5 years are censored, while 78.5 percent (22/28) of those lasting more than 24 years are censored. To take a simple example of how this can affect even the most basic conclusions of research, if we had decided to drop the censored observations we would have calculated the mean duration of contractions to be 3.96 years. If we include the censored observations, we find that the mean duration is 6.05 years. Even this estimate is surely an underestimate of the expected duration of a crisis, because the real duration of censored episodes is unobserved. If we were to fit the simplest possible duration function—with an exponentially distributed hazard rate—to this data, we would estimate an expected duration of 7.21 years.
Growth Collapses
381
Inferences about the behavior of countries during crises will thus end up being automatically biased to reflect the performance of economies that are more successful at dealing with crises. The key contribution of this paper is to analyze the determinants of the duration of economic contractions using econometric methods designed explicitly to deal with censored observations. These methods, broadly grouped under the labels ‘‘duration analysis’’ and ‘‘survival analysis,’’ have gained increasing prominence in modern economics primarily through their application in microeconometric settings.7 Their key defining characteristic is the joint use of information on duration of censored and uncensored episodes in derivation of the likelihood function. Despite their obvious appeal for the study of many macroeconomic phenomena, their application in macroeconomic analysis has been limited. The largest proportion of papers that use duration analysis in macroeconomic settings study duration dependence of business cycles in developed countries (Mills 2001; Bodman 1998; Di Venuto and Layton 2005; Diebold and Rudebusch 1990; Mudambi and Taylor 1991; Sichel 1991). Applications to developing countries are scant. One recent exception is Mora and Siotis (2005), who estimated a conditional duration model of recessions in a sample of twenty-two emerging markets. Their specification is limited by the small size of their sample and the fact that they consider only external factors. The rest of the paper is organized as follows: section 15.2 discusses our definition of crises and our data set. Section 15.3 takes a first look at the summary statistics of crises. Section 15.4 examines the factors that coincide with the onset of crises, while section 15.5 discusses the results of regressions to account for the duration of crises. Section 15.6 concludes. 15.2
Defining the Event
For the purposes of the analysis in this paper, we define a crisis as an interval that starts with a contraction of output per worker and ends when the value immediately preceding the decline is attained again. Thus a crisis that occurs between times t and t þ j has by definition an average growth rate equal to 0 during that period, and negative during the period between t and any t þ j e for any e < j. A crisis cannot start if the country is already in crisis. In other words, if yt < yt1 , a crisis will start only if either (1) there is no ytj > yt1 , or, (2) if there is a ytj > yt1 , there is a ytp > ytj with p < j. This definition is illustrated in figure 15.2. This definition has several advantages for our object of study. First, it covers a selected group of growth decelerations. In particular, it covers growth decelerations where growth goes from being positive (as it must be before reaching a peak) to being negative between t and
Time series
1955–1993/ 1950–1993
1960–1989
1965–1996
1960–1992 (some until 1985)
1960–1999
1960–2001/ 1960–2000 Unbalanced panel
Paper
Ben-David and Papell (1998)
Rodrik (1999)
Ali and Elbadewi (1999)
Pritchett (2000)
Ros (2005)
Cerra and Sexena (2005)
Table 15.1 Summary of Existing Literature
192/154
70 developing countries
111
62 developing countries
110
74
Countries
WDI/ PENN 6.1
PENN 6.1
PENN 5.6
WDI
WDI
PENN 5.5
Database
Annual
Annual
—
10–21 years
15 years
Annual
Freq
Not breaks, but events of crisis
Not breaks, but events of crisis
One or zero breaks, depending on significance
One and the same for all countries in 1975 One and the same for all countries (1965– 1974)—(1975– 1996)
One or zero breaks, depending on significance
Breaks
Real GDP growth
Real per capita GDP
Real GDP, PPP 1985
Real per capita GDP growth
Real per capita growth, PPP
Real per capita PPP
GDP used
Recessions
Collapse
Plateau, mountain, hill, plains
Collapse
Collapse
Slowdown (if D growth but growth > 0). Meltdown (if growth < 0)
Taxonomy
Omit
Dichotomous classification; in practice equivalent to truncation
Method does not apply; in practice they are ommitted
Not pertinent
Not pertinent
Method does not apply; in practice they are ommitted
Crisis that did not end
382 Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
Data 1980– 2004 (But Y trend built since 1960)
1960–2001
1960–2003
1980–2003
Cespedes and de Gregorio 2005
Reddy and Minoiu 2006
Blyde, Daude, and Fernandez Arias (2006)
Calvo, Izquierdo, and Talvi (2006)
31 (emerging markets covered by EMBI)
71
119
71
WDI/IMF WEO
PENN 5.6
WDI
WDI/IMF WEO
Annual
Annual
Annual
Annual
Not breaks, but events of crisis
Not breaks, but events of crisis
Not breaks, but events of crisis
Not breaks, but events of crisis
Real GDP
Real PPP
Moving average ð1=3fðt 1Þ þ ðtÞ þ ðt þ 1Þg of Per capita GDP, constant LCU
Real GDP
Episode of output contraction
Collapse
Real income stagnation
Product contraction episode or recession
No cases in main sample
Omit
Truncate
Omit
Growth Collapses 383
384
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
Figure 15.2 Graphical Example of Crisis Definition
any t þ j e. But it does not cover all growth decelerations. In particular, we wanted to exclude the growth decelerations that one would find naturally in neoclassical growth models, such as those generated by convergence to the steady state. Our definition allows us to restrict ourselves to episodes where growth decelerates to a negative rate. Our interest in negative growth rates is related to our specialization to duration analysis. Although episodes of negative growth are a standard area of interest in macroeconomics, sustained episodes of negative growth are much harder to explain. Although it is certainly possible to account for negative growth using a neoclassical model, this requires that the economy suffer strong and sustained deteriorations in its fundamentals that are sufficiently strong to offset the effects of technical progress for periods of one or more decades. Thus long, sustained episodes of negative growth are only consistent with very large adverse changes in fundamentals, or dynamics in which those fundamentals do not recover. In order to calculate our crisis indicator, we used GDP per person of working age—henceforth GDPW—in constant local currency units from World Bank (2006). We use working-age population because it is the best widely available proxy for the size of the labor force, which most closely matches the concept of labor input in growth theory. We use local currency units because we are not interested in comparisons of levels among countries, and differences in purchasing power parity (PPP) adjustment factors across times can be arbitrary and have unintended consequences (see Reddy and Minoiu 2006). Our crisis definition immediately suggests several measures of crisis intensity that we calculate and use in the rest of the analysis. These are: 1. Duration: the number of years elapsed between the beginning and the end of the crisis.
Growth Collapses
385
2. Peak-trough ratio: the ratio between the value of GDPW immediately preceding the crisis and the lowest value it attains during the crisis, expressed as a ratio of the peak value. 3. Integral measure of years of lost output: this is the sum of all the gaps between peak and the GDP for each year of the crisis. It is an approximation to the integral above the output series and below a horizontal line drawn at the peak output, expressed as a fraction of peak output. 15.3
Collapses Big and Small
We start out by examining the general characteristics of crises and how they vary across regions. These summary statistics are presented in table 15.2. The first striking fact that we observe is that there is a wide dispersion of crisis characteristics in the world sample. The majority of crises observed do indeed appear to be of the typical business-cycle type: the median crisis duration is just two years, while the median peak-trough ratio and integral measure of years lost are respectively four and six percent of precrisis GDP. These median values, however, come from a highly skewed distribution. The average duration of crises, at 6.05 years, is three times as high as the median duration, while the mean-to median ratios of the peak-trough ratio and product-years lost are respectively 2.6 and 15.9. High dispersion of crises characteristics is also a feature of interregional variations. Although all regions appear to be hit by some short-lived recessions, longlived recessions are much more prevalent in the developing world. Thus, while the mean duration of a recession in industrialized countries is only 2.52 years, in Latin America it is 6.88 years and in Central and Eastern Europe it is 9.74 years. The median peak-to-trough ratio of a crisis, for example, is 6 times as large in Africa and 29 times as large in Central and Eastern Europe as in the industrialized world. These differences are striking, but what is even more striking is that they almost surely underestimate the magnitude of the differences in crisis duration, both around the world median and across regions. The reason is a simple one, and will form the foundation for much of our analysis. Shorter crises are much less likely to be censored than long crises. Almost by definition, the longer a crisis is, the more likely that it will be interrupted, either by the end of the sample or by an interruption in data reporting. Therefore we can never actually observe the complete duration of very long crises. It is hard to exaggerate the magnitude of this difference. In our sample, only 19 out of 387 crises (4.9 percent) that last less than five years are censored, while 67 of 148 crises (45.3 percent) lasting more than five years are censored. Take the example of Venezuela, whose GDP per working-age population was 39.4 percent lower in 2004 than in its peak in 1970.
535
Total
0.08 0.02
42 34 46
Central and Eastern Europe
East Asia and Pacific
63 535
Middle East and North Africa
Total
0.07
46
East Asia and Pacific 90
34
Central and Eastern Europe
109
42
Asia
Industrialized Countries
151
Product-years lost (integral) Africa
Latin America and Caribbean
0.31
63 535
Middle East and North Africa Total
1.00
1.16
1.05
2.71
0.51
1.42
0.12 0.11
0.10
90 109
Industrialized Countries
Latin America and Caribbean
0.29
0.08
151
Asia
0.13
6.05
6.88 5.13
2.52
3.78
9.74
4.79
8.14
Mean
Africa
Peak to trough
90
46
East Asia and Pacific 109 63
34
Central and Eastern Europe
Latin America and Caribbean Middle East and North Africa
42
Industrialized Countries
151
Africa
Number of observations
Asia
Duration of crisis
Table 15.2 Summary Statistics of Crises by Region
2.36
2.88
2.35
0.16
0.54
2.81
1.16
2.94
0.20 0.16
0.13
0.03
0.08
0.24
0.12
0.16
8.02
8.92 7.56
2.88
4.23
6.38
5.71
10.09
Standard deviation
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00 0.00
0.00
0.00
0.00
0.00
0.00
0.00
1
1 1
1
1
1
1
1
Minimum
0.01
0.01
0.01
0.01
0.02
0.06
0.02
0.02
0.01 0.01
0.01
0.01
0.02
0.06
0.02
0.02
1
1 1
1
1
1
1
1
25th percentile
0.06
0.06
0.06
0.02
0.07
2.15
0.07
0.10
0.04 0.04
0.05
0.01
0.05
0.29
0.04
0.06
2
3 2
2
2
12.5
2
3
Median
0.44
0.20
0.62
0.06
0.26
4.03
0.26
0.93
0.12 0.13
0.15
0.03
0.11
0.45
0.08
0.19
7
7 4
3
5
15
7
13
75th percentile
16.40
12.73
13.78
1.16
2.46
9.79
5.51
16.40
0.91 0.95
0.62
0.13
0.35
0.77
0.52
0.95
43
34 27
19
20
19
24
43
Maximum
386 Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
Growth Collapses
387
Even if Venezuela were to experience extremely high growth after 2004, its crisis is likely to last considerably longer than 34 years. One possible mechanism for dealing with this problem would be to drop crises that have not ended from the sample. As discussed in our review of the literature, this is the standard choice in much of the literature dealing with this problem. However, it is a highly inefficient solution, as it entails throwing out a very high proportion of the crises that are most interesting for our purposes: those that are very long. Therefore we adopt the alternative solution in this paper, which is to use the estimation techniques of duration analysis (also called survival analysis in the biometric literature), which are explicitly designed to deal with censored duration times. The essence of duration analysis is to explicitly consider unfinished crises as arising out of the same distribution as finished crises. All countries are assumed to be characterized by a (possibly time-dependent) probability of leaving a crisis at any moment of time given that they are still in the crisis state. That probability— called the hazard rate—is affected by country-specific characteristics that can be summarized by a vector of independent variables that may include countryspecific effects. The behavior over time of this rate is very interesting. If the hazard rate is increasing over time, it means that as time elapses, the probability that a country will recover to its precrisis GDPW level increases. This is the type of behavior that one would expect if the precrisis level of GDPW was an equilibrium out of which the economy was perturbed by a temporary shock. However, if we see that the longer an economy spends in a crisis the harder it is for it to get out of it—as would be implied by a declining hazard rate—this would suggest either that the economy suffered strong blows to its fundamentals—so that its steady-state level of GDPW shifted downward—or that it jumped to an inferior equilibrium. Furthermore, in the presence of positive technological change one would always expect that, if enough time has elapsed after the initial shock, an economy would return to its precrisis level of GDP even if it initially suffered an adverse shock to its fundamentals. The reason is that for a given level of fundamentals, the probability that an economy hits any level of GDPW with positive technological change must tend to 1 as time increases. Declining hazard rates over the very long run would thus be consistent not only with an initial adverse shock, but rather with continuously deteriorating fundamentals as may occur because of the political system’s endogenous reaction to the adverse shock. A first intuitive way to summarize the information in our data set regarding the characteristics of crises is thus to plot the unconditional hazard functions. These hazard rate estimates are plotted in figure 15.1. They are derived as smoothed kernel density estimates of the Nelson-Aalen cumulative hazard function and track
388
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
Figure 15.3 Smoothed Hazard Function by Region
the probability of exiting a crisis conditional on being in the set of countries that have not exited the crisis when t periods have elapsed since falling into it. The remarkable result that emerges from figure 15.3 is that hazard rates, both within regions and for the world as a whole, do not appear to be increasing. Rather, they are either flat or decreasing over time. Two distinct patterns appear to emerge. The first one is that of industrialized countries and East Asia and the Pacific, which have two humps. Within these groups, countries either get out of their crises quickly, or get out later. Even the second group, however, tends to get out of crises much earlier than those in many other regions. The rest of the regions—including the pooled world sample—have hazard rates that are generally flat or decreasing, although confidence intervals obviously become wider as one reaches higher durations. Hazard rates also tend to be much lower in nonindustrial countries, reaffirming the conclusion that crises are much more likely to be of the short duration business-cycle type in industrialized countries than in underdeveloped regions. The confidence intervals drawn around the unconditional region-level hazard rates in figure 15.1 are quite wide. This is simply a reflection of the fact that very long crises are sufficiently rare within each region so as to make it difficult for us
Growth Collapses
389
Figure 15.4 Smoothed Hazard Function: Industrial and Nonindustrial Countries
to precisely estimate the hazard rate. In figure 15.4 we show the result of pooling the sample of nonindustrial countries and comparing it with that of industrial countries. The hazard rate for developing countries is clearly declining up at least to a period of thirty years since entry into crisis. It is also considerably lower than that of industrial countries. Whereas most industrial countries have a probability higher than 20 percent of leaving the crisis during each of the first few years in it, for developing countries that probability stays below 10 percent. The fact that substantial interregional differences across survival functions exist can be tested systematically through several standard tests for equality of survivor functions. These tests are reported in table 15.3. Column 1 shows the result of testing the null hypothesis that all regions have the same survival function. It is thus the statistical counterpart of figure 15.3. Column 2 tests the null that industrial countries have the same hazard function as developing countries, forming the statistical counterpart of the comparison in figure 15.2. Lastly, column 3 evaluates the null hypothesis that all groups of nonindustrial countries have the same survival function. All three homogeneity hypotheses are easily rejected. The data thus indicates that there is substantial interregional heterogeneity in the recovery from adverse shocks. The logical question that this analysis leads us to regards the source of these differences across regions. Is it a reflection of the fact that different regions are hit by
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
390
Table 15.3 Tests for Equality of Survivor Functions
All regions
Industrialized vs. nonindustrialized
Only nonindustrialized
Log-rank
54.49***
33.78***
21.31***
Cox
38.24***
20.78***
16.72***
Wilcoxon
37.36***
20.05***
17.84***
Tarone-Ware
45.38***
25.95***
19.82***
Peto-Peto
43.76***
24.58***
19.72***
different types of crises? Or is it rather a consequence of the fact that regions differ in their capacity to react to adverse shocks? In order to answer these questions, we must understand what factors drive countries to fall into crises and what factors determine the duration of crises once a country has entered into them.8 These are the questions that we tackle in the next two sections. 15.4
Why Do Countries Fall Into Crises?
We approach the study of the determinants of crisis occurrence through the estimation of random effects probit regressions on a panel of countries. The basic idea of this specification is to allow us to understand the potential relative triggers of a country falling into a crisis. In essence, we investigate whether and how different possible instigators correlate with the incidence of the crisis. We look at a battery of potential causes of crises, ranging from the usual suspects—natural disasters, wars, sudden stops, and export collapses—to other less conventional factors. Any exercise of this type may be subject to several types of specification bias. Simultaneity bias is one—though not necessarily the most important one—of them. Others include omitted variables, inadequacy of the linear specification, and incorrect assumptions about the error covariance structure. In the case of some potential explanatory variables—such as natural disasters—endogeneity may be less of a problem than some of these other biases. We do not make an effort to search for appropriate instruments for all of our explanatory variables because we view our exercise as a primarily exploratory attempt to investigate what factors coincide with the onset of crises, rather than to test tightly specified causal hypotheses. If our exercise is successful, it would help build a typology of crises according to the key factors that occur at the same time as the crisis. Our baseline specification will thus be Pit ¼ Fðb 0 Xit þ hi Þ
ð15:1Þ
Growth Collapses
391
where Pit is the probability that country i falls into a crisis at time t, Xit is a k 1 vector of conditioning variables (generally including a constant term), hi is a country-specific effect, b is the k 1 vector of parameters to be estimated, and Fð:Þ is the standard normal distribution. The random effects probit specification models hi as following a Nð0; s 2 Þ distribution. Note also that our event of interest is whether a country enters a crisis or not. According to our definition of crises, a country is obviously a candidate for entering a crisis only if it is not already in one. Therefore we exclude from the sample all country-years in which the country is in the midst of a crisis. This decision is based on the fact that these country-years contain no relevant information about the process of entering into crises. Neoclassical growth theory views output collapses as arising out of adverse shocks that either move the steady-state level of income or alter the per capita stock of physical and human capital. It is thus logical to start by looking at significant disruptions of an economy’s productive framework that may either affect its capacity to convert inputs into outputs or directly affect its stock of accumulated productive assets. Several candidates come to mind. Perhaps the first two are natural disasters and wars. These tend to constitute large, generally exogenous shocks that generate significant disruptions to a society’s capacity to produce. They will also commonly directly affect the capital stock. The speed of some postwar recoveries is indeed a commonly cited observation in defense of the conditional convergence hypothesis. Two other potential candidates are export collapses and sudden stops in capital flows. The latter has been well developed in the literature, particularly through the pioneering work of Guillermo Calvo. Export collapses, while much less studied, tend to crop up in the analysis of many episodes of collapse (see Hausmann and Rodrı´guez 2006). Our data for natural disasters is drawn from the International Disaster Database, maintained by the Office of U.S. Foreign Disaster Assistance (OFDA) and the Center for Research on the Epidemiology of Disasters (CRED), which has information on the physical and human damage caused by 14,877 natural disasters that occurred between 1960 and 2006. We define a substantial occurrence of natural disasters if either the number of people affected by all natural disasters occurring in a given year is greater than 1 percent of the population, or the number of people killed by all natural disasters occurring in a given year is greater than 0.1 percent of the population. Our proxy for natural disasters will be an indicator variable that will be 1 if there was a substantial occurrence of natural disasters in t, t 1, or t 2. Regarding the occurrence of wars, we draw our data from Kristian Gleditsch’s (2004) Expanded War Data Set, which covers all inter- and intrastate wars between and within independent states since 1816. We build an indicator variable that equals 1 if the country was involved in an interstate or civil war in t, t 1, t 2,
392
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
or t 3. We also build separate dummies for civil and interstate wars respectively. Declines in exports are measured using data on merchandise exports from World Bank (2006). We use the log difference in exports between t and t 5 as our indicator of an export performance. For the purposes of defining sudden stops in capital flows, we closely follow the definition of Calvo, Izquierdo, and Mejı´a (2004). According to their definition, a sudden stop is a year-on-year decline in capital inflows containing at least one year in which the decline exceeded two standard deviations from its sample mean. The sudden stop starts when the fall exceeds one standard deviation from the sample mean and ends when it is above one standard deviation. Our measure of private capital flows comes from World Bank (2006) and consists of private debt and nondebt flows. Note that our measure differs from two other measures used by Calvo, Izquierdo, and other coauthors in some of their work. In particular, it differs from the systemic sudden stops (Calvo, Izquierdo, and Talvi 2006) measure given by the episodes of sudden stops that coincide with increases in the aggregate EMBI spread. Aside from the difficulty in obtaining a measure of capital market turmoil relevant for nonemerging-market economies, our key reason for using the broader category is our interest in the broader phenomenon of declines in capital flows. Our measure also differs from the definition found in Calvo, Izquierdo, and Talvi (2006) that combines falls in capital flows with output collapses. The rationale for this is quite simple: we are attempting to understand the capacity of sudden stops to predict output collapses, whereas most of the work of Calvo and his coauthors is concentrated on understanding the dynamics of output collapses that coincide with a decline in capital flows.9 We shall discuss the sensitivity of our results to alternative measures of sudden stops later. Aside from these natural candidates, we try a number of additional explanatory variables that may be associated with the onset of crises. We measure the level of a country’s democracy by its score on the polity index (Marshall, Jaggers, and Gurr 2004). We also measure political transitions by the change over time in its polity index. We use a measure of the log of 1 plus the inflation rate as a proxy for macroeconomic instability. We also attempt to control for a set of additional potential explanatory variables such as years of primary, secondary, and total schooling (from Barro and Lee 2000), the rule of law (ICRG 1999), life expectancy at birth, percent of the population that is urban, and number of telephone mainlines per capita (from World Bank 2006). All of our estimates include region and decade dummies. One additional variable of interest that we will study is the measure of the value-weighted density of the unexploited product space elaborated by Hausmann and Klinger (2006). This measure is designed to capture the sophistication of the goods that an economy could produce—but is not producing—with its productive assets. It is built as a weighted average of the sophistication of all po-
Growth Collapses
393
tential export goods, where the weights are given by the distance between these goods and the economy’s present export basket. The measure of distance in the product space is calculated based on the frequency with which particular goodpairs are exported by the same country, while the measure of sophistication is given by the average income of the countries that export that good, which we call PRODYjt , as originally proposed by Hausmann, Hwang, and Rodrik (2007). More formally, let the proximity between two goods in the product space be given by the minimum of the conditional probabilities of exporting each one of those goods given that you are exporting the other one: jijt ¼ minfpðxit j xjt Þ; pðxjt j xit Þg;
ð15:2Þ
where pðxit j xjt Þ is the probability that you have revealed comparative advantage in good i at time t given that you have revealed comparative advantage in good j at time t. Let xcjt be an indicator variable that takes the value 1 if country c has a revealed comparative advantage greater than 1 in good j at time t, and 0 otherwise. Then the option value of a country’s unexploited export opportunities can be measured by open_ forestct ¼
X X jijt P ð1 xcjt Þxcit PRODYjt : i jijt i j
ð15:3Þ
open_ forest thus captures the flexibility of an economy’s export basket in that it measures the value of the goods that it could be producing with the inputs that it currently devotes to its export production. open_ forest is particularly appropriate for thinking about an economy’s capacity to react to adverse export shocks. To fix ideas, suppose that an economy’s exports of good i were to disappear overnight. This could happen, for example, as a result of the exhaustion of a natural resource, of the emergence of a new lower-cost supplier in international markets, or of the invention of a cheap substitute for that good. We know that this economy must shift resources into a new export sector. jijt can be interpreted as our best guess of the probability that a country will shift resources into good j, and jijt ð1 xcjt Þ can be seen as our best guess of the probability that it will export a good j that it is not already exporting. jijt ð1 xcjt ÞPRODYjt is the expected value (measured in terms of the sophistication of exports) from exporting that good, making open_ forest the weighted average of that expected value over all goods that the economy currently exports. In other words, open_ forest reflects the expected value of an economy’s next-best export basket if it moved out of its current basket of exports. To the extent that many crises are precipitated by declines in an economy’s key export sectors, we expect open_ forest to be a good indicator of the economy’s flexibility in moving to a new export basket in the face of those declines. We
394
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
thus expect open_ forest to be a significant determinant of the duration of crises. Open_ forest may also be important in stopping crises before they materialize. The reason is that export declines in traditional sectors may occur at the same time as new sectors are moving in to absorb unused resources. The higher the productivity of the newer export sectors, the less likely that the initial export collapse will cause the economy to enter a period of negative growth. In order to test these hypotheses, we will include open_ forest in the probit regressions of this section as well as the duration regressions of the next section. Our baseline results are presented in table 15.4. Column 1 presents the result of regressing the probability of falling into crisis on the log of PPP-adjusted real GDP per working-age population and a set of continent and time dummies. The GDP term is negative, indicating that richer countries are less prone to economic crises, although the coefficient is not significant, in contrast to several of the continent dummies, which are highly significant. In column 2 we add the log change in real merchandise exports. Its coefficient is strongly significant with the expected negative sign, while the coefficient on GDP remains insignificant. The next column adds wars, natural disasters, and sudden stops. While wars and sudden stops are highly significant with the expected sign, the result on natural disasters is surprising. The coefficient is far from statistical significance (p ¼ .36) and furthermore has the wrong sign. This is particularly surprising since natural disasters are the one variable in the data set about whose endogeneity we are less worried. In order to confirm that its coefficient is not being distorted by the endogeneity of other explanatory variables, we reestimated the equation, dropping all variables except for natural disasters and time and continent dummies. This exercise (not shown in the table) still gives an insignificant, though positive, coefficient (p ¼ .47). Column 4 adds four additional variables: inflation, political transitions, open_ forest, and the level of democracy. The first two are strongly significant, open_ forest is borderline significant, and democracy is clearly insignificant. All have the expected sign: inflation and political change are associated with greater propensity toward crises, while open_ forest and democracy are associated with lower crisis prevalence. Wars now drop to borderline insignificance. This appears to be more the result of reduced sample size: if we reestimate the equation of column 3 for the same number of observations as in column 4, we get a very similar coefficient as when we include the additional variables (.451, t-stat ¼ 1.7). Finally, in the last column, we drop the clearly insignificant natural disasters and democracy variables. The coefficient signs and significance tests are unaffected, with the exception of wars, which goes from borderline insignificance to borderline significance. Declines in merchandise exports, sudden stops, high inflation episodes, and political transitions are individually strong predictors of the onset of crises, while wars and open forests have a weaker but still significant association with crisis onset.
Table 15.4 Random Effects Probit Regressions, All Countries Dependent variable: Probability of falling into a crisis Log GDP per working-age person
(1) 0.014 (0.24)
(2)
(3)
(4)
(5)
0.051 (0.85)
0.002 (0.03)
0.044 (0.44)
0.015 (0.15)
Latin America
0.491 (4.01)***
0.590 (4.61)***
0.610 (4.48)***
0.250 (1.23)
0.238 (1.2)
Africa
0.409 (2.58)***
0.547 (3.32)***
0.479 (2.64)***
0.072 (0.24)
0.047 (0.16)
0.136 (0.73)
0.025 (0.13)
0.022 (0.1)
0.303 (1.08)
0.351 (1.28)
East Asia and Pacific
0.073 (0.53)
0.217 (1.54)
0.187 (1.11)
0.122 (0.54)
0.162 (0.72)
Central and Eastern Europe
0.190 (1.24)
0.236 (0.96)
0.171 (0.55)
0.105 (0.26)
0.185 (0.46)
Middle East and North Africa
0.430 (3.04)***
0.149 (0.64)
0.124 (0.58)
1970s
0.310 (2.59)***
1980s
0.319 (2.93)***
South and Central Asia
2000s 1990s
0.231 (2.1)**
Log change in real merchandise exports
0.547 (3.78)***
0.492 (3.19)***
0.494 (1.47) C0.177 (1.65)*
0.049 (0.42)
C0.486 (3.77)***
C0.298 (2.12)**
C0.287 (2.67)***
C0.206 (1.77)*
C0.446 (5.01)***
C0.450 (4.73)***
War
0.655 (2.97)***
Natural disaster Sudden stop
0.527 (1.64)
0.026 (0.19) 0.495 (1.48)
0.394 (1.23)
0.104 (0.72)
C0.433 (3.04)***
C0.416 (2.95)***
0.415 (1.56)
0.468 (1.81)*
C0.121 (0.92)
0.070 (0.41)
0.167 (2.1)**
0.236 (2.32)**
0.227 (2.24)**
0.999 (3.23)*** 0.317 (2.48)**
1.009 (3.28)*** 0.365 (2.92)***
Log of inflation Change in polity indicator Open forest
C0.148 (1.73)*
Democracy
0.003 (0.27)
C0.161 (1.91)*
Constant
C1.293 (2.14)**
C1.382 (2.26)**
1.073 (1.56)
0.007 (0)
Observations
2001
1877
1688
1054
158
147
136
83
83
0.0% 3.0%
0.3% 5.3%
1.4% 6.0%
5.7% 7.5%
6.1% 7.9%
Countries Percent crises predicted Pseudo-R 2
0.944 (0.67) 1062
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: * 10%, ** 5%, *** 1%.
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
396
What can we say about the global significance of these variables? There are various ways to address this question. One is by noting that the addition of the explanatory variables drives down the significance of the continent and time dummies. All continent dummies except for South and Central Asia and Central and Eastern Europe are significant and positive in column 1. Since the omitted category is industrialized countries, this indicates a higher unconditional probability of falling into crisis for nonindustrialized countries. By column 5, those effects have disappeared—indeed, the South and Central Asia dummy has turned significantly negative. It thus appears that our explanatory variables account for the differences between the developing and developed worlds in the incidence of crises.10 A second way to address this question is by looking at some goodness-of-fit indicators. These are reported in the bottom two rows of table 15.4. The result of this exercise is not as encouraging. The first column shows the percentage of crises that are accurately predicted by our models. Even our most satisfactory model of column 5 only predicts 6.11 percent of crises adequately. This is also reflected in the pseudo-R 2 measure of McFadden (1974), which is given by Rp2 ¼ 1
Gu ; G0
ð15:4Þ
with Gu and G0 respectively denoting the log-likelihood of the unrestricted model and the model with only a constant. Rp2 hovers between .03 and .079, indicating a poor capacity of the model to predict the onset of crises. The goodness-of-fit results, however, should be interpreted with caution (see Wooldridge 2001, 465–466 for a discussion). The timing of a crisis is likely to be a very uncertain event, and it would indeed be surprising if we were able to correctly predict it a very high number of times. Indeed, the flip side of our inability to adequately predict the onset of crises is that the models are very good at predicting the nonoccurrence of crises. Since the entry into a crisis is a rare event (compared to staying out of the crisis), almost all prediction rules are naturally going to be very conservative. An alternative way to ask about the quality of our model’s predictive capacity is by looking at the times that the model correctly predicts a crisis (that is, when the estimated probability exceeds 0.5) as a percentage of the total number of times it predicts a crisis. In our baseline model of column 5, this percentage is a much more reasonable 52.4 percent. The economic significance of our coefficient estimates can best be interpreted by studying the marginal effect of changes in the explanatory variables on the estimated probability of a crisis. These effects are displayed in table 15.5, which also shows the effect of a one-standard-deviation increase in each of the explanatory variables. By this metric, the largest single effect comes from inflation: a one-
Growth Collapses
397
Table 15.5 Marginal Effects of Explanatory Variables, Baseline Estimation dp/dx Continuous variables Log of per worker GDP Log change in manufacturing exports Lof(1 þ Inflation) Open forest
sd(x)
(dp/dx) sd(x)
0.0035
1.0401
0.0036
0.0972
0.5699
0.0554
0.2360
0.3347
0.0790
0.0376
0.9898
0.0372
Indicator variables War
0.1338
0.1337
0.0179
Sudden stops
0.0549
0.4374
0.0240
Political transitions
0.0955
0.4416
0.0422
Latin America
0.0605
0.3864
0.0234
Africa
0.0113
0.4273
0.0048
Asia East Asia and Pacific
0.0713 0.0354
0.2587 0.3307
0.0185 0.0117
Central and Eastern Europe
0.0130
0.0392
0.3307
Middle East and North Africa
0.0305
0.2948
0.0090
1980s
0.0062
0.4163
0.0026
1990s
0.0239
0.4163
0.0099
2000s
0.0896
0.3150
0.0282
standard-deviation increase in our inflation indicator increases the probability of a crisis by 7.9 percent. The effect of manufacturing exports, however, is also substantial: a one-standard-deviation decline in the rate of growth of manufacturing exports causes an increase of 5.54 percentage points in the probability of a crisis. For indicator variables, a more natural metric is to think of the effect of the variable changing from one to zero. By this metric, a war is by far the most destructive single event, causing an increase in the probability of a crisis of 13.38 percentage points. By contrast, a political transition costs 9.55 percentage points and a sudden stop costs a 5.49 percentage point-increase in the probability of crises. These coefficients should be interpreted with the already mentioned caveat about causality. In section 15.4 we showed that there were substantial interregional differences —particularly between industrial and developing regions—in the characteristics of crises. In tables 15.6 and 15.7 we look at this issue more systematically by splitting the sample between developing economies and industrial economies. We indeed find important differences in the results in the two subsamples. Export declines, inflation, political transitions, and open_ forest retain their effect in the subsample of developing countries. Curiously, wars and sudden stops lose some
Table 15.6 Random Effects Probit Regressions, Developing Countries Dependent variable: Probability of falling into a crisis Log GDP per working-age person
(1) 0.013 (0.21)
(2)
(3)
(4)
(5)
0.057 (0.85)
0.012 (0.15)
0.112 (0.98)
0.067 (0.61)
Latin America
0.329 (2.02)**
0.871 (3.3)***
0.841 (2.54)**
0.426 (1)
0.496 (1.18)
Africa
0.246 (1.43)
0.841 (3.04)***
0.735 (2.12)**
0.259 (0.53)
0.307 (0.64)
South and Central Asia
C0.335 (1.68)*
0.288 (0.98)
0.182 (0.5)
0.128 (0.27)
0.088 (0.19)
East Asia and Pacific
0.107 (0.62)
0.500 (1.84)*
0.446 (1.28)
0.044 (0.1)
0.088 (0.2)
Middle East and North Africa
0.261 (1.46)
0.847 (3.05)***
0.751 (2.18)**
0.323 (0.72)
0.383 (0.87)
1970s
0.236 (1.67)*
0.422 (2.78)***
0.239 (1.45)
0.153 (0.39)
0.156 (0.4)
1980s
0.330 (2.56)**
0.285 (2.1)**
0.208 (1.43)
0.216 (0.57)
0.212 (0.56)
1990s
0.292 (2.24)**
0.210 (1.51)
0.091 (0.61)
0.153 (0.41)
0.134 (0.36)
C0.498 (5.17)***
C0.509 (4.87)***
C0.565 (3.42)***
C0.531 (3.27)***
0.645 (2.44)**
0.326 (1.03)
0.403 (1.33)
Log change in real merchandise exports War
0.072 (0.52)
0.144 (0.79)
0.105 (1.06)
0.205 (1.57)
0.185 (1.44)
Log of inflation
0.817 (2.49)**
0.826 (2.54)**
Change in polity indicator
0.251 (1.84)*
0.309 (2.34)**
Natural disaster Sudden stop
Open forest Democracy
C0.163 (1.69)* 0.005 (0.5)
0.178 (1.88)*
Constant
C1.129 (1.85)*
C2.160 (3.11)***
C1.637 (2.02)**
C0.207 (0.12)
0.351 (0.2)
Constant
3.953 (3.21)***
3.801 (3.37)***
3.527 (3.55)***
3.273 (2.9)***
3.341 (2.9)***
Observations
1379
1277
1129
686
694
133
122
111
64
64
0.0% 2.4%
0.8% 5.8%
1.9% 6.1%
8.4% 7.4%
6.6% 7.7%
Countries Percent crises predicted Pseudo-R 2
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: * 10%, ** 5%, *** 1%.
Growth Collapses
399
Table 15.7 Random Effects Probit Regressions, Industrialized Countries Dependent variable: Probability of falling into a crisis Log GDP per working-age person
(1)
(2)
0.227 (1.15)
0.104 (0.47)
0.164 (0.58)
0.483 (1.02)
0.476 (1)
0.178 (0.69)
0.001 (0)
0.385 (0.86)
0.349 (0.78)
0.730 (1.71)*
0.696 (1.24)
0.621 (1.15)
5.672 (0)
5.310 (0)
Log change in real merchandise exports War Natural disaster
(3)
(4)
(5)
Sudden stop
0.373 (2.49)**
0.407 (2.09)**
0.403 (2.09)**
1970s
0.409 (1.28)
4.988 (0)
4.969 (0)
1980s
0.363 (1.53)
5.023 (0)
5.014 (0)
1990s
0.021 (0.09)
5.225 (0)
5.174 (0)
Log of inflation
11.970 (3.98)***
11.345 (3.98)***
Political transitions
0.356 (0.6)
0.195 (0.34)
0.184 (0.55)
0.144 (0.44)
Open forest Democracy (polity index) Constant Observations Countries Percent crises predicted Pseudo-R 2
0.323 (0.78) 0.156 (0.07)
3.247 (1.08)
12.656 (0.01)
9.905 (0)
622
600
559
368
368
25
25
25
19
19
0.0% 1.4%
0.0% 3.4%
2.5% 5.7%
5.3% 7.3%
6.6% 7.7%
1.146 (0.56)
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: * 10%, ** 5%, *** 1%.
400
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
significance in this exercise—but in the case of wars the effect again seems to come from the restriction of the sample. By contrast, exports, political transitions, and open_ forests appear not to be relevant in industrial countries. Inflation retains strong significance, with a much higher absolute coefficient estimate that reflects the much smaller ranges of variation of this variable in developed economies. Wars—which in the case of developed economies are almost always overseas interstate conflicts such as the Gulf War—are also insignificant in this subsample. While these differences are certainly interesting, the evidence that the datagenerating process is fundamentally different across regions is not all that strong. The developed country sample is smaller, so it is logical to expect broader confidence intervals. All of the variables that we found to be significant in the broader sample have the same sign in both subsamples, and in most cases—with the notable exception of the inflation rate—the coefficient estimates are strikingly similar. The similarity of these coefficient estimates suggests that the key reason for the difference in the frequency of crises across regions comes not so much from differences in the way in which these crises are generated, but rather from differences in the distribution of the underlying determinants. Another potential source of structural differences may come from the fact that very lengthy or costly crises may be associated with different factors from those that generate shorter crises. Table 15.8 examines this hypothesis by splitting the sample between short and long crises. We split these in two dimensions: crisis duration of five years (columns 1 and 2) and crisis duration of more than 75 percent of the last precrisis year’s GDP (columns 3 and 4). This exercise provides some very interesting results. Regardless of whether one uses the duration or the lost output splits, one is struck by the similarity of the coefficient estimates for exports, wars, sudden stops, inflation, and political transitions. There are, however, striking differences between the effects of open forests and the Latin America dummy across subsamples. These suggest that open forests and some unobserved characteristics of Latin America may not be so much a predictor of crisis onset (for which its significance in the whole sample is at best weak), but rather of crisis duration. The bulk of the estimates, however, suggest that one can get thrown into short and long crises for very similar reasons. The substantial difference, therefore, may be in how one recovers from these crises. The next three tables include a series of additional robustness tests for our baseline specification. Table 15.9 studies the effect of adopting alternative definitions of capital flows. One problem with the capital flows window measure (which is discussed at length in Calvo, Izquierdo, and Mejia 2004) is that the decline in capital flows may be caused by an increase in export capacity. In column 1 we use a measure that combines the decline in capital stock with the condition that imports must also have declined. The coefficient on this measure, while positive, is not
Growth Collapses
401
Table 15.8 Random Effects Probit Regressions by Intensity of Crisis Dependent variable: Probability of falling into a crisis
(1) Duration < 5 years
Log GDP per working-age person
0.088 (0.79)
0.193 (1.33)
0.105 (0.98)
0.277 (1.67)*
Log change in real merchandise exports
C0.275 (1.73)*
C0.569 (2.59)***
C0.358 (2.33)**
C0.568 (2.35)**
War
0.428 (1.49)
0.419 (1.07)
0.345 (1.2)
0.652 (1.56)
Sudden stop
0.203 (1.82)* 0.868 (2.57)**
0.150 (0.91) 0.899 (2.08)**
0.235 (2.19)** 0.965 (2.97)***
0.061 (0.31) 0.838 (1.79)*
0.308 (2.16)**
0.468 (2.54)**
0.290 (2.16)**
0.494 (2.31)**
Log of inflation Political transitions
(2) Duration > 5 years
(3) (4) Integral < 0:75 Integral > 0:75 GDP-years output years
Open forest
0.020 (0.19)
Latin America
0.121 (0.52)
0.947 (2.85)***
Africa
0.108 (0.31) C0.548 (1.75)*
0.473 (1.02) 0.182 (0.39)
0.003 (0.01) 0.473 (1.61)
0.705 (1.22) 0.163 (0.25)
East Asia and Pacific
0.373 (1.49)
0.489 (1.25)
0.280 (1.2)
0.833 (1.59)
Central and Eastern Europe
0.370 (0.82)
0.471 (0.75)
0.477 (1.05)
1.125 (1.62)
Middle East and North Africa
0.076 (0.32)
0.362 (0.93)
0.107 (0.48)
0.496 (0.94)
1970s
0.366 (1.09)
0.409 (1.21)
4.937 (2.1)**
1980s
0.302 (0.94)
0.179 (0.8)
0.364 (1.13)
5.140 (2.16)**
1990s
0.198 (0.62)
0.027 (0.11)
0.350 (1.09)
4.569 (1.9)*
0.872 (0.53)
3.216 (0)
South and Central Asia
C0.411 (3.5)***
0.023 (0.23)
C0.512 (3.96)***
0.050 (0.23)
1.192 (2.61)***
5.374 (0)
2000s Constant
0.393 (0.23)
Observations
1004
933
1023
921
81
83
81
83
Countries Percent crises predicted Pseudo-R 2
0.8% 4.8%
1.459 (0.71)
3.9% 20.6%
2.1% 5.1%
2.6% 27.2%
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: * 10%, ** 5%, *** 1%.
Table 15.9 Random Effects Probit Regressions, Alternative Sudden Stop Definitions Dependent variable: Probability of falling into a crisis Log GDP per working-age person
(1)
(2)
(3)
(4)
0.027 (0.28)
0.036 (0.38)
0.055 (0.54)
0.049 (0.48)
C0.409 (2.91)***
C0.400 (2.82)***
C0.360 (2.3)**
C0.369 (2.37)**
War
0.457 (1.76)*
0.459 (1.77)*
Sudden stop 1
0.118 (1.02)
Log change in real merchandise exports
Sudden stop 2
0.540 (1.97)**
0.564 (2.06)**
0.060 (0.62)
Sudden stop 3
0.164 (1.59)
Sudden stop 4
0.221 (1.9)*
Log of inflation
1.014 (3.3)***
1.010 (3.29)***
0.985 (3.13)***
0.954 (3.03)***
Political transitions
0.363 (2.91)***
0.357 (2.86)***
0.369 (2.82)***
0.368 (2.81)***
Open forest
C0.160 (1.89)*
C0.165 (1.95)*
C0.201 (2.27)**
C0.197 (2.22)**
Latin America
0.226 (1.14)
0.225 (1.14)
0.204 (0.99)
0.193 (0.93)
Africa
0.073 (0.25)
0.087 (0.29)
0.002 (0.01)
0.022 (0.07)
South and Central Asia
0.327 (1.2)
0.315 (1.15)
0.278 (0.99)
0.298 (1.06)
East Asia and Pacific
0.127 (0.57)
0.107 (0.48)
0.117 (0.52)
0.144 (0.63)
Central and Eastern Europe
0.218 (0.54)
0.234 (0.58)
0.134 (0.38)
0.129 (0.37)
0.116 (0.54) 0.477 (1.44)
0.110 (0.51) 0.467 (1.41)
0.129 (0.59) 0.608 (1.67)*
0.094 (0.43) 0.625 (1.72)*
1980s
0.491 (1.54)
0.478 (1.5)
0.627 (1.8)*
0.639 (1.83)*
1990s
0.366 (1.15)
0.353 (1.11)
0.461 (1.32)
0.495 (1.42)
Constant
0.395 (0.27)
0.374 (0.25)
0.505 (0.33)
0.548 (0.36)
Middle East and North Africa 1970s
Observations Countries Percent crises predicted Pseudo-R 2
1061
1061
1004
83
83
83
1002 83
0.0611 0.0760
0.0556 0.0753
0.0599 0.0836
0.0539 0.0845
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: * 10%, ** 5%, *** 1%.
Growth Collapses
403
significantly so. This may be because the combination of these two criteria is very stringent. In column 2 we relax it by defining a sudden stop to be any decline in capital flows that coincides with an import decline. This measure appears to be very poorly correlated with the onset of crises (p ¼ .538). Column 3 uses a definition based on total (as opposed to just private) capital flows, which we measure as the sum of the trade balance and the decline in reserves, combined with a decline in imports. This measure does somewhat better, nearing statistical significance (p ¼ .113). If we make this last definition somewhat more stringent by requiring declines in total capital flows to exceed 3 percent of GDP and import declines to exceed 5 percent of initial import values, significance increases slightly (p ¼ .058). We have carried out a substantial number of additional tests with many alternative definitions, and find that, while it is certainly possible to come up with definitions of sudden stops that are significantly associated with the onset of crisis, such a conclusion is not robust to changes in the way in which we define the event. An additional conclusion that can be drawn out of table 15.8 is that the incidence of our significant explanatory variables does not change with different choices of sudden stop indicators. Changes in exports, political transitions, high inflation, wars, and open_ forest maintain their patterns of association with the onset of crises in all alternative specifications. To this moment we have assumed that the country-specific effect hi in equation 15.1 is uncorrelated with the explanatory variables, giving rise to the random effects probit specification. This assumption is of course questionable, but relaxing it is problematic because of the well-known incidental parameters problem. An alternative is to use the fixed effects logit specification, where the coefficient vector b p can be estimated with n-consistency. This specification, however, is not without its cost. As discussed in detail by Wooldridge (2001, 492), it requires the conditional independence of the dependent variable given the explanatory variables. In our context, this implies assuming that the probability of a crisis is independent of the number of crises that have occurred in the past. Table 15.10 shows the results of this specification. There are important changes as well as similarities with the random effects probit specification. The key similarity lies in the coefficients for merchandise exports, the capital-flow window definition of sudden stops (which retain strong statistical significance), and wars (which are significant in the broader sample and borderline significant once one adds additional controls). The effects of inflation and political transitions are preserved, although with lower p-values than under the probit specification. The striking difference, however, lies in the changes in the log of per capita GDP, which is now strongly negative and significant—indicating that richer countries have less propensity to experience crises—and open_ forest, which is now positively, though insignificantly, related to the onset of crises.
404
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
Table 15.10 Fixed Effects Logit Specification Dependent variable: Probability of falling into a crisis
(1)
(2)
(3)
(4)
(5)
C2.137 (3.77)***
C1.704 (2.89)***
C1.611 (2.43)**
C3.593 (3.51)***
C3.607 (3.53)***
C0.925 (4.64)***
C0.931 (4.41)***
C1.245 (4.04)***
C1.242 (4.02)***
War
1.037 (2.54)**
0.840 (1.72)*
0.844 (1.73)*
Natural disaster
0.031 (0.11)
0.210 (0.6)
Sudden stop
0.352 (2.2)**
0.528 (2.54)**
0.528 (2.54)**
Log of inflation
1.334 (1.95)*
1.333 (1.95)*
Change in polity indicator
0.460 (1.8)*
0.454 (1.78)*
Open forest
0.347 (0.81)
0.352 (0.82)
Log GDP per working-age person Log change in real merchandise exports
1970s
C0.660 (1.83)*
1980s
0.119 (0.42)
0.125 (0.55)
0.394 (1.6)
0.599 (2.01)**
0.606 (2.04)**
1990s
0.176 (0.73)
0.335 (1.14)
0.454 (1.44)
1.089 (2.61)***
1.103 (2.65)***
0.232 (0.62)
0.457 (1.14)
0.651 (0.82)
0.660 (0.83)
2000s Observations Countries Percent crises predicted Pseudo-R 2
1906
1761
1571
986
145
126
116
75
986 75
0.00% 1.74%
0.00% 3.32%
0.00% 4.55%
0.00% 8.47%
0.00% 8.41%
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: * 10%, ** 5%, *** 1%.
Our last battery of robustness tests is displayed in table 15.11, where we study the effect of adding additional potential explanatory variables to the probit specification. In this table, we include average years of primary, secondary, and total schooling as measures of human capital’s effect on propensity to fall in crises. Neither of these measures is significant (columns 1–3). Neither is a measure of institutions (the rule of law), of physical infrastructure (telephone mainlines per capita), of urbanization, or of life expectancy. We can summarize the results of this section as follows: a number of variables appear to be associated with the onset of crises. In terms of robustness, the vari-
Growth Collapses
405
able that comes out on top is the log change in merchandise exports, which has come out as significant in all the specifications in which it is included, with the exception of the subsample of developed countries. In terms of economic significance, a one-standard-deviation increase in inflation appears to be associated with much more damage than a similar increment in any other variable. Most specifications coincide in a significant effect of the capital-flows window definition of sudden stops, as well as political transitions, on the probability of a crisis occurring. The effects of wars, initial income, and residual continent or time dummies are much more variable to specification. Particularly, while open_ forest comes out as a significant predictor in some specifications, its coefficient tends to be weak and its sign is reversed in the conditional logit specification. 15.5
How Do Countries Get Out of Crises?
In this section we analyze the determinants of crisis duration. Most existing contributions in the literature do not deal with the problem of censoring that naturally arises in the analysis of contractionary episode duration. As discussed in the introduction, the standard solution taken in the literature addressing this issue is to drop or truncate those observations. Either solution is inappropriate. Dropping the observations biases the sample toward short duration episodes, while truncating them inadequately represents crises as having shorter durations than in reality. The results presented in this section deal with the problem of censored observations by adopting a duration analysis approach. Specifically, if we have n countries with t1 . . . tn crises duration, we concentrate on finding the estimate of the probability density function f ðtÞ with associated survival time SðtÞ that maximizes the likelihood function L¼
Y
f ðti Þ di Sðti Þ 1di
ð15:5Þ
i
where di is an indicator variable that takes the value 0 if the peak per worker GDP has not been reached by the last observation in the sample. Broadly speaking, there are two approaches in the literature to estimating equation 15.5. One is to specify a parametric functional form for f ðtÞ and to estimate the parameters of that form. The second one is to use a nonparametric approach to estimation of f ðtÞ. The latter is commonly associated with estimation of the Cox proportional hazards model. Although the nonparametric approach is more flexible, it can lead to more imprecise estimates of the hazard function than a correctly specified parametric form. We will present versions of both models in this section.
0.138 (1.28)
0.312 (2.23)** 0.137 (1.27) 0.083 (1.49)
Change in polity indicator
Open forest
Years of primary schooling
Life expectancy at birth, total (years)
Urban population (% of total)
Telephone mainlines (per 1,000 people)
Rule of law
Total years of schooling
Years of secondary schooling
0.314 (2.24)**
1.152 (3.59)***
Log of inflation
0.059 (0.73)
1.062 (3.3)***
0.254 (2.37)**
0.244 (2.27)**
Sudden stop
0.765 (2.77)***
C0.533 (3.16)***
0.131 (0.98)
(2)
0.767 (2.78)***
C0.539 (3.2)***
0.199 (1.41)
(1)
War
Log change in real merchandise exports
Log GDP per working-age person
Dependent variable: Probability of falling into a crisis
Table 15.11 Random Effects Probit Regressions: Alternative Controls
0.052 (1.4)
0.134 (1.24)
0.312 (2.23)**
1.096 (3.44)***
0.246 (2.29)**
0.768 (2.78)***
C0.548 (3.24)***
0.206 (1.41)
(3)
0.013 (0.16)
0.066 (0.47)
0.531 (3.13)***
0.720 (1.95)*
0.248 (1.81)*
0.658 (2.03)**
C0.459 (2.15)**
0.182 (1.23)
(4)
0.001 (1.27)
0.123 (1.41)
0.334 (2.59)***
0.930 (2.99)***
0.202 (1.95)*
0.475 (1.82)*
C0.447 (2.96)***
0.104 (0.85)
(5)
0.004 (0.77)
C0.181 (2.04)**
0.367 (2.94)***
0.971 (3.12)***
0.230 (2.27)**
0.476 (1.84)*
0.024 (1)
C0.374 (2.25)**
0.038 (0.16)
1.128 (2.28)**
0.443 (2.93)***
0.910 (2.15)**
C0.487 (2.04)**
0.077 (0.34)
0.061 (0.44) C0.407 (2.88)***
(7)
(6)
0.016 (1.63) 0.044 (0.94)
0.002 (1.2)
0.288 (1.72)*
0.012 (0.03)
0.023 (0.04) 0.110 (0.21)
0.283 (1)
0.083 (0.22)
0.975 (1.66)*
0.654 (3.28)***
0.968 (2.09)**
C0.872 (2.35)**
0.174 (0.36)
(8)
406 Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
0.027 (0.12) 0.064 (0.14)
0.067 (0.28) 0.080 (0.18) 0.054 (0.22)
0.024 (0.1) 0.203 (0.43) 0.023 (0.09)
East Asia and Pacific
Central and Eastern Europe
Middle East and North Africa
951 74 6.54% 8.45%
Observations
Countries Percent crises predicted 74 7.19% 8.25%
951
74 5.88% 8.42%
951
0.880 (0.53)
0.156 (1)
0.019 (0.13)
61 6.54% 10.64%
672
2.351 (1.04)
0.148 (1.07)
0.531 (1.83)*
0.054 (0.12)
0.013 (0.05)
0.055 (0.14)
0.738 (2.47)** 0.497 (1.05)
80 5.81% 7.73%
1037
0.193 (0.12)
0.377 (1.1)
0.021 (0.13)
0.054 (0.38)
0.043 (0.18)
0.347 (0.81)
0.237 (1)
0.428 (1.47)
0.129 (0.56) 0.043 (0.14)
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: * 10%, ** 5%, *** 1%.
Pseudo-R 2
0.805 (0.49)
Constant
0.324 (0.2)
0.173 (1.11)
0.173 (1.12)
1990s
2000s
0.026 (0.17)
0.022 (0.14)
1980s
1970s
0.049 (0.2)
0.202 (0.67)
0.214 (0.71)
0.174 (0.58)
South and Central Asia
Africa
0.264 (1.24) 0.064 (0.2)
0.259 (1.18) 0.047 (0.15)
0.312 (1.48) 0.027 (0.08)
Latin America
83 6.11% 8.01%
1062
1.258 (0.69)
0.388 (1.21)
0.524 (1.62)
0.063 (0.27) 0.497 (1.48)
0.173 (0.43)
0.208 (0.9)
0.375 (1.35)
0.178 (0.83) 0.006 (0.02)
77 7.41% 10.83%
522
4.587 (1.73)*
0.476 (0.88)
0.643 (1.17)
0.141 (0.39) 0.548 (0.94)
0.022 (0.05)
0.278 (0.77)
0.741 (1.55)
0.040 (0.12) C0.940 (1.73)*
50 6.00% 13.12%
369
2.652 (0.56)
0.043 (0.18)
0.162 (0.26)
0.288 (0.38)
0.218 (0.44)
0.180 (0.27)
0.334 (0.54) 1.114 (0.97)
Growth Collapses 407
408
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
Another key issue has to do with how to handle country level heterogeneity in this framework. For analogous reasons to those of panel data estimation with binary dependent variables, fixed effects estimators are not consistent for duration models (Andersen, Klein, and Zhang 1999). Two alternative approaches exist. One is to assume that countries have differing propensities to experiencing crises. These propensities—called frailties—are analogous to the random effects of panel data models. An alternative approach is to use the fact that in the presence of repeated events, the Cox proportional hazards model parameter estimates converge to a value that can be interpreted meaningfully, but the estimated covariance matrix is inappropriate for hypothesis testing (Lin and Wei 1989 and Struthers and Kalbfleisch 1984). Variance-corrected models modify the covariance matrix of the Cox model in order to be able to carry out appropriate tests. Before proceeding to the statistical tests, we start out by looking at the characteristics of crises according to the events associated with them. The summary statistics associated with these different types of crises, as well as their associated unconditional hazard functions, are shown respectively in table 15.12 and figure 15.5. If different types of crises correspond to different types of shocks, then we would expect that the patterns of recoveries associated with different crises would also differ. In particular, this would be true if we believe that some crisis triggers have nonpermanent effects on the determinants of steady-state income. As we have discussed previously, in this case hazard functions should be clearly increasing. This is precisely the feature that Calvo and his coauthors have argued characterizes some sudden stops of capital flows. Similarly, to the extent that the level of democracy appears to be irrelevant for crisis onset, political regime transitions should have a transitory effect on the level of income: they should create havoc Table 15.12 Characteristics of Crises by Coinciding Initial Events Number of observations
Duration
Peak to trough ratio
Lost years of GDP
OLS trend growth (end to endþ5)
Substantial change in merchandise exports
95
6.51
11.6%
0.98
1.1%
Wars
32
6.53
16.5%
1.50
2.1%
Natural disasters
51
7.51
13.0%
1.30
1.0%
High inflation
62
6.60
10.2%
0.97
1.7%
Political transition
118
6.69
11.0%
1.04
1.7%
Sudden stop High open forest (>14)
109 75
5.83 3.08
8.8% 4.6%
0.86 0.23
1.6% 2.3%
Growth Collapses
409
Figure 15.5 Hazard Function by Type of Crisis: Alternative Specifications
during the time of the transition, but after they occur one would not expect there to be a permanent effect. What is interesting about figure 15.5 is that it shows that declining hazard rates appear to characterize many different types of crises. Indeed, all splits appear to be characterized by the same overall pattern: a short initial period of increasing hazard rates, and a much longer period of strongly declining rates. The relative magnitudes are also similar across characteristics. The one striking difference is open_ forest. Countries with very high open forests have crises of much lower duration and consequently display a higher probability of exiting the crisis at any one moment. Recall that of the variables that we have used to carry out the splits, open_ forest (along with natural disasters) was not robustly associated with the onset of crisis. This figure suggests that it may be directly associated with crisis duration. This hypothesis can be tested more systematically by looking at the effect of these factors in duration regressions. We do this in the rest of the section. We start out by looking at the most common parametric duration model, which is the Weibull distribution. Essentially, we estimate the hazard rate for country i as a function of a baseline hazard and the covariates
410
hi ðt j XÞ ¼ h0t vi expðbXÞ;
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
ð15:6Þ
where h0t ¼ pt p1 . The parameter p characterizes the shape of the Weibull distribution, with p < 1 corresponding to a decreasing hazard function. The ni are the country-level frailty terms that are assumed to follow a Gamma distribution. Estimates are displayed in table 15.13. Column 1 shows the effect of simply running a regression of duration on log of per worker GDP. The hazard specification is presented, so that a positive coefficient implies that increasing the variable in question leads to a higher probability of leaving the crisis. Not surprisingly, per worker income appears to be positively related to the probability of exiting a crisis. However, as column 2 shows, this result is not robust to controlling for region and time dummies. In column 3 we introduce open_ forest into the regression. We find that it is very strongly correlated with crisis duration. A one-standard-deviation increase in open_ forest implies an increase of 84.5 percent in the probability of leaving the crisis (expð.543Þ sdðopen_ forestÞ). Column 4 introduces controls for the level of democracy and sudden stops. The democracy control can be taken as a naive test of Rodrik’s (1999) hypothesis that countries with better institutions for conflict management have an easier time adjusting to negative shocks. Framed this way, the hypothesis gets weak support in our data: the effect of democracy is positively related with the probability of leaving the crisis, though the coefficient is borderline significant and not very robust. However, Rodrik’s hypothesis is somewhat more nuanced— see the additional results in table 15.13. Calvo and his coauthors have suggested that the recoveries associated with some types of sudden stops may be associated with faster recoveries. The estimates in column 4 show that this is not the case for collapses in capital flows, generally speaking. The next columns of table 15.13 introduce other potential determinants of crisis duration. The first logical candidates for this are the different determinants that we used in the probit analysis of the previous sections. If how you fall into the crisis matters for postcrisis behavior, then we should expect crisis durations to differ significantly for crises that were initiated by different events. We have already seen in figure 15.3 that there is little indication of this being the case in unconditional hazard rates by group, but we now verify this within the framework of the parametric Weibull specification. In effect, the estimates in column 4 show that different factors that were significantly associated with the onset of crises are not associated with crisis duration. This is the case of wars, inflation, and political transitions—as well as of natural disasters, which did not prove significant in the probit analyses. One may expect that both open forests and democracy may make the society more capable of responding to particular types of shocks rather than uniformly
Growth Collapses
411
Table 15.13 Duration Regressions, Weibull Specification with Frailty Dependent variable: Years in crisis Log GDP per working-age person
(1)
(2)
(3)
0.254 0.238 0.263 (2.61)*** (1.45) (1.31)
Open forest
0.543 (3.61)***
(4)
(5)
(6)
(7)
C0.492 (1.93)*
C0.744 (2.61)***
0.258 (1.29)
C0.497 (1.98)**
0.578 (3.03)***
0.731 (3.37)***
Democracy (polity index)
0.039 (1.66)*
0.035 (1.38)
Sudden stop
0.004 (0.02)
0.139 (0.62)
Log change in real merchandise exports
0.429 (1.34)
War
0.621 (1.15)
Natural disaster
0.099 (0.25)
Log of inflation
0.189 (0.29)
Change in polity indicator
C0.290 (1.01)
Change in exports*open forest
0.457 (2.54)**
0.530 (2.46)** 0.043 (1.64)
C1.288 (0.5)
C1.964 (0.55)
0.112
0.157
(0.54)
(0.56) 0.026 (0.53)
Polity*change in merchandise exports Polity*sudden stops
0.000 (0.01)
Constant
C4.102 (4.51)***
N
330
330
Time Dummies
No
Yes
Continent Dummies
No
Yes
2.061 C5.764 (1.22) (1.91)*
C4.245 (1.18)
C3.171 (0.78)
C4.656 (1.44)
C2.653 (0.7)
229
188
175
227
188
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: * 10%, ** 5%, *** 1%. Representation, hazard rate with region and decade dummies (not shown).
412
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
increasing the probability of exiting all types of crises. Open_ forest, for example, should have a positive effect on an economy’s capacity to adapt to export collapses, while democracy may make societies more capable of adapting to external shocks such as sudden stops. The last two columns of table 15.13 evaluate these hypotheses, and find little support. Neither an interaction between open forests and changes in exports nor the interaction term between democracy and the two external shock indicators (changes in exports and sudden stops) are significant. This may of course reflect the relative coarseness of these multiplicative terms to capture complex nonlinearities. Table 15.14 tests a battery of additional possible correlates of crisis duration. These include an alternative indicator of nonsystemic sudden stops in column 1 (the same one in column 4 of table 15.8, which captures large decreases in total capital flows that coincide with substantial import declines), an indicator of trade policy (column 2), a measure of human capital (total years of schooling, column 3), a measure of financial deepening (liquid liabilities in GDP, column 4), a general measure of openness (trade/GDP ratio, column 5), and a measure of social modernization (life expectancy, column 6). We also try two measures of the idea that institutions may have an effect on crisis duration. The first one introduces an interaction between our indicator of democracy and the terms of trade shock, thus capturing the idea that democracies are better able to adapt to adverse terms of trade shocks. This effect is insignificant. Another specification uses the interaction between the Gini index and 1, minus a scaled democracy variable. This is closest to Rodrik’s (1999) precise specification. It gets strong support in the data, with a significant negative coefficient, suggesting that when there are high levels of social conflict, better institutions for conflict management are associated with shorter crises. In one last specification we include the regressors of columns 1–7 together (column 9). Most of these terms are significant, while the open_ forest indicator remains strongly significant. In the next table we adopt as our baseline specification a regression including time and region dummies, open_ forest, and the log of initial GDP, and we evaluate whether the strength of the coefficient on open forests is at all dependent on the parametric specification of the hazard function. We use four alternative parameterizations: the exponential, the Gompertz, the log-logistic, and the log-normal. In reading table 15.15, it is important to bear in mind that the last two specifications do not accept a hazard rate interpretation and are thus reported in acceleratedfailure time modes, so that the dependent variable is the duration of the crisis. Thus, a positive effect in the hazard representation is analogous to a negative effect in the accelerated-failure time representation. That is, in fact, what we find: open_ forest has a positive, significant effect in the hazard rate representations and a negative, significant effect in the failure time representations. The result that
Growth Collapses
413
open_ forest decreases the time necessary to escape a crisis is robust to the parameterization adopted. Figure 15.6 displays the conditional hazard rates that emerge from the five parametric specifications that we have estimated (with the controls of table 15.13). These are the estimated hazard rates for an observation with the expected random effect ni ¼ 1. In contrast to the unconditional hazard rates of figures 15.2– 15.5, they are not affected by the changing composition of the population and reflect the estimated probability that a particular country will exit the crisis. For comparison purposes, we also report the unconditional hazard rate of a Cox model without shared frailties. Except for the exponential form, which is constrained to be constant, all our estimates of the hazard rates give declining or roughly flat functions in time. In the next two tables we turn toward estimation in the framework of a variance-corrected Cox proportional hazards model. In particular, we specialize to the conditional risk-set model of Prentice, Williams, and Peterson (1981, henceforth PWP). The basic idea of the PWP model is to stratify by event number, so that the conditional risk set for experiencing crisis k is the number of countries that have experienced k 1 crises in the past. The model is stratified by number of crises in order to obtain the corrected variance estimates. Tables 15.15 and 15.16 repeat the estimation exercises of tables 15.12 and 15.13 using the PWP specification. The results are broadly similar. Open_ forest is always significant, with the only exception being the last column of table 15.17, which has a very small number of observations (here it has a borderline p value of .142 with 60 observations). None of the other potential covariates emerge as significant. As we have argued above, declining hazard rates may indicate the presence of multiple equilibria—with negative shocks leading countries to shift to inferior equilibria—or adverse permanent productivity shocks. A valid question to ask at this stage is whether there is evidence that this phenomenon is due to changes in fundamentals. One way to tackle this question is by asking whether the triggers that appear to have sent the economy into the crisis have returned to precrisis levels at the time periods during which we are observing declining hazard rates. Figure 15.7 presents one such exercise. In it we calculate the average paths for countries with crisis duration greater than or equal to ten years for five of the variables we have found to be significantly associated with the onset of crises: wars, exports, capital flows, inflation, and political transitions. The evidence is mixed, and thus interesting. By the tenth year of the crisis, the fraction of countries in the midst of a war has returned to precrisis levels. The number of countries undergoing political transitions has also declined, though not to precrisis levels. Capital flows have actually gone up in comparison to their precrisis levels. This is interesting not only because it suggests that the decline in capital flows is not the cause
C1.661 (3.68)*** 1.129 (1.53) 0.747 (1.08) 0.949 (1.59)
0.535 (3.06)*** 0.051 (0.24) C1.714 (3.68)*** 1.134 (1.51) 0.430 (0.58) 0.443 (0.76) C1.809 (2.15)** 0.428 (0.85)
Open forest
Sudden stop 4
Latin America
Africa
South and Central Asia
East Asia and Pacific
Central and Eastern Europe
C1.960 (2.81)*** C1.587 (2.23)**
1980s
1990s
Tariff rate
2000s
C1.460 (1.97)**
1970s
Middle East and North Africa
0.291 (1.18)
0.172 (0.74)
Log GDP per working-age person
1.199 (1.07) 2.426 (0.31)
C0.399 (1.66)*
0.465 (1.69)*
C2.177 (2.36)** 0.145 (0.29)
0.590 (3.13)***
(2)
(1)
Dependent variable: Years in crisis Representation—Hazard
Table 15.14 Weibull Specification, Alternative Controls
C0.478 (1.9)*
C1.512 (1.84)*
C1.865 (2.33)**
1.117
1.093 (1.33) 0.307 (0.66)
1.105 (1.42) 0.218 (0.43) 0.476 (1.6)
0.467 (0.93)
0.364 (0.59)
C1.141 (1.87)*
C1.458 (3.62)***
0.576 (3.47)***
0.168 (0.85)
(4)
0.324 (0.68)
0.348 (0.52)
0.763 (1.19)
C1.461 (3.45)***
0.582 (3)***
0.351 (1.28)
(3)
C1.204 (2.02)**
C1.593 (2.74)***
0.762 (1.22)
C1.878 (2.6)*** 0.528 (1.06)
0.746 (1.35)
0.909 (1.4)
C1.181 (1.8)*
C1.670 (3.93)***
0.638 (4.07)***
0.335 (1.64)
(5)
C0.808 (2.95)*** 0.373 (1.34)
0.132 (0.09)
1.009 (1.33) 0.386 (0.82)
0.688 (1.32)
0.828 (1.32)
C1.370 (2.16)**
C1.635 (3.77)***
0.578 (3.39)***
C0.366 (1.75)*
(7)
0.438 (0.29)
1.083 (0.69)
C2.481 (2.75)*** 1.350 (1.63)
0.737 (0.92)
C2.101 (1.83)*
0.196 (0.13)
C2.083 (2.46)**
0.475 (1.7)*
C1.518 (2.49)**
(6)
0.421 0.51
0.472 (1.67)*
0.619 (2.34)**
1.442 (1.99)** 0.323 0.57
0.431 0.7
0.879 1.32
0.546 0.76
1.307 (2.80)***
0.728 (3.48)***
0.568 (2.25)**
(8)
C95.204 (2)**
0.664 (0.88)
C1.402 (2.42)**
C3.947 (2.19)** 0.107 (0.09)
C3.911 (2.24)**
C4.622 (2.55)**
1.389 (0.7)
C2.609 (2.22)**
2.058 (2.64)***
C5.013 (3.43)***
(9)
414 Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
189
N
177
C6.156 (1.85)* 179
C6.223 (1.83)*
0.100 (1.26)
191
C5.707 (1.85)*
0.002 (0.35)
223
C5.332 (1.74)*
0.003 (1.08)
97
1.657 (0.31)
0.126 (1.66)*
183
C4.799 (1.59)
0.169 (0.52)
178
0.025 (2.54)** 4.963 1.53
60
9.409 (0.99)
1.653 (1.38)
0.090 (0.69)
0.025 (1.46)
0.418 (2.03)** 0.012 (0.75)
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: * 10%, ** 5%, *** 1%. Regression (9) comes from log file of table 15.17.
4.948 (1.42)
Constant
Gini*(1-Democracy)
Interaction between polity and terms of
Life expectancy at birth, total (years)-
Openness
Liquid liabilities/GDP
Total years of schooling
Growth Collapses 415
416
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
Table 15.15 Duration Regressions, Alternative Specifications Dependent variable: Years in crisis Distribution representation Log GDP per working-age person Open forest
(1) Exponential hazard
(2) Gompertz hazard
0.236 (1.27)
0.105 (0.77)
(3) Cox AFT
(4) Log-normal AFT
0.300 (1.62)
0.323 (1.58)
C0.368 (2.97)***
C0.346 (2.55)**
0.517 (3.68)***
0.396 (3.66)***
Latin America
C1.504 (3.83)***
C1.051 (3.48)***
0.777 (2.2)**
0.769 (2.07)**
Africa
C1.143 (1.86)* 0.817 (1.32)
0.668 (1.55) 0.509 (1.24)
0.720 (1.24) 0.518 (0.88)
0.774 (1.25) 0.506 (0.82)
East Asia and Pacific
0.203 (0.45)
0.174 (0.53)
0.133 (0.35)
0.122 (0.33)
Central and Eastern Europe
C1.711 (2.49)**
C1.297 (2.32)**
1.108 (1.96)**
Middle East and North Africa
0.437 (0.96)
0.173 (0.53)
0.064 (0.18)
0.047 (0.13)
1970s
0.462 (1.83)* 0.340 (1.6)
0.239 (1.07) 0.222 (1.16)
C0.538 (2.5)** 0.256 (1.39)
C0.565 (2.57)** 0.272 (1.41)
South and Central Asia
1980s 2000s Constant N
1.240 (1.98)**
1.218
1.229
0.604
0.610
(2.06)**
(2.21)**
(1.18)
(1.3)
C5.623 (1.98)** 229
C5.194 (2.55)** 229
2.922 (1.16) 229
2.290 (0.87) 229
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: * 10%, ** 5%, *** 1%.
for declining hazard rates, but also because it is suggestive that the marginal product of capital in equilibrium has not gone down.11 On the other hand, we find that the average inflation levels have gone up while the share of exports in GDP has gone down during the crisis. Both of these are consistent with the hypothesis that the economy enters some type of economic tailspin both in its export capacity and in the quality of its macroeconomic policy during prolonged crises. There are several ways in which we can interpret the strength of the open_ forest variable. At a general level, open_ forest is an indicator of an economy’s flexibility. It measures the possibilities that an economy has of moving to the production of other goods, weighted by the sophistication of these goods. It thus combines a hypothesis that flexibility is important with the hypothesis that countries develop
Growth Collapses
417
Figure 15.6 Conditional Hazard Function: Alternative Specifications
producing rich-country goods, as suggested by Hausmann, Hwang, and Rodrik (2007). In table 15.18 we use a measure of open forests that does not weigh goods by their PRODY, thus implicitly assigning goods equal value. The absolute value of the coefficient and its significance are very similar to those obtained by open_ forest. When both variables are introduced in the regression (not shown) neither of them is significant at 5 percent, suggesting that they are too collinear to distinguish between the alternative hypotheses they represent. To the extent that pure density is a simpler explanation, Occam’s razor would suggest sticking with it. Column 2 looks at other potential measures of flexibility. One is a Herfindahl index of export concentration. The idea is that countries with more concentrated export sectors will have a harder time reacting to adverse shocks as it will be more difficult for other industries to expand. It is possible that open_ forest is simply capturing the effects of having a diversified export structure. The result shown in columns 2–4 are surprising: export concentration seems to be associated with a higher, not lower, probability of exiting the crisis. A similar fact appears to be true about the log of pPopulation, another measure of the size of the economy and of its possible flexibility (column 3). In any case, the coefficient on open forests is robust to the inclusion of these alternative indicators
C0.585 (1.85)* C0.576 (2.2)** C1.531 (4.2)*** C0.731 (2.43)**
South and Central Asia
East Asia and Pacific
Central and Eastern Europe
Middle East and North Africa
0.107 (0.57) 0.178 (0.6)
1990s
2000s
Sudden stop
Democracy (polity)
Open forest
0.183 (1.03)
1980s
1970s
C1.298 (4.31)***
Africa
0.098 (0.83)
0.108 (1.09)
0.355 (3.17)***
1.274 (3.76)***
0.029 (1.71)* 0.044 (0.24)
0.350 (2.5)**
0.780 (2.41)**
0.091 (0.45)
0.107
0.278 (0.67)
0.249 (0.79) 0.280 (1.27) 0.030 (0.15)
C0.605 (2.28)**
0.410 (0.92)
0.127 (0.45) C1.002 (1.83)*
0.425 (1.27)
0.384 (1.14)
0.032 (1.63) 0.000 (0)
0.388 (2.75)***
0.283 (2.08)**
1.230 (3.57)***
0.240 (1.03) 0.057 (0.29)
0.529 (1.34) 0.597 (1.46) 0.564 (1.41)
0.254 (0.81)
0.745 (1.5)
0.489 (1.29) 0.198 (0.44)
0.161 (0.57)
0.390 (1.17)
C0.747 (1.75)*
C1.010 (3.53)***
0.093 (0.8)
(6)
0.373 (0.83)
0.535 (1.38)
0.407 (0.64)
0.408 (0.66)
C0.690 (1.66)*
C0.341 (2.06)**
(5)
C0.728 (2.17)**
0.247 (1.47)
(4)
C0.793 (2.29)**
C0.989 (3.51)***
(3)
(2)
C1.184 (4.97)***
0.186 (2.74)***
(1)
Latin America
Log GDP per working-age person
Dependent variable: Years in crisis Representation, hazard
Table 15.16 Duration Regressions, PWP Stratified Cox Model
0.032 (1.49) 0.010 (0.05)
0.307 (1.92)*
C0.779 (2.33)**
C0.675 (1.98)** C0.880 (2.88)***
0.236 (0.54)
C0.575 (1.73)*
0.395 (0.9)
0.471 (1.39)
0.477 (0.75)
C0.831 (2.34)**
C0.269 (1.66)*
(7)
418 Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: * 10%, ** 5%, *** 1%.
4.2%
4.4%
230 3.9%
4.2%
191
4.3%
0.1%
Pseudo-R2
2.2%
535
N
0.155 (0.69) 0.027 (0.79)
1.939 (0.67)
0.007 (0.28) 175
0.108 (0.64)
1.357 (0.63)
Polity*sudden stops
Polity*change in merchandise exports
Change in exports*open forest
191
0.138 (0.57)
Change in polity indicator
233
0.070 (0.11)
Log of inflation
535
0.040 (0.12)
Natural disaster
War
0.288 (0.8) 0.022 (0.05)
Log change in real merchandise exports
Growth Collapses 419
C1.142 (3.68)*** C0.821 (1.83)* 0.528 (1.51) C0.728 (1.93)*
0.381 (2.86)*** 0.050 (0.29) C1.114 (3.35)*** 0.666 (1.37) 0.303 (0.87) 0.307 (0.89) C1.195 (2.31)** 0.240 (0.64)
Open forest
Sudden stop 4
Latin America
Africa
South and Central Asia
East Asia and Pacific
Central and Eastern Europe
C1.374 (4.45)*** C1.231 (3.44)***
1980s
1990s
Tariff rate
2000s
C1.278 (3.28)***
1970s
Middle East and North Africa
0.202 (0.07)
0.093 (0.51)
Log GDP per working-age person
0.183 (0.74)
0.143 (0.6)
0.797 (1.96)* 3.602 (0.65)
0.255 (0.9)
C0.499 (1.66)* 0.041 (0.11)
0.139 (0.4)
0.244 (0.64)
0.485 (1.09)
C0.869 (2.83)***
0.389 (2.24)**
0.207 (0.14)
(3)
0.263 (1)
C1.245 (1.88)* 0.120 (0.37)
0.394 (3.03)***
(2)
(1)
Dependent variable: Years in crisis Representation, hazard
Table 15.17 PWP Stratified Cox Model, Alternative Controls
C1.143 (2.56)**
C1.303 (3.55)***
C1.006
C0.656 (2.33)** 0.184 (0.52)
0.411 (1.28)
0.221 (0.6)
C0.804 (1.88)*
C0.941 (3.22)***
0.345 (2.5)**
0.147 (0.40)
(4)
C1.227 (3.56)***
C1.295 (4.38)***
C0.948 (2.9)***
C1.047 (1.97)** 0.375 (1.13)
0.519 (1.57)
0.483 (1.46)
C0.795 (1.9)*
C1.133 (3.83)***
0.430 (3.71)***
0.196 (0.17)
(5)
C1.211 (1.93)*
C1.554 (2.31)**
0.545 (0.67)
C1.132 (2.03)** 0.848 (1)
0.409 (0.81)
C1.748 (2.47)**
1.186 (0.94)
C1.593 (2.94)***
0.736 (2.97)***
C0.573 (0.47)
(6)
0.523 (0.74)
0.198 (0.82)
50.989 (0.85)
0.568 (0.94)
1.392 (1.24) 0.270 (0.24)
2.351 (1.52)
C2.658 (1.92)*
1.803 (0.91)
C1.998 (2.49)**
1.272 (1.47)
C1.890 (0.12)
(8)
C0.400 (1.85)*
C0.483 (1.89)* 0.160 (0.4)
0.485 (1.22)
0.536 (1.6)
C0.868 (1.84)*
C1.017 (3.13)***
0.434 (2.87)***
C0.260 (0.41)
(7)
420 Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
224 5%
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: * 10%, ** 5%, *** 10.
5%
8%
100
5%
183
5%
5%
5%
190
C0.015 (0.28)
Pseudo-R2
191
0.003 (1.43)
N
179
0.003 (1.07)
0.057 (0.22) 177
0.054 (0.94)
Interaction between polity and terms of
Life expectancy at birth, total (years)-
Openness
Liquid liabilities/GDP
Total years of schooling
12%
60
0.125 (0.13)
0.003 (0.03)
0.008 (0.81)
0.049 (0.56) 0.005 (0.43)
Growth Collapses 421
422
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
Figure 15.7 Evolution of Covariates during Crises
of flexibility, as well as of land area (column 4). In column 5 we present an additional robustness test, which is to control for the magnitude of the initial crisis by introducing measures of the magnitude of the initial shock. In particular, we control for initial and lagged GDP growth, initial and lagged export growth, and the initial terms of trade shock. Since some of these measures will be correlated with the dependent variable by definition, we do not use this set of controls more broadly in this paper. Open_ forest is robust to the inclusion of these controls. 15.6
Concluding Comments
This paper has analyzed episodes during which economic growth decelerates to negative rates in a sample of 180 developing and developed economies. We identify 535 episodes of output contractions. The distribution of these episodes is highly skewed: while the median duration is 2 years, more than a quarter of them last more than 7 years and roughly 14 percent last at least 15 years. Developing countries are much more likely to experience prolonged contractions than are industrial countries.
Table 15.18 Alternative Hypotheses Dependent variables: Years in crisis
(1)
(2)
(3)
(4)
(5)
Log GDP per working-age person
0.249 (1.23)
0.23 (1.15)
0.365 (1.75)*
0.38 (1.78)*
0.606 (1.66)*
Open forest Open forest (Density only)
1.1 (2.89)*** 0.586 (3.68)***
Herfindahl
0.917 (3.54)*** 1.891 (1.65)*
Log of population
1.286 (4.27)*** 2.491 (2.15)** 0.245 (2.70)***
Area (sq. km)
1.334 (4.15)*** 2.772 (2.28)** 0.254 (2.42)** 0.018 (0.25)
Growth in terms of trade at t ¼ 1
7.445 (4.54)***
Growth in GDP at t ¼ 1
30.394 (4.69)***
Growth in merchandise exports at t ¼ 1
0.638 (0.64) 5.519 (0.78)
Growth in GDP at t ¼ 0 Growth in merchandise exports at t ¼ 0 Constant
2.642 (2.48)** 0.266 (0.11)
1.916 (0.72)
1.513 (0.51)
0.325 (0.11)
9.602 (1.80)*
Latin America
1.557 (3.61)***
1.603 (3.72)***
1.768 (4.04)***
1.798 (4.03)***
1.054 (1.59)
Africa
1.187 (1.76)* 0.825 (1.22)
1.244 (1.86)* 0.783 (1.17)
1.167 (1.73)* 0.453 (0.67)
1.201 (1.73)* 0.475 (0.68)
0.613 (0.58) 1.428 (1.54)
East Asia and Pacific
0.177 (0.37)
0.164 (0.35)
0.278 (0.58)
0.271 (0.56)
0.985 (1.27)
Central and Eastern Europe
1.757 (2.41)**
1.713 (2.36)**
1.686 (2.23)**
1.576 (1.97)**
Middle East and North Africa
0.428 (0.87)
0.63 (1.26)
0.695 (1.37)
0.738 (1.41)
0.16 (0.21)
1970s
0.879 (1.43) 1.701 (2.93)***
0.875 (1.43) 1.648 (2.83)***
0.918 (1.51) 1.71 (2.93)***
0.255 (0.98) 0.542 (2.43)**
0.615 (0.72) 1.096 (1.44)
1.31 (2.20)**
1.221 (2.05)**
1.177 (1.97)**
South and Central Asia
1980s 1990s 2000s
2.426 (2.33)**
0.716 (0.91) 1.161 (1.93)*
Observations
229
229
229
225
130
Number of groups
100
100
100
97
62
Note: Absolute value of z statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.
424
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
We have studied the factors that coincide with the onset of these crises. In terms of statistical significance, we find that the change in exports is the variable most strongly associated with the probability of suffering a crisis—at least in developing countries. A one-standard-deviation decrease in the growth rate of merchandise exports implies a 5.47 percentage point increase in the probability of a crisis. In terms of economic significance, a one-standard-deviation increase in inflation appears to be slightly more damaging, though the coefficient is somewhat less precisely estimated. Wars, sudden stops, and political transitions also tend to coincide with the onset of crises. The duration of crisis is somewhat more difficult to predict. Surprisingly, the variables that we find to be significantly associated with the probability of a crisis occurring do not appear to be related to crisis duration. One variable that we find to be robustly associated with crisis duration—aside from continent and decade effects—is a measure of the density-weighted value of a country’s alternative export basket suggested by Hausmann and Klinger (2006). We take this measure to be an indicator of the flexibility of an economy’s productive apparatus in adapting to external shocks. This intuition is confirmed by case studies of collapse episodes in developing countries that emphasize the role that poor performance in the nontraditional export sector plays in deepening growth collapses.12 We also find evidence that an interaction between democracy and inequality, as suggested by Rodrik (1999), affects the duration of crises. Our results leave open several avenues for future analysis. On the one hand, it would be desirable to refine the duration model’s predictive capacity. One possible avenue for doing this would be to adopt a model of time-varying covariates. We have shied away from that alternative because we find it easier to believe in the exogeneity of changes that took place before the onset of the crisis than we would in changes that occur during the crisis. Another avenue is to explore the possible channels through which open_ forest is correlated with crisis duration. Our first tentative attempts to get at this issue failed to find a significant interaction between export collapses and open_ forest. Several explanations could account for this fact. Open_ forest may be a more general measure of the economy’s adaptability to several types of productivity shocks, a multiplicative interaction may be too coarse to capture generalized nonlinearities, or open_ forest may be proxying for some unmeasured country-specific effect. Further research could help discern among these potential competing hypotheses. We also find that decreasing conditional and unconditional hazard rates are a pervasive characteristic of our estimation. While this is not necessarily surprising for unconditional rates, as it may be a consequence of the changing composition of the population, it is definitely counterintuitive when these hazard rates are conditioned on estimated country random effects and covariates. Even though
Growth Collapses
425
decreasing hazard rates can be accounted for within a neoclassical model as a result of substantial, permanent shocks to output, the depth and duration of some recessions in this sample appear hard to explain. To take just one example, there is widespread agreement among many observers of the Bolivian economy that its institutional, political, and macroeconomic framework was more solid at the beginning of this century than in the mid-seventies.13 However, GDP per workingage person was 14.9 percent lower in 2004 than it was in 1978, despite the fact that world productivity surely increased during this period. Further investigation of the characteristics of recessions may allow us to find ways to discriminate between alternative interpretations of this type of phenomenon. Notes We would like to thank Masami Imai, Cameron Shelton, Sanjay Reddy, Eduardo Zambrano, and seminar participants at Harvard University for their valuable comments and suggestions. Alejandro Izquierdo and Bailey Klinger kindly provided data used in this analysis. The usual disclaimer applies. 1. The formal definition of these concepts will be introduced in section 3. 2. Calvo, Izquierdo, and Mejı´a (2004), Calvo, Izquierdo, and Loo-Kung (2005), Calvo (2005), Calvo, Izquierdo, and Talvi (2006). 3. An extensive literature has developed attempting to explain growth volatility. See Ramey and Ramey (1995), Imbs (2002), and Aghion and Banerjee (2005) for discussions. 4. The definition of recoveries used by Cerra and Saxena (2005) differed from ours in that they define a trough as any year in which growth changes from negative to positive. In practice, this implies that they underestimate the length of ‘‘double dip’’ and ‘‘n-tuple dip’’ crises. 5. Reddy and Minoiu (2006) also use the first ‘‘turning point’’ after a crisis to date the end of the crisis, which is subject to the same objection made regarding the Cerra and Saxena technique. 6. Hausmann, Pritchett, and Rodrik (2005), in contrast, have looked at episodes during which growth accelerates. Their main finding is that accelerations are not well explained by macroeconomic policy reforms. 7. Two useful recent surveys are Hosmer and Lemeshow (1999) and Box-Steffensmeier and Jones (2004). 8. Analytically, it is important to distinguish between the causes that lead countries to fall into crises and the reasons that their recovery speeds differ. Although there could be similarities between both processes—and crises generated by large shocks may be more difficult to get out of—they may well be very different. For example, in a related analysis, Collier and Hoeffler (2004) have found that the causes that lead countries to fall into civil wars are very different from those that determine the duration of those wars. 9. Another difference with our definition is that we use annual data, while Calvo and his coauthors, who study the short-run dynamics of sudden stops, use monthly data. 10. This result is not an artifice of sample reduction either: running the regression of column 1 for the sample in column 4 gives significantly positive continent dummies for Latin America, Africa, and the MENA region.
426
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
11. Note that with this specification, we cannot tell if that is because of lower levels of the capital stock or because an improvement in productivity. 12. See, for example, Hausmann and Rodrı´guez (2006) on Venezuela and Auty (2001) on Saudi Arabia. 13. See Morales and Sachs (1992) and Jimenez Zamora, Candia, and Mercado Lora (2005).
References Aghion, P., and A. Banerjee. 2005. Volatility and Growth. Oxford: Oxford University Press. Ali, Ali A., and Ibrahim A. Elbadewi. 1999. Inequality and the Dynamics of Poverty and Growth. Center for International Development at Harvard University, Working Paper 32. Andersen, P. K., J. P. Klein, and M. J. Zhang. 1999. ‘‘Testing for Centre Effects in Multi-centre Survival Studies: a Monte Carlo Comparison of Fixed and Random Effects Tests. Statistics in Medicine 18, no. 12: 1489–1500. Auty, R. M. 2001. The Political Economy of Resource-driven Growth. European Economic Review 45: 839–846. Barro, R. 1989. ‘‘A Cross-country Study of Growth, Saving, and Government.’’ Working Paper No. 2855, NBER, Cambridge, MA. Barro, Robert J., and Jong-Wha Lee. 2000. ‘‘International Data on Educational Attainment: Updates and Implications.’’ Working Paper No. 42, Center for International Development, Harvard University, Cambridge, MA. Ben-David, D., and D. H. Papell. 1998. ‘‘Slowdowns and Meltdowns: Postwar Growth Evidence from 74 Countries.’’ Review of Economics and Statistics 80, no. 4: 561–571. Blyde, Juan, Christian Daude, and Eduardo Fernandez Arias. 2006. ‘‘Growth Collapses and Productivity Destruction.’’ Mimeo., Inter-American Development Bank, Washington, D.C. Bodman, Philip M. 1998. ‘‘Asymmetry and Duration Dependence in Australian GDP and Unemployment.’’ Economic Record 74, no. 227: 399–411. Box-Steffensmeier, J. M., and B. S. Jones. 2004. Event History Modeling: A Guide for Social Scientists. New York: Cambridge University Press. Calvo, G. A. 2005. ‘‘Crises in Emerging Market Economies: A Global Perspective.’’ Working Paper No. 10520, NBER, Cambridge, MA. Calvo, G. A., A. Izquierdo, and R. Loo-Kung. 2006. ‘‘Relative Price Volatility under Sudden Stops: The Relevance of Balance-Sheet Effects.’’ Journal of International Economics 69, no. 1: 231–254. Calvo, G. A., A. Izquierdo, and L.-F. Mejia. 2004. ‘‘On the Empirics of Sudden Stops: The Relevance of Balance-Sheet Effects.’’ Working Paper No. 10520, NBER, Cambridge, MA. Calvo, Guillermo A., Alejandro Izquierdo, and Ernesto Talvi. 2006. ‘‘Phoenix Miracles in Emerging Markets: Recovering without Credit from Systemic Financial Crises.’’ Working Paper No. W12101, NBER, Cambridge, MA. Cerra, Valerie, and Sweta Chaman Saxena. 2005. ‘‘Growth Dynamics: The Myth of Economic Recovery.’’ Working Paper No. WP/05/147, IMF, Washington, D.C. Ce´spedes, Luis Felipe, and Jose de Gregorio. 2005. ‘‘Recuperaciones Cı´clicas: Evidencia Internacional.’’ Mimeo., Banco Central de Chile. Available at http://www.bcentral.cl/esp/estpub/otrasconf/otras/ pdf/Cespedes_De_Gregorio.pdf.
Growth Collapses
427
Collier, P., and P. Hoeffler. 2004. ‘‘Greed and Grievance in Civil War.’’ Oxford Economic Papers 56, no. 4: 563–595. Diebold, Francis X., and Glenn D. Rudebusch. 1990. ‘‘A Nonparametric Investigation of Duration Dependence in the American Business Cycle.’’ Journal of Political Economy 98, no. 3: 596–616. Di Venuto, Nicholas, and Allan Layton. 2005. ‘‘Do the Phases of the Business Cycle Die of Old Age?’’ Australian Economic Papers 44, no. 3: 290–305. Galton, Francis. 1889. Natural Inheritance. New York: MacMillan. Gleditsch, Kristian Skrede. 2004. ‘‘A Revised List of Wars Between and Within Independent States, 1816–2002.’’ International Interactions 30: 231–262. Hausmann, R., J. Hwang, and D. Rodrik. 2007. ‘‘What You Export Matters.’’ Journal of Economic Growth 12, no. 1: 1–25. Hausmann, R., and B. Klinger. 2006. ‘‘Structural Transformation and Patterns of Comparative Advantage in the Product Space.’’ Working Paper No. 128, Center for International Development, Harvard University, Cambridge, MA. Hausmann, R., L. Pritchett, and D. Rodrik. 2005. ‘‘Growth Accelerations.’’ Journal of Economic Growth 10, no. 4: 303–329. Hausmann, R., and F. Rodrı´guez. 2006. ‘‘Why Did Venezuelan Growth Collapse?’’ Mimeo., Center for International Development, Harvard University, Cambridge, MA. Hosmer, D. W., and S. Lemeshow. 1999. Applied Survival Analysis: Regression Modeling of Time to Event. New York: Wiley. International Country Risk Guide. 1999. PRS Group. CD-ROM. Imbs, J. 2002. ‘‘Why the Link between Volatility and Growth is Both Positive and Negative.’’ Discussion Paper No. 3561, CEPR, London. Jimenez Zamora, Elizabeth, Gaby Candia, and Marcelo Mercado Lora. 2005. ‘‘Economic Growth, Poverty and Institutions: A Case Study of Bolivia.’’ Mimeo., Global Development Network, World Bank, Washington, D.C. Lin, D. Y., and L. J. Wei. 1989. ‘‘The Robust Inference for the Cox Proportional Hazards Model.’’ Journal of the American Statistical Association 84: 1074–1078. Marshall, M. G., K. Jaggers, and T. R. Gurr. 2004. ‘‘Polity IV Dataset.’’ Electronic File. College Park: University of Maryland. McFadden, D. 1974. ‘‘The Measurement of Urban Travel Demand.’’ Journal of Public Economics 3: 303– 328. Mills, Terence C. 2001. ‘‘Business Cycle Asymmetry and Duration Dependence: An International Perspective.’’ Journal of Applied Statistics 28, no. 6: 713–724. Mora, Ricardo, and Georges Siotis. 2005. ‘‘External Factors in Emerging Market Recoveries: An Empirical Investigation.’’ European Economic Review 49, no. 3: 683–702. Morales, Juan Antonio, and Jeffrey Sachs. 1990. ‘‘Bolivian Debt Management.’’ In Developing Country Debt and Economic Performance, ed. Jeffrey Scahs, 252–255. Chicago: University of Chicago Press. Mudambi, Ram, and Larry W. Taylor. 1991. ‘‘A Nonparametric Investigation of Duration Dependence in the American Business Cycle: A Note.’’ Journal of Political Economy 99, no. 3: 654–656.
428
Ricardo Hausmann, Francisco Rodrı´guez, and Rodrigo Wagner
Prentice, R. L., B. J. Williams, and A. V. Peterson. 1981. ‘‘On the Regression Analysis of Multivariate Failure Time Data.’’ Biometrika 68: 373–379. Pritchett, Lant. 2000. ‘‘Understanding Patterns of Economic Growth: Searching for Hills among Plateaus, Mountains, and Plains.’’ World Bank Economic Review 14, no. 2: 221–250. Ramey, G., and V. A. Ramey. 1995. ‘‘Cross-Country Evidence on the Link between Volatility and Growth.’’ American Economic Review 85, no. 5: 1138–1151. Reddy, Sanjay G., and Camelia Minoiu. 2006. ‘‘Real Income Stagnation of Countries, 1960–2001.’’ Mimeo., Columbia University, New York. Rodrik, D. 1999. ‘‘Where Did All the Growth Go? External Shocks, Social Conflict, and Growth Collapses.’’ Journal of Economic Growth 4, no. 4: 385–412. Ros, Jaime. 2005. ‘‘Divergence and Growth Collapses: Theory and Empirical Evidence.’’ In Beyond Reforms: Structural Dynamics and Macroeconomic Vulnerability, ed. J. A. Ocampo, 211–232. Palo Alto: Stanford University Press. Sichel, Daniel E. 1991. ‘‘Business Cycle Duration Dependence: A Parametric Approach.’’ Review of Economics and Statistics 73, no. 2: 254–260. Struthers, C. A., and J. D. Kalbfleisch. 1986. ‘‘Misspecified Proportional Hazard Models.’’ Biometrika 73, no. 2: 363–369. Wooldridge, J. G. 2001. Econometric Analysis of Cross-Section and Panel Data. MIT Press, Cambridge, MA. World Bank. 2006. World Development Indicators 2006. Electronic File. Washington, D.C.: The World Bank.
VI
The Man Behind the Mind
16
The Columbia Years Edmund S. Phelps
So, Guillermo: this is your life! I remember someone advising me in a similar situation a while back that in a speech of this type accuracy about the subject’s life is not important. Style is important. So it’s often better to simplify here and embroider there. There is a lot of bad advice out there, too. You may have seen the Orson Welles movie in which the Charlton Heston character, south of the border on his honeymoon, accomplishes prodigious feats that save the town from bad guys, deeds his wife and others know nothing of.1 Commenting, the Marlene Dietrich character asks, ‘‘What does it matter what you say about someone?’’ Let me tell you, in Festschrift speeches it matters a helluva lot! I won’t go over with you the story line of that movie. Today it’s the story line of Guillermo Calvo we want to stay focused on. But I do remember what I think was the movie’s theme: some important stuff that some people do doesn’t get recognized. Well, for today, anyway, that won’t be true of the stuff Guillermo has done. The Early Years We all know that Guillermo grew up in Buenos Aires, hometown of his wife Sara—and, as a matter of fact, of my wife Viviana. After attending the University of Buenos Aires, he went for a doctorate in economics at Yale, where I got my doctorate. After a spell of teaching in South America, he joined the Economics Department at Columbia, which I had done a year earlier. (Later he went to Penn, which I had also done.) Not a great deal is known about Guillermo’s Yale period. Legend has it that, one day, Guillermo raised his hand in James Tobin’s class to point out a difficulty with something in the lecture. According to the legend, Guillermo was right—and their relationship went nowhere from there. Curiously, around the same time I
432
Edmund S. Phelps
also had a clash with Jim in a crowded seminar room over a book of mine, Fiscal Neutrality toward Economic Growth. Maybe all these parallel experiences were omens that Guillermo and I were destined to come together in a common cause. In fact, Guillermo also took a graduate course of mine on growth economics. It was my first graduate course and I would arrive at every meeting with about forty equations. (I needed to have them in front of me because I was unable to remember them.) I thought it was clear from the look on his face that he understood everything. He sat by the east wall of the Cowles Foundation classroom, near Susan Rose (later Susan Rose Ackerman). Though not bound to accuracy, as I said, I confess that I don’t actually recall any chats with him, and yet I know that we knew each other somehow. Little did I know that in six or seven years’ time we would become allies in writing a new chapter of macroeconomics and close friends. I left Yale in January 1966, starting my work on unemployment and inflation at LSE [London School of Economics], Penn, and CASBS [Center for Advanced Study in the Behavioral Sciences] at Stanford. Guillermo, after writing his dissertation with Tjalling Koopmans, left to teach in South America under a program of the Ford Foundation. The Years at Columbia I reached Columbia in September 1971, knowing that the Economics Department was in a deep hole but hoping there was some way of climbing out. Three of us— Kelvin Lancaster, the chairman, Ronald Findlay, and I—were determined to turn the department around. But how to do it? The emergence of Guillermo was our first big break—though I did not appreciate how big it was going to be, particularly for me. As I remember it, I received one day that fall an international telegram from Guillermo: am leaving the andes stop need to find position in research university stop can you provide assistant professorship? stop guillermo calvo
I set about persuading my colleagues that Guillermo was extremely bright and highly talented. I’m not sure I tried very hard to spread the word to other universities. In any case, it appeared that no other university had bid, at least not one Guillermo wanted to go to, and we landed him for delivery in January 1973. (At the same time I worked strenuously to recruit John Taylor, even to the point of hanging out at a Stanford party to get to know him better.) John was signed up too, for delivery six months after Guillermo. Someone told me that one of them, I am not sure which, was paid $500 more than the other for some reason, and this
The Columbia Years
433
differential continued untouched as long as they were both at Columbia. It was an extraordinarily bureaucratic place in those days. The years with Guillermo and John at Columbia were enormously productive. I had wondered whether my future would be something of a let-down after the ferment of the 1960s and the great luck I had in that decade. And when they arrived at Columbia nothing much happened at first: I was doing optimal taxation in a Rawlsian vein, and Guillermo was finishing up old work. John, however, drew me back into wage and price setting, and when I got stuck I would sometimes chat about the matter with Guillermo. By 1975, we were all on the same page. We did nonsynchronous price and wage setting, a modern version of contract theory, optimal monetary stabilization theory, and some nonmonetary modeling, too. The main effort was the building of New Keynesian theory (in Michael Parkin’s term.) Toward the end of the decade Stanley Fischer told me he felt Columbia had the best department for macroeconomics in the country. Robert Lucas exclaimed, ‘‘Ned, you have a whole school here.’’ Maybe there was no day in the 1970s when I had a mountain high like the feeling I had a couple of times in the 1960s. But there was an extraordinary atmosphere of continual challenge, accomplishment, and the sense of a carpet of new problems and discoveries ahead. When John Taylor left in 1979 it was sad, and when Guillermo left too, in 1986, it was painful. Some of the Papers Of the many papers that Guillermo wrote or cowrote, I particularly treasure four that I was tacitly or explicitly involved in. The most conspicuous of them was his paper on price stickiness, which became an instant classic. At my urging, John had looked at my 1968 paper on money-wage dynamics and wrote down in that spirit a difference-equation system for the wage rates of the various vintages— this quarter, last quarter, and the two before that; meanwhile I was studying a similar system with which to analyze the problem of optimal disinflation. I became almost obsessed with trying to get this system into a continuous-time framework but failed because I hadn’t thought of what would be Guillermo’s trick of supposing that the old prices fade away with a half-life. It was a perfect paper because you had to read it carefully and think about it a while before you could feel that you understood it. The model had the property that a step-increase of effective demand, such as a fiscal stimulus, would cause both employment and the inflation rate to jump; then both would recede back to their steady-state levels, as the price level approached its new steady-state level. An implication was that the stronger the stimulus, and thus the higher the level to which employment jumped, the faster the inflation rate would be falling (following its jump)—this despite the accelerationism inherent in the natural rate construction contained in the
434
Edmund S. Phelps
model. (Some thought Guillermo had got the sign wrong.) It is true that this theorem could have been proven by John and me using our period model of discrete price or wage setting, but we had not seen this delicious implication. Some policy economists, understanding that the price level tends to restabilize, given an inflation stabilization policy in the central bank, conclude that the natural rate is no bar to fiscal stimulus after all, since the inflation rate cannot explode; but this is an evident non sequitur. It is still impossible to sustain by effectivedemand stimulus an employment path bounded above its natural path. The Calvo solution is a demonstration that, when the economy is driven off its natural course, the equilibrium path immediately heads toward home. Of the other three papers, one was the paper on time consistency, which I hope the Phelps-Pollak paper helped to inspire, though perhaps it did not. Another was the paper on general equilibrium in a customer-market economy, which I finally persuaded Guillermo to do with me. That paper turned out to be a forerunner to the many nonmonetary models of structural slumps and structural booms that I began constructing in the last years of the 1980s. By that time, to my embarrassment, I had forgotten about the earlier paper with Guillermo, although I can say that I did it differently the second time around. The joint effort I would like especially to recount is the one on contract theory. I had been invited to write a paper on indexation of money wages and prices. It was suspected that, with enough sophistication, every price and wage would be indexed to the consumer price level and, when that is done, a step-decrease in effective demand would trigger an immediate drop in all prices and money wages, which would save employment from a Keynesian recession. Some of my ruminations led me to criticize what I called the neo-neoclassical theory of wage contracts. I argued that the states of the world in that theory are not of this world: in the real world, at least the capitalist world, the effective state at the firm is to an important degree in the employer’s mind. And the employees cannot see into the head of their employer. So if he tells them the new state is such that they should accept a wage cut without striking or cutting back on their performance, they will not be able necessarily or generally to see all or even much of the evidence that has led the employer to declare the new state; the employer himself may have very little evidence, and be unable to convey much of his observations and impressions. This led me to discussions with Guillermo aimed at a modern contract theory in which the wage has to be a function of observables, such as the layoff level and the level of profits at the firm. We sketched such a model in the appendix to the paper I wrote. Guillermo did the analysis required to derive what was in effect a wage curve relating the wage called for by the contract to the number of employees in the laid-off status. In retrospect, I saw that this was the wrong place to launch such a basic innovation.
The Columbia Years
435
In his Columbia years Guillermo also found time to write plenty of other papers. A favorite of mine is his reworking of Tobin’s ‘‘dynamic aggregative model,’’ which had the distinguishing feature that it dispensed with Keynes’s investment demand function (whose inverse gives the marginal efficiency of investment), replacing it with the marginal productivity of capital function. The paradox of this Keynesian landmark was its tacit implication that consumer demand was powerless to stimulate output and employment—there was 100 percent crowding out. In the Calvo version, a positive shift in the propensity to consume becomes expansionary through its impact on rational expectations of resulting inflation over the future. I wondered at the time whether the paper was a helping hand stretched out to Keynesians or a sort of joke. For if a consumer demand stimulus boosts aggregate demand and employment only to the extent it is expected to cause prolonged inflation (and against a backdrop of fixed-money wage rates), it is not apt to be a popular policy tool. The Personal Side I haven’t forgotten that during our years together at Columbia Guillermo and I were pretty close—especially when I was between marriages and later when he was between marriages. Whether it was about opera or about life, it was usually Guillermo I talked to. And he often had some insight or perception to share. I remember one evening when he and I went to Mozart’s Magic Flute at the Met. At some point the talkative birdman, Papageno, who had been gagged for awhile, is ungagged and says, ‘‘Ah, now I am Papageno again.’’ That line hugely amused Guillermo. His response to it made me realize that it expressed the theme of the opera—that there is nothing as central to us as the ideas and values that we hold and need to profess. I also realized that there was nothing more important to Guillermo than developing and getting out his thoughts. He is outstandingly serious about his ideas and his values. I think we are alike in that way. Someone wrote recently that I am mild-mannered. Guillermo is also mild-mannered. We are like Clark Kent. Beneath Guillermo’s genial exterior is a rigorous and fierce mind that has no patience with sloppy thinking and clueless modeling. I think I am the same—at least I hope so. This must be why we became very close. When he was speaking at my own festschrift celebration in 2001, Guillermo said, introducing his junior coauthors, that they were my ‘‘grandchildren.’’ That makes him my son. Guillermo: I am a proud father. Note 1. The film is A Touch of Evil, with Charlton Heston, Janet Lee, Orson Welles, Marlene Dietrich, and Akim Tamiroff.
17
The Practitioner Roque B. Ferna´ndez
It is a great honor for me to be present at this event in recognition of the contributions of Guillermo Calvo to economic theory, as well as to practical policymaking. As have most of the members participating in this conference, I had the privilege of interacting with him in different moments of my professional life. First, when I was student at the University of Chicago; second, at the time where we were creating the University of CEMA in Buenos Aires; and third, when I assumed a place on the Central Bank of Argentina Board of Directors, and later on as Economic Minister. Most participants are acquainted with his very significant theoretical work, so I will focus my presentation on the role of Guillermo Calvo as an economic practitioner or adviser rather than as a theoretical economist. At the time of my being in the Central Bank, I asked the IMF for technical assistance, suggesting that I would need the technical advice of Guillermo Calvo, and the IMF conceded my request. That was the end of the eighties and beginning of the nineties when Argentina was suffering a hyperinflation. In short, there were two fundamental problems: one, the conventional fiscally ridden monetary expansion, and the other, the Central Bank support of the financial system. As the dynamics of hyperinflation evolved, monetary expansion was mostly devoted to preserving the financial system, and the Central Bank acted as a lender of last resort to commercial banks whose liabilities were short-term deposits indexed to wholesale prices, consumer prices, and a special index contructed by averaging nominal interest rates paid to depositors. Reserve requirements were at the margin almost 100 percent, and were remunerated by the government at at a rate that kept pace with the prevailing rate of inflation. The peso devaluated erratically, sometimes overshooting the rate of inflation, and depositors often ran against commercial banks to withdraw pesos to buy dollars. Jointly with other members of the Central Bank Board I had worked out some ideas for rescheduling deposits, but we were entangled with several complex issues on bank regulation. Guillermo Calvo helped to disentangle the
438
Roque B. Ferna´ndez
complexities to reach a very simple proposal. The idea consisted of offering depositors a government bond denominated in dollars in exchange for short-term peso deposits (most deposits were seven-day deposits). This idea was enacted in 1990. The government bond was a ten-year bond (BONEX 89), and the exchange was not a voluntary exchange; except for special exceptions (small deposits, old people, or people affected by serious illness) the exchange was compulsory. As a result of the financial reform, in a few months the rate of inflation was significantly reduced and Argentina started a period of sustained growth with price stability that lasted for more than ten years. The dollar-denominated bonds were fully paid, and depositors reestablished the confidence in the financial sector. Although the idea helped solve a chronic problem of more than twenty years of high and volatile inflation, and it was accepted over time as a reasonable solution, initially there was some resistance, and Guillermo Calvo had to endure a personal situation totally unexpected in his role of economic adviser. After participating in a lunch meeting with the Argentine Bankers Association (ADEBA) where he did not talk or answer any question related to financial reform, one of the bankers— who had anticipated that the rescheduling of bank deposits was a potential policy decision—decided to present a complaint to the Argentine IMF representative. Essentially the banker blamed the IMF for sending an adviser who had not thought enough about the complexity of the situation and was giving policy makers misleading technical assistance that could jeopardize the stability of the Argentinean banking system. Was the banker right? Policy makers could be wrong, and could be responsible for taking quick decisions, perhaps without full information, and perhaps without enough time to think about what they are doing, but economic advisers are supposed to know everything beforehand. They are supposed to have studied and evaluated a similar situation in the past, and they are supposed to have the relevant economic theory at hand to answer all the questions policy makers might have. Was Guillermo Calvo really qualified for the job? Had he really thought enough about what he was recommending? Some of his thoughts at the time were documented in his papers ‘‘A Delicate Equilibrium: Debt Relief and Default Penalties in an International Context’’ (1989) and ‘‘Debt Relief and Debt Rescheduling: The Optimal-Contract Approach’’ (Calvo and Kaminsky 1991). In what follows I will comment on some of the topics on policy-making covered in those papers as well as on some personal recollections on several policy issues, but limiting myself just to the topic of emerging-markets sovereign debt. Rescheduling of deposits in Argentina at the end of the eighties and beginning of the nineties was essentially a rescheduling of government debt. Assets of commercial banks were mostly government obligations, either because commercial banks financed government spending, or because high reserve requirements (fully
The Practitioner
439
remunerated by the government) were enacted to sterilize government-related monetary expansion. That was not the only government debt; there was also defaulted debt on syndicated loans from international commercial banks. Therefore rescheduling of deposits was the first step in dealing with the problem of the debt that started in 1989 and ended with a Brady bonds exchange for old debt in 1993. Argentina was not alone; several other countries were having a debt problem and Guillermo Calvo observed two fundamental types of events: first, the international debt problem that arose at the start of the eighties was pretty much around in the early nineties, with the financial community reacting to the first signs of distress by launching a coordinated action through multilateral institutions under the assumption that high interest rates and debtors’ term-of-trade deterioration were only passing difficulties, which a payments’ rescheduling would help to overcome. Second, after impressive adjustment efforts, debtor countries were hardly able to collect the resources necessary to resume normal payment of their debt obligations. Attempting to collect higher revenues led in some cases to negative growth and record-high inflation. He conjectured that, sooner or later, a feeling would start to emerge that it could be unrealistic to expect debtor countries to repay their debts in full. At the end of the eighties the debt-reduction view gained momentum with the Brady Proposal, according to which debtors and creditors were encouraged to find mutually acceptable ways to reduce the value of contractual debt. At the same time several market operators warned that debt reduction could become a tug-of-war process, giving the wrong incentives to future debtors. If that were the case, it would be very hard for new lending to resume in a significant way. To address this difficult issue Calvo and Kaminisky tried to go behind the veil of formal contracts and obtain information about implicit contract clauses that might account for the possibility of debt reduction. On sovereign debt contracts, they assumed that the premium over a risk-free interest rate was equivalent to an insurance premium the debtor had to pay in the good states of nature in order to cover the lender against partial repayment in bad states. They envisioned debtors and lenders as engaging in optimal consumption-smoothing contracts, which were not fully state-contingent only because there were ‘‘verification’’ costs (as defined by Townsend 1979). That is, in ‘‘good’’ states of nature not subject to verification, the debtor is obligated to pay the principal and contractual interest rate, which is given by a risk-free rate plus the prespecified interest-rate (or risk) premium. On the other hand, in ‘‘bad’’ states of nature where verification is activated, a fixed cost is paid by the borrower, in exchange for which he is only obligated to pay just a reduced amount of principal and interest. In this view, debt reduction could not be rationalized as an incentive to play an incompetent macroeconomic
440
Roque B. Ferna´ndez
policy. The implicit contract allowed for rescheduling or sizable debt reductions totally consistent with rational full precommitment. Strategic defaults were ex ante ruled out by sizable verification costs assuring debtors’ willingness to pay. Most policy makers, as well as high-ranking multilateral officials, thought that modeling implicit rescheduling and default with the standard debt contract seemed highly stylized and unrealistic. But now we could argue, for example, that at the beginning of the nineties with a hyperinflation state of nature, rescheduling deposits from seven days to ten years was contemplated in a sort of an implicit contract. Today we have additional information we did not have at the time of the policy action. The rescheduling was legally challenged, and the Supreme Court of Argentina ruled in favor of the rescheduling given the prevailing (stateof-nature) financial conditions of Argentina. Of course it would be hard to tell that Brady bonds, or global bonds (as were denominated later on), had implicit clauses for rescheduling or debt reduction. On the contrary, collective action clauses allowing for rescheduling, interest, or debt reduction were excluded in emerging-market bonds under New York governing law. But it is nevertheless true that during the nineties emerging markets were able to issue a significant amount of debt under New York governing law that in several cases was eventually defaulted, rescheduled, and significantly reduced. If debt reduction is explicitly ruled out, how can economists assume implicit covenants for debt reduction? Is there any provision in the legal tradition or governing law of financial centers to make explicit what we consider implicit in modeling optimal contracts? Again, today we can show the existence of implicit covenants for debt reduction, and not only that, today those implicit covenants have been accepted in the courts under the New York governing law. Under the law of the state of New York (see Choi and Gulati 2003) there are two fundamental groups of legal provisions that are particularly relevant in sovereign debt contracts. One group provides for unanimous action clauses (UACs) requiring approval of a hundred percent of bondholders before any haircut is enacted. This means that just one holdout may cause efforts to modify the payment obligations of a bond to fail. The other group are generally denominated exit consents, covering matters such as negative pledges, governing law, submissions to jurisdiction, and listing provisions, that can be used to implicitly circumvent the unanimity requirement. Exit consents can be modified with simple majority (51 percent), or in some instances with special majority (66 percent) of the outstanding bonds. To illustrate, imagine the case of outstanding bonds having UACs covering just payments dates for principal and interest, and the sovereign in financial distress
The Practitioner
441
has managed to work out a coalition with a majority of domestic bondholders to enact a restructuring deal. Suppose that external bondholders, either original bondholders or in a coalition with some domestic bondholders and vultures (internal and/or external), refuse to go along with the deal because they have expectations of a better deal. Exit consents could be used by the sovereign in coalition with domestic bondholders to change old bond provisions prior to their exit from the old bonds. Through exit consents, domestic bondholders exchange their old bonds for the new restructured bonds, simultaneously consenting to changes in the terms of the old bonds not covered under UAC provision. Through changes in the governing law, listing provisions, and other terms, the exit consent procedure can diminish dramatically the value of the old bonds. Changing the governing law from that of New York to that of the home country, which could be less sympathetic to holdout behavior, would make it harder for the holdouts to sue. Alternatively, exit consent may be used to rescind the issuer’s waiver of sovereign immunity or its consent to jurisdiction in New York, both of which would complicate the ability of holdouts to bring suit to enforce their rights under the old bond covenant (See Ferna´ndez 2003 and Ferna´ndez and Ferna´ndez 2004). Obviously, implicitly there could be a lot more than one can imagine in legal contracts, and Calvo was right in his theoretical assumption of an implicit debt reduction possibility. Either implicitly or explicitly, covenants with debt reduction are not possible in good states of nature; debt reduction is only possible in bad states of nature. I will not enter into the very relevant criticism that most of the time bad luck is endogenous, or that bad luck is the natural outcome of wrong economic policies. I will not enter because I mostly agree with that criticism, and, obviously, policy makers should be very careful to apply the right economic policy to avoid endogenous bad luck. But some times exogenous bad luck happens, and debt rescheduling or debt reduction is an unwanted result. And this brings me to the last topic of this presentation. Is there any other possibility to deal with true bad luck? As the nineties went through, bad states of nature were renamed as sudden stops. And capital flows volatility became an important item in the research agenda of Guillermo Calvo. If someone asks anyone knowledgeable from the media who is Guillermo Calvo, the most probable answer is ‘‘an economist that predicted the Tequila Crisis.’’ Although being able to predict has its own merits, sometimes it becomes the sad part of the economic practitioner. On some occasions, for a professional economist, being able to predict a sudden stop is equivalent to a physician being able to predict an incurable illness. The adviser would never know if a policy maker will be able to survive if the prediction of a sudden stop is accurate.
442
Roque B. Ferna´ndez
In a recent work, analyzing the default crisis of 2001–2002, Calvo, Izquierdo, and Talvi (2003) pointed out that Argentina, being an economy extremely closed to international trade, was extremely vulnerable to a sudden stop such as the one that followed the Russian crisis. But again, in this respect Argentina was not alone, and it is relevant to mention that sudden stops have also affected other emerging economies producing important financial crises, as was the case in East Asian countries in 1997. As reported by Andrew Crockett (2002), the international experience in the aftermath of the East Asian crisis exhibits two new features. First, East Asia has become a very large net capital exporter. The external demand generating the export surplus has undoubtedly facilitated the region’s recovery from the crisis. And countries with current account deficits, particularly the United States, have benefited from capital inflows to finance their deficits. Nonetheless, it is hard to believe that East Asia should be an exporter of savings in the longer term, or that the U.S. deficit is indefinitely sustainable. The second feature is that the region is exporting safe capital while importing risky capital. This international exchange of risk is restoring and strengthening national and corporate balance sheets in Asia and is thereby rendering the region’s economies more resilient, and perhaps less vulnerable to sudden stops. Still, this pattern has drawn the criticism that it has impeded the development of East Asia’s own bond markets. East Asian countries are small open economies from a trade account point of view, while Argentina and other Latin American countries are small open economies from a capital account point of view. Nevertheless, I believe that exporting safe capital and importing risky capital is a common feature to East Asia and Latin America after the crisis of the nineties. Therefore it is interesting to evaluate the situation of a post-financial crisis or post-default dynamics of capital flows in these emerging economies. First, one should ask the question of what is wrong in exporting safe capital and importing risky capital? As I said, the criticism is that it impedes the development of a domestic bond market. To some extent this is not totally true, as we have observed that countries in open default (Argentina at the time of this conference) have managed to have new bonds with a significantly lower country risk than old (defaulted) bonds. It is also true that there are expectations of successful negotiations with defaulted bondholders, and that the new bonds survive thanks to restrictions to the local financial system. But even in the case that new local bonds could not be issued, it is not clear what is the welfare loss (if any) of importing the services of financial intermediation. One could argue that local savers are better off importing the services of a governing law under a foreign jurisdiction until structural reforms assuring stability of property rights are enacted in
The Practitioner
443
the home country. One could also argue that nonsubsidized private local borrowers could also be better off taking international loans collateralized abroad with their own savings or export goods. In sum, one could imagine very realistic situations where a fully open financial system with unrestricted capital account flows is the best possible outcome, while institutional reforms are enacted to assure a sound development of local financial and capital markets. Second, one should also ask the question if there are fundamental reforms in the international financial institutions that could help to alleviate the volatility of capital flows to emerging economies. Should sudden stops require a fundamental review in the explicit or implicit legal provisions governing sovereign debt contracts? This is a question that has received considerable attention in later years and there have been two confronting views. One would argue in favor of a more active role for international financial institutions along the traditional lines of financial liquidity assistance, and other would argue the opposite, in the sense that international financial assistance could make things worse. For example, Dooley (2000) has argued that liquidity assistance of the type proposed by supporters of the idea of an ‘‘international lender of last resort’’ would not eliminate sudden stops; on the contrary, it would promote larger sudden stops. He has persuasively argued that models of corporate finance discussing the standard private debt contract (similar, but not quite the same as Calvo and Kaminisky) are the appropriate theoretical frameworks if sovereign immunity is accounted for. Output losses following default (or ‘‘verification costs’’ in Calvo terminology) are the result of a second-best optimum of an incomplete contract and not a mistake in contract design. In this debate Calvo has been very careful in his advice and has not endorsed the view of an international lender of last resort. In his work ‘‘Explaining Sudden Stops, Growth Collapse, and BOP Crisis,’’ Calvo argued that policy makers should aim at improving fiscal institutions to avoid endogenous bad luck, as argued before. Lowering the fiscal deficit is highly effective in the medium term, but could be counterproductive in the short run if it relies on higher taxes. The key assumption is that higher taxes lower the after-tax marginal value productivity of capital, pushing the economy to a low-growth equilibrium. International financial institutions could help to break the stalemate by offering loans for reform. If successful, the loans will be fully repaid because fiscal reform would place the economy on a high-growth path. To conclude I want to express my gratitude to all the people involved in preparing this conference and also to Guillermo Calvo. Over the years I benefited not only from his technical advice, but also from his friendship and support at times when I needed it most.
444
Roque B. Ferna´ndez
References Calvo, Guillermo. 1989. ‘‘A Delicate Equilibrium: Debt Relief and Default Penalties in an International Context.’’ In Analytical Issues in Debt, eds. J. A. Frenkel, M. P. Dooley and P. Wickmam. Washington, D.C.: International Monetary Fund. Calvo, Guillermo. 2002. ‘‘Explaining Sudden Stops, Growth Collapse, and BOP Crisis: The Case of Distortionary Output Taxes.’’ Mimeo., IADB, University of Maryland, and NBER. Calvo, G., A. Izquierdo, and E. Talvi. 2003. ‘‘Sudden Stops, The Real Exchange Rate and Fiscal Sustainability: Argentina’s Lessons.’’ Working Paper No. 9828, NBER, Cambridge, MA. Calvo, Guillermo, and Graciela Kaminsky. 1991. ‘‘Debt Relief and Debt Rescheduling: The OptimalContract Approach.’’ Journal of Development Economics 36: 5–36. Choi, Stephen, and Mitu Gulati. 2003. ‘‘Why Lawyers Need to Take a Closer Look at Exit Consents.’’ International Financial Law Review, September: 15–18. Crockett, Andrew. 2002. ‘‘Capital Flows in East Asia Since the Crisis.’’ Bank for International Settlements. Speech given at the ASEAN Plus Three Meeting, Beijing, October 11. Dooley, Michael. 2000. ‘‘Can Output Losses Following International Financial Crisis be Avoided?’’ Working Paper No. 7531, NBER, Cambridge, MA. Ferna´ndez, Roque. 2003. ‘‘Crisis de Liquidez y Default Estrate´gico en el Servicio de la Deuda Soberana.’’ Academia Nacional de Ciencias Econo´micas, Diciembre. Available at http://www.cema.edu.ar/ u/rbf/. Ferna´ndez, Katherina, and Roque Fernandez. 2004. ‘‘Willingness to Pay and the Sovereign Debt Contract.’’ Universidad del CEMA, March. Available at http://www.cema.edu.ar/u/rbf/. Townsend, R. M. 1979. ‘‘Optimal Contracts and Competitive Markets with Costly State Verification.’’ Journal of Economic Theory 21: 265–293.
18
In His Own Words: An Interview with Guillermo Calvo Enrique G. Mendoza
Guillermo Calvo is one of the most influential economists in the field of international macroeconomics in the last thirty years. He has produced seminal articles in every area of macroeconomics and international economics he has worked in, including his early classic articles on capacity utilization and time inconsistency; his 1980s works on efficiency wages, price stickiness, and policy credibility; and his recent studies on sudden stops and emerging-market crises. Yet the defining feature of Guillermo Calvo’s contribution to our profession is not the depth and wide scope of the economic theories he has developed, but the central emphasis he puts in all his work on the role of economics as a tool for understanding reality and improving the quality of human life. Guillermo Calvo’s passion for the policy implications of economic theory is obvious to anyone that has met him since his days as Senior Advisor of the Research Department of the IMF in the mid-1980s. This feature of his professional interests was much less obvious to those who interacted with him during his early years as an important figure of the rational expectations revolution. It was probably hard to see that behind the highly technical treatment presented in his articles at that time was an author who had his feet soundly set on the ground and focused on understanding how society could benefit from the renaissance of macroeconomic theory that was taking place. Interestingly, Michael Rothschild did figure out the true nature of Guillermo Calvo in those early years. When the University of California in San Diego tried to hire Calvo in the mid-1980s, Rothschild explained to Calvo that he was an excellent fit for San Diego because he was a particular type of theoretician: ‘‘Most theoreticians make theory out of theory,’’ Rothschild noted, but Calvo was different because he made ‘‘theory out of reality, taking what is really out there in a much wider and complex form than a well-developed but narrow theory.’’ The following pages are excerpts from three interview sessions that Guillermo Calvo and I had at his office in the Inter-American Development Bank in spring 2003. These interviews provide a clear picture of Calvo as the theoretician of
446
Enrique G. Mendoza
reality that Michael Rothschild described. The recollection of the conversation with Rothschild is one of many fascinating memories of Calvo’s personal life and professional career that emerged during our meetings. We taped and transcribed the three meetings in Spanish, and then Calvo and I together edited that material to produce this much shorter interview in English. The Beginnings EM: Guillermo, you were born and raised in Argentina. I imagine that the experiences you went through growing up in turbulent Argentina played a central role in developing your interest in economics, and in a particular class of economic problems. Would you like to say a few words about how growing up in Argentina shaped your interest in the economics profession? GC: When I was finishing high school in Buenos Aires there was no school of economics to speak of. However, I was very lucky because my father, who worked at the central bank, brought me economics books from the library there, under the advice of some of his friends who had worked with Rau´l Prebisch. I found the material strange and fascinating at the same time. One day he brought me the General Theory and I almost decided that economics was not for me! Fortunately, in school there was a course called economics. As programmed, it consisted of some kind of history of economic thought, but, once again, I got lucky. The course was taught by Julio Olivera who had a passion for Walras. As a result, we spent the whole semester discussing the Walrasian system. I could hardly believe that a subject that appeared impossible to tackle when reading Keynes suddenly became so clear and elegant. One year later I joined the central bank and worked under my teacher’s son, Julio H. G. Olivera, whose orders to me were, essentially, go to the library, get a copy of Allen’s Mathematics for Economists and Hicks’ Value and Capital and don’t leave your room until you are done with them. I felt like I had reached nirvana! In addition, there was a seminar series in which we discussed some key papers coming out in journals like the JPE [ Journal of Political Economy]. One of the first such papers was Samuelson’s overlapping generations model. Quite frankly I must say that I became a little dizzy trying to read papers on the frontier of economics, but I found it very encouraging that everything we read had a solid mathematical basis. Thus, even though the deeper economics often escaped me, I felt I had a firm grasp of the reasoning in each one of the papers, and that, I felt, was good enough. Did growing up in such a fascinating economic laboratory, Argentina, influence my decision to go into economics? Maybe, because before I learned economics it was very hard for me to follow the public debate, which was heavily peppered with economic terms. But I believe
In His Own Words: Interview with Guillermo Calvo
447
that what really hooked me at the beginning was the beautiful theory, and finding out that emotion could be subject to mathematical analysis. EM: Let’s talk about your decision to go to graduate school. Can you describe how early in your college years you decided that you would go for an economics PhD abroad, and how you ended up choosing to attend Yale? GC: I had practically no college years because, as I told you, when I started there was no school of economics. Thus, I enrolled in accounting, which I found quite uninteresting. However, at the University of Buenos Aires, where I attended, Professor Julio H. G. Olivera ran a series of seminars that his best students attended, and I was accepted because I worked with him at the central bank. (By the way, Rolf Mantel and Miguel Sidrauski also attended those seminars). In those seminars we read Hick’s Value and Capital, Samuelson’s Foundations, and Koopmans’ Three Essays on the State of Economic Science. With a group of adventuresome classmates I also independently read the recently published Debreu’s Theory of Value, which required learning some basic topology. The Koopmans-Debreu duo (both of whom were at Yale at the time) put Yale among my favorite places to go, and made me think of graduate school, not as the next step in a professional career, but as a door to heavenly intellectual delights! However, scarcely having onethird of college under my belt made my chances of acceptance in graduate school extremely slim. Moreover, my willingness to learn the intricacies of accounting was declining at an alarming speed. Thus, there came a time when I felt that I had fallen into a cruel trap, no exit in sight. However, once again my good fortune gave me a hand. This time in the form of the USAID [United States Agency for International Development], which offered scholarships in a program at Yale! This was an MA program but the best students had a chance of being considered for their PhD program. I was accepted on the basis of recommendation letters, and nobody seemed to care that I was far from graduation. As I was later told, this had been simply an oversight of the admissions committee! EM: What about your years at Yale? Can you give us a quick review of the basic facts (what years where you there, who were some of your classmates, who were the leading members of the faculty), and more importantly, can you tell us about your mentors and your dissertation, and reflect back on how the experience at Yale matched your pre-graduate school aspirations? GC: When I got to Yale in 1964, Debreu had already left for Berkeley, but Koopmans and Herb Scarf were there. Moreover, Ned Phelps taught growth theory, and David Cass was a young assistant professor who paced around the department like a caged animal, full of ideas and projects. In my second year, Joe Stiglitz showed up and the place caught fire! At the same time, we had the strong presence of Tobin and several other famous names. In my particular case, the presence
448
Enrique G. Mendoza
of Carlos Dı´az-Alejandro was very important, because he was in the process of writing his economic history of Argentina, a masterpiece. It was very good to have someone with whom to talk plain, but deep, economics. Besides, Carlos was a great guy who knew how difficult it is to plunge into a new culture and language, and did his best to lighten the burden. Thus, the straight answer to your question is that I was fully satisfied with my experience, but I had not realized how lonely you could become as a foreigner. Classmates? Edmar Bacha was my buddy. We struggled together to disentangle the mysteries of a language that none of us had a good command of, a problem that became especially acute when we had to learn about the inventions of the Industrial Revolution in an economic history course (the Flying Shuttle? . . . we still don’t know what it is!). Ted Truman was another classmate, but I must admit that I socialized very little (I was already married), and there was not a lot of scientific interaction among students. EM: So the year is now 1967 (you were 26 years old), but you finish your PhD at Yale at the age of 33. Why did it take you so long? What were the placement options that you contemplated for the beginning of your career, and which one did you take? GC: In 1967 I went to a summer program in mathematical economics run by Hirofumi Uzawa, thanks to Koopmans’ recommendation. That was a wonderful three months, when I started to work on my dissertation on optimal growth in a vintage capital model. Uzawa had a lighter touch than Koopmans, and thus encouraged me to attack problems without, at first, being too careful about the mathematical details. That was all very positive. However, the Vietnam War was still raging, and campuses had become ground zero for antiwar demonstrators. That plus my own misgivings about the Vietnam War made me feel more and more at a disconnect with the world around me. As a result, I started to look for options outside the U.S. That is why in 1968 I left without finishing my thesis and headed for Lima, Peru, where the Ford Foundation sponsored a postgraduate program at the central bank. Soon after I arrived, there was a military coup, and in 1969 I headed off to Bogota´, Colombia, another Ford Foundation teaching post. I stayed there until December 1972. I wrote several papers inspired by many interesting issues in those countries . . . but no thesis. Remember, this is way before the Internet, and way before mathematics became so easy to type! Thus, sending a draft to your advisor via snail mail was a painful and not very promising endeavor. In a way, I had fallen into another trap. But my good fortune prevailed. Carlos Dı´az-Alejandro heard that Columbia was looking for assistant professors without success, and mentioned my name to his friend (and now mine) Ron Findlay. In short, I visited Columbia and my good rapport with wonderful people
In His Own Words: Interview with Guillermo Calvo
449
like Ron Findlay and Kel Lancaster was immediate. Two or three days later I got a telegram (no faxes, no e-mails, mind you) in Bogota´ with an offer as a lecturer, which would be transformed into assistant professor on writing my dissertation. That I did in 1974, at the ripe age of 33! With the benefit of hindsight, however, this long gestation period was very beneficial because I learned a lot of math to tackle an existence proof in my dissertation. (Such proof, by the way, never saw the light of day because it had become so technical that Koopmans thought he would have to ask the math department for an adviser. To make sure that there were no serious existence problems, he did some informal checking with some mathematicians there who felt my proof was okay, which led Koopmans to give me the green light, allowing me to assume existence). Equally important, in the years prior to actually writing my dissertation I was exposed to a set of very rich economic issues firsthand, which served me as a source of inspiration for many years to come. Thanks to that detour in the real world, I realized how relevant economics was for understanding, and occasionally solving, important problems. At that point, my ‘‘marriage’’ to economics had become complete: beauty and relevance converged! Time Consistency and Other Early Works (The Columbia Days) EM: I have the impression that your Columbia days were very important for your personal and professional life. Can you describe for us the environment that you were exposed to there and how it affected your developing career? GC: Columbia was a crucial step in my career. I had a dream team of colleagues: Ned Phelps, Phil Cagan, Ron Findlay, Bob Mundell, Jagdish Bhagwati, Carlos Rodriguez, John Taylor, Stan Wellisz, Maury Obstfeld, Carlos Dı´az-Alejandro, and more. This outstanding group of scholars provided the right atmosphere to develop abstract ideas. You see, the real world is very good for inspiration, but real-world people, no matter how intelligent they are, have very little patience for theory. I am sure that if I had stayed in the real world my career would have been very different. EM: Together with the classic article by Kydland and Prescott, your article on time inconsistency is credited with making one of the most substantial contributions of the rational expectations revolution to economics, both theory and policy. Can you tell us about the gestation process of the idea itself and also about the process leading to the publication of the paper? GC: My friend Assaf Razin claims that time inconsistency was an obsession with me as far back as 1967 (during a seminar organized by Uzawa in Chicago that Assaf also attended). In fact, I used to tease my younger siblings when we were
450
Enrique G. Mendoza
children with time-inconsistency tricks, so Assaf is likely to be right (although I, quite honestly, do not recall. Was I so sickly obsessive that I did not notice?). In any case, soon after I arrived at Columbia, and prompted by Ned Phelps, I became interested in the theory of justice (Utilitarians, Rawls, and all that). I wrote a paper on the time inconsistency of Rawlsian maximin principle, for example. Simultaneously, I spent many hours with Carlos Rodriguez trying to disentangle the implications of Auernheimer’s recently published JPE paper on an honest government rule for money creation where, it turns out, one can find the seeds of macro time inconsistency. First I wrote an open-economy time inconsistency example, which was published in the Journal of Money, Credit and Banking [ JMCB], and later the better known closed-economy example published in Econometrica. Motivation for writing the latter I owe to Alvin Marty who, learning about the open-economy paper, told me that if I wanted to reach mainstream macroeconomists, I should make the case in terms of a closed-economy model. He was certainly right. It shows, incidentally, how U.S.-oriented macro was at the time (and still is in some places). In any case, when I was working on macro time inconsistency, Ed Prescott showed up at Columbia to give one of his seminal papers on the subject, but I was out of town. When I came back, I asked a senior colleague about Ed’s paper, and he said that he could not quite figure out. Thus, I placed the paper on my to-read list, and did not read it until, during the publication process, the referee (who turned out to be Ed, on his own recognition) pointed an accusing finger at my oversight. This I quickly fixed, the paper was published, and the rest is history! EM: It is interesting that at times a seminal idea in economics leads to interpretations that are in conflict with those of the originator of the idea, particularly when it comes to economic policy. For example, Mundell’s arguments on currency areas and exchange rates have been and are still used by a large number of economists to favor flexible exchange rates, while Mundell himself sees the same arguments as a strong argument in favor of fixed exchange rates. Do you feel that something similar happened with the notion of time inconsistency? To be more precise, what is your own view on the key policy lessons one should take from time inconsistency? GC: It shows that a policy maker could be a saint and still be a damned liar during the election campaign. In other words, time inconsistency cannot be simply dismissed as a character failure, or a political party failure, which the electorate should be able to weed out. Moreover, the examples show that the negative effects of time inconsistency could be reduced, and even eliminated, by devising new policy instruments or institutions. Hence, there are potentially important lessons for policy and institutions. I believe that the independence of central bank issue
In His Own Words: Interview with Guillermo Calvo
451
got a big boost from this literature. I would say the same thing about the political economy literature. EM: On the theoretical side, I find it very interesting that while some researchers went on to study time-consistent policy in dynamic games, a large chunk of the research you did in the second half of the 1980s focused more on the macroeconomic consequences of having to live with time-inconsistent policies (or with the ‘‘lack of credibility’’ of policies, as you referred to it). This use of the word ‘‘credibility’’ is sometimes questioned because the credibility literature that your work started is not based on time-consistent dynamic games. Can you tell us why your interests flowed in a different direction, and what your opinion is of the proper use of the word ‘‘credibility?’’ GC: Good question! I believe there is such a thing as premature formalization, and also believe something like that happened with time inconsistency. Once the profession grabbed hold of time inconsistency, researchers started to play games with it, games not necessarily inspired by the real world or policy relevance, but games mostly inspired by the need to become noticed and be published. Don’t take this as a criticism of the literature. I am a firm believer in pure research. But volume does not make relevance, or even indicate the priorities set by the profession, especially the seriously committed side of the profession. Speaking of myself, after finishing my papers on time inconsistency, I felt that I wanted to go back to simpler grounds where I could better understand the dynamics implied by time inconsistency. Thus, knowing that in a full-fledged model I could generate time inconsistency, I decided to simplify that part by inserting phenomena that resembled time inconsistency in an exogenous manner and focus more sharply on their dynamic implications. That is partly why my policy temporariness papers came to life. But another important reason was that I was trying to explain why orthodox stabilization programs in Latin America in the late 1970s and early 1980s did not work out. I was afraid that public opinion would swing to a crazy heterodox extreme and all fiscal discipline would be thrown out the window. The credibility papers helped to show that when credibility is imperfect, even good orthodox policies could lead to unsatisfactory results. Let me add that my sense is that time inconsistency is not an everyday problem but something that policy makers are drawn to in extreme circumstances. Leo Leiderman and I, for instance, showed in an AEA [American Economic Association] paper that one could rule out time inconsistency from the monetary policy of several developing countries. We did that by showing that the first-order restrictions of time-consistent optimal policy could not be rejected. This suggests that time inconsistency as a daily phenomenon may not be relevant, but does not rule out that one can find it once in a while—and in a big way! I don’t think the formal
452
Enrique G. Mendoza
literature has explored this avenue, with the exception of a few papers, like one by Bob Flood and Peter Isard (‘‘Monetary Policy Strategies,’’ IMF Staff Papers 37: 612–632, 1989), which assumes the existence of a fixed cost in exploiting time inconsistency. EM: There were other important areas you worked on during the early part of your professional career. From my own work, I remember in particular your classic AER [American Economic Review] paper on capacity utilization. Can you tell us more about this paper and some of the other work that you were involved with in the 1970s? GC: This paper was inspired by the debate in Colombia about capacity utilization. Several empirical studies suggested that capacity utilization was very low there, and I wanted to check what basic theory would say about it. However, if anything, my paper thickened the plot because the model implies that capacity utilization should increase with the real exchange rate. Thus, if underdevelopment goes hand in hand with high interest rates (which was the conventional wisdom in that world of low capital mobility), then Colombia should rather display high capacity utilization. On the other work, I already mentioned the theory of justice, but what really absorbed my attention in the early 1970s was the theory of supervision. I wrote several of those papers on the subject with Stan Wellisz. Stan, by the way, is one of my admired figures at Columbia University, and at the time greatly helped me to put relevance into my math scribbling. The focus in these papers was to explain hierarchical ladders in an organization, and structural unemployment. Based on the imperfect supervision paradigm, I wrote one of the first papers on what was later called efficiency wage hypothesis to explain unemployment. However, I eventually stopped working on this field because I felt that the next necessary step was empirical analysis at the micro level. At the time, I could not find a partner to help me plunge into that uncharted territory, and I felt that doing it by myself would have distracted me too much from the macro issues that still captured my imagination. Sticky Prices and Noncredible Policy (Penn and the IMF Transition) EM: To continue on the track of your professional career, you moved from Columbia to Penn in 1986. What were the motivations for your move? Can you summarize for us your experience at Penn? GC: Maury Obstfeld and I left for Penn at the same time. The department of economics at Columbia was in a shambles; young assistant professors like Maury
In His Own Words: Interview with Guillermo Calvo
453
were overburdened with thesis projects and got little administrative support. There were fights among the faculty, some of which became public and made the department a butt of jokes in the profession. Leaving was a very hard decision for my wife and me because we both loved New York and had many good colleagues at Columbia. My stay at Penn was pleasant but short, however. Along came an attractive offer to visit the IMF Research Department, and later a permanent offer, right after the collapse of the Berlin Wall. Being at the Fund meant having a front seat to observe this colossal drama. EM: Your 1983 JME [ Journal of Monetary Economics] article on staggered prices is one of the most widely cited articles in the recent literature on general equilibrium models with nominal rigidities. Readers will be very interested in learning what motivated you to develop the ideas you proposed in this paper. GC: Ned Phelps and John Taylor had made big strides on that kind of model for several years (as I recall they were already hard at work on these issues in 1973). However, I did not get very interested in it, because I was trying to understand high-inflation countries where, I thought, price stickiness should not be a big issue. All of that changed in 1981 when Argentina abandoned its ‘‘Tablita’’ stabilization plan and underwent a series of large devaluations. To my surprise, unemployment took a big jump, and the real exchange rate suffered a sizable increase. I could not understand this in terms of flexible-price models, which led me to pay more attention to the Phelps-Taylor approach. (Let me add that nowadays I would first turn my attention to imperfections in the credit market, rather than price stickiness, but that was the early 1980s and the conventional wisdom in theory circles—although not necessarily among development economists—was that the capital market was not a source of problems in developing countries). Since, once again, my main interest was to understand the dynamics of imperfect credibility, I made simplifying assumptions in order to be able to formulate the sticky-prices model in continuous time (where one can use phase diagrams) and start from utility functions. The latter was important for my purpose, because credibility analysis becomes too ad hoc if you start from demand functions. Let me add a vignette. I started working on this approach when I visited Chicago during the spring of 1981. However, I never made a presentation on it in the Money Workshop, where price stickiness was not part of the conventional wisdom, to put it mildly. I guess I was afraid of being butchered on the spot! You can imagine my surprise when much later I saw mainstream macroeconomists using the framework as a matter of fact. Let me add, that the paper you are referring to is a closed-economy macro paper that I wrote after writing the open-economy version motivated by Argentina’s experience. I wonder what would have happened to my citation count if I had stopped at the open-economy paper!
454
Enrique G. Mendoza
EM: What is your opinion of the recent literature on dynamic general equilibrium models with nominal rigidities? GC: It is motivated by the failure of pure real business cycle models, and it is a natural development. I have contributed to it also with Oya Celasun and Michael Kumhof (trying to account for inflation inertia). Like cash-in-advance, pricestickiness models fill a vacuum in general equilibrium theory without which one cannot even begin to address some basic policy issues in monetary economics. Unfortunately, the microfoundations are still weak. EM: We arrive at the mid-1980s. This is the time when two of your most famous papers on the macrodynamics of the imperfect credibility of economic policy appeared, one in the JPE in 1986 and one in the JMCB in 1987. The JPE paper aimed to explain the consumption booms associated with failed disinflation programs based on exchange-rate management, and the JMCB paper integrated this idea together with a Krugman-style model of currency crisis in an intertemporal optimization framework. Together with the paper that Helpman and Razin published in the AER around the same time (approaching the issue from a different perspective), these papers were the originators of the large literature on the real effects of exchange-rate-based stabilizations. Can you describe what sparked your interest to begin writing about this particular issue? GC: Once again, Argentina’s devaluation showed to me very clearly that nominal devaluations could result in real devaluation, and all kinds of important real effects could follow. That was the motivation. In my work, real effects followed from lack of full credibility, and that is how, in a way, I was tying up this line of research with time inconsistency. I insisted on bringing in utility and production functions because, once again, it is very hard to analyze credibility problems starting from ad hoc demand functions. EM: I would like to switch now for a moment to your transition from Penn to the Research Department of the IMF, which took place in 1987–1988. I wonder, what motivated you to consider leaving academia for the IMF? Can you describe the duties you performed at the Fund and give us your impressions about how this experience influenced your research? GC: I first went to the Fund as a visiting scholar, and later was offered the position left vacant by the departure of Max Corden. The head of research was Jacob Frankel, and he gave me full rein to concentrate on research and visit the field as much as I deemed necessary. The unexpected bonus was guys like you, Carlos Ve´gh, Pablo Guidotti, and Carmen Reinhart, with whom I did many research projects. It was hard to go back to academia under those circumstances, when it appeared that I had been granted a NSF [National Science Foundation fellowship]
In His Own Words: Interview with Guillermo Calvo
455
for life, an airline ticket that would never expire to see the world after the collapse of the Berlin Wall, and an army of first-rate collaborators! EM: You and I met at the IMF when I arrived there in 1989. My recollection, in line with what you just said, is that at that time you were the head of an active group of researchers in international macroeconomics. I recall two important research programs in which you were involved. One of them was your joint work with Carlos Ve´gh, in which you were pursuing further exchange-rate-based stabilization models to analyze the real exchange rate and introduce your staggered pricing setup; the other was your solo work and your joint work with Pablo on self-fulfilling expectations models of public debt. Can you give us a short overview of these two research programs? GC: Carlos Ve´gh gave me the extra kick I needed to bring my open-economy price-stickiness models to full fruition. He brought a lot of enthusiasm and creativity to this endeavor, and I am glad he did because the final product could be called Mundell-Fleming Mark II. With Pablo we worked on the optimal currency denomination and term structure of public debt. This stemmed partly from my AEA 1988 paper on public debt (Servicing the Public Debt: The Role of Expectations, 647–671), which, in turn, was inspired by the repeated failures of Brazil and Argentina to lower inflation to reasonable levels. The original conjecture was that peso-denominated debt was behind this kind of problem because lack of credibility kept nominal interest high, resulting in unsustainable fiscal deficit (if inflation had actually been lowered). This research led us naturally to explore dollarization and term structure. Interestingly, a solution to the original high-interest problem was dollarization of public debt, while the current evidence and research makes one seriously question such a dollarization policy. Missing from those papers, however, is the sudden stop phenomenon that has been prominent in recent financial crises, and leads one to take dollarization with a great deal more caution. EM: At the beginning of the 1990s the nature of the problems affecting international capital markets was changing dramatically. We went from the debt crisis and the problems of disinflation to the bonanza of the first half of the 1990s and the surge of inflows of foreign capital into what would be called later the ‘‘emerging markets.’’ Amid the chorus of optimistic voices praising this phenomenon as an outcome of the painful years of stabilization and reform in developing countries, a paper that you coauthored with Leo Leiderman and Carmen Reinhart argued that this was probably not the case. Would you like to elaborate on the details of your argument? GC: This line of research surged almost fortuitously from several trips I took around Latin America for the Fund at the beginning of the 1990s. In country after
456
Enrique G. Mendoza
country that I visited people talked about capital inflows and explained them by some change in domestic policy. The conventional wisdom at the Fund also attached primary responsibility to domestic policy, particularly the fact that several countries were beginning to undertake structural reform, and the implementation of the Brady Plan. I became suspicious about this line of explanation as I realized that capital inflows were taking place in countries that were pursuing very different policies and, besides, that some beneficiaries of those inflows were still dealing with serious domestic security issues (for example, the Shining Path in Peru). Thus, I conjectured that there must be a common factor in all of these episodes, and that it was likely to lie outside the region; for example, the U.S. Carmen Reinhart (who was working at the Fund’s Research Department at the time), Leo Leiderman (who was a visiting scholar there), and I joined forces to explore this conjecture and, sure enough, we found strong evidence of the relevance of external factors. Nowadays several other papers have replicated our results but, at the time, we got a lot of flak, particularly at the Fund where the prevailing view was that if a country did its homework, the capital market would reward it with stable capital inflows. Our concern was that the flows might be reflecting low interest rates in the U.S., which could be reversed very rapidly (like in the early 1980s) and cause financial chaos. Unfortunately, we were right. The Tequila crisis in Mexico 1994–1995 took place after U.S. interest rates started to rise. EM: While at the IMF you also worked on the issues related to the transition economies moving out of socialism. In your opinion, what were the main challenges that these economies were facing? If you had to grade the contribution that the West and the international financial organizations made to the process of transition, what grade would you give it? Did you have the impression that the tools of economic theory that we had at our disposal were useful in this context, or did you find yourself looking for a new toolbox? GC: Transition economies provided us a colossal experiment where everything, all aspects of the economy, had to be thought out at the same time. The policy issues were at the polar opposite of fine-tuning. Even new institutions had to be established. My impression, however, is that the Fund treated some of these cases as if they were economies with well-running capitalist institutions. Thus, for example, bank credit was drastically curtailed without paying much attention to the resulting credit crunch, given that banks in those countries were, by far, one of the main credit suppliers (coupled with inter-enterprise credit). This led Fabrizio Coricelli and me to write a paper on Poland, and a couple of other papers on inter-enterprise credit (the main alternative to bank credit) in transition economies. But these pieces had little impact. Fortunately, several transition economies have grown new institutions, and whether or not we were right is now a largely irrelevant issue.
In His Own Words: Interview with Guillermo Calvo
457
EM: Now let’s talk about your own transition from the IMF to the University of Maryland in 1994. What motivated your departure from the Fund, and what factors made you decide to move to the University of Maryland? GC: The atmosphere at the Fund had deteriorated since Jacob Frenkel’s departure. This was partly due to losing his protective umbrella, but also to a phenomenon that is quite common in bureaucratic institutions whose main focus is not research: an occasional tendency to devote more resources to the main tasks of the organization, and away from research. Besides, the offer at Maryland was very tempting. The New World after the Mexican Crash of 1994 (Maryland and the IDB) EM: Early in 1994 you prepared some comments for a Brookings Papers article that the late Rudi Dornbusch coauthored with Alex Werner on Mexico’s slow growth performance despite its impressive reform record. Your comments went on a different track and argued that there were causes of grave concern over the possibility of a balance-of-payments crisis in Mexico because of large and growing financial imbalances in the Mexican economy. These comments became the seeds of your well-known paper ‘‘Varieties of Capital Market Crises’’ and of our two 1996 joint papers on the Mexican crisis (one published in AER Papers & Proceedings and the other in the JIE [ Journal of International Economics]). The comments also led to the New York Times article on your work under the heading ‘‘The Prophet of Doom That Was Right.’’ Can you describe the key points of your argument and the thought process that led you in this direction? GC: Rudi and Alex seemed to believe that Mexico could be put on the right track by devaluing 15 percent, and I thought they were leaving out of the picture a key ingredient: financial stocks. In particular, I feared that devaluation would breed distrust in the minds of investors and lead to a run, which is what happened. When it happened I was so excited that I worked around the clock on the varieties paper, and a first version was ready in about three days. The NYT note brought me a level of notoriety that I had never known, expected, or looked for. But I cannot deny that it was fun—and profitable! In Argentina when I get an interview, they still refer to me as the guy who anticipated the Tequila. This is funny, of course. But even funnier is the fact that although I may have anticipated the Mexican crisis, I did not anticipate the Tequila—because what distinguishes the Tequila from a garden-variety balance-of-payments crisis is the contagion that the Mexican crisis provoked around the globe. EM: In our 1996 papers, we argued strongly that Mexico’s crisis was not an isolated phenomenon and instead ought to be seen as the first case of a new breed of
458
Enrique G. Mendoza
global capital-markets crises. We wrote that these crises could occur even if the ‘‘observable’’ fiscal accounts and other so-called fundamentals were in good shape, and that instead phenomena like the anticipation of banking crises and financial contagion in international capital markets could play a central role. My recollection is that authors like Tim Kehoe and Hal Cole or Andres Velasco and Roberto Chang shared our views, but for the majority of the profession it took the Asian crisis of 1997 and the Russian debacle of 1998 to accept this notion. With the large number of crises that have now occurred, do you think that the facts validated the views we expressed in 1996? GC: Yes. Capital markets are incomplete, especially so in emerging markets. In that context, capital inflow episodes do not necessarily generate efficient solutions but at least do not generate catastrophes while inflows last. As credit increases on average, borrowing from a new source can easily offset a cut in a line of credit from another. Thus, under those conditions it is unlikely that serious systemic problems would develop. On the other hand, as capital flows dry up or are reversed, there are few efficient ways to deal with the situation. For the private sector we have bankruptcy regulations, which may be efficient for big firms, but create havoc everywhere else. Moreover, there is no equivalent to Chapter 11 for sovereign countries. Thus, there is a strong element of self-fulfilling expectations when capital goes out—which need not be correlated with traditional fundamentals like fiscal deficits. For the capital outflow to cause damage, the country must display significant debt levels. This explains why all of these crises have followed periods of significant capital inflows. EM: It appears that, compared to the 1990s, we have now entered a very different period in global capital markets in which non-FDI [foreign direct investment] inflows into emerging markets have all but dried up. What lessons should we draw from the crises of emerging markets? For example, do you think developing countries should try to manage or control capital inflows? What role should international financial institutions play in the new era of globalized capital markets? GC: Controls on capital inflows should be a policy of last resort. I believe there is a role for the G7 here, a ‘‘traffic control’’ role, in which a new institution or facility is created to ensure that the price of emerging market debt is not subject to large swings, one way or the other. In a recent paper in Economia, the journal of the Latin American and Caribbean Economic Association (LACEA), I called the new institution EMF, Emerging Market Fund. In this fashion, there will be some control at a global level, helping to prevent global systemic problems, but capital would still flow freely across emerging markets. The alternative of each country establishing its own regulation could give rise to problems akin to competitive devaluations. The end result could be a worldwide investment allocation that
In His Own Words: Interview with Guillermo Calvo
459
could hardly be called efficient. Besides, without international cooperation, it is very hard for an individual country to establish effective controls on capital flows. EM: From the standpoint of economic theory, explaining the phenomenon that you labeled the sudden stop (a sudden, large reversal of the current account and capital inflows coupled with Great-Depression-size collapses in domestic production, absorption, and relative prices) remains a major challenge. We published a short article in the 2000 AER Papers & Proceedings suggesting some lines of research based on various forms of informational and financial-market imperfections that seemed promising venues to account for sudden stops. In your opinion, how far are we in developing a sound theory of sudden stops, and are you confident at this point that we can use this theory to propose policy measures that can be useful to prevent sudden stops? GC: Developing a sound theory of sudden stops requires making assumptions on the nature of market imperfections. Thus, it will be hard, if not impossible, to develop a general theory of sudden stops. Therefore, the goal I have set for myself is to develop a set of examples (a bit what I did in my varieties paper), continue to look closely at emerging markets, and conduct some econometric testing. Unfortunately, the typical theorist in the profession, who feeds himself on theory, will likely have a tendency to try to find explanations in complete-markets theory, because developing that theory can be done without crossing Rio Grande. Thus, if crises cease to be as frequent as they have been in the last decade or so, I am afraid that the profession will tend to commit sudden stop theory based on market imperfections to the trash can. I said ‘‘I am afraid’’ because, as a result, future policy makers (including the IMF) are likely to make the same mistakes they made in the 1990s. EM: In 2001 you took a two-year leave of absence from your Maryland position to take on the job of Chief Economist and head of the newly created Research Department of the Inter-American Development Bank. Can you tell us, what lessons have you learned from your experience as Chief Economist and head of research of an international financial organization? GC: What is exciting about RES at IDB is that it represents the only policyoriented research unit in all of Latin America that has a serious regional and global view. Moreover, through a series of networks, RES is in close contact with key policy makers in the region. I believe this kind of activity is very productive because, to the extent that countries choose the globalization route, they have to engage in serious, professional dialogue with other countries pursuing the same goals at a regional and global level. Moreover, providing countries with a global view is very useful because in a globalized environment countries are highly vulnerable to external shocks. In the absence of a global view, typically
460
Enrique G. Mendoza
the domestic-policy debate tends to ignore global factors and focus mostly on domestic-issues. Politicians love grabbing each other’s throats! Looking Ahead EM: In this last section of the interview, I want to ask you some frequently asked questions about the economics profession and its future. In your opinion, which are the most influential contributions to macroeconomics and international economics over the last thirty years that you have been an active researcher in these fields? GC: I, for one, benefited greatly from the so-called rational expectations revolution. The reason is that it gave me a framework in which I could tackle credibility-type issues. Prior to that, one has to recall that macroeconomists thought that monetary theory could not be formulated under rational expectations. In his famous and very influential hyperinflation paper, for example, Cagan shows that as the economy converges to RE (‘‘perfect foresight’’ in his paper because uncertainty was not explicitly modeled), the steady-state equilibrium became unstable. In particular, the model could display hyperinflation with a constant money supply! As a result, until the early 1970s, most papers assumed adaptive expectations with sufficiently long lags. It could be argued that AE reflects some kind of lack of credibility—people don’t pay attention to policy announcements, and prefer to extrapolate from the past. But aside from the fact that AE would be just one particular form of lack of credibility, there is the fundamental issue that one simply would not know the implications of full credibility, because the latter would imply rational expectations! On the other hand, in the field of international finance, I believe the work of Fleming and Mundell was very important because it showed the macroeconomic relevance of the capital account of the balance of payments for small economies. In addition, it helped to extricate the field from its obsession with not looking outside the womb, and only focusing on closed-economy models. EM: If I ask you to name one or two economists that you consider your favorites, or the ones that have influenced your own work the most, who would you name and why? GC: Here is my short list: Mundell, Phelps, Lucas-Sargent-Wallace. I already explained why Mundell is on the list. Phelps was my teacher at Yale, and I found him fascinating. He really helped me grow when we were colleagues at Columbia. The last trio I have also implicitly alluded to before. The three of them, however, have not only shown that rational expectations is an interesting hypothesis but have done fundamental research that helped to establish RE as a valid scien-
In His Own Words: Interview with Guillermo Calvo
461
tific discipline. By doing that, incidentally, they also opened the door for eventually rejecting the RE hypothesis. EM: In terms of the social value of our profession, what do you think have been the most important contributions of the fields you work on to improving social welfare? GC: It is very hard to tell. I sometimes think that countries learn from their own mistakes (and sometimes not even that!), and there is little an economist can do to help. Having said that, however, I believe the main contribution of macro has been to teach basic issues like the price level has a lot to do with monetary variables; expectations are key and, thus, for policy effectiveness, credibility is central; external factors are important; the financial sector plays a critical role for macro stability, and financial vulnerabilities are behind major catastrophes in recent emerging market crises, and so on. EM: If you had a genie that offered to give you the answer to two unresolved economic problems, which ones would you choose? Or to put it differently, in which areas do you think we should direct our efforts as researchers in the next few years? GC: The two areas where we have made some progress but I feel there is a lot ahead of us are growth and poverty alleviation. Maybe triggering growth will take care of the latter, but poverty alleviation is so important that we should not wait any longer and should put lots of research effort in that direction. EM: Finally, and since you are one of the masters of time inconsistency, if you were given the chance to start over from any of the different periods of your career that we covered here, would you reoptimize and move in a different direction than you did? GC: Not really. I am glad that I have straddled academia and policy circles. I have learned a lot and, above all, I’ve had a lot of fun! EM:
Thanks a lot for this fascinating review of your career.
Reprinted from ‘‘Toward an Economic Theory of Reality: An Interview with Guillermo A. Calvo,’’ by Enrique G. Mendoza. Macroeconomic Dynamics 9, 1 (February 2005): 123–145.
Publications of Guillermo A. Calvo
Books Debt, Stabilization, and Development. With R. Findlay, P. Kouri, and J. B. de Macedo. 1989. Oxford: Basil Blackwell. Eastern Europe in Transition: From Recession to Growth? Proceedings of a Conference on the Macroeconomics of Adjustment. With M. I. Blejer, F. Coricelli, and A. H. Gelb. 1993. Washington, D.C.: World Bank. Money, Exchange, Rates and Output. 1996. Cambridge, MA: MIT Press. Private Capital Flows to Emerging Markets After the Mexican Crises. 1996. Washington, D.C.: Institute for International Economics. The Debt Burden and its Consequences for Monetary Policy. Proceedings of a Conference held by the International Economic Association at the Deutsche Bundesbank. 1998. London: Macmillan. Money, Capital Mobility, and Trade: Essays in Honor of Robert Mundell. Edited with Rudi Dornbusch and Maurice Obstfeld. 2001. Cambridge, MA: MIT Press. Emerging Capital Markets in Turmoil: Bad Luck or Bad Policy? 2005. Cambridge, MA: MIT Press.
Articles ‘‘The Mirage of Exchange Rate Regimes for Emerging Markets Countries.’’ With F. Mishkin. Journal of Economic Perspectives, 2003. Also in Emerging Capital Markets in Turmoil: Bad Luck or Bad Policy? ed. G. Calvo. Cambridge, MA: MIT Press 2005. ‘‘Sudden Stops, the Real Exchange Rate and Fiscal Sustainability: Argentina’s Lessons.’’ With A. Izquierdo and E. Talvi. In Monetary Unions and Hard Pegs, eds. V. Alexander, J. Me´litz, G. M. von Furstenberg. Oxford: Oxford University Press, 2003. Also in Emerging Capital Markets in Turmoil: Bad Luck or Bad Policy? ed. G. Calvo. Cambridge, MA: MIT Press, 2005. ‘‘Relative Price Volatility under Sudden Stops: The Relevance of Balance-Sheet Effects’’ With A. Izquierdo and R. Loo Kung. Journal of International Economics 69, no. 1, June 2006. ‘‘The resolution of global imbalances: Soft landing in the North, sudden stop in emerging markets?’’ With E. Talvi. Journal of Policy Modeling 28, 2006. ‘‘Sudden Stops and Phoenix Miracles in Emerging Markets.’’ With A. Izquierdo and E. Talvi. American Economic Review 96, no. 2, May 2006.
464
Publications of Guillermo A. Calvo
‘‘Inflation Inertia and Credible Disinflation—The Open Economy Case.’’ With O. Celassum and M. Kumhof. Journal of International Economics, forthcoming. ‘‘Efficient and Optimal Utilization of Capital Services.’’ American Economic Review 65, no. 1, March 1975. ‘‘Optimal Growth in a Putty-Clay Model.’’ Econometrica 44, no. 5, September 1976. ‘‘Optimal Maximin Accumulation with Uncertain Future Technology.’’ Econometrica 45, no. 2, March 1977. ‘‘A Model of Exchange Rate Determination under Currency Substitution and Rational Expectations.’’ With C. A. Rodriguez. Journal of Political Economy 85, no. 3, April 1977. ‘‘The Stability of Models of Money and Perfect Foresight: A Comment.’’ Econometrica 85, no. 5, October 1977. ‘‘Some Notes on Time Inconsistency and Rawl’s Maximin Criterion.’’ Review of Economic Studies 45, no. 1, February 1978. ‘‘Urban Unemployment and Wage Determination in LDCs: Trade Unions in the Harris-Todaro Model.’’ International Economic Review 19, February 1978. ‘‘Supervision, Loss of Control and the Optimum Size of the Firm.’’ With S. Wellisz. Journal of Political Economy 86, October 1978. ‘‘On the Optimal Acquisition of Foreign Capital through Investment of Oil Export Revenues.’’ With R. Findlay. Journal of International Economics 8, no. 4, November 1978. ‘‘On the Indeterminacy of Interest Rates and Wages with Perfect Foresight.’’ Journal of Economic Theory 19, no. 2, December 1978. ‘‘Ex-post Behavior of Firms Offering Optimal Employment Contracts.’’ Economic Letters 1, no. 3, 1978. ‘‘Optimal Seigniorage from Money Creation—An Analysis in Terms of the Optimum Balance of Payments Deficit Problem.’’ Journal of Monetary Economics 4, 1978. ‘‘On Models of Money and Perfect Foresight.’’ International Economic Review 20, no. 1, February 1979. ‘‘Optimal Population and Capital Over Time: The Maximin Perspective.’’ Review of Economic Studies 46, no. 1, February 1979. ‘‘Fiscal Policy, Welfare and Employment with Perfect Foresight.’’ Journal of Macroeconomics 1, Spring 1979. ‘‘Quasi-Walrasian Theories of Unemployment.’’ American Economic Review 69, no. 2, May 1979. ‘‘The Incidence of a Tax on Pure Rent: A New (?) Reason for an Old Answer.’’ With L. J. Kotlikoff and C. A. Rodriguez. Journal of Political Economy 87, August 1979. ‘‘Hierarchy, Ability and Income Distribution.’’ With S. Wellisz. Journal of Political Economy 87, no. 5, October 1979. ‘‘Employment-Contingent Wage Contracts.’’ With E. S. Phelps. Journal of Monetary Economics, CarnegieRochester Conference Series no. 5, 1977. Reprinted in Studies in Macroeconomic Theory: Employment and Inflation, ed. E. S. Phelps. New York: Academic Press, 1979. ‘‘Capital Accumulation and Welfare: A Note.’’ Economic Letters 4, no. 2, 1979. ‘‘Tax-Financed Government Spending in a Neoclassical Model with Sticky Wages and Rational Expectations.’’ Journal of Economic Dynamics and Control 2, no. 1, February 1980.
Publications of Guillermo A. Calvo
465
‘‘Apertura Financiera, Paridad Mo´vil y Tipo de Cambio Real’’ (Financial Opening, Flexible Parity and the Real Exchange Rate). Ensayos Econo´micos, Central Bank of Argentina, December 1980. ‘‘Technology, Entrepreneurs and Firm Size.’’ Quarterly Journal of Economics 95, no. 4, December 1980. ‘‘Capitalizacio´n de las Reservas y Tipo Real de Cambio.’’ (Capitalization of Reserves and Real Exchange Rate). Cuadernos de Economia, Chile, April 1981. ‘‘Devaluation: Levels vs. Rates.’’ Journal of International Economics 11, 1981. ‘‘On the Time Consistency of Optimal Policy in a Monetary Economy.’’ Econometrica 46, no. 6, November 1978. Reprinted in Rational Expectations and Econometric Practice, eds. R. E. Lucas Jr. and T. J. Sargent. Minneapolis: University of Minnesota Press, 1981. ‘‘Involuntary Unemployment and Inventories: An Exploratory Model of Equilibrium and Pure Competition.’’ Journal of Macroeconomics 4, no. 3, Summer 1982. ‘‘Growth and Inflationary Finance: Variations on a Mundellian Theme.’’ With D. P. Peel. Journal of Political Economy 91, no. 5, September 1983. ‘‘A Model of Non-Walrasian General Equilibrium: Its Pareto Inoptimality and Pareto Improvement.’’ With E. S. Phelps. In Macroeconomics, Prices and Quantities, ed. J. Tobin. Washington D.C.: The Brookings Institution, 1983. ‘‘International Factor Mobility and National Advantage.’’ With S. Wellisz, Journal of International Economics 14, 1983. ‘‘Staggered Contracts and Exchange Rate Policy.’’ In Exchange Rates and International Macroeconomics, ed. J. A. Frenkel. Chicago: University of Chicago Press, 1983. ‘‘Staggered Contracts in a Utility-Maximizing Framework.’’ Journal of Monetary Economics 12, September 1983. ‘‘Trying to Stabilize: Some Theoretical Reflections Based on the Case of Argentina.’’ In Financial Policies and the World Capital Market: The Problem of Latin American Countries, eds. P. Aspe A., R. Dornbusch, and M. Obstfeld. Chicago: University of Chicago Press, 1983. ‘‘Demand Fluctuations, Inventories and Capacity Utilization.’’ With F. E. Thoumi, Southern Economic Journal 50, no. 3, January 1984. ‘‘Macroeconomic Implications of the Government Budget: Some Basic Considerations.’’ Journal of Monetary Economics 15, no. 1, January 1985. ‘‘Reserves and the Managed Float: A Search for the Essentials.’’ Journal of International Money and Finance 4, no. 1, March 1985. ‘‘The Inefficiency of Unemployment: The Supervision Perspective.’’ Quarterly Journal of Economics 100, no. 2, May 1985. ‘‘Currency Substitution and the Real Exchange Rate: The Utility Maximization Approach.’’ Journal of International Money and Finance 4, 1985. ‘‘Fractured Liberalism: Argentina Under Martinez de Hoz.’’ Economic Development and Cultural Change 34, no. 3, April 1986. ‘‘Temporary Stabilization: Predetermined Exchange Rates.’’ Journal of Political Economy 94, no. 6, December 1986. ‘‘On the Costs of Temporary Liberalization/Stabilization Experiments.’’ In Economic Reform and Stabilization in Latin America, eds. M. Connolly and C. Gonzalez. New York: Praeger Publishers, 1986.
466
Publications of Guillermo A. Calvo
‘‘Welfare, Banks, and Capital Mobility: The Case of Predetermined Exchange Rates.’’ In Structural Adjustment and the Real Exchange Rate in Developing Countries, ed. Sebastian Edwards. Chicago: University of Chicago Press, 1986. ‘‘Balance of Payments Crises in a Cash-in-Advance Economy: Current Account and Real Exchange Rate Implications with Perfect-Foresight Dynastic Families.’’ Journal of Money, Credit and Banking 19, February 1987. ‘‘Real Exchange Rate Dynamic with Fixed Nominal Parities: Structural Change and Overshooting.’’ Journal of International Economics 22, no. 1-2, February 1987. ‘‘On the Costs of Temporary Policy.’’ Journal of Development Economics 27, no. 1-2, December 1987. ‘‘On the Economics of Supervision.’’ In Incentives, Cooperation and Risk Sharing, ed. H. R. Nalbantian. New Jersey: Rowman and Littlefield, 1987. ‘‘Costly Liberalizations.’’ International Monetary Fund Staff Papers 35, September 1988. ‘‘Servicing the Public Debt: The Role of Expectations.’’ American Economic Review 78, no. 4, September 1988. ‘‘Optimal Time-Consistent Fiscal Policy with Uncertain Lifetimes.’’ With Maurice Obstfeld, Econometrica 56, March 1988. An extended version was published in Economic Effects of the Government Budget, ed. Elhanan Helpman, Assaf Razin, and Efraim Sadka. Cambridge, MA: MIT Press, 1988. ‘‘Anticipated Devaluations.’’ International Economic Review 30, no. 3, August 1989. ‘‘Is Inflation Effective for Liquidating Short-Term Nominal Debt?’’ International Monetary Fund Staff Papers 36, no. 4, December 1989. ‘‘Controlling Inflation: The Problem of Non-Indexed Debt.’’ In Debt, Adjustment and Recovery: Latin America’s Prospect for Growth and Development, eds. S. Edwards and F. Larrain. New York: Basil Blackwell, 1989. ‘‘Perspectives on Foreign Debt.’’ In Spanish, with Eduardo Borensztein. El Trimestre Econo´mico, December 1989. ‘‘A Delicate Equilibrium: Debt Relief and Default Penalties in an International Context.’’ In Analytical Issues in Debt, ed. Jacob A. Frenkel. Washington, D.C.: International Monetary Fund, 1989. ‘‘Incredible Reforms.’’ In Debt, Stabilization and Development, eds. Guillermo Calvo, Ronald Findlay, Pentti Kouri, and Jorge Braga De Macedo. New York: Basil Blackwell, 1989. ‘‘Credibility and Nominal Debt: Exploring the Role of Maturity in Managing Inflation.’’ With Pablo Guidotti. International Monetary Fund Staff Papers 37, no. 3, September 1990. ‘‘Time Consistency of Fiscal and Monetary Policy.’’ With M. Obstfeld. Econometrica 58, no. 5, September 1990. ‘‘Interest Rate Policy in a Small Open Economy: The Predetermined Exchange Rates Case.’’ With Carlos Ve´gh. IMF Staff Papers 37, December 1990. ‘‘Indexation and Maturity of Government Bonds: An Exploratory Model.’’ With Pablo Guidotti, in Capital Markets and Debt Management, eds. R. Dornbusch and M. Draghi. New York: Cambridge University Press, 1990. ‘‘From Centrally-Planned to Market Economy: The Road from CPE to PCPE.’’ With Jacob Frenkel. IMF Staff Papers 38, June 1991.
Publications of Guillermo A. Calvo
467
‘‘Debt Relief and Debt Reschedule: The Optimal Contract Approach.’’ With Graciela Kaminsky. Journal of Development Economics 36, July 1991. ‘‘Credit Markets, Credibility and Economic Transformation.’’ With Jacob A. Frenkel. Journal of Economic Perspectives 5, Fall 1991. ‘‘The Perils of Sterilization.’’ IMF Staff Papers 38, no. 4, December 1991. ‘‘Financial Aspects of Socialist Economies: From Inflation to Reform.’’ In Adjustment and Growth in Reforming Socialist Economies: Lessons from Experience, eds. Vittorio Corbo, Fabrizio Coricelli, and Jan Bossak. Washington, D.C.: World Bank, 1991. ‘‘Optimal Maturity of Nominal Government Debt: The First Tests.’’ With Pablo Guidotti and Leonardo Leiderman. Economic Letters 35, no. 4, 1991. ‘‘Temporary Stabilization Policy: The Case of Flexible Prices and Exchange Rates.’’ Journal of Economic Dynamics and Control 15, 1991. ‘‘Optimal Inflation Tax under Precommitment: Theory and Evidence.’’ With Leonardo Leiderman. American Economic Review 82, no. 1, March 1992. ‘‘Stabilizing Previously-Centrally-Planned Economy: Poland 1990.’’ With Fabrizio Coricelli. Economic Policy 0, no. 14, April 1992. ‘‘Optimal Maturity of Nominal Government Debt: An Infinite-Horizon Model.’’ With Pablo Guidotti. International Economic Review 33, no. 4, November 1992. ‘‘Are High Interest Rates Effective for Stopping High Inflation? Some Skeptical Notes.’’ The World Bank Economic Review 6, no. 1, 1992. ‘‘Obstacles to Transforming Centrally-Planned Economies: The Role of Capital Markets.’’ With Jacob A. Frenkel. In Transition to a Market Economy in Central and Eastern Europe, Proceedings of the OECDWorld Bank Conference, Paris, 1992. ‘‘Stagflationary Effects of Stabilization Programs in Reforming Socialist Countries: Enterprise-Side vs. Household-Side Factors.’’ With Fabrizio Coricelli. World Bank Economic Review 6, no. 1, 1992. ‘‘Transformation of Centrally-Planned Economies: Credit Markets and Sustainable Growth.’’ With Jacob A. Frenkel. In Central and Eastern Europe: Roads to Growth, ed. G. Winkler. Washington, D.C.: IMF, 1992. ‘‘Capital Inflows and Real Exchange Rate Appreciation in Latin America: The Role of External Factors.’’ With L. Leiderman and C. Reinhart. IMF Staff Papers 40, no. 1, March 1993. ‘‘Output Collapse in Eastern Europe: The Role of Credit.’’ With Fabrizio Coricelli. IMF Staff Papers 40, no. 1, March 1993. ‘‘On the Flexibility of Monetary Policy: The Case of the Optimal Inflation Tax.’’ With P. Guidotti. Review of Economic Studies 60, no. 3, June 1993. ‘‘Exchange-Rate-Based Stabilization under Imperfect Credibility.’’ With Carlos Ve´gh. In Proceedings from I.E.A. Conference on Open Economy Macroeconomics. New York: MacMillan, 1993. ‘‘Management of Nominal Public Debt: Theory and Applications.’’ In The Political Economy of Government Debt, eds. Harrie Verbon and Frans van Winden. Amsterdam, Holland: North Holland, 1993. ‘‘Trade Reforms of Uncertain Duration and Real Uncertainty: A First Approximation.’’ With Enrique G. Mendoza. IMF Staff Papers 41, no. 4, December 1994.
468
Publications of Guillermo A. Calvo
‘‘Inflation Stabilization and Nominal Anchors.’’ With C. Ve´gh. Contemporary Economic Policy 12, no. 2, April 1994. ‘‘Money Demand, Bank Credit, and Economic Performance in Former Socialist Economies.’’ With M. Kumar. IMF Staff Papers 41, no. 2, July 1994. ‘‘The Capital Inflows Problem: Concepts and Issues.’’ With L. Leiderman and C. Reinhart. Contemporary Economic Policy 12, no. 3, July 1994. ‘‘Capital Inflows to Latin America: The 1970s and the 1990s.’’ With C. Reinhart and L. Leiderman. In Economics in a Changing World, Proceedings of the Tenth World Congress of the International Economic Association, Vol. 4, ed. E. L. Bacha. London: Macmillan, 1994. ‘‘Credibility and the Dynamics of Stabilization Policy: A Basic Framework.’’ In Advances in Econometrics, Sixth World Congress, Vol. II, Proceedings from the VI World Meeting of the Econometric Society, ed. C. A. Sims. Cambridge, UK: Cambridge University Press, 1994. ‘‘Interenterprise Arrears in Economies in Transition.’’ With F. Coricelli. Empirica 21, 1994. ‘‘Stabilization Dynamics and Backward-Looking Contracts.’’ With C. Ve´gh. Journal of Development Economics 43, no. 1, 1994. ‘‘Targeting the Real Exchange Rate: Theory and Evidence.’’ With C. Reinhart and C. Ve´gh. Journal of Development Economics 47, 1995. ‘‘Fighting Inflation with High Interest Rates: The Small-Open-Economy Case under Flexible Prices.’’ With C. Ve´gh. Journal of Money, Credit, and Banking 27, no. 1, March 1995. ‘‘The Management of Capital Flows: Domestic Policy and International Cooperation.’’ In The International Monetary and Financial System: Developing-Country Perspectives, ed. G. K. Helleiner. New York: St. Martin’s Press, 1996. ‘‘Inflows of Capital to Developing Countries in the 1990s.’’ With L. Leiderman and C. Reinhart. Journal of Economic Perspectives 10, no. 2, Spring 1996. ‘‘Capital Flows and Macroeconomic Management: Tequila Lessons,’’ International Journal of Finance Economics 1, 1996. ‘‘What Role for the Official Sector?’’ With M. Goldstein, in Private Capital Flows to Emerging Markets after the Mexican Crisis, eds. G. Calvo, M. Goldstein, and E. Hochreiter. Washington, D.C.: Institute for International Economics, 1996. ‘‘Capital Flows in Central and Eastern Europe: Evidence and Policy Options.’’ With R. Sahay and C. Ve´gh, in Private Capital Flows to Emerging Markets after the Mexican Crisis, eds. G. Calvo, M. Goldstein, and E. Hochreiter. Washington, D.C.: Institute for International Economics, 1996. ‘‘Disinflation and Interest-Bearing Money.’’ With C. Ve´gh. Economic Journal 106, no. 439, November 1996. ‘‘Monetary Policy and Interenterprise Arrears in Post-Communist Economies: Theory and Evidence.’’ With F. Coricelli. Journal of Policy Reform 1, 1996. ‘‘Petty Crime and Cruel Punishment: Lessons from the Mexican Debacle.’’ With Enrique G. Mendoza. American Economic Review 86, no. 2, May 1996. ‘‘Reflections on Mexico’s Balance of Payments Crisis: A Chronicle of Death Foretold.’’ With E. Mendoza. Journal of International Economics 41, September 1996. ‘‘Growth, Debt, and Economic Transformation: The Capital Flight Problem.’’ In New Theories in Growth and Development, eds. F. Coricelli, M. DiMatteo and F. H. Hahn. Cambridge, UK: Macmillan, 1997.
Publications of Guillermo A. Calvo
469
‘‘Varieties of Capital-Market Crises.’’ In The Debt Burden and its Consequences for Monetary Policy, eds. G. Calvo and M. King. New York: Macmillan, 1998. ‘‘Why is ‘the Market’ so Unforgiving? Reflections on the Tequilazo.’’ In Financial Reform in Developing Countries, eds. J. M. Fanelli and R. Medhora. New York: Macmillan, 1998. ‘‘Uncertain Duration of Reform: Dynamic Implications.’’ With A. Drazen. Macroeconomic Dynamics 2, no. 4, December 1998. ‘‘Capital Flows and Capital-Market Crises: The Simple Economics of Sudden Stops.’’ Journal of Applied Economics I, November 1998. ‘‘Inflation Stabilization.’’ With C. Ve´gh. In Handbook of Macroeconomics, eds. J. Taylor and M. Woodford. Amsterdam, Netherlands: North Holland, 1999. ‘‘Empirical Puzzles of Chilean Stabilization Policy.’’ With Enrique G. Mendoza, in Chile: Recent Policy Lessons and Emerging Challenges, eds. G. Perry and D. Leipziger. Washington, DC: World Bank, 1999. ‘‘Capital-Market Crises and Economic Collapse in Emerging Markets: An Informational-Frictions Approach.’’ With Enrique G. Mendoza. American Economic Review 90, May 2000. ‘‘Contagion, Globalization, and the Volatility of Capital Flows.’’ With Enrique G. Mendoza, in Capital Flows and the Emerging Economies, ed. Sebastian Edwards. Chicago: University of Chicago Press, 2000. ‘‘When Capital Flows Come to a Sudden Stop: Consequences and Policy.’’ With Carmen Reinhart, in Reforming the International Monetary and Financial System, eds. Peter B. Kenen and Alexander K. Swoboda. Washington, D.C.: International Monetary Fund, 2000. ‘‘Balance of Payments Crises in Emerging Markets: Large Capital Inflows and Sovereign Governments.’’ In Currency Crises, ed. Paul Krugman. Chicago: University of Chicago Press, 2000. ‘‘Betting Against the State.’’ Journal of International Economics, June 2000. ‘‘Rational Contagion and the Globalization of Securities Markets.’’ With Enrique G. Mendoza. Journal of International Economics 51, no. 1, June 2000. ‘‘Fixing For Your Life.’’ With C. Reinhart. Brookings Trade Forum, 2000. ‘‘Capital Markets and the Exchange Rate: With Special Reference to the Dollarization Debate in Latin America.’’ Journal of Money, Credit and Banking 33, no. 2, May 2001. ‘‘Economic Policy in Emerging Markets’’ (in Spanish). Moneda y Cre´dito, 212, 2001. ‘‘On Dollarization.’’ Economics of Transition Journal 10, no. 2, 2002. ‘‘Contagion in Emerging Markets: When Wall Street is a Carrier.’’ Proceedings from the International Economic Association Congress, Vol. 3, Buenos Aires, Argentina, 2002. ‘‘Fear of Floating.’’ With C. Reinhart. Quarterly Journal of Economics 117, no. 2, 2002. ‘‘A Theory of Inflationary Inertia.’’ Volume in honor of Edmund S. Phelps, eds. P. Aghion, R. Frydman, J. Stiglitz and M. Woodford. Princeton, N.J.: Princeton University Press, 2003. ‘‘Explaining Sudden Stops, Growth Collapse and BOP Crises: The Case of Distortionary Output Taxes.’’ IMF Mundell-Fleming Lecture, IMF Staff Papers, 2003. Also in Emerging Capital Markets in Turmoil: Bad Luck or Bad Policy? ed. G. Calvo. Cambridge, MA: MIT Press 2005. ‘‘The Mirage of Exchange Rate Regimes for Emerging Markets Countries.’’ With F. Mishkin. Journal of Economic Perspectives 17, no. 4, 2003.
Index
Acemoglu, D., 330, 361 Ackerman, Susan Rose, 432 Alesina, A., 122, 124, 130 Allen, R. G. D., 446 Amadeus database, 338 American Economic Association, 451, 455 American Economic Review, 452, 454, 457, 459 Andersen, P. K., 408 Andorra, 125 Angeletos, G.-M., 249 Arellano, Cristina, 109, 112 Argentina, xiv, xix, 74, 103, 105, 446, 448 Calvo policy for, 437–438 Cavallo and, 76 central bank independence and, 78 credibility and, 108 debt rescheduling and, 440–442 deposit rescheduling in, 438–439 dollarization and, 122 double mismatch and, 217 financial intermediation and, 252, 260, 264, 272 globalization and, 173, 176, 182, 187, 203 government debt and, 251 inflation policy and, 75–76, 437, 455 movable collateral and, 266 nominal rigidities and, 454 private bond markets and, 269–270 recovery and, 218–219, 221, 226–241 Tablita plan and, 453 Tequila Effect and, 99 wealth redistribution and, 96 Argentine Bankers Association (ADEBA), 438 Asia. See also Specific country crisis in, xvii, 99, 101, 103, 173, 183, 200, 442, 458 double mismatch and, 217 East Asian Four and, 219, 221 fiscal health of, 217–222 rapid recovery of, 218
Asian Bond Market Initiative, 269 Asian Development Bank, 264, 266, 269 Asset market, 223–224, 241–244 Asymmetry, 5, 16–17, 249–250 Auernheimer, Leonardo, xv, 41–68 Bacha, Edmar, 448 Backus, David K., 7 Baldwin, R., 280 Banco Central do Brasil, 83–84 Bank for International Settlements (BIS), 251, 270–272 Bank of Canada, 78 Bank of England, 81 Banks advanced vs. emerging economies and, 72–74 Argentina crisis and, 75–76 central, 71, 81 (see also Central banks) financial intermediation and, 249–252, 260, 264, 266–272 foreign exchange and, 73 inflation targeting and, 79–84 institutional development and, 74–79 runs on, 73 sudden stops and, 75 Barings, 88 Barro, R. J., 122, 124, 130, 134, 379, 392 Barro-Lee model, 134 Bassetto, M., 41 Batini, Nicoletta, 72, 77 Begg, D., 41 Ben-David, D., 379–380 Berg, A., 355 Bermuda, 125, 151n10 Bernanke, Ben S., 77, 79 Bhagwati, Jagdish, 449 Blanco, Herminio, 108 Blejer, Mario I., 88
472
Bloomberg.com, 199 Blundell, R., 145 Bodman, P. M., 381 Brady bonds, 101, 251, 272 Brady Plan, 439, 456 Brash, Donald T., 77 Brazil, xvi, 455 central bank independence and, 82–84 double mismatch and, 217 globalization and, 172–173, 176, 178, 183, 187, 198, 200 inflation targeting and, 79–84 Lula and, 82–84 private bond markets and, 270 recovery and, 219, 221 Brookings Papers ( journal), 457 Buiter, Willem, xv, 41 Bulgaria, 338 Burnside, Craig, 75 Cagan, Phillip, 449, 460 Call bonds, 43 Calomiris, Charles, 75 Calvo, Guillermo, 97, 116 Argentina crisis and, 437–438 background of, 431, 446–449 balance-of-payment crises and, 96 capital income taxation and, 279 Columbia University and, xiii–xix, 432–435, 448–453 contract theory and, 434, 438–440 credibility and, 107, 452–457 credit’s role and, 327 crisis recovery and, 217 debt reduction and, 438–443 elegant simplicity of, 41 financial intermediation and, 250, 272 Fraga and, 80 future policy and, 460–461 hard currencies and, 97 inflation targeting and, 71–74, 77, 79–80, 85, 87– 89 influence of, xiii–xix, 3, 433–435, 437–443, 445 Inter-American Bank and, 459–460 International Monetary Fund (IMF) and, 437– 438, 445, 453–456 liability dollarization and, 73 macroeconomic pessimism and, 159–160 mathematics and, 446, 448, 452 Mendoza interview with, 445–461 Mexican crisis and, 173, 457–460 Mundell-Fleming model and, 21
Index
optimal policy rules and, 18 output collapses and, 380, 392, 400, 408 papers of, 433–435, 438, 443, 445, 453–455, 457 pegging and, 121 Penn and, 452–453 rational expectations (RE) and, 460–461 recessions and, 377–378 recovery and, 218 shock transmission and, 175 sticky prices and, 433, 445, 452–457 sudden stops and, 99–101, 104–105 switching effects and, 3–4 as theoretician of reality, 445–446 time consistency and, 449–452 transition economies and, 456 Calvo, Sara, 431 Canada, 78, 176, 260 Capital flows, 95, 456 balance-of-payment crises and, 96 capital mobility and, 279–293 contagion and, 98–106 continuous time analysis and, 42–46 convergence hypothesis and, 391 crises recovery and, 217–241 financial intermediation and, 249–252, 260, 264, 266–272 fiscal theory of the price level (FTPL) and, 42– 64 globalization and, 173–175, 279–293 (see also Globalization) inflation targeting and, 71–90 Lahiri-Singh-Ve´gh model and, 23–38 liability dollarization and, 73 macroeconomic pessimism and, 159–168 Mundell-Fleming model, 21–38 output collapses and, 377–425 (see also Output collapses) perfect mobility and, 21 sudden stops and, 73–75, 96–114, 378, 392, 394, 443, 445 taxes and, 279–293 transition economies and, 327–345, 349–376 (see also Transition economies) Capital mobility arbitrage condition and, 288–289 globalization and, 279–293 households and, 280–284 labor and, 279–286 median voter equilibrium and, 287–289 Pareto optimality and, 285 policy tools and, 286–287 political economy and, 279–293
Index
production and, 284–286 taxes and, 279–293 Caprio, Gerald, Jr., 80–81 Cardoso-Lecourtois, M., 337 Carlstrom, C. T., 51, 53 Carstens, Agustin, xiii Cash-in-advance (CIA) constraint credit markets and, 332–338 financial intermediation and, 249 fixed exchange rates and, 35–37 flexible exchange rates and, 35–36 intuition and, 36 Lahiri-Singh-Ve´gh model and, 24–25, 34–38 Nontrader’s binding constraint and, 34–36 Trader’s binding constraint and, 36–37 Cass, David, 447 Castellanos, J., 269 Cavallo, Domingo, 76 Cavallo, Michelle, 112 Celasun, Oya, 454 Center for Advanced Study in the Behavioral Sciences (CASBS), 432 Center for Research on the Epidemiology of Disasters (CRED), 391 Central Bank of Argentina, 437 Central banks accountability and, 71, 81 advanced vs. emerging economies and, 72–74 bank runs and, 73 dollarization and, 121–122 fiscal theory of the price level (FTPL) and, 43–46 independence of, 78, 82–84, 450–451 independent currency unions (ICUs) and, 126– 127 inflation targeting and, 71, 79–84, 87 local-currency pricing (LCP) and, 4–7 pegging and, 55–62 Pou and, 76, 78 Cerra, V., 379 Cespedes, Luis, 21, 112, 217 Chang, Roberto, 21, 96, 112, 217, 458 Charoenseang, J., 218 Chatsuripitak, N., 268–270 Cheap talk, 300–301, 309–310 Chile, xvi, 73–75, 173 financial intermediation and, 260, 267 globalization and, 176 inflation targeting and, 79–80, 84–86, 113 wealth redistribution and, 96 China, 103, 200 Choi, S., 440 Christiano, L. J., 41, 51, 53, 218
473
Claessens, S., 218 Cobb-Douglas production function, 160 Cochrane, J., 41 Cold War, 279 Cole, Harold L., 96, 458 Collateral, 264, 266 Colombia, 182, 176 Columbia University, xiii, xix, 432–435, 448–453 Common currency comparative analysis of, 127–129 effects of, 123–127 empirical model and, 130–132 estimator techniques and, 125 growth and, 121–124 inflation and, 132, 135, 137, 147–148 macroeconomic performance and, 130–139 moral hazard and, 131–132 nonparametric analysis and, 145–147 outcome equations and, 134–135, 147–148 robustness analysis and, 145–149 treatment equation and, 133–134, 146 volatility and, 137, 139 Communante Financiere de l’Afrique (CFA), 122, 127, 148–149 Consumption capital mobility and, 279–293 cash-in-advance (CIA) constraint and, 249 equilibrium and, 28–29 (see also Equilibrium) financial intermediation and, 249–252, 264, 266– 272 Lahiri-Singh-Ve´gh model and, 23–38 local-currency pricing (LCP) and, 4–7 macroeconomic pessimism and, 159–168 Obstfeld model and, 5–18 velocity shocks and, 31–32 Contagion causes of, 100–101 credibility and, 97–104 ex ante distribution and, 104–105 global capital markets and, 104–106, 171–172 information costs and, 104–105 measurement of, 99 rational, 104–106 sudden stops and, 98–106 Continuous time analysis, 42–46 Contract theory, 434, 438–440 Contreras, B., 41 Convergence hypothesis, 391 Corden, W. M., 130, 454 Coricelli, Fabrizio, xviii, 327–347 Corsetti, G., 174, 217 Costa Dias, M., 145
474
Costa Rica, 173 Cottarelli, C., 328 Council for Mutual Economic Assistance (CMEA), 355 Covariance matrix, 131 Cowles Foundation, 432 Cox proportional hazards model, 413 Cranston, R., 266–267 Crawford, V., 309 Credibility, 445, 460 Calvo and, 452–457 contract enforcement and, 267 emerging economies and, 95–97, 107–114 financial frictions and, 107–114 Mendoza-Uribe model and, 107–108 policy implications and, 114–117 property registration and, 267 sticky prices and, 452–457 sudden stops and, 96–104, 107–114 time consistency and, 449–452 Credit markets aggregate effects and, 340–341 cash-in-advance (CIA) constraints and, 332–338 chain system and, 335–338 credit-to-GDP ratios and, 330 cross-country regressions and, 339–345 depth of, 327 empirical analysis of, 338–345 financial sector development and, 328–332 growth during transition and, 332–338 inter-enterprise arrears and, 335–338 interest rates and, 332–333 labor and, 332–333, 337 liberalization and, 327, 331 liquidity and, 333–338 literature on, 328, 330 output collapses and, 377–425 (see also Output collapses) reform and, 327, 335–338 shocks and, 336–337 stability and, 327 threshold model and, 341–344 trade credit and, 335–338 transition economies and, 327–345 underdevelopment of, 327–332 volatility and, 330, 337 World Development Indicators (WDI) and, 338 Crises, xvi–xvii, 443 Argentina, 103 (see also Argentina) Asian, xvii, 99, 101, 103, 173, 183, 200, 442, 458 balance-of-payments, 96, 114–115 Brazil, 79–84
Index
causes of, 390–405 Chile, 79–80, 84–85, 96 chronological data on, 204–213 deteriorating fundamentals and, 176 double mismatch and, 217 Ecuador, 85, 103, 172 emerging market economies and, 98–104 (see also Emerging market economies) event definition and, 381–385 financial intermediation and, 249–252, 260, 264, 266–272 globalization and, 171–175 (see also Globalization) Japan, 172–173 literature on, 217 macroeconomic pessimism and, 159–168 Mexico, xvi, 85, 96, 98–99, 103, 187, 441, 457–460 output collapses and, 377–425 (see also Output collapses) recovery and, 217–241 (see also Recovery) regression analysis and, 390–405 Russia, 73–74, 80, 99, 101, 103, 174, 458 shock transmission and, 171–172 systemic risk and, 175–176 Tequila, 99, 187, 441, 457–460 Thailand, 172–173 time inconsistency and, 95–96 Turkey, 85 wealth redistribution and, 96 Croatia, 338 Crockett, Andrew, 442 Cukierman, Alex, 78 Cumulative liberalization index (CLI), 354 Currency, xvi bank runs and, 73 common unions and, 121–124 (see also Common currency) credibility and, 95–104, 107–117 credit markets and, 327–345 crises recovery and, 217–241 devaluation and, 62 dollarization and, 21, 121–122, 455 (see also Dollarization) Dornbusch model and, 223 duration analysis and, 384 fiscal theory of the price level (FTPL) and, 46–52 floating fears and, 73, 85 globalization and, 200 hard, 97–117, 121–153 (see also Pegging) independent currency unions (ICUs) and, 121– 129, 133, 139, 143, 146–150 inflation targeting and, 72–73
Index
information costs and, 107–117 institutional development and, 74–79 International Monetary Fund (IMF) and, 123 liability dollarization and, 73–74, 109 macroeconomic fundamentals and, 219–222 Mendoza-Uribe model and, 107–108 optimum currency area and, 133 rational herding and, 105 seignorage and, 121 substitution and, 72–73 sudden stops and, 107–114 treatment equation and, 133–134 Czech Republic, 176 Daniel, B., 41 Debreu, G., 447 Debt, xiv, xvii, 97 asymmetric information and, 249–250 Bank for International Settlements (BIS) and, 251 Calvo policy and, 438–443 capital mobility and, 286–287 collateral and, 113 as countercyclical shock absorber, 252 crises recovery and, 217–241 (see also Crises) developing countries and, 251–270 domestic vs. external, 251–260, 271–272 double mismatch and, 217 exchange rate fluctuations and, 84–87 fiat money and, 41 financial intermediation and, 249–252, 260, 264, 266–272 fiscal theory of the price level (FTPL) and, 41–64 forgiveness of, 166 GDP ratios and, 252–260 globalization and, 174–175 (see also Globalization) highly indebted poor countries (HIPCs) and, 320 inflation targeting and, 73–74 institutional development and, 74–79 institutions and, 260–267 legal weaknesses and, 264–267 lemons problem and, 174 movable collateral and, 264, 266 pegging and, 55–62 private bond market and, 268–270 procyclicality and, 252 rescheduling of, 440–442 transition economies and, 327–345, 360 (see also Transition economies) unanimous action clauses (UACs) and, 440–441 de Gregorio, Jose, 80 Deliberative democracy, 301–302
475
Dell’Ariccia, G., 328 del Valle, C., 269 De Melo, M., 355 Democracy deliberative, 301–302 output collapses and, 394, 410, 412 values and, 315 Denizer, C., 355 Depreciation. See Currency Desai, Padma, xvii, 217–246 De Soto, H., 264 Devereux, Michael, xiv, 4, 9, 13, 16–17 Dı´az-Alejandro, Carlos, 448–449 Dı´az-Gime´nez, J., 249 Diebold, F. X., 381 Dietrich, Marlene, 431, 435n1 Di Venuto, N., 381 Djankov, S., 267, 330, 335 Doe, Samuel, 125 Dollar, D., 361 Dollarization, xvi, 455. See also Common currency advanced economies and, 121–122 CFA franc and, 127 description of, 121–122 Ecuador and, 150 El Salvador and, 150 growth and, 122 increase of, 122 independent currency unions (ICUs) and, 124, 139, 143, 150 interest rates and, 122 labor mobility and, 126 liability, 73–74, 109 Mundell-Fleming model and, 21 Panama and, 125, 130 Puerto Rico and, 125 strict, 139–144 tax evasion and, 125–126 treatment equation and, 146 treatments and, 146 volatility and, 143 Dornbusch, Rudiger, 122, 223, 457 Double mismatch, 217 Double threshold model, 343–344 Drazen, Allan, xviii, 295–324 Dupor, B., 41 Duration analysis, 424 correlates for, 412 crises recovery and, 405–422 equilibrium and, 387–390 event definition and, 381–385 exponential parameter and, 412–413, 416–417
476
Duration analysis (cont.) Gini index and, 412 Gompertz parameter and, 412–413, 416–417 hazard rates and, 387–390, 408–422 Herfindahl index and, 417 human capital and, 410 log-logistic parameter and, 412–413, 416–417 log-normal parameter and, 412–413, 416–417 multiple equilibria and, 413, 416 Nelson-Aalen cumulative function and, 387–388 output collapses and, 378–381, 394–405, 408–422 Prentice-Williams-Peterson model and, 413, 418–421 regional effects and, 378–379, 385–390 time-specific effects and, 378–381, 394–405 Weibull specification and, 410 Durdu, C. B., 116 Dynamic aggregative model, 435 East Asian Four, 219, 221 East Caribbean Currency Area (ECCA), 122, 127, 148–150 Economia ( journal), 458 Economic and Monetary Union (EMU) countries, 77 Economist, The ( financial magazine), 279 Ecuador, 75, 85, 101, 103, 150, 172 Education labor and, 280–284 public discussion and, 299–302, 312 Edwards, Sebastian, xvi, 80, 121–156 Eichenbaum, Martin, 75 Eichengreen, Barry, 73, 122 Elasticity, 6, 112, 114–115, 159–162 El Salvador, 122, 150 Elster, J., 301 Emerging market economies. See also Transition economies advanced economies and, 72–74 borrowing constraints and, 109–110 capital markets and, 457–460 contagion and, 98–106 credibility and, 95–104, 107–117 crises lessons from, 98–104 crises recovery and, 217–241 financial intermediation and, 251–270 globalization and, 171–213 (see also Globalization) independent currency unions (ICUs) and, 121– 127, 133, 139, 143, 146–150 inflation targeting and, 71–90 institutional development and, 74–77 liability dollarization and, 109
Index
output collapses and, 387–390, 422–425 (see also Output collapses) ownership issues and, 295–321 policy implications and, 114–117 signaling and, 106 stability and, 95 sudden stops and, 98–104 Emerging Market Fund (EMF), 458 Emerging Markets Bond Index (EMBI), 99, 392 Engel, Charles, xiv, 4, 9, 13, 16–17, 123, 125 Enron, 73 Equations baseline crisis specification, 390 capital mobility model, 281–289, 292–293 Cobb-Douglas, 160 covariance matrix, 131 domestic production, 160 duration analysis, 405 empirical growth model, 131–132 enterprise liquidity, 333 equilibrium Fisher condition, 43 Euler, 12, 17, 44 export proximity, 393 fiscal theory of the price level (FTPL), 43–47, 51, 54, 57 flexible-price equilibrium, 8–9 inflation model, 147 Lahiri-Singh-Ve´gh, 23–37 local-currency pricing (LCP), 5, 7 macroeconomic pessimism, 160–165 maximum likelihood function, 405 Obstfeld model, 5, 7–14 pseudo R2 measure, 396 recovery model, 223–225, 241–243 sticky-price equilibrium, 10–14 strong-form globalization, 192 threshold model, 341 transitional credit growth, 332–333, 336, 341 weak-form globalization, 191 Wilcoxon test, 191 worker consumption, 161 Equilibrium, 113, 434, 438, 460 arbitrage condition and, 288 asset price continuity principle and, 60–61 consumption, 28–29 duration analysis and, 387–390, 413, 416 exchange rates and, 16–18 (see also Exchange rates) fiscal theory of the price level (FTPL) and, 43–62 Fischer condition and, 43 flexible-price, 8–9 Kolmogorov-Smirnov test and, 178, 182
Index
Lahiri-Singh-Ve´gh model and, 27–29 leisure time and, 282 local-currency pricing (LCP) and, 4–10 macroeconomic pessimism and, 159–168 median voter, 287–289 money market, 27 multiple, 413, 416 Nash, 18n6 nominal rigidities and, 453–454 Obstfeld model and, 5–18 optimal policy rules and, 14–18 output collapses and, 387–390 Pareto optimality and, 285 pegging and, 55–62 political economy model and, 280–293 preset price, 9–14 recovery and, 225–231, 241–244 reform and, 166–167 steady-state, 49–52, 63 sudden stops and, 110, 112 transversality and, 28 velocity shocks and, 27–28 Equity market issues and, 99, 112–115, 172, 175, 182, 199, 266, 285 open economies and, 112–115 volatility and, 172 Estonia, 176 Euler equations, 12, 17, 44 European Bank for Reconstruction and Development (EBRD), 266, 355, 362, 372 European Union (EU), 280, 290–291, 335 Exchange rates, xiv–xv, 97, 217, 438 asymmetric responses and, 16–17 cash-in-advance constraints and, 24–25, 34–38 crises recovery and, 217–241 Devereux-Engel model and, 4, 13 Dornbusch model and, 223 fixed, 9–14, 30–36 flexible, 3, 8–9, 16–18, 29–33, 35 floating fears and, 73, 80, 85 fluctuation issues and, 84–87 Friedman model and, 3–4 globalization and, 176 (see also Globalization) hard currencies and, 97–117 inflation targeting and, 79–87 institution building and, 362–372 Lahiri-Singh-Ve´gh model and, 23–38 liability dollarization and, 73 local-currency pricing (LCP) and, 4–10 Mundell-Fleming model and, 21–38 Obstfeld model and, 5–18
477
optimal regimes for, 21–38 pegging and, 55–62, 121 (see also Pegging) switching effects and, 3–4 transition economies and, 355–360 Expanded War Data Set (Gleditsch), 391 Fearon, J., 301, 312 Ferna´ndez, Roque B., xiv, xix, 437–444 Fiat money, 41, 45 Financial intermediation asymmetric information and, 249–250 Bank for International Settlements (BIS) and, 251 Calvo and, 250, 272 cash-in-advance (CIA) constraint and, 249 contract enforcement and, 267 default scenario and, 250–251 government debt and, 249–252, 260, 264, 266– 272 International Monetary Fund (IMF) and, 249– 250 movable collateral and, 264, 266 obtaining credit and, 267 production and, 250 property registration and, 267 real estate and, 249–250 stress test approach and, 249 Financial Times, 199 Findlay, Ron, 432, 448–449 Finland, 176, 260 Fiscal Neutrality toward Economic Growth (Phelps), 432 Fiscal theory of the price level (FTPL), 74 adjusted fiscal surplus and, 52–55, 66n14 asset price continuity principle and, 60–61 call bonds and, 43 continuous time analysis and, 42–46, 65n3 equilibrium and, 43–62 Euler equation for, 44 fiat money and, 41, 45 government bonds and, 43–52, 63 growth and, 43–52 inflation and, 43–45, 48–50 instantaneous adjustments and, 59 literature on, 41–42 monetary rule and, 46–55 nominal interest rule and, 55–62 nominal money treatment and, 52–53 once-and-for-all increase and, 50 pegging and, 41–42, 51, 55–63 primary surplus and, 49–50 taxes and, 44 transversality and, 45–54, 58
478
Fischer, Stanley, xviii, 121–122, 349–376, 433 Fisher, I., 114 Fisher condition, 43 Fisher’s principle of randomization, 191 Fisman, R., 339–342 Fitzgerald, T., 41, 51, 53 Fixed effects logit analysis, 403 Fleisig, H. W., 264, 266–267 Fleming, M. J., 21, 270, 460 Flood, Robert, 452 Forder, James, 78 Ford Foundation, 448 Foreign aid, 349–376 Foreign direct investment (FDI), 101, 458 Foundations of Economic Analysis (Samuelson), 447 Fraga, Arminio, 78, 80, 82 Frailties, 408 France, 176, 291 Frankel, J. A., 122–123, 125, 175 French Guyana, 126 Frieden, J., 134 Friedman, Milton, 3 Fuerst, T. S., 51, 53 G7 countries, 176, 178, 192, 200 Galton, F., 379 Game theory heterogeneous public and, 312–315 homogeneous public and, 303–308 public uncertainty and, 308–312 social welfare maximization and, 303–308 special interest groups (SIGs) and, 298–299, 313–315 time consistency and, 449–452 Garber, Peter M., 108 Garro, A. M., 266 Gelb, A., 355 Germany, globalization and, 172, 176–178, 182–183, 186, 193–194, 198, 200 tax harmonization and, 290–291 Ghosh, A., 123 Gini index, 412 Giovanetti, G., 249 Gleditsch, K., 391 Globalization, 202 capital income taxation and, 279–293 central vs. peripheral countries and, 173–175 chronological market demographics for, 204– 213 contagion and, 171–172
Index
deteriorating fundamentals and, 176 determinants and, 176–199 equity and, 199 financial centers and, 173, 175 Fisher’s principle of randomization and, 191 Kolmogorov-Smirnov test and, 178, 182, 187 lemons problem and, 174 liquidity and, 174–175 literature on, 176 long-term capital management (LTCM) and, 171–172, 175, 178, 200, 203 measurement and, 176–199 nongovernmental organizations (NGOs) and, 296 origins of, 199–201 political-economy model for, 280–293 prices and, 174–175, 199 shock transmission and, 173–175, 182–183, 192– 194, 198–199 spillover and, 172–173, 182, 186–187, 200 strong-form, 172–173, 203 systemic risk and, 175–176 trade competition and, 174 weak-form, 172–173, 201 Wilcoxon test and, 187 World War I era and, 279 Goldfajn, Ilan, 78, 82 Gompertz parameter, 412–413, 416–417 Goods market, 224, 241–244 Government. See also Debt asymmetric information and, 249–250 bonds and, 43–52, 63, 76, 268–270, 272, 438, 440–441 central bank independence and, 78, 82 cheap talk and, 300–301, 309–310 crises recovery and, 217–241 deposit rescheduling and, 438–439 fiat money and, 41, 45 financial intermediation and, 249–252, 260, 264, 266–272 fiscal theory of the price level (FTPL) and, 41–64 GKO bills and, 178 Lahiri-Singh-Ve´gh model and, 27 macroeconomic pessimism and, 159–168 nongovernmental organizations (NGOs) and, 296, 316 pegging and, 55–62 promises of, 96 public uncertainty about, 308–312 social welfare maximization and, 303–308 special interest groups (SIGs) and, 298–299 time inconsistency and, 95–96
Index
unanimous action clauses (UACs) and, 440–441 wealth redistribution and, 96 Great Depression, 459 Greece, 176 Greene, W. H., 124, 131–132 Gross domestic product (GDP). See Growth Growth, xiv, xviii, 432, 443, 447. See also Debt common currencies and, 121–124, 130–139 continuous negative, 380 dollarization and, 122, 139–144 empirical model for, 130–132 export flexibility and, 405–422 financial intermediation and, 249–252, 260, 264, 266–272 fiscal theory of the price level (FTPL) and, 43–52 independent currency unions (ICUs) and, 121– 124, 127–129 International Monetary Fund (IMF) and, 87–88, 295–321 liberalization and, 331 outcome equations and, 134–135 output collapses and, 377–425 (see also Output collapses) peak levels in, 377, 381, 384–385 private bond markets and, 268–270 recovery and, 219–241 Russia and, 349 total factor productivity (TFP), 130 transition economies and, 327–345, 349–376 (see also Transition economies) treatment equation and, 133–134 volatility and, 330 (see also Volatility) Guatemala, 122 Guidotti, Pablo, xvii, 454 Guiso, L., 340 Gulati, M., 440 Gulde, A., 123 Gulf War, 400 Gurr, T. R., 392 Gust, C., 218 Habermas, J., 301–302 Hall, R. E., 361 Hansen, B. E., 341 Haque, B., 41 Hausmann, Ricardo, xviii–xix, 73, 377–428 Havrylyshyn, O., 355 Hazard rates Cox proportional model and, 413 developing countries and, 389 duration analysis and, 387–390 frailties and, 408
479
Gompertz parameter and, 412–413, 416–417 Nelson-Aalen cumulative function and, 387–388 nonindustrial countries and, 388 output collapses and, 387–390, 405, 408–425 pooling and, 388–389 regional analysis and, 387–390 Heaton, J., 270 Heckman, J. J., 124 Helbling, T. F., 349–376 Helicopter drops, 45 Helpman, Elhanan, 107, 454 Herding, 105 Herfindahl index, 417 Herring, R. J., 268–270 Heston, Charlton, 431, 435n1 Hicks, J. R., 446–447 Hidehiko, I., 124 Highly indebted poor countries (HIPCs), 320 Holland, 176 Hong Kong globalization and, 174–176, 200, 203 institution substitution and, 99 private bond markets and, 270 Households capital mobility and, 280–284 labor and, 280–284 Lahiri-Singh-Ve´gh model and, 23–38 Human capital. See Labor Hungary, 85, 176 Hwang, J., 393, 417 Hyperinflation, 50, 52–53, 308, 437, 440, 460 Iceland, 338 Independent currency unions (ICUs) analysis of, 125–127 central banks and, 126–127 comparative analysis of, 127–129 dollarization and, 121–122, 124, 139–144, 150 estimator techniques and, 125 growth and, 121–124 interest rates and, 122 macroeconomic performance and, 133 nonparametric analysis and, 146–147 robustness and, 146–149 treatment equation and, 146 India, 270 Indonesia, 176, 200, 217, 272 Industrial Revolution, 448 Inflation, xvi bad dream story and, 52 common currency and, 132, 135, 137 cyclical, 74
480
Inflation (cont.) devaluation and, 62 disinflation and, 80, 433, 454–455 financial intermediation and, 270–272 fiscal imbalances and, 75–76 fiscal theory of the price level (FTPL) and, 43– 45, 48–50 foreign, 23 Friedman model and, 3–4 hyperinflation and, 50, 52–53, 308, 437, 440, 460 independent currency unions (ICUs) and, 147– 148 literature on, 433–435 nominal rigidities and, 454 output collapses and, 396–397, 403 recovery and, 237–241 social welfare and, 308 switching effects and, 3–4 taxes and, 45 time inconsistency and, 95–96 transition economies and, 355, 360 very rapid, 147 Inflation targeting, xv–xvi, 3 accountability and, 71, 81 advanced economies and, 72–74 Brazil and, 79–84 Calvo and, 71–74, 77, 79–80, 85, 87–89 central banks and, 71, 78, 82–84, 87 Chile and, 79–80, 84, 86, 113 constrained discretion and, 79 critics of, 79 currency substitution and, 72–73 disinflation and, 80 emerging market countries and, 71–90 exchange rate fluctuations and, 84–87 fiscal imbalances and, 75–76 five elements of, 71 flexibility and, 78–79 floating fears and, 73, 80, 85 information-exclusive strategy and, 71 institutional development and, 74–79 International Monetary Fund (IMF) and, 87–89 liability dollarization and, 73, 74 prices and, 71 reform and, 77 shocks and, 78–79, 82–83, 86 stability and, 71, 74–79 sudden stops and, 73–75 transparency and, 71, 79 Information costs, 104–105 asymmetric information and, 249–250 cheap talk and, 300–301, 309–310
Index
financial intermediation and, 249–250 policy implications and, 114–117 public discussion and, 288–316 signaling and, 305 social welfare maximization and, 303–308 sudden stops and, 107–114 uncertainty and, 308–312 Institutions. See also Banks building, 362–372 credibility and, 107–114 defined, 360–361 development of, 74–79 formal/informal, 360–361 government debt and, 260–267 inflation targeting and, 74–79 legal issues and, 264–267, 360–362 moral hazard and, 75 nongovernmental organizations (NGOs) and, 296, 316, 361 organizations and, 361, 372 output collapses and, 404 public discussion and, 298–316 rapid privatization and, 349 reform and, 362–372 regression analysis and, 372–373 role of, 360–362 strong fiscal, 74–77 strong monetary, 77–79 substitution and, 98, 107–114 transition economies and, 360–374, 372–373 Institutions Matter (World Bank), 360 Insurance, 96–97, 439 Inter-American Development Bank, xix, 269, 445, 459–460 Interest rates, 80–81, 83, 85 asset price continuity principle and, 60–61 asymmetric responses and, 16–17 crises recovery and, 217 Devereux-Engel model and, 4 dollarization and, 122 globalization and, 176 independent currency unions (ICUs) and, 122 Lahiri-Singh-Ve´gh model and, 26–27 local-currency pricing (LCP) and, 4–10 macroeconomic pessimism and, 160 nominal rule, 55–62 Obstfeld model and, 5–18 parity conditions and, 26–27, 62, 107 pegging and, 41–42, 51 recovery and, 218, 231–234 shock transmission and, 175 sudden stops and, 101
Index
transitional credit growth and, 332–333 verification costs and, 439 International Disaster Database, 391 International financial institutions (IFIs), 349–350. See also Institutions building of, 362–372 role of, 360–362 International Financial Statistics (IFS), 134, 272 International Monetary Fund (IMF), xvii, 437, 452 Brazil and, 81, 84 Calvo and, 437–438, 445, 453–456 cheap talk and, 300–301, 309–310 common currency and, 123 deliberative democracy and, 302 financial intermediation and, 249–250, 268 Guidelines on Conditionality and, 316–317 highly indebted poor countries (HIPCs) and, 320 Independent Evaluation Office (IEO) and, 317– 318 inflation targeting and, 87–89 institution building and, 363, 366–367 nongovernmental organizations (NGOs) and, 296, 316 private bond markets and, 268–269 program ownership and, 295–321 Public Information Notice and, 311 Reinhart and, 456 Research Department and, xiii, xviii social welfare maximization and, 303–308 special interest groups (SIGs) and, 298–299, 313–315 stress test approach and, 249 sudden stops and, 99 transition economies and, 456 transparency and, 317 International Standard Industrial Classification (ISIC), 338 Investment arbitrage condition and, 288–289 capital markets and, 457–460 capital mobility and, 279–293 crises recovery and, 217–241 domestic production and, 160–161 labor and, 279–286 macroeconomic fundamentals and, 225–231 macroeconomic pessimism and, 159–168 market clearing and, 162 program ownership and, 295–321 rational herding and, 105 systemic risk and, 175–176
481
Ireland, 125, 151n10 Isard, Peter, xviii, 295–324, 452 Italy, 176 Izquierdo, Alejandro, 121, 442 inflation targeting and, 73, 77 institution substitution and, 97, 99 output collapses and, 379–380, 392, 400 Jaggers, K., 392 Jamaica, 317 Japan, xvii, 260 financial intermediation and, 252 globalization and, 172–174, 176, 178, 182–183, 186, 193, 198–200 Jazbec, Bostjan, xviii, 327–347 Johnson, S., 361 Jonas, Jiri, 85 Jones, C. I., 361 Journal of International Economics, 457 Journal of Monetary Economics, 453 Journal of Money, Credit and Banking, 450, 454 Journal of Political Economy, 446, 454 J. P. Morgan, 99, 251, 272 J-shaped pattern, 108 Kaminsky, Graciela L., xvi–xvii, 74, 99 Calvo and, 438–439, 443 contract theory and, 439–440 financial intermediation and, 252 globalization and, 171–215 recovery and, 217 Kehoe, Timothy J., 96, 458 Keynes, John Maynard, 18, 433–435, 446 King, Mervyn, 83 Kiser, S. L., 218 Kisselev, Kate, 112 Kiyotaki, N., 337 Klein, J. P., 408 Klein, Michael W., 108 Klingebiel, D., 81, 218 Klinger, B., 392, 424 Kocherlakota, N., 41, 51, 53 Kolmogorov-Smirnov test, 178, 182, 187 Koo, J., 218 Koopmans, T., 432, 447–449 Korea, 74, 99, 106, 176, 200, 217, 260 Kraay, A., 361 Krogstrup, S., 280 Krugman, P., 217, 280 Kuhn-Tucker condition, 34 Kumhof, Michael, xvii, 114, 249–277, 454 Kydland, 449
482
Labor arbitrage condition and, 288–289 capital mobility and, 279–286 contract theory and, 434 credit markets and, 332–333, 337 education and, 280–284 efficiency and, 280–281 elasticity and, 161 hazard rates and, 410 IMF Guidelines and, 317 market clearing and, 162 maximum likelihood function and, 405 output collapses and, 404–405, 410 productivity and, 160–161 (see also Production) skilled, 280–284 taxes and, 279–286 Laeven, L., 218 Lagrangians, 34 Lahiri, Amartya, xv, 21–39 Lahiri-Singh-Ve´gh model cash-in-advance constraints and, 24–25, 34–38 equilibrium and, 27–29 fixed exchange rates and, 30–33 flexible exchange rates and, 29–33 government and, 27 interest rates and, 26–27 intuition and, 36 Kuhn-Tucker condition and, 34 Lagrangians for, 34 output shocks and, 32–34, 37 prices and, 23 traders/nontraders and, 23–27 transversality and, 28 types of shocks and, 27–28 utility maximization and, 25–27 velocity shocks and, 31–32, 37 Lancaster, Kel, 432, 449 Landerretche, Oscar, 80 La Porta, R., 267 Latin America, xiv. See also Specific country crises recovery and, 219–222 financial intermediation and, 266–267 fiscal health of, 217–222 peak growth levels and, 377 Latin American and Caribbean Economic Association (LACEA), 458 Laubach, F., 77, 79 Laxton, Douglas, 72, 77 Layton, A., 381 Learner, E. E., 134 Lee, J. W., 218, 392, 435n1 Leeper, E., 41
Index
Legal issues government debt and, 264–267 institutions and, 360–362 New York governing law and, 440–441 property rights and, 361 unanimous action clauses (UACs) and, 440– 441 Leiderman, Leo, 100–101, 175, 455–456 Leisure time, 282 Lemons problem, 174 Leone, Alfredo M., 88 Lerrick, Adam, 97, 116 Levy-Yeyati, E., 123 Lewis, A., 171 Li, S., 114 Liability dollarization, 73–74, 109 Liberalization, 327, 331 index of privatization (LIP) and, 354 transition economies and, 354, 362–372 Liberia, 125 Lichtenstein, 125 Lin, D. Y., 408 Liquidity, 97, 114–115 credit markets and, 333–338 globalization and, 174–175 output collapses and, 412 (see also Output collapses) private bond markets and, 268–270 transitional credit growth and, 333–334 Loans. See Debt London School of Economics, 432 Long-term capital management (LTCM), 171– 172, 175, 178, 200, 203 Loo-Kung, R., 379 Lopez-de-Silanes, R., 267 Love, I., 339–342 Lucas, Robert E., Jr., 23, 46, 116, 433–434, 460 Lula, 82–84 Lyons, R., 175, 187, 203 McCallum, B. T., 41, 51, 53 McFadden, 396 McLiesh, C., 267, 330 Maddala, G. S., 124, 131 Magendzo, I. Igal, xvi, 121–156 Magic Flute (Mozart), 435 Malaysia, 174, 176, 187, 200, 217 Manakit, P., 218 Manin, B., 312 Mantel, Rolf, 447 Marimon, R., 249 Marion, Nancy P., 108
Index
Markets access issues and, 21–22 asset price continuity principle and, 60–61 asymmetric information and, 249–250 capital mobility and, 279–293 chronological crisis data on, 204–213 common currency effects and, 123 contagion and, 98–106 Council for Mutual Economic Assistance (CMEA) and, 355 crashes and, 176 credibility and, 95–104 credit, 327–345 crises recovery and, 217–241 decentralized, 335 degree of integration and, 182–183 derivative, 268–270 double mismatch and, 217 emerging countries and, 71–90 equilibrium and, 25–27 (see also Equilibrium) equity and, 99, 112–115, 172, 175, 182, 199, 266, 285 export flexibility and, 379, 392–394, 400, 403– 422 financial intermediation and, 249–252, 260, 264, 266–272 Gini index and, 412 globalization and, 171–215 (see also Globalization) goods vs. assets, 22 hard currencies and, 97–117 Herfindahl index and, 417 imperfection in goods, 21–22 incomplete insurance, 96–97 inflation targeting and, 71–90 Lahiri-Singh-Ve´gh model and, 23–38 liability dollarization and, 73–74 liberalization and, 327, 331 local-currency pricing (LCP) and, 4–7 macroeconomic pessimism and, 159–168 Mundell-Fleming model and, 21–38 Obstfeld model and, 5–18 output collapses and, 377–425 (see also Output collapses) pegging and, 55–62 prices and, xvi (see also Prices) private bond, 268–270 segmented asset, 22–38 stability mechanisms and, 55–62 strong-form, 191–199 sudden stops and, 73–74 (see also Sudden stops)
483
transition economies and, 349–376 (see also Transition economies) type of distortion and, 22 weak-form, 178–191 Marshall, M. G., 392 Martinique, 126 Marty, Alvin, 450 Masson, P., 77 Masten, Igor, xviii, 327–347 Mathematics, 446, 448, 452. See also Equations Mathematics for Economists (Allen), 446 Meade, J. E., 122, 130 Mejı´a, L.-F., 379, 392, 400 Meltzer, Allan H., 97, 116 Mendoza, Enrique G., xvi, xix, 79 Calvo interview and, 445–461 institutions and, 95–119 Menze, J., 266 Mercosur countries, 187 Merrill Lynch, 272 Metzler, L. A., 41 Mexico, 75, 110, 113 credibility and, 108 crisis of, xvi, 85, 96, 98–99, 103 financial intermediation and, 252 globalization and, 173, 175–176, 203 macroeconomic pessimism and, 159 markets after 1994 crash and, 457–460 Tequila Effect and, 99, 187, 441, 457–460 wealth redistribution and, 96 Milesi-Ferretti, Gian Maria, 99 Mill, John Stuart, 121 Mills, T. C., 381 Minella, Andre´, 78, 82 Minoui, C., 380, 384 Mishkin, Frederic S., xv, 71–93 Mitra, Pritha, xvii, 217–246 Mody, A., 349–376 Monaco, 122, 125 Monetary policy. See Policy Money market, 224, 242–244 Moody’s, 200 Moore, J., 337 Mora, R., 381 Moral hazard common currency and, 131–132 institutional development and, 75 output collapses and, 408–409, 424–425 Morande´, Felipe, 80 Morocco, 317 Mozart, Wolfgang Amadeus, 435 Mudambi, R., 381
484
Mundell, Robert, 21, 133, 150, 449, 460 Mundell-Fleming model, xiv–xv Calvo and, 21 dollarization and, 21 flexible exchange rates and, 22 imperfection in goods markets and, 21–22 Lahiri-Singh-Ve´gh model and, 23–38 open economy and, 21–28 perfect mobility and, 21 shocks and, 22 sticky prices and, 21 traders/nontraders and, 22 Mussa, Michael, 87 Nash equilibrium, 18n6 National Bank of Hungary, 85 National Monetary Council, 81 National Science Foundation, 454 Natural disasters, 391 Nelson-Aalen cumulative hazard function, 387– 388 Neumeyer, P., 112 Neut, Alejandro, xvi, 159–169 New Keynesian theory, 433 New York, 440–441, 453 New York Times, 289, 457 Nicolini, J. P., 249 Niepelt, D., 41 Nobel Prize, xiii, xix, 116 Nongovernmental organizations (NGOs), 296, 316, 361 Nonparametric analysis, 145–147 Non-Ricardian policy, 47, 52 Norway, 176, 338 Null hypothesis, 178, 182, 187, 191 Oates, W. E., 279–280 Obstfeld, Maurice, xiv–xv, 279 Calvo and, 449, 452–453 exchange rates and, 3–19 fiscal theory of the price level (FTPL) and, 51, 53 Occam’s razor, 417 Olivera, Julio H. G., 446–447 Open economies crises recovery and, 217–241 government debt and, 251–270 institutional development and, 74–79 Lahiri-Singh-Ve´gh model and, 23–38 macroeconomic pessimism and, 159–168 Mundell-Fleming model and, 21–28 ownership issues and, 295–321 Ordinary least squares (OLS) analysis, 148
Index
Organisation for Economic Co-operation and Development (OECD), 125, 289 Ostry, J., 123 Outcome equations, 134–135, 147–148 Output collapses, 32–34, 37 big, 385–390 causes of, 390–405, 424 comparative analysis and, 379 convergence hypothesis and, 391 democracy and, 394, 410, 412 duration analysis and, 378–390, 394–422, 424 equilibrium and, 387–390 event definition and, 381–385 export flexibility and, 379, 392–394, 400, 403– 422 fixed effects logit specification and, 403 frailties and, 408 Gini index and, 412 goodness-of-fit indicators and, 396 hazard rates and, 387–390, 405, 408–425 Herfindahl index and, 417 human capital and, 404–405, 410 inflation and, 396–397, 403 institutions and, 404 literature on, 379–381 long, 385–390 maximum likelihood function and, 405 natural disasters and, 391 Nelson-Aalen cumulative function and, 387– 388 peak-trough ratios and, 377, 385 political changes and, 400, 403 postwar data and, 380 predicting, 392, 394–405 Prentice-Williams-Peterson model and, 413, 418–421 pseudo R2 measure and, 396 recovery from, 405–422 regional effects and, 378–379, 385–390 regression analysis and, 390–405, 408–422 short, 385–390 small, 385–390 statistical approach and, 379, 390–405 sudden stops and, 378, 392, 394 time-specific effects and, 378–381, 394–405 unexploited product space and, 392–394 unfinished episodes and, 380–381 war and, 378, 380, 387, 390–408 Weibull specification and, 410 World Development Indicators (WDI) and, 377 Overlapping generations model, 446 Oviedo, P., 112
Index
Ownership cheap talk and, 300–301, 309–310 concept of, 295–299 creation of, 295–296 defined, 295 heterogeneous public and, 312–315 highly indebted poor countries (HIPCs) and, 320 homogenous public and, 303–308 International Monetary Fund (IMF) and, 295– 321 nongovernmental organizations (NGOs) and, 296, 316 public discussion and, 299–316 special interest groups (SIGs) and, 298–299, 313–315 World Bank and, 296–297, 318 Paasche, B., 112 Pakistan, 317 Panama, 122, 125, 130 Panizza, Ugo, 73 Papell, D. H., 379–380 Pareto optimality, 285 Park, Y. C., 218 Parkin, M., 433 Parrado, E., 122 Parsley, David, 99 Pegging, 63 asset price continuity principle and, 60–61 Calvo and, 121 CFA franc and, 127 Chile and, 80 contagion and, 98–106 credibility and, 97–104, 107–114 dollarization and, 121–122 equilibrium and, 55–62 fiscal theory of the price level (FTPL) and, 41– 42, 51, 55–62 hard vs. soft, 121 independent currency unions (ICUs) and, 121– 129, 133, 139, 143, 146–150 interest rates and, 41 nominal rule, 55–62 nonparametric analysis and, 145–147 sudden stops and, 98–114 transition economies and, 355, 360 volatility and, 137, 139, 143, 146–147, 152n17 Penn, 452–453 Perri, Fabrizio, 112 Perry, G., 260, 264 Peru, 176 Pesenti, P., 174
485
Petersen, M., 340 Peterson, A. V., 413 Phelan, C., 41, 51, 53 Phelps, Edmund S., xiii, xix, 460 on Calvo, 431–435 Columbia University and, 431–435, 447, 449–450 growth theory and, 447 nominal rigidities and, 453 Phelps-Pollak paper, 434 Philippines, 176, 200, 260, 270, 317 Poland, 176, 260, 270 Policy, xiv, 455 Brady Plan and, 439, 456 Calvo’s influence on, xiii–xix, 3, 433–435, 437– 443, 445 capital mobility and, 286–287 central bank independence and, 78, 82–84, 450– 451 competition, 362–372 contract theory and, 434, 438–440 credibility and, 95–104, 107–114 (see also Credibility) crises recovery and, 217–241 debt forgiveness and, 166 debt rescheduling and, 440–442 deposit rescheduling and, 438–439 emerging vs. advanced economies and, 72–74 exchange rates and, xiv (see also Exchange rates) expansionary, 166 fiat money and, 41, 45 financial intermediation and, 249–252, 260, 264, 266–272 fiscal theory of the price level (FTPL) and, 41–64 floating fears and, 73 foreign aid and, 349–376 future and, 460–461 hard currencies and, 97–117 IMF and, 87–89 (see also International Monetary Fund (IMF)) inflation targeting, 71–90 institutional development and, 74–79 Lahiri-Singh-Ve´gh model and, 23–38 legitimacy and, 305 liability dollarization and, 73–74, 109 (see also Dollarization) liberalization, 327, 331, 354, 362–372 macroeconomic pessimism and, 159–168 non-Ricardian, 47, 52 optimal rules for, 14–18 pegging and, 41, 55–63 (see also Pegging) privatization, 96, 349, 354, 362–372 procyclicality and, 15
486
Policy (cont.) program ownership and, 295–321 public discussion and, 288–316 recovery and, 217–241 reform and, 166–167 (see also Reform) Ricardian, 54–55, 65n11, 107 social welfare and, 303–308 time inconsistency and, 95–96 transition economies and, 349–376 (see also Transition economies) transparency and, 79, 317 unanimous action clauses (UACs) and, 440–441 verification costs and, 439 welfare analysis and, 14–16 Political economy budget constraints and, 291–293 equilibrium derivation and, 291–293 globalization and, 279–293 model for, 280–293 Politics, 451 capital mobility and, 279–289 Cold War and, 279 democracy and, 301–302, 315, 394, 410, 412 globalization and, 279 output collapses and, 400, 403 program design and, 302 special interest groups (SIGs) and, 298–299, 313 voters and, 286–289, 302 Ponzi games, 109 Portugal, 260 Posen, A., 77, 79 Pou, Pedro, 76, 78 Powell, Andrew, 75, 124 Prebisch, Rau´l, 446 Prentice, R. L., 413 Prescott, Edward, 449–450 Pesenti, P., 217 Prices, xiv–xvi, 97, 115–116 asset price continuity principle and, 60–61 credibility and, 452–457 crises recovery and, 223–225 Devereux-Engel model and, 4, 13, 17 exchange rates and, 3 (see also Exchange rates) fiat money and, 41, 45 fiscal theory of the price level (FTPL) and, 41–64 Friedman model and, 3–4 globalization and, 174–175, 199 inflation targeting and, 71 instantaneous adjustments and, 59 Lahiri-Singh-Ve´gh model and, 23 lemons problem and, 174 local-currency (LCP), 4–10
Index
macroeconomic pessimism and, 159–168 Mundell-Fleming model and, 21–38 Obstfeld model and, 5–18 sticky, 21, 249, 433, 445, 452–457 sudden stops and, 110, 112 (see also Sudden stops) switching effects and, 3–4 Principles of Political Economy (Mill), 121 Pritchett, L., 379 Privatization reform and, 96 too-rapid, 349 transition economies and, 354, 362–372 Production capital mobility and, 284–286 Cobb-Douglas function and, 160 constant-returns-to-scale function and, 284–285 export flexibility and, 379, 392–394, 400, 403–412 financial intermediation and, 250 Herfindahl index and, 417 local-currency pricing (LCP) and, 4–7 macroeconomic pessimism and, 159–168 market clearing and, 162 Obstfeld model and, 5–18 output collapses and, 377–425 (see also Output collapses) Pareto optimality and, 285 sudden stops and, 110, 112 (see also Sudden stops) unexploited product space and, 392–394 Property rights, 361 Przeworski, A., 305, 312, 315 Pseudo R2 measure, 396 Public discussion bias and, 300 cheap talk and, 300–301, 309–310 common ground and, 300–301 constraints and, 299 deliberative democracy and, 301–302, 315 drawbacks of, 315–316 education and, 312 functions of, 303–315 fuzzy thinking and, 296–297 heterogeneous public and, 312–315 homogenous public and, 303–308 institutional capacity and, 299 need for, 296–297 nongovernmental organizations (NGOs) and, 296, 316 program design and, 302–303 public uncertainty and, 308–312 reform and, 298–299
Index
signaling and, 305 social welfare maximization and, 303–308 special interest groups (SIGs) and, 298–299, 313–315 understanding role of, 299–302 voting and, 302–303 Public Information Notice, 311 Puerto Rico, 125 Purchasing power parity (PPP), 384 Rabanal, Pau, 88 Race-to-the-bottom prediction, 280 Rajan, R., 338–343 Rank-sum test, 187 Rational expectations (RE), 460–461 Rawlsian maximin principle, 450 Razin, Assaf, xvii Calvo and, 449, 450, 454 globalization and, 279–294 institutions substitution and, 99, 101, 107 Real estate, 249–250 Rebelo, Sergio, 75, 107 Recession, xviii–xix, 108. See also Output collapses inflation targeting and, 75, 86 Keynesian, 434 macroeconomic pessimism, 159–168 Recovery aggregate supply and, 224–227 asset market and, 223–224, 241–244 corporate governance and, 218 datasets for, 226–229 depreciation rates and, 223 Dornbusch model and, 223 double mismatch and, 217 equilibrium and, 225–231, 241–244 exchange rate depreciation and, 234–236 goods market and, 224, 241–244 growth rates and, 219–241 inflation and, 237–238, 241 interest rates and, 231–234 literature on, 217–218, 244n1 macroeconomic fundamentals and, 219–231 model for, 223–244 money market and, 224, 241–244 output collapses and, 405–422 prices and, 223–225 private sector and, 218 rapid Asian, 217–218 simulations of, 225–226, 229–231 slow Latin American, 218 Reddy, S. G., 380, 384
487
Reform, 443 credibility and, 95–97, 107–114 credit markets and, 327, 335–338 inflation targeting and, 77 institutions and, 362–372 macroeconomic pessimism and, 166–167 public discussion and, 298–299 special interest groups (SIGs) and, 298–299 transition economies and, 362–372 Regression analysis credit markets and, 339–345 exponential parameter and, 412–413, 416–417 fixed effects logit and, 403 Gompertz parameter and, 412–413, 416–417 hazard rates and, 408–422 Herfindahl index and, 417 institutions and, 372–373 log-logistic paramter and, 412–413, 416–417 log-normal parameter and, 412–413, 416–417 output collapses and, 390–422 Prentice-Williams-Peterson model and, 413, 418–421 transition economies and, 372–373 Weibull specification and, 410 Reinhart, Carmen, xvi–xvii, 74, 454 Calvo papers and, 455 crises recovery and, 217 financial intermediation and, 252, 264, 270 globalization and, 171–215 inflation targeting and, 73, 80, 85 International Monetary Fund (IMF) and, 456 private bond markets and, 260 sudden stops and, 99–101 switching effects and, 3–4 Reinstein, A., 269 Rhee, C., 269 Ricardian policy, 54–55, 65n11, 107 Rigobon, Roberto, 99 Risk. See also Hazard rates; Investment credibility and, 97–104, 107–114 Mendoza-Uribe model and, 107–108 systemic, 175–176 Robinson, J., 361 Rodriguez, Carlos, 449, 450 Rodriguez, Francisco, xviii–xix, 377–428 Rodrik, D., 217, 361, 393, 410, 412, 417 Rogoff, K., 51, 53, 264 Roland, G., 330 Roldos, J., 218 Romania, 338 Rose, A. K., 122–123, 125 Rose, Susan, 432
488
Rothschild, Michael, 445–446 Roubini, Nouriel, 112, 174, 217 Rudebusch, G. D., 381 Russia, 106, 114, 338 bond default of, 174 credit markets and, 331 crisis of, 73–74, 80, 99, 101, 103, 174, 458 double mismatch and, 217 financial intermediation and, 252 GKO government bills and, 178 globalization and, 172, 176, 178, 183, 186–187, 192, 198, 200 growth in, 349 rapid privatization and, 349 sudden stops and, 101, 103 transition economies and, 349–376 (see also Transition economies) Russian virus, 174 Sachs, J., 134 Sachs-Warner index, 134 Sack, B., 270 Sadka, Efraim, xvii, 101, 279–294 Sahay, Ratna, xviii, 349–376 Sala-i-Martin, X., 134 Salomon Brothers, 175 Samuelson, Paul, 446–447 Sargent, T., 41, 49, 74, 460 Savastano, Miguel, 75, 80–81, 87, 264 Savastono, M., 77 Saxena, S. C., 379 Scarf, Herb, 447 Schmidt-Hebbel, Klaus, 72, 80 Schmitt-Grohe, S., 249 Schmukler, S., 175, 187, 203 Schwartz, Gerd, 88 Seignorage, 51, 107–108, 116, 121, 123, 139 Senegal, 317–318 Serven, L., 264 Sharma, S., 77 Shleifer, A., 267, 330 Shocks, xv, xvii asymmetric responses and, 16–17 credit markets and, 336–337 crises recovery and, 217–241 Dornbusch model and, 223 globalization and, 171–175, 182–183, 192–194, 198–199 inflation targeting and, 78–79, 82–83, 86 information costs and, 105 local-currency pricing (LCP) and, 5 macroeconomic aggregates and, 110
Index
macroeconomic pessimism and, 159–168 Mundell-Fleming model and, 22 Obstfeld model and, 9 optimal policy rules and, 14–18 output, 32–34, 37 (see also Output collapses) permanent, 18n7 signaling and, 106 sudden stops and, 110, 112 (see also Sudden stops) technology, 9, 13, 15–16 types of, 22–23 velocity, 27–28, 31–32, 37 world liquidity, 114 Sichel, D. E., 381 Sidrauski, Miguel, 447 Signaling, 106 Silk road, 203 Simpson, J., 266 Sims, C., 41 Singapore, 174–176, 200 Singh, Rajesh, xv, 21–39 Siotis, G., 381 Siu, H. E., 249 Slovakia, 176 Smith, Gregor W., 7 Smith, Katherine A., 97, 109, 112 Sobel, J., 309 Social welfare, 303–308 South Africa, 73 Spain, 176, 200 Special interest groups (SIGs), 298–299, 313–315 Spillover, 172–173, 182, 186–187, 200 Stability Brazil and, 79–84 Chile and, 79–80, 84 credit markets and, 327 emerging market economies and, 95 hard currency and, 98 inflation targeting and, 71, 74–84 institutional development and, 74–79 Mendoza-Uribe model and, 108 rational herding and, 105 recovery and, 217–241 Tablita plan and, 453 time inconsistency and, 95–96 Standard & Poor’s, 200 Stanford University, 432 Stiglitz, Joseph, 447 Stokes, S. C., 312, 315 Stress test approach, 249 Sturzenegger, F., 123–124 Subramanian, A., 361
Index
Sudden stops, 73–75, 443, 445 causes of, 100–101 contagion and, 98–106 credibility and, 97–104 equilibrium and, 110, 112 foreign direct investment and, 101 information costs and, 107–114 institution substitution and, 107–114 interest rates and, 101 International Monetary Fund (IMF) and, 99 liability dollarization and, 109 output collapses and, 378, 392, 394 rational herding and, 105 Tequila Effect and, 99 wealth effects and, 107 Summers, L. H., 121 Svensson, Lars O., 79 Swagel, 280 Sweden, 176 Swift, Jonathan, 171 Switzerland, 260 Tablita stabilization plan, 453 Talvi, Ernesto, 442 hard currency pegs and, 121 inflation targeting and, 73, 77 institution substitution and, 97, 99 output collapses and, 379–380, 392 Tamiroff, Akim, 435n1 Tanner, Evan, xvii, 249–277 Taxes, xiv, xvii capital income taxation and, 279–293 European Union and, 280, 290–291 evasion of, 125–126 financial intermediation and, 249–250, 273n3 fiscal theory of the price level (FTPL) and, 44 harmonization and, 290–291 inflation, 45 institution building and, 362–372 international competition and, 279–280, 289– 291 labor and, 279–286 Mexico and, 110 political-economy model for, 280–293 race-to-the-bottom prediction and, 280 social welfare maximization and, 307–308 stochastic distortion and, 97 Taylor, A. M., 279 Taylor, John, 432–434, 449, 453 Taylor, L. W., 381 Technology, 9, 13, 15–16 Teles, P., 249
489
Tenev, S., 355 Tequila crisis, 99, 187, 441, 457–460 Terrorism, 159 Thaicharoen, Y., 361 Thailand, xvii double mismatch and, 217 financial intermediation and, 252 globalization and, 172–176, 183, 186–187, 192– 193, 198–199 recovery and, 218–219, 221, 226–241 Theory of Value (Debreu), 447 Three Essays on the State of Economic Science (Koopmans), 447 Tille, C., 174 Time inconsistency, 449–452 Tobin, James, 41, 431–432, 435, 447 Todd, P. E., 124 Total factor productivity (TFP) growth, 130 Touch of Evil, A (film), 431 Townsend, R. M., 439 Trade credit, 335–338 Transition economies, 456 Council for Mutual Economic Assistance (CMEA) and, 355 credit markets and, 327–345 empirical analysis of, 338–345, 353–354 European Bank for Reconstruction and Development (EBRD) and, 355, 362, 372 financial sector development and, 328–332 foreign aid and, 349–376 geography and, 354–355 inflation and, 355, 360 initial condition effects and, 354–355 institutional roles and, 360–374 liberalization and, 354, 362–372 output growth and, 332–338, 350–353 policy performance and, 355–360 privatization and, 354, 362–372 reform and, 362–372 regression analysis of, 372–373 scale of, 349 Soviet bloc and, 349 Transparency inflation targeting and, 71, 79, 82–83 International Monetary Fund (IMF) and, 317 volatility and, 137 Transversality, exchange rates and, 28 fiscal theory of the price level (FTPL) and, 45– 54, 58 Treatment effects model, xvi, 133–134, 146 Trebbi, F., 361
490
Triple threshold model, 341–342 Turkey, 75, 85, 99, 176, 260 Uganda, 307–308 Ukraine, 176, 338 Unanimous action clauses (UACs), 440–441 United Kingdom, 150, 176 United States central bank independence and, 78 credit markets and, 339 globalization and, 172, 176, 178, 182–183, 186, 198, 203 government debt and, 251 inflation and, 130 interest rate decline in, 101, 103 movable collateral and, 266 shock transmission and, 175 United States Agency for International Development (USAID), 447 University of Buenos Aires, 447 University of California, 445 University of CEMA, 437 University of Chicago, 437 University of Maryland, 457, 459 Uribe, Martin, 107, 109, 112, 249 U.S. Foreign Disaster Assistance (OFDA), 391 Utilitarianism, 450 Utility capital mobility and, 282 credibility and, 97–104, 107–114 Lahiri-Singh-Ve´gh model and, 25–27 political-economy model and, 280–293 transitional credit growth and, 332–335 Uzawa, Hirofumi, 448–449 Valderrama, D., 112 Vales, Rodrigo O., 80 Value and Capital (Hicks), 446–447 van Rijckenghem, C., 183 van Rooden, R., 355 Van Wincoop, E., 122 ‘‘Varieties of Capital Market Crises’’ (Calvo), 457 Ve´gh, Carlos A., xv, 107 Calvo and, 454–455 exchange rates and, 21–39 financial intermediation and, 250, 252 growth in transition economies and, 353–355 inflation targeting and, 72, 74 pegging and, 41, 122 recovery and, 217 Velasco, Andre´s, 21, 96, 112, 159–169, 458
Index
Velocity shocks, 27–28, 31–32, 37 Venezuela, 176, 182, 200, 385, 387 Verification costs, 439 Vietnam, 308, 448 Vishny, R., 267 Vladkova-Hollar, I., 328 Volatility, 124, 130 asset prices and, 159–168 common currency and, 137, 139 credit markets and, 330, 337 openness and, 137 pegging and, 137, 139, 143, 146–147, 152n17 strict dollarization and, 143 Voters, 286–289, 302 Wagner, Rodrigo, xviii–xix, 377–428 Wallace, N., 41, 49, 74, 460 Wall Street, 73, 174 Wall Street Journal, 199 Walrasian system, 446 Wars, 448 output collapses and, 378, 380, 387, 390–408 World War I era and, 279 World War II era and, 279 Wealth effects, 107 Weder, B., 183 Wei, L. J., 408 Weibull specification, 410 Welfare maximization of, 303–308 optimal policy rules and, 14–16 public discussion and, 303–308 Welles, Orson, 431, 435n1 Wellisz, Stan, 449, 452 Werner, Alex, 457 Wilcoxon test, 187 Williams, B. J., 413 Wolf, H., 123 Woodford, M., 41, 52–55, 74 Wooldridge, J. M., 124, 396 World Bank, xviii, 134, 266–267 credit markets and, 338 duration analysis and, 384 export data and, 392 institutions’ role and, 360–361 ownership and, 296–297, 318 private bond markets and, 268–269 social welfare maximization and, 307–308 transition economies and, 372 World Development Indicators (WDI), 134, 338, 377 World Pension Association, 264
Index
World War I era, 279 World War II era, 279 Yale University, 431–432, 447–448, 460 Yan, I., 114 Yue, Z. V., 112 Yuen, Chi-Wa, 101 Zhang, M. J., 408 Zingales, L., 338–343
491
