Privatization and Its Limits in Central and Eastern Europe: Property Rights in Transition (Studies in Economic Transition)

Monetary Policy in Transition Inflation Nexus Money Supply in Postcommunist Russia

' Milan Nikolic

Studies in Economic Transition General Editors: Jens Hölscher, Reader in Economics, University of Brighton; and Horst Tomann, Professor of Economics, Free University Berlin This series has been established in response to a growing demand for a greater understanding of the transformation of economic systems. It brings together theoretical and empirical studies on economic transition and economic development. The postcommunist transition from planned to market economies is one of the main areas of applied theory because in this field the most dramatic examples of change and economic dynamics can be found. The series aims to contribute to the understanding of specific major economic changes as well as to advance the theory of economic development. The implications of economic policy is a major point of focus. Titles include: Lucian Cernat EUROPEANIZATION, VARIETIES OF CAPITALISM AND ECONOMIC PERFORMANCE IN CENTRAL AND EASTERN EUROPE Irwin Collier, Herwig Roggemann, Oliver Scholz and Horst Tomann (editors) WELFARE STATES IN TRANSITION East and West Bruno Dallago (editor) TRANSFORMATION AND EUROPEAN INTEGRATION The Local Dimension Hella Engerer PRIVATIZATION AND ITS LIMITS IN CENTRAL AND EASTERN EUROPE Property Rights in Transition Hubert Gabrisch and Rüdiger Pohl (editors) EU ENLARGEMENT AND ITS MACROECONOMIC EFFECTS IN EASTERN EUROPE Currencies, Prices, Investment and Competitiveness Oleh Havrylyshyn DIVERGENT PATHS IN POST-COMMUNIST TRANSFORMATION Capitalism for All or Capitalism for the Few? Jens Hölscher and Anja Hochberg (editors) FINANCIAL TURBULENCE AND CAPITAL MARKETS IN TRANSITION COUNTRIES Jens Hölscher and Anja Hochberg (editors) EAST GERMANY’S ECONOMIC DEVELOPMENT SINCE UNIFICATION Domestic and Global Aspects Mihaela Kelemen and Monika Kostera (editors) CRITICAL MANAGEMENT RESEARCH IN EASTERN EUROPE Managing the Transition

Emil J. Kirchner (editor) DECENTRALIZATION AND TRANSITION IN THE VISEGRAD Poland, Hungary, the Czech Republic and Slovakia Tomasz Mickiewicz ECONOMIC TRANSITION IN CENTRAL EUROPE AND THE COMMONWEALTH OF INDEPENDENT STATES Milan Nikolic MONETARY POLICY IN TRANSITION Inflation Nexus Money Supply in Postcommunist Russia Julie Pellegrin THE POLITICAL ECONOMY OF COMPETITIVENESS IN AN ENLARGED EUROPE Stanislav Poloucek (editor) REFORMING THE FINANCIAL SECTOR IN CENTRAL EUROPEAN COUNTRIES Gregg S. Robins BANKING IN TRANSITION East Germany after Unification Johannes Stephan ECONOMIC TRANSITION IN HUNGARY AND EAST GERMANY Gradualism and Shock Therapy in Catch-up Development Johannes Stephan (editor) TECHNOLOGY TRANSFER VIA FOREIGN DIRECT INVESTMENT IN CENTRAL AND EASTERN EUROPE Hans van Zon THE POLITICAL ECONOMY OF INDEPENDENT UKRAINE Adalbert Winkler (editor) FINANCIAL DEVELOPMENT IN EASTERN EUROPE The First Ten Years

Studies in Economic Transition Series Standing Order ISBN 0–333–73353–3 (outside North America only) You can receive future titles in this series as they are published by placing a standing order. Please contact your bookseller or, in case of difficulty, write to us at the address below with your name and address, the title of the series and the ISBN quoted above. Customer Services Department, Macmillan Distribution Ltd, Houndmills, Basingstoke, Hampshire RG21 6XS, England

Monetary Policy in Transition Inflation Nexus Money Supply in Postcommunist Russia Milan Nikolic

© Milan Nikolic´ 2006 All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission. No paragraph of this publication may be reproduced, copied or transmitted save with written permission or in accordance with the provisions of the Copyright, Designs and Patents Act 1988, or under the terms of any licence permitting limited copying issued by the Copyright Licensing Agency, 90 Tottenham Court Road, London W1T 4LP. Any person who does any unauthorized act in relation to this publication may be liable to criminal prosecution and civil claims for damages. The author has asserted his right to be identified as the author of this work in accordance with the Copyright, Designs and Patents Act 1988. First published in 2006 by PALGRAVE MACMILLAN Houndmills, Basingstoke, Hampshire RG21 6XS and 175 Fifth Avenue, New York, N.Y. 10010 Companies and representatives throughout the world. PALGRAVE MACMILLAN is the global academic imprint of the Palgrave Macmillan division of St. Martin’s Press, LLC and of Palgrave Macmillan Ltd. Macmillan® is a registered trademark in the United States, United Kingdom and other countries. Palgrave is a registered trademark in the European Union and other countries. ISBN-13: 978–1–4039–8743–3 ISBN-10: 1–4039–8743–2 This book is printed on paper suitable for recycling and made from fully managed and sustained forest sources. A catalogue record for this book is available from the British Library. Library of Congress Cataloging-in-Publication Data Nikoli´ c , Milan, 1962– Monetary policy in transition : inflation nexus money supply in postcommunist Russia / by Milan Nikoli´ c. p. cm. — (Studies in economic transition) Includes bibliographical references and index. ISBN 1–4039–8743–2 (cloth) 1. Monetary policy – Russia (Federation) 2. Inflation (Finance) – Russia (Federation) 3. Money supply – Russia (Federation) I. Title: Inflation nexus money supply in postcommunist Russia. II. Title. III. Series. HG1080.2.N54 2006 339.530947—dc22 2005044648 10 9 8 7 6 5 4 3 2 1 15 14 13 12 11 10 09 08 07 06 Printed and bound in Great Britain by Antony Rowe Ltd, Chippenham and Eastbourne

Contents List of Tables

vii

List of Figures

ix

Acknowledgements

x

Introduction

xi

1 The Emergence of Open Inflation and Stabilization Efforts in Postcommunist Russia Introduction The monetary roots of inflation in postcommunist Russia The fiscal theory of price level and its application to Russia Frameworks for analyzing the impact of monetary and fiscal policies on macroeconomic stabilization Interaction of fiscal and monetary policies during transition 2 Inflation Expectations in Postcommunist Russia Introduction A brief overview of the literature on inflation expectations in postcommunist Russia and early reactions to open inflation Methodology and model specification Results Alternative modelling Forecasting inflation Rationality of expectations of inflation Policy implications and further research Conclusion 3 The Relationship between Money Supply and Inflation in Postcommunist Russia Introduction A brief overview of Russian monetary policy and the related literature Data and methodological overview Results

v

1 1 2 11 13 17 23 23

24 29 32 39 42 45 49 50 52 52 53 55 59

vi

Contents

Dividing the sample Conclusion 4 Money–Inflation Causality in Transition Economies: The Case of Russia Introduction Data and methodological overview The Wiener–Granger causality model and the lag-length selection criteria Empirical results Policy considerations Conclusion

68 74 76 76 78 81 83 92 96

5 The Role of International Financial Institutions in Stabilizing the Russian Economy Introduction The role of the IMF in the global financial architecture The Washington Consensus The IMF’s involvement in Russia: general policy considerations Quantitative indicators of capital flows and the IMF’s involvement in the European transition economies and the Russian Federation Common criticisms of the IMF programmes Specific policy considerations of the IMF’s involvement in postcommunist Russia Alternative medicine: more prescriptions, fewer injections Conclusion

119 120

Conclusions and Directions for Further Research

122

Appendix

126

Notes

129

Bibliography

139

Index

148

97 97 98 99 100

102 114 115

List of Tables 1.1 1.2 1.3 1.4 1.5 1.6 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10 2.11 2.12 2.13 2.14 3.1 3.2 3.3 3.4

OLS estimates of the distributed lag model of inflation, Russia, February 1992 to September 1994 CBR credit flows, 1992–95 Russian budget deficit, 1992–98 Seignorage and inflation tax on monetary aggregates, Russia, 1991–98 Budget balance, USSR, 1985–90 Simple empirical assessment of the non-interest budget constraint, Russia, 1991–98 Lag length selection for the autoregressive distributed lag model of inflation OLS estimates of the autoregressive distributed lag model of inflation, February 1992 to July 1998 Breusch–Godfrey serial correlation LaGrange Multiplier test Ljung–Box Q-statistics for the squared residuals Normality test for the autoregressive distributed lag model Choice of ARCH/GARCH model Maximum likelihood estimate of inflation, ARCH(1,0) ARMA(6,1) model of inflation ARMA(6,1) estimate of inflation with the structural break included Unit root tests for 1 and 2 Wald test of the coefficient restrictions: (c2, )  (0,1) Tests for zero correlation and non-significant zero mean Test for lack of serial correlation and non-significant zero mean OLS estimates of the strong form efficiency test, February 1992 to July 1998 Unit root tests, February 1992 to July 1998 Seasonal dummies and optimum lag length (l*) selection OLS estimates of the distributed lag model of inflation, in first differences, February 1992 to July 1998 Maximum likelihood estimates of the distributed lag model of inflation vii

4 7 8 9 17 21 32 33 33 34 34 35 35 40 41 47 47 48 48 48 57 57 60 61

viii List of Tables

3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 3.13 3.14 4.1 4.2 4.3 4.4 4.5 4.6 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9

Q-statistics for the standardized and squared residuals OLS estimates of the autoregressive distributed lag model of inflation Maximum likelihood estimates of the autoregressive distributed lag model of inflation Q-statistics for the standardized and squared residuals Scaled recursive Chow test (SRCT) and one-step forecast stability test (O-SFT) for the period January–October 1994 ARCH(1,0) autoregressive distributed lag model of inflation involving m2x OLS of the first differences of the distributed lag model of inflation in the sub-periods Q-statistics for the standardized and squared residuals Maximum likelihood estimation of the distributed lag model of inflation in the sub-periods Q-statistics for the standardized and squared residuals Unit root tests; February 1992 to July 1998 Lag lengths selected by means of the FPE and the Bayesian estimation criteria Granger causality F-statistics for the lags specified by various criteria F-tests of the FPE lag specifications Significance levels derived from Granger causality tests of  and m2 Significance levels derived from Granger causality tests of  and m2x Net capital flows into East European transition economies, by type of flow, 1993–1998 Net capital flows to the CETE-5 and Russia, by type of finance, 1993–98 Debt indicators for the transition economies, 1990–98 Net IMF lending to the transition economies, by facility, 1990–98 Foreign currency disbursements to the Russian federal government, 1994–98 Multilateral debt relief agreements with official creditors, 1990–98 Financial relief through restructuring and payment deferrals International Rescue packages, East Asia, Russia and Brazil, July 1997 to October 1998 Capital flight from Russia, 1994–98

62 64 65 66 67 68 71 72 73 74 79 84 84 86 88 90 105 106 108 109 110 111 112 113 119

List of Figures 1.1 1.2 1.3 2.1 2.2

2.3 2.4 2.5 2.6 2.7 3.1 3.2 5.1 5.2

Lagged monetary growth (M2) and inflation (), Russia, 1992–94 Fit of the distributed lag model of inflation, Russia, 1992–94 Money multiplier, Russia, 1992–94 Velocity of circulation of M2, Russia, 1992–98 Inflation and a one-time change in the intercept of a trend stationary process, Russia, February 1992 to July 1998 Tests on the parameter stability of the autoregressive distributive lag model of inflation Recursive coefficient estimates of the autoregressive distributed lag model of inflation Actual and ARMA(6,1) fitted inflation Static forecast of the ARMA(6,1) model with the structural break included, August 1992 to July 1998 Static forecast of ARCH(1,0) model of inflation, November 1992 to July 1998 Recursive coefficient estimates of m2 in the distributed lag model Correlation between inflation () and lagged ruble broad money (m2), February 1992 to July 1998 Capital flows into transition economies, 1993–98 Flows of FDI to transition economies, 1990–98

ix

4 5 6 29

30 37 38 42 43 44 69 70 104 107

Acknowledgements This book was originally conceived as a PhD dissertation and was defended in 2003 under the title Money Supply–Inflation Relationship in Postcommunist Russia at University College London. I cannot do justice here to all those who helped me in so many ways to complete the manuscript, so I would like to express my gratitude to those most closely involved. On a personal level I benefited considerably from discussions with my supervisors Alan Smith and particularly Tomasz Mickiewicz. I also benefited from discussions with my former supervisors Stanislav Gomulka and particularly Jacek Rostowski, who was very influential in the first stages of my research and during the coauthoring work. Similarly I am very grateful to Jens Hölscher and John Driffill for comments and suggestions on earlier versions of the text. Special thanks go to my former colleague Peter Westin, who as editor of Russian Economic Trends kindly provided me with up-to-date time series on the extended money supply in Russia. I would like to thank John Bonin (editor of the Journal of Comparative Economics) and a number of anonymous referees for useful comments and suggestions on a previous version of part of the manuscript. I am also grateful to all my colleagues, my students at the University of Surrey and other academics and friends who contributed material and answered numerous questions. Thanks in this regard go to Lukasz Konopielko, Janice Bell, Venla Sipila, Dejan Djokic, Slavo Radosevic, Roman Zyla, Chris Flockton, Vesna Perovic-Karadzic and many others. I thank Elsevier and the Journal of Comparative Economics for permission to reprint previously published material. Finally, many thanks to Ivan and Luka, whose company has been enormously inspiring and therapeutic – I dedicate this book to them. MILAN NIKOLIC

x

Introduction Numerous empirical studies have been devoted to inflation processes and have demonstrated consistent money–price relationships in various market economies. However while their propositions may be valid for market economies they do not seem hold for the majority of transition economies. Apart from in Russia and Poland, no systematic evidence of a money–price relationship was detected in transition economies in the first half of the 1990s, thus undermining the conventional monetarist view in the transitional context. Among other things this implies that the traditional tools used to stabilize and control inflation in advanced market economies may not be appropriate for transition economies. This book scrutinizes the inflation process in postcommunist Russia, and the strength, dynamics and causality of the relationship between inflation and various monetary aggregates. In particular it investigates whether lagged inflation has been an important determinant of price rises, whether there is a significant relationship between inflation and the monetary aggregates, and whether the lower inflationary environment that emerged in 1994 and 1995, caused the transmission of monetary impulses to future inflation to become both slower and weaker. It also aims to shed additional light on the choice of lag selection criteria for causality testing on the one hand, and the suitability of monetary aggregates for influencing and controlling inflation via policy instruments on the other. Since each part of the analysis contributes to the evaluation of the suitability of the stabilization measures used in Russia, the part played by the international financial institutions under whose influence stabilization was conducted is inevitably brought under the spotlight. In this context the dual role of the International Monetary Fund (IMF) as the main coordinator of Western assistance to Russia and the main guide to Russian economic policy merits special attention. In Chapter 1 the analysis of macroeconomic instability in postcommunist Russia emphasizes the necessity of coordinating fiscal and monetary policies. Although Russia may not have a fiscally dominant economic regime, the continuing difficulty of balancing the government budget has led to monetization of the deficit. The analysis reinforces the notion that it is monetization of the deficit, rather than the deficit itself, that has affected the price level in Russia. Both the theoretical framework xi

xii

Introduction

presented in this chapter and the empirical analysis demonstrate the need for fiscal correction and that there may not be a simple monetary cure for inflation. Chapter 2 analyzes the formation and rationality of inflationary expectations in Russia, where successive stabilization policies have tended to ignore inflation persistence. In the absence of a surveys on inflationary expectations and time series of financial market indicators the analysis relies on inflation history. A review of the literature on inflation expectations and rationality is followed by details of our methodology and our criteria for model selection. The empirical findings from two competing models are presented and both models are rigorously scrutinized for their forecasting ability. The findings clearly demonstrate that lagged inflation was a very important determinant of the rate of inflation in Russia and this ought to have been considered when designing stabilization policies in the first half of the 1990s. Chapter 3 analyzes the strength and dynamics of the relationship between inflation and monetary aggregates. The chapter begins with a consideration of the literature on the topic and Russia’s monetary policy during the period in question. Next the data, methodology and criteria for model selection are outlined. This is followed by an empirical investigation of the significance of the relationship between money and inflation. In order to ascertain whether the lower inflationary environment in 1994 and 1995 slowed and weakened the transmission of monetary impulses to future inflation we then split the sample into two periods and conduct a number of empirical and diagnostic tests. It is found that broad money growth had the strongest correlation with inflation. However this relationship was unstable and sensitive to changes taking place in the new economic and institutional environment, and the average speed at which changes in the growth of ruble broad money were transmitted to inflation increased. Chapter 4 builds on the analysis in Chapter 3. It aims to shed additional light on the choice of lag selection criteria in causality testing, and investigates the suitability of monetary aggregates for influencing and controlling inflation. The results show unequivocally that the choice of lag length in distributed lag models is crucial to the outcome of causality testing. Of the various ad hoc and statistical criteria for optimum lag length selection, Akaike’s (1970) final prediction error criterion is found to outperform all others. The analysis reveals that there was feedback or bilateral causality between inflation and broad money in postcommunist Russia, which brings into question the wisdom of choosing money-based stabilization measures.

Introduction xiii

Chapter 5 examines the role played by international financial institutions in Russia’s stabilization efforts and particularly the involvement of the International Monetary Fund (IMF). It is argued that, contrary to the common perception, the financial and technical assistance provided to Russia by the IMF in the 1990s was less than generous. It consistently underestimated the amount of assistance needed for successful stabilization, disbursed far fewer resources than had been pledged by the West and had no desire for debt forgiveness. Moreover most of the funds that were disbursed appear to have been wasted. It is notable that all of Russia’s stabilization programmes had IMF approval and all of them failed. It is concluded that the IMF made a significant number of policy mistakes that exacerbated the country’s economic problems. Admittedly the transition process was a unique experience for Russia and errors were inevitable. Nevertheless, given the enviable resources of the IMF it could have done much better and at least some of the socioeconomic pains could have been avoided.

This page intentionally left blank

1 The Emergence of Open Inflation and Stabilization Efforts in Postcommunist Russia

Introduction By the beginning of the last decade of the twentieth century it had become clear that the task of mending the system of central planning was beyond even the resourceful and mighty Soviet Union. Substantial overhauls in 1957, 1965 and 1975, as well as piecemeal reforms between 1985 and 1989, failed adequately to address the chronic inefficiencies of the system. In the face of the rapid deterioration of the economy after the dissolution of the Soviet Union and the breakdown of overall authority, the newly installed Russian government embarked on a programme of radical economic reforms at the end of 1991. Russia’s leap into market reforms began on 2 January 1992. The cornerstones of this endeavour resembled the Polish reforms1 launched two years earlier and included a general decontrol of prices and trade, stabilization of the domestic currency and privatization.2 Perhaps the least successful aspect of the Russian reformers’ endeavours was their failure to regain macroeconomic stability, and high and volatile inflation characterized the economy throughout the 1990s. Price rises are usually analyzed within the framework of the classic quantity equation. However in the theoretical analysis of the determinants of inflation an appealing new strand of research emphasizes the role of fiscal policy. Proponents of this approach (Cochrane, 1998; Canzoneri et al., 1998; Sims, 1994; Woodford, 1994) hold the view that there can be two regimes for price determination: the so-called ‘monetary dominant regime’ and the ‘fiscal policy regime’. In the former the price level is determined by the quantity equation. Thus monetary policy de facto determines the price level while fiscal policy is said to ‘remain reactive’ (Komulainen and Pirtilä, 2000). In the latter regime the price level is 1

2

Monetary Policy in Transition

determined by the government’s intertemporal budget constraint. In other words, if future surpluses are likely to fall short of financing the debt, prices must be adjusted upwards, thereby reducing the real value of government debt. Hence in this regime fiscal policy determines the price level and monetary policy is reactive. That is, when there are changes in the price level the money supply is adjusted to bring the money demand equation into balance. This chapter considers the emergence of inflation in postcommunist Russia from the viewpoint of these regimes and their interaction. The chapter stops short of empirically determining the dominant regime because the lack of reliable data on the budget deficit would render the results unreliable, given the short time series, and because empirical studies (for example Komulainen and Pirtila, 2000) have already rejected fiscal dominance as an explanation of inflation in postcommunist Russia. This is not to say that fiscal deficit does not matter for inflation. On the contrary, as the traditional macroeconomics literature suggests and as Komulainen and Pirtilä (2000) have found, it is the inflationary method of financing the deficit rather than the deficit per se that affects the price level. The following section discusses the monetary roots of inflation following the liberalization of prices. The third section briefly describes the new fiscal theory of price level, provides an overview of the related literature and applies the theory to postcommunist Russia. The fourth section presents a theoretical framework for analyzing the interaction between monetary and fiscal policies, and the fifth analyzes Russia’s fiscal stance during transition and the repercussions of this on monetary policy. The final section summarizes the findings.

The monetary roots of inflation in postcommunist Russia Price liberalization and the emergence of open inflation Russia’s market reforms began on 2 January 1992 with the liberalization of prices. However this measure was not only tardy but also far too timid, with serious economic, intellectual, social and political consequences (Åslund, 1995). One of most serious consequences was sustained high inflation. Although views on the causes of inflation differ there is a consensus among economists about the detrimental effects that inflation has on domestic economies.3 In transitional economies, high inflation distorts relative prices, discourages investment, inhibits growth (Fisher et al., 1996), generates uncertainty about key prices,

Inflation and Stabilization Efforts in Postcommunist Russia 3

encourages unproductive activities aimed only at hedging against inflation, fosters a general climate of uncertainty and lack of trust in government policies, and hurts the most deprived groups in society (Hernández de Catá, 1995). The failure of successive Russian governments to control inflation has been a focal point of the academic literature on systemic transformation.4 Hence since an analysis of macroeconomic stabilization is crucial to understanding of Russia’s economic transition we pay particular attention to this topic. The announcement of the sweeping liberalization of previously controlled prices is often regarded as the beginning of shock therapy in Russia. The chief purpose of liberalizing prices was to eliminate the excessive demand for artificially low priced goods and the consequent queues. Although a sizable monetary overhang had prevailed in the Soviet economy since the introduction of currency reform in 1961 it had become particularly acute in 1991. In that year the monetary authorities had created as much as 137.5 billion rubles, more than the total of 133.8 billion printed between 1961 and 1990 (Ekonomika i Zhizn, vol. 10, 1992). Another reason for price liberalization was to alter production to suit consumers’ preferences rather than state prescriptions. About 80 per cent of producer prices and approximately 90 per cent of consumer prices became free in value terms (Åslund, 1995). However price controls remained on food and energy, which resulted in huge subsidies to these sectors and contributed to the continuing budget deficit. Contrary to the prediction by various institutions including the Russian government and the International Monetary Fund, prices rose in January 1992 by as much as 245 per cent.5 The failure of these institution to envisage the size of the price jump upon liberalization can be explained by inappropriate use of models of the demand for financial assets and reliance on the velocity of savings rather than that of cash, because the former proved to be much more variable than the latter (Gros and Steinherr, 1995). In any event the elimination of monetary overhang should not have caused inflation per se since inflation is by definition a continuous increase in the price level. Thus a one-off jump in prices should not have caused inflation. As it happened price rises in Russia attained double digits each month in 1992 and 1993, except in August 1992 (Nikolic, 2001). There is considerable evidence that the price rises after liberalization were due to increases in the money supply.6 More precisely the inflation rate followed variations in the growth of the broad money supply four months earlier (Figure 1.1). As the growth of M2 in 1992–94 is a reasonably good proxy for the rise in prices (Table 1.1 and Figure 1.2), inflation appeared to be a monetary phenomenon in the early years of transition.7

4

Monetary Policy in Transition

40 M2 (–4)

30

p

20 10 0 July 1992

January 1993

July 1993

January 1994

July 1994

Figure 1.1 Lagged monetary growth (M2) and inflation (), Russia, 1992–94 (%) Source: Nikolic (2000a).

Table 1.1 OLS estimates of the distributed lag model of inflation, Russia, February 1992 to September 1994 Variable

Coefficient

M2 M2(1) M2(2) M2(3) M2(4) M2(5) M2(6) M2(7) M2(8) D1 D6

0.025028 0.328145 0.180520 0.212921 0.257276 0.211170 0.250186 0.038532 0.018118 3.668939 0.651321

R2

0.706866

Adjusted R2

0.462587

Standard error of regression Sum squared residuals Log likelihood Durbin–Watson statistic

2.221126

Std error 0.099480 0.094653 0.117613 0.117796 0.084087 0.100409 0.090307 0.083230 0.081146 2.114631 1.919754

59.20080

Mean dependent variable Standard deviation of dependent variable Akaike information criterion Schwarz criterion

43.50815 1.666928

F-statistic Probability (F-statistic)

t-statistic 0.251587 3.466824 1.534870 1.807534 3.059635 2.103103 2.770400 0.462960 0.223274 1.735026 0.339273

Probability 0.8056 0.0047 0.1508 0.0958 0.0099 0.0572 0.0169 0.6517 0.8271 0.1083 0.7403 0.650348 3.029836 4.739840 5.282902 2.893686 0.042489

Note: D1 and D6 are seasonal dummies, denoting seasonal variations in January and June, respectively. Source: Nikolic (2000a).

If inflation was indeed a consequence of the growth of the money supply, the obvious question is what drove this growth? The answer can be found in an analysis of the balance sheet of the Central Bank of Russia (CBR). The accounting identities of the CBR imply that (1) an increase in net domestic assets (NDA) and net international reserves

Inflation and Stabilization Efforts in Postcommunist Russia 5 10 5 0 4

–5

2

–10

0 –2 –4 Jan. 1993 Apr. 1993 July 1993 Oct.1993 Jan. 1994 Apr. 1994 July 1994 Residual

Actual

Fitted

Figure 1.2 Fit of the distributed lag model of inflation, Russia, 1992–94 Source: Nikolic (2000a).

(NIR) were fundamental determinants of base monetary growth (MB); and (2) base monetary growth, often referred as high-powered money, was the fundamental determinant of broad monetary growth (M2), which in turn was a fundamental determinant of inflation (), as stated above and elaborated in the next chapter. CBR activities thus drive inflation. However there are two complicatory factors. The first complication is related to a change in the monetary base. An increase in net international reserves and domestic credits does not necessarily imply a one for one increase in the monetary base. The difference lies in a non-credit component of net domestic assets: other items net (OIN). The latter includes all other unclassified assets of the CBR and its net worth. More specifically, MB  NIR  NDA

(1.1)

NDA  NCD  OIN

(1.2)

NCD  NCG  NCB  NFSR  MB  NIR  OIN

(1.3)

where NCG, NCB and NFSR are variations in the CBR’s net credits to the central government, commercial banks and former Soviet republics respectively. The second complication arises from the relationship between base money and broad money, which depends on the currency in circulation (C), deposits (D) and commercial banks’ reserves (R).8 If there is a change

6

Monetary Policy in Transition 4.00

3.50

3.00

2.50

2.00

Fe b. 19 98

19 97 Ju ne

.1 99 6 ct O

Fe b. 19 96

19 95 Ju ne

.1 99 4 ct O

Fe b. 19 94

19 93 Ju ne

.1 99 2 ct O

Fe b. 19 92

1.50

Figure 1.3 Money multiplier, Russia, 1992–94 Source: Calculations based on data in Russian Economic Trends (Russian European Centre for Economic Policy, 1993–99).

in the ratio of broad money to base money, referred to as the ‘money multiplier’, base monetary growth and broad monetary growth will diverge. Typically, in periods of loose monetary policy the money multiplier falls, and vice versa. As Figure 1.3 shows, after a brief period of volatility at the beginning of 1992 the money multiplier fell dramatically between October 1992 and October 1993. Needless to say this coincided with a period of a rapid credit expansion by the CBR. Additional reasons for the changes in the money multiplier during the first years of transition were changes in interest rates, the reserve requirement and fluctuations in the sizable excess reserves held by commercial banks, which in turn reflected, or at least in part, the inefficiency of the payment system in Russia.9 Following a period of relative stability, the money multiplier began to rise at the end of 1994 and in general continued its upward trend during the rest of the period under consideration. During this time monetary policy was relatively tight, prices were stable and the efficiency of the payment system improved. CBR credits Whatever the complications arising from changes in the money base and the money multiplier, and from the relationship between broad money and inflation, it is clear that credits issued by the CBR were the main cause of the growth of the money supply. Table 1.2 provides details of the credit issued in 1992–95.

Inflation and Stabilization Efforts in Postcommunist Russia 7 Table 1.2 CBR credits flows, 1992–95 (percentage of GDP)

Commercial banks 1992 1st quarter3 2nd quarter 3rd quarter 4th quarter 1993 1st quarter 2nd quarter 3rd quarter 4th quarter 1994 1st quarter 2nd quarter 3rd quarter 4th quarter 1995 1st quarter 2nd quarter

Government (Ministry of CIS states Finance)

Other1

Total

GDP2

15.0 8.3 10.2 14.1 18.5 3.6 7.9 6.9 5.4 0.9 1.8 2.0 2.2 1.7 1.6

8.7 3.1 10.3 16.7 4.6 1.6 6.1 4.2 1.7 0.0 0.7 0.1 0.3 2.5 0.0

13.5 0.0 12.7 15.0 15.7 8.7 9.3 6.7 5.6 5.3 7.7 6.7 10.9 9.0 5.5

0.5 0.8 0.5 0.3 1.6 0.7 4.3 1.0 0.4 0.5 0.1 1.4 1.0 2.8 1.9

36.6 12.2 33.7 46.0 37.2 14.6 27.5 16.7 12.3 6.8 8.9 10.1 12.4 11.1 5.3

18,063 1,832 2,703 5,042 8,486 161,700 13,200 22,000 48,500 78,000 630,000 85,700 116,800 183,500 244,000

2.0 0.6

0.0 0.0

1.4 0.1

0.0 0.1

3.3 7.0

252,000 354,000

Notes: 1. CBR credits to enterprises, to other (non-federal) forms of government and to extrabudgetary funds. 2. Rubles bn, current prices. 3. Average monthly increase in flow of credits relative to December 1991. Sources: Calculations based on data in Russian Economic Trends (RECEP, 1995) and Granville (1995).

Net credit to the government The first type of credit, net credit to the government, was used to finance the budget deficit. In normal circumstances there are three ways to finance a budget deficit: money creation, internal debt or bonds and external debt. The undeveloped domestic financial market and the low credibility of the ruble prevented the government from financing its deficit by bond creation in the early stages of the reform, and this, combined with the non-arrival of pledged international assistance, left the government with no option but to increase the money supply.10 Table 1.3 shows the size of the budget deficit in the period 1992–98 Monetization of the budget deficit was mainly achieved by varying the monetary base, which amounted to an implicit taxation. The two most common measures used to evaluate the real value of the revenues

8

Monetary Policy in Transition Table 1.3 Russian budget deficit, 1992–98 (percentage of GDP) 1992

1993

1994

1995

1996

1997

1998

18.9

7.3

10.4

6

8.9

7.6

8

Source: EBRD (2000).

that a government can obtain by money creation are seigniorage (SE) and inflation tax (IT). Seigniorage is usually defined as: SE 

M GDP

(1.4)

where M is an aggregate of money supply such as currency in circulation (C), base money (MB) or broad money (M2), and GDP is gross national product.11 Conversely inflation tax is imposed on holders of monetary wealth and can be defined as follows: M GDP (C  D(  i)) IT  GDP

IT 

(1.5) (1.6)

where , D and i represent the inflation rate, ruble deposits and interest rate respectively.12 As Table 1.4 shows, both seigniorage and inflation tax amounted to a significant proportion of GDP, particularly during the first two years after liberalization. The trouble with monetary financing from the government’s point of view was that economic agents opted to hedge against tax on their monetary wealth. They did so mainly by converting rubles into foreign currency, US dollars in particular. Hence the tax base (MB/GDP) as a proportion of GDP (in this case base money, MB) shrank, as shown in Table 1.4.13 If the government had continued to maintain a constant level of monetary financing inflation would have accelerated. In other words if the tax base had shrunk further, taxes would have had to be increased in order to maintain the same level of monetary financing. An additional consequence of low monetization was that it made the economy very sensitive to capital flows. An annual capital flight of several billion US dollars, or more than a third of the entire stock of ruble M2 valued in US dollars, could have caused serious macroeconomic problems and threatened the entire financial system. Thus the government faced a stark choice: to find a non-monetary way of financing the

161.0 2,506.1 840.0 204.4 128.6 21.8 10.9 84.5

97.2 1,526.0 875.0 311.4 197.7 47.8 14.7 27.8

 annual average

1.4 19.0 171.5 610.7 1,585.0 2,200.0 2,602.0 2,685.0

191 1,678 13,278 35,698 80,800 103,800 130,400 187,800

182 2,235 16,691 48,000 103,800 130,900 164,500 210,400

958 6,400 32,601 97,800 220,800 288,300 374,100 448,300

13.00 11.76 9.73 7.86 6.55 5.95 6.32 7.84 7.83 6.76 3.67 2.85 1.05 1.02 2.14

10.81 8.43 5.13 3.52 1.23 1.29 1.71

28.64 15.28 10.68 7.76 3.07 3.30 2.76

14.2 11.0 6.5 3.6 1.0 0.6 4.6

52.8 24.0 12.7 7.5 2.3 1.3 9.7

GDP C MB M2 MB/GDP SEC SEMB SEM2 ITMB ITM2 (rubles tril) (rubles bn) (rubles bn) (rubles bn) (%) (%GDP) (%GDP) (%GDP) (%GDP) (%GDP)

Sources: Calculations based on EBRD (2000); RECEP (1993–99).

Notes:   inflation rate; GDP  gross domestic product; C  cash in circulation; MB  monetary base; M2  ruble broad money; SE  seigniorage; IT  inflation tax: derived from the monthly figures of economic aggregates (RET, 1993–99) according to the formula IT  (M* / GDP)  100. Unlike in Layard and Richter (1994), therefore, interest receipts from deposits are not accounted for.

1991 1992 1993 1994 1995 1996 1997 1998

 (end of year)

Table 1.4 Seignorage and inflation tax on monetary aggregates, Russia, 1991–98

9

10 Monetary Policy in Transition

budget deficit if the latter could not be slashed, or to let inflation accelerate. From May 1993 the government issued various securities to help finance the deficit. These securities covered 1.5 per cent, 10 per cent, and 40 per cent of the budget deficit in 1993, 1994 and 1995 respectively (Granville, 1995). External financing also helped to reduce the deficit, although this was considerably less than had been pledged (Nikolic, 2002). In early 1995 the CBR was prohibited from financing the budget deficit as this was essential to the stabilization of the economy in the short to medium term. Net credit to commercial banks Net credit to commercial banks, in addition to providing liquidity to banks through the refinance rate, were also channelled to enterprises as a non-budgetary subsidy. These credits were allocated according to political bargaining rather than market considerations (Granville, 1995). In the dispute between the government and the Duma over jurisdiction of the CBR the Duma had gained the upper hand. The Duma's reasons for increasing the issuance of credits by the CBR to enterprises via commercial banks were to maintain employment and help ailing industries and agriculture. These credits included military conversion subsidies and working capital credits (ibid.). Clearly macroeconomic stabilization was not high on the CBR’s list of priorities during this period. CBR credit to commercial banks amounted to 15 per cent and 3.6 per cent of GDP in 1992 and 1993 respectively (Table 1.2). Most of these credits were earmarked for ailing enterprises and agriculture. Since the credits were issued via commercial banks the responsibility for their repayment lay with the banks. However no action was taken until at least 1994 against a bank or a firm that could not reimburse its loan (ibid.). Moreover since the real interest rates on the credits were negative, enterprise credits amounted to grants. In November 1993 positive real interest rates were introduced and banks started to refuse to channel the credits. As commercial banks were slow to forward the funds to enterprises they built up huge excess reserves (RECEP, 1995). For example their excess reserves were about four times more than the required reserves in 1992 and double the required reserves in 1993 (Granville, 1995). This puzzling phenomenon can be partly explained by the inefficiency of the financial system, the large credit risk and the opportunity the banks had to delay channelling the credit in order to buy foreign assets that would yield considerable gains in the short term given the high inflation rate. It is very likely that the inflation tax imposed by the government on the

Inflation and Stabilization Efforts in Postcommunist Russia 11

non-interest bearing funds was passed on to deposit holders in the form of lower saving rates. Net credit to former Soviet republics Both cash and non-cash credits were provided to ‘near abroad’ fomer Soviet republics to enable them to continue trading with Russian enterprises. The cash credits were intended to minimize the severance of links between enterprises in these states, thus constraining trade shocks and falls in output. In addition to cash credits, many non-Russian enterprises were able to obtain unauthorized credits (‘non-cash’ rubles) from the CBR via national central banks. The dual money system, characterized on the one hand by an unlimited supply of non-cash credits and on the other by a hard constraint on the delivery of cash, was particularly costly to Russia in the first year of transition. In 1992 CBR credits to other former Soviet republics amounted to at least 8.5 per cent of Russian GDP if the delivery of cash is excluded, and 11.6 per cent otherwise (Granville, 1997). The substantial increase in the money supply inevitably resulted in higher prices and hampered the stabilization efforts. The destabilizing effects of the ruble zone prevailed until November 1993. In July that year the CBR had suddenly withdrawn pre-1993 ruble notes and this, together with the collapse of negotiations between Russia and Kazakhstan in November 1993, effectively sealed the fate of the old ruble area.

The fiscal theory of price level and its application to Russia The prevailing view in the classic macroeconomics literature is that inflation has monetary determinants such as the money supply and the exchange rate. In contrast, at the heart of the new fiscal theory of price level is the notion that the price level must adjust to equilibrate the real value of nominal government debt with the present value of surpluses (Komulainen and Pirtilä 2000). However the quantity theory and fiscal theory of price determination are not mutually exclusive but are differing strands of the same theory (Cochrane, 2000). The origins of the new fiscal theory can be found in the work of Sargent and Wallace (1986) and their notion of interaction between monetary and fiscal policies. The theory was further developed and formalized by Woodroff (1994), Sims (1994), Burnside et al. (1998) and Canzoneri et al. (1998). Unsurprisingly the new theory has attracted considerable criticism. The critics charge that since the theory rules out the possibility of

12 Monetary Policy in Transition

government default the model leads to overdetermination of the price level in the fiscal regime (Buiter, 1999). It is also argued that the model is misspecified since it assumes that the government’s intertemporal budget constraint needs to be satisfied only in equilibrium (Buiter, 2002). Similarly Cochrane (1998) argues that if a government violates the present budget constraint the model cannot be empirically tested since prices will react and the off-equilibrium price sequence will remain hidden. Despite these criticisms the new fiscal theory has found application in transition economies since most of them have experienced persistent budget deficits and high inflation. Among the theoretical papers on the subject it is worth mentioning the work of Begg and Wyplosz (1999), who have used the theory to show the necessity of prudential stabilization in efforts in Central and East European countries that wish to join the European Union. According to Fakin and de Crombrugghe (1997), after their dramatic shift away from subsidies in the early years of transition the countries of Central Europe still show signs of unsuccessful fiscal adjustment, insufficient deficit reduction and loose expenditure policy. Although fiscal issues were usually overshadowed by monetary ones in transition economies, interest in fiscal policy intensified in the wake of the Russian financial crisis in August 1998. A large number of authors have stressed the importance of controlling the fiscal deficit to achieve price stability (see for example Barbone and Marchetti, 1995; Budina and van Wijnbergen, 1997; Buiter, 1997; Fakin and de Crombrugghe, 1997; Coricelli et al., 1998; Cottareli and Boyle, 1999; Dabrovski, 1999). Because of its consistently high level of inflation during transition prior to the financial crisis, the perils of the unsustainable Russian budget deficit were emphasized by Cheasty and Davis (1996), Lopez-Claros and Alexashenko (1998) and Sinel’nikov-Murilev and Trofimov (1998), among others. In the wake of the crisis there was little doubt among researchers that the cause of the country’s financial meltdown was its poor fiscal performance coupled with a number of internal and external factors (Nikolic, 2000a; Desai, 2000; Sutela, 2000; Kharas et al., 2001). However there is little empirical evidence that inflation in Russia was due to its fiscal deficit. Using vector-autoregressive models Komulainen and Pirtilä (2000) analyzed whether fiscal deficits had caused inflation in Romania, Bulgaria and Russia. While there was some evidence that the fiscal deficit had increased inflation in Bulgaria, they rejected the null hypothesis that fiscal deficit was significant determinant of

Inflation and Stabilization Efforts in Postcommunist Russia 13

inflation in both Romania and Russia. According to the authors this finding did not support the presence of fiscally dominant regimes in these countries. Rather the price levels in Russia and Romania were determined in monetarily dominant regimes by the quantity theory of money. This is not to say that fiscal policy does not matter. On the contrary it does, as traditional macroeconomics would suggest. Komulainen and Pirtilä’s findings simply underline the notion that the inflationary method of financing the deficit, rather than the deficit itself, affects the price level.

Frameworks for analyzing the impact of monetary and fiscal policies on macroeconomic stabilization Predictably the expansionary monetary policy of the CBR was not conducive to macroeconomic stabilization. Unlike in other transition economies, the mainly half-hearted stabilization programmes, implemented in Russia in 1991–94 had a very little success. Indeed the mid 1995 stabilization effort culminated in the severe financial crisis of August 1998. It has become conventional wisdom that these programmes failed because they were not supported by structural and institutional reforms (Gavrilenkov and Kuboniwa, 1997). In the absence of such reforms most of the macroeconomic problems arose from the fiscal side. Subsequently monetary policy was used as an adjustment tool. It follows that in order for a stabilization programme to succeed, fiscal and monetary policies should be interrelated and coordinated. The interaction between these policies can be analyzed using the Sargent and Wallace (1976, 1986) framework. Fiscal and monetary framework In Sargent and Wallace’s (1976, 1986) framework the interaction between policy makers (the monetary and fiscal authorities) and the public (private agents) is endogenized. The fiscal authority takes decisions on public expenditure and tax rates, while the monetary authority decides about the composition of government debt. They do so by maximizing their objective function subject to the constraints imposed by the behaviour of the public. Meanwhile the public decides about consumption, investment and employment and is assumed to pay the imposed taxes. The rational expectations hypothesis (Muth, 1961) postulates that future rates of inflation are conditioned by agents’ perceptions of long-term monetary and fiscal policies. The agents are expected to change their

14 Monetary Policy in Transition

‘strategy of rules’ whenever there is a regime change in government policies (Sargent and Wallace, 1976). Private agents limit government action and determine the government’s budget constraint by choosing how much and what combination of the elements of government debt to hold. Hence monetary and fiscal policies must be coordinated because the government faces a budget constraint. It follows that there is no purely monetary cure for inflation. The budget constraint is thus becomes pivotal to the analysis of inflation stabilization in Sargent and Wallace’s framework. The government’s budget constraint (Sargent and Wallace, 1976, 1986) is represented by: Gt  Tt 

(Ht  Ht1)  (Bt  Bt1)  rt1Bt1 t

(1.7)

where Gt is real government expenditure at time t, Tt represents real taxes net of transfers (apart from interest payments on government debt) at time t, Ht is a stock of base money at time t, t is the price level at time t, Bt is the real value of a one-period government bond issued at t and paid off at (t  1) and r is the net real rate of interest. Equation 1.7 specifies that a real primary or non-interest budget deficit (Dt  Gt  Tt) at time t, can be financed by increasing the monetary base and/or issuing to the public an interest-bearing debt. Following Keynesian tradition, this framework assumes that the public is willing to hold interestbearing government debt on the same terms as private debt. The upper – bound (B) of the public’s willingness to accumulate real interest-bearing government debt (Bt) is assumed to be constrained by the total wealth in – the country. In practice (B) is often far lower than the total wealth. Similarly the public’s willingness to accumulate the other government debt – the stock of base money – is determined by Cagan’s (1956) function of demand for money. This describes the demand for real base money as a decreasing function of the expected rate of inflation, as follows:

冤 冥

Ht t1 t  1  2 Et t

1 2  0

(1.8)

where Et[·] is the value of [·] expected to prevail by the public as of time t. When solving Eqation 1.8, t can be expressed solely in terms of the expected future value of Ht:

兺冢

冣

1 2 j EtHtj  1 j0 1

(1.9)

Inflation and Stabilization Efforts in Postcommunist Russia 15

The price level at time t is determined by the interaction between the public’s preference for holding high-powered money (parameters 1 and 2) and the expected path of high-powered money now and in the indefinite future. It follows that the government deficit can influence the price level solely through the effects on the expected path of high-powered money. As pointed out by Sargent (1993) this is the crux of the monetarist supposition that inflation is always a monetary phenomenon. However the government deficit and the path of high-powered money are not necessarily rigidly linked because the government can cover its deficit by interest-bearing government debt, or at least temporarily and to a point, as shown in Equation 1.7. Hence under the Sargent’s system (Equations 1.7 and 1.9) the inflationary consequences of the government deficit depend on the government’s strategy for servicing its debt (ibid.). There are two alternative debt-servicing strategies: a strict Ricardian regime, and the Friedman (1989) rule. The former regime is not inflationary, or at least not initially, but the latter is. In the strict Ricardian regime the government always finances its entire deficit (surplus) by issuing (retiring) interest-bearing government debt. Sargent (1993) shows that this kind of regime has no effect on the price level as long as the deficit is of a temporary nature, since it is not permitted to affect the level of base money. In contrast, according to the Friedman (1989) rule deficits are always entirely financed by issuing additional base money. Subsequently the time path of government deficits affects the time path of the price level via an increase in the money supply, as shown in Equation 1.7 (Sargent, 1993). There are a number of combinations of the above regimes that can be used. For example Bryant and Wallace (1980) and Sargent and Wallace (1986) have described debt-servicing regimes that are intermediate between Ricardo’s and Friedman’s. In these regimes, issuing high-powered money will affect the price path sooner or later. The time lag of the effects depend on the parameters 1 and 2 in Equations 1.8 and 1.9 (Sargent, 1993). The idea that monetary and fiscal policies are interrelated and must be coordinated is further developed by Sargent and Wallace (1986). Their extended analysis is based on the premise that in the absence of fiscal authority help, fighting current inflation by means of tight monetary policy will lead to higher inflation in the future. According to Sargent and Wallace, fiscal correction is the necessary and sufficient condition for stabilization. The limits of monetary policy are further illustrated by Sargent and Wallace’s ‘unpleasant monetarist arithmetic’ (ibid.). The authors argue

16 Monetary Policy in Transition

that financing the debt by money and debt, thus shifting towards tight monetary policy, will lead to an increase in government debt in the future. That increase will be caused by an increase in government’s interest burden. Given that there is a limit on the public’s willingness to absorb the debt, the government will find it difficult to sell its debt to the public except at falling prices, which will in turn increase expenditure. In order to avoid insolvency the government will have to resort to inflationary money creation to finance its deficit. Thus the public’s perception that the government is unable to meet the intertemporal budget constraint will cause an increase in the inflation rate above the level initially expected by the agents. In other words inflation will only occur if the present value of the government debt is not equal to the present value of all the government surpluses expected in the future. The only way to avoid inflation is to ensure that the fiscal programme meets the intertemporal budget constraint and is accepted as credible evidence that inflation will not be used to finance the budget. Should long-term government policies violate the budget constraint a higher inflation rate will inevitably follow, as in the rational expectations model. In sum, Sargent and Wallace’s ‘unpleasant arithmetic’ postulates that the central bank will be prevented from successfully fighting inflation if the fiscal authority persists in running a net-of-interest budget deficit. Open economy framework The analysis so far has concentrated on a closed economy. In an open economy the government budget constraint (Equation 1.7) is also determined by international economic transactions. These include financial and real resources transfers between trading nations. Thus the government budget constraint for a small open economy becomes: Gt  Tt 

(Ht  Ht1)  (Bt  Bt1)  rt1Bt1 t et(B*t  B*ti)  r*t1etB*t1  et(R*t  R*t1)

(1.10)

This can be rearranged as Ht Dt  rt1Bt1  r*t1etB*t1    Bt  et(B*t  R*t ) t

(1.11)

where Dt  Gt  Tt is the real government primary or non-interest budget deficit as described above, et is the real exchange rate at time t, B*t is foreign public debt at time t, R*t is the foreign currency reserve at

Inflation and Stabilization Efforts in Postcommunist Russia 17

time t, and r*t is the interest rate on foreign debt at time t. Hence Equation 1.11 states that the excess of real government spending, G, at time t, and domestic debt, rB, plus foreign debt service, r*B*t , over real tax receipts, T, must be financed by one of four means: printing money, running down the foreign currency and/or gold reserves, or selling public debt to the domestic sector or the foreign sector. Running down the foreign currency reserves, if these are sizable, will only bring temporary relief. Moreover it will adversely affect the exchange rate. Similarly borrowing from the public or from abroad may induce domestic spending, which will in turn affect the budget deficit.

Interaction of fiscal and monetary policies during transition Budget distortions in the late years of communism The budgetary problems and fiscal crises that have plagued Russia’s economic transformation have their roots in the late years of communism. The failure of central planning partly reflected the state’s inefficiency in balancing its budget. Among the important factors behind the mounting budget distortions at the end of 1980s were the fall in the world prices of raw materials, growing military expenditure due to the arms race and the war in Afghanistan, unfavourable economic policy decisions in the mid 1980s (for example the attempt at new industrialization and the anti-alcohol campaign which sharply reduced budget revenues), and a series of large-scale disasters (for example Chernobyl and the earthquake in Armenia). The deterioration of state finances, as manifested in falling revenues and increased expenditure, is shown in Table 1.5.

Table 1.5 Budget balance, USSR, 1985–90 (rubles bn)

1985

1986

1987

1988

State budget revenue 367.7 366.0 360.1 378.9 State budget expenditure 386.0 415.6 429.3 459.5 Balance  18.3  49.6  69.2  80.6 Balance as a percentage of GDP  2.4  6.2  8.4  9.2 Source: Robinson (2001).

1989 (estimated)

1990

398.8 479.8  81.0

429.9 489.9  60.0

 8.8

 6.2

18 Monetary Policy in Transition

Russia’s fiscal performance during transition The Russian reformers intended to tackle the budget deficit in the first stage of the reforms. President Yeltsin was confident that the deficit could be slashed and announced that ‘in 1992 [it] should be almost non-existent or minimal’ (Sovetskaya Rossiya, 29 October 1991). In the event the reformers proved overly optimistic and the results fell far short of expectations. As Table 1.3 has shown, the budget deficit persisted and marred the government’s efforts to stabilize the economy throughout the transition period. The main problem with the expenditure side of the state budget was the government’s inability to stop subsidy payments being made by local authorities and the CBR (Robinson, 2001). Conversely tax collection was weak. According to official data, the tax revenues of Russia’s enlarged budget (including the consolidated revenues of the federal and regional budgets and the extrabudgetary funds) fell from over 44 per cent of GDP in 1992 to less than 30 per cent in 1996 (RECEP, 1997). The situation improved slightly in 1997 as a result of the payment of tax arrears by several large enterprises, most notably Gasprom. The key factors in the deteriorating revenue performance were as follows (ibid.): ● ●

●

●

●

A decline in output and profits. A shrinkage of the tax base due to tax exemptions, tax deferrals and other tax concessions. Statutory tax rates were reduced under the profit tax and VAT reforms, so that the excess wage tax was eliminated and export tariffs were phased out. A deterioration in tax discipline, a sharp increase in tax arrears,14 and tax evasion.15 A large share of tax revenues was received in kind, which made fiscal policy less flexible.

The government’s fiscal position was further undermined by the maintenance of the outdated and opaque tax code. The 1991 Law on the Principles of The Tax System was eventually replaced by a new tax code, but only its first general part was implemented and then not until the beginning of 1999. The inadequate budget revenues in 1992 and 1993 severely undermined the the government’s stabilization effort. More importantly for the reform process, the government’s poor fiscal record had an adverse effect on the reformist parties in the parliamentary election of December

Inflation and Stabilization Efforts in Postcommunist Russia 19

1993, and eventually the prime minister and finance minister were forced out. According to one Russian newspaper the shortfall in tax collection in 1993 amounted to 30 per cent (Rossiiskie vesti, 5 March 1994), and by the autumn 1994 it had risen to as much as 50 per cent (Segodnya, 27 September 1994). This dramatic decline in revenues could not be offset sufficiently by slashing expenditure, so the ballooning budget deficit exerted unbearable pressure on the exchange rate, which was slipping away from the CBR’s control. With reserves running low the CBR was unable to prevent what has been dubbed Black Tuesday (11 October 1994), when the ruble – US dollar exchange rate fell by 28 per cent, jumping from R3,000 to almost R4,000 per dollar in one day (RECEP, 1994). Black Tuesday brought home the depth of the reform’s failures in respect of budget financing over the previous three years and it was realized that the days when the CBR could issue large credits without causing high inflation were over. As a consequence in 1995 the CBR was barred from financing the budget deficit. In the new policy environment the government drew on domestic debt and borrowing from abroad. Short-term treasury bills Gosudarstvennye Kratkosrochnye Obligatsiis (GKOs) and fixed coupon bonds Obligatsii Federal’nykh Zaemovs (OFZs) had been introduced in 1993, but they only took off in the aftermath of Black Tuesday as a result of a rise in the CBR’s discount rate. By 1996 the stock of GKOs and OFZs amounted to about 11 per cent of GDP (RECEP, 1993–99), causing a major concern for the monetary authorities. The interest rate on these securities rose to over 100 per cent in the same year and the state’s domestic debt tripled (Izvestya, 15 October 1997). By the autumn of 1996 commercial banks no longer had sufficient liquidity to finance the government debt market since the stock of GKOs and OFZs exceeded the total stock of ruble deposits in the banking system (EBRD, 1998). In order to gain control over the growing deficit the government had to find new pool of funds at a lower cost of borrowing. Given the liquidity problem in the Russian banking sector the government’s only option was to seek external sources of deficit financing in order to roll over debt repayments into fresh bond issues and expand the debt market at a lower cost (Robinson, 2001). As a consequence the GKO market was opened to foreigners in 1996. The demand for GKOs peaked in 1997, when foreign investors acquired US$19.7 billion, or about 34 per cent of the entire stock of GKOs and OFZs (ibid.). The high demand for these securities lowered the interest rate on government debt to less than 10 per cent in the summer of 1997 (ibid.). Most of the foreign money invested in 1996

20 Monetary Policy in Transition

and 1997 was in short-term debt that matured in less than a year. Maintaining investors’ confidence was therefore crucial to the roll-over into new GKO issues of matured debt. Investors were aware that the accumulating deficits would push up the government’s debt servicing costs in the future and therefore increase the size of future deficits. As the government’s finances, and thus the exchange rate target, were viewed as unsustainable the currency risk premium rose, resulting in upward pressure on interest rates (Nikolic, 2000a).16 This in turn led to an increase in future debt servicing costs and so thus increased future expected budget deficits. This spiral was exacerbated by the adverse effects of the higher interest rates on growth, the dwindling foreign exchange reserves and the less than credible government policies, all of which eventually led the country into a variant of the debt trap (ibid.). The dire state of the government’s finances and the economy as a whole was further undermined by additional external and internal factors (ibid.). The most important of the external factors, prompted by the 1997 Asian crisis, was the growing reluctance of international capital investors to invest in countries such as Russia, which was plagued by macroeconomic and structural weaknesses, particularly overdependence on short-term capital inflows (RECEP, 1998). Another adverse external factor was the fall in oil and other commodity prices. Among the domestic factors the most important were the excessively large budget deficit and the unsustainable build-up of ruble-denominated debt. These were due to poor tax collection, unproductive government expenditure, the small and weak domestic capital market, the weak and inefficient banking system, poor corporate governance and the continued accumulation of payment arrears (ibid.). The failure of policy makers to address these issues in the relatively favourable investment climate of 1997 led in August 1998 to Russia’s worst financial crisis in its transition period and contributed to global financial turbulence. The ruble was effectively devalued and left floating and the government defaulted on its GKOs (Nikolic, 2000a).17 Empirical indicators of the government’s non-interest budget constraint Using a simplified version of Sargent and Wallace’s (1976, 1986) framework we shall estimate the budget constraint faced by the Russian government between 1992 and 1998, excluding the interest accruing on the government’s domestic and foreign debt. The reason for this exclusion is that reliable data on the build-up of interest on government debts are

Inflation and Stabilization Efforts in Postcommunist Russia 21 Table 1.6 Simple empirical assessment of the non-interest budget constraint, Russia, 1991–98

1991 1992 1993 1994 1995 1996 1997 1998

GDP (rubles bn)

Dt* (rubles bn)

1,400 19,000 171,500 610,700 1,585,000 2,200,000 2,602,000 2,685,000

– 4,929.47 38,710.52 71,829.47 69,331.03 228,479.41 176,541.81 900,879.59

Dt Ht (% GDP) (rubles bn) – 25.94 22.57 11.76 4.37 10.39 6.78 33.55

– 2,053 14,456 31,309 55,800 27,100 33,600 45,900

Bt (rubles bn)

Rt* ($ bn)

Bt* ($ bn)

– – 207.63 12,532.37 60,975.70 163,393.10 147,767.20 2,176.51

8.200 0.744 4.354 1.869 10.386 2.841 1.936 5.305

– 10.60 33.30 9.80 1.20 4.60 1.10 57.60

et (rubles/$) 169.2 414.5 1,247.0 3,550.0 4,640.0 5,570.0 5,974.0 21,140.0

* Calculated by averaging end of period exchange rate. Sources: EBRD (1999); RECEP (1993–99); IMF (1999); Nikolic (2002); and author’s calculations.

not available. Hence Equation 1.11 is simplified as follows: H Dt    Bt  et(B*t  R*t) t

(1.12)

Equation 1.12 states that the real government deficit, D, at time t must be financed by one of four means: printing money, Ht, running down international currency (and/or gold) reserves, et(B*t  R*t), selling public debt to the domestic sector, Bt, or selling public debt to the foreign sector, B*t. As Table 1.6 shows, the non-interest nominal budget deficit was very significant in 1992–98. The money supply was the principal means of financing the deficit in 1992–95, and although monetization of the deficit continued in the subsequent years domestic borrowing took over as the principal source of funds. In 1996–98 domestic borrowing was about twice as large as the increase in base money. In 1998 alone, even though the government could not borrow after its default in August that year, the increase in domestic borrowing amounted to 14.4 per cent of GDP, a similar increase to that in the previous year. Thus even from the limited evidence presented in Table 1.6 it is clear that the financing of the government deficit was following an unsustainable path. International financial investors could clearly see that the government’s finances were in disorder and that the government would default on its obligations sooner or later.

Conclusion Our analysis of macroeconomic instability in postcommunist Russia has revealed the necessity of coordinating fiscal and monetary policies, in that

22 Monetary Policy in Transition

the monetary authority cannot successfully fight inflation if the fiscal authority persists in running a net-of-interest budget deficit. Needless to say controlling inflation is even more difficult if the monetary authority subordinates macroeconomic stability to other socioeconomic aims. Although Russia does not have a fiscally dominant economic regime, its continuing inability to balance the budget has resulted in monetization of the deficit. In the early years after price liberalization the budget deficit was mainly financed by money creation, but in the following years domestic borrowing exceeded the level of money creation. This eventually became unsustainable, leading the economy into financial crisis in August 1998. Both periods were characterized by a lack of macroeconomic stability. Both the theoretical framework presented in this chapter and the simple empirical analysis demonstrate the need for fiscal correction. In this framework fiscal correction is the necessary and sufficient condition for stabilization. This is not to say that the Russian fiscal regime has dominated the monetary one. Rather we wish to stress that it was monetization of the deficit, rather than the deficit itself, that has affected the price level in postcommunist Russia.

2 Inflation Expectations in Postcommunist Russia

Introduction The central role of inflation expectations has long been recognized in macroeconomic theorizing and stabilization policy analysis. Wage bargaining, price setting, asset allocation and investment all depend on inflationary expectations in one way or another. However interest in the matter in the former socialist economies arose only during the liberalization attempts at the beginning of transition. Given economic agents lack of experience with open inflation in the majority of transition economies, few studies have been conducted on the formation of expectations in these economies. Hence the objective of this chapter is to analyze the formation and rationality of inflationary expectations in postcommunist Russia, a transition economy in which stabilization policies initially ignored inflation persistence. In the absence of a sample survey of inflationary expectations and an appropriate time series of financial market indicators we are forced to rely on inflation history. The findings of this chapter are in sharp contrast to the claim that lagged inflation is relatively unimportant in explaining inflation in transition economies (Cottarelli et al., 1998; Coorey et al., 1998; Cottarelli and Boyle, 1999). The chapter is organized as follows. The next section provides a brief overview of the literature and the origins of inflation and inflationary expectations in Russia. In the third section the data, methodology and criteria for model selection are outlined. The empirical results are reported and discussed in the fourth section, while the fifth considers an alternative model. The sixth and seventh sections are devoted to tests of the forecasting ability of the models and the rationality of expectations, respectively. Some policy implications and suggestions for further 23

24 Monetary Policy in Transition

research are outlined in the eighth section, and the ninth summarizes the findings.

A brief overview of the literature on inflation expectations in postcommunist Russia and early reactions to open inflation Early analyses of the formation of inflation expectations were characterized by an emphasis on the weighted average of past changes (Fisher, 1930) and the part played by an exogenous psychological factor — that is, ‘animal spirit’ (Keynes, 1936). In other words, expectations were assumed to be subject to a high degree of inertia but also to unexplained waves of optimism or pessimism. After the Second World War expectations were modelled in a deterministic manner, mainly assuming that the expected value of a certain variable could be proxied by its observed values in the recent past (Cagan, 1956; Frenkel, 1973; Holden and Peel, 1977). Alternatives to the arbitrary modelling of price expectations were based on financial market indicators or survey data. The former method measured inflationary expectations by means of a financial indicator, such as a forward interest rate (Mishkin, 1990; Svensson, 1993; Frankel and Lown, 1994; and Söderlind, 1995), while the latter drew on survey data, for example from the Livingston Survey in the United States or the Gallup poll in the United Kingdom.1 The limitations of these types of modelling gave impetus to development of the rational expectation hypothesis (REH) (Muth, 1961; Sargent and Wallace, 1976). Although the REH could be applied to a wide range of economic variables, most theoretical and empirical studies focused on the formation of inflationary expectations. Interest in the role of expectations in market economies arose following disillusionment with the validity of the Phillips curve and in light of the stagflations of the 1970s.2 It was argued that when high and growing unemployment coexisted with rapidly accelerating inflation, prices would rise because people expected them to rise (Carlson and Parkin, 1975). Since the formation of this expectation was central to inflation persistence (Ball, 1991), this topic received scrupulous attention by the designers of stabilization policies, particularly but by no means exclusively in inflation-targeting regimes. Various models of expectations were developed, particularly for the United States. These models can be broadly classified as those which fell into the adaptive expectations hypothesis (Fisher, 1930; Cagan, 1956; Frenkel, 1973), hybrid models of expectations,3 the REH (Muth, 1961; Poole, 1976;

Inflation Expectations in Postcommunist Russia 25

Sargent and Wallace, 1976; Shiller, 1978) and the direct measurement of inflationary expectations in the form of sample surveys (Turnovsky, 1970; Carlson and Parkin, 1975; Carlson, 1977; Figlewski and Wachtel, 1981) or financial market indicators (Mishkin, 1990; Svensson, 1993; Söderlind, 1995). Researchers paid great attention to the Livingston data published by The Philadelphia Enquirer. Reportedly the most striking feature of this series was its poor forecasting ability (Wachtel, 1977). For example actual inflation exceeded expectations in over 70 per cent of the surveys conducted in the period 1947–73. This poor predictive power was demonstrated by a low R2, which did not exceed 30 per cent, and the large constant term of 2.05 per cent (ibid.). With regard to the rationality of the Livingston data, opinions differed. While many researchers found that the data were consistent with rationality (for example Turnovsky, 1970; Mullineaux, 1978, 1980), others rejected the REH even in its weak form, except for the specific period 1957–71 (for example Pesando, 1975; de Menil and Bhalla, 1975; Carlson, 1977; Holden and Peel, 1977; Figlewski and Wachtel, 1981).4 In contrast to market economies, where the formation of both consumers’ and producers’ inflation expectations have received considerable attention, the topic has sparked little interest in transition economies. Although inflation expectations are not at the core of their analyses, Hoggarth (1996), Allen et al. (1996), Korhonen and Pesonen (1998) and Nikolic (2000a) have all found that one-month lagged inflation is among the most important determinants of inflation in the Russian economy. Quite the reverse has been found for transition economies as a whole by Cottarelli et al. (1998) and Coorey et al. (1998). The scant interest in the topic in transition economies can perhaps be explained by the lack of experience with open inflation in the majority of these economies, the lack of survey data on expectations of inflation and the unavailability of financial and time series data of inflation. In the absence of means of independently measuring expectations, we postulate a scheme for generating expectations in terms of observable variables. First, however, a brief description of the history of inflation in postcommunist Russia is in order. In the presence of a monetary overhang, the liberalization of prices on 2 January 1992 was bound to cause an explosion in prices. Indeed in the first month of liberalization, prices rose by about 245 per cent, which far exceeded of what had been predicted by the International Monetary Fund (IMF) and the Russian government. This jump in prices can be analyzed with the help of the quantity equation. An often-used variant

26 Monetary Policy in Transition

of this equation is Pigou’s (1927) income version: MV  PY

(2.1)

where M is the total quantity of money in the economy, V is the velocity of income circulation, defined as the average number of times a unit of currency turns over in the course of financing the year’s final activity, P is an implicit price deflator and Y is national income. If prices in Russia were fixed at time t but liberalized at time t  1 the jump in prices should be equal to Pt1 (Mt1/Mt)(Vt1/Vt)  Pt Yt1/Yt

(2.2)

Subsequent analysis of short-term price movements after liberalization can safely ignore changes in income and money supply because there is no reason to expect that the government would have printed money at the same time as liberalizing prices. By the same token there is no a priori reason to expect that income would have changed dramatically because of liberalization. After all it would have taken time for these changes to materialize. Hence, for, say, the first month after liberalization one can assume that Yt1/Yt and Mt1/Mt were both approximately close to one. It follows that prediction of the jump in prices required only prediction of an increase in velocity. It is in this area that serious miscalculations occurred. While the IMF and the Russian prime minister respectively estimated that the size of the monetary overhang was about 50 per cent (Gros and Steinherr, 1995) and that prices would rise by 100 per cent (Rossiiskaya Gazeta, 3 July 1992), prices actually jumped by as much as 245 per cent in January 1992, as noted earlier. One of the probable causes of this miscalculation was the choice of model. Sophisticated models of the demand for financial assets had given quite good longterm results in developed market economies (Gros and Steinherr, 1995). However these models were unsuitable for Russia and other transition economies, where households essentially had only three assets: cash, savings deposits and foreign currency. Another probable cause of the miscalculation was reliance on the velocity of bank deposits rather than the velocity of cash, in that the former proved to be much more variable than the latter (ibid.). Since empirical studies of the demand for money in transition economies were very limited and even non-existent in some newly independent states, the degree of velocity was seriously misjudged in

Inflation Expectations in Postcommunist Russia 27

subsequent stabilization programmes. The designers of the stabilization programmes in the Baltic states, Russia and other countries of the former Soviet Union thought that inflation would not respond immediately to monetary tightening. Therefore they projected an increase in the velocity of broad money in the first quarter of the programme, and assumed that it would subsequently stabilize or even decline. In the event, velocity continued to increase in subsequent quarters in almost all of the economies where stabilization was not successful. Whatever the initial jump in prices may have been and however calculated, price adjustment caused by elimination of the monetary overhang would not have resulted in inflation if prices had not continued to rise in the following months. It would merely have resulted in a one-off change in the price level, because by definition inflation is a continuous rise in the level of prices. In the event prices continued to rise by double digits every month in 1992 and 1993, apart from in August 1992. After considerable volatility in 1994 and 1995, prices gradually stabilized during the following years, although price rises never fell to a single annual digit. In the years after liberalization the main reason for inflation was a series of supply shocks. These arose from a combination of shortages, monopolistic pricing, the financing of the budget deficit, the excessive issuance of credits and a substantial decline in the value of the ruble in the foreign exchange market. Economic agents in Russia had hardly any experience of monetary policy in a market economy, but the continuous increase in velocity indicates that they had noted that the government permitted the economy to adjust to supply shocks by increasing the general level of prices,5 and that they expected a similar response to future shocks. The Russian people, however, took time to adjust to the new inflationary environment and to protect the real value of their assets and income. The government took advantage of this delay by imposing an extraordinarily high inflation tax. In 1992 (except January) and 1993, when inflation was at its highest, the average monthly inflation tax amounted to as much as 32.1 per cent and 15.5 per cent of GDP respectively.6 Over time inflationary expectations became deeply embedded. Typically businesses reacted by demanding higher prices for their good and services. In turn trade unions demanded higher wages and benefits for their members in order to maintain real income in the face of persistent inflation. Hence most economic agents in Russia came to accept inflation as permanent. It appears that the rise in prices was the result of both inertia and expectations. The former was a consequence of the rigid price and wage

28 Monetary Policy in Transition

determination system inherited from the previous economic regime and the latter a consequence of the considerable depreciation of the ruble, which itself was largely due to the consistently high money supply.7 In any stabilization attempt the former is best broken by freezing wages and the latter by anchoring expectations to an effective intermediate target. Since in the absence of a well-developed financial market, economic agents’ best available proxy for a current inflation rate, except inflation history, is a prevailing level of market exchange rate, the latter seems to be an obvious target. Anchoring the exchange rate gives strong and transparent signals that future inflation is likely to be low. This is particularly relevant for transition economies such as Russia, where evident shifts in the demand for money and unstable velocity reduced the effectiveness of money as a target. Nevertheless none of the Russian stabilization programmes had heterodox elements or the exchange rate was used as an intermediate target until July 1995. At that time a currency band of 4300–4900 rubles per US dollar was introduced, and this was later modified into a sort of a crawling peg. While the official programme did not envisage price or wage controls, the Russian authorities applied these measures in autumn 1995 (Bofinger et al., 1997). This was accompanied by the imposition of control over the budget deficit, so that the stabilization of inflation became an achievable aim, or at least in the medium term. It is clear that the introduction of the currency band, some wage control and more restrictive monetary and fiscal policies dampened inflationary expectations. This assertion is confirmed by the fact that the velocity of the circulation of M2 steadily declined after the implementation of the stabilization programme (Figure 2.1).8 Although quantity theory assumes that velocity and real growth are stable in the long term, this may not be the case in the short term.9 The period prior to and after stabilization in Russia in July 1995 is a clear case in point. In the years prior to stabilization inflation was clearly sticky – that is, the tightening of monetary policy was not followed by an immediate fall in inflation, as manifested in the increase in velocity. Between February 1992 and June 1995 M2 and inflation grew on average by 13.2 per cent and 16 per cent respectively. Thus inflation rose about 3 per cent more than the money supply. Possible explanations of the divergence between the growth of money and prices include changes in the demand for real balances, exogenous movements in the exchange rate, wages, the current terms of trade, the future stance of monetary policy, the growth of interenterprise credits and other arrears, the remaining administrative prices and finally measurement errors.10

Inflation Expectations in Postcommunist Russia 29 1.2

1.0

0.8

0.6

0.4 1992

1993

1994

1995

1996

1997

1998

Figure 2.1 Velocity of circulation of M2, Russia, 1992–98

The generally upward trend in the velocity of M2 in this period was, however, most likely caused by a gradual improvement of the payment system, expectations of higher future inflation and the growing ability of economic agents to protect their monetary holdings from depreciation. Conversely the mid 1995 stabilization programme dampened inflation expectations, so that the money supply could be increased somewhat without necessarily causing inflation. Moreover the tightening of monetary and fiscal policies, positive rates of return on domestic assets, the stable real exchange rate and the containment of capital flight increased confidence that inflation would remain low, so that velocity declined significantly in the following years. Indeed the average monthly growth of the money supply in this period was 2.32 per cent while inflation rose by just 1.66 per cent. Judging by the trend in velocity before and after June 1995 it appears that inflation was expected to rise in the former period and to decline in the latter. In both cases economic agents could have used their previous experience of inflation to anticipate its current level within general trends in the two periods. Thus, since inflation persistence was strong throughout the entire period, it seems promising to analyze the expected rate of inflation by its recent history.

Methodology and model specification The data in this study consist of monthly observations of the consumer price index published by the Russian European Centre for Economic

30 Monetary Policy in Transition

Policy and cover the period between February 1992 (the month after price liberalization) and July 1998. The cut-off date is dictated by availability of data. The inflation path and its descriptive statistics, as well as a one-time change in the intercept of a trend stationary process, are shown in Figure 2.2. An interesting feature of the descriptive statistics is that the Ljung–Box tests reject the null hypothesis of zero autocorrelation. This is an indication that the series contain a large degree of persistence and perhaps of near unit root process. However the standard tests for unit root and the KPSS test for stationarity do not reject the null hypothesis of stationarity.11

40 Residual Inflation Trend

30 20

20 10 10

0

0

–10

–10 –20 1992

1993

1994

Series: Observations: Mean: Median: Maximum: Minimum: Standard Deviation: Skewness:

1995

p 78 9.184463 5.680609 37.99595 –0.347032 9.199365 0.925905

1996

1997

Kurtosis: Jarque-Bera: Probability Jarque-Bera: Q20-statistic: ProbabilityQ20 ADF C,T(4) PP C,T(9) KPPS C,T(7)

1998

2.919567 11.16594 0.004 464.98 0.000 –4.4430* –4.8795* 0.1100*

Figure 2.2 Inflation and a one-time change in the intercept of a trend stationary process, Russia, February 1992 to July 1998 Note: ADF and PP  augmented Dickey–Fuller and Phillips–Perron standard unit root tests, respectively; KPPS  the Kwiatkowski, Phillips, Schmidt and Shin (1992) tests for stationarity; an asterisk indicates significance at the 1 per cent level; C, T() indicates that a constant term and/or a linear trend has been introduced and the maximum lag length of the dependent variable (in parenthesis).

Inflation Expectations in Postcommunist Russia 31

In addition, the Jarque–Bera normality test shows that inflation contains a sizable non-normal distribution. While kurtosis of the series is near mesokurtic (2.92), non-normal distribution is probably caused by the considerable skewness to the right (0.93). The positively skewed distribution lends credence to the hypothesis that inflation persistence may be driven by relative price adjustment (Coorey et al., 1998). Administered price increases may have supported higher inflation rates in the presence of downward sticky prices. The underlying assumption in this chapter is that inflation in postcommunist Russia was determined to a large extent by its lagged values (Nikolic, 2000a). As mentioned earlier there is clear evidence that inflation in Russia was sticky in the period prior to June 1995. The idea of sticky nominal prices and wages is one of the corner stones of macroeconomics. It implies, among other things, that in the short term monetary policy affects economic activity rather than prices. Apart from a probable asymmetry in price and wage stickiness, a major limitation to proxying expectations solely by past behaviour is the implicit assumption that expectations are not affected by other factors, such as a major shift in government policy, discrete exchange or terms of trade changes and exogenous shocks. By definition such expectations are not rational. As long as changes from year to year are smooth, expectation proxies based solely on past behaviour are likely to work well in terms of explanatory power. However in the event of sudden and large fluctuations there is clear risk that the model may perform poorly. Despite these shortcomings, in the absence of experience with economizing with real money balances in a market economy with an immature financial system, a recent monthly inflation rate is the best available proxy for current monthly inflation for most economic agents in Russia, particularly with regard to the few years following the liberalization of prices. A very simple autoregressive distributed lag (ADL) model based on inflation history and capturing inflationary expectations is given in Equation 2.3. t  c0 

X

兺 

i1

i

ti

 e0, t

t  1, … , T

(2.3)

where c0 is a constant,  is a coefficients of lagged inflation, subscripts i and t are the lag length and the current time period respectively,  is an inflation rate, X is the optimal lag length of autoregression, and e0, t is the stochastic error term that follows the classical assumptions; namely it has zero mean, constant variance and is not autocorrelated (that is,

32 Monetary Policy in Transition Table 2.1 Lag length selection for the autoregressive distributed lag model of inflation SC 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

5.18 5.20 5.23 5.19 5.14 4.96 4.97 5.00 4.80* 4.87 4.94 5.00 5.08 4.95 5.03

* Denotes the smallest value of the Schwarz criterion.

it is white noise). This model can be thought of as belonging to the family of models from Hicks’ classical analysis of the factors that affect the formation of expectations.12 Prior to estimating expectations it is essential to establish an optimal lag length for the autoregression, which amounts to determining an integer X in Equation 2.3. Of a number of possible statistical and ad hoc criteria, minimizing the Schwarz criterion of predictive accuracy is favoured because it not only produces the most parsimonious model but is also asymptotically consistent.13 Application of the Schwarz criterion (SC) for X up to 15 produced the results presented in Table 2.1.

Results Having determined the appropriate lag length for the ADL model of inflation we apply the ordinary least squares (OLS) method to Equation 2.3. However application of the OLS (Table 2.2) does not produce efficient estimates since the error term in the final expression does not seem to follow the classical assumption specified above. More specifically, although the diagnostic statistic does not indicate any deficiency of the model, further tests utilizing the LaGrange Multiplier and correlograms

Inflation Expectations in Postcommunist Russia 33

of the squared residuals, plus Jarque–Berra tests (Tables 2.3–2.5), reveal that the magnitude of the residuals is related to the magnitude of recent residuals. In other words there is autoregressive conditional heteroscedasticity (ARCH) in the residuals. Although this does not invalidate the standard OLS inference, ignoring ARCH effects may result in loss of efficiency. Table 2.2 OLS estimates of the autoregressive distributed lag model of inflation, February 1992 to July 1998 Variable

Coefficient

t-statistic

0.001 1.068 0.191 0.212 0.368 0.207 0.045 0.015 0.076 0.022

0.001 9.466* 1.181 1.294 2.290* 1.500 0.380 0.130 0.646 0.267

c t1 t2 t3 t4 t5 t6 t7 t8 t9 R2 R2 adjusted Standard error Log likelihood ratio

0.943 0.934 2.123 144.445

Probability 0.999 0.000 0.242 0.201 0.026 0.139 0.705 0.897 0.521 0.791

Schwarz criterion F-statistic Probability (F) Durbin– Watson

4.800 108.773 0.000 2.075

Table 2.3 Breusch–Godfrey serial correlation LaGrange Multiplier test Lag

F-statistic Probability Observations  R2 Probability

1

2

3

4

5

6

7

8

7.91* 0.01 7.28* 0.01

4.14* 0.02 7.68* 0.02

2.66 0.06 7.53 0.06

2.41 0.06 9.01 0.06

1.87 0.11 8.87 0.11

1.52 0.19 8.84 0.18

1.26 0.29 8.73 0.27

2.66* 0.02 17.72* 0.02

Note: The asterisks indicate significance at the 5 per cent level or better. The F-statistic is an omitted variable test for the joint significance of all lagged residuals. Because the omitted variables are residuals and not independent variables, the exact finite sample distribution of the F-statistic under H0 is not known, but we still present the F-statistic for comparison purposes. The observation  R2 statistic is the Breusch–Godfrey LaGrange Multiplier test statistic. This is computed as the number of observations multiplied by the (uncentred) R2 from the test regression. Under general conditions the LaGrange Multiplier test statistic is asymptotically distributed as 2( p).

34 Monetary Policy in Transition Table 2.4 Ljung–Box Q-statistics for the squared residuals Lag

Q-statistic Probability

1

2

3

4

5

6

7

8

9

7.68* 0.01

7.71* 0.02

7.73* 0.05

9.22* 0.06

9.88* 0.08

10.19* 0.12

10.37 0.17

22.89* 0.00

25.55* 0.00

15

16

17

18

26.82* 0.03

27.08* 0.04

27.09 0.06

27.09 0.08

Lag 10

11

12

13

14

Q-statistic 25.61* 25.61* 26.37* 26.69* 26.73* Probability 0.00 0.01 0.01 0.01 0.02 * Significant at the 5 per cent level or better.

Table 2.5 Normality test for the autoregressive distributed lag model Skewness Kurtosis Jarque–Bera Probability

0.348* 7.668* 64.028* 0.000

Note: Asterisks indicate significance at the 5 per cent level or better. The asymptotic standard errors of the skewness and kurtosis coefficients are 兹(6/T) and 兹(24/T) respectively, where T is the sample size. An asterisk associated with the coefficients of skewness and/or kurtosis indicates significance and implies that the coefficient is more than twice its standard error. The Jarque–Bera normality statistic is distributed as chi-square with two degrees of freedom.

In order to account for the ARCH effects we have utilized the ARCH models proposed by Engle (1982) and Bollerslev (1986).14 In the first step an appropriate specification of the ARCH model had to be determined. To that end we relied on minimizing the Schwarz criterion of predictive accuracy. We looked at the ARCH(1), ARCH(2) and ARCH(3) models and, as it has become a convention, the GARCH(1,1), GARCH(1,2), GARCH(2,2) and GARCH(2,1) models. According to Bera and Higgins (1993) a data set that requires a model of order greater than these presented above is very rare. The results of the search for the appropriate ARCH model are presented in Table 2.6.

Inflation Expectations in Postcommunist Russia 35 Table 2.6 Choice of ARCH/GARCH model ARCH Schwarz criterion

1,0

2,0

3,0

1,1

1,2

2,2

2,1

4.008*

4.120

4.147

4.126

4.063

4.113

4.069

* Minimum value of the Schwarz criterion for the given ARCH models.

Table 2.7 Maximum likelihood estimate of inflation, ARCH(1,0)

c0 t1 t2 t3 t4 t5 t6 t7 t8 t9

Coefficient

z-statistic

Probability

0.006 1.138 0.149 0.097 0.010 0.050 0.107 0.087 0.019 0.022

0.062 11.170* 0.919 0.766 0.102 0.559 1.493 1.558 0.283 0.620

0.951 0.000 0.358 0.444 0.919 0.576 0.135 0.119 0.777 0.535

Variance equation: C ARCH(1) R2 R2 adjusted Standard error Log likelihood Durbin–Watson Q18 (stand. res.) Probability Q18 Jarque–Bera (JB) Probability JB

0.106 1.609 0.933 0.920 2.340 112.873 2.223 13.467 0.763 2.108 0.349

0.979 2.824* F-statistic Probability (F) Q18 (sq. res.) Probability Q18 Skewness Kurtosis ARCH LM8 Probability LM8 RESET F (l) Probability F (1)

0.328 0.005 72.498 0.000 23.240 0.182 0.078 2.158 9.578 0.296 0.048 (0.057) 0.827 (0.811)

* Significant at the 5 per cent level or better.

Having determined the optimal specifications of the ARCH model we increase the efficiency of the estimate of Equation 2.3 by means of the maximum likelihood method. The estimates of the ARCH(1) are presented in Table 2.7.15 The diagnostic statistic presented in Table 2.7 does not reveal any deficiency of the ARCH(1) model of inflation in Russia for the given period. The fit of the model is reasonably good and there is no sign of

36 Monetary Policy in Transition

autocorrelation in the residuals. Although the only significant lag in the autoregression appears to be lag 1, the Wald test rejects the null hypothesis that the coefficients of the rest of the lags are jointly zero.16 Given the enormous changes that took place in Russia and the numerous structural shifts during the stabilization period, it would be interesting to examine whether the parameters of the model presented in Table 2.7 are stable across the various subsamples of data. The majority of stability tests, however, cannot be performed on models estimated by the likelihood method. Since most of these tests can be used with least squares and two-stage least squares regressions we subjected the OLS estimates of Equation 2.3, as presented in Table 2.2, to a plethora of recursive least squares stability tests. The tests included: a recursive residuals test (RRT), a CUSUM test based on the cumulative sum of the recursive residuals, a CUSUM of square test, a one-step forecast test (O-SFT), an N-step forecast test (N-SFT) and a recursive OLS coefficient test (ROLSCT). As shown in Figures 2.3 and 2.4, none of the stability tests except the CUSUM square test could reject the null hypothesis of no structural break in the data at the conventional level of significance.17 In contrast to the results of the rest of stability test, the plot of the CUSUM of square test shows that at the beginning of 1994 the parameters of the equation changed significantly. Aided by the plot of the O-SFT, we can see that the period between early 1994 and early 1995 is least successfully captured by the equation. This particular period was characterized by the uncertainty brought about by the stop and go nature of the various stabilization efforts and the exchange rate collapse in October 1994. Thus the performance of the model in this period is hardly surprising, particularly in light of the fact that we have already established that the inflation series has a structural break in 1994–95. In order to account for this we could include a dummy variable for the structural break plus a dummy for the external shock of October 1994. This would improve the fit by about 3 percentage points, but would inevitably be less parsimonious and would require further diagnostic testing. Nevertheless the CUSUM square tests indication of a structural break could not be supported by any of the other stability tests mentioned above. In particular, since the recursive residuals in the RRT fall inside the  2 standard error band there is no indication of instability in the parameters of the equation. Similarly the lower portion of the plot (vertical axis) of the O-SFT shows the probability values of sample points where the hypothesis of parameter constancy would be rejected at the 5,

Inflation Expectations in Postcommunist Russia 37 30

10

20 5 10 0

0 –10

–5 –20 –30

–10 1994

1995

1996

Recursive residuals

1997

1994

1998

± 2 Standard errors

1995

1996

CUSUM

1.4 1.2 1.0 0.8 0.6 0.4 0.2 0 –0.2

1997

1998

5% significance 10 5 0 –5 –10

0 0.05 0.10 0.15 1994

1995

1996

CUSUM of squares

1997

1994

1998

1995

1996

One-step probability

5% significance

1997

1998

Recursive residuals

10 5 0 –5 –10

0 0.05 0.10 0.15 1994

1995

1996

N-Step probability

1997

1998

Recursive residuals

Figure 2.3 Tests on the parameter stability of the autoregressive distributive lag model of inflation

10 or 15 per cent levels. While this plot can help us to identity periods when the equation is least successful, it shows that parameter constancy cannot be rejected at the 5 per cent level. By the same token the N-SFT, which uses the recursive calculations to carry out a sequence of Chow forecast tests, does not indicate instability in the parameters of the equation.18 Much of the same applies to the CUSUM, which is based on the cumulative sum of the recursive residuals. Since the cumulative sum does not go outside the area between the two 5 per cent critical lines there is no indication of parameter instability. Finally, the RCE enables

38 40

2.5 2.0

20

1.5 1.0

0

0.5 0

–20

–0.5 –40

–1.0 1994

1995

1996

Recursive c0 estimates

1997

1998

1994

1995

1996

Recursive p5 estimates

± 2 Standard errors

2.0

0.8

1.5

0.4

1997

1998

± 2 Standard errors

0 1.0 –0.4 0.5 –0.8 0

–1.2 1994

1995

1996

Recursive p1 estimates

1997

1998

1994

1.5

1995

1996

Recursive p6 estimates

± 2 Standard errors

1997

1998

± 2 Standard errors

1.0

1.0 0.5

0.5 0

0

–0.5 –0.5

–1.0 –1.5

–1.0 1994

1995

1996

Recursive p2 estimates

1997

1998

1994

3

1.5

2

1.0

1

0.5

0

0

–1

–0.5

–2

1995

1996

Recursive p7 estimates

± 2 Standard errors

1997

1998

± 2 Standard errors

–1.0 1994

1995

1996

Recursive p3 estimates

1997

1998

1994

2

1995

1996

Recursive p8 estimates

± 2 Standard errors

1997

1998

± 2 Standard errors

0.8

1

0.4

0 0 –1 –0.4

–2

–0.8

–3 –4

–1.2 1994

1995

1996

Recursive p4 estimates

1997

1998

± 2 Standard errors

1994

1995

1996

Recursive p9 estimates

1997

1998

± 2 Standard errors

Figure 2.4 Recursive coefficient estimates of the autoregressive distributed lag model of inflation

Inflation Expectations in Postcommunist Russia 39

us to trace the evolution of estimates for any coefficient as more and more of the sample data are used in the estimation. Not only do the plots of the selected coefficients not go beyond the two standard error bands, but also there are no dramatic jumps in the plots except slight ones at the beginning of the period, which are likely to reflect the limited number of observations used to estimate them at the start of the recursion. As a matter of fact the plots of the individual coefficients are remarkably smooth and linear. Moreover Ramsey’s (1969) regression specification error test (RESET) does not indicate any deficiency of the model.

Alternative modelling An alternative and more parsimonious method of modelling time series is the popular Box–Jenkins method, which is technically known as the autoregressive integrated moving average (ARIMA). ARIMA models encompass the generic family of models associated with the broad term ‘adaptive expectations hypothesis’. The emphasis of these new forecasting tools is not on constructing single or simultaneous equation(s) models but on analyzing the probabilistic, or stochastic, properties of economic time series on their own according to the philosophy ‘let the data speak for themselves’. Hence the conditions for optimal forecasts can be generalized to include the nature of shocks, leading to the general proposition that forecasts should be revised by a weighted average of all previous errors and not just by a fraction of the last one. In this way the ARIMA models provide a general framework for the most efficient forecasting of inflation, when no information other than past inflation is considered (Feige and Pearce, 1976). Given this the ARIMA models appear to be the most appropriate ones for our analysis. Since our time series is stationary – that is, it does not need to be integrated – we can model Equation 2.3 as autoregressive (AR) or moving average (MA) processes, or a combination of the two (ARMA). The first step of our analysis was to identify an approximate structure of the model by finding out appropriate values of the AR term (p) and the MA (q) term. To that end we utilized the chief tools for identification: The autocorrelation function (ACF), partial autocorrelation function (PACF) and the resulting correlograms. This analysis indicates that both the ACF and the PACF decay exponentially and have significant spikes.19 Hence the most promising model is an ARMA process. Significant spikes at lags 1 and 6 in the PACF indicate an AR(6) process while quite a few significant lags in the ACF indicate a complex MA process. In order to

40 Monetary Policy in Transition Table 2.8 ARMA(6,1) model of inflation Coefficient c0 AR(6) AR(1) MA(1)

3.506 0.067 0.878 0.337

R2

0.913

R2 adjusted

0.909

Standard error Log likelihood Durbin–Watson Q20-residual Probability Q20 Inverted AR roots

2.504 166.187 2.061 19.452 0.303 0.960 0.22  0.52i

Standard error

t-statistic

Probability

8.754 0.056 0.066 0.124

0.400 1.204 13.349* 2.708*

0.690 0.233 0.000 0.009

Mean dependant variable Std deviation of dependant variable Schwarz criterion F Probability F Q20-residual squared Probability Q20 squared 0.45th  0.46i 0.45  0.46i

Inverted MA root

8.128 8.318 4.854 238.536 0.000 21.178 0.219 0.22  0.52i 0.54 0.34

* Significant at the 5 per cent level or better.

identify the optimum model we again utilized the Schwarz criterion of predictive accuracy. We tested the AR(6) in combination with various MA terms. With the help of ACF, PACF and the Schwarz criterion we identified ARMA(6,1) as the most promising model. Having tentatively identified the appropriate p and q terms in our ARMA model we then estimated the parameters of the AR and MA terms included in the model. The results of this estimate (obtained by OLS) and the diagnostic statistics are presented in Table 2.8. After estimating the ARMA(6,1) model we conducted diagnostic tests to check whether it was specified correctly. We assumed that the random error terms in the actual process were normally distributed and independent. It followed that, if the model had been specified correctly, the residuals from the model should resemble a white noise process. None of the individual autocorrelation and partial autocorrelations turned out to be significant, nor did the Box–Pierce Q20 statistics of residuals and square residuals (Table 2.8). The reciprocal roots of the AR and MA polynomials had a modulus of no greater than one.20 Hence the white noise residuals obtained by the model indicated acceptance of the model. However when testing for the stability of the parameters of the model, the Chow break point (SBT) test indicated that there was a structural

Inflation Expectations in Postcommunist Russia 41 Table 2.9 ARMA(6,1) estimate of inflation with the structural break included

c1 SBFEB94 DO94 AR(1) AR(6) MA(1)

Coefficient

Standard error

t-statistic

Probability

14.391 9.661 2.515 0.779 0.103 0.588

3.972 1.971 1.093 0.080 0.063 0.114

3.624* 4.901* 2.301* 9.725* 1.629 5.161*

0.001 0.000 0.025 0.000 0.108 0.000

R2

0.939

R2 adjusted

0.934

Standard error 2.138 Log likelihood 153.730 Durbin – Watson 2.074 16.078 Q20 0.518 Probability Q20 18.582 Chow FT1995:06LR Probability Chow FT 0.997 Inverted AR roots 0.93

Mean dependent variable Standard deviation of dependent variable Swartz criterion F Probability F Q20 squared Probability Q20 squared RESET1 Probability RESET1 0.47  0.52i 0.25  0.56i

0.47  0.52i

Inverted MA root

8.128 8.318 4.627 201.801 0.000 17.743 0.405 1.403 0.236  0.25  0.56i 0.600 0.590

* Significant at the 5 per cent level or better.

break at the beginning of 1994.21 This finding came as no surprise, since we had already established that the inflation series had a structural break at that point. We therefore modified the ARMA(6,1) model to take account of the structural break (SBFEB94).22 In addition, following Nikolic (2000a), a dummy variable, DO94, was included in the model to account for exogenously induced inflation that occurred in 1994 as result of the exchange rate crisis of October 1994.23 Thus: t  c1  1AR(1)   1AR(6)  1MA(1)  1SBFEB94  1DO94  e1, t

(2.4)

where c1 is a constant, e1, t is an error term with the classical properties described above, and 1, 1, 1, 1 and 1, are the coefficients of the respective variables described above. Estimates of Equation 2.4 and the accompanying diagnostics statistics do not reveal any deficiency of the model, as shown in Table 2.9 and Figure 2.5. The fit of the model is good and the model seems to predict turning points reasonably well.

42 Monetary Policy in Transition 30 20 10

10

5

0

0

–10

–5 –10

1993

1994 Residual

1995

1996 Actual

1997

1998

Fitted

Figure 2.5 Actual and ARMA(6,1) fitted inflation

Forecasting inflation Perhaps the most important use of ARMA models is to forecast future values of the sequence of a dependent variable. Since our objective is to examine the formation of inflation expectations based on inflation history we shall test the performance of the ARMA(6,1) model in this regard. This is done by applying static forecasting, or one-step-ahead forecasting, (Figure 2.6). The first two forecast error statistics depend on the scale of the dependent variable. These can be used as a relative measure to compare forecasts for the inflation series across different models; the smaller the error the better the forecasting ability of that model, according to that criterion. The remaining two statistics are scale invariant. The Theil (1961) inequality coefficient always lies between zero and one, where zero indicates a perfect fit. The mean squared forecast error can be decomposed into the bias proportion, which tells us how far the mean of the forecast is from the mean of the actual series; the variance proportion, which tells us how far the variation of the forecast is from the variation of the actual series; and the covariance proportion, which measures the remaining unsystematic forecasting errors.24 In our model the bias proportion is extremely small, indicating that the mean of the forecasts does a good job of tracking the mean of the dependent variable. In other words, since the extent to which the average values of simulated and actual series deviate from each other is

Inflation Expectations in Postcommunist Russia 43 40

pF ± 2 Standard errors

30

20

10

0

–10 1993

1994

Forecast: Actual: Observations: Root mean squared error: Mean absolute error:

1995 pF p 72 2.0466 1.4205

1996

1997

Mean absolute percentage error: Theil inequality coefficient: Bias proportion: Variance proportion: Covariance proportion:

1998

78.594 0.0896 0.0000 0.0548 0.9452

Figure 2.6 Static forecast of the ARMA(6,1) model with the structural break included, August 1992 to July 1998

negligible there is no indication of systematic error in the model. Furthermore the somewhat larger, but still very small, the variance proportions indicate that most of the bias is concentrated on the covariance proportions. Hence our one-step-ahead forecast of the ARMA(6,1) model of inflation seems to perform reasonably well. We have also applied a one-step-ahead forecast to the estimated ARCH(1) model described earlier. The procedure included computing a static forecast of the mean, its forecast standard error and the conditional variance. The upper graph in the Figure 2.7 shows the forecast of the dependent variable from the mean equation, plus the two standard deviation bands. The lower graph is the forecast of the conditional variance. In general variance shown in the lower graph declines over the forecast period but it has three notable peaks. As could be expected, the largest peak is at the point of the structural break in the inflation series, as discussed above. The next largest peak is in October 1994, the time of the exchange rate crisis. The final peak (early 1995) perhaps reflects

44 Monetary Policy in Transition 40

pF ± 2 Standard errors

30 20 10 0 –10 –20 1993

1994

1995

1996

140

1997

1998

Forecast of variance

120 100 80 60 40 20 0 1993

1994

Forecast: Actual: Observations: Root mean squared error: Mean absolute error:

1995 pF p 69 2.1265 1.3832

1996

1997

Mean absolute percentage error: Theil inequality coefficient: Bias proportion: Variance proportion: Covariance proportion:

1998 64.981 0.0929 0.0109 0.0007 0.9884

Figure 2.7 Static forecast of ARCH(1,0) model of inflation, November 1992 to July 1998

the inflation shock that followed the liberalization of energy prices in this period. Since the ARCH term is not close to one, the volatility shocks, although notable, are not for the most part persistent and the forecasts of the conditional variance converge to the steady state reasonably quickly. In conclusion it seems that both models give good static forecasts of inflation.

Inflation Expectations in Postcommunist Russia 45

Rationality of expectations of inflation Under rationality the long-term response of expected inflation to the actual rate of inflation should be equal to one, as economic agents cannot systematically be fooled (Muth, 1961). In other words the expectations of agents are said to be rational if they coincide with the true mathematical expectations conditioned on all relevant information available at the time the forecast is made. Two tests are used to test the rationality of expectations: a test for unbiasedness and a test for efficiency. The former tests whether forecasts are unbiased estimates of the actual inflation rate, and the latter tests whether forecasts incorporate all the available relevant information. The latter typically distinguishes between ‘weak form efficiency’ and ‘strong form efficiency’. The weak form efficiency test involves testing whether the expectations error can be explained by expectation errors in previous periods. Conversely strong form efficiency (orthogonality) implies that economic agents make efficient use of all available information in forming expectations of the inflation rate. In order to analyze the underlying principle behind each of the rationality tests, let t be the inflation rate at time t and te the rate of inflation during period t expected by agents at the end of period t  1; It contains all relevant information at the end of t1. If te is the rational expectation (RE) of t, then  te  E[t兩It]

(2.5)

where E is the conditional operator. Equation 2.5 implies that te is an unbiased predictor of t, which can be tested by running the regression t  c2   et  e2, t

(2.6)

and testing the joint hypothesis that (c2, )  (0, 1). Acceptance of the null hypothesis would be consistent with rationality. However tests based on Equation 2.6 are rather weak, since the rational expectation hypothesis implies much more than unbiasedness. For expectations to be rational they must include all the relevant information available at the time they are formed. This information is likely to include the past history of inflation, previous inflation forecasts (and thus forecast errors) and other variables that could influence inflation, such as money supply, unemployment, fiscal variables and so on. Typically this is the objective of the tests for efficiency.

46 Monetary Policy in Transition

As stated above, the weak form efficiency test involves testing whether the expectations error can be explained by expectation errors in previous periods. Let j, t  t  te represent the measured forecast error, where j  1 for ARCH(1,0) and j  2 for the ARMA(6,1) model presented above. The hypothesis of zero correlation and non-significant zero mean can be tested by regressing j, t on lagged values of itself. That is, we estimate j, t 

K

兺 

i1

i

j, t1

 e3, t

(2.7)

and test the null hypothesis H0 : i  0, i  1, …, K for a range of choices K. Assuming that past price information is readily available and clearly relevant (that is, ti is contained in the It1 for i  1), an obvious test of rationality is that j, t cannot be correlated with lagged rates of inflation. The Equation 2.8 is estimated and the following null hypothesis is tested: H0 : i  0, i  1, …, K for a range of choices K. j, t 

K

兺 

i1

i

ti

 e4, t

(2.8)

Strong form efficiency or orthogonality implies that economic agents make use of all available information in forming expectations about the inflation rate. This hypothesis can be tested by looking for lack of correlation between forecast errors (j, t) and other variables contained in It1.25 j, t  zti   e5, t

(2.9)

where zti is a vector of information variables dated i  1 or earlier (extended broad money supply, m2x, in our case),26  is a vector of coefficients and the null hypothesis (consistent with rationality) H0 :   0, i  1 is tested. Each of the tests described above tests for violation of rationality in a particular direction. The tests produce various assessments of rationality and are therefore presented separately. Each test is performed on both the ARCH(1) and ARMA(6,1) models. However forecast error series are required to be stationary in their regressions, so, both series have been tested for stationarity and found to be stationary at the conventional 5 per cent level of significance. The results of the standard unit root tests are presented in Table 2.10. The diagnostic statistics in Table 2.11 clearly demonstrate that the null hypothesis of coefficient restriction in Equation 2.6 cannot be

Inflation Expectations in Postcommunist Russia 47 Table 2.10 Unit root tests for 1 and 2

ADF Specification1 LM1 (LM4)2 PP Specification1

1

2

6.57* 0, 0(1) 1.98 (2.82) 9.26* 0, 0(3)

4.82* 0, 0(2) 0.35 (6.25) 9.26* 0, 0(3)

Notes: * (**) significant at the 1 (5) per cent level. 1. Indicates whether a linear trend (T) and/or a constant term (C) has been introduced and the maximum lag length of the dependent variable (in parenthesis). 2. Chi-square values of an LM test on first (fourth) order autocorrelation. The critical values at the 5 per cent confidence level are 3.84 (9.49).

Table 2.11 Wald test of the coefficient restrictions: (c2, )  (0, 1)

F-statistic Probability Chi-square Probability

ARCH(1, 0)

ARMA(6,1)

0.735 0.484 1.469 0.480

0.428 0.654 0.856 0.652

rejected in either of the models. Hence expectations of inflation are unbiased in both models. Similarly, neither the Wald test for the null hypothesis H0 : i  0, i  1 (Equation 2.7) nor the LaGrange Multiplier (LM) test for zero correlation in forecast errors could reject the notion of rationality of expectation of inflation (Table 2.12).27 Moreover the null hypothesis from Equation 2.8 could not be rejected, as shown in Table 2.13.28 Nevertheless the strong form efficiency test (Equation 9) rejects the notion of rational expectation of inflation for both models. Table 2.14 presents the OLS estimates of the inflation regressed forecast errors (j, t) on the slightly modified monetary model in Nikolic (2000a).29 Efficient forecasting requires that the coefficients of all the information variables are equal to zero. A non-zero coefficient implies that forecasters could have improved their predictions by better exploiting the information set

48 Table 2.12 Tests for zero correlation and non-significant zero mean

LM 1 (4) Probability 1(4) Wald F Probability Wald 2 Probability

ARCH(1, 0)

ARMA(6,1)

0.000 (3.008) 1.000 (0.556) 0.945 0.334 0.945 0.331

0.345 (6.249) 0.557 (0.181) 1.154 0.334 3.463 0.326

Table 2.13 Test for lack of serial correlation and nonsignificant zero mean

LM 1 (4) Probability 1(4) Wald F Probability Wald 2 Probability

ARCH(1, 0)

ARMA(6,1)

0.947 (5.200) 0.331 (0.267) 1.282 0.262 1.282 0.258

2.293 (2.293) 0.682 (0.682) 0.343 0.560 0.343 0.558

Table 2.14 OLS estimates of the strong form efficiency test, February 1992 to July 1998 ARCH(1,0)

t1 m2xt1 m2xt2 m2xt3 m2xt4 DO94 SBFEB94 R2 R2 adjusted Standard error Durbin–Watson F Probability F LM 1(7) Probability LM 1(7)

ARMA(6,1)

Coefficient

t-statistic

Coefficient

t-statistic

0.144 0.077 0.058 0.096 0.054 6.146 0.628

3.248* 1.997 1.537 2.573* 1.432 3.360* 2.305*

0.104 0.060 0.093 0.146 0.038 2.142 0.522

2.952* 1.927 2.739* 4.369* 1.124 1.253 2.109*

0.356 0.294 1.790 2.247 5.722 0.000 0.849 0.357

* Significant at the 5 per cent level or better.

(16.920)* (0.018)

0.390 0.334 1.682 2.272 6.927 0.000 3.097 0.078

(16.267)* (0.023)

Inflation Expectations in Postcommunist Russia 49

It1 or its subset. Rationality also requires non-correlated error terms in these equations. An autocorrelated forecast error would imply that predictions could be improved by simply taking account of this phenomenon when generating predictions. Since the values of the F-statistics for both models are not significant we can reject the hypothesis that the coefficients of all the information variables are equal to zero. Similarly, although the Breusch–Godfrey serial correlation LM test does not indicate the presence of a first-order serial correlation for either of the models it does so for the seventh order. Furthermore rationality is undermined by the reasonably high coefficients of determination. Thus having established the non-zero coefficients of the information variables and the presence of autocorrelation we can reject the rationality hypothesis in its strong form.

Policy implications and further research The persistence of inflation30 in the Russian economy not only highlights the accommodating nature of Russian monetary policy but also has strong implications for stabilization policy. The latter was neglected prior to the summer of 1995, with adverse but predictable outcomes. Orthodox money-based stabilization programmes are not usually noted for their high credibility and transparency. Given the persistently high inflation, this stabilization approach might have not been the most appropriate one to achieve rapid disinflation in Russia, particularly if there was little commitment to fiscal adjustment. The slowness with which money-based programmes reduce inflation, when they do at all, and their high costs in terms of output and unemployment do not favour their implementation in countries where chronic inflation is persistent. In contrast exchange-rate-based heterodox programmes are able quickly to break up inertia and inflationary expectations and therefore bring down inflation to low levels. Since it appears that inflation in Russia has been a result of both inertia and inflationary expectations, exchange-rate-based heterodox stabilization might have been appropriate, although there would still have been difficulties. Such a programme would have required sizable foreign currency reserves, which Russia did not have at the beginning of the stabilization attempt. In addition some heterodox elements, such as wage and price controls, may have been difficult to enforce. Furthermore they could have caused considerable rigidity in relative prices and proved hard to phase out. Despite these shortcomings it appears that this type of programme, with external financial assistance,

50 Monetary Policy in Transition

would have performed better than orthodox money-based stabilization if introduced earlier. As the experience of July 1995 indicates, such a programme would have had a clear dampening effect on inflationary expectations and inertia. Needless to say a strong commitment to fiscal adjustment, as well as it having a high likelihood of being followed, would need to be a conditio sine qua non for the success of such a programme, as indeed is the case for any stabilization programme. Although both of our models exhibit desirable statistical properties, the volatility of inflation in Russia during the sample period indicates that the Markov-switching model would be a suitable alternative.31 In contrast to conventional modelling, this approach allows for the possibility of structural change. and for two or more regimes to prevail over time. That is, there is a regime (or a state) when inflationary expectations are low and one or more regimes when they are high. Shifts between the regimes (timing of structural breaks) occur in a probabilistic fashion and are therefore endogenous rather than being imposed by the researcher. Such a model may well be able to pick up endogenous shifts in the level of inflationary expectations and produce superior results. It could also be complemented by explanatory variables such as unemployment, money supply and output gap. It is likely that such a model may prove rational not only in the weak sense but also in terms of the strong efficiency criteria described above.

Conclusion Contrary to earlier assertions, this chapter has clearly demonstrated that lagged inflation in the Russian economy was a very important determinant of the rate of contemporaneous inflation. In both the ADL and ARMA models of inflation expectations, lagged inflation accounted for more than 90 per cent of the variations in contemporaneous inflation. Economic agents could easily have utilized both these simple models to make consistent forecasts of the inflation rate in one month’s time. Although they would have been unbiased and weakly efficient the forecasts would not have been rational in terms of the strong form of efficiency. That is, in addition to previous inflation rates, economic agents could have used other information at their disposal to improve their forecasts. Having identified a high degree of inflation persistency in the Russian economy, this chapter has suggested that instead of money-based stabilization measures, exchange-rate-based stabilization measures with heterodox elements might have been more suitable for Russia in terms of dampening inflationary expectations and dispelling

Inflation Expectations in Postcommunist Russia 51

inflationary inertia. The findings indicate that inflationary expectations in a transition economy may not be different from those in a market economy, but the institutional and historical legacies of the former socialist system should not be overlooked. Finally, one has to be aware that the inability of autoregressive models to predict sharp downturns and upturns limits their value for forecasting. In particular they depend on government policies remaining more or less constant during the period under consideration. In other words, and following the ‘Lucas critique’, expectations obtained by mechanical extrapolation of the past value of a variable may fail to capture changes in the formation of expectations due to a policy change (Lucas, 1981). Furthermore measurement errors, too few observations and the questionable reliability of Russian statistical sources could further undermine the accuracy of estimated inflationary expectations. However in the absence of long-term series of financial market indicators and sample surveys, the forecasting alternatives discussed in this chapter are particularly suitable for short-term forecasts.

3 The Relationship between Money Supply and Inflation in Postcommunist Russia1

Introduction The relationship between the growth of monetary aggregates and price rises is one of the best documented relationships in the economics literature. In the simple analytical framework of Fisher’s (1963) variant of . the quantity equation, prices (P) may increase as a result of growth of the . . money supply (M), an increase in the velocity of money (V) or a decline in . the total physical volume of transactions (T), assuming in each case that other two factors remain unchanged:2 ˙  V˙  T˙ P˙  M

(3.1)

In practice changes in the other two factors may either add to or offset the initial disturbance, so empirical researches often adopt the convention . . of naïve quantity theory (Locke, 1823), where V and T are treated as . . constants and P varies in direct proportion to M. Such a representation makes explicit the role of time, and thereby facilitates the study of the effect of monetary change on the temporal pattern of inflation. Numerous empirical studies have found consistent money–price relationships in various market economies. Monetarists argue that these studies support their claim that inflation is produced only when there is a more rapid increase in the quantity of money than in output, and therefore it is a purely monetary phenomenon (Friedman, 1989). While this proposition may be valid for market economies it does not hold for the majority of transition economies. Except in the case of Russia and Poland, no systematic pattern in the money–price relationship has been detected in transition economies, thus undermining the conventional monetarist view in the transitional context (Economic Commission for Europe, 1995). 52

The Relationship between Money Supply and Inflation 53

Although it would be interesting to examine this relationship in all transition economies with the benefit of the more extensive data now available, this chapter focuses solely on Russia. The main objective is to scrutinize the strength and dynamics of the relationship between inflation and monetary aggregates in postcommunist Russia. In particular we shall examine whether this relationship is significant and whether the lower inflationary environment that emerged in Russia from 1994, and especially in 1995, caused the transmission of monetary impulses to future inflation to become both slower and weaker. Evidence of the strength of this relationship and insights into the dynamics may, in certain conditions, be an important aid for the designers of monetary policy.3 The chapter is organized as follows. The next section briefly discusses Russia’s monetary policy to date and reviews the literature on the topic. In the third section the data, methodology and criteria for model selection are outlined. The empirical results are reported and discussed in the fourth section, and the fifth analyzes the relationship between money and prices during the early and later phases of transition. The final section summarizes the findings.

A brief overview of Russian monetary policy and the related literature Unlike the majority of countries in Central and Eastern Europe, which were able to achieve relative macroeconomic stability soon after beginning their transition to a market economy, Russia took a long time to bring down inflation to acceptable levels. A number of, mainly halfhearted, stabilization programmes were successively implemented between the autumn of 1991 and the spring of 1994 to tighten monetary conditions and tame inflation, but these were followed by periods of monetary expansion and price rises. Thus macroeconomic stabilization was characterized by ups and downs rather than continuous improvement (Koen and Marrese, 1995). An examination of the balance sheet of the Central Bank of Russia (CBR) reveals that the main cause of the growth of the money supply after price liberalization in January 1992 was the provision of CBR credits to the government to finance the budget deficit, to commercial banks and to former Soviet republics. These credits were priced well below the market clearing interest rate, and the consequently high demand for loans at what amounted to negative real interest rates forced the CBR to follow the International Monetary Fund’s (IMF) recommendation and impose quantity constraints on credit growth in early 1993.

54 Monetary Policy in Transition

These constraints were accompanied by quantitative targeting of various aggregates of the money stock, a measure that had been introduced in 1992 and continued for years thereafter. The setting of a positive real interest rate in early 1993 was followed by the collapse of the ruble zone and a steady increase in political support for stabilization, which shifted the main goal of monetary policy from the short-term maintenance of output to the control of inflation. The ambitious March 1994 programme initially succeeded in bringing down inflation to single digits, but by the autumn of that year inflation rate had returned to double digits. Restrictive monetary policy reinforced by the unsustainably high interest rates led to the relaxation of monetary and fiscal policies in the summer of 1994. This in turn led not only to the exchange rate collapse in October 1994 but also to recognition that the CBR could not issue large credits without causing high inflation. Despite CBR interventions to stabilize the value of the ruble there was no explicit exchange rate target until 1995. In July that year an exchange rate band of 4,300–4,900 rubles per US dollar was announced as an integral part of a well-prepared stabilization programme. The signalling properties of an exchange rate target with respect to future inflation were strong, particularly in a country where foreign currency, predominantly US dollars, was extensively used as a unit of account. However the success of the exchange rate stabilization programme was by no means certain unless the fundamentals were brought under control, including the government deficit, which was successfully reduced from 18.9 per cent in 1992 to a respectable 4.9 per cent in 1995 (Cheasty and Davis, 1996). This considerable reduction in the budget deficit and the increased use of capital markets and foreign loans to finance the remaining deficit were the main reasons for the success of the 1995 stabilization programme in the short term. However, macroeconomic stability is always fragile in transition economies such as Russia and the situation subsequently deteriorated, as discussed towards the end of the section on Russia’s fiscal performances in Chapter 1. Although inflation remained well below 2 per cent per month in the three years after the 1995 stabilization, prices soared by more than 38 per cent in September 1998 and almost 12 per cent in December of the same year. Thereafter matters improved, with prices undergoing a downward trend and inflation averaging just over 4 per cent in the first five months of 1999. We turn now to the literature on the transmission of monetary impulses to prices. Russia and Poland can be viewed as special cases among transition economies in that their money supply has had a systematic

The Relationship between Money Supply and Inflation 55

impact on the inflation rate. Many researches (Easterly and Vieira da Cunha, 1994; Fisher, 1994; Koen and Marrese, 1995; Hoggarth, 1996; Allen et al., 1996; Korhonen, 1996, 1998; Korhonen and Pesonen, 1998 claim that since price liberalization at the beginning of 1992 inflation in Russia has clearly had monetary roots,5 and that ruble broad money (M2) has had a stronger influence on inflation than other monetary aggregates. Working with limited data sets these authors found that the acceleration of inflation after price liberalization corresponded closely to the growth of M2, with a lag of up to four months. However their findings differ in respect of the influence of the lags on inflation and in terms of the time period covered by the analysis.6 The atypically short time lag between increased money supply and inflation in Russia,7 according to Hoggarth (1996), conformed to the tenet that the length of the time lag between money growth and inflation is in direct proportion to the stage of transition of a country. Similarly Buch (1998) claims that the speed of transmission of monetary impulses into prices is inversely related to the sophistication of a country’s financial system, and this process was comparatively fast in Russia because its financial system was not very sophisticated. Hoggarth (1996) and Allen et al. (1996) suggest that a temporary fall in the inflation rate not only reduces the extent to which money growth affects future inflation but can also, by establishing a lower inflationary environment, permanently lengthen the time lag of monetary policy.8 Consequently the lag between M2 growth and inflation in Russia increased from three to four months in 1992–93 to six months in the lower inflationary environment of 1994 (Hoggarth, 1996). This finding is supported by Korhonen and Pesonen (1998), who have found that the lag between money growth and prices in Russia between January 1992 and October 1997 shifted gradually from lags of one to four months to longer lag lengths. Similarly Alan et al. (1996) conclude that, in the lower inflationary environment in both Russia and Poland in 1993–94, money velocity became more uncertain, with the implication that a change in broad money growth had a weaker but more prolonged influence on future inflation. This has implications for major economies where the present low rates of inflation could in principle increase the time lag of monetary policy (ibid.).

Data and methodological overview The method and the objectives of this analysis differ from those of previous studies. First, we employ the longest time series available to

56 Monetary Policy in Transition

date, covering the period from the liberalization of prices in January 1992 until the Russian financial crisis of August 1998. Second, in addition to using data for M2, which is said to have the strongest link to inflation, we also analyze the narrower and more easily controlled monetary aggregates: ruble cash (M0), the monetary base (MB) and extended broad money (M2X). Third, instead of arbitrarily choosing the lag length of money supply aggregates or applying the rule of thumb method, we employ statistical criteria to select the optimum lag length. Fourth, various statistical tests and diagnostic statistics are applied to test the performance and robustness of the model. Finally, measures of summary statistics are employed in order to quantify, albeit tentatively, the impact of changes in the growth of money on the temporal response of prices. The data used in this study come form Russian Economic Trends RECEP, (1993–99, 1994, 1995, 1996, 1997, 1998).9 The data consist of monthly observations of the consumer price index (CPI) and monetary aggregates: ruble currency outside the banks (M0); base money (MB), which is defined as M0 plus the minimum reserve requirements of the CBR on commercial banks; ruble broad money (M2), defined as M0 plus ruble deposits in the banking system; and extended broad money (M2X), defined as M2 plus foreign exchange deposits with domestic banks. As any empirical work on time series requires, all the variables have been examined for stationarity and cointegration. Not surprisingly, no variable in its original form was found to be stationary, nor were any of the original times series of monetary aggregates cointegrated with the CPI. In order to obtain stationary time series we employed the growth rates of each variable, which was approximately equivalent to combining differencing transformation, or ‘operator’ as it is often termed, with power transformation.10 The growth rates of CPI, M0, MB, M2, and M2X, denoted as , m0, mb, m2 and m2x respectively, were found not to exhibit a unit root; that is, all series were stationary at the conventional 5 per cent level of significance.11 The results of the standard unit root tests – augmented Dickey–Fuller (ADF) and Phillips–Perron (PP) – are presented in Table 3.1. In order to examine seasonality patterns in the data, all the variables have been regressed against a constant, a trend and seasonal dummies, Dk‘s. The latter were set equal to 1 for the month in which seasonal variation was observed and 0 otherwise; k  1 … 11 denotes the month in which the value for a dummy was set equal to 1, for example D1  January, D2  February and so on. While inflation series do not contain any seasonal component, all of the money supply measures but M2X do, as reported in Table 3.2. An additional dummy variable, DO94, has been included to account for the exogenously induced inflation that

The Relationship between Money Supply and Inflation 57 Table 3.1 Unit root tests, February 1992 to July 1998

ADF Specification1 LM1 (LM4)2 PP Specification1

m0

mb

m2

m2x



2.73* 0,0(11) 0.00 (7.55) 8.47* C,T(3)

5.19* 0,0(11) 1.03 (7.47) 6.67* C,T(3)

3.93* 0,0(11) 1.02 (3.04) 6.60* C,T(3)

4.02* 0,0(11) 1.26 (3.24) 6.63* C,T(3)

4.07** C,T(1)3 3.54 (9.01) 4.46* C,T(3)

Notes: *(**) significant at the 1 (5) per cent level. 1. Indicates whether a linear trend (T) and/or a constant term (C) has been introduced and the maximum lag length of the dependent variable (in parenthesis). 2. Chi-square values of an LM test on fist (fourth) order autocorrelation. The critical values at the 5 per cent level of confidence are 3.84 and (9.49). 3. Alternatively for the specifications 0,0 (13) ADF is 2,45. Also significant at the 5 per cent level, LM(1)  0.67, LM(4)  1.21.

Table 3.2 Seasonal dummies and optimum lag length (l*) selection OLS method

Independent variable

Seasonal dummies

SC ( l*)

AIC ( l*)

m0 mb m2 m2x

D1, D4, D6 D1, D4, D6, D7 D1, D6

6.2607 (8) 5.8492 (4) 5.9104 (8) 5.6556 (13)

5.8110 (8) 5.4674 (5) 5.4928 (8) 5.1203 (13)

m0 mb m2 m2x

D1,D4,D6 D1, D4, D6, D7 D1, D6

4.6918 (9) 4.5849 (5) 4.4063 (9) 4.5120 (9)

4.0411 (15) 4.1587 (6) 3.9206 (9) 4.0382 (12)

Equation 3.3

Equation 3.4

Note: The dependent variable is the rate of inflation (t ) in all estimates.

occurred in 1994 as result of the exchange rate crisis. This variable takes the value of 1 for October 1994 and zero otherwise. Following earlier studies (Easterly and Vieira da Cunha, 1993; Koen and Marrese, 1995; Hoggarth, 1996; Allen et al., 1996; Korhonen and Pesonen, 1998) changes in prices () are modelled as a function of each of the aggregates of money supply (m). All the variables are in growth form and the following model is estimated: t  c0 

N

兺 m

j0

j

tj



S

兺 D

k1

k

kt

 DO94t  et

(3.2)

where c0 is a constant; the betas are parameters of money supply to be estimated; subscripts j and t denote the lag length and the current time period respectively; N is the maximum number of lags; Dkt and k

58 Monetary Policy in Transition

are the seasonal dummies and their respective coefficients; S is the number of seasons; DO94 and  are the October 1994 dummy and its coefficient; and e is the stochastic error term that follows the classical assumptions, namely it has zero mean, constant variance and is not autocorrelated. This model is a version of the distributed lag (DL) model associated with the monetarist tradition and could be viewed as a restricted form of the Saint Louis equation.12 Like similar monetary models, it is not intended to explain each and every fluctuation in the rate of change of prices, but rather is a shorthand description of the fundamental inflation process.13 The equation implies that disequilibrium in the money market, for a given stock, is eliminated only by changes in the general level of prices. Unlike in structural models the exclusion of non-monetary factors in Equation 3.2 reflects the view that the potential exogenous shift variables have only temporary or short-term effects on the rate of inflation.14 In empirical analyses it is common practice to adopt some a priori restrictions on the beta coefficients in models such as the one depicted by Equation 3.2. This is usually done by assuming that the betas follow a systematic pattern. Thus researchers often assume that the beta coefficients either decline geometrically (the Koyck DL model),15 or follow a cyclical pattern (the Almon or polynomial DL model with or without restrictions).16 Since these models impose artificial patterns on the distribution of the beta coefficients over time that are not likely to coincide with their true distribution, their usefulness for our purpose is limited. Therefore as our goal is to establish the true distribution of the beta coefficients, which reflects the relative importance or weight of each lagged value (j) of m, we proceed without imposing any restrictions on the distribution of the betas.17 Inclusion of a single, one-term, lagged dependent variable (t1) as an explanatory variable yields an autoregressive dynamic model (Hoggarth, 1996; Allen et al., 1996; Korhonen and Pesonen, 1998). This model is specified by Equation 3.3: t  c0  1t1 

N

兺 m

j0

j

tj



S

兺 D

k1

k

kt

 DO94t  et

(3.3)

This procedure can be justified by observing that t  1 contributes a great deal to the contemporary inflation rate, and thus the model captures the significant inflationary inertia prevalent in the Russian economy. Since values of the beta coefficients in Equations 3.2 and 3.3 are likely to be sensitive to the choice of the lag length, careful attention must be

The Relationship between Money Supply and Inflation 59

paid to the manner in which lag length is set. In the majority of empirical studies on this subject, either an ad hoc approach or a rule of thumb method is applied to specify the lag length. In contrast we employ the Schwarz (Bayesian) criterion (SC) and the Akaike information criterion (AIC) to determine the most appropriate lag structure for the model.18 Implementing the SC and AIC tests for the selection of the optimal lag length, l*, in Equations 3.2 and 3.3, with the maximum number of lags (N), set at 15, yields the results shown in Table 3.2.

Results Although not presented here, estimations of the distributed lag (DL) model in Equation 3.2 using the OLS method produced a high R2 of over 80 per cent for each of the monetary aggregates. R2 was greater than the Durbin–Watson statistic in all cases, pointing to the presence of serial correlation in the data. Following Maddala (1992), we re-estimated Equation 3.2 in first differences using the OLS method, and dropping the intercept term for all monetary aggregates. The results are presented in Table 3.3. The performance and the diagnostic statistics of the distributed lag model presented in Table 3.3 suggest that the OLS method may not give satisfactory results in that the statistics reveal that our assumptions about the error term, e, may not be satisfied. More specifically, e was assumed to have zero mean, constant variance and no autocorrelation. Visual inspection of the forecast error revealed their large variability over time. This suggests that the variance of the forecast error was not constant but varied from period to period. In other words there seemed to exist a particular kind of heteroscedasticity in which the variance of the regression error depended on the volatility of the errors in the recent past. Indeed the null hypothesis of no correlation in the variance of the OLS residuals, formally tested by the LaGrange Multiplier (LM) test for autoregressive conditional heteroscedasticity (ARCH), was rejected for all monetary aggregates at the conventional level of significance. We also calculated the coefficient of skewness (a test of symmetry) and kurtosis (a test for fat tails) for the estimated residuals and performed a Jarque-Bera normality test.19 Since, in the presence of heteroscedasticity, biased and inconsistent estimation of the variances of the OLS parameters render statistical inferences invalid, Engle (1982) suggests that the use of an ARCH model will improve efficiency. In such a model, Equation 3.2 has to be complemented by an additional equation that relates the variance ( t2) of the

60 Monetary Policy in Transition Table 3.3 OLS estimates of the distributed lag model of inflation, in first differences, February 1992 to July 1998 m0 Coeff.

mb t-stat.

Coeff.

m2 t-stat.

Coeff.

m2x t-stat.

Coeff.

t-stat.

 0.096 1.93 0.045 0.71 0.046 0.92 0.024 0.60 1 0.159 3.08* 0.089 1.40 0.243 4.75* 0.048 0.99 0.118 2.39* 0.068 1.08 0.144 2.53* 0.018 0.35 2 0.079 1.76 0.043 0.67 0.161 2.76* 0.114 2.28* 3 0.121 2.67* 0.077 1.48 0.211 4.32* 0.161 3.43* 4 0.167 3.31* – – 0.203 3.80* 0.118 2.54* 5 0.177 3.57* – – 0.193 3.92* 0.028 0.60 6 0.108 2.68* – – 0.063 1.41 0.033 0.82 7 0.008 0.26 – – 0.006 0.16 0.006 0.16 8 ; 7.796 3.84* 6.741 3.22* 7.188 4.32* 6.358 3.47* 4.027 3.32* 1.021 0.99 1.976 2.32* 1 1.552 1.43 0.349 0.30 – – – – 4 0.264 0.27 0.521 0.52 0.461 0.61 – – 6 – – 0.414 0.46 – – – – 7 0.41 0.28 0.55 0.40 R2 0.28 0.17 0.47 0.31 Adjusted R2 Standard error. 1.86 2.00 1.60 1.82 Fstatistic 3.19* 2.56* 6.44* 4.38* Log likelihood 133.60 140.21 123.72 133.90 Durbin–Watson 1.66 1.75 1.61 2.04 LM 1 (4) 0.36 (1.62) 0.00 (4.643) 1.97 (6.45) 0.00 (1.028) ARCH LM (lag) 9.256 (1)* 6.83 (1)* 7.599 (1)* 12.46 (1)* Skewness 0.324* 0.125 0.214 1.030* Kurtosis 3.392 4.110* 3.054 6.959* Jarque–Bera 1.651 3.724 0.534 57.251* SC (l*) 4.670 (8) 4.678 (4) 4.322 (8) 4.495 (8) Long run multiplier 1.03 0.23 1.27 0.50 Mean lag 3.71 2.85 3.56 4.20

Note: Asterisks indicate significance at the 5 per cent level or better. Asterisks associated with the coefficients of skewness and kurtosis imply that the coefficient exceeds twice its standard error. The asymptotic standard errors of the skewness and kurtosis coefficients are 兹(6/T) and 兹(24/T) respectively, where T is the sample size. The Jarque–Bera normality statistic is distributed as chi-square with two degrees of freedom.

error term (et) to the amount of volatility observed in the recent period. This ARCH(p) model is written in the following form:20 2t  0  1e2t1  2e2t2  … pe2tp

(3.4)

The presence of ARCH effects can be, and usually is, corrected by the maximum likelihood estimation (MLE). Assuming normality and ARCH errors, the MLE estimates are determined by an iterative search, which amounts to maximizing the log likelihood function (l) of the following form: lt   0.5 log t2  0.5 e2t /t2

(3.5)

The Relationship between Money Supply and Inflation 61

An alternative ARCH specification can be obtained by minimizing the Schwarz criterion of predictive accuracy. In our case, both criteria provide similar or equivalent information on the exact linear specification of the conditional variance function for all monetary aggregates, apart from the monetary base.21 However neither criterion seems to give an ARCH/GARCH specification that makes growth of the monetary base a good proxy for inflation. Specification of the equation involving mb, obtained by the Schwarz criterion, which is more parsimonious, is presented in Table 3.4 along with the other orders of the ARCH/GARCH models for the rest of the monetary aggregates.22 Table 3.5 lists the Q-statistics for the standardized and squared residuals. The performance and diagnostic statistics obtained form the distributed lag model of inflation corrected for ARCH effects, reveal that like the monetary base, the growth of ruble cash money does not seem to be a Table 3.4 Maximum likelihood estimates of the distributed lag model of inflation m0 Coeff.  1 2 3 4 5 6 7 8 ; 1 4 6 7 0 1 2 1 R2 Adjusted R2 Standard error F-statistic Log likelihood Run test Skewness Kurtosis Jarque–Bera SC (l*) Long run mult. Mean lag

0.029 0.028 0.006 0.022 0.003 0.018 0.018 0.065 0.014 7.083 0.676 0.505 0.035 – 0.010 0.025 0.582 0.538 0.167 0.089 2.288 0.653 95.142 0.058* 0.202 3.119 0.512 3.801 (8) 0.15 4.76

mb t-stat. 1.12 1.22 0.29 1.14 0.12 1.09 1.23 3.64* 1.81 0.50 1.17 1.79 0.14 – 0.78 0.20 2.29* 4.22*

Coeff. 0.017 0.007 0.030 0.009 0.016 – – – – 4.925 0.533 0.023 0.321 0.003 0.020 0.223 – 0.780 0.182 0.007 2.185 1.041 108.855 0.534* 0.234 3.235 0.787 3.953 (4) 0.03 0.47

* Significant at the 5 per cent level or better.

m2 t-stat.

m2x

Coeff.

t-stat.

0.71 0.037 0.23 0.113 0.94 0.036 0.32 0.084 0.76 0.150 – 0.120 – 0.076 – 0.023 – 0.000 1.32 10.493 0.58 0.968 0.04 – 1.01 0.302 0.01 – 1.44 0.173 1.56 1.078 – 1.014 9.01* 0.621

2.06* 3.68* 1.44 3.38* 10.35* 5.46* 3.80* 1.28 0.03 8.18* 2.36* – 1.25 – 1.22 2.53* 3.62* 3.67*

0.450 0.334 0.295 0.177 1.841 1.989 2.893* 2.12* 99.785 104.290 0.493* 0.163* 0.353* 0.017 2.183 3.304 3.352 0.270 3.874 (8) 3.882 (8) 0.57 3.99

0.64 3.81

Coeff.

t-stat.

0.036 1.33 0.072 1.54 0.058 1.43 0.101 2.85* 0.137 8.35* 0.115 8.37* 0.040 2.88* 0.048 2.34* 0.029 0.96 6.670 0.37 – – – – – – – – 0.008 0.79 1.008 8.18* 0.892 8.73* 0.906 12.97*

5

6

7

8

9

10

11

12

13

14

0.70 1.01 3.06 4.05

1.61 2.16 1.81 1.94 7.75 10.48 6.22 6.26

* Significant at the 5 per cent level or better.

0.63 0.99 3.06 0.50

2.55 1.96 11.11 6.36

4.75 2.24 11.11 6.43

4.79 2.60 11.20 6.49

4.84 2.73 11.25 6.92

5.14 3.00 11.35 7.74

6.04 4.41 11.37 7.76

6.18 4.41 11.37 7.97

6.23 4.63 11.55 9.78

4

Squared residuals m0 0.00 0.43 mb 0.01 0.46 m2 0.51 2.96 m2x 0.10 0.25

3

lag

12.05 26.44* 9.94 18.45

2

Q-statistics for the standardized and squared residuals

Standarized residuals m0 3.02 3.02 3.50 5.38 6.33 7.16 7.18 7.35 7.43 10.24 11.08 11.08 11.08 mb 2.08 3.97 9.46* 16.4* 16.5* 16.78* 17.94* 18.05* 18.24* 22.76* 22.76* 25.30* 25.41* m2 0.14 0.23 0.42 1.15 2.41 2.42 2.83 2.83 3.89 4.69 7.83 7.85 8.68 m2x 0.12 0.15 0.15 1.12 2.13 6.56 7.36 8.92 8.94 9.57 9.67 13.95 13.96

1

Table 3.5

6.43 4.81 12.50 9.80

12.10 26.45* 10.41 18.76

15

6.46 4.89 12.61 10.95

13.06 26.47* 12.73 18.82

16

18.17 27.93 15.28 21.79

18

7.02 7.53 4.91 6.80 14.56 15.38 11.00 11.07

16.45 27.13 14.02 20.05

17

62

The Relationship between Money Supply and Inflation 63

good proxy for inflation. Although the coefficient for seven months’ lagged ruble cash seems to be highly significant, the overall test of the significance of the regression line (that is, the F-statistic) indicates that collectively the coefficients are not different from zero. In addition the Q-statistics for the standardized residuals for mb are significant, implying the presence of autocorrelation. In contrast the diagnostic statistics for the two measures of broad money supply, m2 and m2x, suggest that both can explain variations in the inflation rate. Neither the Q-statistics for the standardized and squared residuals nor the Run test statistics justify rejection of the null hypothesis of no significant correlation. While the coefficient of skewness associated with the equation containing m2 marginally exceeds twice its standard error, neither the ARCH LM test (not reported) nor the Q-statistics for the squared residuals reveal the presence of ARCH effects. With regard to the variations in the current inflation rate, the growth of lagged m2 explains just under a half of them, somewhat more than are explained by m2x. Most of the coefficients and all of the ARCH/GARCH terms are statistically significant for both aggregates. However the significantly negative coefficient of current m2 seems to run counter to economic theory. Nevertheless the general requirement for all betas to be positive is somewhat too stringent. It may be possible that a number of short-term shocks – such as administrative price adjustments for oil and gas, exchange rate swings, changes in the monetary policy regime, or individual income policy decisions – may render the lag structure unstable if the sample period is relatively short. Perhaps more importantly and more likely, inflation may initially overreact to a change in the money supply and compensating adjustments may be needed, as appears to have been the case with our sample. As with the DL model, the OLS method does not seem appropriate for the autoregressive distributed lag (ADL) model (Equation 3.3). The ARCH LM test and the Q-statistics for the squared residuals (not reported), indicate the presence of ARCH effects for all monetary aggregates (Table 3.6). The exact order of the ARCH and GARCH models is determined by the procedure employed for the DL model. Both the l and the Schwarz criterion suggest the identical order of ARCH/GARCH models. The estimates from the ADL model (Equation 3.3) corrected for ARCH effects, and the Q-statistics for the standardized and squared residuals are presented in Tables 3.7 and 3.8 respectively. The ARCH LM test (not presented) and the Q-statistics for the residuals of each of the aggregates of the money supply do not reveal the presence

0.96 0.95 1.76 96.60* 128.00 3.01 (3.91) 10.66 (1)* 0.002 2.866 0.052 4.692 (9)

1.37 15.23* 1.69 1.69 0.95 0.16 1.35 1.87 0.63 1.64 1.96* 0.28 4.18* 3.84* 0.52 0.50 —

0.558 0.869 0.089 0.057 0.036 0.007 0.046 0.074 0.020 0.060 0.066 0.009 8.088 4.466 0.624 0.570 —

Coeff.

mb

0.96 0.95 1.78 117.76* 130.658 4.56* (7.62) 8.76(1)* 0.371* 4.055* 4.786 4.585 (5)

0.631 0.779 0.030 0.095 0.019 0.019 0.071 0.019 — — — — 7.150 2.797 0.071 0.261 0.229

Notes: * Significant at the 5 per cent level or better. 1. Heteroscedasticity consistent covariances (White, 1980).

R2 R2 adjusted Standard error F-statistic Log likelihood LM 1 (4) ARCH LM (lag) Skewness Kurtosis Jarque–Bera SC (l*)

c   1 2 3 4 5 6 7 8 9  1 4 6 7

t-stat.

Coeff.

m0

1.62 12.50* 0.44 1.88 0.37 0.29 1.60 0.36 — — — — 3.80* 2.73* 0.06 0.23 0.26

t-stat.

m2

0.972 0.964 1.563 132.924* 120.261 3.735 (9.859)* 9.230(1)* 0.093 3.018 1.001 4.406 (9)

0.545 0.876 0.038 0.178 0.072 0.033 0.075 0.024 0.003 0.099 0.037 0.014 7.043 2.195 — 0.086 —

Coeff.

Table 3.6 OLS estimates of the autoregressive distributed lag model of inflation

1.47 14.92* 0.72 3.68* 1.60 0.80 1.82 0.54 0.06 2.36* 0.91 0.36 4.22* 2.60* — 0.10 —

t-stat. 2.03* 8.80* 0.75 1.14 0.48 3.09* 2.99* 0.10 1.08 0.55 0.04 0.65 13.17* — — — —

t-stat.

0.965 0.957 1.720 127.027* 128.141 0.153(1.951) 16.449 (1)* 0.143 5.801* 22.79* 4.5120 (9)

0.401 0.818 0.052 0.055 0.020 0.108 0.085 0.004 0.058 0.034 0.002 0.030 6.509 — — — —

Coeff.

m2x1

64

The Relationship between Money Supply and Inflation 65 Table 3.7 Maximum likelihood estimates of the autoregressive distributed lag model of inflation m0

c   1 2 3 4 5 6 7 8 9  1 4 6 7 0 1 2 1

mb

m2x1

m2

Coeff.

t-stat.

Coeff.

t-stat.

Coeff.

t-stat.

Coeff.

t-stat.

0.157 0.930 0.029 0.008 0.033 0.034 0.015 0.030 0.007 0.021 0.048 0.027 6.930 0.952 0.558 0.007 — 0.002 0.275 — 0.690

1.28 25.09* 1.00 0.39 1.61 1.40 0.76 1.46 0.78 1.10 2.18* 1.80 1.54 1.69 1.34 0.02 — 0.27 1.74 — 6.44*

0.033 0.782 0.046 0.019 0.022 0.022 0.029 0.065 — — — — 8.926 1.539 0.718 0.091 0.236 0.000 2.684 — —

0.66 51.91* 3.30* 3.84* 3.33* 1.86 3.13* 6.07* — — — — 0.39 6.12* 4.13* 0.77 2.91* 0.02 4.18* — —

0.313 0.886 0.017 0.041 0.064 0.034 0.068 0.044 0.027 0.029 0.026 0.013 6.657 1.400 — 0.502 — 0.001 2.319 0.006 —

6.28* 32.89* 0.95 2.55* 3.19* 1.71 6.01* 3.28* 1.94 3.39* 1.94 0.98 0.72 6.07* — 2.79* — 0.06 3.90* 0.15 —

0.462 0.798 0.028 0.047 0.010 0.071 0.070 0.009 0.036 0.001 0.004 0.028 7.418 — — — — 0.202 1.456 — 0.006

5.21* 18.65* 0.99 3.40* 0.68 9.39* 5.89* 0.51 1.91 0.06 0.30 1.48 34.88* — — — — 2.78* 4.95* — 0.802

R2 0.941 Adjusted R2 0.919 Standard error 2.357 F-statistic 44.014* Log likelihood 84.849 Run test sig. 0.546* Skewness 0.176 Kurtosis 3.110 Jarque–Bera 0.389 SC (l*) 3.626 (9)

0.948 0.934 2.131 68.689* 91.242 0.043 0.312* 2.548 1.705 3.565 (5)

0.962 0.950 1.858 76.679* 89.812 0.546* 0.054 2.255 1.632 3.708 (9)

0.960 0.949 1.867 85.906* 100.484 0.118* 0.087 2.837 0.164 3.894 (9)

Notes: * Significant at the 5 per cent level or better. 1. Heteroscedasticity consistent covariances (White, 1980).

of ARCH effects. Nevertheless the performance of the models containing m0 and mb is somewhat inferior to those which include the broad money variables. Less generally, in the Table 3.7 there is no positive significant coefficient of m0 and the model containing mb seems to be plagued by autocorrelation. Even in the model containing m2 a number of negative values for the coefficients for money, of which two are significant, run counter to the postulates of economic theory. Hence one may tentatively conclude that the ADL model containing m2x, corrected for ARCH effects, is a pretty good determinate of inflation in postcommunist Russia.23

5.40 1.65 1.94 3.97

1.51 1.26 4.61 2.43

5

5.52 6.31 2.90 3.99

1.57 1.32 4.86 6.11

6

5.91 6.31 3.74 4.89

1.57 5.89 5.05 6.71

7

Note: * Significant at 5 per cent or level better. 1. Heteroscedasticity consistent covariances (White, 1980).

1.63 0.13 0.51 1.39

4.76 0.47 1.92 3.95

4

Squared residuals m0 0.51 0.78 mb 0.04 0.13 m2 0.01 0.34 m2x 0.95 1.37

3

1.42 0.94 4.55 2.38

2

9

10

lag

7.05 8.86 9.55 7.28 7.53 7.53 3.98 3.98 4.01 6.33 7.58 7.58

3.04 3.54 4.65 6.15 6.89 9.95 5.11 5.19 6.50 7.22 7.61 7.67

8

Q-statistics for the standardized and squared residuals

Standardized residuals m0 1.19 1.25 1.30 mb 0.64 0.65 0.66 m2 0.42 1.19 1.85 m2x 0.62 1.10 1.85

1

Table 3.8

9.71 7.56 4.56 8.09

5.26 10.06 7.23 8.37

11

9.77 7.58 5.28 9.60

5.43 10.08 7.84 11.05

12

11.79 7.66 5.96 10.40

5.49 10.27 8.61 12.71

13

11.80 8.12 5.98 11.21

8.11 10.62 12.09 16.05

14

8.82 10.67 15.52 16.25

16

11.99 12.33 18.50 18.95 6.25 6.92 11.24 11.58

8.12 10.63 13.44 16.05

15

12.87 18.96 7.40 12.09

10.61 10.72 16.59 16.68

17

13.01 19.33 8.79 16.41

10.61 10.72 21.81 17.08

18

66

The Relationship between Money Supply and Inflation 67 Table 3.9 Scaled recursive Chow test (SRCT) and one-step forecast stability test (O-SFT) for the period January–October 1994

m2 (DL) m2x (ADL)

Jan.

Feb.

Mar.

Apr.

May.

Jun.

Jul.

Aug.

Sep.

Oct.

1.16

1.18

1.2

1.16

1.15

1.18

1.05

1.00

–

–

1.27

–

–

–

–

–

–

–

1.01 (6.64) (6.08)

Notes: The values in parenthesis are the O-SFT values of the recursive residuals that lie outside the two standard error bounds. As the Chow statistic and the critical values of F are functions of time, we divided the Chow value by its 5 per cent critical value from the tables of F to yield an SRCT for recursion. Values greater than unity imply that the null hypothesis of no structural change between periods t-1 and t would be rejected at the 5 per cent level of significance.

Having established that the monetary aggregates m2 and m2x are the best proxies for inflation in the DL and ADL models respectively, to test whether the inflation functions had undergone any structural change24 we subjected both models to stability tests.25 The results of these tests are somewhat mixed, as demonstrated by the matrix of potential breaks in Table 3.9. All of the stability tests based on recursive estimations indicate that both models provide a poor fit for the October 1994 observation. This suggests either the presence of an outlier, or an exceptional value for t or an alteration in the structural parameters of the model. Thus inclusion of the dummy variable DO94 for October 1994 seems justifiable. However the SRCT indicates a lack of stability for the parameters of m2 in the DL model throughout 1994. The SRCT also suggests a structural break in February and October 1994 for the ADL model that includes m2x. The extent to which inclusion of the DO94 dummy affects the stability of the parameters of the models is unclear since these tests cannot be applied with this dummy included. In order to overcome this difficulty we applied the dummy variable technique (in the additive form) to test for structural stability. The dummy took the a value of 1 for observations on and after the suspected structural break and zero otherwise. The results of this test do not justify rejection of the null hypothesis of no structural break in 1994 in the DL model that includes m2 and the DO94 dummy. Hence this DL model provides a reasonably good representation of inflation. However the same test does point to a structural break in February 1994 for for ADL model that includes m2x. This is hardly surprising given that the Perron (1989) test presented earlier suggested a structural break in the inflation series in that month. Hence after accounting for the structural break and implementing of the exact

68 Monetary Policy in Transition Table 3.10 ARCH(1,0) autoregressive distributed lag model of inflation involving m2x c





1

2

3

4





Coefficient t-statistic

4.819 8.90*

0.648 25.27*

0.011 0.60

0.084 3.77*

0.049 1.80

0.035 1.55

0.102 6.69*

10.461 4.842 0.248 1.151 4.53* 11.18* 1.86 3.16*

R2 0.924

Adj.R2 0.912

S.E. 2.467

F Log 76.46* 119.38

Run 0.251*

Skew. 0.271

Kurto. 2.505

J–B 1.662

SC 3.866

0

1

(l*) 4

Lag 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Q-standardized 3.44 3.60 5.12 7.93 9.83 11.87 11.87 12.76 12.77 12.78 13.18 13.18 13.34 13.44 Q-squared 0.57 2.25 3.68 3.68 3.78 6.04 6.94 8.79 8.81 8.81 9.55 9.59 11.63 11.77 residuals Notes: Significant at the 5 per cent or level better.  is a coefficient of the dummy for the structural break in February 1994.

order of the ARCH term, the best final ADL model of inflation that includes m2x is ARCH(1,0) (Table 3.10).26 The time path of the recursive OLS estimates reveals that, after initial instability in the parameter estimates due to the small number of observations, all of the money coefficients of the variables in the DL and ADL models are either constant or decline slightly over time. This can be seen in Figure 3.1, which adds to the evidence on the diminishing influence of money on prices.27

Dividing the sample In order to compare the influence of the money supply on inflation during the two phases of transition we have divided the sample into two parts. The first (period I) includes the observations before October 1994 and the second (period II) the observations thereafter.28 The choice of the dividing point is governed by the fact that the one-step forecast test indicates the existence of a structural break in the models in October 1994.29 Furthermore it has been claimed (Hoggarth, 1996; RECEP, 1994) that in the second half of 1994 inflation started to follow the six-months lag in m2 growth instead of the three to four months lag that had prevailed in 1992–93. In order to verify this claim we have plotted the most highly correlated lags of broad money growth with the inflation rate for the two sub-periods. Figure 3.2 and the correlation statistics therein suggest that not only did a shift occur in the second part of 1994 but also the lagged response of prices to money became even longer than six months.

69 0.4

0.6

0.2

0.4

0

0.2

–0.2

0

–0.4

–0.2 1994

1995 m2

1996

1997

1994

1998

1995 m 2(–5)

±2 Standard errors

0.7

0.7

0.6

0.6

0.5

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

1996

1997

1998

±2 Standard errors

0

0 1994

1995 m 2(–1)

1996

1997

1998

1994

1995 m 2(–6)

±2 Standard errors

0.6

1996

1997

1998

±2 Standard errors

0.4 0.3

0.4 0.2 0.2

0.1 0

0 –0.1 –0.2

–0.2 1994

1995 m 2(–2)

1996

1997

1998

1994

1995 m 2(–7)

±2 Standard errors

0.8

1996

1997

1998

±2 Standard errors

0.4

0.6

0.2

0.4 0 0.2 –0.2

0

–0.4

–0.2 1994

1995 m 2(–3)

1996

1997

1998

1994

1995 m 2(–8)

±2 Standard errors

1996

1997

1998

±2 Standard errors

0.6 0.5 0.4 0.3 0.2 0.1 0 1994

1995 m 2(–4)

1996

1997

1998

±2 Standard errors

Figure 3.1 Recursive coefficient estimates of m2 in the distributed lag model

70 Monetary Policy in Transition 35

Correlation 0.59

30 25 20 15 10

Correlation 0.62

5 0

p

m2(t – 7)

May 98

Dec. 97

Jul. 97

Feb. 97

Sep. 96

Apr. 96

Nov. 95

Jun. 95

Jan. 95

Aug. 94

Mar. 94

Oct. 93

May 93

Dec. 92

Jul. 92

Feb. 92

–5

m2(t – 4)

Figure 3.2 Correlation between inflation () and lagged ruble broad money (m2), February 1992 to July 1998

Indeed the correlation coefficients (r) between inflation and lagged money growth suggest that the dominant response of the former to the growth of latter shifted from four months (r  0.59) prior to late summer 1994 to seven months (r  0.62) thereafter. Coincidentally or not, the dividing point is near enough to the time that Russia embarked on a path of relatively sustained monetary stability, which may arguably be taken as a break in the economic policy regime. According to the OLS estimates of the DL model involving m2 and m2x for the two sub-periods (Tables 3.11 and 3.12) for period I only m2 seems to represent the inflation process.30 Thus the model for period I appears to have all the properties deemed desirable by economic and statistical theory: the fit of the model is fairly good, all of the significant coefficients of money are positive and the diagnostic statistics do not indicate any model deficiency. Table 3.11 also reveals that both aggregates of broad money supply introduce ARCH effects in period II. However the specification applied to the entire period may not be appropriate for the sub-periods. Hence a new search for both the optimal lag structure using the Schwarz criterion and the exact order of the ARCH model for the two sub-periods has been performed. The tests indicate that the model involving m2 in period I, as presented in Table 3.11, is the optimal specification.

1.485 3.392

0.707 0.463 2.221 2.894* 43.509 1.667 0.763* 0.16 (4.42) No 4.28 (8)

0.025 0.328 0.180 0.213 0.257 0.211 0.250 0.039 0.018 3.668 0.651

0.25 3.47* 1.53 1.81 3.06* 2.10* 2.77* 0.46 0.22 1.73 0.34

t-stat.

0.639 4.675

0.379 0.196 1.113 2.073 62.365 1.262 0.058* 4.748* (9.065) 23.858 (4)* 3.70 (8)

0.015 0.135 0.021 0.006 0.075 0.096 0.128 0.115 0.060 0.855 0.597

Coeff. 0.327 1.600 0.301 0.103 1.170 1.182 1.598 1.737 1.270 1.234 1.521

t-stat.

Nov. 1994–July 19981

1. Heteroscedasticity consistent covariances (White, 1980).

Notes: * Significant at the 5 per cent level or better.

Long-run multiplier Mean lag

R2 Adjusted R2 Standard error F-statistic Log likelihood Durbin – Watson Run test (sig.) LM 1 (4) ARCH LM SC (l*)

 1 2 3 4 5 6 7 8 1 6

Coeff.

Feb. 1992– Sep. 1994

m2

0.209 5.042

0.394 0.048 2.956 1.140 51.853 2.214 0.498* 0.00 (1.387) No 5.736 (8)

0.066 0.002 0.020 0.106 0.109 0.116 0.048 0.031 0.021 – –

Coeff. 0.75 0.02 0.20 1.12 1.17 1.35 0.53 0.41 0.30 – –

t-stat.

Feb. 1992– Sep. 1994

0.085 0.214 0.049 0.062 0.158 0.187 0.179 0.142 0.065 – –

Coeff.

1.142 4.079

1.83 3.19* 0.67 0.85 2.18* 2.61* 2.63* 2.21* 1.40 – –

t-stat.

Mar. 1994– July 1998

0.405 0.272 1.059 3.06* 61.403 1.190 0.038 7.439* (8.497) 4.40 (1)* 4.227 (8)

m2x

Table 3.11 OLS of the first differences of the distributed lag model of inflation in the sub-periods

71

72 Monetary Policy in Transition Table 3.12

Q-statistics for the standardized and squared residuals Lag

m2 (I) m2 (II) m2x (I) m2x (II)

Standardized squared Standardized squared Standardized squared Standardized squared

1

2

3

4

5

6

7

8

9

10

11

0.25 0.41 5.52* 3.75 1.22 3.46 7.60* 4.77*

3.79 0.44 6.58* 5.22 1.23 5.54 7.62* 5.24

4.64 0.52 6.59 5.23 2.79 5.71 7.93* 5.25

4.69 2.76 7.07 5.23 3.15 6.87 7.97 5.35

4.99 2.79 7.09 5.23 3.55 7.75 7.98 5.35

5.28 3.94 7.10 5.33 3.56 9.00 9.28 5.55

– – 7.27 5.34 – – 10.73 5.63

– – 8.03 5.35 – – 11.11 5.76

– – 8.38 5.42 – – 11.23 5.80

– – 8.40 5.90 – – 11.24 6.17

– – 11.52 5.91 – – 11.77 8.16

* Significant at the 5 per cent level or better.

The remaining specifications and estimates are presented in Table 3.13 and the Q-statistics for the residuals in Table 3.14. Although the results of the Schwarz criterion lag selection are not reported here, Table 3.13 indicates that this criterion favors the model of inflation that includes only the current and one-month lagged values of either m2 or m2x. After correcting for ARCH effects the model does not comply with economic theory. That is, while the individual coefficients of the lagged values of both aggregates of broad money are individually significant, the overall significance of the coefficients (F-test) of all monetary aggregates for the two sub-periods is not different from zero. Hence these statistics indicate that there is not a stable linear relationship between inflation and m2x in period I, nor between inflation and either of the m2 or m2x in period II. In order to capture the intensity and speed of the transmission of monetary impulses to future inflation we have used several measures of summary statistics. These statistics, reported in Table 3.11, reveal important differences between the estimates with regard to the strength of the influence of the models’ coefficients on the dependent variable in the two sub-periods. The following analysis focuses on m2, since in a distributed lag model this monetary aggregate – despite its shortcomings in respect of period II – outperforms other aggregates in explaining inflation. Most notably, the percentage of the total variation in the inflation rate explained by the regression model (R2) for period I is almost twice that of its corresponding value for period II. The estimates of the coefficients of m2 are also considerably greater for period I than for period II. This difference is equal to the difference of the long-run multipliers which are in fact sums of the coefficients of money. The long-run

The Relationship between Money Supply and Inflation 73 Table 3.13 Maximum likelihood estimation of the distributed lag model of inflation in the sub-periods m2

m2x

Nov. 1994– July 19981 Feb. 1992– Sep. 1994

 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 6 0 1 1

Coeff.

t-stat.

0.009 0.047 – – – – – – – – – – – – – 0.406 0.254 0.005 0.103 1.046

0.35 2.22* – – – – – – – – – – – – – 1.90 2.65* 0.69 1.73 10.48*

R2 0.125 Adjusted R2 0.013 Standard error 1.249 F-statistic 0.908 Log likelihood 42.714 Durbin–Watson 1.364 LM 1 (4) NA ARCH LM No SC (l*) 2.49 (1)

Coeff. 0.163 1.274 0.762 0.299 0.659 0.603 0.363 0.173 0.482 0.083 0.133 0.275 0.104 0.244 0.134 – – – – – 0.942 0.537 2.216 2.328 19.459 3.229 13.556*() No 4.789 (14)

t-stat.

Mar. 1994– July 1998 Coeff.

t-stat.

0.17 0.004 1.50 0.064 1.32 – 0.62 – 1.28 – 1.30 – 1.76 – 0.52 – 2.32* – 0.58 – 0.61 – 1.82 – 0.34 – 1.40 – 1.08 – – – – – – 0.002 – 0.102 – 1.036

0.12 2.11* – – – – – – – – – – – – – – – 0.18 1.42 8.53*

0.140 0.054 1.208 1.625 46.206 1.342 NA No 2.477 (1)

Notes: * Significant at 5 per cent level or better. 1. Bollerslev–Wooldridge (1992) robust standard errors and covariance.

multipliers represent the change in the long run value of inflation caused by a unit increase in the growth of the money supply. The multiplier value of 1.49 for m2 in period I indicates that a 1 per cent increase in the growth of m2 is reflected in a 1.49 per cent increase in prices. In period II

74 Monetary Policy in Transition Table 3.14

Q-statistics for the standardized and squared residuals Lag

m2 (II) m2x (I) m2x (II)

Standardized squared Standardized squared Standardized squared

1

2

3

4

5

6

7

8

9

10

11

3.90* 0.52 8.26* 3.82 2.25 0.14

4.14 1.22 14.34* 3.82 2.25 1.42

6.62 2.56 15.59* 3.84 2.84 1.48

7.38 2.72 16.09* 3.84 3.29 1.54

7.51 2.80

7.72 3.71

7.72 4.28

7.76 4.83

7.93 4.95

8.22 7.09

8.30 7.63

3.65 1.75

4.32 2.66

4.52 2.78

4.63 3.57

4.73 3.57

4.75 5.16

4.76 5.25

* Significant at 5 per cent level or better.

the multiplier is considerably lower (0.639 per cent) and has a noticeably smaller impact on the inflation rate, as suggested by the recursive coefficient estimates in Figure 3.1. Moreover, in period II the multiplier is not statistically different from zero. However, even though the multiplier is notably different from 1 in both sub-periods, the Wald test cannot reject the null hypothesis of m2 being reflected one-for-one in price rises in either of the periods, or indeed for the entire period.31 The summary statistics presented in Table 3.11 also indicate a lengthening of the lag between money and prices. As a measure of speed the mean lag, expressed in the number of months in which prices respond to money on average, appears to be significantly greater in period II (4.68) than in period I (3.39).32 Similarly, while the inflation rate has the highest correlation with money growth four months earlier in period I (0.318), he highest correlation in period II (0.624) is with money growth seven months earlier, as shown in Figure 3.2.33

Conclusion Our analysis of the relationship between various monetary aggregates and inflation supports the earlier claim that broad money growth appears to have been strongly correlated with inflation in postcommunist Russia. However this relationship proved to be unstable and sensitive to the changes that were taking place in the new economic and institutional environment. In addition to other evidence of changes in this relationship, the summary statistics presented in this chapter suggest that the average speed of transmission of changes in the growth of ruble broad money to inflation increased from just over three months to just short of five months when Russia embarked on a path towards macroeconomic stability soon after the exchange rate crisis of October 1994.

The Relationship between Money Supply and Inflation 75

Similarly the summary statistics reveal that changes in the growth of broad money had a considerably greater impact on prices in the period prior to October 1994 than in the period thereafter. Furthermore the overall lack of significance of the coefficients of money in the later period points to a break in the systematic money – price relationship that prevailed for two and a half years after price liberalization. In contrast the impact of changes in the previous month’s inflation rate on current inflation did not abate, reflecting the considerable inflation inertia in the Russian economy and the persistence of inflation shocks. This finding calls for the inclusion of a one-month lagged inflation variable in the inflation model. Hence the autoregressive distributed lag model of inflation that includes m2x avoids the shortcomings of some of the models in the literature. It also provides a reasonably good shorthand description of the fundamental inflation process in Russia. Although our sample period of six and a half years is the longest considered so far, it may still be considered too short for meaningful analysis, particularly in these times of major transformation not only of Russia’s economic system but also of the entire country and region. In addition, as pointed out by many researchers, the accuracy of Russian statistics must be treated with caution and we should not rule out the possibility of model misspecification due to erroneous data. Therefore the validity of the conclusions drawn from this analysis rests on the adequacy of the sample length and the accuracy of the data for model specification.

4 Money–Inflation Causality in Transition Economies: The Case of Russia

Introduction Despite numerous empirical studies of the role of money in the evolution of market economies, the topic sparked considerable controversy when the new economic environment was created in the former socialist economies of Central and Eastern Europe. Among the unsettled policy issues that were unwittingly extended to transition economies were those related to Granger causality testing and the suitability of the chosen macroeconomic stabilization programmes. This chapter aims to shed additional light on the choice of lag selection criteria for causality testing on the one hand, and the suitability of monetary aggregates for influencing and controlling inflation via policy instruments in transition economies on the other.1 At one end of the spectrum of opinions on monetary policy, money is viewed only as a passive adaptor to business conditions and has little independent influence. At the other end, an opposing hypothesis holds that monetary actions, as measured by movements in the monetary aggregates, have a lasting effect on nominal variables such as GNP, output and price level. An influential tenet of the latter school of thought, branded monetarism, is the view that inflation, defined as a persistent increase in the general level of prices, results solely from expansion of the money stock at a rate that exceeds the increase in the amount of money demanded in the economy. In contrast to empirical evidence that supports the monetarist view for the majority of market economies, no systematic pattern for a money–price relationship has been detected in transition economies, apart from Russia and Poland, thus undermining the conventional monetarist view in the transitional context (Economic Commission for Europe, 1995). Hence the traditional tools used for 76

Money-Inflation Causality 77

economic stabilization and controlling inflation in advanced market economies may not be appropriate for transition economies. Given that high inflation is costly in terms of loss of output (Fisher et al., 1996), inappropriate policies to combat inflation may not only be detrimental to the welfare of societies in transition but may also cause an unnecessary delay in the transition process, which in turn may exacerbate the loss of welfare in the long term.2 The suitability of monetary policy, and particularly the appropriateness of monetary targeting in a transition economy such as Russia, is best settled by empirically testing the usefulness of the aggregates of money supply as an intermediate target for controlling inflation. In practice an intermediate target is effective if it (1) has significant explanatory power in determining the goal variable, (2) is exogenous or causally prior to the goal of policy actions, and (3) is reasonably controllable by policy makers. Since the first of these criteria appears to have been fulfilled in Russia, or at least in the early years following price liberalization, it remains to be determined whether broad money aggregates are causally prior to inflation and whether they are controllable by policy makers.3 The results of empirical investigations of the causal relationship between money supply and inflation, or indeed any other macroeconomic variable (for example income), seem to be largely determined by the length of their distributed lags in a bivariate model. Despite numerous empirical studies (Hsiao 1981; McMillin and Fackler, 1984; Thornton and Batten, 1985; Jones, 1989) there is no consensus on the criteria for selecting the lag length in causality testing. On the one hand Hsiao (1981) and Thornton and Batten (1985) claim that the statistical criterion given by Akaike (1970) is superior not only to the other statistical criteria but also to the ad hoc specifications that are common in literature. On the other hand Jones (1989) favours one of the ad hoc criteria and implies that the results obtained by Thornton and Batten (1985) were data specific. It seems that a good way of settling the controversy is to utilize an independent data set, for example for Russia instead of the United States, and evaluate the empirical results obtained by various criteria as well as an extensive portion of a lag space. This amounts to an empirical testing of the monetarist paradigm that in a transition economy the money supply causes inflation without a feedback. The remainder of this chapter is organized as follows. The following Section discusses the data and methodology employed in the analysis. In the third section the model and the criteria for its selection are outlined. The empirical results are reported and discussed in the fourth

78 Monetary Policy in Transition

section, and the fifth briefly reviews a few policy considerations in respect of Russia’s experience with monetary targeting. The final section summarizes the findings.

Data and methodological overview As with most analyses of inflation and money supply processes in Russia, this study uses data from Russian Economic Trends, published by the Russian European Center for Economic Policy. The choice of monetary variables is based on the finding that there was a systematic relationship between money supply and inflation in the first few years after price liberalization in January 1992 (Nikolic, 2000a). The data consist of monthly observations of the consumer price index (CPI), ruble broad money (M2), and extended broad money (M2X), from February 1992 to July 1998. M2 is defined as the sum of currency outside banks (as reported by the Central Bank of Russia’s money circulation department) and ruble deposits in the banking system, defined as the sum of demand, savings and time deposits of enterprises, the state insurance company, voluntary organizations and individuals, plus deposits for long-term capital investment, apart from government deposits. M2X comprises M2 and foreign exchange deposits with commercial banks. As any empirical work on time series requires, all the variables have been examined for stationarity and cointegration. As expected, no variable in its original form was found to be stationary, nor were any of the original time series of the monetary aggregates cointegrated with the CPI. In order to obtain a stationary time series we employed the growth rates of each individual variable, which was approximately equivalent to combining differencing transformation – or ‘operator’ as it is often termed – with power transformation.4 Since policy deliberations are couched in terms of growth rates, the use of growth rates seems appropriate. The growth rates of each time series CPI, M2, and M2X, denoted as , m2 and m2x respectively, were checked for the existence of a unit root by means of standard augmented Dickey–Fuller (ADF) and Phillips–Perron (PP) tests. In addition, since standard unit root tests are not very powerful – that is, they cannot distinguish between unit roots and near unit root stationary processes – we employed a more powerful test to test the null hypothesis of stationarity against the alternative of a unit root. This test was devised by Kwiatkowski, Phillips, Schmidt and Shin (1992) and is named after them (the KPSS test). The results of the standard unit root tests and the KPSS test for stationarity are presented in Table 4.1.

Money-Inflation Causality 79 Table 4.1 Unit root tests; February 1992 to July 1998  l 1 2 3 4 5 6 7 8 9 10

m2

m2x

ADF

PP*

KPSS

ADF

PP*

KPSS

ADF

PP*

KPSS

4.07 3.86 3.49 4.44 3.12 2.47 1.94 2.18 2.56 2.29

4.46 4.38 4.46 4.51 4.51 4.49 4.45 4.41 4.39 4.36

0.179 0.138 0.120 0.111 0.108 0.108 0.110 0.112 0.114 0.116

4.96 5.67 5.96 7.00 4.71 4.02 3.88 4.00 2.93 2.37

6.62 6.67 6.60 6.55 6.43 6.41 6.44 6.47 6.50 6.48

0.056 0.050 0.050 0.054 0.066 0.080 0.096 0.108 0.115 0.114

4.835 4.303 4.133 4.336 3.255 2.525 3.077 5.453 5.604 3.635

5.924 5.932 5.928 5.941 5.865 5.813 5.828 5.875 5.902 5.912

0.141 0.122 0.114 0.111 0.114 0.117 0.117 0.115 0.113 0.112

Critical values 1 per cent 5 per cent 10 per cent

ADF

PP

KPSS

4.082 3.469 3.161

4.080 3.468 3.161

0.216 0.146 0.119

* The lag truncation (l) for Bartlett kernel for , m2, and m2x is 10, 1 and 3, respectively, although Newey and West (1987) suggest 3 for all series.

All of the ADF test statistics were obtained by including a constant and a linear time trend (trend) into the regressions. It appeared that the ADF statistics were sensitive to the choice of truncation lag (l) for the inflation series. The critical values of the ADF test clearly allowed rejection of a null hypothesis of a unit root for the first four lags at a conventional level of significance.5 Following Perron (1989), l was selected if the t-statistic for the coefficient of a lagged difference of a dependent variable ( i) was greater than 1.60 in absolute value and the t-statistic on i for i l was less than 1.60. This procedure yielded l4, for which the null of a unit root could be rejected at the conventional level of significance. However the Schwarz criterion for determining the optimum l suggested a lag of 8, for which the null of a unit root could not be rejected. Nevertheless, neither the constant nor the trend term were significant in this specification (or indeed for any l 4) indicating misspecification of the deterministic part of the regression. Furthermore the inflation rate time series might have had a structural break at the beginning of 1994, when the inflation rate decreased markedly without a significant change in the trend (slope) over the period (Nikolic, 2000a). Applying the modified unit root test (Perron, 1989), the presence of a structural break was confirmed and the unit root hypothesis rejected. Following Perron’s unit root test, a dummy variable, Du, was included to

80 Monetary Policy in Transition

account for this break and the following regression was estimated:

  15.32  0.15trend  4.52Du  0.611  (4.59)

(3.71)

(4.23)

9

兺 

i1

i

1

(4.1)

(4.88)

where the dummy variable Du1 after January 1994 and 0 otherwise, and the t-ratios are in brackets. The values of the test statistics for the dummy variable (4.23) and for the one period lagged inflation (4.88) imply a structural break and rejection of the unit root hypothesis, respectively. Hence both the ADF and the PP test statistics showed that the inflation series was a trend stationary process. Similarly, m2 was found to be a trend stationary process since both, the Perron method (l2) and the Schwarz criterion (l4) produced truncation lags that ultimately resulted in rejection of the null of a unit root. The same reasoning applied to the m2x series since the Perron method and the Schwarz criterion produced l4 and l10, respectively, for both of which the null of a unit root was also rejected. Equally importantly, the KPSS test statistics could not reject the null hypothesis of stationarity for any of the series at any tested truncation lag length, except for lag 1 in the inflation series. This last point did not question the acceptance of stationarity even for the inflation series. This is because a conventional rule – the so-called l8 rule, lINT[8(T/100)1/4] – in the KPSS test, set l equal to 7, and therefore the hypothesis of stationarity could not be rejected. Hence, all the series under consideration were found to be trend stationary. In order to examine seasonality patterns in the data all the variables were regressed against a constant, a trend and the seasonal dummies, Dis. The dummies were set equal to 1 for the month in which the seasonal variation was observed and 0 otherwise, and i1 … 11, denoted the month in which the value for a dummy was set equal to 1, for example D1January, D2February and so on. While the inflation and the m2x series did not contain any seasonal component, the m2 series exhibited seasonal patterns in January and June of each year.6 In addition a dummy variable DO94 was included to account for the exogenously induced inflation that occurred in October 1994 as result of the exchange rate crisis commonly referred as Black Tuesday (see Chapter 1).7 The dummy variable took the value of 1 for October 1994 and zero otherwise.

Money-Inflation Causality 81

The Wiener–Granger causality model and the lag-length selection criteria While the significant lead-lag relationship between money supply and inflation in Russia has been confirmed by a number of studies,8 not much has been revealed about the cause and effect of this relationship. A common means of detecting the direction of causality is to utilize the test that was initiated by Wiener (1956) but is better known as the Granger (1969) causality test. The Granger test procedure is based on the premise that if predictions of variable Y obtained using past values of Y and another variable, X, are statistically superior to forecasts obtained using only past values of Y, then X is said to ‘cause’ (in Granger’s sense) Y. The standard representation of the Granger variant of the causality test can be specified and implemented as follows. Let (t, mi,t) represent discrete, linearly indeterministic, stationary, bivariate time series on inflation and the growth of the ith money supply at time t. The Granger test to determine the causal relationship between the inflation series (t) and growth of the money supply (mi,t), involves estimating the following reduced-form bivariate distributed lag model of finite order: (L) 冤m 冥  冤AC (L) t

i,t

a c

冥冤m 冥  冤ee 冥

Bb(L) Dd(L)

t

1t

i,t

2t

(4.2)

where t1, …, T; mim2, m2x; A(L), B(L), C(L) and D(L) are one-sided lag polynomials of order a, b, c and d respectively; and the error terms, e1 and e2 are assumed to be uncorrelated with zero means and constant variances (that is, they are white noise).9 Examination of the Granger causality between mi and  amounts to testing the following null hypotheses: B(L)  0 and C(L)  0. If neither can be rejected, then mi and  are independent series. If both are rejected, then there is ‘feedback’ between mi and . If the former hypothesis is rejected but the latter is not, there is unidirectional causality running from mi to , whereas if the latter is rejected and the former is not, the reverse applies. Tests of the hypothesis mentioned above crucially hinge on the unknown parameters a, b, c and d, which represent lag lengths for the one-sided polynomials A(L), B(L), C(L) and D(L) respectively. Since the Granger causality test seems to be sensitive to the choice of a lag for the polynomials, the approach chosen to determine the lag length can be crucial to the outcome of the test. Essentially there are two alternatives.

82 Monetary Policy in Transition

The first consists of so-called ad hoc approaches that are non-statistical in nature and include arbitrary lag specifications and rule-of-thumb lag length specifications. The most popular lag lengths in the former are 4-4 and 8-8 for the dependent and independent variable respectively, while the latter, advocated by Geweke (1978), favours a smaller lag length for the independent variable. Extending these practices, when the arbitrary approach was analyzed we employed lag lengths of 4-4, 8-8, 12-12 and 16-16 for the dependent and independent variables respectively. The corresponding lag specifications for the rule-of-thumb method were 4-2, 8-4, 12-8 and 16-8. Undoubtedly there could be a variety of specifications and one could argue that when dealing with monthly data, as in our case, it might be necessary to consider longer lag lengths than those mostly used for quarterly data. However, Nikolic (2000a) shows that the average lag between variations in the money supply and the price level was far shorter in postcommunist Russia than in developed market economies. Hence the choice of the lag length specifications for the ad hoc approaches is justified. The second alternative involves a number of statistical criteria presented in the literature (Hsiao, 1981; Batten and Thornton, 1983; McMillin and Fackler, 1984; Thornton and Batten; 1985; Jones, 1989; Cheng, 1996).10 We shall employ the two that are most commonly used: Akaike’s (1970) final prediction error (FPE), initially suggested by Hsiao (1981) for use in causality testing, and Geweke and Meese’s (1981) Bayesian estimation criterion (BEC).11 These two criteria provide interesting extremes in the balance/efficiency trade-off, as pointed out by Thornton and Batten (1985). The former tends to favour unbiasedness over efficiency by selecting, on average, lags that are too long in large samples. While the latter is asymptotically efficient it tends to select lags that are too short in finite samples. We pay particular attention to the FPE criterion, which received a different evaluation by Jones (1989) from those by Hsiao (1981) and Thornton and Batten (1985). The crux of the matter is whether the FPE outperforms other statistical search criteria as well as the ad hoc ones. In order to determine this it is necessary to follow the five steps in Hsiao’s (1981) method, as described in the Appendix. This is a stepwise procedure based on Granger’s concept of causality and the FPE criterion, which is employed as a practical means to identify the order of lags of each variable in a bivariate autoregressive process. According to Hsiao the great advantage of the criterion is that it balances the risk due to the bias that is introduced when a lower order is selected and the risk due to

Money-Inflation Causality 83

the increase of variance that results when a higher order is selected. Moreover the method does not require all the variables to have identical lag lengths. Thus unlike methods that impose this artificial restriction, which cannot only reduce the efficiency of the procedure but may also bias the value, the FPE method is free from such problems. Furthermore the procedure, as well as being a reasonably powerful test of exogeneity (causality), allows a finer specification of the system equations without using an arbitrary damping factor. Moreover an economically meaningful hypothesis can be formulated and tested by means of reduced form estimates (ibid.).

Empirical results Equation A.2 in the Appendix, or rather the subset of equations derived from that equation, were first estimated using the OLS method with the inclusion of an exogenous shock dummy DO94, seasonal dummies, Di, where appropriate, and a linear time trend.12 The residuals obtained from the regressions appeared to negate the white noise assumption. More specifically, the Ljung–Box Q-statistics for standardized and squared residuals were used to check for the existence of serial correlation (or the specification of the equation) and for autoregressive conditional heteroscedasticity (ARCH) respectively. In addition the LaGrange Multiplier (LM) test for ARCH and the Jarque–Bera test for skewness and kurtosis were used when an ARCH effect was suspected. When an ARCH was detected in Equations A.3 to A.10, the equations were re-estimated by means of the ARCH or generalized ARCH (GARCH) models proposed by Engle (1982) and Bollerslev (1986) respectively. The ARCH/GARCH specifications were obtained by minimizing the Schwarz criterion for predictive accuracy.13 The same procedure was also used to estimate the ad hoc models described above. We looked at the ARCH(1), ARCH(2) and ARCH(3) models and, as it has become a convention, at the GARCH(1,1), GARCH(1,2), GARCH(2,2) and GARCH(2,1) models. Only in the few instances when these models failed adequately to model an ARCH/ GARCH term did we use a more complex ARCH/GARCH structure. Finally the Ljung–Box Q-statistics for standardized and squared residuals and the Jarque–Bera statistics were used to test the adequacy of the ARCH/GARCH models. These statistics were insignificant in all cases.14 The lag-length selection results, obtained by unbiased and consistent estimations of variances employing the FPE and BEC criteria, are shown in Table 4.2.15

84 Monetary Policy in Transition Table 4.2 Lag lengths selected by means of the FPE and the Bayesian estimation criteria Dependent variable/ independent variable /m2 /m2x m2/ m2x/

FPE

BEC

10/13 10/15 12/2 14/21

10/1 10/7 12/2 12/12

Table 4.3 Granger causality F-statistics for the lags specified by various criteria Dependent/ independent variable /m2 /m2x m2/ m2x/

Arbitrary 4-4

8-8

2.67* 2.10* 6.97* 4.13* 8.01* 2.97* 2.35 1.90

Rule-of-thumb

12-12

16-16

4-2

3.75* 4.83* 1.08 5.41*

6.12* 0.91 4.98* 0.61 1.05 3.44* 4.29* 2.09

Statistical

8-4

12-8

16-8

FPE

BEC

3.87* 11.69* 3.30* 0.36

4.97* 1.45 1.44 1.10

7.04* 3.82* 1.26 2.18

4.37* 3.92* 3.61* 5.53*

6.08* 3.60* 3.61* 4.56*

* Significant at the 5 per cent level or better.

According to Hsiao (1981) the magnitudes of the pairs of FPEs for controlled and manipulated variables determine the direction of causality, as explained in the Appendix. However, as noted by Thornton and Batten (1985), this procedure could be interpreted as applying a higher than conventional significance level and it requires the computation of F-statistics to attain conventional significance levels. Hence the standard F-tests for Granger causality were performed on both the statistical and the ad hoc lag-length specifications (Table 4.3). As can be seen from Table 4.3 the standard F-test for Granger causality produced differing and at times contradictory results across the specifications. In particular the results of all the arbitrary specifications indicate causality running from both aggregates of broad money to inflation. However the commonly used specifications, 4-4 and 8-8 suggest unidirectional causality for extended broad money, and bilateral causality for ruble broad money, while the 12-12 and 16-16 specifications suggest the contrary.

Money-Inflation Causality 85

Similarly the matrix of the rule-of-thumb specifications produced quite contradictory results. In particular the specifications 8-4, 12-8 and 16-8 indicate that causality ran from ruble broad money supply to inflation, but only in the 8-4 specification was this bilateral causality. The specification 4-2, in contrast, indicates that none of the aggregates of broad money caused inflation while inflation caused only ruble broad money. Extended broad money Granger caused inflation only in the 8-4 and 16-8 specifications. None of the rule-of-thumb specifications indicate that inflation caused extended broad money. However in contrast to the ad hoc approaches, both statistical criteria indicate bidirectional causality between both of the monetary aggregates and inflation.16 Given the variety of results of Granger causality produced by the various specifications, the obvious question is which of the specifications provides a correct account of Granger causality, or which criterion produces the best model specifications. One way to determine this is to subject all the specifications to an ordinary F-test. Since our objective is to evaluate the performance of the FPE relative to other criteria, all the FPE-selected models have been compared with those selected by the ad hoc models and the BEC. Following Thornton and Batten (1985), if the F-statistics for FPE-selected models that were of a higher order (H) than the alternative model were significant at the conventional 5 per cent level, the FPEspecified model would be favoured over the alternative one. Conversely if the F-statistics for FPEs of a lower order (L) than the alternative were not significant the FPE would be preferred model. The results of this comparison (Table 4.4), indicate that the FPE criterion outperforms the other specifications without exception. The results in Table 4.4 not only reveal the comparative superiority of the FPE-selected model over the ad hoc models and the BEC, but also have an implication for the results of the Granger causality test reported in Table 4.3. That is, the FPE criterion reveals the models that correctly identify Granger causality between inflation and both the aggregates of broad money supply in Russia. Therefore, as Table 4.3 suggested, there is feedback or bilateral causality between both ruble broad money and inflation and extended broad money and inflation in Russia. Our analysis clearly illustrates that the outcome of causality testing is sensitive to the choice of lag-length. Thus contrary to Jones’s (1989) claim, the arbitrary lag length specifications that are most commonly used in the literature – 4-4 and 8-8 – and the other arbitrary specification can produce misleading results, as indicated by Thornton and Batten (1985). The rule-of-thumb methodology produces even less satisfactory

10.18* (H) 7.33* (H) (NN) 22.28* (H)

14.39* (H) 8.31* (H) (NN) 23.71* (H)

8-8 5.62* (H) (NN) 0.63 (L) 3.68* (H)

12-12 1.28 (L) 0.64 (L) 0.86 (L) (NN)

4-2 12.97* (H) 10.26* (H) 11.53* (H) 22.55* (H)

8-4 9.09* (H) 4.82* (H) (NN) 12.86* (H)

12-8

Rule-of-thumb

1.81 (L) (NN) 0.35 (L) 7.42* (H)

16-8

0.54 (L) (NN) 0.43 (L) (NN)

16-16

3.21* (H) 3.20* (H) – 5.79* (H)

BEC

Statistical

Notes: * Significant at the 5 per cent level or better. NN  non-nested models that were not tested. A dash (–) indicates the same lag length as one chosen by the FPE.

/m2 /m2x m2/ m2x/

4-4

Arbitrary

F-tests of the FPE lag specifications

Dependent/ independent variable

Table 4.4

86

Money-Inflation Causality 87

results. In short, none of the ad hoc approaches has produced a complete set of satisfactory results for the given sample. In contrast, the statistical criteria FPE and BEC have performed very well. To illustrate further the extent of the sensitivity of the Granger causality test to the choice of lag length an extensive search of the lag space was performed. The lag search was conducted over all possible combinations of up to 15 lag lengths for the dependent and independent variables.17 In line with Thornton and Batten (1985) and Jones (1989), we found marginal significance levels for the computed F-statistics in the Granger tests (Tables 4.5 and 4.6).18 The results in Tables 4.5 and 4.6 add further evidence to the finding revealed by Table 4.3; that is, causality tests between inflation and both aggregates of money supply are heavily dependent on the lag length specification. Thus, it is clear that using ad hoc approaches to determine the order of lags in these tests can produce seriously misleading results, as was the case with our sample. Finally, as the FPE criterion appears to be superior to the other methods of model specification considered here, we shall combine all the single equation specifications obtained by the FPE criterion in order to identify the system. Given the existence of feedback or bilateral causality between m2 and m2x on the one hand and  on the other, a bivariate feedback model for both of the broad money aggregates and inflation fits the data best. Hence for the m2 and  pair we choose the following:

冤m2 冥  冤cc 冥  冤AC (L)(L) t

10

1

t

2

2

冥冤m2 冥 

B13(L) D12(L)

t

t

e  冤DD 冥  冤DD 冥  冤DO94 0 冥 冤e 冥 1

6

1t

1

6

2t

(4.3)

where the ci are constants, A(L), B(L), C(L) and D(L) are one-sided lag polynomials of orders 10, 13, 2 and 12, respectively; Di and DO94 are the seasonal and October 1994 dummies respectively, and the ei are the white noise error terms described earlier. For the m2x and  pair we choose:  冤m2x 冥  冤cc 冥  冤CA t

1

t

2

10

(L) (L)

21

 e  冥冤m2x 冥  冤DO94 0 冥 冤e 冥

B15(L) D14(L)

t

1t

t

2t

(4.4)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

 on m2, Lags of m2

0.115 1.000 0.465 0.121 1.000 0.000* 0.000* 0.000* 0.000* 0.050 0.017* 0.465 0.461 0.506 0.049*

1

0.108 0.874 0.965 0.887 0.982 1.000 0.992 0.867 0.001* 0.029* 0.007* 0.009* 0.024* 0.044* 0.079

2

0.279 1.000 1.000 0.968 1.000 1.000 0.944 1.000 0.003* 0.055 0.046* 0.512 0.841 0.463 1.000

3

1.000 1.000 1.000 0.875 0.133 0.999 0.243 0.183 0.044* 0.094 0.300 0.004* 0.217 0.058 0.031*

4 1.000 1.000 1.000 0.001* 1.000 0.091 0.001* 0.381 0.994 0.933 0.015* 0.664 0.093 0.125 0.000*

5 0.144 1.000 1.000 0.093 1.000 0.001* 0.398 0.219 0.042* 1.000 0.167 0.994 0.038* 0.018* 0.001*

6 0.000* 1.000 0.666 0.394 0.912 0.001* 0.000* 0.281 0.076 0.021* 0.001* 0.000* 0.001* 0.000* 0.001*

7 0.000* 1.000 0.774 0.253 0.020* 0.315 0.029* 0.706 0.061 0.986 0.012* 0.000* 0.000* 0.002* 0.000*

8

Lags of  9 0.000* 0.020* 0.041* 0.048* 0.000* 0.020* 0.002* 0.010* 0.002* 0.000* 0.002* 0.000* 0.000* 0.002* 0.015*

Table 4.5 Significance levels derived from Granger causality tests of  and m2

0.000* 0.026* 0.003* 0.140 0.040* 0.154 0.004* 0.070 0.669 0.002* 0.000* 0.000* 0.001* 0.002* 0.012*

10 0.002* 1.000 0.350 0.003* 0.000* 0.042* 0.231 0.000* 0.000* 0.000* 0.000* 0.001* 0.001* 0.002* 0.002*

11

0.012* 0.036* 0.613 0.308 0.073 0.094 0.000* 0.002* 0.725 0.000* 0.000* 0.000* 0.001* 0.002* 0.003*

12

0.042* 0.001* 1.000 1.000 0.017* 0.119 0.002* 0.005* 0.001* 0.000* 0.001* 0.000* 0.002* 0.004* 0.004*

13

0.048* 0.032* 0.266 0.004* 0.000* 0.000* 0.000* 0.003* 0.000* 0.000* 0.000* 0.003* 0.004* 0.007* 0.008*

14

0.051 0.016* 0.020* 0.031* 0.049* 0.006* 0.022* 0.028* 0.005* 0.001* 0.074 0.001* 0.003* 0.005* 0.008*

15

88

1.000 1.000 0.172 0.000* 0.808 0.002* 0.006* 0.010* 0.002* 0.021* 1.000 1.000 0.978 0.999 1.000

1

1.000 0.793 0.007* 0.000* 0.003* 0.001* 0.003* 0.010* 0.004* 0.027* 0.107 0.160 0.891 0.510 1.000

2

1.000 0.161 0.172 0.002* 0.013* 0.001* 0.006* 0.029* 0.013* 0.042* 0.095 0.076 0.055 0.101 0.140

3

* Significant at the 5 per cent level or better.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

m2 on , Lags of 

1.000 0.039* 0.002* 0.004* 0.023* 0.002* 0.006* 0.026* 0.020* 0.053 0.111 0.128 0.061 0.119 0.149

4 1.000 0.055 0.055 0.111 0.184 0.028* 0.038* 0.106 0.181 0.258 0.374 0.392 0.189 0.246 0.215

5 1.000 0.005* 1.000 1.000 0.017* 0.995 0.016* 0.138 1.000 0.164 1.000 0.277 0.069 0.105 0.172

6 1.000 0.012* 1.000 1.000 0.027* 1.000 0.013* 1.000 1.000 1.000 1.000 0.294 0.081 0.120 0.178

7 1.000 0.033* 0.078 0.063 0.044* 0.074 0.060 0.025* 0.095 0.098 0.174 0.333 0.110 0.161 0.212

8

Lags of m2

0.180 0.037* 0.087 0.139 0.229 0.308 0.262 0.053 0.087 0.077 0.146 0.309 0.144 0.207 0.262

9 1.000 0.049* 0.820 0.196 1.000 0.482 0.200 0.042* 0.064 0.088 0.159 0.325 0.160 0.187 0.222

10 0.194 0.100 0.320 0.785 0.346 0.290 1.000 0.333 0.062 0.123 0.179 0.346 0.178 0.202 0.244

11 0.820 0.035* 0.082 0.049* 0.074 0.102 0.157 0.208 0.245 0.320 0.359 0.401 0.126 0.180 0.190

12 0.917 0.052 0.116 0.076 0.074 0.098 0.153 0.224 0.076 0.116 0.161 0.226 0.140 0.200 0.210

13 0.887 0.070 0.149 0.106 0.103 0.154 0.225 0.314 0.115 0.138 0.193 0.262 0.153 0.178 0.128

14 0.914 0.056 0.124 0.086 0.076 0.131 0.188 0.271 0.159 0.211 0.289 0.376 0.265 0.310 0.165

15

89

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

 on m2x Lags of 

0.021* 0.580 0.447 0.007* 0.023* 0.010* 0.012* 0.000* 0.005* 0.031* 0.035* 0.170 0.605 0.123 0.676

1

0.021* 0.614 0.306 0.027* 0.085 0.000* 0.430 0.563 0.067 0.414 0.083 0.157 0.414 0.572 0.103

2

0.022* 1.000 0.538 0.006* 0.030* 0.000* 0.170 0.002* 0.002* 0.042* – 0.344 0.632 0.094 0.446

3

0.023* 1.000 0.001* 0.025* 0.019* 0.007* 0.326 0.077 0.023* 0.221 0.082 0.042* 0.286 0.533 0.002*

4 0.002* 0.898 0.224 0.010* 0.072* 0.010* 0.117 0.048* 0.017* 0.154 0.024* 0.055 0.968 0.689 0.300

5 0.001* 0.089 0.012 0.000* 0.053 0.010* 0.152 0.072 0.003* 0.155 0.303 0.011* 0.003* 1.000 0.001*

6 0.000* 0.045* 0.000* 0.000* 0.000* 0.000* 0.012* 0.126 0.000* 0.124 0.011* 0.041* 0.000* 0.999 0.002*

7 0.023* 1.000 0.000* 0.000* 0.000* 0.000* 0.039* 0.060 0.014* 0.025* 0.004* 0.025* 0.053 0.298 0.002*

8

Lags of 

1.000 1.000 0.000* 0.000* 0.001* 0.004* 0.000* 0.001* 0.002* 0.178 0.083 0.046* 0.001* 0.934 0.004*

9

Table 4.6 Significance levels derived from Granger causality tests of  and m2x

0.528 1.000 0.023* 0.028* 0.290 0.025* 0.002* 0.010* 0.021* 0.051 0.082 0.012* 0.020* 0.037* 0.006*

10

1.000 0.385 0.002* 0.000* 0.003* 0.016* 0.005* 0.011* 0.015* 0.066 0.159 0.002* 0.004* 0.011* 0.008*

11

0.658 1.000 0.163 0.072 0.970 1.000 0.007* 0.205 0.288 0.438 0.035* 0.002* 0.004* 0.009* 0.006*

12

0.115 1.000 0.412 0.001* 0.352 0.255 0.001* 0.097 0.091 0.016* 0.040* 0.002* 0.003* 0.008* 0.009*

13

0.899 0.473 0.879 0.145 0.095 0.224 0.001* 0.104 0.308 0.130 0.434 0.004* 0.006* 0.011* 0.015*

14

1.000 0.394 – 0.023* 0.020* 0.365 0.030* 0.919 0.222 0.470 0.444 0.003* 0.007* 0.011* 0.013*

15

90

– – – 1.000 1.000 1.000 1.000 1.000 0.009* 0.681 0.262 – – – –

1

1.000 0.436 0.015* 0.273 1.000 1.000 1.000 0.139 0.299 0.620 0.530 0.001* 0.004* 0.420 0.000*

2

0.474 1.000 0.340 1.000 0.472 1.000 1.000 0.153 1.000 1.000 0.872 0.225 0.876 0.038* –

3

1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.019* 1.000 0.935 0.919 0.463 0.339 –

4 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.028* 0.615 0.821 0.288 0.809 0.083 0.001*

5 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.048* 0.162 0.160 0.151 – 0.000* 0.000*

6 1.000 1.000 1.000 0.593 0.676 1.000 0.052 1.000 0.998 1.000 0.997 0.300 0.299 0.001* 0.001*

7 1.000 0.156 1.000 0.854 0.907 0.928 1.000 1.000 0.231 1.000 1.000 0.187 0.845 0.004* 0.004*

8

Lags of m2x

1.000 1.000 1.000 0.946 0.932 0.830 0.603 0.995 0.780 1.000 1.000 0.006* – 0.000* 0.000*

9 1.000 0.464 0.661 0.800 0.582 0.696 0.800 0.793 0.119 0.032* 0.012* 0.007* 0.013* 0.001* 0.000*

10 1.000 1.000 1.000 1.000 0.629 0.068 0.136 0.030* 0.074 0.001* 0.015* 0.007* 0.945 0.008* 0.001*

11 0.568 0.711 0.535 0.603 0.322 0.428 0.005* 0.054 0.378 0.022* 0.007* 0.002* 0.265 0.003* 0.001*

12 1.000 1.000 1.000 1.000 1.000 1.000 0.162 0.646 0.059 0.035* 0.002* 0.194 0.017* 0.019* 0.000*

13

1.000 1.000 1.000 1.000 1.000 1.000 1.000 0.749 0.978 0.382 0.013* 0.004* 0.032* 0.012* 0.001*

14

1.000 0.529 1.000 0.807 1.000 1.000 0.571 0.137 0.376 0.362 0.002* 0.000* 0.000* 0.000* 0.000*

15

Notes: * Significant at the 5 per cent level or better. A dash indicates that no ARCH/GARCH or OLS estimate gave white noise residuals so no reliable decision in respect of Granger causality could be made.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

m2x on , Lags of 

91

92 Monetary Policy in Transition

Policy considerations As noted earlier, for a variable to be useful as a policy target, unidirectional causation from it to a goal variable is a necessary condition. Hence the finding that there is feedback or bilateral causality between monetary aggregate M2 and inflation severely limits ability of the former to be an effective policy target for the latter. The existence of such a relationship means that movements in the target variable may reflect the combined influence of policy actions and movements in the goal variable that are not directly attributable to desired policy changes. In other words, if the target variable both causes and is caused by the goal variable then it does not provide an unambiguous signal of where actions are headed. The counter argument is that the authorities may be able to control the money growth rate but have reacted to past inflation by partially accommodating it. It may indeed be the case that feedback causality is the result of accommodating inflation. Does this mean that strict control of the money growth rate would have stabilized inflation in Russia at the beginning of its transition? Had the velocity of money been constant and the institutional infrastructure of a competitive market economy been in place, the answer to this question would possibly have been positive. However, as documented above and shown in Figure 2.1, the velocity of money was far from constant and the necessary institutional infrastructure was not in place during first half of the 1990s. In addition it would be utterly unrealistic to expect Russia to have had a strict monetary policy and a balanced budget in the midst of a transformational recession when no domestic borrowing was possible and the majority of the promised international loans were not forthcoming. Lack of financial discipline in this period meant that monetary tightening resulted in a considerable increase in bartering, the widespread use of money surrogates and the accrual of all sorts of arrears. Widespread dollarization illustrated the diminishing role of the ruble as a unit of account. In such an economic environment, had the government not been a large agent paying and demanding payment in rubles, the ruble would have been endangered as a medium of exchange. Notwithstanding all this the monetary authorities adhered to monetary targeting for a considerable length of time. More specifically, in the ‘Guidelines on the monetary policy of the CBR’ presented to the Russian parliament in early 1992, the Gaidar government opted to target the quarterly growth rates of M2. The government not only fail to comply with its agreed targets but also exceeded them by far, which unsurprisingly resulted in high inflation during 1992 and 1993. One of the reasons for

Money-Inflation Causality 93

this failure was that the total disorder of monetary data made such targets unreliable. In 1993 targets were set for credit to commercial banks, to the budget and to former Soviet republics. In addition floor levels were set for net international reserves and gross reserves, and ceilings for the CBR’s net domestic assets and net credit to the enlarged government, that is, credit targets (Buch, 1998). Additional target levels for M2 and the monetary base followed in 1994. Under the auspices and guidance of the IMF the Russian monetary authorities continued this approach in 1995. M3 was used not only as the main intermediate target for monetary policy but also for calculating the targets for the monetary base and the CBR’s net domestic assets (Bofinger et al., 1997).19 In July 1995 the authorities officially announced a corridor for the ruble–dollar exchange rate, which was adjusted the following year and finally modified into a sort of crawling peg. This action on the part of policy makers did not exclude the practice of monetary targeting.20 However according to Bofinger et al. (ibid.) the strong overshooting of the monetary targets meant that this stabilization programme could not be classified as money-based stabilization but rather as exchange-rate-based stabilization, which turned out to be successful in short to medium term. The poor results of the stabilization efforts prior to 1995 provided ammunition for critics of the design and implementation of the IMFsupported policies. The suitability of the orthodox money-based stabilization programme was thus put under a spotlight. It was suggested, that money-based stabilization programmes implemented in highinflation countries such as Chile in 1974–1975 and Argentina and Brazil in early 1990 could not be recommended to transition economies. The slowness with which money-based programmes reduced inflation and their adverse effects on output and employment not only rendered them unsuitable for use in high-inflation countries but also their recessionary effects tended to be immediate. This was mainly due to shortterm downward rigidity in prices and wages and lack of confidence in the government’s resolution to stick to targets. The next serious reservation was related to the essential precondition for money-based stabilization: stable monetary velocity. Given the severe limitations of monetary targeting in the OECD countries, it was considered that money-based approaches were unlikely to be suitable for short-term stabilization in transition economies where the demand for money was not stable or at least during the early years of transition (ibid.).21 Furthermore it was argued that monetary targets carried the danger of Dornbusch-style exchange rate overshooting (Fisher, 1986). Moreover even if monetary velocity could be forecast perfectly it would have to forecast both real

94 Monetary Policy in Transition

output and prices in order to set a sensible target for the monetary aggregate. This task would be particularly difficult at a time when there was considerable doubt about real output and prices, as was the case during early transition (Begg, 1997). It was thought that in the face of unstable money demand an exchange rate anchor could not only help to end the inflationary spiral, thereby enhancing confidence and helping with remonetization, but also induce a greater commitment to fiscal adjustment, which was notably lacking in Russia.22 With the benefit of hindsight it appears that despite its shortcomings the exchange rate peg, with occasional realignments, should have been introduced earlier to stabilize the economy. Alternatively money targeting might have been used as a medium-term policy instrument, as practiced by the Bundesbank and the Swiss National Bank, rather than for short-term stabilization purposes. As it was the IMF team and Prime Minister Gaidar vastly underestimated inflation in January 1992. Inflation was assumed to be 100 per cent but in fact it was 245 per cent, and as a consequence the government’s credibility was dented (Gomulka, 1995). Among the other policy errors of the IMF were support for a common currency in and monetary union among most of the countries of the former Soviet Union, stubborn insistence on macroeconomic policies and stabilization, and neglect of liberalization, privatization and institutional change (ibid.), not to mention economic growth. It is unclear what implications monetary targeting might have had for fiscal responsibility, which was as important as the choice of aggregate targets, if not more so. On the other hand it has been argued (Begg, 1997) that money-based stabilization worked surprisingly well in other transition economies, including Albania, Slovenia, Latvia and Lithuania. Along the same lines Buch (1998) argues that the results of monetary targeting were not entirely disappointing in Russia, for example it helped the CBR to reduce growth in domestic assets, and particularly lending to the government. In addition, and contrary to instability in the demand for money in the majority of transition countries over time (Begg, 1997), the demand for money in Russia was relatively stable in the long term, although short-term fluctuations were substantial (Buch, 1998). Buch also argues that the CBR should have aimed to move to a monetary target in the medium term. In should be noted that when monetary targeting was introduced in Russia the predicted demand for money was hardly based on solid foundations. This was related not only to the great uncertainty that prevailed during the years after price liberalization in January 1992, but also to the

Money-Inflation Causality 95

lack of meaningful time series of economic aggregates under the new policy regime. For the same reasons, policy makers could not scrutinize the suitability of monetary aggregates according to the criteria suggested above, apart from using a very short time series that was bound to produce unreliable results. In addition, in the pursuit of multiple nominal anchors, the failure of one should not have prejudiced the entire programme. For any nominal anchor, fiscal responsibility really mattered, as evidenced by the financial crisis of August 1998. Our analysis leads us to conclude that, given the existence of bilateral or feedback causality between broad money supply and inflation, use of the former to control the latter had severe limitations in postcommunist Russia. Our results point to the accommodating nature of Russian monetary policy, as indeed was the case in the Visegrad countries in the early years of transition (Rostowski and Nikolic, 1998). However, had the government not been accommodating inflation in the early years of transition, in the prevailing environment of poor financial discipline and underdeveloped economic infrastructure it would have endangered the role of the ruble as a medium of exchange. Bofinger et al. (1997) show that, because of the questionable stability of the demand for money, the use of monetary targeting for short-term stabilization purposes rather than as a medium-term policy guidelines as practised by Bundesbank and the Swiss National Bank, was a completely misconceived concept for transition countries. Money-based stabilization programmes developed for stable economies would have resulted in a liquidity crunch (Calvo and Végh, 1992) and high interest rates, which would have adversely affected output. In light of these critiques the continuation of quantitative targeting of the money stock does seem to have been an unsuitable policy prescription.23 Having found that monetary aggregates were not exogenous or causally prior to inflation in Russia as well as unstable velocity of money, it would be somewhat redundant to analyze the controllability of these two aggregates of the broad money supply. However it is worth mentioning that the CBR may have had difficulty controlling the broad money supply even if the broad money supply happened to be prior to the goal variable and with no feedback. More specifically, since changes in broad money result from the cumulative effects of changes in the monetary base and the money multiplier, the latter may have caused control and information problems for the CBR (Buch, 1998). The former would have emerged as a result of the bank’s inability accurately to predict changes in the portfolio structure of households and the excess reserves of commercial banks, and the latter because the CBR, and any

96 Monetary Policy in Transition

other central bank for that matter, could observe the behaviour of money stock only after a time lag.

Conclusion This chapter has assessed various criteria for the optimal choice of lag length in causality testing and found that Akaike’s (1970) FPE criterion outperforms both the ad hoc and the statistical criteria considered. The results also indicate that arbitrary lag length specifications, including those most commonly used – 4-4 and 8-8 – may produce seriously misleading results in causality testing. This could have severe consequences for economic policy, particularly in cases where intermediate targeting is exercised. Identification of the direction of causality between broad money and inflation in postcommunist Russia has enabled us to assess the appropriateness of monetary policy in this economy. An intermediate target is appropriate if it reasonably determines a goal variable, if it is exogenous or causally prior to the goal of policy actions, and if it is controllable by policy makers. Assuming the first prerequisite for a stable relationship between a target (ruble broad or extended broad money supply) and a goal variable (inflation) is fulfilled, as claimed in the literature, this study has demonstrated that neither of the targets is exogenous in the case of Russia. In other words the existence of feedback or bilateral causality between inflation and both aggregates of broad money diminishes their suitability for monetary targeting. This is because neither of them can provide an unambiguous signal of where policy actions are headed. Hence it would be a redundant exercise to test the controllability of the monetary aggregates. In light of these findings it appears that the use of monetary targeting for short-term stabilization purposes, as applied by the Russian monetary authorities under the strong auspices of the IMF, might not have been the wisest policy recommendation. Finally, we have to express reservation about the Russian statistics used in this study. In addition there were conceptual problems related to the change from a socialist to a market economy, and we had at our disposal only a limited number of observations. Moreover the period covered was characterized by a large reduction in economic activity, unprecedented transformation of the economic system and sociopolitical institutions, a new economic policy regime and advances in financial innovations. Because of all this our results and conclusions should be treated with some caution.

5 The Role of International Financial Institutions in Stabilizing the Russian Economy1

Introduction Between the end of Czarist Russia in 1917 and the early 1950s about a third of the world’s population experienced a transition from a market economy to central planning. In the late 1980s, in the run-up to the collapse of the Soviet empire, the reverse process was set in motion in most of the former socialist countries: transition from a planned to a market economy. Their long history of central planning and lack of experience of free-market economics made assistance by international financial institution essential to the transition economies. It was expected that this assistance would be both financial and consultative. Because of their considerable experience with structural reforms in Latin America the Bretton Woods institutions, guardians of the international financial system – the World Bank and the International Monetary Fund (IMF) – were the obvious candidates to help the transition economies in Central and Eastern Europe with their endeavors. These institutions, and particularly the IMF, were involved to varying degrees of success with the transition of these economies, including the Russian Federation. Given the considerable difficulty and protraction of Russia’s transition efforts, many aspects of the part played by the IMF are controversial. It is often claimed that the IMF, as the main guide and coordinator of the West’s assistance to Russia, performed less than optimally (see for example Sachs, 1997; Sanders, 1998; LaRouche, 1999; Soros, 2000). Its positive efforts in providing credibility to this major transition economy in the 1990s were in general counteracted by its failure to disburse pledged loans at critical times for successful transition, particularly at the beginning of its involvement. In addition the shortcomings of IMF-guided stabilization programmes in the first few 97

98 Monetary Policy in Transition

years of transition (see Chapters 2 and 4) were coupled with neglect of institution building (Kolodko, 1999), not to mention disregard for equitable growth of the Russian economy. Furthermore the IMF made a number of policy errors that contributed to the delay in stabilizing the economy. The remainder of this chapter is organized as follows. The next two sections briefly discuss the role of the IMF in the global financial architecture and the Washington Consensus respectively. The fourth section analyzes the general policy considerations of the IMF’s involvement in Russia, and the fifth outlines the quantitative indicators of capital flows in European transition economies and Russia. The sixth section is devoted to common criticisms of the IMF programmes and the seventh outlines specific policy considerations of the IMF’s involvement in postcommunist Russia. The reasons for developing countries to seek IMF assistance are considered in the eighth section, which is followed by a summary of the findings.

The role of the IMF in the global financial architecture The World Bank and particularly the IMF have been the cornerstones of the global financial system for more than five decades. During this time the results of the involvement in various stabilization programmes have been mixed. The World Bank has responded to financial crises with recommendations to strengthen policy regimes and financial support. Although it is not intended to act as a lender of last resort and is not primarily designed to fight crises, its participation has been required because of the structural origins of the crises and the enormous impact they have had on income distribution and poverty. With macroeconomic stability as its chief aim the IMF has played a leading role in the creation and execution of these programmes. Despite differences related to country-specific characteristic, most of the IMF programmes have three common and complementary elements: (1) securing sustainable external financing, (2) the adoption of demandrestraining measures, especially in the early stages of a programme, and (3) the implementation of structural reforms. The availability of external financing determines the magnitude and pace of the adjustment process. At the outset of a programme a country that experiences balance of payment difficulties can usually borrow only limited funds. IMF guidelines require countries not to have an ex ante external financing gap, to be up to date with their debt service commitments, and to eliminate any external debt arrears that have accumulated prior to programme approval (Mussa and Savastano, 1999).

The Role Played by International Financial Institutions 99

Demand-restraining measures, which are typically understood as measures to tighten monetary and fiscal policies, are the best known but controversial elements of typical IMF programmes. The intention of the architects of such programmes is to bring aggregate demand into line with expected output and the availability of external financing, and thus to establish a sustainable current account. In order to facilitate external adjustment the programme architects may also opt for alteration of the nominal exchange rate. In recent times the IMF has tended to stress the tightening of monetary rather than fiscal policy in countries with weak financial systems when investors lose their confidence. The purpose of this is to prevent currency crises, but the IMF’s record in this area has not been impressive. Typical programmes to alleviate structural and institutional rigidities are aimed at facilitating the efficient allocation of resources, and in so doing to smooth the progress of economic growth. Structural reforms may include changes to a variety of activities and vary from country to country. Typically the key structural priorities for transition economies are privatization and the building of market institutions.

The Washington Consensus The IMF’s involvement in steering the former planned economies of Central and Eastern Europe along the path to a market economy constituted the biggest challenge of its existence. The IMF took the lead in assisting the efforts of Western donor organizations and countries. The starting point for policy advice to these countries was the so-called Washington Consensus, a body that comprised the US Treasury, the IMF and the World Bank. It was the product of Latin America’s structural crises in the 1980s and operated according to the motto ‘Liberalize and privatize as quickly as possible, and be tough in fiscal and monetary matters’. Restructuring would follow in a later phase. Since the assumption was that transition economies were in macroeconomic disequilibrium, much like the Latin American economies, stabilization was the foremost priority. Demand-rather than supply-side management was the preferred order of the day. External and internal liberalization, coupled with privatization and stabilization measures, were expected to transform the transition economies into fully fledged market economies. The Washington Consensus’s motto was to a large extent promoted in Russia via IMF-guided reforms. The main components of the Russian reform were rapid price liberalization, liberalization of the foreign exchange market and convertibility of the ruble, extensive foreign

100 Monetary Policy in Transition

trade liberalization, macroeconomic stability measures, privatization and other systemic, structural and institutional reforms. Safety nets and external assistance were additional supportive features (Gomulka, 1995). Discussions of the reforms usually centre on the issues of rapid or shock versus gradual pace of reforms or the extent of these two. Since the chief concern of the IMF has traditionally been short- and mediumterm macroeconomic stability, we shall concentrate on this issue. The results of the macrostabilization efforts offer a way of judging the degree of the IMF’s success in Russia according to its own criteria.

The IMF’s involvement in Russia: general policy considerations The part played by the IMF in developing economies was subject to scrutiny by structuralist theoreticians during 1970s and 1980s (for example Taylor, 1988), and inevitably its role in the Russian transition received particular attention in the wake of the financial crisis of August 1998. The articles on this, which were usually critical for their own ends, distinguished between the structural deficiencies of the IMF in the global financial architecture (Soros, 2000) and the specific policy mistakes it made in Russia (Sachs, 1997). The IMF’s involvement in Russia began after its experimentation with the early reformers at the start of the 1990s. Together with the Russian government it devised a stabilization plan and the first official IMFsupported programme was unveiled in June 1992, the same month in which Russia formally rejoined the Bretton Woods institutions. Halfhearted stabilization efforts had already been made since autumn 1991, but with poor results. In this period the IMF’s position was that the Soviet ruble should continue to be common currency for the successor states of the Soviet Union, that Russia should have a balanced budget, and that – since inflation was viewed as a purely monetary phenomenon – the money supply should be kept under tight control. Various monetary targets were imposed and the inflation rate was supposed to fall below 5 per cent per month. Notably, during the first few years neither the exchange rate nor a wage rate target served as a nominal anchor. Hence an orthodox money-based stabilization strategy was chosen by the IMF for Russia. The poor results of the first stabilization programme led to the introduction of a replacement one in June 1993. However this suffered a similar fate to its predecessor and was succeeded by a more ambitious programme in March 1994. While the inflation rate of 1526.0 per cent in 1992 did fall to 875.0 per cent in 1993 and 311.4 per cent

The Role Played by International Financial Institutions 101

in 1994 (EBRD, 1998) it was still very high and acted as testimony to the failure of the stabilization efforts. In addition, as discussed in Chapter 2, on 11 October 1994 Russia suffered its first full financial crisis in postcommunist times. Consequently it was clear that stabilization had to take a new form. After thorough preparation an exchange-rate-based stabilization programme with heterodox elements was implemented in June 1995. This programme was relatively successful but only in the short to medium term.2 Macroeconomic stability is always fragile in transforming economies like the Russian Federation in which the fiscal deficit was not been brought under control and averaged close to 8 per cent per annum up to 1998 (Nikolic, 2000a). The dire position of the government’s finances and the economy as a whole was exacerbated by other internal and external factors (see Chapter 1 for details of these and their consequences). Russia’s Failure to achieve macroeconomic stability during the first years after liberalization attracted sharp criticism of the IMF from various quarters. This is not to say that successive Russian governments were not immune from criticism. Their chief fault was their consistent refusal to reform. Yet the reluctance of the G7 countries to involve themselves in reforms and provide financial assistance to Russia when it was most needed was equally to blame, and the snail’s pace of the World Bank’s involvement contributed to the failure. Nonetheless the Russian government and the IMF squandered numerous opportunities to stabilize the economy, with disastrous consequences for the welfare of the Russian people. Critics charged that these two bodies never missed an opportunity to miss an opportunity (Sachs, 1997). During this time there was a common understanding between the West (via the IMF) and the Russian government: the West would pretend to aid Russia while it would pretend to act (Granville, 1995). Likewise the IMF promised loans and the Russian government promised reforms. This turned out to be pseudo lending for pseudo reforms. The end result of Russia’s transformation endeavour in the 1990s may well bear out the criticisms of the IMF’s programmes. Although there are doubts about their accuracy, between 1992 and 1995 Russian official statistics recorded a fall in GDP of 42 per cent and a fall in industrial production of 46 per cent – far worse than the contraction of the US economy during the Great Depression. Critics claim that the consequence of the IMF’s prescriptions for the decontrol of prices, radical economic and financial liberalization and the indiscriminate opening of markets to imported products was to transform Russia into a raw materials producer, rather than an agro-industrial economy (LaRouche, 1999).

102 Monetary Policy in Transition

Soros (2000) argues that the IMF was not in Russia to aid economic development but merely to serve as a bill collector for the world financial community. It has also been suggested that a considerable proportion of the IMF loans was used to bail out international creditors, creating what is now recognized as a moral hazard (ibid.; Sanders, 1998). Much of the rest went to corrupt government officials, bureaucrats and certain businessmen. The effect of all this was devastating for the Russian people. Between 1991 (the time of the Soviet Union’s collapse) and 1995 real income plummeted 40 per cent. A quarter of all Russians were living below the subsistence level and nearly a third below the poverty level. Three quarters were barely surviving on the average income of $100 per month. A quarter of the labour force received their wages late, in kind or not at all (Sanders, 1998). Meanwhile the average life expectancy for men fell to 59, a decline of seven years. Finally, criticism ought to have been directed at the political economy of the IMF’s involvement in Russia, yet the literature has not been very forthcoming on this matter. Specifically, given the large gap between the promised funds and those delivered, as well as the timing of some deliveries, one cannot help but think that the IMF and the US Treasury, which calls the tunes at the IMF, were used to spoon-feeding Russia to the point of no return. Throughout the transition period the Russian reformers were given just enough aid and at just the right time for them to remain in power and ensure that Russia did not revert to central planning and autarky. According to Thacker (1999) contrary to the expectation that the IMF would become less politicized after the end of the Cold War, the influence of politics actually increased after 1990. Thacker claims that political realignment towards the United States, the largest power in the IMF, increased a country’s probability of receiving an IMF loan. He concludes that the actions of multilateral organizations are still driven by the political interests of their more powerful member states.

Quantitative indicators of capital flows and the IMF’s involvement in the European transition economies and the Russian Federation Capital flows to Russia from 1989 to 1993 were shaped by Western governments’ determination to make the transition ‘stick’, coupled with a wait-and-see approach by private sources of funds. In addition to the finance provided by the IMF and the World Bank, bilateral credits were

The Role Played by International Financial Institutions 103

extended to Russia by seven major industrial countries. Moreover official creditors, under the auspices of the Paris Club, and debt deferrals by commercial banks offered a comprehensive debt relief package for Russia, but not until the late 1990s. When Russian’s economic performance improved and the transition progressed, private capital began to enter the market, first tentatively and then with great speed. Hence the sequence was as follows: official financing, FDI, non-guaranteed bank loans, dedicated equity funds, international bond issues, direct local stock and money market investments (EBRD, 1998). Despite the fact that by 1993 virtually all the East European economies and the Russian Federation were IMF members, in general they received a smaller and declining share of financial resources relative to developing countries in the first half of the 1990s (UN and ECE, 2000). Although in 1990–98 many of the transition economies attracted capital inflows3 of about 5 per cent of GDP, in line with that received by developing economies, a significant number, including Russia, failed to do so (ibid.) Indeed as Figure 5.1 and Table 5.1 show, net capital inflows, including ‘errors and omissions’, were negative in Russia in that period.4 The variations in the size of the capital flows to transition economies reflect their degree of economic reform. In effect, access to official funds was often conditional on the implementation of structural reforms and sound macroeconomic policies. As Table 5.1 shows, most of the foreign capital (about 60 per cent) was attracted by the early reforming countries: the Czech Republic, Hungary and Poland. In contrast, although the exact size of the volatile capital flows to Russia is uncertain, Figure 5.1 indicates that the inflow of capital per capita was negligible while the outflow was substantial. The former consisted of reported financial inflows, while the latter was the sum of recorded flows plus errors and omissions, which included unrecorded capital flows – a synonym for capital flight. Russia’s total capital outflows, including unrecorded capital, averaged about 3 per cent of GDP in 1993–98 (UN and ECE, 2000). This large outflow was made possible by the large current account surplus and foreign borrowing. The growth of capital flows to transition economies during the 1990s was associated with a change in their distribution (Table 5.2). While the five leading Central European transition economies (the CETE-5)5 underwent a steady decrease in official financing and had even repaid their IMF debts by 1996, the Russian Federation received increased funds after 1994. These were mostly IMF credits and grant aid, which accounted for most of the official financial flows to transition economies. Additional flows were German transfers to the former Soviet Union as part of the

–500

Source: UN and ECE (2000).

* Including errors and omissions (capital flight). Russian Federation*

Russian Federation

Uzbekistan

Tajikistan

Ukraine

Republic Of Moldova

Bulgaria

Albania

Kyrgyzstan

Georgia

Belarus

Kazakhstan

Turkmenistan

Armenia

Azerbeijan

Romania

Latvia

FYR Macedonia

Poland

Bosnia-Herzegovina

Slovenia

Lithuania

Slovakia

Estonia

Croatia

Hungary

Czech Republic

104

2500

2000

1500

1000

500

0

Figure 5.1 Capital flows into transition economies, 1993–98 (US dollars per capita)

105 Table 5.1 Net capital flows into East European transition economies, by type of flow, 1993–19981 Trade flows US$ billions Albania Bosnia-Herzegovina3 Bulgaria Croatia Czech Republic Hungary Poland Romania Slovakia Slovenia FYR Macedonia Estonia Latvia Lithuania Armenia Azerbeijan Belarus Georgia Kazakhstan Kyrgyzstan Republic of Moldova Russian Federation Tajikistan Turkmenistan Ukraine Uzbekistan Total4 Russian Federation5

Per capita

Private flows Per GDP2

Total Long-term

0.9 3.8 2.5 7.6 22.7 20.5 32.4 12.4 8.3 2.2 1.5 2.4 1.5 5.1 1.7 4 3.8 1.8 6.1 1.5 1.1 40.8 0.8 1.7 8 2.9

298 1,082 292 1,686 2,208 2,017 837 550 1,547 1,094 748 1,646 595 1,389 473 528 366 348 372 325 252 277 129 392 156 124

111 – 62 250 169 204 112 87 163 78 175 218 106 223 229 256 62 105 77 141 113 40 136 148 47 59

26 – 47 159 154 207 80 44 148 108 20 131 109 63 14 99 14 6 76 26 43 21 29 156 23 39

39 – 1 137 112 160 61 42 73 108 5 100 103 55 12 99 10 6 71 23 40 17 23 110 17 30

91.9

231

39

56

43

2.6

18

3

21

17

Notes: 1. Total flows are the sum of the capital and financial accounts plus errors and omissions, as reported in the national balance of payments statistics. Total private flows include FDI, long-term private guaranteed and non-guaranteed debt, short-term debt and portfolio equity flows. 2. Per $100 GDP in 1997. These are purchasing power parity (PPP) estimates of GDP. 3. 1994–98. 4. Excluding Bosnia-Herzegovina. 5. Excluding errors and omissions from total flows (capital flight). Source: UN and ECE (2000), p. 149.

106 Monetary Policy in Transition Table 5.2 Net capital flows to the CETE-5 and Russia, by type of finance, 1993–98 (percentage of GDP) CETE-5

Russian Federation

1993–95

1996–98

1993–95

1996–98

Capital transfers1 FDI Long-term debt External bonds IMF Short-term funds Portfolio investment2 Short-term flows Errors and omissions

1.8 2.3 0.9 1.4 0.5 1.3 0.6 0.3 0.4

0.1 2.6 0.3 0.4 0.1 2.3 0.6 0.9 0.8

0.2 0.4 0.7 0.1 1.1 2.0 0.2 0.1 1.9

0.1 0.7 0.3 1.5 0.9 2.4 4.0 3.9 2.5

Total net flows Total flows (US$ bn)

6.2 40.5

5.2 45.3

2.1 16.5

2.1 24.3

Notes: 1. Includes debt write-offs under debt restructuring agreements; this was especially important for Poland during 1993–95. 2. Excludes external bonds. Source: UN and ECE (2000), p. 151.

German reunification agreement. As the transition progressed private flows (FDI, long-term debt and short-term debt) began to dominate the capital market. An important proportion of these were so-called nondebt-creating inflows, notably FDI. On average the share of FDI in total net inflows was higher in transition economies than in developing countries.6 Despite being among the top 10 recipients (the group of countries that received more than 70 per cent of FDI flows) per capita FDI in Russia was disappointingly low. It rose significantly after the opening of the Russian economy, but thereafter political instability and the poor business climate deterred many foreign investors. Although the trend for Russia was generally upward during 1990–98, much more FDI was received by the CETE-5 countries and the Central and East European transition economies as a whole (Figure 5.2). This difference was even more pronounced if per capita FDI is considered instead. On average the transition economies received cumulative FDI inflows of $439 per capita between 1989 and 1997, but Russia received only $63 in the same period (EBRD, 1998); many times less than that received by all the transition economies but the Former Yugoslav Republic of Macedonia and somewhat

The Role Played by International Financial Institutions 107 25,000 20,000 15,000 10,000 5,000 0 1990

1991

1992

1993

1994

1995

1996

1997

1998

All Central and East European transition economies CETE-5 Russia Figure 5.2 Flows of FDI to transition economies, 1990–98 (US$ millions) Source: UN and ECE (2000) and author’s own calculations.

less than the average amount received by the Commonwealth of Independent States ($84). At the end of 1997 the total FDI stock in Russia was only half the amount invested in Hungary in 1989–97. The flow of capital to Russia rose to a peak in 1997. This was a result of Russia’s improved economic policy environment and financial position, plus the successful rescheduling of agreements with Paris and London Club creditors in 1996 and 1997 respectively. However most of these inflows were short-term investments in government securities and equities and Russia became increasingly vulnerable to shifts in market sentiment. This was manifested in an increase in the already sizeable capital outflows following the onset of the Asian crisis in late 1997. The widening of the gap between the inflow and outflow of capital culminated after the August 1998 crisis in Russia and the latter default on its short-term securities. Our analysis of flows of capital in Russia during transition would be incomplete without considering external debt and its implication for the stabilization of the economy. Table 5.3 compares the external debt of and debt indicators for the Central and East European transition economies with those for Russia in the 1990s. Notably, Russia’s total external debt equalled the combined debt of all the other transition economies in 1994 and surpassed it in 1998. In

108 Table 5.3 Debt indicators for the transition economies, 1990–98, (US$ billions)

Russia:1 Total external debt (TED) Long-term debt Concessional Bilateral Multilateral Official non-concessional Bilateral Multilateral IMF credit World Bank credit Private creditors Bonds2 Commercial banks2 Short-term debt Memorandum item IMF credits/TED (per cent) Central and Eastern Europe: Total external debt Long-term debt Concessional Bilateral Multilateral Official nonconcessional Bilateral Multilateral IMF credit Private creditors of which Bonds2 Commercial banks2 Short-term debt

1990

1991 1992 1993

59.8 48.0 0.0 0.0 0.0 5.9 5.5 0.4 0.0 0.0 42.1 1.9 17.9 11.8

67.8 55.2 0.7 0.7 0.0 8.8 8.4 0.4 0.0 0.0 45.6 1.9 16.8 12.6

0.0

0.0

109.3 117.7 91.1 102.0 5.2 4.9 5.1 4.7 0.1 0.2 36.6 47.6

1994

1995 1996

1997 1998

78.4 111.7 121.5 120.3 124.9 126.0 183.6 65.2 103.4 111.6 110.0 112.8 120.0 165.2 1.0 2.1 2.5 2.5 2.5 2.3 2.3 1.0 2.1 2.5 2.5 2.5 2.3 2.3 0.0 0.0 0.0 0.0 0.0 0.0 0.0 10.8 54.6 64.5 64.3 72.9 74.2 85.3 9.3 50.8 58.8 52.7 57.6 55.7 59.4 0.5 1.3 1.5 2.0 2.8 5.3 6.6 1.0 2.5 4.2 9.6 12.5 13.2 19.3 0.0 0.0 0.6 1.5 2.6 5.3 6.4 53.4 46.7 44.6 43.2 37.4 43.5 77.6 1.7 1.6 1.8 1.1 1.1 4.6 16.0 18.5 15.9 16.4 16.7 15.6 29.3 29.3 13.1 8.3 9.9 10.4 12.1 6.1 18.4 1.3

2.2

3.5

8.0

10.0

10.5

10.5

113 116.8 121.5 138.3 139.8 141.0 156.3 99.9 104.1 109.5 120.8 121.7 116.5 127.8 14.3 13.6 12.1 13.6 12.8 10.5 11.5 14.2 13.4 11.9 13.3 12.5 10.1 11.0 0.1 0.2 0.2 0.3 0.3 0.4 0.5 38.6 39.6 41.9 42.3 40.0 37.0 36.5

28.7 6.6 1.3 49.3

34.7 7.9 5.0 49.5

24.3 8.9 5.4 47.0

24.9 9.3 5.4 50.9

25.1 11.1 5.7 55.5

26.9 12.3 3.1 64.9

25.9 11.9 2.2 68.9

23.5 11.0 2.5 69.0

23.5 10.6 2.4 79.8

5.0 34.7 18.2

6.7 33.6 15.7

7.4 30.8 13.1

11.7 29.1 12.7

28.1 14.4 12.0

30.9 16.6 17.5

29.1 20.2 18.1

26.1 21.1 24.5

28.3 21.4 28.5

Debt indicators (per cent): Ratio of external debt to GNP 10.3 12.5 18.6 29.1 37.9 35.3 29.6 28.8 69.4 Russia1 Central and Eastern Europe 38.8 65.1 58.2 50.9 47.3 43.5 41.7 44.4 44.2 Ratio of external debt to exports 73.8 124.8 143.0 169.8 156.7 129.5 119.3 121.4 207.1 Russia1 Central and Eastern Europe 179.4 214.9 165.5 141.0 121.8 103.3 98.3 100.2 103.0 Ratio of debt service to exports 14.6 42.8 2.5 3.3 4.4 6.4 6.7 6.4 12.1 Russia1 Central and Eastern Europe 20.9 19.5 16.8 11.7 14.4 13.6 13.5 14.1 15.9 Notes: 1. Soviet Union Prior to 1992. 2. Government or government-guaranteed debt only. Source: UN (2000); IMF (1999).

The Role Played by International Financial Institutions 109

addition Table 5.3 implies that, in contrast to the situation in the other economies, the importance of official credit to Russia, and that by the IMF in particular, grew over time. However, and again in contrast to the other economies, official credit was not significant in Russia at the beginning of transition, which arguably was when Russia needed it most. One can also argue that the extent of IMF lending to the transition economies was hardly excessive (Table 5.4). As can be seen in Table 5.5 actual foreign currency disbursements relative to gross national product (GNP) and total external debt (TED) were not particularly high, even during the period when they reached their peak. Moreover as far as Russia was concerned, the size of official disbursements rarely matched what had been promised, as will be discussed later. IMF foreign currency disbursements to Russia from the beginning of transition until the end of 1998 totalled just over $20 billion. It was not only IMF credits that were important for Russia’s financial consolidation. As mentioned earlier a growing part was played by private financing and other sources of official finance, particularly bilateral and multilateral credits, as shown in Table 5.3. This might not have

Table 5.4 Net IMF lending to the transition economies, by facility, 1990–98 (US$ billion) 1990 1991 1992 1993 1994 1995 1996 1997 1998 Regular facilities Repayment terms 3–5 years (credit tranche) 3.5–7 years (SFF/EAP) 4–10 years (EFF) Concessional facilities (ESAF) Additional facilities Compensatory financing STF Total Memorandum items: Selected characteristics of higher conditionality lending agreements Number initiated during year Average length (months) Total amount committed (US$ billion)

0.1

2.0

1.9

0.1

4.4

3.8

2.2

0.4 0.3 0.0

1.0 0.2 0.8 0.0

1.8 0.0 0.1 0.0

0.1 0.5 4.9 0.0 0.3 0.0 0.0 0.0 0.5 0.0 0.0 0.1

1.2 0.0 2.6 0.2

0.0 0.8 0.0 0.0 2.2 3.9 0.2 0.2

0.0 0.0 0.1

1.5 0.1 0.0 0.0 3.5 1.8

0.0 0.7 0.6 0.2 2.0 2.8 0.9 0.0 2.1 2.3 4.7 3.6

0.1 2.9 0.0 0.5 2.3 5.5

3 12 1.6

5 12 4.9

9 18 1.6

7 21 2.1

6 12 1.5

0.2

8 18 2.1

12 13 9.2

12 28 13.2

3.1

6 32 3.4

Notes: SFF Supplemental Financing Facility, EAPEnlarged Access Policy, EFFExtended Fund Facility, ESAFEnhanced Structural Adjustment Facility, STFSystemic Transformation Facility, Source: UN (2000).

110 Monetary Policy in Transition Table 5.5 Foreign currency disbursements to the Russian federal government, 1994–98 (US$ million) 1994

1995

1996

1997

1998

Total

Multilateral IMF1 World Bank EBRD Other Bilateral Tied Untied Bonds2 Suppliers/other commercial Total Total excluding IMF

1,931 1,544 280 6 101 2,057 2,057 0 0 507 4,496 2,952

6,319 5,450 826 43 0 1,554 1,554 0 0 93 7,966 2,515

4,940 3,758 1,107 75 0 3,280 1,090 2,190 1,000 0 9,220 5,462

4,777 2,019 2,699 59 0 1,375 1,375 0 3,549 1,136 10,836 8,817

7,519 6,240 1,219 60 0 2,110 2,110 0 9,615 156 19,399 13,160

25,486 19,011 6,131 243 101 10,376 8,186 2,190 14,164 1,892 51,918 32,907

Memorandum items: GNP Total external debt (TED) IMF disbursment/GNP IMF disbursment/TED Multilateral/GNP Multilateral/TED

320,580 121,500 0.48 1.27 0.60 1.59

391,784 120,300 1.39 4.53 1.61 5.25

472,297 124,900 0.80 3.01 1.05 3.96

489,583 126,000 0.41 1.60 0.98 3.79

225,216 183,600 2.77 3.40 3.34 4.10

Notes: 1. Full amount of IMF purchases. In 1998 part of this amount was disbursed directly to the Central Bank of Russia. 2. The figure for 1998 includes $3,700 of Eurobonds purchased by residents. Data on resident purchases in other years were not available. Source: IMF (1999) and author’s calculations.

come had Russia not received the green light for its reforms by the IMF. In addition, prior to the August 1998 crisis Russia had come to key rescheduling agreements with Paris and London Club creditors (Table 5.6). The agreements shown in Table 5.6 allowed a substantial lengthening of the maturity structure of sovereign debt7 and reduced debt service pressures. Moreover after two years the London Club wrote off a third of the debt due in February 2000, as well as accepting an interest service reduction (Hishow, 2001). Consequently the real payments amounted to just 40 per cent of the due payment, as shown in Table 5.7. It has been claimed that the reduced debt service expenditure, totalling $80 billion, or about half of the Russian GDP for 2000, represented a significant contribution by the West to the balance of payments (ibid.). Moreover Hishow argues that the size of Russia’s external debt may allow it to abuse its dominant position as a debtor and default on its external obligation, as it did with its internal debt in the wake of the August 1998 crisis (ibid.). The worry for the West is that Russia may

1 Jan. 91 1 Jan. 91 1 Jan. 91 1 Jan. 91 1 Jan. 91

Date of agreement

2 Apr. 933 2 June 94 3 June 95 15 Apr. 96 6 Oct. 97

1 Jan. 93 1 Jan. 94 1 Jan. 95 1 Jan. 96 1 Jan. 97

Start date 12 12 12 Stock n.a.

Length (months) Yes No No No No

Arrears included 100 100 100 100 n.a.

Share of debt consolidated (per cent) 14,497 7,100 6,400 40,200 32,500

Amount consolidated (US$ mill.)

10/0 15/2 15/4 21/5 25/0

Maturity (years/ months)

6/0 2/9 2/10 2/11 6/0

Grace (years/ months)

Repayment terms2

Sources: IBRD and World Bank (1999); Hishow (2001).

Notes: 1. The figures in this table are commitment values (amounts of agreed debt relief). They should correspond to the disbursement figures (minus debt forgiveness, when applicable). All the agreements were negotiated through the Paris Club or the London Club. 2. Maturity is measured here from the end of the consolidation period to the date of the final amortization payment; the grace period is the time between the end of the consolidation period and the date of the first amortization payment. The secretariat of the Paris Club measured grace and maturity from the midpoint of the consolidation period. 3. Agreement followed the deferral signed in January 1992 by the former Soviet republics.

Contract cut-off date

Consolidation period for current maturities

Table 5.6 Multilateral debt relief agreements with official creditors, 1990–981

111

112 Monetary Policy in Transition Table 5.7 Financial relief through restructuring and payment deferrals (US$ billion) Debt service

1993 1994 1995 1996 1997 1998 1999 2000 Total

Saving

Debt to service ratio (%)

Saving

Due

Real

(duereal)

Due

Real

(duereal, %)

20.7 20.2 20.2 18.1 13.7 13.1 17.5 17.5 41.0

3.6 4.6 7.1 7.1 7.4 7.8 11.0 13.0 61.6

17.1 15.6 13.1 11.0 6.3 5.3 6.5 4.5 79.4

31.8 29.8 24.4 20.0 15.4 17.5 25.0 23.3 –

5.5 6.8 8.6 7.8 8.3 10.4 15.7 17.3 –

26.3 23.0 15.8 12.1 7.1 7.1 9.3 6.0 –

Source: Hishow (2001).

either impose a unilateral moratorium or try to achieve an infinite debt restructuring, which in the end may result in the same thing. It could be argued that the continuation of lending to Russia despite its unsustainable policies may be partly based on expectations of support from the international financial community. In other words, there is an implicit moral hazard, with creditors continuing to invest in Russia in the belief that it is too big to be allowed to fail. Reportedly, in the wake of the 1997–98 crisis the IMF’s sceptical officials were persuaded by the US Treasury to provide a loan package on the ground that Russia was ‘too nuclear to go bust’ (Hale, 1998). It is difficult to determine the optimal level of financial assistance for any country, and particularly for Russia given its military might and vast economic resources. The IMF faced the same dilemma when determining the size of its rescue package after the financial crisis. On the one hand sufficient finance was necessary to meet Russia’s anticipated foreign exchange needs, restore market confidence and ensure the success of the stabilization programme. On the other hand the provision of very large amount of finance risked a moral hazard. Between July 1997 and October 1998 the international community pledged about $187 billion to support Indonesia, Korea, Russia, Brazil and Thailand (Table 5.8). However due to these countries poor performance with the IMF-agreed reform programmes the actual disbursements amounted to about one third of the pledged funds for all countries. In the case of Russia, due to the disappointing results of its reforms only the first tranche of the IMF loan

The Role Played by International Financial Institutions 113

($4.5 billion) was disbursed. That represented about 7 per cent of total disbursements, or approximately a quarter of the sum pledged to Russia. Following mounting fiscal and financial difficulties, the August 1998 ruble devaluation, unilateral restructuring of the GKO debt, the 90-day moratorium on private debt repayments and the ongoing weakness in oil prices, the flow of finance was cut off for a while. Looking at Table 5.8 one cannot help but notice that the funds pledged to Russia were considerably less than those funds designated for all the other countries but Thailand. Moreover the actually funds disbursed to Russia were much less than those received by the other countries. This has led Hale (1998) to argue that the Russian package was too small to be effective. According to Hale, had the package for Russia ($22 billion) been similar to that for Mexico in 1995 ($40 billion), investors would probably not have fled. Furthermore, had more of these fund been disbursed investors would have been less sceptical. In the event investors judged that, given Russia’s delicate financial situation, the pledge of $22 billion would not be enough to bail Russia out. The evidence presented here gives credence to the claim that Western aid to Russia was not sufficient to enable Russia to go through transition smoothly (Sachs, 1997; Hale, 1998). In fact the West’s contribution to Russia could hardly be called aid as it was clear that the credits would have to be repaid sooner or later. However, although the credits were cheaper than could be obtained from the financial markets, the Russian

Table 5.8 International Rescue packages, East Asia, Russia and Brazil, July 1997 to October 1998 (US$ billion) Funds pledged

Funds disbursed

World World IMF Multilateral1 Bank Bilateral Total IMF Bank Others Total Indonesia Republic of Korea Thailand Russia Brazil

11.2 20.9 4.0 11.2 18.0

10.0 14.0 2.7 1.5 9.0

5.5 10.0 1.5 1.5 4.5

26.1 23.3 10.5 9.9 14.5

47.3 6.8 1.3 58.2 18.2 5.0 17.2 3.0 0.8 22.6 4.5 41.5 4.62

Total

65.3

37.2

23.0

84.3 186.8 37.1

7.1 18.3 62.5

Notes: 1. World Bank, Asian Development Bank and Inter-American Bank. 2. Package approved in December 1998, first IMF disbursement in January 1999. Source: IBRD and World Bank (1999).

1.4 9.5 4.0 27.2 8.9 12.7 4.5 4.0 8.6

114 Monetary Policy in Transition

government was expected to accept the conditions laid down by IMF ideologues and policy makers.

Common criticisms of the IMF programmes Although every society has it own priorities the broad objectives of economic policy do not usually differ. These include a high rate of growth, low inflation, the alleviation of poverty, social stability, an adequate supply of public goods and services, and as little income inequality as possible. Criticisms of the IMF take various forms and follow several lines. In general critics consider that some elements of typical IMF programmes are discordant with the broad objectives of economic policy. One line of criticism is that the macroeconomic approach underlying the IMF’s approach to stabilization is fundamentally wrong (Taylor, 1988). Taylor’s criticism is rooted in a contrasting view about the nature of inflation, the relative importance of fixed-price and flexi-price markets, the part played by forced saving and output adjustment, the dynamics of economic growth and the sensitivity of specific balance of payments and financial linkages to policy measures. Another line of criticism is that IMF is a hermetic institution whose standard stabilization package is not sufficiently responsive to the everchanging conditions in the global economy and the evolution of professional thinking. More specifically the IMF’s financial programming, based on Polak’s (1957) model, is viewed as rather dogmatic, somewhat outdated and ill-suited for crises, such as the collapse of central planning and the financial crises in Asia and Latin America in the 1990s (see Taylor, 1988 for details). Other criticisms relate to the structural elements of reform packages. The critics charge that the IMF staff lack both expertise and a mandate to convey advice and design conditionality on structural issues.8 Feldstein (1998) argues that the legitimate political institutions of a country should determine that country’s economic structure and the nature of its institutions, rather than being governed by the IMF via its conditionality provisions. Instead the IMF’s role should be limited to providing technical advice and limited financial assistance. According to Feldstein (1998), if a nation is in a desperate need for short-term financial help it should not give the IMF ‘the moral right to substitute its technical judgments for the outcomes of the nation’s political process’. Soros (2000) identifies two interconnected asymmetries in the IMF’s operating practices: a disparity between crisis prevention and crisis intervention, and a disparity between treatment of lenders and borrowers.

The Role Played by International Financial Institutions 115

The first disparity stems from the fact that the IMF cannot provide any debt relief to debtor countries during a crisis since that could have devastating effects on financial markets. Only after the crisis has been weathered can any debt relief be provided. The second disparity is explained by the political economy of the IMF. Soros asserts that the international financial architecture is skewed towards the centre. This implies that the countries at the centre of the global financial system control the IMF. Therefore it would go against the national interest of controlling shareholders if the IMF penalized lenders. The net effect of this is that the burden of adjustment is placed mainly on the borrowing countries by compelling them institutionally to service their debt, which usually stretches them to the limits of their capacity. This feature of the IMF reportedly played an important part in shaping investors, expectations in Russia in 1998 (ibid.). That is, many investors continued to buy Russian treasury bills (GKOs) despite the fact that fiscal and monetary indicators clearly indicated an imminent crisis. Their actions were influenced by the view that Russia was too important to be denied an IMF bailout. In the event, recognition of the moral hazard inherent in the IMF’s method of operating made a bailout politically unacceptable (ibid.). Thus Russia was doomed to default and the GKO holders to financial disaster.

Specific policy considerations of the IMF’s involvement in postcommunist Russia The questions ‘what went wrong’ and ‘who lost Russia’ have been prominent in the literature for several years. While answering these questions might be useful for policy analysts and others on both sides of the Atlantic it is beyond the scope of this chapter.9 Instead we shall list specific errors made by the IMF in Russia. First, IMF policies contributed to short-termism in policy making and diverted attention from strategic policy making. The IMF’s priorities, for the IMF, irrespective of changes in conditionality, seemed to be related to short-term financial performance criteria. That is, the IMF tended to impose quarterly ceilings on the nominal value of the fiscal deficit, quarterly or even monthly ceilings on the expansion of net domestic credit by the central bank, and quarterly floors on net international reserves. This tended to fix the focus of policy makers on the very short term. Second, there was a huge underestimation of corrective inflation after the liberalization of prices in January 1992. While the IMF and the Russian prime minister respectively estimated that the monetary

116 Monetary Policy in Transition

overhang would be about 50 per cent (Gros and Steinherr, 1995) and the price jump no more than 100 per cent (Rossiiskaya Gazeta, 3 July 1992), prices actually jumped 245 per cent in January 1992 (see Chapter 2, for the probable causes of this miscalculation). This error clearly undermined the credibility of the Gaidar government’s budget for the first quarter of 1992. Third, the IMF’s support for the ruble zone, an arrangement whereby 15 former Soviet Republics continued to use a common unconvertible currency – the Soviet ruble – after the break-up of the Soviet Union in December 1991 was a costly mistake. This arrangement was intended to minimize the severance of links between enterprises in these states, thus reducing the magnitude of trade shocks and falls in output. In addition it is thought that the IMF considered the non-Russian republics to be unfit to manage their own currencies (Sachs, 1997). According to the IMF, it tried to limit inflation by providing a ‘set of rules for a coordinated monetary policy’ (Hernández-Catá, 1995). The establishment of the ruble zone probably did absorb some of the trade shocks between these states and safeguarded some non-Russian industries from immediate collapse. As discussed in Chapter 1, these industries were able to obtain unauthorized credits (‘non-cash’ rubles) from the Central Bank of Russia via their national central banks. However in July 1993 the CBR withdrew the pre-1993 ruble notes, thereby ensuring the demise of the ruble zone. Fourth, the financial support provided to Russia was insufficient to bring about stabilization in the early stages of the reforms. There is some disagreement on this point among economists. For example Sachs (1997) argues that the West should have financed the Russian budget deficit to the order of 5 per cent of GDP, while the IMF maintained that Russia should aim for a balanced budget. According to Sachs the expectation that, in the midst a deep crisis, Russia could have reduced it budget deficit to below the levels that prevailed in most OECD countries was utterly unrealistic. Along similar lines Portes (1994) argues that the main policy error was overemphasis on macroeconomic policy. Gomulka (1995) takes a different stand and suggests that, while helpful, external financial support was not essential to stabilization. Similarly, since the financing of the budget deficit usually creates a debt which needs to be paid in the future Hernández-Catá (1995) has recommended against this course of action. Whatever the merits of larger or smaller external assistance, the fact remains that the West, principally via the IMF, consistently failed to live up to its promise from the very start of its involvement with Russia. At the

The Role Played by International Financial Institutions 117

beginning of 1992 the Russian government was promised US$24 billion of aid, to be disbursed in the second half of that year, but very little of this was received. Neither the $0.719 billion IMF standby arrangement, nor the $6 billion exchange rate stabilization fund, nor the funds associated with an official debt rescheduling deal materialized. An interim IMF credit of $1 billion was disbursed in August 1992, but this was for reserve purposes only (Granville, 1995). Moreover the $670 million of assistance approved by the World Bank in 1992 was not disbursed until the end of 1993. The failure of the West to the deliver the promised aid happened when political climate was adverse for the reformist government, and at the end of 1992 the prime minister was forced to resign. Subsequently, the reforms lost their appeal and there was an unmistakable reduction in the momentum of the reform at the beginning of 1993 (ibid.). The $28 billion aid package announced in July 1993 suffered a similar fate to that announced in April 1992. Neither of these packages was ever properly elaborated or delivered (Sachs, 1997). Sachs further argues that both the G7 and the IMF failed to understand that Western financial assistance for the Russian budget was a conditio sine qua non to achieve financial stabilization. In July 1997 the IMF pledged yet another package of $22 billion, but in the event only $4.7 billion was disbursed. As investors viewed this package as too small to bail out Russia they duly fled. The IMF would argue that the reason for the failure to disburse all of the pledged funds was the consistent failure of the Russian government to hit the agreed targets. However others argues that the targets were unrealistic (ibid.) and the government was never likely to achieve them. Fifth, contrary to the partial cancellation of the Polish debt, the West wrote off none of the old Soviet debt and Russia was expected to service it. This put a considerable strain on the already deteriorating government finances. Debt forgiveness always hinges on good will, and this appears to have been lacking in the Western democracies. Sixth, the IMF failed to incorporate in its programme a non-monetary means of financing the budget deficit (ibid.). This could have included treasury bills and bonds. The trouble with these instruments is that their rate of return has to be very high to make them attractive. This in turn increases the burden on the government’s finances and enlarges the deficit in the long turn. Unless the rate of return is reduced significantly the government’s finances may be viewed as unsustainable, prompting a run on currency or even default. Despite this unpleasant arithmetic, non-monetary financing of the budget is largely perceived as less detrimental to the national economy than a monetary one. Yet there was no

118 Monetary Policy in Transition

significant non-inflationary financing of the deficit in Russia in the first years of the transition. Seventh, the IMF’s failure to recognize the presence of both inflation inertia and inflationary expectations before the summer of 1995 led to the adoption of money-based stabilization measures when an exchangerate approach with heterodox elements would have been more suitable. Admittedly this would have required more foreign exchange reserves than Russia had at the time, but this is exactly where external financial support should have come into play. After all, one of the roles of the IMF is to provide short-term liquidity to countries with ailing finances. Finally, as discussed in Chapter 4 the existence of bilateral causality between inflation and broad money in postcommunist Russia meant that monetary targeting was unsuitable. Yet monetary targeting was a centre piece of the early stabilization efforts guided by the IMF. In effect monetary targets were imposed in a non-monetary economy (Soros, 2000). Consequently not only were the targets bound to be missed, but even it they had been met inflation would not necessarily have been brought under control. Most of the errors listed above were related to the IMF’s approach rather then to implementation failures. However the fourth error could be classified as an error in both implementation and design although Gomulka (1995) argues that most of the errors were in implementation rather than design.10 The IMF would argue that its recommendations were good but were never fully implemented. As pointed out by Stiglitz (2002), in economics it is impossible to follow a prescription precisely. Instead policy makers should consider what is feasible to achieve, given the complexity of political processes and the fallibility of individuals (ibid.). Whatever the causes of the errors they significantly hindered the stabilization efforts and made them more difficult and costly (Sachs, 1997). As the Russian public had lost confidence in the ruble there were massive capital flights throughout the 1990s (Table 5.9). This loss of confidence also made it difficult to finance the budget deficit by non-monetary means. In particular the public was unwilling to buy Russian treasury bills, except at enormously high interest rates. Likewise the decline of the broad money to GDP ratio increased the inflationary consequences of financing the budget deficit via the Central Bank of Russia. Moreover as Sachs (ibid.) argues because of the delay in stabilizing the economy there was a rise in tax evasion and tax exemption. This resulted in a significant decline in tax revenues, thus further exacerbating the budgetary problem. Another consequence of the delay was that regional governments

The Role Played by International Financial Institutions 119 Table 5.9 Capital flight from Russia, 1994–98 (US$ billion) 1994

1995

1996

1997

1998

273

8,751

5,674

8,030

9,237

Source: IMF (2000).

captured an ever-larger share of total revenue, at the expense of the federal government and the financing of the budget deficit (ibid.). A particularly strong criticism of the IMF’s conduct in Russia came from the Nobel Prize winner Joseph Stiglitz, who at the time was chief economist at the World Bank. Stiglitz (2002), argues that the IMF helped to create an ‘ersatz capitalism’ that instead of providing incentives for wealth creation and economic growth was more conducive to asset stripping. This particular kind of ‘Wild East’ capitalism was, according to Stiglitz, the result of various mistakes by the Washington Consensus, such as neglect of competition policy, errors in the sequencing of reforms, excessive zeal in fighting inflation, and putting aside the issues of poverty, inequality and social capital. The lack of a competition policy and the failure to create the required infrastructure for a market economy enabled privatized firms to establish monopolies and cartels. Price liberalization before privatization wiped out savings and imposed stabilization, which in turn prevented growth because it led to high interest rates and asset stripping since it was too expensive for businesses to expand. Asset stripping was also a consequence of privatization being done the ‘wrong way’ and of the liberalization of the capital markets. Rapid privatization was also responsible for the decline in income and the increase in inequality. Stiglitz further argues that the IMF’s loans to Russia were harmful because they set back the reform agenda. They made the country more indebted and impoverished, contributed to the overvalued exchange rate and involved deliberate intervention in the politics of the country.

Alternative medicine: more prescriptions, fewer injections Given the policy mistakes and the controversial role of the IMF in Russia, one wonders why developing countries should need IMF support. Hale (1998) offers three reasons why such support is important. First, the IMF provides macroeconomic and other policy advice that politicians can sell as their own. Even though the public is aware that the US Treasury and the G-7 countries heavily influence the IMF, it still

120 Monetary Policy in Transition

gives an impression of autonomy so that its highly specialized and wellknown officials can make policy advice more politically acceptable to borrowers. Second, in the present global financial environment the IMF acts as a lender of last resort, similar to the role of the central bank during domestic banking crises. Consequently it can help troubled economies to alleviate their short-term liquidity problems and restore investors’ confidence. Third, the IMF can initiate microeconomic reforms that might otherwise be politically unacceptable. The emphasis of such reforms is on non-inflationary economic growth. In addition to the above, the role of the IMF as a provider of credibility cannot be overemphasized (Cottarelli and Curzio, 1998). Instead of delegating monetary policy to a domestic entity, an alternative way of enhancing the credibility of adjustment policies in developing countries is to submit these policies for the approval of a supranational organization such as the IMF. The growing part played by the IMF as a provider of credibility rather than a lender of resources is supported by three pieces of evidence (ibid.). First, the share of net IMF credit in total net external financing (including FDI) to developing countries fell from 4.25 per cent during the 1980s to less than 1 per cent in 1990–96. Second, the number of precautionary programmes has increased over the years: by the end of 1996 one third of standby and extended fund facility (EFF) arrangements were precautionary; that is, they had been undertaken without any intention of drawing funds. Third, the ratio of actual borrowing to potential borrowing in all outstanding IMF arrangements has declined since the 1980s, while the number of countries with IMF-approved programmes has risen to a historical peak. Overall, the IMF’s stamp of approval has been in demand to enhance authorities’ credibility and give a clear signal to investors that a country is relatively safe for investing.

Conclusion Transforming a world superpower into a market economy after three quarters of a century of central planning was an unprecedented task with an unpredictable outcome. This process has taken longer than anticipated and the results have been less than desirable. But despite being characterized as a country of ‘robber capitalism’ rather than a market economy, Russia is now far removed from days of central planning and autarky. Today’s Russian-style capitalism was brought about by Russian reformers supported by the West and guided by the IMF. The

The Role Played by International Financial Institutions 121

IMF prescribed the medicine and the West spoon-fed Russia just enough to pass the point of no return. The downside of the process, as argued in this chapter, was that the prescription was often inappropriate and the medicine quite inadequate. Even when the prescription was right the patient was usually given too little too late. One should bear in mind that a large share of responsibility for the outcome of the reforms lies with Russia’s leading politicians. Not only were they errant but it was also in their interest to postpone true reforms. The ‘reformers’ sins aside, the IMF has been less than generous with its financial and technical assistance. It has consistently underestimated the amount of the western assistance needed for successful stabilization in Russia. Because of conditionality clauses it has disbursed far fewer resources than were pledged, not to mention the lack of desire for debt forgiveness. All the Russian stabilization programmes had IMF approval and all of them failed. Finally, the IMF arguably made a significant number of policy mistakes that aggravated the long-suffering Russian economy. Admittedly the transition process was a huge undertaking and errors were inevitable. Nevertheless, given the reputation and enviable resources of the IMF, one cannot help but think that it could have done much better and that at least some of Russian’s socioeconomic pains during transition could have been avoided.

Conclusions and Directions for Further Research This book has scrutinized critically and rigorously monetary issues in postcommunist Russia and in particular the relationship between inflation and various monetary aggregates as well as the role of the International Monetary Fund in that process. Methodologically the emphasis has been on the application of empirical analysis and rigorous econometric testing of the results. All the main empirical findings that emerge from the book have implications for economic policy. The empirical results imply that macroeconomic policies adopted in Russia under the auspices of the IMF during the period under consideration may not have been optimal. In particular, the study argues that lack of coordination of fiscal and monetary policies considerably contributed to the failure of all stabilization programmes implemented prior to August 1998. In addition, having identified a high degree of inflation persistency in the Russian economy, an unstable relationship between money supply and inflation with transmission of monetary impulses to future inflation becoming both slower and weaker, and the existence of feedback or bilateral causality between inflation and aggregates of broad money, this study suggests that, instead of money-based stabilizations, exchange-rate-based stabilization might have been more suitable for this transition economy. Coordination of fiscal and monetary policies and macroeconomic stability, though necessary, are not sufficient conditions for the success of economic reforms. Liberalization, privatization and creation of market institutions are also cornerstones of reform. Throughout the 1990s Russia struggled to put these cornerstones in order. Partial and tardy liberalization was a fertile ground for state intervention, corruption and rent seeking. Privatization on the other hand, was rapid but unreal. In the absence of market institutions privatization was an exercise on an unprecedented scale of transferring state assets, including natural resources, to a small minority of incumbents. The process resulted in greater concentration of ownership than in any advanced market economy. Insider privatization in the 1990s, instead of being efficiency enhancing, ended up as an asset-stripping and rent-seeking episode with a loss of almost half of national output (while the USA lost just about a third during the Great depression of the 1930s). 122

Conclusions and Directions for Further Research 123

One of the reasons that the Russian reforms were not successful in the 1990s was the lack of independence of monetary policy and the multitude of conflicting objectives it had. These objectives included providing liquidity to ailing enterprises in Russia and ‘near abroad’, recapitalizing the banking sector, and financing the government budget deficit. Such conflicting objectives formed a poor environment for the creation of market instruments for the implementation of monetary policy. Even so, indirect monetary instruments such as open market operations, a Lombard facility, deposit auctions and reserve requirements slowly evolved, though mostly by trail and error, between 1995 and 1998. However, after the August 1998 financial crises and default on domestic debt, the development of these instruments was reversed. In the subsequent period the central bank mainly relied on raising the legal reserves ratio and intervention in the foreign exchange market. Targeting monetary aggregates was the main instrument of Russian monetary policy during the 1990s, except for the period 1995–98, when an exchange rate band was maintained. Even though various stabilization programmes were implemented in this period under auspices of the IMF, and all of them failed, the performance of the economy was worse under money-based programmes: failure to stabilize inflation, severe contraction of economic activity and diminishing importance of all functions of domestic currency. Even the exchange rate programme implemented in 1995, after promising results in the medium term, ended up in August 1998 in the most serious Russian transformation crisis to date. The programme caved in as a result of adverse international climate with regard to investment and commodity prices and an unsustainable build up of domestic debt in an economy which neglected structural reforms. Government default on domestic debt, a 90-day moratorium on some foreign debt repayments, and currency devaluation in August 1998, seemingly catastrophic, turned out to be a blessing in disguise for the Russian economy. The devaluation stimulated export and import substitution, implying that the IMF-sponsored stabilization programmes maintained the overvalued currency in the previous period. Devaluation was accompanied by improving terms of trade on the back of increase in world market prices of oil, gas and nonferrous metals and an increase in tax collections. Eventually, after a lost decade, Russian economy returned to sustained growth in 1999, averaging close to 7 per cent per annum in real terms between 1999 and 2005. Excessive budget deficit during the 1990s turned into budget surplus in every year for the period 2000–05. International reserves have increased dramatically. Although inflation was brought down to between 10 and 22 per cent per annum in the

124 Conclusions and Directions for Further Research

period, it proved extremely difficult to reduce to a single digit. While the nominal exchange rate with the US dollar is virtually unchanged at the end of 2005 from the beginning of 2000, real exchange rates have appreciated during the period in excess of 50 per cent. The strong currency allowed a considerable monetization of the economy so that money supply increased sevenfold in the period 1999–2005. However, the exchange rate appreciation eradicated all the gains from the devaluation of 1998 and the resulting import-substitution stimulus to the economy. Excessive supply of foreign currency and the increasing gold and foreign exchange reserves, when unsterilized, cause excessive growth of the money supply. That in turns put pressure on prices. The monetary authorities in this period had contradictory objectives: to reduce inflation and control exchange rate appreciation. It has been a difficult balancing act. Allowing increases in foreign exchange and gold reserves is inflationary, whereas buying excess foreign currency strengthens the currency, thus undermining competitiveness of domestic producers on the world market. The monetary authorities could use different methods to sterilize excessive supply of foreign currencies. They could do it either through a ministry of finance (ministry either keeps its tax receipt with a bank or buys foreign currency), through the banking system (increasing reserves) or issuing its own bonds. Nevertheless none of the methods is without a problem. The first option is not transparent and could undermine bargaining position of the government in foreign debt negotiations while the second and third options would crowd out private investment. In January 2004 the Stabilization Fund was created to sterilize the excessive supply of foreign currency. The idea was to tax and save oil revenues arising from oil prices above $20 per barrel. However, as the overall economic environment continued to be favourable on the back of rising oil prices, monetary authorities seem to prefer exchange rate stability to bringing inflation to a single digit. In the medium to long run, if the central bank is to adhere to its goal of doubling GDP in the early years of the next decade, one of the prerequisites is to reduce inflation. Empirical findings that emerge from the book unequivocally show that the short-term interrelation between money aggregates and inflation has been weakening in the new economic environment. Hence, monetary targeting could not be efficient even in the bank had adhered to its own prediction margins of the growth of money supply. Therefore the bank will need to switch from targeting monetary aggregates to inflation targeting. For the successful policy of inflation targeting it will have to enhance the role of interest rates in monetary policy as well as to minimize its presence on the domestic foreign exchange market and switch from a dirty to a free

Conclusions and Directions for Further Research 125

floating exchange rate regime. Furthermore, if the inflation targets of 7–8.5 per cent in 2006 and 5.5 per cent by 2008 are to be achieved the bank may need to set ceilings to natural monopoly and administrative price rises. All of these measures need to be accompanied by the further development of the financial market and appropriate government fiscal policies, trade policies, tariffs in particular, and structural reforms. Russia, as an open economy whose international trade makes as much as a half of its GDP, is very sensitive to external economic conditions. In the last several years Russia has enjoyed large and sustained terms of trade gains particularly because of a very high price of hydrocarbons. Since the hydrocarbons contribute about one-quarter to GDP and comprise slightly more than a half of exports, a change in their price affects growth to a considerable extent. Nevertheless, Russia has recently experienced a slight slowing of economic growth despite sustained if not rising world market prices of hydrocarbons. Conceivably Russia has exhausted excess capacity in the economy and currency appreciation undermined price competitiveness of domestic producers on world market, as mentioned above. Significantly, investment in the industry is below average in transition economies. Increasing investment climate in the economy in order to transform it from a commodity export economy to a refined products economy will be one of the biggest challenges of post-transition Russia. This is likely to be a slow process but indispensable because primary production economy based on extraction of natural resources may not be conducive to sustained economic growth. Moreover, such an economy is highly vulnerable to adverse external shocks. Monetary and fiscal policies have to be designed to take account of potential adverse external shocks. On the fiscal side, Russian policymakers have maintained conservative assumptions in respect of the price of hydrocarbons, and this resulted in establishing a large Stabilization Fund. This needs to be accompanied with a wider range of monetary sterilization measures. The central bank intends to let the exchange rate float and that should ensure more flexibility in case of adverse terms of trade shocks. Although the central bank seems to have favoured exchange rate stability to disinflation, the order of preferences may need to change in the future. That will also help ensure downward expectations of inflation. It is also essential to expand the market for the currency dominated debt as well as continuation of issuing the securities. The development of the bond market would further enhance the effects of interest rate policy on the money market. It is of paramount importance that the central bank maintains its independence to choose a set of monetary instrument as it sees fit in highly uncertain external and internal environments.

Appendix In order to test the null hypothesis H0:B(L)  0 the first step of Hsio’s (1981) procedure is to specify the own lag length of . This is best achieved by means of the following equation: t  c0  A(L)t  et

(A.1)

Determination of the own lag length for  could be achieved by varying a lag in the autoregression (Equation A.1) from 1 to x, where x is the highest order lag, c0 is a x a constant, A(L) is a distributed lag polynomial so that A(L)  a1AaL , L is the lag a operator so that L t  ta and et is a zero mean white noise error term. Thus the first step of the Hsiao procedure is to regress t on ta to obtain the sum of squared residuals (SSR) from the sample T, where a  1, 2,…, x. The obtained SSR is used to compute the final prediction error (FPE), as defined by Akaike (1970), for each autoregression described in Equation A.2:

兺

FPE(a) 

(T  a  1) (T  a  1)



SSR(a)

(A.2)

T

Thus if x in Equation A.1 were set at 15, as in our test, 15 FPEs would be obtained from Equation A.2.1 The order of a lag of a one-dimensional autoregressive process that yields the smallest FPE is chosen and denoted as FPE(a*). Having found the order (a) of A(L), the second step of Hsio procedure is to determine whether monetary variables enter the  equation. This amounts to testing the null hypothesis H0:B(L)  0. Thus the second step begins with an estimation of the bivariate: t  c0  A(L)t  B(L)mi,t  et

(A.3)

where mi,t are the monetary variables m2 and m2x considered one at the time, and B(L) is a distributed lag polynomial defined in a similar manner to A(L). In Equation A.3 t is therefore used as the controlled variable, with the order of lag(s) set at a*, and mi, t is treated as the manipulated variable. From Equation A.3 we again obtain the SSR needed to compute the FPEs of the controlled variable according to the formula given by Equation A.4. As in step one we vary the order of lags of mi,t over b  1, 2, …, x and determine the order that yields the smallest FPE, say b*. The corresponding two-dimensional FPE(a*, b) is FPE(a*,b) 

(T  a*  b  1) (T  a*  b  1)



SSR(a*, b)

(A.4)

T

In step three of Hsio’s procedure the smallest FPEs from steps one and two are compared and it is concluded that money Granger-causes inflation (mi ⇒  ) if 126

Appendix 127 / ) if the FPE(a*, b*)  FPE(a*), and money does not Granger-causes inflation (mi ⇒ reverse is true. In other words, if the FPE obtained from step two is smaller than the FPE obtained from step one then the optimal model for predicting inflation () will be the one that includes a* lagged  and b* lagged mi. If, however, the reverse is true a one-dimensional autoregressive representation for  should be used. By the same token, according to Hsiao the null hypothesis H0:B(L)  0 is rejected if FPE(a*, b*)  FPE(a*) and not rejected if the reverse is true. In step four of the Hsiao procedure, steps one to three are repeated for the mi processes. Firstly the FPEs resulting from the treatment of each variable of broad money supply as a one-dimensional autoregressive process are found. Secondly, having specified the order of autoregressive operator on mi, mi and  are respectively treated as a controlled and a manipulated variable. As above, this is in order to compute the optimum lag of the latter and the FPEs of the former. The purpose of this exercise is to test whether inflation Granger-causes money supply. Equations A.1 to A.4 are then transformed to reflect changes in the treatment of variables  and mi, resulting in Equations A.5 to A.8. Needless to say this is tantamount to testing the null hypothesis H0:C(L)  0.

mi,t  c0  D(L)mi,t  et FPE(d) 

(T  d  1) (T  d  1)



(A.5) SSR(d)

(A.6)

T

where d is the order of lag(s) of the one-dimensional autoregressive money supply process varying from 1 to x, and the d that yields the smallest FPE is chosen and denoted as FPE(d*). mi,t  c0  D(L)mi,t  C(L)t  et FPE(d*,c) 

(T  d*  c  1) (T  d*  c  1)



(A.7)

SSR(d*,c)

(A.8)

T

Where d* is the optimum number of lags for mi,t, (computed from Equation A.6) and c is the order of lag(s) varying from 1 to x, where the c that yields the smallest FPE is denoted as c*. The order of the one-dimensional autoregressive process of mt, obtained by using the FPE(d*) criterion in Equation A.6, is compared with the corresponding two-dimensional FPE(d*,c) obtained from Equation A.8. If the former is less than the latter a one-dimensional autoregressive representation for mi is used. This is equivalent to acceptance of the null hypothesis H0:C(L)  0. If the reverse is true we say that inflation Granger-causes money supply ( ⇒ m), and the optimal model for predicting mi is the one that includes d* lagged values of mi and c* lagged . Naturally this amounts to rejection of the same null hypothesis. Thus the combinations of FPEs obtained from Equations A.2, A.4, A.6 and A.8 in the Hsiao (1981) procedure lead to the identification of four theoretical cases of Granger-causality between money and inflation: 1. Unidirectional causality from m to  (m ⇒  ):FPE(a*, FPE(d*, c*) FPE(d*), or B(L)  0 and C(L)  0.

b*)

 FPE(a*) and

128 Appendix 2. Unidirectional causality from  to m ( ⇒ m):FPE(d*, c*)  FPE(d*) and FPE(a*, b*) > FPE(a*), or C(L)  0 and B(L)  0. 3. Feedback or bilateral causality, from m to  and  to m at the same time (m ⇔ ): FPE(a*, b*)  FPE(a*) and FPE(d*, c*)  FPE(d*), or B(L)  0 and C(L)  0. 4. Independence – that is, no causality from m to  or vice versa / ): FPE(a*, b*) FPE(a*) and FPE(d*, c*) FPE(d*), or B(L)  0 and C(L)  0. (m ⇔ From a practical point of view, or the standpoint of a policy maker, only variables that satisfy the conditions in numbers 1 and 2 above may be utilized as intermediate targets, via the instrument variables, for controlling the goal variable.2 In other words for a measure to be useful as an intermediate target, as well as being correlated with a goal variable and reasonably controllable by policy makers, it should also be exogenous or causally prior to the goal variable and have no significant feedback from the goal variable. Thus for a variable to be useful as a policy target, unidirectional causation from it to the goal variable is a necessary condition, although not necessarily a sufficient one. That is not the case with numbers 3 and 4 because the variables there are not exogenous. Finally, in step five of the Hsio procedure we combine all the equation specifications in order to identify the system. We pay particular attention at the direction of causality, as outlined in the four cases above. Thus if for example the goal variable is chosen to be , then the usefulness of mi as a policy target rests on the direction of causality with the goal variable . An analogous formula for the Bayesian estimation (BEC) criterion, in the univariate distributed lag case, as described by Equation A.1, can be expressed as follows:3 BEC(a) 

SSR(a) T



aSSR(x) / Tln T Tx1

(A.9)

where the variables are as described above and the minimum BECs correspond to the optimal lag lengths (a*). The formula could be modified in a straightforward manner to account for the bivariate distributed lag case for both aggregates of money supply.

Notes 1 The Emergence of Open Inflation and Stabilization Efforts in Postcommunist Russia 1. For the differences between these two programmes see Dabrowski (1997) and Hanson (1999), among others. 2. To be precise, the government did not produce an overall plan or even a consistent programme, but rather a policy package that relied on decrees, laws and mini programmes. 3. See Dornbusch et al. (1990), Fisher (1993), Burdekin et al. (1995) and Heyman and Leijonhufvud (1995), among others. 4. See Åslund (1995) for references. 5. See Nikolic (2002). 6. See Nikolic (2000a) for references. 7. For a detailed analysis see ibid. 8. More formally, the money multiplier can be defined as follows: M2/MB  [(C  D)/(C  R)]  [(C/D)  1]/[(C/D)  (R/D)]

9. 10. 11.

12. 13. 14.

15.

16.

17.

where C  currency in circulation, D  deposits and R  commerical banks’ reserves. For details see Granville (1995). CBR credits to the government were prohibited in 1995 prior to an introduction of the June stabilization programme. This is a flow definition of seigniorage that corresponds to the amount of goods and services the government can obtain by issuing additional money or forcing commercial banks to hold more reserves. Providing that the inflation rate and real money balances remain constant (steady state), seigniorage is equal to inflation tax. The share of base money in GDP in developing countries varies from country to country but averages about 15 per cent. In 1996 the stock of identifiable arrears to the consolidated budget were estimated at close to R128 trillion, or 5.7 per cent of GDP, while the arrears to the extrabudgetary funds were close to R100 trillion, or 4.3 per cent of GDP (RECEP, 1997). The State Tax Service reported that at the beginning of 1997 about one-third of enterprises were not paying any taxes, and only 16.6 per cent of enterprises were paying taxes that were due (RECEP, 1997). The currency risk premium is defined as the difference between the interest rate for the same maturity instrument of ruble-denominated domestic debt (taking into account the depreciation expected under the current exchange rate policy) and dollar-denominated government debt. In 1998 the official exchange rate rose from R5.96/$ to R20.65/$, representing a depreciation of 246 per cent. Between August 1998 and the end of March 1999 it depreciated by 287 per cent from R6.24/$ to R24.16/$ (RECEP, 1993–99). 129

130 Notes

2 Inflation Expectations in Postcommunist Russia 1. For a description of the Livingston data see Turnovsky (1970), Carlson (1977) and Figlewski and Wachtel (1981). For Gallup Poll surveys see the seminal work by Carlson and Parkin (1975). 2. For theoretical discussions see Friedman (1968), Phelps (1968) and Lucas and Rapping (1969). 3. Hybrid models of expectations can be further classified as the adaptiveregressive or habitat models (see for example Modigliani and Sutch, 1966; Modigliani and Shiller, 1973; Frenkel, 1976) and partially rational expectations models (for example Walters, 1971; Feige and Pearce, 1976; Brinner, 1977; Spitäller, 1978). 4. For a detailed review of the literature see Gramlich (1983) and Ball (1991). 5. A similar accommodation to varying degrees also prevailed in the three leading reformers in Central Europe (Poland, Hungary and the Czech Republic) at the beginning of transition (Rostowski and Nikolic, 1998). 6. Inflation tax (IT ) has been calculated using monthly data (Russian European Centre for Economic Policy, 1993–99) according to the formula IT  M2  (inflation rate – interest rate). Layard and Richter (1994) calculate that in 1992 (quarters two and four) and 1993 (quarters one to three) inflation tax was 29.9 per cent and 25 per cent respectively. 7. A pair-wise Granger causality test, including inflation, wages, exchange rate and money supply variables up to six lags, confirms this conjecture. In addition the correlation between inflation and contemporaneous and lagged nominal wages is quite strong (in excess of 50 per cent for up to six lags), indicating a backward indexation process in Russia. Very similar results are obtained for the correlation between inflation and contemporaneous and lagged exchange rates. 8. The velocity of circulation has been overestimated in the Russian economy because GDP has been underestimated, given the size of the grey economy. As the grey economy becomes more established and measurable, the measured economy will grow faster than the material economy, whose velocity will appear to be falling rather faster than is actually the case. However the pronounced decline in velocity leaves little doubt that the decline is due not only to the more precise measurement of GDP but also to an increase in the average time that economic agents hold money balances. 9. Buch (1999) has found that quantity theory holds for Russia in the long term. 10. For details of the factors that affected the demand for money see Anderson and Citrin (1995). 11. Nikolic (2000a, 2000b) has also shown that the inflation series, calculated as monthly growth rates of the Consumer Price Index in Russia for the same period, is a trend stationary process with at least one structural break and no seasonal components. Since the break point is assumed to be unknown a priori, the break point that gives the least favourable result for the null hypothesis of a unit root using the test statistic is chosen. In other words the break point is chosen to minimize the one-sided t statistic. These and all the other estimations in this analysis, are done in EViews (v. 3.0).

Notes 131 12. The other two models in this taxonomic classification are those which capture stochastic shocks (wars, crop failure, major natural disasters and so on) and structural developments, including monetary, fiscal, income and exchange rate policies. 13. The test statistic is given by SC  2L/n  k log n/n, where k is the number of estimated parameters, n is the number of observations and L is the value of the log likelihood function using the k estimated parameters. In order to select the most appropriate model we choose the values that minimize the Schwarz criterion. Analogous to other information criteria, the Schwarz criterion is based on minus two times the average log likelihood function, adjusted by a penalty function. 14. For the sake of brevity we shall use ARCH to refer to both ARCH and generalized ARCH (GARCH) models. For details of ARCH/GARCH modelling see Nikolic (2000a). 15. The hypothesis that, in addition to the lagged inflation rate, the expectation of future inflation also depends on its variability (defined as the absolute change in the rate of inflation) was also tested and the coefficient of variability of inflation was not found to be significant. 16. F-statistic  0.7649 (probability  0.6347), log likelihood ratio  24.7217 (probability  0.00173). 17. The conventional level of significance throughout this chapter is 5 per cent. 18. In contrast to the single Chow forecast test, this test does not require the specification of a forecast period — it automatically computes all feasible cases, starting with the smallest possible sample size for estimating the forecasting equation and then adding one observation at a time. 19. The graphs of the correlogram and partial correlogram are not presented but can be obtained from the author upon request. 20. If p has a real root whose absolute value exceeds one or a pair of complex reciprocal roots outside the unit circle (that is, with a modulus greater than one) it implies an explosive autoregressive process. Conversely if q has reciprocal roots outside the unit circle the MA process is non-invertible, which makes interpreting and using the MA results difficult. 21. LR  26.4416, probability  0.000026. 22. The dummy variable SBFEB94  1 after January 1994 and zero otherwise. 23. The dummy variable DO94 takes a value of 1 for October 1994 and zero otherwise. At the beginning of October 1994 the Central Bank of Russia lost control of the exchange rate and with reserves running low it was unable to prevent the ruble exchange rate from plummeting on Black Tuesday (see chapter 1). While there have been many financial and other crises in Russia during transition the magnitude of the Black Tuesday crisis, if not taken into account, renders a simple ADL model of the money – price relationship unstable (Nikolic, 2000a). A similar exogenous shock but of greater magnitude seems to have taken place in August 1998. It would be necessary to take this into account in a similar manner if the period after July 1998 were modelled. 24. Note that the bias, variance and covariance proportions add up to one. 25. An alternative test for the hypothesis of strong rationality can be performed by adding variables contained in It1 in addition to the lagged expectation

132 Notes

26.

27.

28. 29.

30.

31.

errors on the right-hand side of Equation 2.7 and testing for joint significance (Pesaran, 1987). The choice of variable(s) contained in It1 depends on the available information and the cost of acquiring and processing it. It follows that rational agents will set the marginal cost of acquiring and processing information to be equal to the benefits of acquiring it. We assume that one of the most relevant variables in inflation is broad money supply (m2x), as in Nikolic (2000a). If it turns out that this variable does not systematically influence expectations of inflation, then an alternative variable will be included in It1 and tested. Conversely if expectations turn out not to be rational there is no point in including other variables in the set. The Schwarz criterion of predictive accuracy was used to choose integer K in Equation 2.7. Consequently K  1 and K  3 have been set for the ARCH(1,0) and ARMA(6,1) models respectively. According to the Schwarz criterion k  1 for both models. That is, contemporaneous values of m2x are excluded from the regressions because it is assumed that agents do not have this information at their disposal. The same models are estimated, including constant term, but the results and conclusions are not significantly different. Similarly autoregressions of higher order are estimated – that is, 9 lags for inflation and 3 lags for m2x, as specified by the Schwarz criterion, but the results in respect of rationality are not qualitatively different. Since our models are reduced form models, however, we cannot say that inflation was persistent in a structural sense, or whether it was the result of economic policies implemented at the time and shocks hitting the Russian economy. See Hamilton (1989, 1990) and Ricketts and Rose (1995).

3 The Relationship between Money Supply and Inflation in Postcommunist Russia 1. I am indebted to the editor of the Journal of Comparative Economics, John Bonin, and two anonymous referees for their useful comments on an earlier version of this chapter, which was published in the Journal of Comparative Economics, vol. 28 (2000), pp. 108–23. 2. The dots over the variables denote the rate of growth. 3. This subject is beyond the scope of this chapter, but see Nikolic (2000b). 4. However this view is not shared unanimously by Russian economists. For example Koen and Marrese (1995) singled out Petrakov who maintains that inflation in Russia was fundamentally caused by the structural deformities of the economy. 5. For a comprehensive survey see Korhonen and Pesonen (1998). 6. In Western economies the lag between monetary policy changes and inflation is much longer. For example Carlson (1980) reports that the change in the money supply in the United States in 1995–69 was felt over a period of 20 quarters with a mean lag of about 11 quarters. Similarly Hoggarth (1996) estimates that in the UK the lag is two to three years and that monetary policy effects are broadly similar across the G7 countries. 7. As Hoggarth (1996) notes, the short-run Phillips curve may be flatter at low rates of inflation than at high rates.

Notes 133 8. The first observation for all the series except MB is February 1992. For the MB series the first observation is May 1992. All of the estimations were done in EViews (version. 3.0), except the Run test, which was done in Minitab (version 9.2). 9. With this combination the rates of growth are equivalent to differences of logarithms, since  log (xt)  log (xt)  log (xt  1)  log (xt/xt  1)⬵ (xt/xt  1)1  (xt  xt  1)/xt  1 as long as the ratio xt/xt  1 is reasonably small. 10. The inflation rate time series might have had a structural break at the beginning of 1994 when the inflation rate decreased markedly without a significant change in the trend (slope) over the period. Applying the modified unit root test (Perron, 1989), the presence of a structural break was confirmed and the unit root hypothesis rejected. Following Perron’s proposed unit root test we included a dummy variable to account for a potential structural break. Hence the following regression was estimated:   15.32  0.15trend  4.52Du  0.611  (4.59)

11. 12.

13. 14.

15. 16.

17.

(3.71)

(4.23)

9

兺 

i1

i

1

(4.88)

where Du is a dummy variable, so that Du  1 after January 1994 and 0 otherwise; the t-ratios are in brackets. The values of the test statistics for the Dummy variable (4.23) and the one period lagged inflation (4.88) imply a structural break and rejection of a unit root, respectively. See Andersen and Karnosky (1974). An alternative approach for future research aimed at establishing a causal relationship could apply cointegration analysis to this restrictive functional form. The arguments about the superiority of structural or monetarist models are beyond the scope of this chapter. Since the Koyck DL model, although neat, is obtained by a purely algebraic process, and is therefore devoid of any theoretical underpinning, it is often rationalized as an adaptive expectation model or a partial adjustment model, or a combination of both. For application of the Almon lag technique see Almon (1965); for problems with the application of this technique see Schmidt and Waud (1973). We recognize that in models such as the one depicted in Equation 3.2, multicollinearity is likely to appear. In other words, since successive values (lags) tend to be highly correlated, the standard errors tend to be large and therefore the estimation of coefficients is likely to be less than fully efficient. Nevertheless multicollinearity violates no regression assumptions. Unbiased, consistent estimates will occur. The only effect of multicollinearity is to make it hard to obtain coefficient estimates with a small standard error. Moreover multicollinearity may not pose a serious problem when R2 is high and all the regression coefficients are significant, as revealed by high t values. The test statistics are given by SC  2L/n k log n/n and AIC  2L/n 2k/n respectively, where k is the number of estimated parameters, n is the number of observations and L is the value of the log likelihood function using the k estimated parameters. In order to select the most appropriate model, we chose

134 Notes the values that minimized the SC and AIC. Among various statistical criteria, SC has superior large-sample properties and therefore is favoured when it suggests a different lag selection from the AIC. 18. A zero value of the skewness coefficient and a large value (over 3) of the kurtosis coefficient suggests the existence of an ARCH effect. Similarly a large value for the Jarque–Bera test would indicate rejection of the null hypothesis of normality and would also be consistent with the existence of ARCH effects. 19. Equation 3.4 can be further generalized into a general autoregressive conditional heteroscedasticity model, GARCH(p,q), in which the conditional variance of e at time t is dependent not only on past squared disturbances but also on past conditional variances: 2t  0  1e2t1  …  pe2tp  12t1  …  q2tq.

20.

21.

22.

23.

24.

By the same token we can introduce  2t (or alternatively the standard deviation t) on the right-hand side of Equation 3.3 in order to test whether inflation is dependent on the 2t or t of the forecast error as well as on changes in money supply. This is referred to as an ARCH-M (ARCH-in-mean) model. For a survey and an overview of the application of ARCH and GARCH models See Bollerslev et al. (1992). Iterative estimation of l indicates that the best specification for an ARCH/GARCH model for the monetary base is GARCH(2,1), while the Schwarz criterion favours GARCH(1,1). The tests for determining the exact order of the ARCH and GARCH models are not reported but are available from the author upon request. We have looked at ARCH(1), ARCH(2) and ARCH(3) models and, as it has become conventional, at GARCH(1,1), GARCH(1,2) and GARCH(2,1) models. A data set that requires a model of an order greater than those presented is very rare (see Bera and Higgins, 1993). However most of the explanation for the current rate of inflation in the model seems to be coming from the one month lagged inflation rate. A unit change in the previous month’s inflation rate seems to be determining as much as 80 per cent of the variations in the current inflation rate. By the same token the large value of the dummy DO94 may dwarf all other coefficients. Nonetheless the exclusion of this dummy does not change much. The fit of the model deteriorates only slightly and the values of the longer lags become negative, suggesting that the model would be more parsimonious if this dummy were excluded. Indeed the Wald test for redundant variables suggests that lags of longer than four should be excluded from the new specification. The usefulness of equations such as those estimated here for policy purposes hinges crucially on the temporal stability of the equation parameters. However stability of the model coefficients is a necessary but not a sufficient condition for the model to be a useful tool for policy makers. This issue is discussed in Nikolic (2000b). The tests were a recursive residuals test (RRT), a one-step forecast test (O-SFT), an N-step forecast test (N-SFT), a scaled recursive Chow test (SRCT), a Chow forecast test (CFT) and a recursive OLS coefficient test (ROLSCT). Since these

Notes 135

25.

26.

27.

28.

29. 30. 31.

32.

tests may be used with OLS and the two-stage OLS regressions, they were performed on the DL and ADL models presented in the Tables 3.3 and 3.6 respectively, after omitting the variable DO94 from the estimation for computational reasons. In order to conserve space only the O-SFT and the SRCT are presented in Table 3.9. In order to make the model more parsimonious the Wald test for redundant variables is applied. This justifies the exclusion of the variables of m2x with lags longer than four months since their joint coefficients are not different from zero. Only the coefficients obtained from the DL model are presented since it could be argued that the lagged inflation rate in the ADL model would impose a geometrically declining lag structure that would smooth out short-term fluctuations in the money supply and consequently affect the distribution of the coefficients of money. The dummy variable DO94, has been omitted from the estimation for computational reasons. The October 1994 observation is omitted since its very large value needs to be modelled by introducing a dummy, as explained earlier. Inclusion of this variable might, not only dwarf other values of the parameters but could also reduce the degrees of freedom in the relatively small subsamples. Arguably an alternative date for the dividing point could be February 1994, as indicated by the SRCT for the ADL model and the observation that the inflation rate had a structural break in this period. However October 1994 is preferred for a practical consideration. Namely an earlier date would reduce an already small sample and make the analysis of period I much less reliable. The models of the divided samples involving m0 and mb are not presented since these two aggregates have not proved to be good proxies for inflation. The test produces chi-squared statistics for periods I and II of 0.517 and 0.625 respectively. The corresponding statistic for the entire period is 0.439. The corresponding statistic for the entire period is 3.56 months. The magnitude of the calculated mean lags for periods I and II, obtained by OLS ‘unremedied’ estimates and their differences are not substantially different from those obtained by the ‘remedied’ estimates. A similar distributed lag model, except for the ARCH term, is used by Carlson (1980) for the US economy in 1955–79. The corresponding statistic for the entire period is 0.654 for a lag of four months. For this particular exercise the lag estimates of m2 prior to February 1992 are taken into consideration. The largest r for inflation and m2, if observations of m2 prior to February 1992 are excluded, applies to a lag of five months for both period I and the entire period, while the highest r for period II is for a lag of seven months, as reported.

4 Money–Inflation Causality in Transition Economies: The Case of Russia 1. Inflation has been chosen as a goal variable since the two main goals of economic policy, full employment and price level stability, are directly linked to nominal GNP. In many transition economies, taming inflation has often

136 Notes

2.

3.

4. 5. 6. 7. 8. 9.

10. 11.

been proclaimed as the chief objective of economic policy. This arguably unidimensional goal was advocated and to an extent dictated by the IMF. However gross national product or gross domestic product can be used as a goal variable, or indeed could be incorporated into the current analysis without difficulty. In the mid 1990s the International Monetary Fund argued that high inflation would be detrimental to the Russian economy in many ways. For example it would generate uncertainty about key prices, including real interest and exchange rates, which in turn would deter long-term credit, investment and growth. In addition, high inflation would damage the economy by encouraging unproductive activities aimed solely at hedging against inflation. Furthermore high inflation would hurt the most deprived social groups that lacked the political strength to protect their incomes against rising prices. Finally, high inflation would contribute to a general climate of uncertainty and lack of trust in government policies, which would encourage speculation against the ruble and capital flight (Hernández-Catá, 1995). Many researchers (for example, Nikolic, 2000a) claim that broad money aggregates (potential intermediate targets) have a significant explanatory power in determining inflation (the goal variable) in Russia in the years after price liberalization in February 1992. However Nikolic also shows that the influence of variations in the former on the latter became considerably weaker but more protracted when the financial system became more sophisticated and the macroeconomic environment became more stable. See Nikolic (2000a). The conventional level of significance throughout this chapter is 5 per cent. m2  18.61  0.27 trend  6.18D1  4.19D6 (15.4) (10.15) (2.79) (2.03) (t-statistics in parenthesis). Failure to take account of this exogenous shock caused the true partial coefficients simultaneously to equal zero (F-test). See also Note 23, chapter 2 See Nikolic (2000a). If the assumption of mutual non-correlation of error terms e1t and e2t holds we can estimate Equation (4.2) using the ordinary least square (OLS) method. Should the error terms and prove not to be white noise, this has to be rectified and the estimation becomes much more complicated and computationally demanding. For some of these criteria see Thornton and Batten (1985) and Geweke and Meese (1981). Unlike Batten and Thornton (1983) and Jones (1989), we refrain from using the Pagano and Hartley (PH) (1981) t-test for lag specification, despite its computational efficiency. The PH procedure produced similar results to the FPE test in both of the studies mentioned above, albeit with less parsimonious lag specifications. Since a potential weaknesses of the FPE criterion is that it selects lag lengths that are too long in large samples (it is asymptotically inefficient) we saw little point in testing a criterion that was likely to select even longer lag lengths. Besides, even in Jones’s (1989) study the PH did not outperform the FPE, and it performed considerably worse than the FPE in Batten and Thornton’s (1983) study.

Notes 137 12. The rationale for including the DO94 and Di dummies is explained above; the linear time (trend) was included in all equations to adjust for possible non-stationary elements. 13. See Nikolic (2000a) for details of the use of (G)ARCH models. 14. In the rare cases when the Q-statistics for the residuals were significant despite the implementation of various ARCH/GARCH specifications, we saw no purpose in presenting the results of causality or the significance of the F-statistics as they would be unreliable. 15. The entire set of model selection specifications is not presented but can be obtained from the author upon request. 16. Despite the perceived superiority of the statistical search procedure for lag-length determination over the ad hoc ones, the former is not immune from criticism because the researcher chooses the maximum lag length arbitrarily. 17. While an extensive search of the lag space would be a reasonably efficient procedure for a bivariate case in which the OLS estimates gave satisfactory results, it would be burdensome and computationally less efficient in the ARCH/GARCH estimations. For example a search of the lag space with 15 lags would require about 1800 regressions in our case. Hence extending the lag space to 21 lags or more, as might be more appropriate for the regression of m2x on , would be very demanding computationally. 18. Our preference for significance level over F-statistics is guided by the fact that the latter are not invariant with regard to degrees of freedom. 19. The same approach was adopted for two other members of the CIS-3: Ukraine and Kazakhstan. 20. On 22 February 1996 the CBR and the government set out a mediumterm strategy for economic policy, agreed with the IMF, that provided Russia with an extended fund facility credit of $10.2 billion. This credit would operate for a longer period (three years) than the stand-by facility granted the previous year. In order to ensure appropriate implementation of the programme by the Russian authorities, the IMF would monitor movement towards the targets specified in the programme on a monthly, rather than a quarterly basis. These targets included ceilings on the net domestic assets, limits on the monetary authority’s net credit to the federal and general governments, and floors on the net and gross international reserves. The latter were to reach $19.7 billion by the end of 1996 (RECEP, 1996). By the end of first quarter of 1996 all the targets in the monetary programme had been met. 21. According to Bofinger et al. (1997), while the Bundesbank uses monetary targeting as a medium-term policy guideline the IMF applies this concept for short-term stabilization purposes. They argued that this practice is inconsistent with the monetarist paradigm, which opposes all types of fine tuning because of the long and variable lags in monetary policy. 22. The lack of political consensus on to whether make large fiscal adjustments was evidenced by the government’s failure to introduce budgetary measures to restrict credit to enterprises and certain sectors, as stipulated by the IMF. 23. The same conclusion applies to Ukraine and Kazakhstan.

138 Notes

5 The Role of International Financial Institutions in Stabilizing the Russian Economy 1. I am indebted to Dr Tomasz Mickiewicz for the useful comments on earlier versions of this chapter. Responsibility for any remaining errors rests with the author. 2. For discussions of this programme see Bofinger et al. (1997) and Nikolic (2000a). 3. Throughout this chapter capital inflow refers to the acquisition of domestic assets by non-residents (plus grants). The sale of domestic assets is defined as a negative capital inflow. Thus the term net capital inflow denotes acquisitions minus sales of domestic assets by non-residents. Conversely capital outflow refers to the acquisition of foreign assets by residents. Sales of foreign assets are defined as a negative capital outflow. Thus the term net capital outflow denotes acquisitions minus sales of foreign assets by residents. 4. ‘Errors and omissions’ are unrecorded capital flows, mainly capital flight. 5. The Czech Republic, Hungary, Poland, Slovakia and Slovenia. 6. Between 1990 and 1998 FDI accounted for 34 per cent of capital flows to developing countries, while between 1993 and 1998 it accounted for 43 per cent of capital flows to the CETE-5 and 26 per cent to the Russian Federation. 7. The sovereign debt alone made Russia one of the most indebted emerging markets in the world. Only Brazil ($220 billion), Korea ($170 billion) and Indonesia ($169 billion) ran bigger debts (Hishow, 2001). 8. See Mussa and Savastano (1999) for references. 9. The interested reader is referred to the works of Åslund (1995), Gros and Steinherr (1995), Soros (2000) and Stiglitz (2002). 10. Apparently the same IMF team had much more success in Poland than in Russia (Gomulka, 1995).

Appendix 1. If the appropriate lag length turns out to be 15 for any variable, prudence requires that the maximum lag length is allowed to extend beyond 15 by at least three lags to check whether longer lags are more appropriate. This is extended further if the longest lag under consideration is again chosen as the optimal one. 2. Depending on the choice of the goal variable, only one of the two can be utilized; that is, if the inflation rate () is a goal variable then only mi in number 1 (unidirectional causality from mi to ) can be utilized as a policy target. 3. For further details see Jones (1989) and Geweke and Meese (1981).

Bibliography Akaike, Hirotugu (1970) ‘Statistical Predictor Identification’, Annals of the Institute of Statistical Mathematics, vol. 22, no. 1, pp. 163–80. Allen, Bill, Glenn Hoggarth and Lionell Price (1996) ‘Monetary Policy and Liberalisation in Poland, Russia and United Kingdom’, in Franco Bruni, Donald E. Fair and Richard O’Brian (eds), Risk Management in Volatile Financial Markets (Dordrecht, Boston and London: Kluwer). Almon, Shirley (1965) ‘The Distributed Lag Between Capital Appropriations and Expenditures’, Econometrica, vol. 33, no. 1, pp. 178–96. Andersen, Leonell C. and Dennis S. Karnosky (1974) ‘A Monetary Interpretation of Inflation’, paper presented to the Conference on Price Behavior, National Bureau of Economic Research in Income and Wealth, 21 November. Anderson, Jonathan and Daniel A. Citrin (1995) ‘The Behaviour of Inflation and Velocity’, in Daniel A. Citrin and Ashok K. Lahiri (eds), Policy Experiences and Issues in the Baltics, Russia, and Other Countries of the Former Soviet Union, IMF Occasional Paper 133 (Washington, DC: International Monetary Fund). Åslund, Anders (1995) How Russia Became a Market Economy (Washington, DC: Brookings Institution). Ball, Laurence (1991) ‘The Genesis of Inflation and the Costs of Disinflation’, Journal of Money, Credit, and Banking, vol. 23, no. 3, pp. 439–52. Barbone, Luka and Dario Marchetti (1995) ‘Transition and the Fiscal Crisis’, Economics of Transition, vol. 3, no. 1, pp. 59–74. Batten, Dallas S. and Daniel L. Thornton (1983) Polynomial Distributed Lags and the Estimation of the St. Louis Equation, Research Paper 63 (St Louis, Mo.: Federal Reserve Bank of St Louis). Begg, David (1997) ‘Monetary Policy During Transition: Progress and Pitfalls in Central and Eastern Europe, 1990–6’, Oxford Review of Economic Policy, vol. 13, no. 2, pp. 33–46. Begg, David and Charles Wyplosz (1999) Monetary and Exchange Rate Policies, EMU and Central and Eastern Europe (London: Centre for Economic Policy Research). Bera, Anil and Matthew L. Higgins (1993) ‘ARCH Models: Properties, Estimation and Testing’, Journal of Economic Survey, vol. 7, no. 4, pp. 305–66. Bofinger, Peter, Heiner Flassbeck and Lutz Hoffmann (1997) ‘Orthodox Money-Based Stabilisation (OMBS) Versus Heterodox Exchange Rate-Based Stabilisation (HERBS): The Case of Russia, The Ukraine and Kazakhstan’, Economic Systems, vol. 21, no. 1, pp. 1–33. Bollerslev, Tim (1986) ‘A Generalized Autoregressive Conditional Heteroskedasticity’, Journal of Econometrics, vol. 31, no. 3, pp. 307–27. Bollerslev, Tim, Ray Chou and Kenneth Kroner (1992) ‘ARCH Modeling in Finance: A Review of the Theory and Empirical Evidence’, Journal of Econometrics, vol. 52, nos 1–2, pp. 5–59. Bollerslev, Tim and Jeffrey M. Wooldridge (1992) ‘Quasi-Maximum Likelihood Estimation and Inference in Dynamic Models with Time Varying Covariances’, Econometric Reviews, vol. 11, no. 2, pp. 143–72. 139

140 Bibliography Brinner, Roger E. (1977) ‘The Death of the Phillips Curve Reconsidered’, Quarterly Journal of Economics, vol. 91, no. 3, pp. 389–418. Bryant, John and Neil Wallace (1980) ‘A Suggestion for Further Simplifying the Theory of Money,’ Staff Report no. 62 (Minneapolis, Minn.: Federal Reserve Bank of Minneapolis). Buch, Claudia M. (1998) ‘Russian Monetary Policy – Assessing the Track Record,’ Economic Systems, vol. 22, no. 2, pp. 105–45. Budina, Nina and Sweder van Wijnbergen (1997) ‘Fiscal Policies in Eastern Europe’, Oxford Review of Economic Policy, vol. 13, no 2. pp, 47–64. Buiter, Willem H. (1997) Aspects of Fiscal Performance on Some Transition Economies under Fund-Supported Programs, IMF Working Paper WP/97/31 (Washington, DC: International Monetary Fund). Buiter, Willem H. (1999) The Fallacy of the Fiscal Theory of the Price Level, NBER Working Paper w7302 (Cambridge, Mass.: NBER). Buiter, Willem H. (2002) ‘The Fiscal Theory of the Price Level: A Critique’, Economic Journal, vol. 112, no. 481, July, pp. 459–80. Burdekin, Richard, Salaman Suyono and Thomas D. Willett (1995) ‘The High Costs of Monetary Instability’, in Thomas D. Willett, Richard Burdekin, Richard J. Sweeney and Clas Wihlborg (eds), Establishing Monetary Stability in Emerging Market Economies (Boulder, CO: Westview). Burnside, Craig, Martin Eichenbaum and Sergio Rebelo (1998) Prospective Definitions and Asian Currency Crises, NBER Working Paper w6758 (Cambridge, Mass.: NBER). Cagan, Philip (1956) ‘The Monetary Dynamics of Hyperinflation’, in Milton Friedman (ed.), Studies in the Quantity Theory of Money (Chicago, Ill.: University of Chicago Press). Calvo, Guillermo A. and Carlos A. Végh (1992) Inflation Stabilisation and Nominal Anchors, IMF Discussion Paper 92/4 (Washington, DC: International Monetary Fund). Canzoneri, Matthew B., Robert E. Cumby and Behzad T. Diba (1998) Is the Price Level Determined by the Needs of Fiscal Solvency?, NBER Working Paper w6471. (Cambridge, Mass.: NBER). Carlson, John A. (1977) ‘A study of Price Forecast’, Annals of Economic and Social Measurement, vol. 6, no. 1, pp. 27–56. Carlson, John A. and Michael Parkin (1975) ‘Inflation Expectations’, Econometrica, vol. 42, no. 166, pp. 123–38. Carlson, Keith M. (1980) ‘The Lag from Money to Prices’, Review, Federal Reserve Bank of St. Louis, vol. 62, no. 10, pp. 3–9. Cheasty, Adrienne and Jeffrey M. Davis (1996) ‘Fiscal Transition in Countries of the Former Soviet Union: An Interim Assessment’, MOCT-MOST, vol. 6, no. 3, pp. 7–34. Cheng, Benjamin (1996) ‘On the Relationship Between Money and Inflation in the United States: Additional Evidence’, Applied Economic Letters, vol. 3, no. 8, pp. 549–52. Cochrane, John, H. (1998) A Frictionless View of US Inflation, NBER Working Paper w6646 (Cambridge, Mass.: NBER). Cochrane, John H. (2000) Money as Stock: Price Level Determination with no Money Demand, NBER Working Paper w7498 (Cambridge, Mass.: NBER). Coorey, Sharmini, Mauro Mecagni and Erik Offerdal (1998) ‘Disinflation in Transition Economies: The Role of Relative Price Adjustment’, in Carlo

Bibliography 141 Cottarelli and Gyorgy Szapary (eds), Moderate Inflation: The Experience of Transition Economies (Washington, DC: IMF and National Bank of Hungary). Coricelli, Fabrizio, Marek Dabrowski and Urszula Kosterna (1998) Fiscal Policy in Transition, Forum Report of the Economic Policy Initiative no. 3 (London: Centre for Economic Policy Research). Cottarelli, Carlo and Peter Boyle (1999) Disinflation in Transition: 1993–97, IMF Occasional Paper no. 179 (Washington, DC: International Monetary Fund). Cottarelli, Carlo and Giannini Curzio (1998) Inflation, Credibility, and the Role of the International Monetary Fund, Paper on Policy Analysis and Assessment 98/12 (Washington, DC: International Monetary Fund). Cottarelli, Carlo, Mark Griffiths and Reza Moghadam (1998) The Nonmonetary Determinants of Inflation – A Panel Data Study, IMF Working Paper WP/98/23. (Washington, DC: International Monetary Fund). Dabrovski, Marek (1997) ‘The First Half of Russian Transformation’, in Anders Åslund (ed.), Russia’s Economic Transformation in the 1990s (London: Pinter). Dabrowski, Marek (1999) Disinflation, Monetary Policy and Fiscal Constraints. Experience of the Economies in Transition, Centre for Soviet and Economic Resarch Foundation (CASE) Report, no. 16 (Warsaw). Desai, Padma (2000) ‘Why Did the Rouble Collapse in August 1998’, American Economic Review, Papers and Proceedings, vol. 90, no. 2, pp. 48–52. Dornbusch, Rudiger, Federico Sturzenegger and Holger Wolf (1990) ‘Extreme Inflation: Dynamic and Stabilization’, Brooking Papers on Economic Activity, vol. 2, pp. 1–4 (Washington, DC: Brookings Institution). Easterly, William and Paulo Vieira da Cunha (1994) ‘Financing the Storm: Macroeconomic Crisis in Russia’, Economics of Transition, vol. 2, no. 4, pp. 443–650. EBRD (1998) Transition Report 1998 (London: European Bank for Restructuring and Development). EBRD (1999) Transition Report 1999 (London: European Bank for Restructuring and Development). EBRD (2000) Transition Report 2000. (London: European Bank for Restructuring and Development). Economic Commission for Europe (1995) Economic Survey of Europe in 1994–1995. (New York and Geneva: Economic Commission for Europe). Engle, Robert (1982) ‘Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation’, Econometrica, vol. 50, no. 2, pp. 987–1007. Fakin, Barbara and Alain de Crombrugghe (1997) Fiscal Adjustments in Transition Economies – Social Transfers and the Efficiency of Public Spending: A Comparison with OECD Countries, Policy Research Working Papers 27/3, no. 1803 (Washington, DC: World Bank). Feige, Edgar L. and Douglas K. Pearce (1976) ‘Economically Rational Expectations: Are Innovations in the Rate of Inflation Independent of Innovations in Measures of Monetary and Fiscal Policy?’, The Journal of Political Economy, vol. 84, no. 3, pp. 499–522. Feldstein, Martin (1998) ‘Refocusing the IMF’, Foreign Affairs, vol. 77, no. 2, pp. 21–33. Figlewski, Stephen and Paul Wachtel (1981) ‘The Formation of Inflationary Expectations’, Review of Economics and Statistics, vol. 63, no. 1, pp. 1–10. Fisher, Irving (1930) The Theory of Interest (New York: Macmillan).

142 Bibliography Fisher, Irving (1963) The Purchasing Power of Money, rev. edn (New York: Augustus M. Kelley). Fisher, Stanley (1986) ‘Exchange Rate versus Monetary Targets in Disinflation’, in Stanley Fisher (ed.), Indexation, Inflation and Economic Policy (Cambridge, Mass.: MIT Press). Fisher, Stanley (1993) ‘The Role of Macroeconomic Factors in Growth’, Journal of Monetary Economics, vol. 32, no. 3, pp. 485–512. Fisher, Stanley (1994) ‘Prospects for the Russian Stabilization in the Summer of 1993’, in Anders Åslund (ed.), Economic Transformation in Russia (London: Pinter). Fisher, Stanley, Ratna Sahay and Carlos A. Vegh (1996) ‘Stabilisation and Growth in Transition Economies: The Early Experience’, Journal of Economic Perspectives, vol. 10, no. 2, pp. 45–66. Frankel, Jeffrey A. and Cara S. Lown (1994) ‘An Indicator of Future Inflation Extracted from the Steepness of the Interest Rate Yield Curve Along its Entire Length’, Quarterly Journal of Economics, vol. 109, no. 2, pp. 517–30. Frenkel, Jacob A. (1973) Inflation and the Formation of Expectations, University of Chicago Report no. 7353 (Chicago, Ill.: University of Chicago). Frenkel, Jacob A. (1976) ‘Inflation Expectations and Some Dynamic Aspects of the Welfare Costs’, in Michael Parkin and George Zis (eds), Inflation in Open Economy, no. 4 (Manchester: Manchester University Press). Friedman, Milton (1968) ‘The Role of Monetary Policy’, American Economic Review, vol. 58, no. 1, pp. 1–17. Friedman, Milton (1989) ‘Quantity Theory of Money’, in John Eatwell, Murray Milgate and Peter Newman (eds), The New Palgrave: Money (London and Basingstoke: MacMillan). Gavrilenkov, Evgeny and Kuboniwa Masaaki (1997) Development of Capitalism in Russia: The Second Challenge (Tokyo: Maruzen). Geweke, John (1978) ‘Testing the Exogeneity Specification in the Complete Dynamic Simultaneous Equations Model’, Journal of Econometrics, vol. 7, no. 2, pp. 163–85. Geweke, John and Richard Meese (1981) ‘Estimating Regression Models of Finite but Unknown Order’, International Economic Review, vol. 22, no. 1, pp. 55–70. Gomulka, Stanislav (1995) ‘The IMF Supported Programs in Poland and Russia, 1990–1994: Principles, Errors, and Results’, Journal of Comparative Economics, vol. 20, no. 3, pp. 316–46. Gramlich, Edward M. (1983) ‘Models of Inflation Expectations Formation: A Comparison of Household and Economist Forecasts’, Journal of Money, Credit and Banking, vol. 15, no. 2, pp. 155–73. Granger, Clive W. J. (1969) ‘Investigating Causal Relationship by Econometric Models and Cross-Spectral Methods’, Econometrica, vol. 37, no. 3, pp. 424–38. Granville, Brigitte (1995) The Success of Russian Economic Reforms (London: Royal Institute of International Affairs). Granville, Brigitte (1997) ‘Farewell, Rouble Zone’, in Anders Åslund (ed.), Russia’s Economic Transformation in the 1990s (London and Washington: Pinter). Gros, Daniel and Alfred Steinherr (1995) Winds of Change (London: Longman). Hale, David D. (1998) ‘The IMF, Now More than Ever’, Foreign Affairs, vol. 77, no. 6, pp. 7–14.

Bibliography 143 Hamilton, James D. (1989) ‘A New Approach to the Economic Analysis of Nonstationary Time Series and the Business Cycle’, Econometrica, vol. 57, no. 2, pp. 357–84. Hamilton, James D. (1990) ‘Analysis of Time Series Subject to Changes in Regime’, Journal of Econometrics, vol. 45, nos 1–2, pp. 39–70. Hanson, Philip (1999) ‘The Russian Economic Crises and the Future of Russian Economic Reform’, Europe–Asia Studies, vol. 51, no. 7, pp. 1141–66. Hernández-Catá, Ernesto (1995) ‘Russia and the IMF: The Political Economy of Macrostabilization’, in Daniel A. Citrin and Ashok K. Lahiri (eds), Policy Experiences and Issues in the Baltics, Russia, and Other Countries of the Former Soviet Union, IMF Occasional Paper 133 (Washington, DC.: International Monetary Fund). Heyman, Daniel and Alex Leijonhufvud (1995) High Inflation (Oxford: Claredonian Press). Hishow, Ognian N. (2001) ‘Russia’s External Debt: Infinite Rescheduling?’, Communist and Post-Communist Studies, vol. 34, no. 1, pp. 114–27. Hoggarth, Glenn (1996) ‘Monetary Policy in Russia’, in Jouko Rautava (ed.), Russia’s Financial Markets and the Banking Sector in Transition, Bank of Finland Studies A:95 (Helsinki: Bank of Finland). Holden, Kenneth and David A. Peel (1977) ‘An Empirical Investigation of Inflationary Expectations’, Oxford Bulletin of Economic and Statistics, vol. 39, no. 2, pp. 291–9. Hsiao, Cheng (1981) ‘Autoregressive Modelling and Money-Income Causality Detection’, Journal of Monetary Economics, vol. 7, no. 1, pp. 85–106. International Bank for Reconstruction and Development and World Bank (1999) Global Development Finance (Washington, DC: IBRD and World Bank). International Monetary Fund (IMF) (1999) Russian Federation: Recent Economic Developments, International Monetary Fund Staff Country Report no. 99/100, (Washington, DC: IMF). International Monetary Fund (IMF) (2000) International Financial Statistics (Washington, DC: IMF). Jones, Jonathan D. (1989) ‘A Comparison of Lag-Length Selection Techniques in Tests of Granger Causality Between Money Growth and Inflation: Evidence for the US, 1959–86’, Applied Economics, vol. 21, no. 6, pp. 809–22. Keynes, John Maynard (1936) General Theory of Employment, Interest and Money. (London: Macmillan). Kharas, Homi, Brian Pinto and Sergei Ulatov (2001) An Analysis of Russia’s 1998 Meltdown: Fundamental and Market Signals, Brookings Papers on Economic Activity (Washington, DC: The Brookings Institution). Koen, Vincent and Michael Marrese (1995) Stabilization and Structural Change in Russia, 1992–1994, IMF Working Paper WP/95/13 (Washington, DC: International Monetary Fund). Kolodko, Grzegorz W. (1999) ‘Transition to a Market Economy and Sustained Growth. Implications for the Post-Washington Consensus’, Communist and Post-Communist Studies, vol. 32, no. 3, pp. 233–61. Komulainen, Tuomas and Jukka Pirtilä (2000) Fiscal Explanations for Inflation: Any Evidence from Transition Economies?, Bank of Finland Institute for Economies in Transition Discussion Paper no. 11 (Helsinki: Bank of Finland).

144 Bibliography Korhonen, Iikka (1996) ‘An Error Correction Model for Russian Inflation’, Review of Economies in Transition, vol. 4/96, pp. 53–61 (Helsinki: Bank of Finland, July). Korhonen, Iikka (1998) ‘An Error Correction Model for Prices, Money, Output, and Interest Rate in Russia’, Review of Economies in Transition, vol. 5/98, pp. 33–44 (Helsinki: Bank of Finland, November). Korhonen, Iikka and Hanna Pesonen (1998) ‘The Short and Variable Lags of Russian Monetary Policy’, Review of Economies in Transition, vol. 4/98, pp. 35–45, (Helsinki: Bank of Finland, April). Kwiatkowski, Denis, Peter Philips, Peter Schmidt and Yongcheol Shin (1992) ‘Testing the Null Hypothesis of Stationarity against the Alternative of a Unit Root’, Journal of Econometrics, vol. 54, nos 1–3, pp. 159–78. LaRouche, Lyndon H. Jr (1999) ‘Lyndon LaRouche, Economist: Scrap the Foolish IMF Package!’, Kommersant -Vlast, 26 January. Layard, Richard and Andrea Richter (1994) Who Gains and Who Loses from Russian Credit Expansion? London School of Economics and Political Science, Centre for Economic Performance, Discussion Paper no. 200 (London: LSE). Locke, John (1823) The Works of John Locke, vol. 5 (London: Thomas Tegg). Lopez-Claros, Augusto and Sergei V. Alexashenko (1998) Fiscal Policy Issues During the Transition in Russia, IMF Occasional Paper no. 155 (Washington, DC: International Monetary Fund). Lucas, Robert (1981) Studies in Business Cycle Theory (Cambridge, Mass.: MIT Press). Lucas, Robert E. Jr and Leonard A. Rapping (1969) ‘Real Wages, Employment and Inflation’, Journal of Political Economy, vol. 77, no. 5, pp. 721–54. Maddala, G. S. (1992) Introduction to Econometrics, 2nd edn (New York: Macmillan). McMillin, Douglas W. and James S., Fackler (1984) ‘Monetary vs. Credit Aggregates: An Evaluation of Monetary Policy Targets’, Southern Economic Journal, vol. 50, no. 3, pp. 711–23. Menil, George de and Surjit S., Bhalla (1975) ‘Direct Measurement of Popular Price Expectations’, The American Economic Review, vol. 65, no. 1, pp. 169–80. Mishkin, Frederic S. (1990) ‘What Does the Term Structure Tell us about Future Inflation.’ Journal of Monetary Economics, vol. 25, no. 1, pp. 77–95. Modigliani, Franco and Robert J. Shiller (1973) ‘Inflation, Rational Expectations and the Term Structure of Interest Rates’, Economica, vol. 40, no. 161, pp. 12–43. Modigliani, Franco and Richard Sutch (1966) ‘Innovations in Interest Rate Policy’, The American Economic Review, Papers and Proceedings, vol. 56, nos. 1/2, pp. 178–97. Mullineaux, Donald J. (1978) ‘On Testing for Rationality: Another Look at the Livingston Price Expectation Data’, Journal of Political Economy, vol. 86, no. 2, pp. 329–36. Mullineaux, Donald J. (1980), “Inflation Expectations and Money Growth in the United States”, The American Economic Review, vol. 70, no. 1, pp. 149–61. Mussa, Michael and Miguel A. Savastano (1999), The IMF Approach to Economic Stabilization, International Monetary Fund Working Paper WP/99/104 (Washington, DC: International Monetary Fund). Muth, John F. (1961), ‘Rational Expectations and Theory of Price Movements’ Econometrica, vol. 29, no. 3, pp. 315–35.

Bibliography 145 Newey, Whitney K. and Kenneth D. West (1987), ‘A Simple Positive SemiDefinite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix’, Econometrica, vol. 55, no. 3, pp. 703–8. Nikolic, Milan (2000a), ‘Money Growth–Inflation Relationship in Postcommunist Russia’, Journal of Comparative Economics, vol. 28, no. 1, pp. 108–33. Nikolic, Milan (2000b), ‘Money–Inflation Causality in Transition Economies: A Case of Russia’, mimeo, University College London, SSEES, June. Nikolic, Milan (2001) ‘Inflation Expectation in Postcommunist Russia’, mimeo University College London, SSEES, January. Nikolic, Milan (2002), Russia and the IMF: Pseudo Lending for Pseudo Reforms, Centre for the Study of Economic and Social Change in Europe, University College London, Working Papers, vol. 6, pp. 1–20. Pagano, Marcello and Michael Hartley (1981), ‘On Fitting Distributed Lag Models Subject to Polynomial Restrictions’, Journal of Econometrics, vol. 16, no. 2, pp. 171–98. Perron, Pierre (1989), ‘The Great Crush, the Oil Shock and the Unit Root Hypothesis’, Econometrica, vol. 57, no. 6, pp. 1361–401. Pesando, James E. (1975), ‘A Note on the Rationality of the Livingston Price Expectations’, Journal of Political Economy, vol. 83, no. 3, pp. 849–58. Pesaran, Hashem M. (1987), The Limits of Rational Expectations (Oxford: Basil Blackwell). Phelps, Edmund S. (1968), ‘Money Wage Dynamic and Labour Market Equilibrium’, Journal of Political Economy, vol. 76, no. 4, pp. 678–711. Pigou, Arthur C. (1927), Industrial Fluctuations, 2nd edn (London: Macmillan). Polak, J. J. (1957) Monetary Analysis of Income Formation and Payments Problems, International Monetary Fund Staff Papers vol. 5 (Washington, DC: International Monetary Fund). Poole, William (1976), Rational Expectations in the Macro Model, Brooking Papers on Economic Activity, vol. 2, pp. 463–505. Portes, Richard (1994), ‘Transformation Traps’, Economic Journal, vol. 104, no. 426, pp. 1178–89. Ramsey, James B. (1969), ‘Tests for Specification Errors in Classical Linear Least Squares Regression Analysis’, Journal of the Royal Statistical Society, series B, vol. 31, no. 2, pp. 350–71. Ricketts, Nicholas and David Rose (1995), Inflation, Learning and Monetary Policy Regimes in the G-7 Economies, Bank of Canada Working Paper no. 95/6 (Bank of Canada). Robinson, Neil (2001), ‘The Myth of Equilibrium: Winner Power, Fiscal Crises and Russian Economic Reform’, Communist and Post-Communist Studies, vol. 34, no. 4, pp. 423–46. Rostowski, Jacek and Milan Nikolic (1998), ‘Macroeconomic Shocks and Policy Responses During Transition: A Cross-Country Comparison’, in Leszek Balcerowicz, Cheryl W. Gray, and Iraj Hoshi (eds), Enterprise Exit Processes in Transition Economies (Budapest: CEU Press). Russian European Centre for Economic Policy (RECEP) (1993–99) Russian Economic Trends (monthly updates) (Moscow: RECEP). Russian European Centre for Economic Policy (RECEP) (1994) Russian Economic Trends, vol. 3, no. 3 (Moscow: RECEP).

146 Bibliography Russian European Centre for Economic Policy (RECEP) (1995) Russian Economic Trends, vol. 4, no. 3 (Moscow: RECEP). Russian European Centre for Economic Policy (RECEP) (1996) Russian Economic Trends, vol. 5, no. 1 (Moscow: RECEP). Russian European Centre for Economic Policy (RECEP) (1997) Russian Economic Trends, vol. 6, no. 4 (Moscow: RECEP). Russian European Centre for Economic Policy (RECEP) (1998) Russian Economic Trends, vol. 7, no. 2 (Moscow: RECEP). Sachs, Jeffrey D. (1997) ‘Why Russia Has Failed to Stabilize’, in Anders Åslund (ed.), Russia’s Economic Transformation in the 1990s (London and Washington: Pinter). Sanders, Bernard (1998) ‘The IMF’s Disastrous Plan for Russia’, Christian Science Monitor, 25 June. Sargent, Thomas J. (1993) Rational Expectations and Inflation, 2nd edn (New York: HarperCollins). Sargent, Thomas J. and Neil Wallace (1976), ‘Rational Expectations and the Theory of Economic Policy’, Journal of Monetary Economics, vol. 2, no. 2, pp. 169–83. Sargent, Thomas J. and Neil Wallace (1986), ‘Some Unpleasant Monetarist Arithmetics’, in Thomas Sargent (ed.), Rational Expectations and Inflation (New York: Harper & Row). Schmidt, Peter and Roger N. Waud (1973) ‘The Almon Lag Technique and the Monetary Versus Fiscal Policy Debate’, Journal of American Statistical Association, vol. 68, no. 341, pp. 11–13. Shiller, Robert (1978) ‘Rational Expectations and the Dynamics of Structure of Macroeconomic Models: A Critical Review’, Journal of Monetary Economics, vol. 4, no. 1, pp. 1–44. Sims, Christopher A. (1994) ‘A Simple Model for Study of the Determination of the Price Level and the Interaction of Monetary and Fiscal Policy’. Economic Theory, vol. 4, no. 3, pp. 381–99. Sinel’nikov-Murilev, Germanovich S. and Yu G. Trofimov (1998) ‘Fiscal Crisis and Macroeconomic Policy in Russia’, Communist Economies & Economic Transformation, vol. 10, no. 2, pp. 189–202. Söderlind, Paul (1995) Forward Interest Rates as Indicators of Inflation Expectations, Stockholm School of Economics Seminar Papers, no. 594 (Stockholm: Stockholm School of Economics). Soros, George (2000) Open Society – Reforming Global Capitalism (London: Little, Brown). Spitäller, Erich (1978) ‘A Model of Inflation and Its Performance in the Seven Main Industrial Countries, 1958–76’, IME Staff Papers, vol. 25. no. 2, pp. 254–77. Stiglitz, Joseph (2002) Globalization and its Discontents (London: Penguin). Sutela, Peka (2000) ‘The Financial Crises in Russia’, in Tuomas Komulainen and Iikka Korhonen (eds), Russian Crises and its Effects (Helsinki: Kikimora). Svensson, Lars E. O. (1993) ‘Monetary Policy with Flexible Exchange Rates and Forward Interest Rates as Indicators’, Banque de France, Cahiers economiques et monetaires, vol. 43, no. 0396–4701, pp. 305–32. Taylor, Lance (1988) Varieties of Stabilization Experience – Towards Sensible Macroeconomics in the Third World (Oxford: Clarendon Press). Thacker, Strom Cronan (1999) ‘The High Politics of IMF Lending’, World Politics, vol. 52, no. 1, pp. 38–75.

Bibliography 147 Theil, Henry (1961) ‘Economic Forecast and Policy’ (Amsterdam: North-Holland). Thornton, Daniel L. and Dallas S. Batten (1985) ‘Lag-Length Selection and Tests of Granger Causality between Money and Income’, Journal of Money, Credit and Banking, vol. 17, no. 2, pp. 165–78. Turnovsky, Stephen J. (1970) ‘Empirical Evidence of the Formation of Price Expectations Hypothesis’, Journal of the American Statistical Association, vol. 65, no. 332, pp. 1441–54. United Nations (UN) (2000) World Economic and Social Survey 2000: Trends and Policies in the World Economy (Washington, DC: United Nations). United Nations and Economic Commission for Europe (UN and ECE) (2000) Economic Survey of Europe, no. 1 (Geneva: UN and ECE). Wachtel, Paul (1977) Inflation Uncertainty and Savings Behavior Since the Mid-1950s, Explorations in Economic Research, vol. 4, Fall, pp. 55–78. Walters, A. A. (1971) ‘Consistent Expectations, Distributed Lags and the Quantity Theory’. The Economic Journal, vol. 81, no. 322, pp. 273–81. White, Halbert (1980) ‘A Heteroskedasticity-Consistent Covariance Matrix and a Direct Test for Heteroskedasticity’, Econometrica, vol. 48, no. 4, pp. 817–38. Wiener, Norbert (1956) ‘The Theory of Prediction’, in E. F. Beckenback (ed). Modern Mathematics for Engineers (New York: McGraw-Hill). Woodford, Michael (1994) ‘Monetary Policy and Price Level Determinacy in a Cash-in-Advance Economy’, Economic Theory vol. 4, no. 3, pp. 345–80. Working Centre for Economic Reforms (1994) Russian Economic Trends (RET), vol. 3, no. 3 (London: WCER). Working Centre for Economic Reforms (1995) Russian Economic Trends (RET), vol. 4, no. 2 (London: WCER).

Index Afghanistan 17 Albania 94, 104, 105 aid 113, 117 anti-alcohol campaign 17 Argentina 93 Armenia 17, 104, 105 arrears external debt 98 other 28 payment 20 tax 18 assets domestic 4, 5, 29, 93, 94, 137, 138 foreign 10, 138 stripping 119, 122 Asian crisis 20, 107 assistance consultative 97 external 100, 116 financial 49, 101, 112, 114 IMF 98 international 7, 97 technical xiii, 121 Western xi, 97, 117, 121 Word Bank 117 autocorrelation 49, 57, 59, 63, 65 function 39 partial 39, 40 zero 30 autoregression 31, 32, 36, 126, 132 balance of payments 98, 105, 110, 114 banking crises 120 system 19, 20, 56, 78, 124 barter 92 Black Tuesday 19, 80, 131 bonds 7, 19, 117, 124 borrowing

domestic 17, 21, 22, 92 foreign 19, 104, 120 Brazil 93, 112, 113, 114, 138 Breusch–Godfrey test 33, 49 Bulgaria 12 budget balance 17 consolidated 129 constraint 12, 14, 16, 20, 21; intertemporal 2, 12, 16 deficit 2, 3, 7, 8, 10, 12, 14, 16, 17, 18, 19, 20, 21, 22, 27, 28, 53, 54, 116, 117, 118, 119, 123; net-of-interest 22; persistent 12 distortions 17 enlarged 18 federal 18 regional 18 revenue 17, 18 Bundesbank 94, 95, 137 capital flows 8, 98, 102, 103, 104, 138; net 105, 106; unrecorded 103, 138 inflows 20, 103 flight 8, 29, 103, 104, 105, 118, 119, 136, 138 markets 54, 119 outflow 103, 107, 138 private 103 capitalism ersatz 119 robber 120 Russian style 120 Wild East 119 causality 76, 83, 135, 137 bilateral 84, 85, 87, 92, 96, 118, 122, 128 direction 81, 84, 85, 96, 128 feedback 87, 92, 95, 96

Index 149

causality – continued Granger 76, 81, 82, 84, 85, 91, 127 lag length selection criterion testing xi, xii, 77, 81, 82, 85, 87, 88, 90, 96, 130 unidirectional 81, 84, 127, 128, 138 Wiener–Granger 81 Central Bank of Russia (CBR) 18, 19, 53, 92, 94, 95, 110, 116, 118, 131, 137 balance sheet 4, 53 budget financing 19, 53 credits 10, 11, 13, 53, 54, 129 interventions 54 net domestic assets 93, 94 reserves 19 Central and Eastern Europe 53, 76, 97, 99 central planning 1, 17, 97, 102, 114, 120 Chernobyl 17 Chile 93 CIS 137 closed economy 16 cointegration 56, 78, 133 commercial banks 5, 6, 10, 19, 53, 56, 78, 93, 95, 103, 129 communism 17 competition policy 119 conditionality 109, 114, 115, 121 constraint budget 2, 12, 14, 16, 20, 21 hard 11 quantity 53, 54 consumer preferences 3 consumption 13 correlograms 32, 39 corporate governance 20 corruption 122 CPI 56, 78 crawling peg 28, 93 credibility 7, 49, 94, 97, 116, 120

credit bilateral 102 cash 11 CBR 6, 7, 10, 11, 53, 129 excessive 27 expansion 6 growth 53 net: to commercial banks 10; to former Soviet republics 11; to government 5, 7, 93, 137 non-cash 11 risk 10 unauthorized 11, 116 working capital 10 crises banking 120 currency 99 financial 98, 114, 123 fiscal 17 other 99, 114, 131 current account 99, 103 currency appreciation 125 band 28 in circulation 5, 8, 56, 129 common 94, 100, 116 crisis 99 devaluation 123 disbursements 109, 110 domestic 1, 123 foreign 8, 16, 17, 26, 54, 124; disbursements 109, 110, 112, 113; reserves 21, 49 overvalued 123 reform 3 risk premium 20, 129 run 117 strong 124, Czech Republic 103, 130, 138 debt cancellation 117 deferrals 103 external 7, 98, 107, 110 financing 2, 16; unsustainable path 20, 22, 123 forgiveness xiii, 117, 121

150 Index

debt – continued government: domestic 11, 13, 14, 17, 19, 21, 113, 123; foreign 17, 20, 123, 124; interest bearing 14, 15; long-term 10, 106; real 2; short-term 106, indicators 107, 108 internal 7, 110 market 19 matured 20 multilateral 111 payment deferrals 112 private 14, 113 public 16, 21 relief 103, 115; multilateral 11 repayment 19, 113, 123; roll-over 19 rescheduling 117 restructuring 112, 113, 117 ruble-denominated 20 servicing 15, 20, 98, 110; regime 15 short-term 20, 105, 106 sovereign 110 Soviet 117 trap 20 demand aggregate 99 excess 3 for financial assets 3, 26 for GKOs 19 for money 2, 14, 26, 28, 93, 94, 95, 130; Cagan’s function 14 deposits auctions 123 foreign exchange 56, 78 holders 11 ruble 5, 19, 26, 56, 78, 129 saving 26, 78 time 78 determinants inflation 1, 25 money 5, 11 differencing transformation 56, 78 discount rate 20

dollarization 92 Duma 10 dummy variable for exogenoity 41, 56, 58, 67, 80, 83, 131, 134, 135 external shock 36 seasonal 4, 56, 83 structural break 36, 67, 68, 79, 80, 131, 133 economic agents 8, 23, 27, 28, 29, 31, 45, 46, 50, 130 development 102 environment 74, 76, 92, 107, 124 growth 94, 99, 114, 119, 120, 125 performance 103 policy xi, 17, 70, 96, 114, 122, 132, 135, 136, 137; decisions 17; error 115 reforms: geneal 101, 117, 120, 121, 122, 123; piecemeal 1; pseudo 101; radical 1, 101; sequencing 119; structural 103, 123, 125 regime xi, 22, 28 resources 112 system 75, 96 transformation 17 efficiency test 45 strong 47, 48 weak 46 empirical analysis xii, 22, 122 findings xii, 12, 76, 122, 124 results 21, 23, 52, 77 studies xi, 2, 24, 26, 52, 59, 76, 77 test xii, 77, 82, 122 employment 10, 13, 93, 135 enterprise credits 10 equilibrium 12 error term 31, 32, 41, 58, 59, 60, 126, 136 estimate 20, 35, 39, 40, 41, 91 excess reserves 95

Index 151

exchange rate 11, 17, 19, 21, 28, 63, 93, 130 appreciation 124 anchor 28, 94 band 54, 123 based stabilization 49, 50, 54, 93, 101, 122, 123 collapse 36, 54 control 131 corridor 93 crawling peg 28, 93 crises 41, 42, 57, 74, 80 floating 125 nominal 99, 124, 131 official 129 overshooting 93 overvalued 119 peg 94 policy 129, 131 real 16, 29 stabilization fund 117 stability 124, 125 target 20, 28, 54, 100 exogenous shock 31, 131, 136 expectations adaptive 39 anchoring 28 error estimating 32, 45, 46 formation 32, 42, 45, 46, 51 generating 25, 51 hypothesis 24 inflationary xii, 23, 24, 25, 27, 29, 31, 47, 49, 50, 51, 118, 125, 132; direct measurement 25 investor 115 models 24, 25; deterministic 24; hybrid 24, 130 rational 13, 16, 23, 31, 45, 130 true mathematical 45 unbiased 45, 47 expenditure debt service 110 government 12, 14, 16, 18, 19, 20 military 17 public 13, 17 export tariffs 18

extended fund facility (EFF) 109, 120, 137 extrabudgetary funds 7, 18, 120 external shock 36, 125 F-test 72, 84, 85, 86, 136 FDI 103, 105, 106, 107, 120, 138 fiscal authority 13, 15, 16, 22 adjustment 12, 49, 50, 94, 137 correction xii, 15, 22 crisis 17 deficit 12, 101, 115 difficulties 113 dominance 2 indicators 115 performance 12, 18, 54 policy xi, 1, 2, 11, 12, 13, 14, 15, 17, 18, 21, 29, 54, 99, 122, 125, 131 programme 16 regime 12, 22 responsibility 94, 95 side 13, 125 stance 2, 18 theory of price level (FTPL) 2, 11, 12 variables 45 financial architecture 98, 100, 115 assets 3, 26 assistance xiii, 49, 101, 112, 114, 117, 121 community 102, 112 consolidation 109 crisis 12, 13, 20, 22, 56, 95, 98, 100, 101, 112, 114, 123, 131 disaster 115 discipline 92, 95 help 114 inflows 103 innovations 96 institutions xi, xiii, 97, 99, 101, 103, 138 investors 21 liberalization 101 linkages 114

152 Index

financial – continued market 7, 28, 113, 115, 125; indicators xii, 23, 24, 25, 51 meltdown 12 performance 115 relief 112 resources 103 stabilization 117 support 98, 116, 118 system 8, 10, 31, 55, 97, 98, 99, 115, 136 transfers 16 turbulence 20 forecast conditional variance 43 dependant variable 43 error 45, 46, 47, 59, 134; autocorrelated 49; mean squared 42; regressed 47; series 46; statistics 42 mean 42 one-step-ahead 43 perfect 93 static 43, 44 test: Chow 37, 131; N-step 134; one-step 36, 67, 68, 134 unbiased 45, 50 Former Soviet republics 5, 11, 53, 93, 111, 116 Friedman rule 15, 52, 130 G7 101, 117, 132 Gallup poll 24, 130 Gasprom 18 global financial turbulence 20 GKO 19, 20, 113, 115 goal variable 77, 92, 95, 96, 128, 135, 136, 138 government bonds 14 budget: balancing xi; constraint 14, 16, 20; deficit xi, 7, 8, 15, 16, 21–2, 53, 54, 123, 124 central 5 credibility 94, 116 credit, net 7, 93, 137

debt 14, 15, 16, 19, 20, 108, 129; composition 13; nominal 11; real 2 default 12, 20, 21, 123 enlarged 93 expenditure 14, 20 federal 110, 119 policy 3, 14, 16, 20, 31, 51, 125, 136 revenue 119 securities 10, 14, 107 spending 17 surplus 16 grants 10, 138 gross domestic product (GDP) 7, 8, 9, 10, 11, 18, 19, 21, 27, 101, 103, 105, 106, 110, 116, 124, 125, 129, 130 gross national product (GNP) 76, 109, 135 hedging 3, 8, 136 heteroscedasticity 33, 59, 64, 65, 66, 71, 134 high-powered money 5, 15 Hungary 103, 107, 130, 138 Indonesia 113, 138 inflation accelerating 10, 24, 55 accommodation 92, 95 causes 2, 3, 12, 29, 85 chronic 49 contemporaneous 50, 58 control xi, xii, 3, 22, 54, 76, 77 corrective 114 cure xii, 14 definition 27, 76 determinants 1, 5, 11, 25, 65 effects on economy 2 emergence 2 estimate 35, 47, 94, 114 exogenously induced 41, 54, 80 expectations: downward 125; formation 25, 42, 46; rational 45, 47; role 23; survey data 25; unbiased 47

Index 153

inflation – continued expected rate 14, 29 fitted 42, 87 forecasting 39, 42, 44, 45, 50 future xi, xii, 28, 29, 50, 54, 55, 72, 122, 131 hedging 3, 136 high 2, 10, 12, 19, 54, 77, 92, 93, 136 history xii, 23, 28, 29, 31, 42 inertia 75, 118 lagged xi, xii, 23, 25, 31, 46, 50, 55, 75, 80, 131, 132, 133, 134 method, financing deficit 2, 13 model 4, 5, 32, 33, 35, 37, 40, 43, 44, 47, 50, 60, 61, 64, 65, 67, 68, 71, 72, 73, 75, 127, 135 monetary phenomenon 100 monetary roots 2 open 1, 2, 23, 24, 25, 129 path 30 permanent 27 persistence xii, 23, 24, 27, 29, 31, 49, 50, 122, 132 process xi, 58, 70, 75 shocks 44, 75 stabilization 28, 29, 92, 93, 123 sticky 31 structural break 41, 43, 67, 79, 133, 135 targeting 24, 124, 125 tax 8, 9, 10, 27, 129, 130 temporal pattern 52 volatile 1, 50 inflationary economic growth 120 environment xi, xii, 27, 55 expectations xii, 22, 23, 24, 25, 27, 28, 32, 49, 50, 51, 118 financing 118 inertia 49, 59 money creation 16 spiral 94

institutions building 98 international financial: IMF xi, xiii, 21, 25, 26, 53, 93, 94, 96, 97, 98, 99, 100, 101, 102, 103, 109, 110, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 136, 137, 138; World Bank 97, 98, 99, 102, 111, 113, 117, 119, 120 interest foreign debt 17 government debt 19, 20 rate: clearing 53; forward 24; market-clearing; nominal 6, 8; policy 125; real 10, 14, 53, 54 securities 19 service 110 international assistance 7 bond issues 103 capital investors 20, 121 community 112 creditors 102 economic transactions 16 financial architecture 115 financial institutions xi, xiii, 97 rescue package 113 reserves 4, 5, 92, 115, 123, 137 trade 125 investment 2, 13, 23, 78, 123, 124, 125, 136 Kazakhstan 11, 137 Keynesian tradition 14 Korea 112, 138 lag length: determination ad hoc xxi, 32, 59, 77, 82, 83, 84, 85, 87, 96, 137; Akaike Information Criteria (AIC) 59, 77, 126, 133, 134; arbitrary 56, 82, 84, 85, 96, 137; BEC 82, 83, 85,

154 Index

lag – continued 87, 128; choice xii, 82, 85, 87, 96; FPE 82, 83, 84, 85, 86, 87, 96, 126, 127, 128, 136, 137; optimal xii, 31, 32, 35, 56, 57, 59, 70, 96, 128, 138; rule-of-thumb method 56, 59, 82, 85; statistical criteria xii, 56, 77, 82, 85, 87, 96, 134, 137; Schwarz (Bayesian) Criteria (SC) 32); order 82, 83, 85, 126, 127 space 77, 87, 137 specification 70, 82, 86, 136 polynomials 81, 87 structure 59, 63, 70, 135 Latin America 97, 114 Latvia 94 lender of last resort 98, 120 lending disbursements 109, 110, 112, 113 liberalization foreign exchange market 99 price 2, 3, 22, 30, 53, 55, 75, 77, 78, 94, 99, 119, 136 trade 100 linear trend 30, 47, 57 liquidity crunch 95 Lithuania 94 Livingston Survey 24 Ljung–Box Q-statistics 34, 61, 62, 63, 66, 72, 74, 83, 137 loan non-guaranteed bank 103 promised 92, 101, 102, 109, 117 pseudo 101 stand by 137 London Club 107, 110, 111 Long-run multiplier 60, 71, 72 Lucas critique 51 Macedonia 106 macroeconomic instability xi, 1, 21 literature 2, 11 problems 8, 13 stabilization 3, 10, 13, 53, 76

stability 1, 22, 53, 54, 98, 100, 101, 122 traditional 13 weaknesses 20 maximum likelihood estimation (MLE) 35, 60, 61, 65, 73 measurement errors 28, 51 Mexico 113 models adaptive expectations hypothesis 24, 39 ad hoc xii, 32, 59, 77, 82, 83, 84, 85, 87, 96, 137 Almon 58, 133 ARCH 33, 34, 35, 43, 44, 46, 47, 59, 60, 61, 63, 65, 68, 70, 72, 83, 131, 132, 134, 135, 137 ARIMA 39 ARMA 39, 40, 41, 42, 43, 46, 50, 132 autoregressive 12, 31, 32, 33, 34, 37, 38, 39, 51, 58, 59, 63, 64, 65, 68, 75, 82, 83, 126, 127, 131, 134 demand for financial assets 3, 26 distributed lag xii, 4, 5, 58, 59, 60, 61, 69, 71, 72, 73, 81, 128, 135; fit 5, 35, 36, 41, 42, 67, 70, 125, 134 GARCH 34, 35, 61, 63, 83, 91, 131, 134, 137 Koyck DL 58, 133 Markov-switching 50 moving average 39 Polak 114 selection xii, 22, 53, 137 monetarist paradigm 77, 137 supposition 15 unpleasant arithmetic 15, 16, 117 monetary policy coordination 122 expansionary 13 framework 13 instruments xi, 79, 117, 123

Index 155

monetary accommodation, partial 92 aggregates xi, xii, 9, 52, 53, 54, 55, 56, 57, 59, 61, 63, 67, 70, 72, 74, 76, 77, 78, 84, 85, 87, 95, 96, 122, 123, 124, 128, 135, 136 authorities 3, 13, 19, 22, 92, 96, 124, 137 determinants 5, 11 expansion 53, 76, 115 financing 8, 9, 117 impulses, transmission xi, xii, 53, 54, 55, 72, 122 non: economy 118; factors 58; means 117, 118; way 8 overhang 3, 25, 26, 27, 116 phenomenon 3, 15, 49, 52, 100 policy: accommodating 49, 92, 95; expansionary 13; future stance 28; loose 6; tight 15, 27, 28, 29, 53, 92, 99, 100 roots, inflation 2, 55 stability 70 targeting 77, 78, 92, 93, 94, 95, 96, 118, 123, 124, 137 tightening 27, 28, 29, 92, 99 wealth 8, 14 monetization economy 124 deficit xi, 7, 21, 22 low 8 re- 94 money base 5, 6, 8, 14, 15, 21, 56, 129 broad xii, 3, 5, 6, 8, 9, 27, 55, 56, 63, 65, 68, 70, 72, 74, 75, 77, 78, 84, 85, 87, 95, 96, 118, 122, 127, 136 creation 7, 8, 16, 22 demand 93, 94, 95, 130 demand equation 2 depreciation 28, 29, 129 extended broad 46, 56, 78, 84, 85, 96

multiplier 6, 95, 129 narrow 56 printing 17, 21 quantity theory 13, 28, 52, 130 stock 54, 76, 95, 96, 103 supply: growth 124; ceiling 93, 115, 137; target 137 surrogates 92 system, dual 11 velocity 3, 26, 27, 28, 29, 52, 55, 92, 93, 95, 130 moral hazard 102, 112, 115 moratorium 112, 113, 123 near abroad 11, 123 net domestic assets 4, 93, 137 new industrialization 17 normality test 31, 34, 59 OECD 93, 116 OFZs 19 open economy framework 16 ordinary least squares (OLS) 4, 32, 33, 36, 40, 47, 48, 59, 60, 63, 64, 68, 70, 71, 83, 91, 134, 135, 136, 137 output adjustment 114 fall 101, 116 gap 50 Paris Club 103, 111 payment system 6, 29 Phillips curve 24, 132 planned economy 97, 99 postcommunist economies xi, xii, 1, 2, 3, 21, 22, 23, 24, 25, 31, 52, 53, 65, 74, 82, 95, 96, 98, 101, 115, 118, 122, 129, 132 poverty 98, 102, 114, 119 price administrative 28, 63, 125 adjustment 27, 31, 63 consumer 3, 29, 56, 78, 130 control 3, 49 downward sticky 31 decontrol 1, 101

156 Index

price – continued deflator 26 determination 1, 11, 12, 28 regime: monetary dominant 1; fiscal policy 1 fiscal theory 2, 11, 12 fixed 114 jump, one-off 3, 27 level: general 27, 58, 76; time path 15 liberalization 2, 3, 8, 22, 23, 25, 26, 27, 30, 31, 44, 53, 55, 56, 75, 77, 78, 94, 99, 115, 119, 122, 136 movements 96 oil 20, 113, 123, 124 path 15, 30 producer 3 relative 49 rigid 27, 49, 93 rise xi, 1, 3, 24, 26, 27, 29, 52, 53, 74, 125 relative 49 stability 12 volatility 27 world market 17, 123 privatization 1, 94, 99, 100, 119, 122 policy-makers 125 quantity equation

1, 25, 52

rational expectations model 16, 129, 130 hypothesis 13 reciprocal roots 41, 131 refinance rate 10 reforms gradual 100 institutional 13, 100 market 1, 2 Polish 1, 117 pseudo 101 shock 100 structural 100, 103, 114, 123, 125 regime change 14 rescue package 112

reserves banks’ 5, 129 excessive 6, 10, 95 foreign currency 4, 5, 17, 19, 20, 21, 49, 93, 115, 118, 123, 124, 131, 137 gold 17, 21 required 10 reserve requirement 6 Ricardian regime 15 Romania 12, 13 ruble convertibility 99 deposits 8, 19, 56, 78 depreciation 28, 29, 129 devaluation 113, 123, 124 non-cash 11, 116 notes 11, 116, 132 Soviet 100, 116 stock 8 zone 11, 54, 116 Russia, Czarist 97 Russian financial crises 12, 13, 20, 22, 56, 95, 100, 107, 110, 113, 122, 123 government 1, 3, 20, 25, 100, 101, 117 reformers 1, 18, 100, 102, 120, 121 transformation crises 123 safety net 100 Saint Louis equation 58 Sargent’s system 15 saving forced 114 deposits 26 velocity 3 rates 11 securities 10, 19, 107, 125 shock therapy 3 simultaneous equation 39 seasonality 56, 80 seigniorage 8, 9, 129 serial correlation 33, 48, 49, 59, 83 Slovenia 94, 138

Index 157

Soviet Union 1, 27, 94, 100, 102, 103, 108, 116 economy 3 republics, former 5, 11, 27, 53, 93, 94, 103, 111, 116 speed of transmission xi, xii, 53, 54, 55, 72, 74, 122 stability tests 36, 67 stabilization exchange rate-based 49, 50, 93, 101, 122 fiscal and monetary framework 13 fund 117, 124, 125 heterodox 28, 49, 50, 101, 118 orthodox money based xii, 49, 50, 93, 94, 95, 100, 118, 122, 123 policy 23, 49 programme 49, 50, 53, 54, 76, 93, 95, 97, 98, 100, 101, 112, 121, 122, 123, 129 stagflation 24 stationarity 30, 46, 56, 78, 80 statistics descriptive 30 diagnostic 32, 35, 40, 41, 46, 56, 59, 61, 63, 70 Durbin–Watson (DW) 59 Ljung–Box Q 34, 61, 62, 63, 66, 72, 74, 83, 137 summary 56, 72, 74, 75 Run test 63, 133 Theil inequality coefficient 42 stochastic error term 31, 58 shocks 131 structural break 36, 41, 43, 67, 68, 79, 80, 130, 133, 135 reforms 97, 98, 103, 123, 125 structuralism 100 subsidies 3, 10, 12, 18 supply shocks 27 Swiss National Bank 94, 95 target credit

93

intermediate 28, 77, 93, 96, 128, 136 quantitative 54, 95 tax arrears 18 code 18 collections 123 concessions 18 deferrals 18 discipline 18 exemptions 18 excess wage 18 evasion 18, 118 implicit 7 profit 18 receipts 17 revenues 18, 118 rate 13, 18 real 14, 17 system 18 test Breusch–Godfrey 33, 49 Chow 37, 40, 67, 131, 34 Dickey–Fuller augmented 30, 56, 78 Efficiency: strong form (orthogonality) 45, 46, 47, 48, 49, 50; weak form 45, 46 F-statistic 33, 49, 63, 84, 85, 87, 131, 137 Jarque-Bera normality 31, 34, 59, 60, 83, 134 KPSS 30, 78, 79, 80 LaGrange Multiplier 32, 33, 47, 59, 83 Ljung-Box 30, 34, 83 Phillips–Perron 30, 56, 78 RESET 39 Perron’s 67, 78, 79, 80, 133 skewness 31, 34, 59, 63, 83, 134 stability 36, 37, 67 unbiasedness 45, 82 unit root 30, 46, 47, 56, 57, 78, 79, 80, 130, 133 Wald 36, 47, 74, 134, 135 zero correlation 46, 47, 48 Thailand 112, 113

158 Index

time series x, xii, 2, 23, 25, 39, 55, 56, 78, 79, 81, 95, 133 trade shocks 11, 116, 125 transformation crises 123 differencing 56, 78 economic 17, 75, 96, 101 recession 92 systemic 3 transition economic xi, 1, 2, 3, 6, 11, 12, 13, 17, 18, 20, 23, 25, 53, 55, 68, 93, 94, 95, 97, 98, 109, 113, 122 economies 25, 26, 28, 51, 52, 53, 54, 76, 77, 93, 94, 95, 98, 99, 100, 102, 103, 104, 105, 106, 107, 108, 125, 135 transmission speed xi, xii, 53, 54, 55, 72, 74, 122 treasury bills 19, 115, 117, 118 trend linear time 47, 57, 79, 80, 83, 137 stationary process 30, 80, 130, 133 unemployment 24, 45, 49, 50 unit root 30, 46, 47, 56, 57, 78, 79, 80, 130, 133

United States 102, 132 USSR 17

24, 77,

vector-autoregressive models (VAR) 12 velocity broad money 26, 27, 29 cash 26 deposits 26 income 26 saving 3 Visegrad countries 95 volatility 6, 27, 44, 50, 59, 60 wages bargaining 23 controls 28, 49 determination 28 excess tax 18 freezing 28 stickiness 31 target 100 Wald test 36, 47, 74, 134, 135 Washington Consensus 98, 99, 119 white noise 32, 40, 81, 83, 87, 91, 126, 136 Yeltsin, Boris

18