Trade, Integration and Economic Development: The EU and Latin America (Schriftenreihe der Österreichischen Gesellschaft für Europaforschung (ECSA Austria) ... Association of Austria Publication Series)

SpringerWien NewYork

European Community Studies Association of Austria (ECSA Austria) Publication Séries Volume 13 Schriftenreihe der Ôsterreichischen Gesellschaft fur Europaforschung (ECSA Austria)

SpringerWienNewYork

Gabriele Tondl (éd.) Trade, Intégration and Economie Development The EU and Latin America

SpringerWienNewYork

Prof. Dr. Gabriele Tondl Research Institute for European Affiars, Vienna University of Economies and Business Administration

Financial support was given by Bundesministerium fur Wissenschaft undForschung, Wien, and the European Commission, DG Education and Culture, Brussels This work is subject to copyright. AU rights are reserved, whether the whole or part of the material is concerned, specifically those of translation, reprinting, re-use of illustrations, broadcasting, reproduction by photocoping machines or similar means, and storage in data banks. © 2008 Springer-Verlag Wien Printed in Austria SpringerWienNewYork is a part of Springer Science + Business Media springer.com Product Liability: The publisher can give no guarantee for ail the information contained in this book. The use of registered names, trademarks. etc. in this publication does not imply, even in the absence of a spécifie statement, that such names are exempt from the relevant protective laws and régulations and therefore free for gênerai use. Typesetting: Caméra ready by editor Printing: Ferdinand Berger & Sôhne Gesellschaft m.b.H., 3580 Horn, Austria Printed on acid-free and chlorine-free bleached paper SPIN: 12064786 Library of Congress Control Number: 2008925001 ISSN 1610-384X

ISBN 978-3-211-75149-7 SpringerWienNewYork

Preface During the past decade, Latin American (LA) countries have become highly open economies. Trade agreements were implemented fostering trade relations both with other Latin American countries and with third countries, among them the main external trading partner, the EU and the US. This policy approach differs significantly from the import substitution policy dominating in LA in the 1970s. LA countries have developed manifold trade relations. At the same time economic integration has also developed in LA and become most advanced in the case of Mercosur. Especially for South America, Europe is an important trading partner and investor. Moreover, the economic integration process accomplished by the European Union (EU) is often seen as a role model for LA integration. For both regions, the EU and LA, it is important to understand economic developments in the region of the trading partner and to know why the other side pursues certain interests in the process of trade negotiations. To deepen the understanding of these issues, is the main objective of this volume. The volume is the proceedings of the International Conference – 7th Arnoldshain Seminar on “Trade and Integration. The EU and Latin America”, organized at Vienna Economics University, August 28 – September 1, 2006. It was the 7th conference of the Arnoldshain Network, a group of academics from the Johann Wolfgang Goethe University Frankfurt, the Universidad Nacional de Cordoba, Argentina, the Universidade di Sao Paolo and the Vienna University of Economics and Business Administration. The Arnoldshain network, named after the location of its first meeting in Germany, was founded under the wish to promote academic cooperation between researchers from Europe and Latin America and to support student exchange between the involved institutions. At the Vienna conference, the network concluded the foundation of the “International Association for Comparative Economics and Integration” which should henceforth become the institutional framework of the Arnoldshain Network to pursue its activities. The foundation meeting was attended by the network’ s initiator, Prof. em. Ulrich Peter Ritter, Prof. em. Roland Eisen (both Johann Wolfgang Goethe University Frankfurt), Prof. Díaz Cafferata, Prof. María-Luisa Recalde, Prof. Ángel Enrique Neder (all Universidad Nacional di Cordoba), Prof. Basilia Aguirre (Uni-

VI

versidade di Sao Paolo), Prof. Gabriele Tondl (Vienna University of Economics and Business Administration) and a number of new members from Europe and LA which contributed to the conference. The conference started with an introductory session where representatives from international institutions, government offices and academia introduced the conference subject and discussed the interests of both regions. Ricardo Santiago, at that time director of the Inter-American Development Bank, Paris, sketched the recent economic developments and trade integration in LA. Alfredo Valladao, Professor and Director of the “Chair Mercosur”, Science Po, Paris, talked about the aims and achievements of Mercosur. Marcel Vaillant, professor at Universidad de la Republica, Uruguay, explained the trade policy objectives of Mercosur vis-à-vis the EU while Philipp Dupuis, Deputy Head of Unit “Trade with Latin America”, Directorate External Trade, European Commission, introduced the EU trade policy and the EU Latin America Agenda, and Andreas Melán, Head of the Latin American and Caribbean Unit, Austrian Foreign Ministry, reported on the latest EU- LA summit in Vienna in spring 2006 and the EU´s foreign policy with Latin America. The conference also hosted a Round Table on “Institutions and Good Governance” with Ulrich Peter Ritter, Peter Eigen,Head of Transparency International, Wolfgang Hetzer, Head of Unit Intelligence, European Anti-Fraud Office, and Basilia Aguirre, Universidade di Sao Paulo. During the other parts of the conference some 30 papers were presented and discussed by academics. Most of these papers are included in this volume. The books starts with the address of the initiator of the Arnoldshain Network and a protagonist of Comparative Economics, on “James Fenimore Cooper: An Astute and Critical Precursor of Comparisons in the Social Sciences”. An introductory chapter by Gabriele Tondl follows explaining trade and integration within Latin America and with its main trading partners. The first part of the book on “Open Economy Macroeconomics” contains three papers: The paper of Sergio Barone and Alberto Díaz Cafferata discusses the relationships between exports, external debts and external solvency in Argentina before the 2002 crisis. Fernando Zarzosa Valdivia presents a general equilibrium model of an open economy with real exchange rate movements which explains the Dutch Disease and functional and sectoral income distribution. The paper of Michael Brei contains an empirical analysis of the impact of current account reversals on relative prices in the

VII

Brazil. Part 2 of the volume deals with “Institutions for Development: The Case of Corruption” and contains a paper of Wolfgang Hetzer on the EU budget asking whether it is a breeding ground for corruption, and a second paper by Basilia Aguirre discussing the multiple faces of wrongdoing in the case of corruption Brazil. The third part of the volume contains seven papers which address “Trade and Integration Issues”. The paper of Laura MárquezRamos and Inmaculada Martínez-Zarzoso investigates empirically the effects of distance in a gravity model of foreign trade. Dierk Herzer looks at the composition of trade in Chile and its relationship with productivity. Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer and Inmaculada Martínez-Zarzoso analyse the role of price competition for the market share of Chile in different European product markets. María Luisa Recalde and Marcelo Florensa investigate whether the implementation of Mercosur has lead to trade creation or trade diversion. Rinaldo Antonio Colomé and Fernando M. Giuliano explain the differences of agricultural policies in the European Union and LA countries and its consequences for agricultural trade. Matteo Grazzi and Antonella Mori discuss the regulations of FDI in some LA countries and their impact on FDI inflows. Ángel Enrique Neder, Julieta Schiro and Jonatan Saúl give an assessment of financial integration in LA. Finally, part 4 deals with “Regional Issues”. It contains a paper by Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata which analyses regional income mobility in Argentina. The second paper by Jorge Alberto Fornero investigates whether Mercosur countries converge in per capita GDP and finally the paper by Basilia Aguirre and Guilherme Dias discusses fiscal federalism in Brasil. The participants of the Vienna conference made a valuable contribution to develop research in the field of the objectives of the International Association for Comparative Economics and Integration. The Association welcomes that most contributions have found its way into this volumes. A special thank is owed to Nurgül Özen for her huge secretarial support to produce the publication. Gabriele Tondl President of the International Association for Comparative Economics and Integration

Table of Contents

Address of the Initiator of the Arnoldshain Seminars Ulrich Peter Ritter James Fenimore Cooper: An Astute and Critical Precursor of Comparisons in the Social Sciences

3

Gabriele Tondl Trade and Integration in Latin America and with Its Main Trading Partners

17

Part 1:

Open Economy Macroeconomics

39

Sergio V. Barone and Alberto M. Díaz Cafferata Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

41

Fernando Zarzosa Valdivia Real Exchange Rate Movements, Dutch Disease and Functional and Sectoral Income Distribution

81

Michael Brei The Impact of Current Account Reversals on Relative Prices: The Brazilian Experience Part 2:

Institutions for Development: The Case of Corruption

111 129

Wolfgang Hetzer The European Budget, A Breeding Ground for Corruption?

131

Basilia Aguirre The Multiple Faces of Wrongdoing – A Closer Look on Corruption

141

Part 3:

149

Trade and Integration Issues

Laura Márquez-Ramos and Inmaculada Martínez-Zarzoso On Distance Effects in Gravity Models – Short Versus Long Distances

151

Dierk Herzer Trade, its Composition and Total Factor Productivity: Cointegration Evidence for Chile

169

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer and Inmaculada Martínez-Zarzoso Chile’s Market Share in the EU Market: The Role of Price Competition in a Panel Analysis Setting

189

María Luisa Recalde and Marcelo Florensa Mercosur: Trade Creation or Trade Diversion? An Application of the Gravity Model and Kalman Filter

221

Rinaldo Antonio Colomé and Fernando M. Giuliano Agricultural Policies and Trade in the European Union and Selected Latin American Countries

249

Matteo Grazzi and Antonella Mori The Regulation of FDI in Latin America and the Caribbean: What Impact on Host Countries?

271

Ángel Enrique Neder, Julieta Schiro and Jonatan Saúl Financial Integration in Some Countries of South America The Use of Interest Parity Conditions as Indicators

307

Part 4:

323

Regional Issues

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata Analysis of Regional Income Mobility in Argentina

325

Jorge Alberto Fornero Do Mercosur Countries Converge in Per Capita GDP and Productivity?

351

Basilia Aguirre and Guilherme Dias Fiscal Reform and Federal Relations

389

List of Authors

413

Address of the initiator of the Arnoldshain Seminars and of the International Association for Comparative Studies in Economics and Integration Ulrich Peter Ritter

James Fenimore Cooper: An Astute and Critical Precursor of Comparisons in the Social Sciences I. Introduction and biography “Had I only read this book when I was still teaching my class on comparative economics.” I thought more than once, when I read Cooper’s memories of his travels in Europe as the first book after my retirement. I suppose most of us know his name, some characters from his books and maybe even some of his books from our days as youngsters. But I suppose few of us know him as a methodologically conscious and truth searching comparatist. That he deserves these attributes I want to show in my paper. In addition I want to make a case for making use of this historical material in particular and historical material in general in teaching comparative economics to our students. Also I want to show that Cooper was aware of problems of comparing that are still relevant for us today, but are often overlooked because of the seeming availability of numerical material. Why am I so enthused? Because I would certainly have used this book as an illustration in my class on comparative economics had I read it before. I learned that Cooper was a comparatist, if there ever was one. In his book “A Residence in France With An Excursion Up The Rhine, And A Second Visit To Switzerland”1, Cooper compares nearly everything from Goethe and Schiller (JFC 2004, Letter 1

A German version was published under the title „Lebensbilder aus Frankreich, den Rheinländern und der Schweiz”, in Braunschweig in 1837, Reprint Kelkheim 2001. The English version was published in the Internet by the Gutenberg Project and will be refered to further on as “JFC 2004”.

4

Ulrich Peter Ritter

XIV), the electoral systems in the US and the Canton of Vaud in Switzerland (JFC 2004, Letter XVII), wine prices and wine cultivation (JFC 2004, Letter XVIII), the differences between American and European hands and feet!!!! (JFC 2004, Letter XXVI), democracy in the US and Switzerland (JFC 2004, Letter XXIII) and more. Some of these comparisons are, of course, very banal, but some of them show a surprising amount of sophistication and awareness of methodological difficulties. In this paper I want to present to you Cooper, the comparatist, demonstrating this by two of his more interesting comparisons: - A comparison of the cost of living in Europe and the United States - The controversy about the costs of republican versus monarchic government, in the literature known as the Public Finance Controversy. But before I do so, let me give you some information about his biography and his works. He was born in Burlington, N. J. on September 9, 1779, three years after the Declaration of Independence, and he died one day before his 72nd birthday on September 14, 1851, 6 weeks before the Coup d’Etat of Louis Napoleon in France, in Cooperstown, N. Y., a town named after his father. Reared in the wild country round Otsego Lake, New York, on the yet unsettled estates of his father William Cooper, a judge and member of Congress, he was sent to school in Albany and in New Haven, where he Yale at the age of fourteen, remaining for some time the youngest student on the rolls. In his junior year Cooper was expelled from Yale after having committed a series of pranks, one of them being the training of a donkey to sit in a professor’s chair (Redekop 1986). Three years afterwards he joined the United States Navy; but after making a voyage or two on a merchant vessel, to perfect himself in seamanship, and obtaining his lieutenancy, he married and resigned his commission in 1811. He then became a settler in Westchester Country, New York. He was thus very familiar with what he was later to describe in his most famous novels, because he grew up in the wilderness in what was then the country of the pioneer settlers, the hunters and trappers and last but not least the various tribes of the Indians that later were to play an important role in his novels. His books on the life on the frontier have largely founded our knowledge and fantasies of the US in those times. His tales are still

James Fenimore Cooper: An Astute and Critical Precursor of Comparisons in the Social Sciences

5

read the world over and still serve as the script for many a movie and TV feature. For him the American Indians were human beings. He did not believe in the stereotype of the cruel savage. As Karl Meyer Frommhold says in his edition of the Leather Stocking tales (1996), posterity owes it in good part to him that the American Indians were not completely exterminated. Cooper was the first and best known American author of his time and one of the most popular 19th century authors. His stories have been translated into nearly all the languages of Europe and into some of those of Asia. Balzac admired him as did Johann Wolfgang von Goethe; Victor Hugo pronounced him greater than the great master of modern romance, Sir Walter Scott, and this verdict was echoed by a multitude of readers, who were satisfied with no lesser title for their favourite than that of “the American Scott.” He has, of course, had also adversaries who critique his enormous vanity and his irritability. Wikipedia (July 2006), making reference to the Encyclopedia Britannica, says: “It is only as a novelist that he deserves consideration. His qualities are not those of the great masters of fiction; but he had an inexhaustible imagination, some faculty for simple combination of incident, a homely tragic force which is very genuine and effective, and up to a certain point a fine narrative power.“ James Fenimore Cooper is generally known as the first American writer to reach worldwide recognition, with his more than 30 books, numerous articles and essays and hundreds of letters, some of which are still not yet published. He is better known for his writings on naval themes and on the life on the big Frontier than as a comparatist. However, he lived for seven years in Europe between 1826 and 1833 and thus at a time when Alexis de Toqueville explored the United States. He then travelled widely in England, France, Germany, Italy and Switzerland. He wrote diaries during his stay that are full of comparisons, some of which deserve even to be called comparative studies. II. A comparison of the cost of living in Europe and the United States The first comparison I want to highlight is a comparison of the costs, or as Cooper says, “expences of living” in Europe and the U.S., published as the 7th of 28 letters of his book (JFC 2004, Letter VII).

6

Ulrich Peter Ritter

Cooper does not take this comparison lightly. He starts his argument with the phrase: “Your question, as to the comparative expense of living at home and of living in Europe, is too comprehensive to be easily answered…” One reason”, he points out, is that “prices vary so materially that it is difficult to make intelligent comparisons” (JFC 2004, Letter VII). That is why he limits himself mostly to a comparison between Paris and New York, a very smart decision in view of the scarcity of comparable data. Right in the beginning he introduces a definition of standards of reference: “so long as one keeps within the usual limits of American life, or is disposed to dispense with a multitude of little elegancies”. This restriction is necessary because “while no money will lodge a family in anything like style, or with suites of rooms, ante-chambers, etc. in New York, for the simple reason, that buildings which possess these elegancies, or indeed with fine apartments at all, have never yet been erected in the country.” He also formulates an initial hypothesis: “If one does this, New York has the advantage over Paris” (JFC 2004, Letter VII). He then goes on to explain why this is so. Cooper answers the question in a differentiated form. He first speaks about lodging, then about food and wine and then about clothes and clothing style. But he also speaks about tariffs, taxes, and habits, as well as the organization of commerce and the motivation of consumers and merchants. As to lodging he points out: “a family can be better lodged in a genteel part of the town for less money, than it can be lodged, with equal room and equal comforts, in a genteel quarter of Paris; always excepting the inferior distribution of the rooms, and other little advantages, such as the convenience of a porter, etc. all of which are in favour of the latter place.” In a footnote he explains this: “[Footnote 17: In New York, the writer has a house with two drawing-rooms, a dining-room, eight bed-rooms, dressing-rooms, four good servants' rooms, with excellent cellars, cisterns, wells, baths, water-closets, etc. for the same money that he had an apartment in Paris, of one drawing-room, a cabinet, four small and inferior bed-rooms, dining-room, and ante-chamber; the kitchens, offices, cellars, etc. being altogether in favour of the New York residence. In Paris, water was bought in addition, and a tax of forty dollars a year was paid for inhabiting an apartment or a certain amount of rent; a tax that was quite independent of the taxes on the house, doors, and windows, which in both cases were paid by the landlord.]” (JFC 2004, Letter VII).

James Fenimore Cooper: An Astute and Critical Precursor of Comparisons in the Social Sciences

7

After this comparison of the cost of lodging Cooper goes on to compare the cost of food of all kinds. “Food is much the cheaper in New York bread alone excepted,” he says. This may not surprise us. But it is less self-understood, when he goes on to say that “Wines can be had, as a whole, better and cheaper in New York, if obtained from the wine-merchant, than in any European town we have yet inhabited. Even French wines can be had as cheap as they can be bought here (i. e. in Paris), for the entrance-duty into the country is actually much less than the charges at the gates of Paris. The transportation from Bordeaux or Champagne, or Burgundy, is not, as a whole, essentially less than that to New York, if indeed it be any less.” (JFC 2004, Letter VII) “All the minor articles of table luxuries, unless they happen to be of French growth, or French fabrications, are immeasurably cheaper in America than here i.e. in Paris.” Clothes are nominally much cheaper here than with us; but neither the French nor the English use habitually as good clothes as we; nor are the clothes generally as well made. You are not, however, to suppose from this that the Americans are a well-dressed people; on the contrary, we are greatly behind the English in this particular, nor are our men, usually, as well attired as those of Paris. This is a consequence of a want of servants, negligent habits, greediness of gain, which monopolizes so much of our time as to leave little for relaxation, and the high prices of articles, which prevent our making as frequent calls on the tailor, as is the practice here.” (JFC 2004, Letter VII) “My clothes have cost me more in Europe, however, than they did at home, for I am compelled to have a greater variety, and to change them oftener. Our women do not know what high dress is, and consequently they escape many demands on the purse, to which those of Paris are compelled to submit. It would not do, moreover, for a French belle to appear every other night for a whole season in the same robe, and that too looking bedraggled, and as jaded as its pretty wearer. Silks and the commoner articles of female attire are perhaps as cheap in our own shops, as in those of Paris: but when it comes to the multitude of little elegances that ornament the person, the salon, or the boudoir, in this country, they are either wholly unknown in America, or are only to be obtained by paying treble and quadruple the prices at which they may be had here ...We absolutely want the caste of shopkeepers as it exists in Europe. By shopkeepers, I mean that humble class of traders who are content

8

Ulrich Peter Ritter

with moderate profits, looking forward to little more than a respectable livelihood, and the means of placing their children in situations as comfortable as their own. This is a consequence of the upward tendency of things in a young and vigorous community, in which society has no artificial restrictions, or as few as will at all comport with civilization, and the buoyancy of hope that is its concomitant. The want of the class, notwithstanding, deprives the Americans of many elegancies and some comforts, which would be offered to them at as low rates as they are sold in the countries in which they are made, were it not for the principle of speculative value, which enters into nearly all of our transactions. In Paris the man or woman who sells a duchess an elegant bauble, is half the time content to eat his humble dinner in a small room adjoining his shop, to sleep in an “entresol” over it, and to limit his profits by his wants. The pressure of society reduces him to this level. With us the thing is reversed, and the consumer is highly taxed, as a necessary result. As we become more familiar with the habits of European life, the demand will gradually reduce the value of these minor articles, and we shall obtain them at the same relative prices, as ordinary silks and shawls are now to be had. At present it must be confessed that our shops make but indifferent figures compared with those of London and Paris. I question if the best of them would pass for more than fourth-rate in London, or for more than third-rate here; though the silk-mercers at home might possibly be an exception to the rule.” (JFC 2004, Letter VII) To finalize his comparison he concludes, “The amount of all my experience, on this point, is to convince me, that so long as one is willing to be satisfied with the habits of American life, which include a great abundance, many comforts, and even some few elegancies, that are not known here, such as the general use of carpets, and that of many foreign articles which are excluded from the European markets by the different protective systems, but which, also, do not know a great many embellishments of living that are common all over Europe, he can get along with a good deal less money in New York, than in Paris; certainly, with less, if he mix much with the world.” (JFC 2004, Letter VII) If we reconsider this comparison we must note that the author in his effort to present a correct, reflected, and plausible result used several methodological gimmicks, which only much later became a standard in comparisons of purchasing power. Thus he introduced a standard object of comparison, i.e., the normal American upper

James Fenimore Cooper: An Astute and Critical Precursor of Comparisons in the Social Sciences

9

class New Yorker. He discussed the institutional effects on differences such as tariffs and taxes. He also mentioned the organization of crafts and commerce as an important factor and points out the importance of differences of habits, style, and motivation. All in all I find his comparison quite differentiated, informative, and reflected. III.

The controversy about the costs of republican versus monarchic government

As a second example of Cooper’s expediency in comparisons I want to use what in the literature about him is often referred to as “the public finance controversy”. It is what today we would call a hypothesis testing or rather a hypothesis refuting comparison. In the third letter of his report (JFC 2004, Letter III) Cooper tries to refute the hypothesis proposed in a 70 page article in Volume 5 of the Revue Britannique by an anonymous author argues that the cost of government in France was lower than in the U.S. i.e., as it was understood at the time, that a republican government is more expensive than a monarchy. If we were to look at this comparison with the eyes of a modern social scientist we could probably find many a flaw in it. Still I want to point out that this was no run of the mill comparison, but a rather sophisticated one if one looks at it from the point of view of the methodological refinement. Cooper goes about his refutation in four steps or on four levels. He criticizes his opponent on the basis of the four fundamental methodological questions in the critical analysis and planning of comparisons (Ritter 1996, pp. 311-313; Ritter 1997, pp. 313-361): Who compares why, what and how. In the first place he looks at the author of the comparison, i. e. the person or institution responsible for it. In the third letter of his Gleanings Cooper writes: “A controversy concerning the cost of government, was commenced some time in November last, under the following circumstances, and has but just been concluded. As early as the July preceding, a writer in the employment of the French government produced a laboured article, in which he attempted to show that, head for head, the Americans paid more for the benefits of government than the French. Having the field all to himself, both as to premises and conclusions, this gentleman did not fail to make out a strong case against us; and, as a corollary to this proposition, which was held to be proved, he, and others of his

10

Ulrich Peter Ritter

party, even went so far as to affirm that a republic, in the nature of things, must be a more expensive polity than a monarchy.” (JFC 2004, Letter III) In one sentence, the subject or author of the comparison was a partisan of monarchy and on the government pay role. This, as Cooper calls it, “extravagant assertion” apparently made a deep impression on the discussion in France and even beyond the borders of that country. It struck people, as Cooper says,” by the boldness of the proposition, as well as by the plausibility of the arguments by which it had been maintained.” (JFC 2004, Letter III) The reasons for making this comparison lay in the discussion and political disputes at that time inside and outside the French parliament. As a matter of fact there was a necessity to lower government spending and our anonymous author spends the first 14 of the 70 pages of his article to show in a very detailed form that to lower government spending the cuts and savings expected from an administrative reform and salary cuts were hardly worth the effort (Lettre 1831, pp. 273–287). He then sets out to explain the real purpose of his comparison. It was to prove that those in favour of the reforms were wrong, when they constantly pretended that the government of the U.S. was “Bon Marche”, i. e., cheap. At a time, when the battle between monarchists and democrats or republicans became an acutely pressing issue the article was, of course, very favourably received by the monarchists all over Europe and was seen by them as another proof or their cause. On the other hand it shocked not only Republicans, but also those who where in favour of constitutional or parliamentary monarchy. It was seen as a blow particularly directed at all protagonists of the American form of government, one of the most respected ones of these being the General Lafayette. In a letter dated November 22 1831, the latter points out to Cooper that he discovered many mistakes in the article, but lacked the time for a detailed refutation and would like to pass this task on to the better hands of Cooper. Cooper also looks critically at the object of the comparison and the method used to describe it. As we shall see, Cooper criticizes not only the data which the anonymous author of the offensive comparison uses but also the manner in which he uses it. And this is, where most of his detailed efforts are made.

James Fenimore Cooper: An Astute and Critical Precursor of Comparisons in the Social Sciences

11

It was clear to our unknown author, that before comparing government expenditures of the two countries in total and per head they had to be made comparable by addition, subtraction and estimation. Some of the topics chosen for this end were: - The definition of government spending, the federal budget not including the expenditures of communities and states, which in the centralized government were part of the central budget; - The state debt and its repayment; - Taxes and duties; - Expenditures for wars; - Expenditures of the US president versus those of the president of the council of ministers in France: - Indemnities per day and travel allowances of state senators and house representatives and of the state secretaries; - their expenditures for their offices and for salaries; - The pay of the officers in the two armies; - The expenditures for the churches, the judicial and the educational system and for the infrastructure i. e. canals and roads. Using an astounding number of statistics and figures the author tries to prove, that per capita expenditures for governmental services were not lower but rather higher in the US than in France. In the year 1829 according to our anonymous author they amounted to 31 frs in France and 35 frs in the US. Cooper criticizes these impressive calculations item by item and tries to show that the author used false or incorrect numbers, presented contradicting facts, mixed up cause and effect, misunderstood or misinterpreted the data or even that there must have been errors in print (JFC 2004, Letter III). I cannot go into the details of these critiques here. I just want to give you two examples. Let us take the national guard as the first example. It is here Cooper says (Lettre 1831, p. 165) that the Revue britannique makes its gravest mistakes. Our anonymous author takes much pain in arriving at comparable figures for the two countries by calculating expenditures for food, clothing, training etc. And he arrives at the conclusion that one man in the militia costs around 500 frs. per year in France as against 2 500 frs. or 5 times as much in the US. Cooper points out the faults in this calculation and shows that the national guard is by far less costly in the US than in France. The flaws he mentions are among others, that in New York state time for service is less than

12

Ulrich Peter Ritter

half of what the other side calculated. Also the value of the time supposed by the anonymous author is only one third. As the anonymous author points out members of the national guard have to clothe and arm themselves at their own costs. But they do not have representative functions and therefore do not have to have fancy uniforms. They can borrow their arms from a neighbour, if they do not have them anyway. (Lettre 1831, pp. 165-167) Also these arms are only needed for training purposes. In case of war, arms are provided by the governments of the state and the federation. A second example is the cost of religious services. While these costs are included in the French budget, the state contributes nothing in New York, where the average salary of a minister does not surpass 400 dollars versus the 1000 Dollars our anonymous author has used for his calculation, these coming from the renting of seats in the church and voluntary contributions. (Lettre 1831, pp. 167168) At the end of his calculations and of his article Cooper comes to the conclusion that altogether a citizen of the State of New York pays per year 14 frs and 1 s for government services from the state and the federation, including payments for schools, clergy and the poor, this is quite a difference to the calculation of 35 frs. in the article in the Revue Britannique, the sum amounting to 40 percent of the sum his adversary had calculated. As we have seen, Cooper goes to great lengths to show where the sets of data cannot be compared with each other and tries to indicate which allowances must be made, mainly because of the cultural contexts themselves, which are so different, and he demonstrates how in each culture the players or citizens place different values on various objects or situations. Thus they might have the same amount of good X, if they wanted it, but the culture does not value X highly enough to want much of it. Cooper’s refutation seems to have more or less ended the public policy debate. In his words: “It was then announced that instructions had been sent to America to obtain more authentic information; and we were promised a farther exposure of the weakness of the American system, when the other side should receive this reenforcement to their logic”. But (Footnote 7): “No such exposure has ever been made; and the writer understood, some time before he quitted France, that the information received from America proved to be so unsatisfactory, that the attempt was abandoned”. Cooper goes on to say that the information since published, has confirmed

James Fenimore Cooper: An Astute and Critical Precursor of Comparisons in the Social Sciences

13

the accuracy of his calculations, “the actual returns varying but a few sous a head from his estimates” (JFC 2004, Letter III). Cooper shows in his calculations the low tax burden of the citizens in the US, using New York as an example. But he refuses to compare them to France. In the last sentence of his letter he writes (Lettre 1831, p.182, translation by the author): “In no way do I want to compare these facts with France because I am sincerely convinced of the in aptitude of a foreigner for such investigations.....the example of the Revue Britannique still being too much present in my mind not to force me oo douter de moi meme for such a task”. And on page 177 he speaks of the fact that his critique shows the difficulties and problems of understanding the details of the customs of a foreign nation and the danger of writing about this subject if one has not personally studied them “from the closeness”. This is a more general piece of wisdom that all comparatist should be aware of when we compare even official data: they do not necessarily represent the same reality in one country as in the other. Cooper makes us aware of how important it is to have data sets that are comparable. It is easy to be wooed into comparing today with all the data which is available, but the problem of finding data that allow valid comparisons still exists. Cooper makes us aware of the pitfalls that are masked today by the elegant sets of figures at our disposal and our tendency to neglect cultural and historical contexts that colour the economic behaviour of the parties involved, making them difficult to compare in a straightforward way. IV.

Conclusion

I come to the conclusion that we cannot look at Cooper’s comparisons as models or scientific achievements. What we must realize is that he was a writer and journalist and not a scientist or university professor. Also, we must consider that the letters we are referring to in this paper were written in the 1830s, a time when social science was in its beginnings and knowledge of statistics was not very developed, not to speak of comparative sciences. That such a science could be developed, was unthinkable of at that time. However, what we can conclude from my little aperçu are three things: First, that it could be worthwhile to invest time to go deeper into the theme treated here, i. e., James Fenimore Cooper as a comparatist, by encouraging further research. Second, when the history not

14

Ulrich Peter Ritter

only of comparative economics, but also of comparative sociology and politics, will be written, it would be only natural to require that a chapter or at least a few paragraphs be included on this precursor. But what is most important to me is the third point: I think that international comparisons with their four subdivisions – comparisons of economic phenomena, comparisons of economic policies, comparisons of economic systems and comparisons of dealing with problems (Ritter 1996, pp. 311-313 and Ritter 1997, pp. 313-361) – should be included in the training of economists. Furthermore, I make a plea for having a history of comparisons included in such a curriculum; and I consider it desirable to make use of historical comparisons in the form of cases or case studies in our teaching, because this would add depth and insight to the learning of our students, help to widen their horizon as it did my own.

References

Cooper, James Fenimore (1970a), Gleanings in Europe: England, Robert E. Spiller (ed.), New York: Oxford UP. Cooper, James Fenimore (1970b), Gleanings in Europe: France, Robert E. Spiller (ed.), New York: Oxford UP. Cooper, James Fenimore (1986), Gleanings in Europe: The Rhine, Historical Introduction by Ernst Redekop and Maurice Geracht, Text established and Explanatory Notes prepared by Thomas Philbrick and Maurice Geracht, Albany: State University of New York Press, 1986. Cooper, James Fenimore, Lebensbilder aus Frankreich, den Rheinländern und der Schweiz, Braunschweig 1837, Reprint Kelkheim 2001. Cooper, James Fenimore (1960), The Letters and Journals of Vol. 6, James Franklin Beard (ed.), Cambridge, Harvard UP. JFC 2004: Cooper James Fenimore (2004), A Residence in France With An Excursion Up The Rhine, And A Second Visit To Switzerland, Paris, 1836, The Project Gutenberg EBook Release Date: July 22, 2004 [E Book #12990], : http://www.gutenberg.org/etext/12990 (last visit: 17.12.2007).

James Fenimore Cooper: An Astute and Critical Precursor of Comparisons in the Social Sciences

15

Lettre (1831), Anonymous, Rapprochement entre les dépenses publiques de la France et celles des États-Unis, Revue Britannique, Paris, pp. 272 – 322. Lettre: Lettre au Général Lafayette sur les dépences publiques des États Unis, Revue des Deux Mondes Paris. Meyer, Frommhold Karl (1996), in: Cooper James Fenimore, Der Lederstrumpf, 23 edition, Reutlingen, pp. 651 – 652. Redekop, Ernest (1986), Introduction to Cooper, James Fenimore, Gleanings in Europe: The Rhine, Albany, State University of New York P. Ritter, Ulrich Peter (1996), Some Basic Considerations and Methodological Decisions in Comparative Research, in: Ulrich Peter Ritter (ed.), Problems of Structural Change in the 21st Century, National and Comparative Research from Argentina, Brazil and Germany, Frankfurt am Main, pp. 311 – 320. Ritter, Ulrich Peter (1997), Vergleichende Volkswirtschaftslehre, 2nd Edition, Munich. Walker, Jeffrey (1993), The Importance of Flotsam and Jetsam in Editing the Unpublished Letters of James Fenimore Cooper, in: James D. Wallace (ed.), James Fenimore Cooper: His Country and His Art, Papers from the 1993 Cooper Seminar (No. 9), The State University of New York College at Oneonta, Oneonta, New York, pp. 52-63. Wikipedia, the Free Encyclopedia: James Fenimore Cooper, : http://en.wikipedia.org/wiki/James_Fenimore_Cooper (last visit: 17.12.2007).

Gabriele Tondl

Trade and Integration in Latin America and with Its Main Trading Partners Abstract Latin America (LA) has opted for trade liberalization and opening up for foreign direct investment in the 1990s. The region has implemented a multitude of trade arrangement within the group. Since LA is a large, emerging market, the United States and the European Union have also a strong interest to establish free trade agreements with that region. Several such agreements have been implemented. In line with these agreements the main trading partners, US and EU, could affirm their trading position in the region but intra-LA trade also increased steadily. I. Recent Economic Developments in LA In the 1950, 1960s and 1970s, LA countries practised an economic regime characterized by market intervention and import substitution policies. Countries wished to shelter the development of own industries and to limit foreign influence. They enjoyed a relatively stable economic growth that reached on average 5 per cent during that period. However, the situation changed completely in the 1980s. In view of the excess liquidity in international capital markets after the first oil price shock, LA countries heavily borrowed on international markets. This brought them into a sever debt crisis in the early 1980s when interest rates substantially increased. Many LA countries slipped into serious recession and stagnation in the 1980s which entered as the "lost decade" in their economic history. Consequently, one after the other had to reconsider its economic model starting to foster market-orientation and to integrate into the world economy. The 1990s became the decade of reforms in LA. Reforms were designed in line with the Washington Consensus, a set of guidelines which the international institutions had proposed for LA. They focused on deregulation, privatization, reduction of government

18

Gabriele Tondl

deficits, macroeconomic stabilization and liberalization of trade and direct investment.1 Consequently the region became ready for free trade negotiations with third countries and within the region and started to search actively to attract international investments, which will be discussed in detail below. Growth resumed in the 1990s reaching on average 3.3 per cent, but this was weaker than expected and certainly below the growth in other dynamic emerging markets. Moreover, the region was not immune to crisis. Mexico went into a deep financial crisis in 1994/95 caused by increasing public debt financing and declining oil prices, which lead to a substantial currency devaluation and economic decline. LA was also heavily affected by the Asian crisis in 1998 and suffered in that period a substantial decline in growth, a tightening of financial markets, consequent pressures on currencies and devaluations. Most dramatic was the devaluation of the Brazilian Real in 1999. Only Chile managed to remain almost unaffected by the Asian crisis. (Singh et al. 2005, Corbo et al. 2005) In the course of economic crises and successive attacks on currencies, countries such as Mexico and Brazil moved from a semifixed exchange rate system to floating exchange rates (ECLAC 1999). Argentina originally had adopted a currency board in 1991 with the aim to control hyperinflation. However, the fixed exchange rate brought the economy under stress when its mighty neighbour economy Brazil devalued its currency and world market prices for Argentina´s exports declined. The country entered into a severe recession. International investors raised the risk score of the country inducing massive capital flight. At the height of this financial crisis Argentina had to declare debt default and unfreeze the fixed link with the dollar in early 2002. A massive devaluation of its currency followed. Despite the profound pro-market reforms in LA in the 1990s, growth performance was disappointing. Moreover, poverty and inequality had not reduced. LA constitutes a big, growing market with 559 million inhabitants in 2005. There are important differences in economic development between LA countries. Argentina, Chile and Mexico lead in 1

Chile started economic reforms already in the late 1970s, earlier and more ambitious than the rest of the region. Thus it experienced a better economic performance and did not suffer a recession in the 1980s but grew at 7.3 per cent in the period 1985-97. (Corbo et al. 2005).

Trade and Integration in Latin America and with Its Main Trading Partners

19

per capita income (see Table 1), while countries like Peru and Venezuela reach just half of that income level. According to ECLAC (Economic Commission for Latin America and the Caribbean), LA reached an annual growth of 4.9 per cent in 2007. Growth was higher in South America (5.5 per cent) than in Central America and Mexico (3.6 per cent). Table 1 shows that among the main LA economies, Argentina, Venezuela, Colombia and Chile were heading in growth in the recent period, whereas the growth rate of Brazil and Mexico was slightly below 3 per cent (see Table 1). Macroeconomic stability has also improved in many LA economies. The major economies, Brazil, Mexico, Chile as well as Colombia and Peru have flexible exchange rate systems and inflation target monetary policies that helped to reduce their inflation rates. Argentina could stabilize its currency in 2003. (ECLAC 2006) Real interest rates have dropped and inflation became below 10 per cent, except for Venezuela (see Table 1). Gross fixed capital formation was fairly stable in the region (see Table 1). However, it declined substantially in Venezuela and increased significantly in Colombia. LA countries lead more prudent fiscal policies during the recent economic cycle and therefore could accelerate the reduction of their external debt. (ECLAC 2006; 2007; and Table 1). After the recent economic upswing, practically all major countries in LA show an increase in labour force participation and a decline in poverty (see Table 1). Finally, one has to note that LA countries have become highly open economies, with rapidly growing export rates reaching about 20 per cent in Colombia and Peru and almost 40 per cent in Chile (see Table 1). It is noteworthy that trade balances also turned into surplus in recent years. Particularly oil and mineral exporters like Venezuela and Chile benefited from high world market prices. Since the second half of the 1990s, foreign direct investment rose steeply, with a short drop in 2003 but an immediate gain thereafter. Annual inflows amounted to 2.2- 6 per cent in the major economies (see Table 1), Chile also leading in this area.

20

Gabriele Tondl

Table 1: Basic economic indicators for the major LA economies (to be continued)

Argentina Brazil Chile Colombia Mexico Peru Venezuela

population (million) 2005 1998 38.7 36.0 186.4 166.0 16.3 14.8 44.9 40.8 103.1 95.3 27.9 2.5.2 26.6 23.4 period average 2003-05 1998-2002

Argentina Brazil Chile Colombia Mexico Peru Venezuela

GDP p.c. in PPP (constant 2000 internat. $) 11778.3 11766.0 7346.4 7126.5 10183.1 9148.9 6285.3 5989.1 9351.3 8929.6 5144.8 4712.6 5376.6 5722.5 goss fixed capital formation (% of GDP) 18.6 16.1 17.6 20.0 21.2 22.1 17.8 14.8 19.3 20.5 18.2 20.3 17.4 23.9

Argentina Brazil Chile Colombia Mexico Peru Venezuela

inflation rate 9.2 4.7 9.4 6.1 2.3 3.7 6.0 10.6 4.4 10.7 2.5 3.3 22.9 22.1

Argentina Brazil Chile Colombia Mexico Peru Venezuela

GDP in billion current US-$ 2005 1998 183.1 299.0 882.5 788.0 118.9 73.1 122.9 98.5 767.7 421.0 79.4 56.6 144.8 91.3 period average 2003-05 1998-2002 real GDP growth % 9.0 -3.1 2.6 1.7 5.5 2.5 4.6 0.5 2.8 3.2 5.2 1.7 6.5 -1.5 external debt (% of exports) 402.4 497.6 209.6 383.0 121.8 165.7 180.6 211.5 84.2 101.6 210.0 351.2 103.2 152.9 real interest rate 1.0 16.2 44.5 57.5 0.1 9.8 7.6 13.0 0.9 5.8 10.3 21.0 -8.8 8.1

Trade and Integration in Latin America and with Its Main Trading Partners

21

Table 1 (cont.): Basic economic indicators for the major LA economies period average 2003-05 1998-2002

Argentina Brazil Chile Colombia Mexico Peru Venezuela

labor force participation rate 71.2 68.2 72.4 71.9 58.7 59.6 74.9 72.8 62.2 62.9 72.1 70.0 72.5 67.8

period average 2003-05 1998-2002 poverty rate 20.2 12.3 21.4 22.7 5.6 9.6 17.8 21.7 11.6 23.7 30.6 34.9 40.1 29.2

exports in % GDP trade balance (% of GDP) 24.9 14.1 7.8 2.3 17.1 11.4 4.2 -0.9 39.7 31.0 7.1 1.0 21.5 18.9 0.0 -1.0 29.1 29.4 -1.8 -2.0 21.2 15.2 2.9 -2.7 37.0 25.3 17.6 5.8 FDI inflows % of GDP Argentina 2.2 3.5 Brazil 2.3 4.6 Chile 6.4 6.8 Colombia 4.6 2.6 Mexico 2.5 3.2 Peru 2.7 2.8 Venezuela 2.2 3.3 Source: Own calculations based on Word Development Indicators 2007. Argentina Brazil Chile Colombia Mexico Peru Venezuela

In the next section we shall discuss the present status of LA trade and integration agreements, both within the region and with its major trade partners, the US and the EU. II. Integration and trade agreements of LA countries A. Intra – LA Integration and Trade Agreements Within LA plans for integration and free trade areas have been repeatedly launched since the 1960s. The wish to strengthen its own political identity, to define its own strategy independently of the

22

Gabriele Tondl

political ambitions of its mighty Northern neighbour, the US, and international institutions has been the major motivation for such plans in LA. The first initiative for a Latin American free trade area goes back to the LAFTA (Latin American Free Trade Association) of 1961, an agreement foreseeing the creation of a free trade zone between South American countries and Mexico originally by 1972, then by 1980. In 1981 ALADI (Asociación Latinoamericana de Integración, Latin American Integration Association) succeeded LAFTA. ALADI aims at the creation of a free trade area, but also considers the possibility of subregional free trade agreements (FTA) under its umbrella. Under ALADI several such subregional free trade agreements were concluded: the Andean Community (CAN), MERCOSUR (Mercado Común del Sur) and the Grupo de los Tres (Colombia, Mexico, Venezuela) (see below). The five Central American countries Costa Rica, El Salvador, Guatemala, Honduras and Nicaragua created the CACM (Central American Common Market) already in 1963, establishing a free trade area (Hummer 2005). Due to the political turmoil in the region and the war between El Salvador and Honduras, the organization was ailing in the 1970s and 1980s but saw a revival since 1991 (Hummer 2005). CACM has established free trade in goods except for sugar cane, coffee, alcoholic beverages and petroleum products (SIECA 2007). Each of its members maintains bilateral free trade agreements with Mexico since the period 1995 – 2001 (SIECA 2007). The Andean Community (CAN, Comunidad Andina) was established in 1988 between Colombia, Peru, Venezuela, Ecudador and Bolivia to create a customs union. Venezuela left the Andean Community in 2006 claiming that it could not pursue its goals with that community and moreover it did not agree with the free trade arrangements that Peru and Colombia planned with the US. The Grupo de los Tres was established between Colombia, Mexico and Venezuela in 1995 to establish free trade. The Mercosur entered into force between Argentina, Brazil, Uruguay and Paraguay in 1991. It is a customs union and aims at creating a common market following the model of European integration. Chile and Bolivia are associated members of the Mercosur since 1996 and covered by free trade regimes. Venezuela signed an

Trade and Integration in Latin America and with Its Main Trading Partners

23

accession protocol to Mercosur in 2007 and is expected to become member in 2008. Chile, originally member in the predecessor of the Andean Community, maintains this mentioned association agreement with Mercosur but also a number of bilateral free trade agreements with other LA countries: with Venezuela (1993), Colombia (1993), Ecuador (1994), Peru (1998) and Mexico (1998) (Hummer 2005). There are several initiatives between LA states to promote intraLA political and economic integration. In 1994, 27 Central- and South American states proclaimed with the declaration of Quito the creation of a free trade area comprising all LA, the LAFTZ (Latin American Free Trade Zone) (Hummer 2005). There are also several projects to promote integration within South America. In 2004, the South American States proclaimed the CSN (Comunidad Sudamericana de Naciones) aiming to create a political, social and economic integration area between South American states. Also in 2004, Venezuela launched ALBA, the Boliviarian Alternative for Latin America, that wishes to establish an integration area characterized by strong social policies, heavy state intervention and renationalization among LA, opposed to the projected FTAA launched by the US (see below). A major element of this initiative is the integration of Latin American oil markets under the leadership of the region´s most important oil producer Venezuela that wishes to diversify away from its primary export destination for oil, the US. At present, Cuba and Bolivia joined Venezuela under this initiative. The principles of this integration project which are centred on state intervention and state ownership stand in sharp contrast to the other integration concepts for LA which are based on economic deregulation and liberalisation. The initiative of Venezuela is likely to separate again the subcontinent and delay intra-LA integration. LA has entered and negotiated trade agreements with the US and the EU (see below). Mexico, Peru and Chile are also involved in the APEC (Asia-Pacific Economic Cooperation) which includes the US, Australia, New Zealand, China and Japan among others and aims to achieve a free trade area by 2010.

o H

(N

Trade and Integration in Latin America and with Its Main Trading Partners

25

B. Free trade agreements LA – US The US has always been interested to assure its political hegemony in LA and to improve political stability in the region, above all in its backyard Central America and the Caribbean Basin. Consequently, these countries show a strong presence of US investors and have mainly exported to the US markets, even before the agreement of free trade areas. In 1993, the US launched an initiative for a regular summit of the 34 states of the Americas (Summit of the Americas). The aim of these summits is the cooperation in the fields of democratization, human rights, social policies, environmental protection, the fight against terrorism, drugs and corruption, and trade liberalization. Since 1994, Mexico is part of NAFTA (North American Free Trade Agreement) which has manifested in a strong engagement of US-investors in outsourcing in Mexico and an extremely high share of Mexico´s exports to the US (over 80 per cent). The NAFTA agreement relates exclusively to trade liberalization. However, cooperation in environment policy and development in labour standards was agreed in two separate agreements. In 2004, the US and the other Central American countries signed the free trade zone CAFTA (Central American Free Trade Agreement). (Hummer 2005) Both agreements guarantee the dominance of the US as major trading partner. The US launched an initiative for the FTAA (Free Trade Area of the Americas) in 2003 which should comprise 34 countries from the Americas, however, many LA countries show reluctance to enter into such arrangements. Many LA countries, above all Brazil, Argentina and Venezuela, fear that the US would dominate such an agreement and that they should foster intra-LA integration. Consequently, progress of the FTAA repeatedly was weakened. The summit of the American States in Argentina in 2005 was accompanied by heavy protests against the FTAA. Consequently, the US attempted to enter into bilateral trade negotiations with several priority LA countries. Shortly after the EUChile trade agreement, the US signed a free trade agreement with Chile in 2004. For the US Chile is an important market for machinery and transport equipment, but also for some agricultural products. The agreement foresees unrestricted market entry for manufactured and agricultural goods and all kind of services, in-

26

Gabriele Tondl

vestment and government procurement. (Office of the United States Trade Representative 2002) The US started also free trade negotiations with Peru, Colombia and Ecuador in 2004 with a wish to secure stability and better prosperity in the region. Free trade agreements were signed with Peru and Colombia in 2006, supposed to enter into force in 2008. They foresee a gradual elimination of access barriers for industrial products (10 years) and agricultural products (15 years) in the case of Peru and an immediate access in the case of Colombia, access to the service sector and free access and protection of investors (Office of the US Trade Representative 2007a, 2007b). The US agreements require the parties to enforce national environmental and labour regulations and compliance with fundamental ILO labour rights. Critics stress that the FTA would yield more satisfactory results for the US than for the LA countries. For example, in the case of Mexico the NAFTA agreement was blamed to have caused a sharp rise in corn prices harming the consumers. Others argued that too little capacity and competence was devoted to negotiations by the LA governments and that the agreements would mainly benefit the exporting sector without improving employment and reducing inequality. The US favours to include Latin America in a common free trade area, i.e. to establish the FTAA, and watches the creation of intra-LA agreements with reservation since that would weaken its political influence in the region and strengthen intra-LA political relations. C. Free trade agreements EU - LA The European Union member countries, particularly its member countries Spain and Portugal, have long historical and cultural ties with LA. Therefore, the EU has increasingly searched to institutionalize the political and economic dialogue with the region since the 1990s. The increasing economic influence of the US in LA and the conclusion of the US free trade arrangements with single countries and groups of countries, such as the NAFTA, has been observed by the EU with much attention. Being aware that it would risk to lose its role in the region as the US moves on to establish FTA, the EU has also searched to intensify political and economic relations with the region. On the economic level, LA constitutes a large and growing market for the EU offering high market potential to export its high

Trade and Integration in Latin America and with Its Main Trading Partners

27

technology products and services and a market that lends itself also to increase presence through foreign direct investment, more and more in the privatizing service sector. Despite that fact, trade with LA accounted for only 0.5 per cent of total EU trade in 2005 (UN Comtrade). LA countries expect to reduce the predominance of trade relations with the US when entering into free trade agreements with the EU. Diversification of trade destinations and FDI ownership is an important goal for LA. Given the trade deficit with the EU in the 1990s, LA has become increasingly concerned to gain market access for its agricultural products in the still highly protected EU markets. Trade with the EU accounts for around 15 per cent of LA trade (UN Comtrade). An institutionalized political dialogue between the EU and the Latin American and Caribbean countries takes place since the Rio summit in 1999 every other year. At a lower level, the EU maintains a special dialogue with Mercosur, the Andean Community, Central America, Mexico and Chile. (European Commission, DG External Trade 2006a; 2006b; 2004) Cooperation programmes of the EU cover the regions as a whole, its subregions and countries. These programmes encompass the fields: social support and enforcement of labour regulations, promotion of regional cooperation and integration in LA, reinforcement of human rights, democracy, good governance and prevention of conflict, cooperation in the field of higher education (ALFA programme), and environmental protection. The EU wishes to enforce an advance in political rights and stability in LA, therefore the agreements foresee that a violation of the commitments of LA in this area would lead to the suspension of the trade agreements. The cooperation agreements have been in force with the Andean Community since 1993, the Central American republics since 1999, the Mercosur since 1999 (in addition bilateral cooperations with the single Mercosur countries have been operated since 1991- 1995), and Mexico in 2000. With Chile an association agreement has been enacted in 2003 which includes also the trade agreements. At bilateral level, one can name the EU–Brazil cooperation agreement established in 1992 as an example. It covers higher education issues and foresees scholarships under the Erasmus Mundi programme and the establishment of European Studies Centers,

28

Gabriele Tondl

social programmes and environmental support to fight deforestation (European Commission, DG External Relations 2007). The Mercosur countries constitute the major trade relation of the EU with LA. For the Mercosur countries the EU is their principal market for agricultural exports, while the EU sells manufactured goods on Mercosur markets. The EU–Mercosur FTA was intended to go beyond the WTO and considered as a single undertaking. The EU has a strong interest to gain unrestricted market access for manufactured products and in the service sector while the Mercosur countries wish to have free access for agricultural products. Both parties still watch important restrictions under the current WTO regime. Since the conflicting interests of the two parties in the WTO negotiations of the Doha Round – Mercosur requests a larger cut in agricultural subsidies then the EU is willing to accept, while the EU demands substantial cuts in tariffs for industrial products and access in services – the negotiations for an EU-Mercosur FTA have come to a standstill. Moreover, the bargaining power of the Mercosur has been repeatedly weakened by political tensions between its member states. The accession of Venezuela to the Mercosur in 2008 brings a new member with a distinctly different political ideology into the group and will once again threaten the political decision power of Mercosur. The planned agreement of the EU with Mercosur envisages to achieve an opening of public procurement, agreements on wines and phytosanitary measures, regulations on investment as well as a dispute settlement mechanism. Given the standstill of the EU-Mercosur negotiations, the EU has searched to enter into bilateral negotiations within the group. Its most important trading partner in Mercosur is Brazil, where it exports machinery and transport equipment. In these sectors relatively high tariffs are still in effect. In addition, the EU is a principal investor in telecommunications, energy, financial services, automotive and agri-food industries. In contrast, the EU imports mainly primary products from Brazil. (European Commission, DG External Trade 2007) For the Andean Community (Colombia, Peru, Ecuador, Bolivia, Venezuela until its exit in 2006) the EU is the second important trading partner after the US and most important FDI source. Countries of the Andean Community presently trade with the EU under the GSP, in addition those combating drug production have duty free access to EU markets in products covered by the GSP and a number of sensitive and even agricultural products. The applica-

Trade and Integration in Latin America and with Its Main Trading Partners

29

bility of the GSP is subject to the implementation of the main international conventions on human, social and environmental regulations. In 2003, plans for a free trade agreement between the EU and the CAN were launched. The start was linked to the attainment of a sufficient level of intra-regional integration within CAN. (European Commission, DG External Trade 2006b) In contrast to this planned trade agreements, the EU maintains two important bilateral trade agreements with Mexico and Chile. The EU-Mexico association agreement includes three pillars, political dialogue, cooperation and trade. It entered into force in 2000. Considering the predominance of the US as a principal trading partner, the EU is subordinate for Mexico. No more than 4 per cent of its exports go to the EU and 10 per cent of imports originate from the EU. In contrast to the trade product structure with other LA countries, EU-Mexican trade comprises in both directions mainly machinery, transport equipment and chemical products. The EU has become an important investor in Mexico, holding 23 per cent of all foreign owned companies in 2003, mainly in the service sector and car production. The EU-Mexico FTA - the first transatlantic agreement of the EU - covers liberalization of trade in manufactured goods, agricultural products (by 2010) and services, and investment (European Commission, DG External Trade 2006c). All these areas are highly important for the EU that wishes to participate in the large Mexican market and to benefit from the NAFTA agreement by investing in Mexico. The association agreement with Chile was implemented in 2003. It covers trade in manufactured and agricultural goods, services, government procurement and investment, going well beyond the WTO commitments. It includes sections on technical regulations and phytosanitary measures. However, the association agreement comprises not only this trade agreement but also agreements on political dialogue and cooperation. The latter includes specific initiatives of the EU to promote education and social support in Chile. The EU imports mining and agricultural products from Chile and exports mainly consumer goods and machinery there. (European Commission, DG External Trade 2006d) Trade of the EU with the CACM is covered by the GSP and the drug regime. While trade with the EU accounts for around 10 per cent of trade for the Central American countries, it is negligible for the EU.

30

Gabriele Tondl

III.

Free trade agreements and trade partners of LA

For historical reasons and promoted by the recent free trade agreements described above, Latin America has developed important trade relations with the US but also more and more within the subcontinent and with the EU. Given their natural abundances, LA countries have a high share of agricultural and mineral products in their exports: Argentina is a major exporter of beef and soybean, Brazil of coffee, meat and sugar cane, Chile exports fruits, wine and fish such as salmon, the Central American countries export pineapples, bananas and coffee. These agricultural products dominate in their exports to the US and the EU. Mexico and Venezuela are major oil producers and Chile is the world´s major copper producer. Oil producers possess also a key product to develop intra-LA trade. However, some Latin American countries have developed a diversified economic structure and export also more advanced manufactures: For example, Brazil exports motor vehicles and pharmaceuticals, Mexico exports machinery and components, to a large extent products of its maquilladora industries, and Costa Rica produces electronic circuits (SIECA 2007). Those manufactures lend themselves to develop export relations with third countries but also within LA. The US is the most important export partner for Central America, the Andean Community and Venezuela, where 30-50 per cent of exports go to the US (see Table 2). In the case of Mexico the exports to the US reach an extreme share of more than 80 per cent. Venezuela and Ecuador are also very focused in their exports on the US market with around 50 per cent of exports. For almost all countries in this group exporting to the US has grown in importance, a development favoured by the free trade arrangements of NAFTA and CAFTA. A remarkable exception is Bolivia which substantially reduced its exports to the US and also to the EU in favour of intra-LA trade. In Central America the EU closely follows the US as export partner but has lost to some extent. However, intra-LA trade relations have well developed and have become even most important for countries such as Guatemala and El Salvador (see Table 2 and SIECA 2007). Exports to South America are negligible but CACM receives 5-6 per cent of its imports from Mercosur and CAN (2007b). The developments in the export structure of Central

Trade and Integration in Latin America and with Its Main Trading Partners

31

America are closely linked with the strengthening of CACM after 1991, bilateral FTA with Mexico and CAFTA. In the Andean Community, the EU is a much less important export partner than the US and has also lost in importance. These developments will be influenced by the implementation of bilateral trade agreements with the US and further trade agreements with the EU which are planned as mentioned above. The only country in CAN which exports equally to the US, the EU and LA is Peru. As already mentioned, the reorientation of Bolivia´s export towards its intra-LA trading partners is most noteworthy. With Venezuela one observes a decline of intra-LA exports which is due to an export decline to Colombia after Venezuela left CAN in 2006. In turn, Venezuela´s exports to the Netherland Antilles, Cuba (the two are not included in the group LA) and China increased. Table 2: Latin America´s export partners (to be continued)

year Central America, Mexico Mexico 95 06 Honduras 95 06 Nicaragua 95 06 Costa Rica 95 06 Guatemala 95 06 El Salvador 95 06

export partners (per cent of total exports) US EU Intra-LA

83.4 84.8 42.6 32.5 41.9 46.0 40.1 42.4 31.3 31.4 17.5 28.2

4.2 4.3 36.4 25.9 32.4 21.6 30.7 17.8 15.7 7.7 30.9 11.9

4.5 4.3 6.4 31.1 20.1 21.1 16.3 11.5 36.4 40.7 44.9 43.0

32

Gabriele Tondl

Table 2 (continued): Latin America´s export partners

year Andean Community Ecuador 95 06 Colombia 95 06 Bolivia 95 06 Peru 95 06 Venezuela (left CAN in 2006) Mercosur Brazil Argentina Paraguay Uruguay

export partners (per cent of total exports) US EU Intra-LA 43.2 53.7 35.6 40.8 31.5 10.0 18.7 21.9

17.9 11.5 24.1 15.4 18.8 6.0 27.7 23.0

18.6 24.5 25.6 26.7 40.2 66.4 18.3 21.7

95

50.8

8.5

22.7

06

48.8

8.0

6.6

95 06 95 06 95 06 95 06

18.8 17.6 8.6 8.8 4.8 3.5 6.0 13.8

27.9 21.5 21.4 17.0 19.4 5.9 20.6 16.9

22.1 23.6 46.3 41.1 64.9 59.2 53.3 35.2

Chile 95 12.6 26.1 (Mercosur associate) 06 15.5 26.5 Source: Own calculations based on UN Comtrade.

19.5 16.4

For the rest of South America, namely for the Mercosur and Chile, the EU is a more important export partner than the US. Argentina, Paraguay and Uruguay have a very high share of intra-LA exports which relates to their membership in Mercosur. Recent small declines in intra-LA trade are related to some re-implementation of trade barriers in Mercosur. Bolivia´s high intra-LA trade can also be explained by its association with Mercosur since 1996.

Trade and Integration in Latin America and with Its Main Trading Partners

33

In LA trade relations are generally fairly concentrated on one single trading partner, e.g. the US in Central America, the Mercosur with Argentina. However, the most advanced LA economies Chile and Brazil managed to diversify their export destinations. They focus on EU trade but are almost equally engaged in intra-LA trade and trade with the US. For the European Union, trade with LA plays still a subordinate role accounting for just 0.46 per cent of its exports. As mentioned above, LA would offer a high market potential for the EU having reached a stable growth performance and a relatively high level of development in some countries. It should therefore be considered as a more attractive market than presently done. In contrast, LA become much more important as an export partner for the US. In 1990, the US exported 11 per cent of its goods to LA, in 2005 this share has climbed to 19 per cent. It should be noted that it is not only the US neighbour Mexico which accounts for this high share, but other LA countries as well. (In 1990, 7 per cent of the US exports went to Mexico, in 2005 13 per cent.) (UN Comtrade) Although trade between LA and the US is favoured by their closer geographical location, there is good reason to suppose that the EU has too much neglected a potential market in LA. In summary, as shown in Prüfer and Tondl (2008), increasing trade openness had a positive effect on LA growth performance. Trading acted as an important channel for an increase in productivity, either because exporters are forced to improve their competitiveness on markets or because imported products introduce new technologies in the countries. IV.

Foreign direct investment in Latin America

In the process of economic reforms and liberalization, LA countries increasingly considered the attraction of foreign direct investment as a key strategy to promote development and growth. Consequently, the stock of foreign direct investment rose steeply since the second half of the 1990s and peaked for example, 84 per cent of GDP in Bolivia and 74 per cent in Chile (see Figure 1). FDI from North America and the EU dominates in the major LA economies (see Figure 2). In the smaller LA countries there is also a significant share of intra-LA FDI. The representation of the EU and the US closely follow their role as a trading partners that was discussed above. Thus we observe that the EU has become the most important investor in most South American countries, such as

34

Gabriele Tondl

Argentina, Bolivia, Chile, Colombia, Ecuador, Peru, Paraguay and slightly in Brazil, whereas North America is traditionally the most important investor in Mexico, Central America and Venezuela. Favoured by the free trade regime of NAFTA and CAFTA, parts of the production chain were relocated from the US to Mexico and Central America to benefit from lower labour costs. Thus a lot of greenfield investment in Mexico and Central America went into the machinery, motor and electronic industry. European investment in LA is predominately market seeking, horizontal FDI, e.g. the automotive investment in Argentina and Brazil. In addition, European companies used privatization of public utilities in LA to invest in the service industry, such as telecommunication. There has also been a rising investment of European banks in South America, in particularly Spanish banks entered into the market. Foreign direct investment had a positive impact on growth rates in LA when accompagnied by political stability and sound legal systems. (Prüfer and Tondl 2008). Foreign owned companies often have higher technological standards and are more productive than local ones. However, they also foster competition in their host countries and thus force local companies to become more productive. Figure 1: Development of FDI stocks in LA countries (share of GDP) 0.9

0.8

ARG BOL BRA CHL COL CRI ECU SLV GTM HND MEX NIC PRY PER URU VEN

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0 90

91

92

93

94

95

96

97

98

99

Source: Own calculations based on Unctad data.

00

01

02

03

Trade and Integration in Latin America and with Its Main Trading Partners

35

Figure 2: EU- vs. North American FDI stock in LA countries in 2003 (share of GDP) 0,3 0,25 0,2 EUR-FDI

0,15

NA-FDI

0,1 0,05

SLV

CRI

GTM

PRY

HND

URU

BRA

M EX

VEN

ECU

NIC

COL

PER

CHL

ARG

BOL

0

Source: Prüfer and Tondl (2008).

V. Conclusions In the 1990s, LA countries became open economies. They actively searched liberalization of trade and opened to foreign direct investment. LA countries concluded a number of free trade agreements with their main external trading partners, the US and the EU. Given the large market potential in LA, both world trading powers competed in negotiating trade agreements with LA. However, LA was also very active in paving the way to intra-LA integration aiming to reduce the predominance of the US in trade relations. Mercosur, the Andean Community and CACM are free trade areas which are well established and which have impressively fostered intra-LA trade. In a number of LA countries, such as Argentina and Bolivia, intra-LA trade has now become most important. Other major LA economies like Brazil and Chile managed well to diversify their export markets. The implementation of FTA also promoted FDI flows into LA. Thus Mexico and Central America received important FDI from US companies which aimed to benefit from lower labour costs in their host economy and the easy trading of components within NAFTA or CAFTA. The motives of EU companies are different. For them access to large markets is the main reason for investing in LA. EULA trade agreements also facilitated that investment.

36

Gabriele Tondl

In summary, one can conclude that becoming more open economies has benefited the economic performance of LA countries.

References

Corbo, Vittorio, Hernández, Leonardo, Parro, Fernando (2005), Institutions, economic policies and growth: Lessons from the Chilean experience, Central Bank of Chile Working Papers no. 317, Santiago. ECLAC-Economic Commission for Latin America and the Caribbean (1999), Economic Survey of Latin America and the Caribbean 1998-1999, Santiago, Chile. ECLAC-Economic Commission for Latin America and the Caribbean (2006), Economic Survey of Latin America and the Caribbean 2005-2006, Santiago, Chile. ECLAC-Economic Commission for Latin America and the Caribbean (2007), Economic Survey of Latin America and the Caribbean 2006-2007, Santiago, Chile. European Commission, DG External Trade (2004), Bilateral Trade Relations Central America, http://ec.europa.eu/trade/issues/ bilateral/regions/central_america/index_en.htm, last visit 02/02/2008. European Commission, DG External Trade (2006a), Bilateral Trade Relations Latin America and the Caribbean, : http://ec.europa.eu/trade/issues/bilateral/regions/lac/index_en.ht m (last visit 02/02/2008). European Commission, DG External Trade (2006b), Bilateral Trade Relations Andean Community, : http://ec.europa.eu/trade/ issues/bilateral/regions/andean/index_en.htm (last visit 02/02/2008). European Commission, DG External Trade (2006c), Bilateral Trade Relations Mexico, : http://ec.europa.eu/trade/issues/ bilateral/countries/mexico/index_en.htm (last visit 17/01/2008).

Trade and Integration in Latin America and with Its Main Trading Partners

37

European Commission – DG External Trade (2006d), Bilateral Trade Relations Chile, : http://ec.europa.eu/trade/issues/ bilateral/countries/chile/index_en.htm (last visit 02/02/2008). European Commission, DG External Relations (2007), Brazil, Country Strategy Paper 2007-2013, : http://ec.europa.eu/ external_relations/brazil/csp/index.htm (last visit 02/02/2008). European Commission, DG External Trade (2007), Bilateral Trade Relations Brazil, : http://ec.europa.eu/trade/issues/bilateral/ countries/brazil/index_en.htm (last visit 16/01/2008). Hummer, Waldemar (2005), Integration in Lateinamerika und in der Karibik. Aktueller Stand und zukünftige Entwicklungen, Verfassung und Recht in Übersee 1/2005, pp. 6. Office of the United States Trade Representative (2002), Free Trade with Chile, : http://www.ustr.gov/Document_Library/ Fact_Sheets (last visit 16/01/2008). Office of the US Trade Representative (2007a), Colombia Free Trade Agreement, : http://www.ustr.gov/Trade_Agreements/ Bilateral/Colombia_FTA/Section_Index.html (last visit 16/01/2008). Office of the US Trade Representative (2007b), Peru Trade Promotion Agreement, : http://www.ustr.gov/Trade_Agreements/ Bilateral/Peru_TPA/Section_Index.html (last visit 16/01/2008). Prüfer, Patricia, Tondl, Gabriele (2008), The FDI-growth nexus in Latin America: The role of source countries and local conditions, CentER WP, Tilburg (forthcoming). SIECA – Secretaria de Integracion Economica Central Americana (2007), State of the Current Central American Economic Situation, Guatemala. Singh, Anoop, Belaisch, Agnes, Collyns, Charles, De Masi, Paula, Krieger, Reva, Meredith, Guy, Rennhack, Robert (2005), Stabilization and Reform in Latin America: A Macroeconomic Perspective on the Experience since the Early 1990s, IMF Occasional Paper no. 238, Washington. UN Comtrade, Commodity Trade Database, United Nations. Unctad, Foreign Direct Investment Statistics, Geneva. World Bank, World Development Indicators 2007, Washington.

Part 1: Open Economy Macroeconomics

Sergio V. Barone and Alberto M. Díaz Cafferata

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina Abstract This paper emphasizes the restriction posed by exports performance as determinant of the perceived capacity to fulfill external debt obligations. A long-run implication is that a degree of openness, determined by structural factors of an economy, sets admissible levels of external debt in foreign currency, and breaking this solvency condition causes CA reversals. Empirical results for Argentina 1949-2004 show: The debt burden exhibits a rising trend throughout the period. CA reversals are preceded in seven out of eight cases by a negative between required and expected exports. Argentina suffers continuous solvency problems after the debt crisis of the early 1980s. Similar results are found when filtered exports series are used. I. Introduction A. Export performance, external balance and crises in Argentine, a long-run historical perspective Why cannot Argentina sustain current account deficits if other countries can? By focusing attention on the role of exports and on a long-term horizon this paper brings to the forefront the influence of structural export capacity in setting a limit for "safe" borrowing levels, as far as it carries information about the ability of this economy to regularly serve debt payments in external currency in the future. We argue that exports may be the key variable in the perception of external solvency. Operating through the interdependencies between real and financial processes for a borrowing country, the external financial weakness of Argentina has been related to the unsatisfactory long-run performance of exports. We explore also

We thank Elena Resk and Emanuel López from the IEF for active and valuable research assistance.

42

Sergio V. Barone and Alberto M. Díaz Cafferata

whether the relationship between exports and external debt may serve as indicator of alert, by pointing out when the stock of debt (or current account deficits) become “excessive”. Our contribution lies in the emphasis placed on the links between exports and external solvency, and the application to Argentina in a historical perspective.1 Structural, slowly changing features of the economy drive the long-run evolution of exports that determine through the solvency condition an "admissible level" of the long-run debt, a restriction which is only broken at the cost of an external crisis. In a nutshell, we argue that a historically rigid exports/GDP ratio is determined by structural factors, and that in consequence the past carries information about future exports performance. Therefore, structural rigidities in exports provide a threshold on agents beliefs regarding the future capacity to pay. B. External crises: multiple causes and weaknesses in the strategy concerning exports and solvency Analysts agree that multiple domestic and external problems interacted to cause the Argentine crisis in 2001-2002. Domingo Cavallo (2003) mentions three main sources of problems since 1999: limited bank credit for the private sector, the Real´s devaluation at a time when Brazil accounted for almost one third of Argentina’s foreign trade, and the over-appreciation of the dollar-pegged peso. A consensus exists that the convertibility peso-dollar was a two-edged sword that helped achieve price stability but also, in combination with fiscal deficits caused real exchange overvaluation and damaged the strategy of exports expansion. Regarding the role of multiple causes, references in the recent literature are Domínguez and Tesar (2005), Bleaney (2004), Izquierdo (2002), Calvo (2005), and Mussa (2002). Kaminsky et al. (1998) provide results of several empirical studies on currency crises. An additional ingredient with policy implications that is worth noticing here, is the peculiar mix of errors and bad luck in the process. Some authors have pointed out policy errors in the evolution of the crisis, but it has been recognized that there was also bad luck in the timing of unexpected unfavorable events in the world economy, complicating otherwise more manageable difficulties. 1

Under the import substitution growth strategy until the mid 1970s, followed by the time of globalization of financial flows.

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

43

Regarding the critical Current Account (CA) reversal, Reinhart (2002) argues that the surge of capital inflows to emerging economies was encouraged by the sustained decline of interest rates in the industrial world and the direction reversed with the tightening of monetary policy in the USA. She points out that "it certainly seems a mystery why these wide swings in capital flows recur", and that "even the best policy mix cannot altogether avoid the eventual reverse of capital". There was also a negative impact of the devaluation by Argentina’s major trading partners which reduced profitability in the tradeable sector; the Russian crisis of August 1998 as well contributed to the unexpected halt in capital flows to emerging markets (Izquierdo, 2002). This paper starts from the interpretation that this type of attention to the short-run may hide the presence of factors belonging to more stable characteristics of the economy that should in consequence be examined in a larger time horizon.2 Therefore, to broaden the perspective it is advisable to adopt a long-run approach and to focus on the structural exports performance. C. Historical evolution of the current account and long-run motives to borrow Let’s provide some relevant background information and interpretation of the Argentine long-run current account (CA) since 1935 to our days. Díaz Cafferata et al. (2005), state the following stylized facts: a) the mean CA is close to zero; b) there are no persistent deficits: changes of sign are found every 3 years (3.35 on average); c) recurrent external crises and frequent reversals of the current account occur every 7.7 years between 1935 and 2004.3

2

3

For example, Reinhart et al. (2003) discusses the influence of reputation built on a country’s history as a factor in the development of an external crisis. A reduction in the CA deficit of at least 2.48 per cent of GDP is called a “reversal”. This percentage was calculated by taking the average of the CA deficits in the whole period. The first reversal was in 1939 and the last one in 2002.

44

Sergio V. Barone and Alberto M. Díaz Cafferata

Table 1: Argentina. Reversions of the Current Account (Changes t 2.48 in percent ) year of reversion 1939 1950 1953 1963 1973 1976 1988 1995 2002

' CA%

-4.51 -2.67 -6.64 -4.07 -3.02 -3.61 -2.82 -2.51 -7.67

' GDP%

before 2.05 -0.05

0.00

-1.98

2.91 -0.30 0.32

after

5.59

9.73

2.36

3.41 -0.67

1.40 -7.71

1.47 -4.42 6.81

8.93

Note: * Average of CA in deficit years. ** In percent average change in GDP in Argentine pesos; the first row refers to the two years before the reversal, the second row to the two years after the reversal. Source: Díaz Cafferata et al. (2005).

This is indeed a puzzling combination of facts, because frequent external crises have happened despite the occurrence of relatively small CA deficits. The authors point out that these stylized facts are not consistent with an expected path under the “stages of the balance of payments” hypothesis, such that the sign of the current account would be influenced by the degree of development: if external savings were used in early stages of development to finance growth, CA deficits would be observed chronologically early during several periods, followed later, after a change of sign, by sustained surpluses when the debt is being paid. Rational motives for international borrowing exist, to smooth consumption or to accelerate growth. Fulfillment of the intertemporal budget constraint would generate a long-run path with early persistent deficits of the CA, a change in sign, and persistent surpluses when the debt is paid.4 Despite the plausibility of the stages hypothesis, there is scant evidence of its occurrence in the international experience;5 numerous empirical studies after Feldstein and Horioka (1980) have found a high correlation between investment and domestic savings, suggesting a reduced contribution of external savings to finance investment. Rather, persistent deficits 4

5

It is possible in theory for a growing economy to run perpetual deficits. Cfr. Obstfeld and Rogoff (1997). On the stages hypothesis and empirical evidence cfr. Halevi (1971); Kindelberger (1968); Calderón et al. (1999); Debelle and Faruqee (1996); Díaz Cafferata et al. (2005).

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

45

are more frequently seen as a signal of danger and are not commonly found. However the hypothesis of stages is not rejected in several cases; a few countries can be mentioned such as Canada, Korea, Ireland, South Korea and Australia among others, that have been able to sustain persistent CA deficits (Milesi-Ferretti and Razin, 1996; Obstfeld and Rogoff, 1997). An interesting point is that the mere accumulation of debt along extended periods does not seem to be necessarily a problem for international lenders. D. Synthesis and scheme of the paper A question is posed about the recurrent external crises in Argentina focusing the explanation on the role of the long-run performance of exports. As shown in Figure 1, during 1949-1976 the exports to GDP ratio is continuously below 10 per cent. Two episodes of trade liberalization after 1976 and in the nineties were accompanied by a rise in the local trend. The working hypothesis is that along these decades the structural exports performance sets admissible levels of external debt through the intertemporal budget constraint of the economy. In this sense, a structural rigidity in the exports to GDP ratio is consistent, ceteris paribus, with an also rigid debt to exports ratio and with small and transitory CA deficits as the casual observation suggests. We will explore if external crises in Argentina are empirically related to the evolution of exports, based on the framework of the intertemporal approach to the CA and the intertemporal budget constraint that provides the specific values of exports, or of the trade balance, consistent with external solvency. In synthesis, we are interested in both the relevant mechanisms linking exports and solvency and to identify empirical magnitudes, helping understand in which theoretical cases, and when in practice, export is the relevant variable among the determinants of solvency, and among the predictors of CA reversal, in terms of the long-run perspective. Are the sustainable CA deficits related to some specific structural characteristics of the economy? More punctually, is there a structural degree of openness that can be related to the size of long-run sustainable CA deficits? Is the exports to GDP ratio a relevant variable regarding the sustainability of CA deficit?

46

Sergio V. Barone and Alberto M. Díaz Cafferata

Figure 1: Argentina. Exports to GDP ratio. Current prices in pesos. Mean and mean plus standard deviation, 1884-2005. 0,35

Max 1919 = 33,78%

0,3

Mean + SD

Mean St. Dev. 1949-2001 8,75% 1,94% 1949-20056 10% 4,96%

0,25

Mean

0,2

Mean - SD

Mean + SD

0,15

Mean 0,1

Mean - SD

0,05

0 1884

1894

1904

1914

1924

1934

1944

1954

1964

1974

1984

1994

2004

Source: Díaz Cafferata and Fornero (2003), Statistical Annex; MECON.

In the rest of this paper, first a succinct exposition of the longrun external solvency, and of the role of exports in the solvency condition is provided in Section II. Section III, discusses how to move from the formal expression to the empirical application and the peculiarities of the Argentine case. Section IV reports empirical exercises to establish relationships between exports, the burden of the debt, and the occurrence of CA reversals. The paper concludes in section V with a summary and policy implications.

6

Since GDP in the denominator includes tradeables and non-tradeables, the exports ratio rises with the devaluation. On January 11th 2002 the exchange rate jumped from 1.0 Peso/U$ during convertibility to 1.7 Peso/U$; as of February 2007 the nominal exchange rate is 3.1 Peso/U$.

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

47

II. External solvency: the role and relevance of long term export performance The frequency and costs of episodes of crisis and default in several countries along the 1980s and 1990s, which constitute the dark side of financial globalization, have caused concern and stimulated theoretical and empirical research to understand the evolution of the crises and extract policy indications for the future. There are different approaches to the issues. One area of intense research has developed with specific interest on currency crises, on which the seminal paper of Krugman (1979) addressed the question of the policy problems to keep a fixed exchange rate. Kaminsky et al. (1998) provide a useful revision of the empirical evidence after the turmoil of the second half of the 1990s and examine the possibility to find leading indicators of the crises. Attention is placed on the exchange market evolutions, the nominal exchange rate, liquidity problems, and short-run causes. Alternatively if emphasis is placed on the CA rather than on the exchange market, the relevant price is the rate of interest and the critical problem is the sustainability of CA deficits. The intertemporal optimization and the dynamic CA is the usual setup in the literature to analyze the behavior of the stock of international debt and to discuss under which conditions a country’s intertemporal budget constraint is satisfied. One issue in this framework is the magnitude of trade surpluses the economy must generate to pay, i.e. the “burden” of the debt for the economy (Obstfeld and Rogoff, 1997; Milesi-Ferretti and Razin, 1996, 1998). An alternative way to pose the question is to ask when the CA deficits become “excessive” (Milesi- Ferretti and Razin, 1996; Lanteri, 2008; Saksonovs, 2006). As regards this matter it has been argued that the difference between countries that are able to run persistent deficits in contrast with other countries that fall in crisis is whether the economy is able to generate enough trade surpluses, i.e. whose present value is enough to repay the debt in the future.7 Milesi-Ferretti and Razin 7

Let us remark at this point the constraint for a country of being unable to issue international bonds in their own currency, named in the literature the "original sin". This economy is in consequence subject to two budget constraints, because additional to required taxing, debt payments need to be done in external currency, such that exports in relation to the external debt or to the flow of interest payments is a

48

Sergio V. Barone and Alberto M. Díaz Cafferata

(1996) call this condition “solvency”; but crises develop not only because of solvency problems. Consider the issue of sustainability: if the current policy stance is maintained, and the turning point from trade deficit to trade surpluses is likely to occur smoothly, without drastic change in consumption and economic activity, these authors define the current policy as “sustainable”. In contrast, “unsustainability” appears when an event triggers a "drastic" policy shift, showing external vulnerability or some lack of capacity to undertake adjustment policies. A related point is that an economy at a certain moment may be solvent, in the sense that the intertemporal condition is fulfilled, but its payment ability be temporarily impaired by liquidity problems. Even when solvency at a point in time is determined by the discounted value of future trade surpluses exceeding the stock of net external debt, flow imbalances also matter. A distinction between liquidity and solvency helps identify the role of exports. For example, the debt-to-GDP ratio is a measure of solvency and the same happens with the long run expected path of exports compared with interest payments in the future; on the contrary, short-term debt over exports and debt service over exports are liquidity measures.8 In other words, from a temporal point of view shifting attention from liquidity towards solvency implies also focusing in longer run determinants rather than on immediate causes. Note that a liquidity problem can be generally solved by rollover or issuing bonds. But an economy must be solvent to have this solution open; and, on turn, solvency depends –among a set of determinants- on exports. To emphasize the point, we argue that historically external crises in Argentina have been caused mainly by the difficulties to export enough in the long run to keep external solvency. Phenomena of external disequilibrium in a sense boil down to solvency, as in the case of the Argentine economy which has recurrently reached the “rigid” long run export constraint even with relatively low debt ratios. This line of reasoning justifies our focus on export performance and on the hypothesis that the long-run export rigidity constrains the “admissible” debt to a structural level. Therefore, only with a

8

signal of the capacity to obtain needed foreign currency to make those payments. Manasse and Roubini (2005).

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

49

structural change in the export ratio the economy might bear a higher long-run debt ratio. Then, why cannot Argentina sustain CA deficits if other countries can? Even when it is clear that there are multiple causes, such that exports appear as only one piece of a complex mechanism, it is also true that this particular piece may become in occasions critical for the external equilibrium of a borrowing country, and our starting point in this paper is that the structural component of exports may be the key for the long-run solvency condition. Miscalculations may be costly. Gandolfo (2001) asserts that many countries have indebted themselves to finance ambitious development programs, and that the reason for the difficulties is that "the hoped-for huge export increases with which to get the foreign exchange to service the debt did not materialize".9 Structural rigidity of the exports ratio, coupled with the fact that imports were tied closely to GDP as the consequence of the import substitution strategy, helps to explain why the economy was capable of using external savings only in scanty amounts and for short periods. We expect to find out how exports as “required” by the long-run external solvency condition compares with actual trend exports, and whether there is a coincidence of observed reversals of the CA when there is a gap between them. If the Argentine crises have been the result of the operation of multiple causes, as the recent literature shows, the required-actual exports gap is not expected to provide a precise advanced signal of crisis. It gives nonetheless indication of external fragility, and the distinction between trends and cycles might help identify relevant thresholds.10 Two dimensions of solvency shall be kept in mind. An indebted economy needs keeping two intertemporal budget constraint restrictions. One is fiscal solvency (with payments capacity derived from tax collection in domestic money and limits to the levels of internal tax pressure). The other is external solvency (concerning net flows of hard currency in balance of payments accounts). By focusing on external solvency we are looking at only the latter of these two restrictions; a critical question is when one or the other, or both, are 9

10

He adds that difficulties were caused partly because of the unfavorable economic situation. There are certainly “country specific” threshold levels at which a gap becomes relevant, showing the presence of other determinants of sustainability, like reputation.

50

Sergio V. Barone and Alberto M. Díaz Cafferata

binding. Assume that with a given GDP and taxing capacity a government is able to collect the needed amount of taxes in domestic money. The problem becomes in this case the requirement to buy external currency in the market or to reduce reserves at specific points in time when payments are due. Our perception is that in our period of observation after the end of the 1940s the dismal exports performance has created permanent strong restrictions to the admissible levels of sustainable external indebtedness of the Argentine economy. A. The intertemporal constraint and the steady state solvency condition: highlighting the role of exports. The CA surplus flow is:11 (1) where NFA is the stock of net foreign assets, Y is the GDP, r is the international interest rate, C is private consumption, G is government current expenditure and I is total investment. Multiplying (1) by minus one and rearranging we get: (2) (1  r ) NFAt C t  Gt  I t  Yt  NFAt 1 which iterated T-times forwardly12,, can be written as: CAt

NFAt 1  NFAt

t T

(1  r ) NFAt  ¦ ( s t

Yt  rNFAt  C t  Gt  I t

1 s t ) (Ys  G s  C s  I s ) 1 r

(

1 T ) NFAt T 1 1 r

In the limit when T o f , (1 1  r ) T NFAt T 1 o 0 , replacing

(Ys  G s  C s  I s )

TBs , where TB is the trade balance, the

constraint becomes: 1 s t (3) ) TB s 1 r If the trade balance for every years s is TBs TBs 1 (1  g ) , where the trade balance grows at the same rate g as the economy, the Equation (3) can be written as: f 1 s t ) (1  g ) s t (4)  (1  r ) NFAt TBt ¦ ( s t 1 r  (1  r ) NFAt

f

¦(

s t

11

Cfr. Obstfeld and Rogoff (1996); Milesi-Ferret and Razin (1996); IMF (1993).

12

We obtain NFAt 1

Yt 1  Ct 1  Gt 1  I t 1 NFAt  2 , by forward  1 r 1 r

substitution in (2). Iterating until t+T, we get (3).

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

51

In the RHS of Equation (4) the trade balance at each point in time s is TBs (Ys  C s  I s  Gs ) , equal to the excess of GDP over absorption or, for an indebted country, the net amount of output the economy transfers to foreigners each period. Operating13 in Equation (4) the solvency condition becomes: 1 r  (1  r ) NFAt TBt ( ) ; or TBt (r  g ) NFAt (5) rg And for an indebted economy net foreign assets is negative, such that D = -NFA and the economy maintains a steady state debt to output ratio Ds Ys . The CA flow deficit identity implies for every time s in the steady state an increase in debt Ds 1  Ds r * Ds  TBs . Writing variables as ratios of GDP with lower case letters, the trade balance consistent with solvency is: tbs (r  g )d s (6) Where s denotes steady state values and Equation (6) provides the trade balance to output ratio (tbs) required to maintain constant the ratio of debt to output ds. Given d and g, the required trade surplus ratio is an increasing function of the rate of interest. B. The rate of interest and the rate of growth Obstfeld and Rogoff (1997), point out that a growing economy can run perpetual CA deficits and still maintain a constant ratio of foreign debt to output, when the country pays out the excess of the interest rate over the growth rate. There is indeed a critical role of the difference between the growth of the economy and the interest rate paid in the external debt. Consider a highly indebted country with d= 0.5, a long-run GDP growth of two percent, and an interest rate of 5 per cent. The required surplus ratio is tbs= (0.05-0.02)0.5= 0.015. If the rate of interest is higher, r=0.08, then tbs= 0.03, and the economy has to double its trade surplus ratio to keep constant its debt ratio. The usual set up assumes g
13

f

If g< r, ¦ ( s t

1  g s t ) converges to 1 (1  (1  g 1  r )) . 1 r

52

Sergio V. Barone and Alberto M. Díaz Cafferata

The trade balance given by Equation (6) may be called the “required” trade balance. The required trade surplus in this sense is the burden of the debt, when the stock of external debt is growing at the same rate g as the GDP, which is itself lower that the relevant rate of interest r. As long as the economy manages to keep a constant amount of absorption relative to GDP, a growing country may increase its debt in absolute terms (i.e. may run CA deficits) because the present value of future surpluses is equal to the initial debt in each period for a similar temporal horizon of the debt. To understand why, consider some limiting cases. i) If the economy is growing and the rate of interest is zero: g>0, r=0, the trade deficit each year s is equal to the CA (because the service account is zero); the debt increases by the amount of the CA deficit at the same rate as the GDP while keeping the debt to GDP ratio constant. ii) Consider the oposite case, when the economy is not growing and there is a positive rate of interest: g=0, r>0. This limiting case of a stagnant economy with a long-run rate of growth g=0, unfortunately is not far from the Argentine experience in some periods;14 the condition implies that the CA deficits (i.e. the long-run CA deficits) shall be also zero, and the service account deficits be compensated by trade surpluses.15 iii) When the economy is growing at the same as the rate of interest r=g>0 the trade balance is equal to zero; there is a permanent roll-over, such that the debt increases in an amount equal to the interest to be pay. iv) In the case of g>r>0 the economy may keep not only the trade balance but also the CA in deficit. The trade balance to output ratio required to fulfill condition (6) is negative, meaning that the economy might run perpetual trade balance deficits given the steady-state debt to GDP ratio ds. In this case the external debt increases beyond the amount of interest payments, still keeping a constant debt to GDP ratio. The two last cases are interesting because they show that it is theoretically possible for a country to keep CA deficits for long 14

15

The GDP per capita in 2003 was the same as in 1973. Cfr. Díaz Cafferata (2005). Further note that an economy with a positive but declining trend growth would find out that increasing exports relative to GDP are required to maintain a given steady-state debt to GDP ratio.

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

53

periods if its growth rate is higher (or equal) than the international rate of interest, and as we know this has been found in practice. But an economy cannot accumulate debt forever if an intertemporal budget constraint must be fulfilled giving rise to the issue of a turning point to which we will come below. C. The role of exports in solvency and the turning point In synthesis, the notion of external solvency implies that a country has the ability to generate sufficient trade surpluses in the future to repay existing debt, i.e. that the present value of transfers to foreigners must be equal to the value of the economy initial debt, such that if the country has run in the past CA deficits, thereby accumulating external debt, the solvency condition requires a turning point from trade deficits to surpluses. Milessi-Ferretti and Razin (1996), indicate that the solvency condition is “silent about timing and nature of this shift”, but nonetheless we shall say something on the matter. Further, external payments capacity is not all that matters. A related question of whether CA imbalances are sustainable is formulated by these authors as follows: the current policy stance is defined as sustainable if the turning point from trade deficit to trade surpluses is likely to occur smoothly, without drastic change in consumption and economic activity, i.e. without a crisis. In both cases the solvency condition requires a turning point from trade deficits to surpluses, and the change of sign in the trade balance occurs smoothly when the CA deficit is sustainable. MilesiFerretti and Razin (1996, p.8) state that “a higher growth rate can facilitate a smoother switch in the trade balance”. To drive attention specifically towards the importance and the role of exports performance, Equation (6) can be rewritten in terms of the required exports ratio to output that maintains stable the ratio of debt to GDP; assume at this point for simplicity that imports are a constant fraction of the GDP, as follows. x (r  g )d  m (7) Equation (7) implies that a country´s exports to GDP ratio must be equal to the imports ratio plus the excess of interest rate over the growth rate for the economy to be solvent. In this case, the export growth is equal to GDP growth, and the economy is in steady state. It is useful to see that given a constant degree of openness measured by the X/GDP ratio (i.e. with exports growing at the same rate as output) the upper limit of the debt to GDP ratio is also given. A strong practical implication of Equation (7) is that, other things

54

Sergio V. Barone and Alberto M. Díaz Cafferata

given, a structural (long-run) value of exports relative to GDP determines uniquely the debt to GDP ratio. To see why, let for simplicity h=1/(r-g) and write d as the dependent variable, d h˜ x  h˜m (8) such that d appears as the limit of the initial debt an economy can bear for a given long-run export performance or, in other words, provides at a given point in time the “admissible” structural debt ratio for a country as a function of the structural exports performance. Interestingly, written in this way, the budget constraint provides indeed the “turning point” Milesi-Ferreti and Razin (1996) asked for. Equation (8) gives in effect a “stop” signal. If d is below the “admissible” value for given exports, the economy may borrow further in the long run. But only to the point where the steady state debt ratio reaches the limit given by the exports performance, because borrowing beyond that limit breaks the long-run solvency condition. Only transitory, shock absorbing, additional debt is allowed for this country.16 Figure 2 presents a comprehensive graphical vision of the issues, in the space of three variables: i) the export ratio X/GDP; ii) the debt ratio D/GDP; and iii) the tax ratio T/GDP. Note that all these variables are steady state values and forward looking. For a given steady state debt ratio do the point v along the line mv provides the burden OxR or exports required. In the lower panel the tax requirement OtR is determined in a similar way. Alternatively, the figure shows the “admissible” debt OdA given the structural export ratio.17,18

16

To highlight these relationships, variables may be written in terms of the, unobserved, permanent (trend: T) and transitory (cyclical, C) components. x xT  xC ; d d T  d C from Equation (7):

d

h ˜ xT  h ˜ xC  h ˜ mT  h ˜ m C and the “admissible” trend debt

is: d T 17

18

h ˜ xT  h ˜ mT .

The structural upper-bound for the tax ratio would similarly point out to the long run solvency condition from the fiscal side. The ratios frequently used by analysts such as D/X or D/Y are not measures of solvency.

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

55

Figure 2: External and fiscal solvency

m+(r*+ ij(d)-g)d

x = X/GDP

w

xR Gap

m+(r*+ ij-g)d

v

S

x

m m’ 0 g’

d

d0

d = D/GDP

g

tR tS t = T/GDP

g+(r*+ ij-g)d

Our approach to the problem of solvency is based on the assumption that agents in the financial markets expect the economy to running into problems when the perceived future structural (longrun) export ratio is below the (expected) burden of the debt. We call this difference the “export gap”, equal in Figure 2 to: Export gap Ox S  Ox R (9) For the empirical application the expected values of the solvency condition are replaced in (8); the formula used for the calculations is in consequence: Export gap Ox s  > ( r * M  g ) d  m @ (10)

56

Sergio V. Barone and Alberto M. Díaz Cafferata

The ability of the economy to maintain at a given point of time a debt ratio d=do depends in consequence on the perception of the future evolution of the export ratio, given the current information. We call xS in the empirical application of section V the “observed” (expected) export ratio, in the sense “proxies” for the steady state that forward looking values are obtained operating with the observed series. It shall be evident at this point that the computations depend heavily on the assumptions about future scenarios at each point in time. The rate of interest relevant to calculate the service of the debt has been written as r r *  M (˜) with two components, the Fed rate and country risk. Rewrite for convenience the solvency condition (7), x (r  q) d  m or x rd  gd  m . If the country risk increases with the debt ratio, rd (r *  M (d ))d ; the dotted line going through m and w assumes M (d ) M ˜ d such that M ˜ d r ˜ d  M ˜ d 2 . Import substitution and government consumption are shifters of the solvency condition: m’<m and/or g’
From the formal solvency condition to empirical measurement

The amount of domestic bonds sold each period to foreigners is equal to the accumulation of external debt, but international buyers of domestic bonds need to perceive they will be paid. How long can the country continue running CA deficits? Extensive periods with CA deficits have been observed in practice. An explanation is the Stages of the Balance of Payments Hypothesis. At the time when the borrowing country is increasing its debt, lenders perceive that in the future the sign of the CA will change; and that the solvent borrowing country will buy the debt back. Consider now a borrowing country that pays, each and every period, the flow of interests of a perpetuity as portrayed in Equation (6). In the steady-state the ratio D/GDP does not rise, but the absolute value of the external debt can indeed increase when the ratio of external debt to GDP is constant and the GDP is growing.

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

57

The balance of trade surplus is used to honor the interest payments. This means that lenders do not panic and try to sell domestic bonds back even when the debt is growing because they are being paid the international rate of interest and they expect the process to continue regularly in the future. When on the contrary, is there an indication that the economy is not going to fulfill the intertemporal budget restriction? The level of exports of the economy, the stability of export revenues, the ratios of exports to debt and the ratio of trade balance to debt are frequently used to assess the capacity of an economy to honor its foreign obligations. Attention is placed here on the role of exports in external solvency in practice, the relevant mechanisms and empirical magnitudes. It has been shown that external solvency depends on an ample set of variables; the focus placed here on exports is driven by the interest of understanding in which cases the performance of exports becomes the relevant variable among the determinants of solvency and among the predictors of CA crises. For the empirical application, the theoretical framework and the intertemporal budget constraint will be used now to provide a historical perspective of the Argentine growth problems and recurrent external crises, and for this purpose a distinction is made between the (unobservable) long-run and cyclical behavior of the relevant variables. A working hypothesis is that the long-run CA is determined by a consistency requirement with the long-run path of exports, and that trouble arise when the CA deviates from this solvency requirement. Are sustainable CA deficits in Argentina related to some specific structural characteristics of the economy? More punctually, is there a structural degree of openness that can be related to the size of long-run sustainable CA deficits? To what extent has the exports/GDP ratio been the relevant variable regarding solvency may be examined estimating whether the historical observed values of exports had been consistent with the intertemporal budget constraint. A policy implication of the discussion is that, as far as the solvency and sustainability of the CA becomes influenced by the degree of openness, at a given point of time it is necessary to monitor the magnitude of the degree of openness (X/GDP), as it renders the sustainable target value of the debt/GDP ratio. The perspective of identifying which value of D/GDP is sustainable given the degree of openness appears as particularly relevant if the

58

Sergio V. Barone and Alberto M. Díaz Cafferata

X/GDP ratio has a structural rigidity. And if the debt of the economy crosses this structural “debt threshold” finds itself in trouble. A. Interaction between real and financial channels, two periods from import substitution to financial globalization For debt payments requirements to be honored, two different kinds of requisites shall be fulfilled. One is the amount of tax collection; the other, the availability of external currency which depends on trade balances and exports performance. The empirical problem we face is to evaluate whether and to what extent the frequent external crises in Argentina are explained by a long-run exports rigidity, and the recurrently growing gap between actual exports and the “required” level of exports which impairs solvency and makes the economy vulnerable even to relatively small increases in external debt. We shall explore if the gap is a predictor of the crises: the reason is that it may be interpreted by the market as an indication of reduced payments capacity below needs. Different indicators of solvency between 1948 and 2004 will be calculated and compared to highlight how each of them behave along time, and at which points their value announces the presence of solvency problems. Low deficits and frequent changes of sign, and periodic reversals of the CA are indication of the inability of Argentina to use external savings, and this feature is one key of its growth problems. Relevant issues include: firstly, how the long-run exports performance, and in practice the structural exports to GDP ratio, contributes to explain the historical characteristics of the external equilibrium. Secondly, more specifically we try to identify in historical series if the presence of trouble in external payments capacity and CA reversals are related to the behavior of exports. Thirdly, from a different perspective, the question for an economy with a given "free trade" exports intensity is the choice of a correct growth strategy and the appropriate use of external financing consistent with a given degree of openness of this particular economy.

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

59

B. Structural long-run conditions A hypothesis related to the presence of long-run characteristics of the economy is that, apart from the incidence of economic policy, there is a structural rigidity in exports performance, related to "permanent" characteristics of the economy, such as the distance to foreign markets, bargaining power in international negotiations, sectorial comparative advantage and specialization, providing a "free trade" long-run exports ratio.19 The observed stable constant X/GDP ratio is a consequence of policy and also of a "structural" degree of openness, such that growing CA deficits break rapidly the solvency condition. A distinction between two different policy approaches to economic development regarding external protection is necessary to understand the relationships between exports performance and external debt in Argentina. Until the seventies, the stock of external debt was largely the sum of "small" amounts of loans by foreign governments and a limited number of large international banks. With a low amount of debt and little volatility of the CA flows the accent on the explanation of the problems to grow was placed on the capacity to export. The mechanism known as the stop-and-go problem is the following:20 with CA flows fluctuating about zero, the external balance is given by the trade balance. If external savings are not used to finance imports, the relevant restriction is: if CA S * 0 , then TB X  m ˜ Y and dTB m ˜ dy CA , such that solvency is: X (r  g ) ˜ d  M ; If also CAt t 0 , then D t 0 , and M X fix

Since the economy was not able to sustain CA deficits and exports were rigid, expansive cycles were aborted because imports grew with GDP and the trade balance went into deficit. In a vicious circle, import substitution policies taxed exporting sectors providing weak signals to invest in potentially dynamic sectors (together with inefficient state intervention in the financial sector), and chronic inflation, coupled with intervention in the financial sector, kept domestic savings at a level around 20 per cent of GDP; investment could not in consequence be higher than domestic savings with a CA moving around zero. In this set up the cost of violating 19 20

Cfr. Díaz Cafferata (1996). On the literature about the mechanism see Brodersohn (1969).

60

Sergio V. Barone and Alberto M. Díaz Cafferata

the external budget constraint was the transitory fall of the low rate of growth, but since the debt-to-GDP and debt-to-export ratios were not allowed to accumulate beyond certain restricted thresholds, the external crises, even if frequent, were not large and persistent. What could the economy do to escape from the import substitution growth problems? A consensus grew gradually among economists that the root of the problem was the anti-export bias of commercial policy, that by disincentivating the expansion of the Agricultural sector prevented the Argentine economy to reach growth rates comparable to similar economies like Australia. A study of EOCEI (1971) transmits the opinion at this time, a quarter century after WWII: the external sector is seen as the restricting factor of growth in Latin America; the unsatisfactory behavior of exports is represented by the fall to a half in the participation in world exports in two decades, from 12 per cent to 6 until 1971. With this diagnosis, the solution appeared to be a policy shift toward restoring incentives for the efficient allocation of factors of production by eliminating distortions in prices created by protection. Contemporaneously with this change in policy recommendations, a global expansion of financial flows brought an additional ingredient to the policy advice: greater facility to use external savings might help dealing with the costs of the structural transformation and speed up the creation of positive effects of the liberalization of goods and factor markets. But, how did the new paradigm work in practice? One consequence of the perceived abundance of funds was that the need to export in order to keep external balance was relegated, hiding an essential long-run problem of external solvency. We shall explore here to what extent the above description is consistent with the facts, working with several representations of the solvency restriction. C. Measurement of the forward looking concept It shall be noted that this way of tackling the problem is forward looking in the sense that it measures how high the trade balance should be in order to avoid solvency problems in future years. The concept of solvency, even in the simple deterministic expression of Equation (6), presents difficulties for the empirical estimation. This equation may be read as a measure of the “burden of the debt" at time t0 in terms of the “required” trade surplus. A rising burden is associated to increasing difficulties to pay, and in consequence the

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

61

economy becomes more vulnerable. It is also to be noticed that solvency is a condition allowing the service of the debt to be paid in the future at every moment t>t0. It is therefore a “forward looking” concept. Our problem is how to apply these theoretical steady-state forward looking solvency relationships on the evaluation of the actual historical path of the variables. The criterion is that an increasing negative gap between the observed and the required trade balance or exports would be an alert signal of a process with difficulties to honor debt payments sooner or later if the tendency is not reverted. It should however be noted that the contemporaneous data is not the forward looking information of Equation (6). From historical time series we want to approximate a solvency condition in each moment of time. By doing this, we are able to obtain a new time series that does not have a simple interpretation as it would be suggested in a first glance by the budget condition presented above. It should be recognized in each estimated value the confluence of two temporal paths: the future and the past. The latter exerts an influence since the values for a particular year are the outcome of the economic structure shaped through the years with internal and external influences, and political decisions. On the other hand, the future path, being the forward looking element, is determined by the model nature and the key assumptions that are specifically used, namely, the economy in a steady state, a constant GDP growth rate and a constant interest rate, among others. Rewrite for convenience the basic expression for solvency (6) already discussed tbs (r  g )d s . In the RHS of the equation the value of the debt relative to GDP, d is assumed as given at time t0, and the exercise is to determine the trade surplus ratio a net debtor has to reach as a function of the steady state rate of interest and GDP growth in the future. In consequence, “solvency” implies an expectation about the future path of the economy. To determine the solvency of an economy at a given point in time is therefore a difficult matter both for economic agents and to economic research. The distinction in between solvency and sustainability suggested by Milesi-Ferretti and Razin (1996) shall be reviewed. Seen in a historical perspective there are two different visions for the analyst. The first one is looking backwards, with hindsight, if the economy was in fact solvent at a given point in time in the past. The other perspective is to look for indications of how, at different points in time in the past, economic agents in the economy assessed the

62

Sergio V. Barone and Alberto M. Díaz Cafferata

degree of solvency. Our estimations do not provide an explicit modeling of this value, but do offer quantitative indication and suggested interpretations on the light of these two perspectives. D. The rate of interest: representative bond, lower and upper limits. Contemporaneous data versus average of future values as forward looking expectations In the standard solvency condition as written in Equation (6) there is only one constant steady state rate of interest, which can be interpreted as the rate paid by a “representative bond”, disregarding the structure of the debt in terms of maturity, type of debt according to the relevant domestic or external agent, FDI, and so on. The choice of the rate of interest to be used in the exercises deserves some discussion. The solvency condition for an agent at a point in time t is forward looking; the equation in the deterministic set-up includes a fix rate of interest to be paid in the future, in steady state, for a representative bond. It is implicitly assumed that the supply of funds is infinitely elastic for a small open economy at the exogenous international rate of interest. For the empirical estimation of solvency, the use of a single rate of interest is a simplification that hides the fact that there is a universe of rates of interest, as well as other potentially relevant characteristics of the debt such as: i) The type of creditor (government agencies, private agents): until the seventies there was a limited number of creditors with a substantial weight of foreign government and agencies, a pattern that shifted to a large number of anonymous small external creditors around the world creating particular difficulties for borrowing countries to renegotiating the debt in problem. ii) The type of debt, such as fixed or indexed rate of interest, financial debt or direct investment. iii) The time profile. For a given net foreign asset position the structure of maturity matters, e.g. the average of the short-term debt to GDP ratio for Argentina that was 4,8 per cent in 19941996 raised to 6,8 per cent in 1997-2000 (Gurtner, 2004). When the international rate of interest increases (Federal Reserve Bond), the current rate of interest overestimates the true rate of interest to be paid by a debtor country, because the rate of interest on past debt is lower, i.d. between 1978-1981 the implicit rate of interest paid for Argentina external debt as estimated by Rodriguez (1986), was lower than the FED’s rate (see Figure 3).

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

63

Figure 3: Argentina 1948 – 2004 (The rate of interest for the solvency condition: lower and upper limits) 60%

50% 2002 40% 1950

1973

1953

30%

1988

1995

1976

1963

20%

10%

0%

1948

1953 1958

1963

1968

1973

Lower limit for the interest rate

1978

1983

1988

1993

1998

2003

Upper limit for the interest rate

Source: Own estimations, based on Rodríguez (1986), IMF – IFS, Ávila (2000) and MECON.

Beyond the type of bonds and time profile, changes in the rate of interest that a given country pays to service the debt may be portrayed in terms of “global” and “country specific” components: i) The “basic” rate of interest is paid in a “risk free” bond, such as the United State treasury Bill or a Federal Reserve Bond, a value that is similar to close substitutes in the international financial market and moves influenced by “global” evolution of supply and demand of financial assets. ii) International arbitrage determines the addition of “country specific” components of the rate of interest, related to fundamentals, which determine the degree of solvency and liquidity. Long-run restrictions, may be the degree of openness or the structure of exports; a developing country exporter of primary commodities is more vulnerable to terms of trade shocks. Increasing government expenditure and fiscal deficits create short time difficulties.

64

Sergio V. Barone and Alberto M. Díaz Cafferata

iii) An additional influence, has been argued, is the history of compliance that punishes countries that have had difficulties in the past, named “debt intolerance”, such that relatively lower levels of debt ceteris paribus are likely to trigger an external crisis in countries with bad “reputation” (Manasse and Roubini, 2005; Reinhart, et al. 2003). iv) It has been noted that there is also an element of “contagion”: belonging to a “club” of emerging countries perceived as sharing capacity (or willingness to pay) makes them vulnerable to contagion (Avila, 200021; Rodríguez, 1986). We will work with only the aggregated value of the debt and will not deal with the effects of its composition. But we still have to decide which interest rate is representative of the value that appears as the forward looking steady state in the solvency equation. Four different values will be used. A distinction we make is between a “lower limit” and an “upper limit” rate of interest paid on a perpetuity to compare an optimistic and a pessimistic set-up. The four different possibilities mentioned above result from the following combination: firstly, a low and a high rate are used taking into account, or not, the presence of risk. The low value is the Fed rate without any changes (i.e. the one for every particular year), which we consider as the “lower limit” for the relevant interest rate; also an “upper limit” for the interest rate is set by adding to the Fed rate an element of country risk, providing this higher rate the unfavorable scenario for the country when analyzing the debt burden issues. Secondly, in both the high and low rate cases, we use two different approaches for representing the theoretical notion of an expected rate. Using current data to calculate the burden or solvency alternatives at each point in time assumes myopic expectations, which is not far from truth when there is a very short relevant horizon and data are annual. An alternative procedure is to obtain a smoothed rate using at each point in time the average of the next ten 21

Cfr. Ávila (2000). He describes the actual rate of interest a country pays as the sum of two components, a world “risk free” rate of interest plus a “country risk” component that is increasing on the magnitude of the debt, providing a total rate which is the supply of external saving the economy faces. Different values of other variables such as exports performance, terms of trade, imports, fiscal solvency, reputation and so on, are likely to shift the curve introducing an element of uncertainty.

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

65

years, as representative of agents expectations together with its long run evolution (assumed to be constant in the theoretical framework). The devise of ten years future average follows the Obstfeld and Rogoff (1997) estimation of the burden of the debt.22 We use as the lower limit the Fed’s discount rate, and as the upper limit the Fed interest rate plus a country risk: i)1948-1969 Fed’s discount rate; assuming expected country risk was not significant for this time period; ii)1970-1980 interest rate from Rodríguez (1986); iii)1981-1997 Fed’s discount rate plus a country risk from Ávila (2000); iv)1998-2004 Fed’s discount rate plus a country risk from MECON. E. The rate of growth of GDP In the solvency equation the horizon is forward looking and the steady state values should therefore provide an estimation at time t in terms of expected future GDP. In this case, we shall use a geometric average of ten years that follow the estimation23 of the future GDP. The constant prices GDP from Díaz Cafferata (2005) is used. F. Sources of data a) Trade Balance in current dollars is from INDEC (www.indec. gov.ar); b) GDP in current dollars from Ferreres (2005); c) External debt in current dollars: From 1948 to 1965: Rodríguez (1986); from 1966 to 2004: Consejo Técnico de Inversiones (2004); d) Interest rate: Fed’s discount rate, from International Financial Statistics, International Monetary Fund; series code: 11160...ZF... .e) GDP growth rate in constant pesos of 1993 from Díaz Cafferata et al. (2005). IV.

Some empirical exercises on the historical solvency of Argentina

We write the solvency condition in alternative ways looking for the best expression to empirically recognize the possible presence in the past of a perception of mounting external difficulties associated to experiences of reversals. To give the simplest specification of the solvency condition as written in Equation (6) an empirical content, 22

23

Since the average mimics large future fluctuations that are not likely to be the type of expectations of agents about future rates, an alternative is to use trend rates For the last periods of the series -1995 and forward- the 10 years previous to the estimation date were used.

66

Sergio V. Barone and Alberto M. Díaz Cafferata

we have to take decisions of two types. Firstly, choosing between different formal expressions of the solvency condition that allow the analyst to distinguish the variables that are assumed as endogenous from those that operate as channels of adjustment. Secondly, selecting the empirical couterpart of the theoretical forward looking path of the relevant variables in steady state, as well as the specific assumptions about the behavior of variables (such as the import ratio, or the risk premium). Also, in the interpretation of numerical results the implication of assumptions such as a single type of bond, and of the lack of uncertainty shall be acknowledged. Four exercises below are aimed at either estimating the “burden of the debt”, or calculating an “observed-required gap”. The burden of the debt is formally the required steady-state payments given a debt to GDP ratio as the exogenous variable; the burden may be in terms of the trade balance surplus (the needed restrain in expediture), or in terms of exports. Note that the “burden” is just the amount of payments to be made. On turn, the “gap” compares this burden with an estimation of the capacity to pay: economic agents may perceive future difficulties to pay from the signal given by two types of gaps, a forward looking negative observed-vs-required gap caused by a weakness in the expected exports performance (or trade balance) given the debt; or a positive observed-vs-admissible gap that emphasize the limits of safe borrowing. Gap between observed and required values of trade balance, or of exports24 assuming an exogenous steady state debt ratio: (i) The burden in terms of needed future trade balances is compared with the forward looking estimation of capacity to pay (“observed”). (ii) Alternatively the negative gap between observed exports and “required” exports consistent with solvency indicates vulnerability. The forward looking imports “m” in Equation (10) must be given a numerical value. Gap between the observed and the “admissible” debt: (i) Rewrite Equation (6) to obtain d tb /(r  g ) to get the admissible debt, assuming the trade balance to GDP ratio as the exogenous variable, and compare with the debt. (iii) You may take the exports ratio as the exogenous variable, and either observed 24

A related issue is whether the relevant variable is the total value of export or the value added in exports, and also the importance of the stability of exports earnings.

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

67

imports, or trend imports to substitute “m”. Vulnerability is suggested by an “observed” debt greater than the “admissible” value. This representation provides a limit to borrowing assuming there is no uncertainty. One further comment on the estimations is pertinent. What we call the “observed” values of exports and of the trade balance (i.e. the forward-looking external payments capacity to be compared with the burden) is calculated from either the annual data or the smoothed series as in exercise 4; the latter are used as a possible replication of expected exports based on the historical trend. Next the results of some of the exercises performed are reported: the burden in terms of trade surplus required (exercise 1); the burden in terms of exports (exercise 2); the gap in terms of exports (exercise 3); and the gap in terms of exports trend (exercise 4). Exercise 1. The burden in terms of trade surplus as a fraction of GDP.25 The standard solvency condition in a deterministic set-up, as in Obstfeld and Rogoff (1997), equates the burden of the debt to the required trade surplus. The higher this burden, the greater the likelihood that the debt is unsustainable, providing warning signals of possible future trouble scenarios. A growing burden (i.e. a rising required trade surplus) suggests indeed increasing difficulties to pay in currency, and announces future needs to lower domestic absorption with the consequent domestic resistence. These signals are more likely to be seen with anxiety by external creditors when the borrowing country suffer from “debt intolerance”. The results of the first exercise are shown in the two parts of Figure 4 which depict the burden of the debt according to the RHS of Equation (6).

25

An alternative perspective of the burden, would be to start from the debt to exports ratio as the exogenous variable. A study by the Policy Development and Review Department of March 23, 2000, about Debt and Reserve- Related Indicators of External Vulnerability, provides in Annex V a Numerical Interpretation of Debt Ratios, in terms of the magnitude of the transfers implied by debt to export ratios. The formula b ( r  n ) D / X 1 indicates the constant fraction of exports that must be transferred to pay "interest after net debt inflows" if the stock satisfies the solvency condition that the present value of the debt goes to zero.

68

Sergio V. Barone and Alberto M. Díaz Cafferata

In both panels (a) and (b) the continuous line show the trajectory of the burden calculated with the contemporaneous rate of interest (a myopic expectations assumption about the forward-looking path) capturing all of its inherent variability; the dashed line instead is the burden using the observed ten years forward average rate of interest (in the spirit of a perfect foresight assumption), which smoothes the fluctuations. The difference between panel (a) and panel (b) is that the former uses the (risk free) lower limit of the contemporaneous and the average rate of interest; the latter on turn considers the upper limit, with a risk premium, increasing the level and influencing heavily the temporal pattern of the burden. It is possible to identify two temporal phases for Argentina. Between 1948 and the mid 1970s the country was under the “import substitution” policy with also substantial restrictions to external financial flows; surprisingly, despite a low debt that remains below 1 per cent of GDP suggesting no insolvency risk, three reversals appear nonetheless in 1950, 1953 and 1963, revealing an imperfect integration to the world financial markets. Figure 4: Argentina 1948-2004, Debt burden in terms of the trade surplus to GDP ratio (to be continued) 6%

a) Lower limit for the interest rate

5%

2002

4% 1950

1953

1963

1973

1976

1988

1995

3%

2%

1%

0%

-1%

1948

1953

1958

1963

1968

1973

Contemporaneous interest rate

1978

1983

1988

1993

1998

-----10 years average interest rate

2003

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

69

Figure 4 (cont.): Argentina 1948-2004, Debt burden in terms of the trade surplus to GDP ratio 30%

25%

20%

15%

90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1980

b) Upper limit for the interest rate

1988 1985

1990

1995

1995

2000

1976 10%

1950 1953

1973

1963

5%

0%

-5%

1948

1953

1958

1963

1968

--- 10 years average interest rate

1973

1978

1983

1988

1993

1998

Contemporaneous Interest rate

In the second phase along almost two decades of financial globalization and rising interest rates, both the risk free rate and the risk premia increase, bringing up the burden in terms of trade balance to extraordinarily high levels, between 10 per cent and 25 per cent of GDP in the 1980s (see figure 4b). The adjustment required to this highly exogenous shock in the balance of trade ratio, to achieve surpluses of this magnitude, is impossible in practice for any economy. Two reversals in 1973 and 1976 appear associated in this new scenario, an increase in the debt coincident with a substantial rise in the international rate of interest until the early 1980s (see Figure 3), and a new framework of financial globalization. At the same time the country shifts to a more open development strategy, lifting restrictions to international capital and commercial flows. The CA deficits and the rise in the rate of interest are reflected in an increase of the debt burden in terms of the trade surplus ratio, with a highest value at the time of the “crises of the debt”. Reversals in 1988, 1995 and 2002, are preceded by a rising burden or coincident with a peak as expected. We shall decide which is the relevant rate of interest to use in the calculations of solvency. Compare the predictive capacity

70

Sergio V. Barone and Alberto M. Díaz Cafferata

of the alternatives, using the lower bound (or risk free) rate of interest and the upper bound rate (with risk) to calculate the burden in terms of trade balance. Until the 1970s the lower and the upper limits for the interest rate provide similar results, as expected, since the risk was low as noted before. After the crisis of the debt the importance of the country risk is evident since the reversals of 1988, 1995 and 2002 are clearly associated with the increasing burden when it is calculated using the upper bound rate (but not with the lower bound). Not to mention the magnitude of almost 80 per cent reached by the burden in the crisis of 2001-2002, partly as a consequence of the devaluation (see note 6) which causes a jump in the debt ratio due to the fall in GDP in dollars given the total external debt in currency values. Exercise 2. The burden in terms of required exports to GDP ratio. Given the debt to GDP ratio, and the observed imports, Equation (6) provides the exports to GDP ratio required to fulfill the solvency condition (i.e., the burden in terms of exports). Note the different profile of the burden in figure 4 (trade surplus) and figure 5 (exports) revealing also the importance of imports ratio “m”. Figure 5 shows the evolution of the burden using the upper limit of the rate of interest. At the beginning there is a high level of the required exports that decreases and stays in a lower level since the mid 1950s. In the import substitution period the burden falls before the reversals in 1950, 1953, 1963 and 1973. The typical mechanism is centered in the balance of trade: a “stop” is caused by increasing imports with rigid exports; the external equilibrium is reached through a devaluation, expenditure adjustment, and recession; after the adjustment, the “go” is allowed by the (transitory) balance of trade surplus and the possibility to borrow again (small amounts). The reversal in the CA (deficit) probably shows a combination of falling imports and reduced willingness to lend to the troubled economy under current conditions. After the adjustment the recovery of the GDP requires higher exports suggesting a "new brand" version of the stop-and-go, now with financial openness and in consequence with more ample movements of the burden.

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

71

Figure 5: Argentina 1948-2004 The required exports to GDP ratio. Upper limit for the interest rate 40% 195 35%

30%

25%

20%

90% 80% 70% 60% 50% 40% 30% 20% 10% 0 1980 %

1985

1990

1995

2000

198

199

195 197

15%

196

197

10%

5%

0% 1948

1953

1958

1963

1968

Required exports, contemporaneous interest rate

1973

1978

1983

1988

1993

1998

Required exports, 10 years average interest rate

After the mid 1970s the burden of the debt to Argentina rises from levels around 5 per cent to magnitudes of the exports ratio between 10 per cent and 20 per cent. There is an international economic phenomenon behind this evolution; a rising exports intensity of the world level together with an abundance of funds, when the Argentine exports grew at the end of the seventies and in the 1990s, was perceived as allowing for higher “admissible” debt levels. With these numbers, if there is long-run level of the export ratio of about 10 per cent of GDP, in order to fulfill the solvency condition, Argentina should increase its export to GDP ratio in between 50 per cent to 100 per cent of its “structural” level. Exercise 3. A negative gap between observed exports and required exports indicates vulnerability. To explore further the role of exports in external solvency, exports performance is compared to required exports, using Equation (10). The imports to GDP ratio in this exercise are the observed contemporaneous values at time t. Writing external vulnerability as a function h of the gap, exports reduce vulnerability.

72

Sergio V. Barone and Alberto M. Díaz Cafferata

h( xo  xR ( D)); h '  0

V

(11)

Figure 6: Argentina 1948-2004 The required exports to GDP ratio 20%

a) Lower limit for the interest rate 2002

15% 1953 10%

1976 1963

1950

1988

1973

1995

5%

0%

1958

1948 1953

1963

1968

1973

1978

1983

1988

1993

1998

2003

-5%

Gap, contemporaneous interest rate

Gap, 10 years interest rate

-10%

b) Upper limit for the interest rate 1953 5%

1973 1950

1963

1976

1995 1988

0% 1948

1953

1958

1963

1968

1973

1978

1983

1988

1993

-5%

-15%

-20%

Gap, 10 years average interest rate

2003

1998

1993

1988

1983

1978

1973

1968

1963

1958

GAP, 1948 - 2004 5% -5% -15% -25% -35% -45% -55% -65%

1948 1953

-10%

Gap, contemporaneous interest rate

1998

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

73

The gap between observed and required exports provides additional information about the solvency condition. When in the historical series the value of observed exports have been lower than required exports our hypothesis is that it would be giving a signal of future difficulties to pay. In this particular exercise the “m” values are the actual imports ratio the interested reader may check that when the exports gap ( xo  x R ) is the same as (tbo  tbR ) gap. Figure 6a depicts the gap for the lower limit rate of interest that shows in the years before the reversal a negative gap in seven out of eight cases; and the more recent reversals had been preceded by a larger number of years with negative value of the gap. When we include the risk (Figure 6b), the gap using the contemporaneous or the ten years forward average is almost continuously in the order of magnitudes of 5 per cent of GDP; this is larger than the previous case, and shows that the economy of Argentina had solvency problems since the "debt crises". The values of negative gaps may also be interpreted as telling that the external debt was “excessive”. Exercise 4. Required exports compared with trend exports Figures 7a and 7b show the required exports, together with the exports to GDP ratio trend. The exercise is similar to the one presented in exercise 3, although this case considers a trend value for the exports to GDP ratio, which reflects an approximation to the structural trade openness of the economy, and this allows us to focus on the long run behavior of the variables. The footnotes statistics of each figure show that in 42 per cent of the years, the solvency condition is not reached in a structural way (Figure 7a), while in Figure 7b the percentage is actually higher (63 per cent). On the other hand, when the gap in both cases is considered as an average, a positive value is obtained for case (a) and a negative for the other one, implying that when the interest rate with risk included is used in the the estimation, the long run solvency condition is not fulfilled as a rule in the country. The analysis of the restriction to manage external financing shall include the understanding of the causes and the process of evolution of the external crises (as represented in this paper by the episodes of reversal) as well as a more comprehensive perspective of the external flows in relation with the financing of investment. The decades of 1980s and 1990s provide interesting evidence for further research on the matter. In the 1980s the jump in the interest rate at the time of the debt crises is reflected in Figure 7b by a rise in the bur-

74

Sergio V. Barone and Alberto M. Díaz Cafferata

den –as seen in Figure 4- and an excess of the required relative to trend exports; this situation with the upper limit for the rate of interest, seems to be the relevant one due to the magnitude of the country risk in these two decades. The gap helps to explain the lack of supply of external financing in the eighties (with an episode of reversal only at the end of the decade in 1998). In the 1990s there is again a period with an excess of required over trend export, with a very different interpretation; in this case the gap was possible thanks to the expectation that a structural change towards a more open economic was a permanent change caused by the policy reforms associated with the “convertibility plan”. In other words, the forward looking trend exports were about the historical experience, providing opportunity for CA deficits, and the reversion of this optimistic expectation at the end of the decade precipitate the economy into the default in 2001-2002. Figure 7: Argentina 1948-2004 Requiered exports, trend exports and the gap between requiered exports and trend exports (ratios of GDP) (to be continued)

a) Lower limit for the interest rate Ratio of number of years with negative gap and total number of years= 0,42 Average of the gap along the whole period: 0,0028 Variation Coefficient = 8,33%

45%

35%

1950 2002 1973

1953 1963

25%

1995

1976 1988

15%

5% 1948

1953

1958

1963

1968

1973

1978

1983

1988

1993

-5%

-15%

-25%

Required exports (10 years average for the interest rate) Observed exports trend GAP (Observed minus required exports)

1998

2003

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

75

Figure 7 (cont.): Argentina 1948-2004 Requiered exports, trend exports and the gap between requiered exports and trend exports (ratios of GDP)

b) Upper limit for the interest rate Ratio of number of years with negative gap and total number of years= 0,63 Average of the gap along the whole period: -0,034 Variation Coefficient = 63,66%

45%

35%

1950

2002

1953 0,35

1988

1973

25%

1995 1976

15%

5% 1948

1953

1958

1963

1968

1973

1978

1983

1988

1993

1998

2003

-5%

-15%

-25%

Required exports (10 years average for the interest rate) Observed exports trend GAP (observed minus required)

V. Conclusions and policy implications

External savings have a potential to help growth, as Argentina itself experienced in the five decades between the end of the XIXth Century and the first part of the XXth Century. If an economy may borrow for a period of time and accumulate capital, and at the same time generates the capacity to pay external debt in foreign currency, the effect of international financial flows may be positive as several countries have experienced in the XXth Century. Why cannot Argentina sustain CA deficits generated by the use of external savings if other countries can? We have argued that focusing attention on the role of exports and a long-term horizon exports performance appears as critical for the capacity to pay the external debt. Substantial policy lessons may be learned from this perspective. An escape from the historical long-run deceleration of growth was

76

Sergio V. Barone and Alberto M. Díaz Cafferata

expected by shifting from import substitution to a strategy more open to external capital and commercial flows. But the automatic beneficial effects of openness should not have been taken for granted. One critical issue has been the break in the characteristics of external indebtedness. As our empirical exercises show, there was a marked increase of three or four times in the burden both in terms of trade surplus ratio required and in terms of the exports ratio; the importance of these variables as indicators of solvency is suggested by the observation that reversals in 1973, 1988, 1995 and 2002, were preceded by a rising burden or coincident with a peak. The large magnitude the burden is seen to reach after only a few numbers of years of CA deficits, may be a dangerous obstacle to economic growth especially for an economy with a low degree of openness like Argentina. A decision of opening the economy and borrowing as in the case of Argentina, requires in consequence a close control of the consistency of expected exports proceedings and required debt payments along time. In this sense a first comment is about the fact that the evolution of the burden of the debt and of the export gap for Argentina is highly sensitive to fluctuations in the rate of interest: either movements in the international free of risk rate or in the risk premium that in both case may be exogenous and independent of the domestic economy. Another feature that appears in the exercises is that the long-run solvency seems to be tied to the capacity to pay in terms of the long-run exports performance. The critical relationship between the real and the financial spheres of the economy is created by the obligation to pay in the future associated to the issuing of debt, that must certainly be matched with the expected future trends of exports. There is no indication at this point of our work that the association between exports and solvency found for Argentina can be used as indicator of the timing of external crises. A related point is that international experience shows that the presence of structural characteristics of the economy determine a long run degree of openness that may be difficult to elevate with only trade liberalization policies. Then, a first question is if the export response to a change in policy will be to put the economy in the expected trend. For a borrowing country to expect future exports to grow above the historical experience, both a higher export growth along the new trend and more stable export proceedings may be necessary. A long and complex transition period of trans-

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

77

formation may be necessary. In other words: a structural transformation in trade. Therefore: ex ante the uncertainty must be incorporated formally in the estimation of expected future exports. Solvency is enhanced and errors are reduced by higher export growth along trend, and lower variability is obtained by a change in the productive structure, as well as sectoral and geographic diversification. Persistent deficits are possible if structural exports are consistent with long-run external savings, and changing permanently the exports to enhance the borrowing capacity to GDP ratio and the sectoral and geographical structure of exports requires adequate longrun incentives and financing of investment in tradeable sectors. We have discussed in this paper only the role of total exports; but also the cyclical variability and possibility of shocks shall be included in the expectation of future exports proceedings. Even along the path of trend exports consistent with solvency, problems may arise from fluctuations in the exports value causing temporary gaps between actual and required export value. Developing economies specialized in the production and export of primary goods and subject to terms of trade fluctuations may be specially vulnerable (Milessi-Ferreti and Razin, 1998). Given the high costs of external crises, we conclude that there is indication that it is advisable for the Argentine economy to keep external indebtedness at levels carefully controlled by the consistency of the burden in terms of long-run exports performance, rather than moving in the limits suggested by evolutions of exports that may be only transitory, and also maintaining prudential margins. The eventual sacrifice of current consumption or future growth is justified by the observed sensitivity of the economy to external shocks and fluctuations. A related lesson is the potential gains to be reached by a more regular growth that makes the expected capacity to honor external debt payments more predictable.

78

Sergio V. Barone and Alberto M. Díaz Cafferata

References

Ávila, Jorge (2000), Riesgo argentino y perfomance macroeconómica, Universidad del CEMA, Buenos Aires, Argentina. Bleaney, Michael (2004), Argentina´s Currency Board Collapse: Weak Policy or Bad Luck?, The World Economy27(5), 699-714. Brodersohn, Mario (1969), Estrategias de estabilización y expansión en la Argentina: 1959-1967, in: Ferrer, Aldo et al. (eds.), Los Planes de Estabilización en la Argentina, Paidos, Buenos Aires, Argentina. Calderón, César, Chong, Alberto, Loayza, Norman (1999), Determinants of current account deficits in developing countries, Banco Central de Chile Documento de Trabajo 51. Calvo, Guillermo A. (2005), Emerging Capital Markets in Turmoil, Bad Luck or Bad Policy?, The MIT Press 2005, Cambridge, Mass. Cavallo, Domingo (2003), How did the Foreign Debt Trigger the Argentine Crisis?, Seminar at Real Institute Elcano, Madrid. Cohen, Daniel (1985), How to evaluate the solvency of an indebted nation, Economic Policy 1 (1), 140-167. Consejo Técnico de Inversiones, Anuario, Several issues. Debelle, Guy, Faruqee, Hamid (1996), What determines the current account? A cross sectional and panel approach, IMF Working Paper WP/96/58. Díaz Cafferata, Alberto M. (1996), How much does a liberalized economy trade?, Primer seminario académico sobre integración económica y comercio, FCE, Universidad Nacional de Córdoba. Díaz Cafferata, Alberto M., Fornero, Jorge (2003), Tendencias y quiebres del grado de apertura exportadora de Argentina y el marco internacional, 1884-2002, XL Reunión de la Asociación Argentina de Economía Política, : www.aaep.org.ar (last visit: May 2007). Díaz Cafferata, Alberto M., Kohn, David, Resk, Elena (2005), Motivo crecimiento y la evolución de la cuenta corriente argentina de largo plazo: 1884-2003, XL Reunión de la Asociación

Rigidities in Openness, Export Performance and Indicators of External Solvency in Argentina

79

Argentina de Economía Política, : www.aaep.org.ar (last visit: May 2007). Domínguez, Kathryn M.E., Tesar, Linda L. (2005), International Borrowing and Macroeconomic Performance in Argentina, NBER Conference on Capital Flows, revised version. Feldstein, Martin, Horioka, Charles (1980), Domestic Savings and International Capital Flows, Economic Journal 90 (6), 314-329. Ferreres, Orlando (2005), Dos siglos de economía Argentina (19102004): Historia argentina en cifras, El Ateneo, Fundación Norte/Sur, Buenos Aires, Argentina. Gandolfo, Giancarlo (2001), International Finance and Open Economy Macoeconomics, Springer Verlag, Berlin. Gurtner, Francois J. (2004), Currency Boards and Debt Traps: Evidence from Argentina and Relevance for Estonia, The World Economy 26 (2), 209-228. Halevi, Nadav (1971), An empirical test of the ‘Balance of Payment Stages’ Hypothesis, Journal of Internacional Economics 1(1),; 103-117 International Monetary Fund IMF, International Financial Statistics, : www.imf.org (last visit: May 2007). International Monetary Fund IMF (1993), Balance of Payments Manual, 5th Edition, Washington. Izquierdo, Alejandro (2002), Sudden Stops, the Real Exchange Rate and Fiscal Sustainability in Argentina, The World Economy 25(7), 903-923. Kaminsky, Graciela L., Lizondo, Saul, Reinhart, Carmen M. (1998), Leading indicators of currency crises, IMF Staff Papers (45) 1. Kindelberger, Charles P. (1968), International Economics, 4th Edition, Irwin, Homewood, Ill. Krugman, Paul (1979), A model of Balance of Payments Crises, Journal of Money, Credit and Banking 11(3), 311-325. Lanteri, Luis N. (2008), Desarrollos teóricos sobre la cuenta corriente y su aplicación al caso argentino, in: A. Díaz Cafferata (ed.), Progresos en economía internacional, Asociación Argentina de Economía Política. Buenos Aires. Manasse, Paolo, Roubini, Nouriel (2005), Rules of thumb of sovereign debt crises, International Monetary Fund WP/05/02.

80

Sergio V. Barone and Alberto M. Díaz Cafferata

Milesi-Ferretti, Gian María, Razin, Assaf (1996), Current Account Sustainability: Selected East Asian and Latin American Experience, NBER Working Paper 5791. Milesi-Ferretti, Gian María, Razin, Assaf (1998), Current Account Reversals and Currency Crises: Empirical Regularities, NBER Working Paper 6620. Mussa, Michael (2002), Argentina and the Fund. From Triumph to Tragedy, Institute for International Economics, Mimeo. Obstfeld, Maurice, Rogoff, Kenneth S. (1997), Foundations of International Economics, MIT Press, Cambridge 1997, Mass. Reinhart, Carmen (2002), Default, currency crises and sovereign credit ratings, The World Bank Economic Review 16 (2), 151170. Reinhart, Carmen M., Rogoff, Kenneth S., Savastano, Miguel A. (2003), Debt Intolerance, NBER Working Paper 9908, : http://www.nber.org/papers/w9908 (last visit: May 2007). Rodríguez, Carlos Alfredo (1986), La deuda externa argentina, CEMA 54, Buenos Aires, Argentina. Saksonovs, Sergejs (2006), The Intertemporal approach to the Current Account and Currency Prices, Darwin College Research Report, Cambridge University DCRR 005.

Fernando Zarzosa Valdivia 1

Real Exchange Rate Movements, Dutch Disease and Functional and Sectoral Income Distribution Abstract In a theoretical GEM model, we show how exogenous productivity, factor endowments, terms of trade and debt shocks, impact on real exchange rates, factors prices and production and consumption levels of a small economy. In addition, we evaluate possible income distribution effects and macroeconomic misalignments of those exogenous shocks, like the Dutch disease. This set up is perfectly suited to the characteristics of the main Latin America economies, Argentina, Brazil, Chile and Mexico were exogenous shocks an exchange rate policies need to be carefully gauged in terms of their impact on real exchange rates, the structure of the economy and the development of the tradable sector. I. Introduction “Real Exchange rates are key relative prices in international finance” (Chinn 2005, p.1); they influence the allocation of resources and expenditure. Dwyer and Lowe (1993, p.1) emphasise that “there is surprisingly little agreement concerning both how to measure and interpret movements in real exchange rates”. Largely, disagreements stem from the fact that the real exchange rate definition may vary according to the context in which it is used. This paper distinguishes two strongly related definitions: (a) The structural real exchange rate and (b) The Purchasing Power Parity (PPP) real exchange rate. Equations (1) and (2) of Table 1 define these real exchange rates.

1

Financial support from the Alfa Programme (Project AML/B7-31197/0666/II-0058-FA- of the European Union) and grant (TEW/ FA060008) from Antwerp University are greatly acknowledged. I also would like to thank Professor Joseph Plasmans for his useful comments, discussions and recommendations.

82

Fernando Zarzosa Valdivia

The structural real exchange rate is based on the relative tradable to non-tradable price and it measures the value of a unit of tradable goods in terms of non-tradable goods.2 According to Driver and Westaway (2004, p.17), this definition is "appropriate for assessing the real exchange rate within countries". However, Dwyer and Lowe (1993, p.1) argue that “despite the domestic role afforded to domestic relative prices in theoretical models of open economies, practical difficulties associated with the estimation of such prices have resulted in widespread adoption of the PPP based measure of the real exchange rate”. The PPP or nominal definition is based on deviations from purchasing power parities between countries. It is defined as the relative price between foreign and domestic goods and it measures the necessary units of domestic goods to buy a unit of foreign goods. There is no single dominant approach to modelling equilibrium real exchange rates. Different authors have used methods ranging from the purely statistical to the purely theoretical but there has been less theoretical work on equilibrium real exchange rates; in part because of its lack of attractiveness and also because of its complexity. All approaches consider real exchange rates as one endogenous variable in a complete macroeconomic system where macroeconomic fundamentals are the key driving variables underlying movements in the exchange rate. Also, different assumptions about the transmission mechanism can potentially have different implications for the behaviour of the equilibrium real exchange rates. Thus, real exchange rates may vary because of differences in consumer preferences or production technologies, the existence of differentiated products, non-tradable goods and market imperfections.3 In any case, the resulting equilibrium real exchange rate might depend on a variety of additional factors including: productivity differentials, terms of trade, factor 2

3

This definition has evolved from the theoretical model for a dependent economy developed by Salter (1959) and Swan (1960; 1963), authors of the so-called “Australian model of balance of payment theory”. “Firms may set prices in their domestic currency (producer currency pricing) and fully pass through any change in the exchange rate into the price they charge in foreign markets. Alternatively, firms may set prices in the buyers’ currency and, at least in the short run, changes in the exchange rate will have no impact in these prices” (Driver and Westaway, 2004, p.19).

Real Exchange Rate Movements, Dutch Disease and Functional and Sectoral Income Distribution

83

endowments, debt shocks, economic policies and windfall discoveries of new resources. The Balassa-Samuelson model (1964) relates productivity shocks to real exchange rate movements. Calderon (2000) and Gay and Pellegrini (2003) apply a general equilibrium model (GEM) to a small economy with tradable goods exogenously given and only affected by productivity shocks. They find that productivity and terms of trade shocks impact on the real exchange rate. Obstfeld and Rogoff (2004) find a similar link applying a GEM model to a world with two countries; they suppose that labour and capital cannot move freely across sectors and that labour productivity and tradable production only change due to productivity shocks. MacDonald and Ricci (2002) apply a GEM model to two countries with market imperfections, labour mobility between sectors and linear production functions; factor productivities are again constant and only affected by productivity shocks. Their results are similar to the Balassa-Samuelson results but their model also explains the law of one price deviation of tradable prices. Like these real exchange rate models, many others assume factor productivities independent of output levels. Exogenous shocks, however, not only impact on real exchange rates but also on factor prices and all goods’ consumption and production. As a result, they may generate macroeconomic misalignment, such as “Dutch disease”. The term Dutch disease refers to the de-industrialization of the manufacturing sectors of an economy as a consequence of windfall discoveries of new resources and productivity, terms of trade, factor endowments and debt shocks.4 In a GEM model for a small open economy, we assume that the tradable sectors operate with diminishing returns to scale and exogenously given total factor productivities (TFP). We aim to explain how exogenous shocks (e.g. productivity improvements, windfall discoveries of new resources and factor endowments, terms of trade and debt shocks) impact on real exchange rates, factor prices and 4

Originally, “the term Dutch disease refers to the adverse effects on Dutch manufacturing of the natural gas discoveries of the 1960s, essentially through the subsequent appreciation of the Dutch real exchange rate” (Corden, 1984, p.359). This process of deindustrialisation of the existing manufacturing was attributed to the upward pressure that the Dutch energy discovery placed on the Guilder and the wage rate (O’Toole, 1998, p.1).

84

Fernando Zarzosa Valdivia

production and consumption levels of an economy. In addition, we evaluate possible income distribution effects and macroeconomic misalignments of such exogenous shocks. Further, we analyse the influence of the initial income distribution and debt-GDP ratios on the magnitude of the exogenous shocks impact on an economy. We also aim to explain why rapidly growing countries are those with larger manufacturing sectors. This model fits in with macroeconomic models of real exchange rate and Dutch disease models. The remainder of the paper is organized as follows. Section II introduces some stylized facts. Section III provides a GEM model for a small economy. It shows a) how exogenous shocks impact on real exchange rates and production, consumption and social welfare levels, b) possible Dutch disease originated by exogenous shocks, c) the influence of income distribution and debt-GDP ratios in the magnitude of the exogenous shocks impact on an economy and d) the income distribution effects of exogenous shocks and the variability of the structural real exchange response to exogenous shocks. An additional sub-section benchmarks our model results to the Dutch disease model results and incorporates imperfections in the non-tradables market. Finally, in Section IV we provide some conclusions and future research suggestions. II. The empirical contex: Some stylized facts from Argentina, Brazil, Chile and Mexico Latin America’s economic prospects heralded new promises in the early 1990s, as ambitious stabilization and reform programs were introduced. These programs were broadly consistent with the socalled Washington Consensus. They reflected a broad shift away from the interventionist and inward looking policies followed in the past. To promote transparency and credibility, monetary policy was typically constrained by a commitment to a fixed exchange rate. Other structural reforms were focused on increasing the role of market forces, e.g. removal of currency restrictions and liberalization of trade and capital inflows. The region’s economic performance and growing productivity in the first half of the 1990s appeared to validate initial expectations. However, signs of fragilities and contagious vulnerability become evident with Mexico’s “tequila” and the Asian and Russian Crises. Following Baldi and Mulder (2004), Argentina, Brazil, Chile and Mexico were unable to accommodate to the constraints of a

Real Exchange Rate Movements, Dutch Disease and Functional and Sectoral Income Distribution

85

fixed exchange rate.5 Ultimately, they abandoned their “fixed” or “quasi fixed” exchange rate regimes and fell into a severe economic and debt crisis, Mexico in 1994, Brazil in 1999 and Argentina in 2002, Chile did not fell into a severe crisis but suffers a slowdown of economic growth and investment in 1997/1998. Being mostly price-takers, tradable goods producers were forced to stem price increases of their products due to the fixed exchange rates. In contrast, non-tradable producers faced no international competition and had fewer incentives to control their output prices. The combined result was a steep rise of the relative price of nontradables to tradables, see Figure 1. Except in Chile, this effect was exacerbated by the entry of substantial short-term inflows of portfolio capital. These contributed to higher final demand, which in turn caused increasingly higher prices of non-tradables compared to tradables. With the abolishment of the fixed exchange rate regimes, the currencies strongly depreciated, capital fled and relative price trends reverted, at least for some time.6 According to Baldi and Mulder (2004, p.23-36), other variables 'empirically explaining' relative price movements are the BalassaSamuelson effect (for all countries), government expenditures (Brazil and Mexico) and terms of trade (Chile and Mexico). These variables and the exchange rate regime, via their impact on relative prices, altered the inter- and intra-sectoral composition of these countries. Figure 2 shows that the share of the tradable sector in employment and GDP fell more than proportionally during the 1990s. Although the share of agriculture and mining seems mostly unaffected by exchange rate regimes, the size of manufacturing was negatively (positively) affected by fixed (flexible) regimes. The authors also argue that exogenous shocks and exchange rate policies should be carefully gauged in terms of their impact on real exchange rates, the structure of the economy and the development of the tradable sector.

5

6

In all four countries fixed exchange rates were introduced to different degrees in the late 1980s-early 1990s in order to overcome hyper- or double-digit inflation. Recall that when a fixed regime is adopted, it is difficult to create enough flexibility elsewhere in the economy in order to compensate for this source of rigidity.

86

Fernando Zarzosa Valdivia

Figure 1: Ratio of price indices of non-tradables to tradables*

* PPI for tradables and CPI for non-tradable. Source: Baldi and Mulder (2004, p.21).

Figure 2: Share of tradable sector in Employment, 1990-2002

Source: Baldi and Mulder (2004, p.24).

III.The model A. Behavioural relationships and macroeconomic conditions We decided to use a theoretical Salter-Swan neoclassical open and micro-founded macro model,7 where representative consumers and producers are assumed to be rational. We assume a world with three goods: two of these goods are supposed to be tradable goods and 7

It allows us focusing our research on a dependent economy like Argentina and others developing countries.

Real Exchange Rate Movements, Dutch Disease and Functional and Sectoral Income Distribution

87

the other one is assumed to be a non-tradable good. “Tradable goods are those with prices determined in the world markets. They consist of exportables, of which the surplus over home consumption is exported; and importables,8 of which the deficiency between consumption and home production is imported,” (Salter, 1959, p. 226). We assume a two-level constant elasticity of substitution (CES) utility function, where exportables, importables and non-tradables are assumed to be imperfect substitutes in consumption; see Equation (3) in Table 1. Our representative consumer maximises her utility facing a restriction given by her budget constraint. The Equations of her optimization problem are set out in Table 1. The first order conditions for utility-maximizing consumers are given by Equations (5) and (6), which define non-tradable and exportable goods demand, respectively. The relative non-tradable to tradable demand depends negatively on the structural real exchange rate and it suggests the existence of a consumption trade-off between tradable and non-tradable goods. The consumption trade-off between exportable and importable goods is reflected by the relative importable to exportable demand and it depends negatively on the international terms of trade. Plus, the tradable expenditure share depends negatively on the structural real exchange rate, see Equation (7). The economy is divided into three sectors, exportables (X), importables (M) and non-tradables (N). These sectors use labour and capital in their production process. Following Rodrik (2006), we assume linear technology for the non-tradable sector but Cobb Douglas for the tradable sectors.9 The representative producer of each sector maximises profits10 taking as given the technology, the 8

9

10

In this model, like in Salter-Swan model, “home produced importables are treated as a special class of exportables: goods which could be sold on world markets, but in fact are not, for we should only have to buy them back”. In this model therefore, production factors are perfect substitutes in the non-tradable sector but imperfect complementary in the tradable sectors. Further, factor productivities are not exogenously given as in Gay and Pellegrini (2003), Obstfeld and Rogoff (2004), MacDonald and Ricci (2002), new trade theory models and others’ models. The profit function is given by the difference between the revenues by selling goods and the cost of producing them (factor rewards).

Fernando Zarzosa Valdivia

88

prices of the goods produced by them and the resource endowments. The Equations of those maximisation problems are set out in Table 2. First order conditions for profit maximizing producers determine the supply of exportable and importable goods; Equation (14) and (15), respectively. Resource allocation between tradable and non-tradable goods depends on the structural real exchange rate, while resource allocation within the tradable sector depends on the terms of trade. Thus, the supply of each tradable good is positively related to the structural real exchange rate, while terms of trade improvements increase the supply of exportable goods but decrease the supply of importable goods. Table 1: Real Exchange Rates and Representative Consumer’s Maximization Problem (to be continued) Structural Real Exchange Rate:

estruct

PT PN

(1)

ePPP

S P* P

(2)

PPP’s Real Exchange Rate: Consumers’ Utility Function: 1 1 E ­° ½E E ° U U º 1- E ª 1- U U 1- U 1- E U C ®J « ª¬G C X  (1- G ) CM º¼ »  (1- J ) C N ¾ ¬ ¼ ¯° ¿° Consumers’ Budget Constraints: PX CX  PM CM  PN CN E Non-tradable goods demand:

(3)

(4)

1

CT CN

1 § J · § PT · E 1 § J · E 1 e  ¨ 1 J ¸¨ P ¸ ¨ 1  J ¸ struct © ¹© N ¹ © ¹

(5)

Exportable goods demand 1

CX CM

1

§ G · § PX · U 1 § G · E 1 ¨ ¸¨ ¸ ¨ ¸ (TT ) © 1  G ¹ © PM ¹ ©1G ¹

(6)

Real Exchange Rate Movements, Dutch Disease and Functional and Sectoral Income Distribution

89

Table 1 (continued): Real Exchange Rates and Representative Consumer’s Maximization Problem Consumers’ tradable expenditure share: E

ET

PT CT

E

PC

§ P · E 1 J¨ T ¸ ©P¹

(7)

Tradable goods Index: 1

CT

ª¬G 1U CXU  (1  G )1 U CMU º¼ U

(8)

Tradable prices index:

PT

U U ª º « G PXU 1  (1  G ) PMU 1 » ¬« ¼»

U 1 U

(9)

Consumer price index (full basket of commodities price):

P

E E ª E 1 E 1 º « J PT  (1  J ) PN » ¬« ¼»

E 1 E

(10)

where:

PT and PN are the domestic prices of tradable and non-tradable goods, respectively P and P* are the foreign and domestic general price level, respectively S is the nominal exchange rate U (or C) is the consumption utility function (or total consumption CES index) CX, CM and CN are exportables, importables and non-tradable goods consumption levels, respectively Ȗ is the preference weight for tradable goods respect to non-tradable goods į is the preference weight for exportables respect to importables ȕ and ȡ are elasticity parameters; ȕ, ȡ < 1 0 < 1/(1-ȕ) < ’ is the inter-class elasticity of substitution between tradable and non-tradable goods, 0 < 1/(1-ȡ) < ’ is the intra-class elasticity of substitution between tradable goods, PX, PM and PN are the prices of exportable, importable and non-tradable goods, respectively E and ET are the representative consumer’s total and tradable expenditure, respectively TT ( PX PM ) are the international terms of trade (relative exportable to importable price) PT and CT are price and consumption indices of tradable goods

Fernando Zarzosa Valdivia

90

In general, our economy is assumed to be a perfectly competitive economy with international price taking behaviour. Prices can be thought of as perfectly flexible and the current account at a ‘sustainable’ level in the sense that it is consistent with eventual convergence to the stock-flow equilibrium. In addition, we assume full employment of a constant supply of labour and capital; furthermore, labour and capital are assumed to be mobile between all sectors (see equations (16) and (17) in Table 2). The equilibrium of an economy with perfectly competitive markets and full employment of its production factors implies no extraordinary profits. Thus, the income generated by all sectors during the period analysed (Gross Domestic product, GDP) must be equal to the factors rewards during that period plus the value of the endowment of tradable goods (see Equation (18) in Table 3). This condition allows us to define the sectoral and functional income distribution of our small economy, see Equations (19) and (20) in Table 3. Table 2: Producers’ Maximisation Problem (to be continued) Exportable sector technology:

X 0  Z X LIXX K \X X

(11)

M 0  Z M LIMM K \MM

(12)

Z N ( Z N L LN  Z N K K N )

(13)

X ( LX , K X ) Importable sector technology:

M ( LM , K M ) Non-tradable sector technology:

N ( LN , K N ) Supply of exportable goods:

X

X0  IX

IX \ X IX \ X ª § IX § 1 · IX \ X § PX · «Z X  estruct ¨ ¨ ¸ ¨ ¸ ¨ ZN « PT ¹ ZN ¹ © © © L ¬

· ¸ ¸ ¹

§ \X ¨ ¨ ZN © K

\X

· ¸ ¸ ¹

º » » ¼

1 1IX \ X

(14)

Supply of importable goods:

M M0  IM \ M IM \ M ª § IM § 1 · § · «ZM  estruct IM \M ¨ PM ¸ ¨ ¨ ¸ ¨ ZN « © PT ¹ © ZN ¹ © L ¬

IM

· ¸ ¸ ¹

§ \M ¨ ¨ ZN © K

\M

· ¸ ¸ ¹

º » » ¼

1 1IM \ M

(15)

Real Exchange Rate Movements, Dutch Disease and Functional and Sectoral Income Distribution

91

Table 2 (cont.): Producers’ Maximisation Problem Full employment condition:

L

LX  LM  LN

(16)

Full capital employment condition:

K

K X  KM  KN

(17)

where: X, M and N are the production levels of exportable, importable and non-tradable sectors, respectively, X0 and M0 refers to exogenously given fungible endowments of exportable and importable goods, respectively

IX

and \ X , I M and \ M are the exportable and importable output elasticities, respectively, lying between zero and one, ZX, ZM and ZN are the total productivity of the production factors employed in exportable, importable and non-tradable sectors. Z N L and Z NK are the specific productivities of the labour and capital factors, respectively, in the non-tradable sector, Li and Ki are the labour and capital employment for sector i, respectively L and K are the labour and capital endowments of this economy, they are given exogenously w and r are the domestic wage and the interest rate, respectively,

Consumers in open economies can consume more tradable goods than their economy produces, but the consumption of nontradable goods is always equal to its domestic production. In other words, tradable and total consumers' expenditure must fit the conditions imposed by the market clearing condition and the current account definition. Additionally, the current account definition and the zero profit, full employment and market clearing conditions impose a macroeconomic restriction on the consumers’ tradable share expenditure; see Equation (21) in Table 3. The income distribution, debt-GDP and current account-GDP ratios indicate the economic structure of our economy since they measure the size of each tradable sector, the size of the tradable sector as a whole, the size of the debt in terms of the GDP and the labour and capital shares in the GDP. In this model, the debt-GDP and Current account-GDP ratios are assumed exogenously given. If the preference weights for tradable goods are the same between different countries, the PPP real exchange rate will change

Fernando Zarzosa Valdivia

92

because of movements in the structural real exchange rates of these countries and the relative prices of traded goods; see Equation (22) in Table 3. Table 3: Macroeconomic Conditions and Real Exchange Rate Relationships Zero profit condition:

PX X  PM M  PN N

GDP

wL + rK  PX X 0  PM M 0 Sectoral income distribution: PX X  PM M

TT

, TX

PX X

, GDP GDP PX X 0 PM M 0 PM M , TX , TM TM GDP GDP GDP Functional income distribution: T L wL / GDP, T K rK / GDP Conditioned consumers’ tradable share expenditure: ET PX X  PM M  r * F  CA 0

E

(18)

(19)

0

PX X  PM M  PN N  r * F  CA

TT  r * f  ca

(20)

(21)

1  r * f  ca Real Exchange Rate relationships:11

eˆPPP

Sˆ  Pˆ *  Pˆ * (1  J )  eˆstruc  eˆstruc  Sˆ  PˆT*  PˆT

(22)

where: r*, F, CA, f(=F/GDP) and ca(=CA/GDP) are the international interest rate, net foreign asset position, current account, debt-GDP and current account-GDP ratios, respectively, T L , T K , T T , T X , T M , T X0 and T M0 are labour / total income, capital / total income, tradable / total income, exportable / total income, importable / total income, exportable endowment value / total income and importable endowment value / total income ratios, respectively. Pˆ * and Pˆ are the foreign and domestic inflation rates, respectively

11

We obtain Equation (22) applying total differential to Equation (9) and (10). Each country faces different Equations (9) and (10).

Real Exchange Rate Movements, Dutch Disease and Functional and Sectoral Income Distribution

93

In addition, if the law of one price (LOOP) holds for the price of tradable goods, then “trends in the relative PPP’s real exchange rate will arise only because of movements in the relative prices of goods within countries” (MacDonald, 1998, p.17) but, “the existence of non-tradable goods will not be sufficient to cause persistent PPP's real exchange rate divergences unless the relative price inflation of tradable to non-tradable goods differs between countries” (Driver and Westaway, 2004, p. 30). In this paper, we assume that the LOOP holds for the tradable prices.12 B. Real exchange rates, Dutch disease and exogenous shocks Production factors and consumers’ expenditures are allocated according to the signals of the structural real exchange rate and terms of trade. In equilibrium, consumer and producer interaction fits the micro- and macroeconomic conditions. Appendix I describes how a non-negative equilibrium is reached for our economy. In addition, it may be useful to put together some of the static comparative results we have derived schematically in table 4.13 Next, we explain those relationships. TFP improvements in the exportable sector diminish the relative cost of producing exportables. This generates the resource movement effect and spending effect, which are associated with the de-industrialisation of the importable sector. The first effect means that increased productivity pushes factor prices up, bidding labour and capital out of the importables and non-tradables sector. The resulting reduction of output in importables is called direct de-industrialisation. Full employment conditions imply that factor prices increments increase GDP and demand for all goods. Non-tradable demand increments push factor and non-tradable prices upwards. This also has the effect of bidding production factors out of both of 12

13

The LOOP may not hold if there are market imperfections or nominal rigidities. The new trade theory, through local currency pricing (LCP) and pricing to market (PTM) models, emphasises that “the source of systematic movements in PPP real exchange rate seems to come from movements in the relative prices of traded goods rather than in the internal price ratio” (MacDonald and Ricci, 2002, p.3). Engel and Rogers (1996) postulate that sticky nominal prices also explain movements on the relative prices of traded goods. Appendix II sets out the equilibrium movement Equations for real exchange rates and GDP. Other formal results are available upon request from the authors.

Fernando Zarzosa Valdivia

94

the tradable sectors (spending effect). This is the so-called indirect de-industrialisation of the importable sector. The total impact on non-tradable output is generally ambiguous but in this model, these TFP shocks reduce it. Further, TFP improvements in any tradable sectors appreciate both real exchange rates and increase real wages and interest rates measured in terms of the full basket of commodities. TFP improvements in the importable sector have similar effects to the ones formerly explained, but in this case, de-industrialization will perform in the exportable sector. Table 4: Response of endogenous variables to exogenous shocks exogenous shocks TFP and TFP TFP specific factor capital factor shocks shocks Int. productivity inflows endowexport- importprices shocks nonchanges ments able able tradable sector sector sector endog. ZX variables

ZM

tradable goods endowments

Z N Z NL Z NK

L

K

TT

r * df - dca

X0

M0

estruct

-

-

+

+

+

+

+

?

-

-

-

e*PPP

-

-

+

+

+

+

+

?

-

-

-

X

+

-

-

?

?

+

+

+

-

+

-

M

-

+

-

?

?

+

+

-

-

-

+

N(=CN)

-

-

+

+

+

+

+

?

+

+

+

CX

+

+

-

?

?

+

+

?

+

+

+

CM

+

+

-

?

?

+

+

?

+

+

+

(C)

?

?

+

?

?

+

+

?

+

+

+

(w/P)**

+

+

+

?

-

-

-

?

+

+

+

(r/P)**

+

+

+

-

?

-

-

?

+

+

+

Note: *Domestic exogenous shocks have similar impact on PPP and structural real exchange rates, see Equation (22). ** w/P and r/P are the real wages and interest rates, respectively, measured in terms of a full basket of commodities.

TFP improvements in the non-tradable sector generate resource movement effects and spending effects favourable to this sector. As a result, the non-tradable production increases but both tradable

Real Exchange Rate Movements, Dutch Disease and Functional and Sectoral Income Distribution

95

production and consumption decrease. In addition, these TFP shocks depreciate the real exchange rates and increase real factor prices measured in terms of the full basket of commodities. Labour biased productivity shocks in the non-tradable sector generate resource movement effects favourable to this sector but pushing wages upwards and interest rate downwards. As a result, the supply of non-tradable goods increases. It pushes real exchange rate upwards and both factor prices downwards. It is not clear whether wages, GDP and tradable production and consumption increase or not, although it is clear that non-tradable production and real exchange rates increase. Real wages measured in terms of the full basket of commodities may increase or not while real interest rates decrease. Capital biased productivity shocks in the nontradable sector have a similar impact on GDP, real exchange rate and production and consumption, but they have opposite effects on wages and interest rates. Factor endowment increments push factor prices downwards but push supply of all goods upwards. The resulting excess supply of non-tradable goods pushes non-tradable and again factor prices downwards. Nonetheless, GDP and demand for all goods increase. The resulting excess demand for non-tradable goods pushes nontradable prices up, but it does not avoid real exchange rates depreciation.14 If the international prices and terms of trade increase, the direct and indirect de-industrialisation will perform on the importable sector through resource movement effects favourable to the exportable sector and spending effects favourable to the non-tradable sector. It is not clear what the impact on real exchange rates will be since terms of trade shocks change tradable and non-tradable prices in the same direction.15 The impact on non-tradable production is 14

15

These results suggest higher real exchange rates in labour abundant countries since in these countries, “the labour-using non-traded services can be produced at a relatively low cost compared to the rich capital abundant country” (Dornbusch, 1985, p.14). Real exchange rate may depreciate to high preferences for exportables (G ĺ1) or larger importables sector (șM ĺșT) but appreciate to larger exportables sector (șX ĺșT). Graphically, production frontiers and indifference maps in the tradable to non-tradable goods space will change since they are related to the tradable production and consumption valued at the initial international prices.

96

Fernando Zarzosa Valdivia

also not clear, therefore it cannot be determined whether these shocks cause Dutch disease or not.16 The impact on real factor prices measured in terms of the full basket of commodities will be opposite to the real exchange rate movements. Exportable endowments increments generate pure spending effects since they may directly be transformed into income. As a result, Dutch disease affects the importable sector. Also, real exchange rates appreciate and all goods consumption and real factor prices measured in terms of the full basket of commodities increase. The effects of importable endowments increments are similar but in this case, Dutch disease affects the exportable sector. Finally, net debt increments addressed to consumption also generate pure spending effects and impact negatively on real exchange rates and production levels of both tradable goods. Nonetheless, if net debt increments are assumed to increase productivities, the results will vary according to the favoured sector.17 C. Does the magnitude of the structural real exchange rate response to exogenous shocks depend on the economic structure of a country? Up until now, models of real exchange rate determination only “imply a role for tastes and technology, as well as the conditions under which one might be more relevant than the other, in particular the inter-sectoral and international mobility of capital” (Garcia, 1999, p.3). They do not take into account the fact that different economies may respond differently to similar exogenous shocks because of their different economic structures. We attempt to include the structure of the economy by including income distribution, debt-GDP and current account-GDP ratios in the model. As a result, the magnitude of the response of an economy to exogenous shocks depends not only on tastes and technology but also on the economic structure. 16

17

However, if tradable endowments are zero and these shocks depreciate the real exchange rates, they will cause Dutch disease. These results suggest lower structural real exchange rates in borrowing countries. Plus, aid flows can be incorporated in the model. If they increase tradable endowments or debt-GDP ratios, they will cause Dutch disease. If they improve TFP, they will allocate resources to the favoured sector. If they increase consumption and cause TFP improvements, they may or may not generate Dutch Disease.

Real Exchange Rate Movements, Dutch Disease and Functional and Sectoral Income Distribution

97

For instance, Table 5 postulates that (i) the magnitude of the structural real exchange rate response to TFP shocks in a tradable sector is positively related to the size of the sector but it is negatively related to the size of the other tradable sector, (ii) the structural real exchange rate response to debt shocks is negatively related to the size of the tradable sector as a whole, (iii) size of debt in terms of the GDP impacts negatively on the structural real exchange response to all kind of shocks, except terms of trade shocks and (iv) larger shares of a factor of GDP imply a bigger response from the real exchange rate to productivity shocks biased to that factor and to endowment increments of that factor. Further, these results suggest that different countries might respond differently to similar shocks because of their different economic structures, even when they have the same technologies and preferences.18 D. Income distribution effects of exogenous shocks and variability of the magnitude of real exchange rates and GDP response to exogenous shocks

Exogenous shocks reallocate resources, which in turn leads to income distribution effects. Thus, if tradable endowments are zero and constant, Table 6 shows schematically the income distribution effects of exogenous shocks. It suggest that (a) TFP shocks in any tradable sector increase the size of the tradable sector as a whole, while any productivity shocks in the non-tradable sector diminish it, (b) TFP shocks biased to a specific tradable sector increase the share of that sector to the tradable and total income, (c) TFP shocks do not change the functional income distribution but biased productivity shocks in the non-tradable sector do, (d) factors endowments shocks increase the non-tradable sector size but, they only 18

For instance, (i) if there are two countries with a similar debt structure, the real exchange rates response to TFP shocks in the non-tradable sector will be larger in the country with the smaller tradable sector, (ii) if there are two countries with the same income distribution ratios, the real exchange rates response to all kinds of exogenous shocks will be smaller in countries with larger debt-GDP ratios and (iii) if there are two countries with similar debt-GDP and equal size for their sectors, the real exchange rates response to labour (capital) endowment shocks will be bigger for the country with larger (smaller) ratio labour / capital endowment.

Fernando Zarzosa Valdivia

98

increase labour shares when the ratio labour / capital endowment increases19 and (e) debt (as a share of GDP) increments diminish the size of each tradable sector and (f) terms of trade improvements reduce the size of the importable sector.20 Table 5: Is the magnitude of the structural real exchange rate response to exogenous shocks independent of the economic structure of a country? distribution ratios functional income distribution

sectoral income distribution

debt structure

r

*

f  ca

Eq. A2.1 parameters*

TT

TX

Ɏ1

?

+

-

-

Ɏ2

?

-

+

-

Ɏ3

-

+

+

Ɏ4

?

?

?

Ɏ5

?

?

?

Ɏ6

?

-

-

Ɏ7

?

-

-

TM

TL

TK

+

+

+

-

+

-

Ɏ8

?-

+

-

?-

Ɏ9

-

-

-

-

Ɏ10

?

-

-

-

Ɏ11

?

-

-

-

Note: *Equation (A2.1) of Appendix II is the structural real exchange rate movement equation

In this model, the magnitude of the structural real exchange rate response to exogenous shocks depends on the initial economic structure of a country. Also, it may be variable because income distribution effects of exogenous shocks change the economic 19

20

Thus, factor endowments drive not only the real exchange rate but also the factorial income distribution of an economy. “Some empirical evidence suggests that factor endowments drive both inequality and the real exchange rate in the long run (see Garcia, 1999). The relative de-industrialisation of this sector will be deeper if these improvements generate Dutch disease.

Real Exchange Rate Movements, Dutch Disease and Functional and Sectoral Income Distribution

99

structure.21 Furthermore, the GDP response to TFP shocks in a specific tradable sector may be larger in countries where the sector is larger, see Equation (A2.3) of Appendix II. Our results also explain two of the stylized facts described by Rodrik (2006): (a) rapidly growing countries are those with larger manufacturing sectors, (b) specialization patterns are not pinned down by factor endowments. Thus, if we assume, like Rodrik (2006), that the exportable sector produces commodities and the importable manufactured goods, the evolution of the TFP of the manufactured sector will be the main driver of economic growth if TFP grows faster in this sector than others. Therefore, rapidly growing countries would be those with larger manufacturing sectors.22 E. Welfare effects of exogenous shocks

If an altruist policymaker and our representative consumer have the same preferences, social welfare movements will be reflected by changes in the representative consumer’s utility. As a result, exogenous shocks will influence the social welfare of a country. Thus, if tradable endowments are zero and constant: a) TFP shocks will improve social welfare levels, b) biased factor productivity shocks in the non-tradable sector however, will not have a clear impact on social welfare because the non-tradable consumption increases but it is not clear whether tradable consumption will increase or not and c) net debt increments will increase the social welfare levels 21

Some results of Table 5 and 6 relationships suggest that (a) current TFP shocks in any tradable sector increase the structural real exchange rate response to upcoming similar shocks, although they decrease its response to upcoming TFP shocks in the other tradable sector, (b) increments in the ratio labour / capital endowment increase the structural real exchange rate response to upcoming labour endowment shocks and (c) current terms of trade improvements may increase the structural real exchange response to upcoming TFP shocks in the exportable sector but reduce its response to upcoming TFP shocks in the importable sector.

22

Formally:

because

If 'Z M

t t

d  '%GDP d  '% Z M

!0Ÿ

)2 t t

t t

d  %'GDP d  %'Z M and

!

d  %'GDP

t t 1

d ) 2 dT M š t0. dT M dZ M

d  %'Z M

t t

Fernando Zarzosa Valdivia

100

through positive effects on all goods consumption. In any case, factor endowments impact positively on the social welfare of a country. Table 6: Do exogenous shocks lead to income distribution effects?

ZX

ZM

ZN

Z NL

Z NK

L

K

TT

șT

+

+

-

-

-

-

-

?+

-

șX

+

-

-

?

?

?

?

+

-

șM

-

+

-

?

?

?

?

-

-

șL

+

-

+

-

șK

-

+

-

+

factor endow ment shocks

¨ capital inflows

TFP shocks importable sector

distr. ratio

TFP & specific factor productivity shocks nontradable sector

international price shocks

TFP shocks exportable sector

exogenous shocks

r * df - dca

Note: *Distribution ratios are defined in Equations (19) and (20).

F. Extensions of the model

1. Benchmarking: comparing the model to Corden and Neary’s (1984) core model Corden and Neary’s core model of Dutch disease puts into context the concerns raised about the adverse effects of productivity improvements, windfall discoveries of new resources and terms of trade on the industrialisation of a small economy.23 Their results suggest that productivity improvements, terms of trade shocks and 23

They assume three sectors, a booming sector (energy), a lagging sector (manufactures) and a non-tradable sector. They also assume that each sector output is produced by a specific factor to that sector, and by labour, which is assumed mobile between sectors and move to equalize its wage.

Real Exchange Rate Movements, Dutch Disease and Functional 101 and Sectoral Income Distribution

windfall discoveries of sector specific resources might generate the Dutch disease. Our model can be benchmarked to the Corden and Neary’s model. Our booming sector is assumed to be the exportable sector, our lagging sector the importable sector and our perfectly mobile factors labour and capital. We additionally assume an exogenously given third specific factor R to each tradable sector.24 Qualitative results of this extension are also shown by table 4 but, in this case, ZX and ZM must change as a consequence of TFP shocks and windfall increments of RX and RM, respectively. Thus, in this extended model, windfall discoveries of specific resources R addressed to a specific tradable sector only reallocate resources favourably to that sector; they do not generate Dutch disease. Plus, if terms of trade shocks do generate the Dutch disease, our results, like Corden and Neary’s results, suggest that immigration (through labour endowment increments) may help an economy to overcome Dutch disease. 2. Structural real exchange rate and imperfections in the non-tradable market If the non-tradable market has a monopolistic structure, nontradable producers will fix their prices in their optimization problem. Non-tradable market imperfections change GDP-factors rewards and GDP–total expenditure relationships. The equilibrium, under this condition, implies higher non-tradable prices, lower nontradable production and higher tradable production25 than their equilibrium levels under perfectly competitive non-tradable markets. Equation (23) relates the structural real exchange rate response to exogenous shocks under perfect and imperfect nontradable markets. 24

25

Production functions of the exportable and importable sectors can be re-defined as follows: X ( LX , K X )

X 0  Z X RMX X LIXX K \X X and

M ( LM , K M )

M 0  Z M RMM M LIMM K \MM

where RX and RM are the specific factor addressed to the exportable or importable sector, respectively and ijX and ijM are exportable and importable output elasticities to the factor R, lying between zero and one. The introduction of imperfections in the non-tradable market increases exportable and importable outputs in a ratio equal to (1–E)/E.

102

Fernando Zarzosa Valdivia

Structural real exchange rate movement equation: I P eˆstruct beˆstruct (23) where: I P and estruct are the structural real exchange rates under perfect e stru ct and imperfect non-tradable markets, respectively P eˆstruct is given by equation (A2.1) of Appendix II

­° ½° (1  r * f  ca ) d (2  E ) ® ¾ t 1 wb wTT  0 * ª º °¯ ¬1  (1  E )(1  T T )  r f  ca ¼ °¿ Non-tradable market imperfections imply higher non-tradable prices, but lower structural real exchange rates than under perfectly competitive non-tradable markets. However, it is not clear whether the structural real exchange rate response to exogenous shocks will be bigger or smaller than in the competitive case. In other words, the impact of exogenous variables on the structural exchange rate is not necessarily bigger under an imperfect non-tradable market. Further, larger imperfections in the non-tradable sector (lower E) imply bigger tradable sectors but a less proportional impact of exogenous shocks in the structural real exchange rate than in the perfectly competitive case. b

IV. Conclusions and further research

We have developed a theoretical model which shows how factor improvements, windfall discoveries of new resources, terms of trade, factor endowments and debt shocks impact on real exchange rates, factor prices and consumption, production and social welfare levels of an small economy. It also evaluates possible Dutch disease and income distribution effects generated by exogenous shocks. In addition, stylized facts of Argentina, Brazil, Chile and Mexico suggest that exogenous shocks and exchange rate policies should be carefully gauged in terms of their impact on the structural real exchange rate, the structure of the economy and the development of the tradable sector. The model also suggest that if the exportable sector produces commodities, the importable sector produces manufactured goods and the TFP grows faster in the importable sector than others, the evolution of the TFP of the manufactured sector will be the main driver of economic growth and therefore, countries with larger manufacturing sectors would grow rapidly.

Real Exchange Rate Movements, Dutch Disease and Functional 103 and Sectoral Income Distribution

Up until now, theoretical models like this one have not taken into account the influence of income distribution and debt-GDP ratios on the response of an economy to exogenous shocks. This model incorporates these factors, and shows that the response of an economy to exogenous shocks differ between countries, even when they have same preferences and technologies. An extended version of our model benchmarks our model to Corden and Neary’s (1984) core model. We also suggest that the introduction of imperfections in the non-tradable market implies lower structural real exchange rates compared to the perfectly competitive case, but it is not clear whether the real exchange rate response to exogenous shocks is bigger or lower than its counterpart in the competitive case. Some of the remaining shortcomings of this theoretical paper are that it does not include a non-linear non-tradable technology and Home Trade Bias Preferences. The introduction of diminishing returns to scale in the non-tradable sector could explain how the response and the degree of response of an economy to exogenous shocks depend also on labour/capital or skill / unskilled labour ratios. Also, the relative factor prices could be endogenously determined. In trade data, “there is evidence of cross-hauling-countries import and export the same commodity, even for the most detailed commodity category” (Thierfelder and Robinson, 2002, p.4). The introduction of product differentiation features for each kind of tradable good may explain the two way trade at the commodity level. Finally, stylized facts of Argentina, Brazil, Chile and Mexico suggest that it is also possible to work out the empirical analysis of the model and its extensions.

104

Fernando Zarzosa Valdivia

References

Balassa, Bela (1964), The Purchasing Power Parity Doctrine: A Reappraisal, Journal of Political Economy, 72, 584-96. Baldi, Anne-Laure, Mulder, Nanno (2004), The impact of exchange rate regimes on real exchange rates in South America, 19902002, OECD Economic Working Paper No. 396, June 2004. Blad, Michael, Keiding, Hans (1990), Microeconomics: Institutions, equilibrium and optimality, North Holland, Elsevier. Calderon, Cesar (2002), Real exchange rate in the long rung and short run: A panel Cointegration approach, Central Bank of Chile Working Paper No. 153. Chinn, Menzie (2005), A primer on real exchange rates: Determinants, overvaluations, trade flows and Competitive Devaluation, National Bureau of Economic Research Working Paper No.11521. Corden, W.Max (1984), Booming sector and Dutch disease Economics: Survey and Consolidation Oxford Economic Papers 36, p. 359-380. (Reprinted in: Corden, W.Max (1992), International Trade Theory and Policy, Edward Elgar) Driver, Rebecca L., Westaway, Peter (2004), Concepts of equilibrium real exchange rates, Bank of England Working Paper No. 248. Dwyer, Jaqueline, Lowe, Philip (1993), Alternative concepts of the real exchange rate: A reconciliation, Reserve Bank of Australia, Research Discussion Paper 9309. Dornbusch, Rudiger (1985), Purchasing Power Parity, National Bureau of Economic Research, Working Paper No. 1891. Engel, Charles, Rogers, John (1996), How wide is the border?, American Economic Review, 86, 1112-1125. Garcia, Pablo, (1999), Income inequality and the real exchange rate, Central Bank of Chile, Working Paper No. 54.

Real Exchange Rate Movements, Dutch Disease and Functional 105 and Sectoral Income Distribution

Gay, Alejandro, Pellegrini, Santiago (2003), The equilibrium real exchange rate of Argentina, Instituto de Economía y Finanzas, Universidad Nacional de Córdoba (Argentina) and Consejo Nacional de Investigaciones Científicas y técnicas (CONICET) MacDonald, Ronald, Ricci, Lucca (2002), Purchasing Power Parity and New trade theory, IMF Working Paper 02/32. Obstfeld, Maurice (2000), International Macroeconomics: Beyond the Mundell-Fleming Model, National Bureau of Economic Research Working Paper 8369. Obstfeld, Maurice (2004), External Adjustment, National Bureau of Economic Research Working Paper No. 10843. Obstfeld, Maurice, Rogoff, Kenneth (2004), The unsustainable US current account: Position revisited, National Bureau of Economic Research Working Paper 10869. O’Toole, Ronnie (1998), The Dutch Disease, Student Economic Review, University of Dublin. Rodrik, Dani (2006), Industrial Development: Stilized Facts and Policies, Draft for the “Industrial Development for the 21st Century”, U.N.-DESA Publication. Salter, W.E.G. (1959), Internal and External Balance, The Role of Price and Expenditure effects, Economic Record.35, 226-38. (Reprinted in: Deepak, Lal (ed.), Development economics, vol. IV., The International Library of Critical Writing in Economics) Shantanayan, Devarajan, Lewis, Jefrey D., Robinson, Sherman (1991), External Shocks, Purchasing Power Parity, and the equilibrium real exchange rate, Development Discussion Paper No. 385, Harvard Institute for International Development, Harvard University.

106

Fernando Zarzosa Valdivia

Appendix I: Equilibrium point: Intuitive analysis with graphic tools

In this part, we extend the graphical analysis developed by Swan (1959) assuming no ad hoc solutions and tradable prices, factor endowments, tradable endowments and debt exogenously given. The interactions of the three markets of this economy are analysed through three kinds of production frontiers and indifference curves, one in the tradable to non-tradable goods space, one in the exportable-importable goods space and another in the non-tradable to importable goods space.26 These indifference curves are strictly convex since our CES utility function is homothetic and fulfils the Inada conditions. Production frontiers are strictly concave since our production functions are standard and well conformed. Indifference curves between tradable and non-tradable goods (UT/N) are related to specific combinations of non-tradable goods and consumers’ tradable expenditure. Each specific value of expenditure in tradable goods is related to a specific exportable-importable indifference curve (UX/M) and importable-(non-tradable) indifference curve (UN/M). Tradable to non-tradable production frontiers (PPFT/N) are related to specific combination of non-tradable output and tradable producers’ income. Each of these combinations is related to a specific importable-exportable production frontier (PPFX/M) and the non-tradable to importable production frontier (PPFN/M). Consumption and production of tradable goods are expressed in monetary values because it makes them comparable in the tradable to non-tradable space.27 26

According to Blad and Keiding (1990), a homogenous of degree k utility function, u(X), is homothetic and fulfils the Inada conditions if u (O X ) O k u ( X ), O  R  , and Lim wu( X ) f š Lim wu( X ) 0. X o0

27

wX

X of

wX

According to Salter (1959, p.227), this procedure effectively “collapses” the three-dimensional transformation function relating exportables, importables and non-traded goods into a two dimensional function.

Real Exchange Rate Movements, Dutch Disease and Functional 107 and Sectoral Income Distribution

Figure A1 tells us that if there is no debt, then the tradable to non-tradable production frontier (PPFT/N) and the line tangent to the exportable-importable production frontier (PPFX/M) are consumers' budget constraints. The “kissing-tangency point” between PPFT/N and UT/N0 determines producers and consumers' equilibrium point in the tradable to non-tradable space - points C=PET/N0. The slope of the tangent line at these points is “equal” to the equilibrium structural real exchange rate. Tradable producers' income and consumers’ expenditure in tradable goods are IT0 and ET0, respectively. Nontradable production is N0 (or CN0). The kissing tangency point between PPFX/M and CAX,M0 determines the equilibrium levels of exportable (X0) and importable (M0) goods -point PEX/M0. The kissing tangency point between CAX,M0 and UX/M0 determines the exportable (CX0) and importable (CM0) consumption levels - point 0 CEX/M . Figure A1: Equilibrium and structural real exchange rate

P

C

UN,M

N

EN,M0

EN,M0

C=P

ET,N0

0

UT,N0

N00 CN0 PPFN,M

PPFT,N0 CM0

M0

ET0 = IT0

M UX,M0

CX

C

EX,M0 PPFX,M0

P

ET , IT

0

X0

EX,M0 CAX,M0

ET ET(CM0 ,CX)

X

0

IT IT(M,X)

Fernando Zarzosa Valdivia

108

Appendix II: Static comparative analysis in the neighbourhood of the equilibrium point: structural and PPP’s real exchange rate and GDP Movement equations28

eˆstruct

) 1 Zˆ X  ) 2 Zˆ M  ) 3 Zˆ N  ) 4 Zˆ N L  ) 5 Zˆ N K  ) 6 Lˆ  ) 7 Kˆ  ) 8 (TT )  ) 9  r * df  dca  ) 10 Xˆ 0  ) 11 Mˆ 0

(A2.1) eˆ P P P (1  J )

1

( ) Zˆ X*  ) 1 Zˆ X )  ( ) *2 Zˆ M*  ) 2 Zˆ M )  ( ) *3 Zˆ N*  ) 3 Zˆ N )  ( ) *4 Zˆ N* L  ) 4 Zˆ N L ) * 1

 ( ) *5 Zˆ N* K  ) 5 Zˆ N K )  ( ) *6 Lˆ*  ) 6 Lˆ s )  ( ) *7 Lˆ*  ) 7 Lˆ )  ( ) *8 TˆT *  ) 8 TˆT )  ª¬ ) *9 ( r * d f *  ca * )  ) 9  r * d f  d ca º¼  ( ) 1 0 Xˆ 0  ) 1* 0 Xˆ 0* )  ( ) 1 1 Mˆ 0  ) 1* 1 Mˆ 0* )

(A2.2)29 ˆ GDP



¼

(T L  T K ) ª ) 1 Zˆ X  ) 2 Zˆ M  (1  ) 3 ) Zˆ N  ) 9 r * df  dca º ¬

 ¬ªT L  (T L  T K ) ) 4 ¼º Zˆ N  ¬ªT K  (T L  T K ) ) 5 ¼º Zˆ N L

K

 ª¬T L  (T L  T K ) ) 6 º¼ Lˆ  ª¬T K  (T L  T K ) ) 7 º¼ Kˆ  ªT X  (T L  T K ) ) 8 X º PˆX  ªT M  (T L  T K ) ) 8 M º PˆM ¬

0

¼

¬

¼

0

 ªT X  (T L  T K ) ) 10 º Xˆ 0  ªT M  (T L  T K ) ) 11 º Mˆ 0 ¬

0

¼

¬

0

¼

(A2.3)

28

29

Equation (A2.1) arises from the relationship between two pre-equilibrium conditions: a) The resulting condition of applying total differential to the Equations (7) and (21) and b) the resulting condition of applying total differential to the ratio tradable / total income taking into account Equations (9), (10), (14), (15), (19) and (20). Equations (A2.1) and (A2.2) have been empirically tested. However, the estimation methods have not taken into account the variability of the magnitude of the real exchange rate response to exogenous shocks.

Real Exchange Rate Movements, Dutch Disease and Functional 109 and Sectoral Income Distribution

where: * refers to the foreign country and a hat over a variable indicates a rate of change ) j ,  j z 9 , )8 M , )8 X are elasticity parameters and ) 9 is a semielasticity parameter )

j

t 0

 j / j z 8,

) 8 dt 0

(1  ) 4 ) š (1  ) 5 ) dt 0, (T L  ) 6 ) š (T K  ) 7 ) t 0 0 d )8 X d 1 and 0 d )8 M d 1

1 E §

)0

)1 )2

· 1 ¨¨ ¸ * E © TT  r f  ca ¸¹ t0 ª1  E § º · 1 « » ¨ ¸ * « E © TT  r f  ca ¹ » 1 J  « TM  TM0 § T X  T X0 ·» «u ¨   (1  TT )(T X 0  T M 0 ) ¸ » «¬ © 1  I X \ X 1  IM \ M ¹ »¼ T X  T X0 0 d )1 d 1 )0 1  I X \ X TM  TM0 )0 0 d )2 d 1 1  IM  \ M

)3 )1 )2 )0 (1TT )(TX0 TM0 ) )1 )2 )10 )11 ) 4 I X )1  IM ) 2  ) 6 ) 5 \ X )1  \ M ) 2  ) 7 ) 6 ) 0TT T L ) 7 ) 0TT T K )8 (1  G )()1  )10 )  G () 2  )11 ) dt 0 1 (1  TT ) t 0 1  r f  ca

)9

)0

)10

) 0 (1  TT )T X 0

)11

(1  TT )) 0T M

)8 X )8 M

)8  G (1  G )()1  )11 )  G (1  ) 2  )11 ), 1  G  )8 (1  G )(1  )1  )10 )  G () 2  )11 )

*

0

0 d )3 d 1

Michael Brei

The Impact of Current Account Reversals on Relative Prices: The Brazilian Experience1 Abstract The present paper is related to the literature on capital account reversals in which a country faces unexpected and collective withdrawals of foreign investments. The associated reduction in the supply of capital leads in succession to a slowdown in economic activity and, most often, to a reversal of the current account deficit and to adjustments in domestic prices. Here, we focus on the impact of current account reversals on domestic prices. We investigate in detail the Brazilian confidence crisis in 2002 and find evidence that a reversal of foreign capital inflows played a key role in triggering the associated economic slowdown, the major currency depreciation and the pronounced increase in the relative price of tradable and non-tradable goods. I. Introduction Emerging market economies that have much of their growth ahead of them can borrow from international capital markets to invest and, therewith, smooth consumption over time. Problems with this mechanism arise when foreign investors collectively reduce their investments from a day to the next. It is particularly dangerous, when capital withdrawals by some key investors trigger a financial panic. Calvo (1998) termed such a situation as a sudden stop of capital inflows. Goldfajn (2001) states with regard to such a situation: „I define a sudden stop as a very large change in the supply of capital. Of course, this sudden stop is always 1

I would like to thank Ernesto Crivelli and the seminar participants at the 7th Arnoldshain Seminar - Trade and Integration: The EU and Latin America - for their useful comments and suggestions. The remaining errors are mine.

112

Michael Brei

in the negative direction. These are also problems with big booms of capital inflows in the sense that you need to know what you are doing with the big influx. But the real problem is when you get billions of dollars less from one year to the other - on the order of ten percent of gross domestic product (GDP) or so. And most of the countries that had crises faced this challenge: Mexico, Asia, Turkey, Brazil, all of them.“ In the recent past, Brazil faced twice collective and large withdrawals of foreign capital. The first episode is investigated by Baig and Goldfajn (2000) who argue that a key factor in triggering the crisis of 1999 was the increased risk aversion on international capital markets with regard to emerging market economies as a result of, amongst other factors, Russia's sovereign default in 1998. The second episode, i.e. the confidence crisis, occurred in 2002 as a result of the following factors: (i) external factors in form of increased high yield spreads in the United States and increased risk aversion on global capital markets caused by the worse-thanexpected crisis in Argentina; and (ii) domestic factors in form of market concerns about the large amount of public debt (72 per cent of GDP) and uncertainties associated with the presidential election in the second half of 2002.2 The deteriorations in the market sentiment resulted in a reduction of capital inflows by US$ 27.8 billion (6 per cent of GDP) during the year 2002. At that time, the capital account was the dominant part of the balance of payments and the current account had to adjust to the capital account restrictions. Over the year, the current account deficit contracted by US$ 15.4 billion (3.3 per cent of GDP). This paper investigates in detail the confidence crisis and addresses the question of whether the reversal in capital inflows was the driving force in triggering the observed recession. A particular focus is dedicated to the impact of the current account reversal on the relative price of traded and non-traded goods. As a theoretical background serves the small open economy model stemming from Obstfeld and Rogoff (2000) who quantify the price 2

The term confidence crisis stems from A. Fraga Neto, Governor of the Brazilian Central Bank at that time. See Banco Central do Brasil (2003).

The Impact of Current Account Reversals on Relative Prices

113

adjustment associated with a potential current account reversal in the United States. In contrast to Obstfeld and Rogoff, we do not impose assumptions on the structural parameters of the model. Instead, we estimate a reduced form of the central equilibrium condition which relates the current account to relative prices. We find evidence that the ratio of the current account deficit to tradable goods consumption prior to the reversal explains large part of the subsequent adjustment in relative prices. The remainder of the paper is organized as follows. In Section II, we present the theoretical model. In Section III, we describe the confidence crisis from a macroeconomic perspective and examine the implications of the theoretical model empirically. The final section concludes. II. A short-run small open economy model Recent literature on emerging market crises emphasizes the important role played by reversals in capital inflows. According to this view, crises are triggered by deteriorations in the risk perception of foreign investors and lenders. Net foreign capital inflows, KI t , are used to finance current account deficits, CADt , and to accumulate international reserves, 'Rt . Abstracting from errors and omissions, the following balance of payments identity holds KI t CADt  'Rt (1) In the event of a slowdown in net capital inflows, adjustments in both the current account deficit and in international reserves may occur. Central banks can sell parts of their international reserves to cushion the impact on the domestic economy, however, when the slowdown in capital inflows is persistent, or large, then international reserves are likely to be insufficient. The involved adjustments in domestic consumption, investment and production can be described by the following national accounting identity CADt I t  Ct  Gt  (Yt  it Dt ) I t  St , (2)

where I t denotes national investment, Ct private consumption, Gt public expenditures, Yt gross domestic product, it the interest rate on net foreign liabilities Dt , and S t national savings. When the slowdown in capital inflows leads to an adjustment in the current account deficit then, simultaneously, adjustments in the other macroeconomic aggregates occur, typically, there are contractions

114

Michael Brei

in investment, consumption and production associated with adjustments in domestic prices. To infer qualitatively the adjustment mechanism such an accounting identity does not help, because the involved variables are interdependent. We use a two-sector small-open-endowment economy model to analyze the economic adjustment in response to a current account reversal. The economy consists of an infinitely lived representative agent who consumes traded and non-traded goods. Domestic production of both traded and non-traded goods is exogenous and equal to YT and YN . Therefore, it is implicitly assumed that capital and labor are not mobile between sectors. There is no domestic bond market and nominal prices are completely flexible.3 The agent derives utility from both traded and non-traded goods according to a constant elasticity of substitution function 1

T 1 T

[J CT T

C

1

T 1 T T T 1

 (1  J ) C N ] T

(3)

with T ! 0 being the elasticity of substitution between traded and non-traded goods, CT and C N the consumption of traded and nontraded goods, and 0 d J d 1 determines the relative size of the traded goods sector in the economy. Given the aggregate price level P , the prices of traded and non-traded goods, PT and PN , and a level of consumption C , the agent minimizes each period expenditures

min PC  (PN CN  PT CT )

CT ,CN

(4)

The first order conditions imply the following relation between the relative price of traded and non-traded goods and the respective levels of consumption PN PT

(

1 J

J

1

)T

1

C  ( N) T CT

(5)

In this setup, the aggregate price level is given by 1

P

( JPT1T  (1  J ) PN1T ) 1T

(6)

Imposing the market clearing condition for non-traded goods, 3

The assumptions of no domestic bonds and exogenous production imply that the optimization problem of the representative agent is static and we drop therefore time subscripts.

The Impact of Current Account Reversals on Relative Prices

i.e. C N

115

YN , the current account deficit can be expressed as

CAD CT  iD  YT (7) where net external debt, D , which is measured in terms of traded goods and the corresponding interest rate, i , are both assumed to be constant. Therefore, balance-sheet effects which affect net foreign liabilities are ruled out. Especially in highly dollarized economies this mechanism plays an important role in amplifying the economic effects, however, here we abstract from such considerations. Given the current account deficit is reduced by  'CAD , the assumptions 'YT 0 and '(iD) 0 combined with the market clearing condition for non-traded goods imply that the whole adjustment is made by a drop in the consumption of traded goods 'CT 'CAD (8) Given flexible prices, it follows from equilibrium condition (5) that the relative price of traded and non-traded goods raises. Obstfeld and Rogoff (2000) quantify the impact of a potential current account reversal in the United States on relative prices by fixing the structural parameters (J ,T ) and the magnitude of the current account reversal. Setting the elasticity of substitution between traded and non-traded goods is crucial, because it determines the response of PN / PT to changes in the consumption of traded goods. Here, we adopt a different approach and estimate the reduced form of equilibrium condition (5) which can be expressed as a loglinear demand function for non-traded goods ln(C N ) E 0  E 1 ln( PT / PN )  E 2 ln(CT ), (9) where E 0 , E1 and E 2 are constant parameters which are nonlinear functions of the structural parameters. Therefore, given the nontraded goods market clears a change in the current account deficit leads to the following adjustment in relative prices

' ln( PT / PN )



E2 'CAD ' ln(CT ), with ' ln(CT ) |  (10) E1 CT

It follows that the impact of a reversal in the current account deficit on the relative price of traded and non-traded goods is completely determined by the change in the current account deficit relative to the consumption of tradable goods. Note that, when the current account deficit is pushed to zero, the effect is proportional

116

Michael Brei

to the current account deficit relative to the consumption of traded goods prior to the adjustment. Calvo et al. (2004) use the ratio of the current account deficit over traded goods consumption as an explanatory variable in a Probit model to determine the key factors of sudden stops and find that the magnitude of this ratio is highly important. To be more precise, the higher this ratio, the more the economy relies on foreign capital to finance domestic consumption of traded goods, and the more vulnerable is the economy to a sudden stop which is associated with sharp changes in relative prices. Taken together, this ratio is a key variable to assess the short-run impact of current account reversals on relative prices. III.

The confidence crisis

In this section, the confidence crisis is investigated from a macroeconomic perspective. In particular, we highlight the importance of a capital inflow reversal in triggering the economic downturn and find evidence that the crisis has much in common with a sudden stop. In addition, we test whether the model can explain the associated change in the relative price of traded and non-traded goods. As indicated in the introduction, deteriorations in the market sentiment regarding to Brazil's economic prospects resulted in a slowdown in foreign capital inflows of US$ 27.8 billion (6 per cent of GDP) in 2002. At the same time, the current account deficit contracted by US$ 15.4 billion (3.3 per cent of GDP). The amount of private foreign loans declined from US$ 12.4 billion in 2001 to US$ 4.4 billion in 2002 and, therewith, the average roll-over rate of outstanding private debt declined from 98 per cent to 33 per cent. Foreign direct investment in 2001 amounted to US$ 22.5 billion and declined to US$ 16.6 billion in 2002.4 This development and its scope were largely unforeseen and the joint result of several elements: the contagion effect (Argentina), the political shock (presidential election) and the external financial shock (high yield spreads in the United States). Brazil responded first by drawing on a precautionary credit line it had set with the IMF in 2001, and when that proved insufficient, it obtained a

4

For more details, see Banco Central do Brasil (2002 and 2003) and Roubini and Setser (2004) p. 61.

The Impact of Current Account Reversals on Relative Prices

117

second major IMF rescue in 2002.5 To test whether Brazil faced indeed a sudden stop according to the definition of Calvo et al. (2004), we use the following identification strategy: (i) the period includes at least one observation where the annual change in quarterly capital inflows lies at least two standard deviations below its sample mean; (ii) the period begins (ends) when the annual fall (rise) in quarterly capital inflows falls (rises above) one standard deviation below its sample mean; and (iii) there is an associated output contraction. Using this definition, we find that the sudden stop period started in the third quarter of 2002 and lasted till the end of the year. Figure 1 shows important macroeconomic variables for the period 1995-2003. As can be seen, the economic adjustment involved an increase in the country risk premium (measured by Brazil's EMBI+ spread), a current account reversal, and large drops in real GDP, consumption and investment. Associated with this the currency depreciated, inflation increased and stock prices collapsed. These findings support our view that the confidence crisis of 2002, with a culmination in the third quarter, incurs typical elements of a sudden stop in capital inflows since the macroeconomic adjustment corresponds largely with the empirical regularities of sudden stops. In the following, we investigate the change in relative prices that was associated with the onset of the sudden stop. From the second to the third quarter of 2002, capital inflows reversed from US$ 9.3 billion to US$ -4.6 billion (3 per cent of GDP). The associated reversal of the current account deficit amounted in total to US$ 6 billion (1.3 per cent of GDP). Figure 2 shows the development of the capital and current account. The relative price of traded and non-traded goods is measured here by the ratio of the wholesale price index (WPI) to the consumer price index (CPI)6 PT / PN WPI / CPI . (11)

5

6

The IMF agreed in August 2002 to a rescue loan package of US$ 30 billion transferred to the Brazilian Central Bank in two stages: 6 billion in 2002 and 24 billion in 2003. This ratio can be seen as a measure of the relative price between goods (many of which are traded and a substantial component of the WPI) and services (most of which are non-traded and a major component of the CPI).

118

Michael Brei

Figure 1: Selected macroeconomic variables for Brazil (to be cont.) a) Current and financial account as a share of quarterly GDP

c) Quarterly depreciation of the nominal exchange rate (Real/$)

e) Percentage deviations of real investment from trend

b) Percentage deviations of real GDP from trend

d) Quarterly inflation (GDP deflator)

f) Percentage deviations of real consumption from trend

The Impact of Current Account Reversals on Relative Prices

119

Figure 1 (continued): Selected macroeconomic variables for Brazil g) Percentage deviations of the Bovespa Stock Index from trend

h) Country risk (EMBI+ spread) for Brazil

Source: Most of the Data was obtained from IPEA-DATA. The other time series were obtained from Uribe and Yue (2006). Real GDP, real investment, real consumption and the rate of unemployment are seasonally adjusted. The trend components for quarterly data are calculated with the HP (1600)-filter, while those for annual data with the HP (100)-filter. Note that the percentage deviations, shares and growth rates are in decimals, e. g. 0.3 denotes 30 per cent.

Figure 2: Current Account and Capital Account Millions of US$

In the quarter under consideration, the relative price increased by 5.2 per cent. Figure 3 shows the development of the respective prices indices and the nominal exchange rate for the period from 1995 to 2004. Domestic consumption and production of traded goods as well as domestic consumption of non-traded goods are measured as summarized in Table 1.7 7

A detailed description of the underlying data is given in the Appendix.

120

Michael Brei

Figure 3: WPI, CPI and the nominal exchange rate

Table 1: Construction of the relevant variables

YT

Domestic production of traded goods: agricultural plus industrial production

CT

Domestic consumption of traded goods: imports plus tradable production minus exports

CN

Domestic consumption of non-traded goods: production of services

The central equation is the reduced form of the equilibrium condition given in Equation (9). Since the only parameters of interest are E1 and E 2 , the equation can be estimated in first differences ' ln(C N ,t ) E 1 ' ln( PT ,t / PN ,t )  E 2 ' ln(CT ,t )  H t (12) The explanatory variables are potentially endogenous, because prices are specified simultaneously with the associated quantities. OLS estimates are, therefore, inconsistent. A consistent estimation method is the generalized method of moments (GMM) estimator which involves the use of instrumental variables that are highly correlated with the original regressors and not contemporaneously correlated with the error term. In our case, the moment conditions can be written as

The Impact of Current Account Reversals on Relative Prices

121

E[' ln(C N ,t )  E1' ln( PT ,t / PN ,t )  E 2 ' ln(CT ,t ) U t ] 0 (13) where U t denotes the instrument set which includes lagged values of the change in relative prices and the consumption of traded goods. A detailed estimation summary is given in Table 2. Note that a constant was specified in the regression. All coefficients are highly significant and have the expected signs. In addition, we perform overidentification tests to check whether the set of instruments valid. On a significance level of 5 per cent, we find no evidence against the selected instruments. The coefficient estimate for E1 which can be interpreted as the demand elasticity of non-traded goods with respect to the relative price is equal to 0.77. The sensitivity of relative prices to a current account reversal is measured by E 2 / E1 and equal to 0.57. This implies that a 10 per cent decrease in the current account deficit relative to tradable goods consumption increases the relative price of tradable goods by 5.7 per cent. Therefore, given this price sensitivity, the larger the current account deficit reversal relative to tradable goods consumption, the larger is the associated price adjustment. Essentially this means that the more an economy relies on foreign capital to finance domestic consumption of traded goods, the more vulnerable it is to sharp changes in relative prices in times of current account reversals.

122

Michael Brei

Table 2: Estimation results for Equation (13) using all information Dependent Variable: Implicit Equation Method: Generalized Method of Moments Sample(adjusted): 1996:1 2005:1 Included observations: 37 after adjusting endpoints Kernel: Bartlett, Bandwidth: Variable Newey-West (17), Prewhitening Simultaneous weighting matrix & coefficient iteration Convergence achieved after: 101 weight matrices, 102 total coef iterations Implicit Equation:

E[' ln(C N ,t )  E1 ' ln( PT ,t / PN ,t )  E 2 ' ln(CT ,t )  G U t ] Coefficient

0

Std. Error

t-Statistic

Prob.

E0

0.769

0.14

5.29

0.000

E1

0.438

0.03

17.11

0.000

0.001

6.56

0.000

0.07

E2

S.E. of regression: 0.079

Sum squared resid: 0.217

Durbin-Watson stat: 2.605

J-statistic: 0.07

Instruments:

U t : xt 1 ,..., xt  4 , xt 6 , xt 9 and a constant.

xt

' ln( PT ,t / PN ,t ) and ' ln(CT ,t )

denotes

In the third quarter of 2002, the current account deficit relative to the consumption of traded goods dropped by 9.8 per cent. With this information, the estimated model implies the following change in the relative price8 E 2 'CADt 1 ' ln( PT ,t 1 / PN ,t 1 ) 0.57* 9.8% 5.6 % (14) ( 0.09 ) ( 0.88 ) E 1 C T ,t The results indicate that a large part of the actual change in the relative price is explained by the model. Although the observed price movement of 5.2 per cent is overestimated, it lies within the 95 per cent confidence interval of the point estimate. 8

In brackets are given the standard deviations of the estimates.

The Impact of Current Account Reversals on Relative Prices

123

In addition, we examine an out-of-sample forecast and estimate the model with information prior to the sudden stop (see Table 3). Given the potential risk of a slowdown in net capital inflows which pushes the current account deficit to zero, the model predicts the following change in the relative price

' ln( PT ,t 1 / PN ,t 1 )

E 2 CADt E 1 CT ,t

0.51* 8.3%

4.2 % (15)

( 0.19 )

(1.57 )

The results underline that the ratio of the current account deficit to tradable goods consumption prior to the sudden stop is an important variable to forecast domestic price developments in the event of a current account reversal. This is an important result since central banks can use this information to predict the potential adjustments in relative prices already in an early stage of a sudden stop. Table 3: Estimation results for Equation (13) using information before 2002:3 Dependent Variable: Implicit Equation Method: Generalized Method of Moments Sample(adjusted): 1996:1 2002:2 Included observations: 26 after adjusting endpoints Kernel: Bartlett, Bandwidth: Variable Newey-West (0), Prewhitening Simultaneous weighting matrix & coefficient iteration Convergence achieved after: 30 weight matrices, 31 total coef iterations Implicit Equation:

E[' ln(C N ,t )  E1 ' ln( PT ,t / PN ,t )  E 2 ' ln(CT ,t )  G U t ] Coefficient

Std. Error

E0

0.857

0.30

2.85

0.009

E1

0.437

0.04

10.19

0.000

E2

0.012

0.004

3.26

0.003

Instruments:

t-Statistic

0

Prob.

S.E. of regression: 0.089

Sum squared resid: 0.185

Durbin-Watson stat: 2.303

J-statistic: 0.505

U t : xt 1 ,..., xt  4 , xt 6 , xt 9 and a constant.

xt denotes ' ln( PT ,t / PN ,t ) and ' ln(CT ,t )

124

Michael Brei

IV.

Conclusions

We use the theoretical model stemming from Obstfeld and Rogoff (2000) to quantify the impact of a current account reversal on the relative price of tradable and non-tradable goods. The model implies that movements in relative domestic prices arise from downward shifts in the consumption of tradable goods that are triggered by the reversal in the current account deficit. The empirical investigation of the confidence crisis reached a few results that are worth summarizing. First, an adverse shock to foreign financing triggered by changes in the market sentiment on global capital markets resulted in an unexpected and sharp slowdown in foreign capital inflows. Associated with this, the current account reversed and large adjustments in domestic production, absorption and prices occurred. Second, the estimation results indicate that the ratio of the current account deficit to traded goods consumption plays an important role in determining the adjustment of relative prices in the event of a current account reversal. This supports the findings of Calvo et al. (2004).

The Impact of Current Account Reversals on Relative Prices

125

Appendix Data Sources BOP statistics:

Banco Central do Brasil: Economia e financas Series temporais - Series especiais - Balanca de pagamentos trimestral

Wholesale price index IPA-10 - indice (ago. 1994 = 100) - (FGV/Conj. (WPI): Economica) Consumer price index INCA - geral - indice (dez. 1993 = 100) - (IBGE(CPI): SNIPC) Industrial production: PIB - industria - R$(milhoes) - IBGE/SCN Trim - Scn4_PIBINDV4 Agricultural production:

PIB – agropecuaria - R$(milhoes) - IBGE/SCN Trim - Scn4_PIBAGPV4

Production of services:

PIB - servicos - R$(milhoes) - IBGE/SCN Trim - Scn4_PIBSERV4

Nominal exchange rate:

Taxa de cambio - R$ / US$ - comercial - compra - media - R$ - BCB Boletim/BP - Bm12_ERC12

126

Michael Brei

References

Abreu, Marcelo de P., Medeiros, Marcelo C., Werneck, Rogério (2003), Formacao de precos de commodities: padroes de vinculacao dos precos internos aos externos, Discussion Paper No. 474 at PUC-Rio. Azevedo, Rodrigo, Bevilaqua, Alfonso (2004), Provision of FX hedge by the public sector: the Brazilian experience, BIS Working Paper No. 24. Baig, Taimur, Goldfajn, Ilan, (2000), The Russian Default and the Contagion to Brazil, IMF Working Paper WP/00/160. Banco Central do Brasil (2002), Meetings of the Monetary Policy Comittee (Copom), 73th to 77th Meeting between June and October 2002. Banco Central do Brasil (2003), Sistema de Metas para a Inflacao: Carta aberta de 21 de Janeiro. Calvo, Guillermo (1998), Capital Flows and Capital Market Crises: The Simple Economics of Sudden Stops, Journal of Applied Economics 1 (1), 35-54. Calvo, Guillermo, Izquierdo, Alejandro, Talvi, Ernesto (2004), Sudden Stops, the Real Exchange Rate, and Fiscal Sustainability: Argentina's Lessons, NBER Working Paper 9829. Calvo, Guillermo, Talvi, Ernesto (2005), Sudden Stop, Financial Factors and Economic Collapse in Latin America: Learning from Argentina and Chile, NBER Working Paper 11153. Goldfajn, Ilan (2001), Roundtable Comments on Monetary and Regulatory Policy, Domestic Finance and Global Capital in Latin America Conference, Federal Reserve Bank of Atlanta. Goldfajn, Ilan, Werlang, Sergio R. (2000), The Pass-Through from Depreciation to Inflation: A Panel Study, Discussion Paper No. 423, PUC-Rio. Goldfajn, Ilan, Hennings, Katherine, Mori, Helio (2003), Brazil's

The Impact of Current Account Reversals on Relative Prices

127

Financial System: Resilience to Shocks, no Currency Substitution, but Struggling to Promote Growth, Banco Central do Brasil.

Minella, Andre, Freitas, Paulo, Goldfajn, Ilan, Muinhos, Marcelo (2003), Regime de Metas para Inflacao no Brasil: Construindo Credibilidade em Contexto de Volatilidade na Taxa de Cambio, Banco Central do Brasil. Roubini, Nouriel, Setser, Brad (2004), Bailouts or Bail-ins? Responding to Financial Crises in Emerging Economies, Washington, DC: Institute for International Economics. Obstfeld, Maurice, Rogoff, Kenneth (2000), Perspectives on OECD Capital Market Integration: Implications for the U.S. Current Account Adjustment, Federal Reserve Bank of Kansas City, Global Economic Integration: Opportunities and Challenges. Uribe, Martin, Yue, Vivian (2006), Country Spreads and Emerging Countries: Who Drives. Whom?, Journal of International Economics 69 (1), 6-36.

Part 2: Institutions for Development: The Case of Corruption

Wolfgang Hetzer

The European Budget, A Breeding Ground for Corruption? I. Introduction In July 2006 the European Anti-Fraud Office (OLAF) published its 6th Activity Report. It covers the period from 1 July 2004 to 31 December 2005. At the end of the 2005 calendar year the Office’s electronic archive (Case Management System – CMS) contained a total of 5 165 case reports. That number includes 1 421 cases opened by OLAF’s predecessor, UCLAF, before 1 June 1999. The annual number of reports of possible irregularities or unlawful acts in connection with the European Union’s budget increased from 735 in 2001 to 857 in 2005. Tip-offs increased by 20 per cent a year in the last two years. Between 2001 and 2005 a total of 3 485 reports were received, each of which was individually evaluated. The average time needed was reduced from 18.6 months in 2001 to 5.3 months in 2005. At the end of 2005, OLAF was investigating 452 cases. A further 226 cases were in the initial assessment process. 233 cases were closed during the year. In 2005, for the first time, more cases (133) were closed with follow-up (e.g. recovery of funds unduly received or improperly used) than without follow-up (100). The average “life cycle” of a case is just under two years. In 2005 a total of 40 internal investigations were opened, as against 23 in 2004. The total number is now 58 cases. A further 30 are at the evaluation stage. In the first full calendar year after the EU-10 enlargement, OLAF opened 24 cases in new Member States and 23 cases in acceding and candidate countries. The year 2005 saw another, special peak reached. OLAF achieved its best ever results (so far), recovering € 203 million that should probably not have been paid to the recipients. Working together with the European Commission’s Directorates-General, the Member States and international donors, OLAF also stepped up its investigations into cases involving external aid.

132

Wolfgang Hetzer

The largest proportion of the cases (35 per cent) originated in Africa. Large-scale investigations also had to be launched into the activities of non-government organisations. OLAF also had cause to devote more of its attention to public procurement procedures. In the field of structural development the European Social Fund is most vulnerable to fraud. False declarations of product origin continue to be the most common fraud in the customs sector. One such case involving textiles diverted from Asia to the Caribbean led to recovery proceedings involving more than € 20 million that had probably been lost to the EU budget. These (incomplete!) statistics make it all too easy for critics to expound theories regarding the extreme susceptibility of the EU to corruption.1 They support their theories with a variety of facts and hypotheses: Subsidies, which experience teaches are particularly attractive to criminals, account for 90 per cent of the EU budget. - The area involving the greatest expenditure, the Common Agriculture Policy, is organised by an extremely complex web of regulations, rather than by free competition. This agricultural protectionism offers greater opportunities for corruption, while the impenetrability of the web of regulations makes control more difficult. - Only a small proportion of the EU budget is managed by the EU institutions themselves; about 80 per cent is administered by the Member States. This intertwining of administrations makes control more difficult. - When it comes to the perception of and importance attached to corruption, there is a North-South divide in the EU. As a result attempts to combat it at EU level tend to be based on the lowest common denominator. - Scandalous cases of “political corruption” at EU level (e.g. legalised fraud involving donations, legalised nepotism and legalised moonlighting by MEPs) undermine the credibility of attempts to combat corruption in the administration and world of business.

1

For full details of the figures visit: http://europa.eu.int/olaf.

The European Budget

133

Other arguments are brought into play as well: the key aspect of the EU budget, it is claimed, is that eighty per cent of it is accounted for by subsidies. However, so the argument goes, subsidising tends “inherently” to encourage people to fabricate the conditions for granting subsidies or conceal details from the taxman in the case of tax concessions. The complexity of the rules governing how subsidies are awarded, which is often not justified by any sensible market requirements and appears to be misled by targeted lobbying, further increases the temptation. In the final analysis, the public seldom perceives EU budget funds as taxpayers’ money. This makes it easier to view the budget as a “self-service shop” and encourages the view that misappropriating funds from the budget is a mere peccadillo. In particular, critics argue, there is no free competition in agriculture, and this increases the scope for corruption. It is true that the prices of EU agricultural products are kept artificially high, often many times the price otherwise paid on world markets. If EU farmers sell their products on the world market, they can claim “refunds” to cover their “losses”. Time and again this leads to large fraud involving exports of, for example, dairy products, beef and cereals. At the same time, the EU’s external customs duties are determined by a whole host of criteria with which it is difficult to prove compliance. The problem of control is exacerbated by the fact that agricultural lobbies in the Member States tend to be particularly influential, which makes it impossible to implement effective rules in the manner desired. The peculiar intertwining of Member States' and supranational powers and duties with regard to implementation and control, it is claimed, also increases susceptibility to corruption. The Member States are said to have no great incentive to combat corruption effectively, as it is “only” the EU that pays the price of corruption that may have an effective “functional” value for the local economy. What is more, claim procedures may prove counterproductive. This is, of course, a classic dilemma. The Commission has a serious interest in combating corruption effectively but lacks the powers to do so. The Member States have the powers but are often not interested. While corruption is becoming increasingly transnational, as critics argue, penalties and law enforcement systems are still organised nationally, with the result that the risk of prosecution and conviction is relatively small. Anti-corruption policy tends to

134

Wolfgang Hetzer

be made up on the hoof because it is usually a reaction to scandals and crises. Consequently, it is often hurried and not thought through. Political corruption in the EU is generally seen as particularly rife. It is a widespread perception that Members of the European Parliament are defrauding millions in the form of donations quite legally, employing spouses and other relatives at the taxpayer's expense perfectly legally and receiving more than one salary. Proposals for far-reaching reforms (for example, standardising the VAT rate, harmonising criminal law or abolishing subsidies) are thought to have no chance of success, and therefore the tendency to engage in corrupt practices at EU level is not expected to decline significantly.2 Suspicions have even been voiced that the lack of scope for control and imposing penalties at EU level is due to a secret plan on the part of each individual Member State to ensure it has the same possibilities to engage in fraud.3 However, as yet there is no reliable evidence for such a strategy, which would be tantamount to a special form of conspiracy. Nevertheless, it is debatable whether the historical background and current objectives of the EU budget have created specific risks and whether the budget’s political implications and characteristics have not helped to create an environment that attracts crime. Nobody should look for quick and easy answers. In the next few years in particular the EU is facing some highly complex tasks. The effective protection of its financial interests will determine whether the EU's taxpayers and voters will retain or regain their faith in the historically necessary continuation of European integration. The following points can be no more than the first modest attempts to characterise the challenges ahead. EU enlargement, in particular, will see structural aid replace agricultural policy as the area of greatest expenditure, as the digression below will show. However, whether this will lead to a rise in crime must remain an open question at this point. Specific risks may definitely not be ruled out. However, a certain basic understanding of the rules and the instruments available is needed before undertaking any estimate of the risk or any steps to prevent corrupt and fraudulent conduct.

2 3

Cf. von Arnim et al. (2006) pp. 46 et seq. Warner (2002) p. 18.

The European Budget

135

II. An example: structural aid The EU’s structural policy aims to support the economies of regions in the Community that are lagging behind. In practice, like sectoral industrial policy, this clearly runs contrary to the principles of a market-based economic policy. However, the commitment has been made in the Treaties to reduce the divide between individual regions in the EU by helping less favoured regions to catch up. When the structural policy was first launched in the sixties Italy was the country that benefited most. In 1973 the United Kingdom and Ireland joined and received money from the European Regional Development Fund. Since the eighties the so-called cohesion countries Greece, Ireland, Portugal and Spain have been receiving funding. The former GDR followed in 1990. Since 2004 the main beneficiaries have been the new Central and Eastern European Member States. Structural policy is planned for seven-year periods. Regional and structural policy is supposed to encourage the harmonious development of the Community as a whole (Article 158 of the EC Treaty). The necessary strengthening of economic and social cohesion has become the area of Community activity that accounts for the greatest expenditure. The instruments designed to reduce regional differences within the EU are the Structural Funds (the Agricultural Fund, the Social Fund, the Regional Development Fund and the financial instrument for fisheries). There are also special funds for certain countries (the Cohesion Fund and the Rural Development Fund). The Regional Development Fund plays a key role. Structural policy is funded from the general EU budget, and Germany is the biggest contributor. Structural aid was indispensable in the lead-up to the accession of the ten new EU Member States. The number involved and the social and economic differences are still posing an enormous challenge for the EU budget. A pre-accession strategy was implemented from 2000 to 2006 with the aim of accomplishing a number of objectives: Three main instruments were set up to support the accession countries: - PHARE was intended to strengthen the institutions in the accession countries to enable them to participate in Community programmes and receive support for regional and social

Wolfgang Hetzer

136

development, the restructuring of industry and small and medium-size companies; - SAPARD was designed to modernise agriculture and support rural development; - ISPA was intended to help improve transport infrastructure and environmental protection. In all, between 2000 and 2006 these three instruments made €21.84 billion available. Given the sums involved, the complexity of the tasks, the actual situation in some (old and new) Member States and the players involved, it is obvious that if criminal operations are conducted, they will be of a special nature, and not just in terms of the expertise needed, the degree of organisation and involvement of high-ranking decision-takers from the worlds of politics, administration and business. The forces of law and order therefore face a particularly daunting challenge because the single European market and the characteristics of the EU budget almost inevitably create a situation where criminals will improve their “skills” and, sooner or later, join forces in a new and particularly dangerous form of organised crime. However, the gap between rich and poor remains large. Incomes in the richest northern regions are two to three times higher than in the poorest regions in the south and east. Within the economic framework as it stands, structural policy can do no more than help the regions to help themselves, and this help often involves cumbersome bureaucratic processes, a situation that is unlikely to change in the foreseeable future because of the need for financial controls. III.

Market or mafia?

From the outset, anyone assessing the threats to the European Union’s financial interests from organised crime faces a dilemma. It is Europol’s view that the situation regarding organised crime is changing dramatically. At the same time Europol’s 2004 European Union Organised Crime Report (OCR) points out that the data contained in the contributions from the Member States used to produce the OCR in 2003 were insufficient to enable the total number of criminal organisations or the number of their active members to be determined. The lack of this empirical background information does not, however, prevent Europol from concluding that the existing criminal organisations have become aware that the

The European Budget

137

scope for crime has increased with the enlargement of the European Union. It also reports that core criminal activities have been extended, with highly skilled support from a variety of professions appearing to play a far greater role. The significance of these findings is limited by the fact that legal definitions and the description of organised crime activities leave something to be desired. The European Commission also believes that it is impossible to determine the extent of organised crime accurately, basically because the data published by the Member States are not even based on a common definition of the phenomenon. However, it agrees with Europol that in the last ten years criminal organisations have succeeded in developing and using extensive international networks to generate enormous profits. To produce a Community-wide risk assessment that can be taken seriously, however, we need harmonised categories and a reliable overall picture of the situation in the areas of activity in which criminal organisations are believed to be operating. There is already a (European) definition of a criminal organisation that can be used as a working basis. It is understood to be a structured organisation, established over a period of time, of more than two persons, acting in concert to commit offences which are punishable by deprivation of liberty or a detention order of a maximum of at least four years or a more serious penalty, whether such offences are an end in themselves or a means of obtaining material benefits and, where appropriate, of improperly influencing the operation of public authorities. Clearly, such an abstract notion cannot produce a clear picture of the material variety of organised criminal operations. Furthermore, this kind of criminal activity is dynamic by nature and cannot be rigid compartmentalised. However, past experience has shown that criminal organisations prefer to operate in certain areas: - Illegal drug-trafficking; - Illegal trafficking in arms and munitions; - People trafficking; - Forgery and product piracy. At European level the conclusion has been reached that international fraud operations are one of the sources of the vast profits made by organised crime. To have any long-term chance of successfully combating the dangers posed by organised crime it is no longer enough simply to agree on a few core definitions as part of a “joint action.” It is already foreseeable that more binding instru-

138

Wolfgang Hetzer

ments will need to be employed that will ultimately harmonise criminal law in the Member States. It might be an initial step in the right direction if the offences that give grounds to assume the existence of a criminal organisation corresponded or were comparable to predicate offences relating to money-laundering. In general, there is no doubt that criminal organisations devise and implement both operations designed to defraud public (European) budgets and webs of corruption. We need to take account of these modes of conduct if we are to minimise risks. This objective will not be achieved if EU Member States cite exceptions for tax offences as an excuse not to cooperate or recognise the abovementioned offences. It should be borne in mind that organised crime not only damages the financial interests of the individual Member State concerned; it threatens the financial interests of the Union as a whole. Therefore, it is only logical that proposals were put forward some time ago under the Corpus Juris project aimed at punishing the creation of a conspiracy set up to unlawfully obtain public funds. The establishment of a European Public Prosecutor could be a further major step towards more effectively combating the risks posed by organised crime to the European Union’s financial interests. Of course, its jurisdiction could be extended beyond this type of offence to encompass other areas of criminal activity in which organised crime also operates. IV.

Conclusions

The circumstances have dramatically increased the importance of an empirically based, hi-tech supported, politically aware, sectorally focussed risk assessment. The challenges involved can be met only if, at least within the framework of the European Union, networked information-gathering and evaluation structures can be developed and the traditional differences between criminal and administrative investigations play a subordinate role or no part at all, at least when it comes to making provision for risks. The current legal and personnel situation of the authorities collectively responsible for protecting the European Union’s financial interests is subcritical. Given the need to improve the effectiveness of prevention and law enforcement with regard to organised crime, this situation therefore has to be included in the list of risk factors. Summing up we can stress the following points:

The European Budget

-

-

-

-

-

-

-

-

139

The EU’s economic system is based on a liberalism that is tempered by the often very different objectives pursued by the Member States. Notwithstanding the continuing differences with regard to economic policy, the creation of a common market by the EU Member States is and will remain the historical and current core element of European integration. The single market and the structure of the European budget offer economic opportunities but also create all sorts of disruptions and risks. The effectiveness of the protection of the EU’s financial interests against unlawful acts (including fraud and corruption) is one of the basic preconditions for the success of the European project. It is becoming increasingly important to counter the asymmetries of integration policy, which can be seen, for example, in the tendency to transfer more powers and funds to the EU while there is still no Community criminal law system. Even if there is not yet sufficient empirical supporting evidence, there are indications that criminal organisations are currently taking advantage of the opportunities for crime offered by the existing structure and will seize the new opportunities that will arise after the EU’s enlargement. If we are to increase the efficiency of preventive and enforcement measures to protect the EU’s financial interests, cooperation must be stepped up as quickly as possible between the existing European institutions and the Member States authorities, but at the same time effective specialised institutions will also need to be created (e.g. a European (Financial) Public Prosecutor). It must be stressed that the dangers posed and damage caused by organised crime extend far beyond direct material losses, as ultimately they can affect the political acceptance of European integration at all levels.

140

Wolfgang Hetzer

References

Arnim, Hans Herbert von, Heiny, Regina, Ittner, Stefan (2006), Korruption – Begriff, Bekämpfungs- und Forschungslücken, FÖV Discussion Papers 33. Warner, Carolyn M. (2002), Creating a Common Market for Fraud and Corruption in the European Union: An Institutional Accident, or a Deliberate Strategy?, EUI Working Paper RSC No. 2002/31.

Basilia Aguirre

The Multiple Faces of Wrongdoing  A Closer Look on Corruption Abstract In this paper we argue that corruption is not a homogeneous phenomenon but that there is a variety of corruption forms and that each form has its own way of influencing economic performance because it acts in different economic spheres. The importance of distinguishing between various kinds of corruption is not only related to its different results but to the fact that the combination of different corruption forms may display different consequences. I. Introduction In Brazil we are being faced with a variety of forms of assaulting public interest. The forms and extent of the occurrences have no precedent in Brazilian history. This indicates that Brazil is going through a process of corruption deepening. “We are seeing a country that putrefies”1 was a recently newspaper title. In August 2006, in Rondônia State (Amazonian Region) more than 60 people from executive, legislative and judiciary power were arrested for having taken part in some kind of corruption action. More than 130 representatives both from Senate and Deputy Chamber were interrogated by the Join Budget Commission under suspicion of involvement in over billing ambulances purchases. Many have been arrested. In 2005, a huge scandal shook the Brazilian Congress and Executive with the indictment of representatives involved in buying votes to pass Executive proposals. The representatives received monthly payments to vote according to the will of the Executive. It became known as “mensalão”.2 We believe that very few countries have reached such an extreme process of corruption and so we have to ask why and how this situation has come up to this point. In order to tackle with the 1 2

O Estado de S. Paulo (SP) – 8/8/2006. This may be translated as “big monthly payment”.

142

Basilia Aguirre

problem we need to better understand the consequences of this process on development and distribution. And finally we should be able to indicate ways out of it. These are no easy tasks. However the very existence of the society is at stake. II. Open questions in the literature of corruption Many authors have been thinking about these questions and a lot of knowledge has been accumulated. Turning our attention to the economic literature we know a couple of things. Some authors indicate that corruption affects economic development (Rose-Ackerman, 1997; Tullock, 1996; Krueger, 1974; Bardhan, 2005) through preventing investments (Mauro, 1995), or reducing tax collection (Tanzi and Davoodi, 1997). Others found a strong relation between corruption and income distribution (Gupta et al. 1998). Corruption reduces investments in education as well as in other social services. As a consequence, human capital accumulation decreases and income distribution is negatively affected. Corruption is also considered to be connected with poverty through the channels of low development levels and inequality. These works indicated the damaging consequences of corruption but they do not go far enough in explaining why corruption spreads and what are the damages and risks of different forms of corruption. Furthermore, the knowledge about how to reduce the level and extent is still in its infancy. The major conclusions of existing studies are, firstly that it is necessary to introduce larger controls over corrupt actions, and secondly that it would be necessary to reduce the size of the state in order to narrow the room for corrupt transactions. It is obvious that these recommendations are not sufficient. Increased controls will only generate other forms of corruption when rules are not complete. Many attempts like this were made in Brazil and the results are usually more damaging than the original solutions.3 The reduction of the role of the state also does not seem to be the correct response. Many countries, like Brazil have reduced the involvement of the state and corruption was not reduced. Moreover, the reduction of the state has limits and as states will always be necessary, corruption will always find its way to come up. So it seems necessary to improve our knowledge on corruption. Although there are many works on the consequences of corruption, few have dealt with the causes of corruption. The traditional litera3

We will turn to this point later.

The Multiple Faces of Wrongdoing

143

ture argues that as people seek to improve their living conditions they will use every chance to increase their income level. Under these circumstances all the situations are alike. Or, all kinds of corruption acts are considered the same. These works do not distinguish between different kinds of corruption and its different consequences on economic activity. In this paper we argue that corruption is not a homogeneous phenomenon that there is a variety of corruption forms and that each form has its own way of influencing economic performance because it acts in different economic spheres. The importance of distinguishing between various kinds of corruption is not only related to its different results but to the fact that the combination of different corruption forms may display different consequences. Moreover, it is likely that certain kinds of corruption lead to a deepening of the corruption process. Finally it should be taken into consideration that the corruption process deepening may be a self-reinforcing process that makes the solution even more complicate. III. The principal forms of corruption In the following we deal with what we call the three basic forms of corruption and discuss what are their features and consequences. We will also argue that the different forms have distinct origins and that it is not possible to treat them as a single phenomenon. This indicates that the instruments to fight corruption need to be different depending on the nature of the case. Observing recent events in Brazil it is possible to identify many forms of corruption. According to Gaviria (2002), “One should distinguish between at least two different forms of corruption. The first form refers to an illegal transaction involving public and private parts. This includes bribe collection by public officials and illegal payments by private business. The second form does not involve private parts and refers mainly to the illegal misappropriation of public property by public officials, including bureaucrats, elected politicians and judges”. (Gaviria 2002, p. 249) This suggests that it is not appropriate to treat all of them in a similar way. In a corruption situation there is more than one interest involved, the corruptor and the corrupted. There is one side that wants something and is willing to pay for it. On the other side there is a person who can deliver what is demanded and would be interested in being paid for that. This line of rationing permits us to formulate a criterion of classification that considers who is willing to pay. Using this crite-

144

Basilia Aguirre

rion it is possible to define three basic corruption kinds: business, political and criminal corruption. The first and usually most discussed form of corruption is what we shall call business corruption. Business corruption is a situation in which firms and entrepreneurs are willing to pay for either the maintenance or the change of a rule or policy. Usual circumstances of this kind are external trade protection policy, subsidies, industrial policy and so on. This kind of corruption will be called decision buying business corruption and it is known in the literature as rentseeking activity. The consequences of this corruption type were analysed mainly by Krueger (1974) and Roley et al. (1996). This is the type of corruption which the economic literature usually considers. In this case a society can bear two kinds of costs. One is the dead weigh cost as a part of the resources that could be used productively is drawn off from production. The other effect is a welfare loss imposed to the rest of the society since it has to buy products much more expensive than otherwise. And a third loss is the reduction in property rights protection. A second common business corruption type is what we will call campaign contribution. This is a situation where business men are willing to contribute to an electoral campaign expecting some form of obedience or loyalty from public representatives. In this case it is possible to observe the dead weigh loss but not the welfare loss. It is important to remark that in either cases the source of money is private. Evidently, already with two quite similar kinds of corruption one can expect two different sorts of consequences. This corruption type is found all over the world. In recent years we have heard about this kind in the US as well as in Germany, France and Japan, among others. The American literature on corruption usually focuses on this corruption type. However there are at least two other kinds that are not so common everywhere. The second corruption kind is what we will call political corruption. This is quite close to what Gaviria (2002) considered as corruption involving only public parts. In this case the demand side is generally represented either by representatives or civil servants, or a combination of the two. In some cases there is the involvement of a third private party who is going to be the intermediate of the transaction. It is possible to identify two main political corruption types. One is what we call vote buying. This is an eminently political transaction since it occurs mainly between Executive and Legislative powers, and sometimes also involves the Judiciary power. It

The Multiple Faces of Wrongdoing

145

happens when the Executive has an interest in the approval of a proposal and does not have enough support for it. So the Executive buys representatives’ votes. This was the case of the above mentioned “Mensalão” in Brazil in 2005. In this case, the source of money may be public and/or private. It will depend on the arrangement to do the payments. If money is private we have again a dead weigh cost but if it is public we have income transfers from public service users to representatives and also to private middlemen. A second kind of political corruption is over-billing public purchases. The recent “Sanguessugas Scandal” in Brazil and the ambulances purchases are examples of this type. Here public money is the only source of money. Also in this case transfers are observed from public service users to bureaucrats, politicians (both from Legislative and Executive) and so-called entrepreneurs that sell the goods and services. The consequences of the two transfers indicated above could be quite different depending on who is being affected by these bad practices. The money could have been used to provide public health care services, or to build roads, or to strengthen the judicial system or to deliver education services. Each one of these public activities has a clearly defined objective. So the effects could be perceived in short or long run production losses, as well as in income distribution. When this kind of corruption becomes generalized, as nowadays in Brazil, it has another damaging effect that is the feeling in the population that every politician does the same and that it does not matter who is going to be elected. In this case democracy itself is at stake. This kind of corruption is not common everywhere. It can occur only where government structures face difficulties to govern. In this case a state weakening process is under way. This process is the very reason for this kind of corruption. It usually occurs in countries where, as Schwartzman (1982) pointed out, business is done inside the state, contrary to the American tradition, where business is done in the private sector. This kind of corruption opens the door to the third corruption kind. Criminal corruption is the third type. It is the most extreme form of corruption. It cannot be present in a situation where the state is strong. The difference from business corruption to criminal corruption resides on the fact that the first is a transaction between public and private agents that are legally established whereas the latter is not. Also in this case, one observes private money buying courts and political decisions as well as financing campaigns. In this way

146

Basilia Aguirre

we can say that criminal corruption has the same two sub-categories as business corruption: campaign contribution and decision buying corruption. However there are three important differences between business and criminal corruption. The first is that in this case money comes from illegal activities as gambling, drug traffic or deforestation. This means that the results of this kind of corruption are decisions that are not legally accepted by law even though they are legitimate by costumes and uses. In some circumstances it became the usual way of doing business. The second is that criminal corruption involves also necessarily Judiciary power. Without the judiciary acceptance it would be impossible that these transactions take place. The third point is that criminal corruption usually seeks to establish a parallel state and to acquire powers that pertain to state monopoly. When successful, it always reaches its goals. Dead weigh losses are also a characteristic of criminal corruption but state weakening is by far its larger threat. When criminal organizations manage to control some sectors or areas of a country, the degree of corruption reaches an extent where all three basic kinds of corruption are intertwined. In this case the cost of combating corruption reaches its highest level because it is almost an operation of war. Criminal corruption extends its arms as far as to prison agents and is usually disseminated all over the judicial and police system. The very nature of property rights protection is blurred since there are no more clear rules about what is to be protected. Criminal corruption subverts the very idea of rule of law because then there will be another set of rules that should be respected and this is not the official set of rules. These circumstances occur only where the state is totally captured by alien interests. In its extreme form one can observe parallel states like in Colombia or in parts or sectors of a country like the “favelas” in Rio de Janeiro, where the authority is in the hands of drug traffic dealers. It is very important to note that the logic of the action in each of these different types of corruption described here is different. And only understanding this logic it will be possible to combat them. The idea of introducing competition among interest groups to fight corruption is a good one when the demand for corruption comes from outside the state. However, when it originates inside the state this may not be a good idea because competition may simply be not available.

The Multiple Faces of Wrongdoing

IV.

147

Conclusion

Economic theory should incorporate the idea that there may be many types of corruption and that each can have different impacts on economic activity. We believe that only in understanding the way in which each type of corruption works, we will be able to offer better insights how to handle corruption and diminish its undesired effects. Finally, we have to be able to recognize that the specific institutional environment in which corruption takes place, may lead to different results. This implies that either a specific kind of corruption will appear in diverse forms in different institutional arrangements or that the interaction between various kinds of corruption will appear in different forms depending on the scenario. We have to conclude that there is a lot of work for future research in this area.

References

Bardhan, Pranab (2006), The Economist’s Approach to the Problem of Corruption, World Development, 34 (2), 341-348. Gaviria, Alejandro (2002), Assessing the effects of corruption and crime on firm performance: Evidence from Latin America, Emerging Markets Review 3, 245–268. Gupta, Sanjeev, Davoodi, Hamid, Alonso-Terme, Rosa (1998), Does corruption affect income inequality and poverty?, IMF Working Paper WP 98/76. Krueger, Anne O. (1974), The political economy of “rent-seeking”, American Economic Review 4 (3), 291-303. Mauro, Paolo (1997), Why Worry About Corruption?, IMF Economic Issues, No 6. Rose-Ackerman, Susan (1997), Corruption and Development, Paper prepared for the Annual Bank Conference on Development Economies, Washington, D.C. Schwartzman, Simon (1982), Bases do Autoritarismo Brasileiro, Rio de Janero Campus, Cap. 6, 115-147.

148

Basilia Aguirre

Tanzi, Vito, Davoodi, Hamid (1997), Corruption, Public Investment, and Growth, IMF Working Paper WP 97/139. Tullock, Gordon (1996), Corruption theory and practice, Contemporary Economic Policy 14 (3), 6-13.

Part 3: Trade and Integration Issues

Laura Márquez-Ramos and Inmaculada Martínez-Zarzoso

On Distance Effects in Gravity Models – Short Versus Long Distances Abstract In this paper, the effect of distance on international trade is investigated for different groups of countries and over time. With this aim, the distance variable is broken down into intervals to find out whether there is a different distance effect for near and far away importers. When the distance “classes” are introduced into the gravity model of trade, the results indicate that the distance effect is significantly higher for developed than for developing countries only for short distances, whereas for long distances (over 2500 km) the distance effect is very similar for both groups of countries. I. Introduction In the empirical literature that analyses the determinants of bilateral trade flows, the gravity model is a widely used and accepted framework. According to the standard gravity model of trade, the bilateral volume of exports between pairs of countries is a function of their incomes, their populations, their geographical distance and a set of dummies. The negative correlation between geographical distance and bilateral trade volumes is one of the most robust empirical findings in economics (Leamer and Levinsohn 1995). However, it is still unclear what is the information embodied in the distance coefficients estimated in gravity regressions. Filippini and Molini (2003) state that “distance is much more than geography: it is history, culture, language, social relations and many other things”.1 Recently, a number of authors have contributed to the debate on the interpretation of distance effects. Factors such as information costs, tastes and preferences, unfamiliarity, and differences in factor endowments have been considered. Loungani, Mody and Razin (2002) show that distance captures more than transport costs and that information costs may be behind the impact of 1

Filippini and Molini (2003) p. 699.

152

Laura Márquez-Ramos and Inmaculada Martínez-Zarzoso

distance on trade. Blum and Goldfarb (2006) find that distance is a good proxy for differences in tastes and preferences. Their results provide a new explanation for the persistent effect of distance in gravity regressions. It suggests that the distance effect in gravity regressions will persist for a number of products even if transportation costs, search costs, and other trade barriers associated with distance are reduced to zero, that is the case to some extent for internet trade. For the distance effect to disappear, there needs to be a homogenization of cultures. Huang (2007) shows that unfamiliarity can explain part of the negative correlation between geographical distance and bilateral trade volumes. He shows that higher uncertainty-aversion leads to lower trade flows to distant partners than gravity models predict. However, the interpretation the author does of the distance coefficient -higher negative coefficients in distance variable are interpreted as becoming less likely to trade with foreign countries that are far away- could be misleading. Finally, Melitz (2007) understands distance as a way to analyse the composition of international trade between different partners. Results show an increasing effect of distance on international trade. Furthermore, the impact of North-South distance has decreased over time. This author points out “at least half the rise in influence has a simple explanation, unrelated to transport costs. It flows from the shift in the composition of trade away from primary goods in agriculture and mining where differences in factor endowment are basic, toward sophisticated and highly differentiated products in manufacturing, where they are not. This shift in composition signifies a movement away from the sort of trade that rises with latitudinal distance and therefore can account for a good part of the rise in the negative coefficient of distance”.2 According to Buch, Kleinert and Toubal (2004) and MárquezRamos, Martínez-Zarzoso and Suárez-Burguet (2007), the magnitude and sign of the distance coefficient is related to the importance of bilateral activities with partners that are far away relative to those that are located nearby. Moreover, the coefficient of distance may differ among developed and developing countries. Márquez-Ramos et al. (2007) show that when controlling for country-heterogeneity the distance coefficient decreased by 13.55 per cent for developed countries and increased by 29.7 per cent for developing countries over the period 1980-1999. The authors classify each group of 2

Melitz (2007) p. 982.

On Distance Effects in Gravity Models

153

countries according to the different scenarios outlined by Buch et al. (2004). Developing countries can be placed in scenario 2,3 since the magnitude of the distance coefficient incre-ases over the period 1980-1999, whereas developed countries can be placed in scenario 3, since the magnitude of the distance coef-ficient decreases over the period 1980-1999. For developing coun-tries, export flows for small distances increase over time, whereas export flows for large distances decrease over time, and therefore trade with countries far away decreases in relation to trade with nearby countries. The opposite applies for developed countries. In order to have a better understanding of the role of distance in gravity equations, different intervals of distance should be considered to disentangle the effect of long distances from the effect of short distances on trade flows. To our knowledge, O’Sullivan and Ralston (1974) are the first authors to consider different distance intervals in gravity models. They examine the distribution of the errors for gravity models in terms of distance (by 100-mile distance intervals), showing that, in their sample, short distance movements are overestimated whereas long distance movements are underestimated. Eaton and Kortum (2002) also consider different distance segments when estimating the determinants of international trade. They estimate a gravity equation in which trade between countries is related to proximity, language, trading areas and distance. Their results show that the distance coefficient is higher in magnitude when goods traded come from more distant locations. The main aim of this paper is to disentangle the information embodied in each of the distance “classes” introduced in gravity regressions. With this purpose, the distance variable is broken down into six intervals and these six indicator variables are introduced in the gravity model of trade. Country-heterogeneity is also taken into account since the analysis is carried out for both developed and de3

Scenario 1: Distance costs decrease proportionally for all countries; all the information about the positive effect of decreasing distance costs is included in the constant term, which is larger. Scenario 2: The distance costs decrease non-proportionally and the decrease is greater for smaller distances, the distance coefficient increases over time. Scenario 3: The distance costs decrease non-proportionally and the decrease is smaller for smaller distances, the distance coefficient decreases over time in absolute terms.

154

Laura Márquez-Ramos and Inmaculada Martínez-Zarzoso

veloping countries. In the remainder of the paper, Section II presents the empirical analysis. Section III discusses the main results and Section IV concludes. II. Empirical analysis A gravity equation is estimated with data for 65 countries in 1980, 1985, 1990, 1995 and 1999 and a total of 4160 (65*64) bilateral trade flows are obtained. The model includes country dummies for exporters and importers (not reported) since the effects of originspecific and destination-specific unobservable market characteristics or multilateral resistance terms4 from both the exporter and the importer countries are considered (Anderson and van Wincoop, 2003). The presence of missing/zero values in the bilateral trade flow data for different years slightly reduces the sample (e.g. in 1999 there were 3433 observations). In order to model countryheterogeneity, a dummy, DP, is interacted with all the independent variables. DP takes a value of one when countries are classified as developed countries and zero otherwise (Márquez-Ramos et al., 2007). OLS estimation is performed on a double log specification for each year considered. The basic estimated equation is: ln X ij G i  O j  D1 ˜ Adjij  D 2 ˜ DP ˜ Adjij  D 3 ˜ PTA  D 4 ˜ DP ˜ PTA  D 5 ˜ FTA  D 6 ˜ DP ˜ FTA  D 7 ˜ CU  D8 ˜ DP ˜ CU  D 9 ˜ SM   D10 ˜ DP ˜ SM  D11 ˜ MU  D12 ˜ DP ˜ MU  D13 ˜ ln Distij 

(1)

D14 ˜ DP ˜ ln Distij  D15 ˜ Lang ij  D16 ˜ DP ˜ Lang ij  uij

where ln denotes natural logarithms. Xij denotes the value of exports from the exporter country i to the importer country j, įi denotes exporter dummies and Ȝj importer dummies. Income and population variables are not added, since in cross-section regressions their effect is reflected in the country specific effects. Adjij is a dummy that takes a value of 1 when countries share the same border and zero otherwise. The effects of regional integration agreements (RIAs) are modelled using five different levels of integration between pairs of countries: Preferential Trade Agreement (PTA), Free Trade Agreement (FTA), Customs Union (CU), Single Market (SM) and Monetary Union (MU). These dummies take the value of 1 when both trading partners belong to the same type of agreement. Langij is a dummy for countries 4

See the Anderson and Wincoop (2003) specification of the gravity model of trade.

On Distance Effects in Gravity Models

155

sharing the same language. Distij is the geographical great circle distance in kilometres between the capitals of country i and j.5 In a second step, equation (1) is estimated considering different intervals for distance. Distance intervals are the same as in Eaton and Kortum (2002). Then, the distance variable is disaggregated in six different intervals (in kilometres): [0, 600); [600, 1200); [1200, 2400); [2400, 4800); [4800, 9600) and [9600, maximum]. In this case, the interpretation of Buch et al. (2006) can be applied to each independent interval. In this estimation, the effect of cultural similarities and knowledge of other countries is controlled since adjacency; RIAs and language are included in the model, thus controlling familiarity effects. Since multilateral resistance terms have been included in the estimated model, country-characteristics such as uncertainly avoidance (Huang, 2007) are controlled for. III.

Estimation results

First, the baseline model specified in equation (1) is estimated with a single distance variable. Table 1 shows the results. Adjacency, integration variables and language display positive and significant coefficients indicating that countries sharing a common border, belonging to the same integration agreement and speaking the same official language, respectively, trade more. The distance coefficient shows the expected negative sign and its magnitude increases by 23 per cent from 1980 to 1999.6 According to Buch et al. (2006) the interpretation of this result is that trade with countries that are far away decreases with respect to trade with countries that are close to the home country, showing that the importance of bilateral activities with partners that are far away relative to those that are located nearby has decreased in the sample. The interaction variable DP*distance is only significant in the first and last years considered, but the magnitude of the coefficient is very small. The differences in the distance coefficient for developed and developing countries are in general not statistically significant.

5

6

Trade data is obtained from Statistics Canada (2001), adjacency and language from CIA and distance is the great circle distances between country capitals. RIAs information is obtained from World Trade Organisation (1995, 2005). Table A.1 in Appendix lists the RIAs considered to construct this variable. This result is frequently found in the gravity literature.

156

Laura Márquez-Ramos and Inmaculada Martínez-Zarzoso

Table 1: Determinants of international trade. The role of distance const. Adj DP*Adj PTA DP*PTA FTA DP*FTA CU DP*CU MU

1980 11.59*** (17.09) 1.05*** (4.86) -1.37*** (-5.00) 0.15 (0.77) 0.05 (0.19) 0.68** (2.30) -0.82*** (-2.65) 3.63*** (2.97) -4.59*** (-3.72) -

1985 12.77*** (18.71) 1.12*** (5.75) -1.56*** (-5.99) 0.14 (0.84) 0.01 (0.03) 0.59*** (2.44) -0.72*** (-2.78) 2.31** (2.26) -2.98*** (-2.88) -

1990 10.83*** (17.85) 1.06*** (5.42) -1.46*** (-5.87) 0.13 (1.06) 0.44* (1.73) 0.23 (0.84) -0.35 (-1.22) 1.03 (0.87) -1.63 (-1.37) -

1995 11.63*** (20.65) 0.88*** (4.90) -1.56*** (-6.86) 0.01 (0.11) 0.61** (2.44) 0.21 (0.96) -0.23 (-0.99) 1.41*** (4.93) -

1999 12.62*** (21.03) 1.03*** (5.69) -1.56*** (-6.70) 0.01 (0.09) 0.45* (1.83) 0.13 (0.67) 0.06 (0.27) 0.83*** (4.30) -

-0.74*** (-5.92) DP*MU -0.23* (-1.80) 0.64*** 0.80*** 0.71*** 0.74*** Lang 0.54*** (4.21) (5.15) (7.08) (6.78) (7.34) -0.19 -0.29** DP*Lang -0.05 -0.17 -0.53*** (-0.28) (-1.01) (-3.20) (-1.34) (-2.16) -1.05*** -1.11*** -1.16*** -1.29*** Dist -1.06*** (-20.44) (-20.23) (-22.53) (-28.56) (-30.73) -0.01 0.01 0.01 0.03** DP*Dist -0.03** (-2.01) (-0.73) (0.74) (0.48) (2.22) 0.77 0.77 0.81 0.85 0.86 R2 No. obs. 3128 3094 3230 3415 3433 Notes: ***, **, * indicate significance at 1 per cent, 5 per cent and 10 per cent, respectively. T-statistics are in brackets. The dependent variable is the natural logarithm of export values (current US$). Distance is also in natural logarithms. The estimation uses White’s heteroscedasticity-consistent standard errors. DP is a dummy variable that takes the value one when the countries are richer than the average in the sample and zero otherwise.

On Distance Effects in Gravity Models

157

Developing countries trade more when they share a common border, although the adjacency coefficient slightly decreases over time. The positive coefficient of the integration variables decreases over time for lower levels of economic integration (PTA, FTA and Customs Union). Since MERCOSUR is not working in practice as a Common Market, it has been considered as a Customs Union. Among all the integration dummies, only the Customs Union dummy present a positive and significant coefficient in all years. Finally, the positive effect on bilateral trade of sharing language increases over time and it is higher for developing countries. Table 2: Determinant of international trade. The role of distance intervals (to be continued) 1985 10.32*** (8.16) 0.76*** (3.81) -0.56** (-2.02) 0.13 (0.78) 0.24 (0.60) 0.35 (1.38) 0.05 (0.17) 2.35** (2.52) -2.28** (-2.41) -

1990 9.33*** (8.34) 0.84*** (4.14) -0.61** (-2.21) 0.15 (1.21) 0.42 (1.64) 0.09 (0.29) 0.18 (0.56) 1.05 (0.90) -1.09 (-0.92) -

1995 9.11*** (8.31) 0.53*** (2.95) -0.36 (-1.43) 0.05 (0.52) 0.51** (2.09) -0.13 (-0.56) 0.34 (1.39) 1.41*** (4.89) -

1999 10.67*** (9.70) 0.73*** (4.03) -0.53** (-2.04) 0.07 (0.59) 0.39 (1.59) -0.10 (-0.51) 0.35 (1.55) 0.71*** (3.56) -

MU

1980 9.75*** (7.75) 0.72*** (3.12) -0.48 (-1.65) 0.16 (0.83) 0.25 (0.88) 0.44 (1.41) -0.16 (-0.47) 3.60*** (2.98) -3.91*** (-3.20) -

-

DP*MU

-

-

-

-

Lang

0.46*** (3.54) 0.05 (0.28)

0.53*** (4.29) -0.04 (-0.25)

0.71*** (6.19) -0.38** (-2.27)

0.63*** (5.94) -0.07 (-0.54)

-0.35*** (-2.77) 0.13 (1.06) 0.68*** (6.83) -0.18 (-1.35)

constant Adj DP*Adj PTA

DP*PTA FTA DP*FTA CU DP*CU

DP*Lang

158

Laura Márquez-Ramos and Inmaculada Martínez-Zarzoso

Table 2 (continued): Determinant of international trade. The role of distance intervals Dist [0, 600) DP*Dist [0, 600) Dist [600, 1200) DP*Dist [600, 1200) Dist [1200, 2400) DP*Dist [1200, 2400) Dist [2400, 4800) DP*Dist [2400, 4800) Dist [4800, 9600)

1980 1985 -0.65*** -0.57*** (-3.09) (-2.97) -0.33*** -0.32*** (-3.13) (-4.33) -0.71*** -0.59*** (-4.06) (-3.53) -0.29*** -0.32*** (-6.24) (-7.22) -0.79*** -0.67*** (-5.03) (-4.48)

1990 -0.81*** (-4.23) -0.26*** (-2.88)

1995 -0.64*** (-3.77) -0.33*** (-4.80) -0.87*** -0.73*** (-5.54) (-5.07) -0.20*** -0.24*** (-4.61) (-7.01) -0.90*** -0.78*** (-6.33) (-5.98)

1999 -0.87*** (-5.09) -0.31*** (-3.95) -0.94*** (-6.61) -0.21*** (-5.90) -1.00*** (-7.68)

-0.14*** -0.17*** -0.08** -0.10*** -0.08*** (-4.53) (-5.42) (-2.53) (-4.04) (-3.09) -0.83*** -0.75*** -0.92*** -0.84*** -1.05*** (-5.72) (-5.39) (-7.11) (-6.94) (-8.86) -0.03 -0.03 -0.01 -0.01 0.00 (-1.09) (-1.20) (-0.22) (-0.48) (-0.03) -0.86*** -0.79*** -0.95*** -0.90*** -1.09*** (-6.45) (-6.13) (-7.97) (-8.07) (-9.95)

DP*Dist [4800, 9600)

-0.07*** -0.04** -0.03** -0.02 0.01 (-4.10) (-2.46) (-2.09) (-1.36) (0.81) Dist [9600, max] -0.90*** -0.82*** -0.99*** -0.92*** -1.11*** (-6.99) (-6.73) (-8.71) (-8.67) (-10.59) DP*Dist [9600, max] 0.00 0.02 0.05*** 0.03** 0.04*** (0.13) (0.84) (2.92) (2.05) (2.94) 0.78 0.78 0.82 0.85 0.86 R2 no. obs. 3128 3094 3230 3415 3433

Notes: ***, **, * indicate significance at 1 per cent, 5 per cent and 10 per cent, respectively. T-statistics are in brackets. The dependent variable is the natural logarithm of export values (current US$). Distance is also in natural logarithms. The estimation uses White’s heteroscedasticity-consistent standard errors. DP is a dummy variable that takes the value one when the countries are richer than the average in the sample and zero otherwise.

Table 2 shows the results when the distance variable is broken down into different intervals. We observe some interesting differ-

On Distance Effects in Gravity Models

159

ences with respect to Table 1. First, the observed heterogeneity in the estimated coefficient for most of the dummies disappears. That is, the coefficients of the interaction variables are insignificant for most of the years in the sample. The adjacency presents the expected positive sign with a higher effect for developing countries and the language dummy increases in magnitude over time. The estimated coefficients for both dummy variables are not always significantly different from those estimated for developed countries. Part of the effect of these variables is now embodied in some of the distance classes. In fact, heterogeneity in the estimated coefficients for the distance intervals is observed for short distances (less than 2400km). For developed countries, distance shows a significantly higher coefficient than for developing countries for the three first distance classes considered. Second, for distances higher than 2400 km, the distance effect does not significantly change for the two groups of countries considered. Table 3 shows the evolution of the distance-classes coefficients over time and for developed and developing countries. Similar to Eaton and Kortum (2002), the distance variable has a higher effect on trade flows for longer distances, but only for developing countries (second part of Table 3), whereas for developed countries the distance effect decreases for longer distances. It can also be observed that the effect of distance increases over time in all the intervals, for both developed and developing countries. The increase over the period 1980-1999 of the magnitude of the distance coefficient is higher for developing countries in all the intervals. Figures 2 and 3 show graphically this evolution for developed and developing countries, respectively. Finally, it seems that certain convergence exists regarding the decreasing non-proportionally distance costs among developed and developing countries. With respect to the lower distance effect for longer distances found for developed countries, an explanation could be found in the changing composition of trade between more distant countries. In the last two decades, bilateral trade in sophisticated-manufactured products (intra-industry trade) has increased more than trade based on comparative advantage. Usually these products are less sensitive to transportation costs since they are high-value added goods. However, for developing countries the distance effect increases with distance. These countries in many cases produce and export basic commodities and the costs of transporting these products increase with distance due to the increasing price of oil.

160

Laura Márquez-Ramos and Inmaculada Martínez-Zarzoso

Table 3: The role of different distance intervals for developed and developing countries. Developed countries Distance [0, 600) [600, 1200) [1200, 2400) [2400, 4800) [4800, 9600) [9600, max] increase in per cent

1980 0.99 0.99 0.93 0.86 0.93 0.89

1985 0.89 0.91 0.85 0.78 0.83 0.81

1990 1.07 1.07 0.98 0.93 0.99 0.95

1995 0.97 0.96 0.89 0.85 0.92 0.89

1999 1.18 1.15 1.08 1.05 1.08 1.07

-10.10

-8.99

-11.21

-8.25

-9.32

increase in per cent

19.19 16.16 16.13 22.09 16.13 20.22

Developing countries increase in per cent

Distance 1980 1985 1990 1995 1999 [0, 600) 0.65 0.57 0.81 0.64 0.87 33.85 [600, 1200) 0.71 0.59 0.87 0.73 0.94 32.39 [1200, 2400) 0.79 0.67 0.90 0.78 1.00 26.58 [2400, 4800) 0.83 0.75 0.92 0.84 1.05 26.51 [4800, 9600) 0.86 0.79 0.95 0.90 1.09 26.74 [9600, max] 0.90 0.82 0.99 0.92 1.11 23.33 increase in per cent 38.46 43.86 22.22 43.75 27.59 Note: Evolution of the distance coefficient is reported. Columns indicate year, and rows distance interval. The last column shows the increase (in per cent) of the distance coefficient over the period 1980-1999.

IV.

Conclusions

Information costs, tastes and preferences, unfamiliarity, and differences in factor endowments have been recently associated to the negative correlation between geographical distance and bilateral trade volumes. However, according to the interpretation of Buch et al. (2004), the distance coefficient is related to the importance of bilateral activities with partners that are far away relative to those that are located nearby. Our results show that breaking down distance into different intervals partially corrects the effects associ-

On Distance Effects in Gravity Models

161

ated to distance, since lower coefficients are found in adjacency and language. Regarding the decreasing non-proportionally distance costs stated by Buch et al. (2004), we find certain convergence among countries since the magnitude of the distance coefficient increases more rapidly for developing than for developed countries and the magnitudes shown in 1999 are for all distance classes and for both groups of countries around unity.

References

Anderson, James E., van Wincoop, Eric (2003), Gravity with gravitas: A solution to the border puzzle, American Economic Review 93, 170-192. Baier, Scott L., Bergstrand, Jeffrey H. (2005), Do free trade agreements actually increase members’ international trade?, Federal Reserve Bank of Atlanta Working Paper 2005-03. Blum, Bernardo S., Goldfarb, Avi (2006), Does the internet defy the law of gravity?, Journal of International Economics 70(2), 384-405. Buch, Claudia M., Kleinert, Jörn, Toubal, Farid (2004), The distance puzzle: on the interpretation of the distance coefficient in gravity equations, Economics Letters 83, 293-298. CIA-Central Intelligence Agency (2003), The World Factbook,  http://www.odci.gov/cia/publications/factbook (last visit: November 2007). Eaton, Jonathan, Kortum, Samuel (2002), Technology, geography and trade, Econometrica 70 (5), 1741-1779. Filippini, Carlo, Molini, Vasco (2003), The determinants of East Asian trade flows: a gravity equation approach, Journal of Asian Economics 14, 695-711. Huang, Rocco R. (2007), Distance and trade: Disentangling unfamiliarity effects and transport cost effects, European Economic Review 51(1), 161-181. Leamer, Edward, Levinsohn, James (1995), International Trade Theory: The Evidence, in: Gene Grossman and Keneth Rogoff

162

Laura Márquez-Ramos and Inmaculada Martínez-Zarzoso

(eds.), Handbook of International Economics, vol. 3, Elsevier Science BV: Amsterdam, 1339-1394. Loungani, Prakash, Mody, Ashoka, Razin, Assaf (2002), The Global Disconnect: The Role of Transactional Distance and Scale Economies in Gravity Equations, Scottish Journal of Political Economy 49(5), 526-543. Márquez-Ramos, Laura, Martínez-Zarzoso, Inmaculada, SuárezBurguet, Celestino (2007), The Role of Distance in Gravity Regressions: Is There Really a Missing Globalisation Puzzle?, Topics in Economic Analysis & Policy 6 (1), 1-25. Melitz, Jaques (2007), North, South and distance in the gravity model, European Economic Review 51(4), 971-991. O’Sullivan, Patrick, Ralston, Bruce (1974), Forecasting Intercity Commodity Transport in the USA, Regional Studies 8, 191-195. Statistics Canada (2001): World Trade Analyzer. The International Trade Division of Statistics of Canada. World Trade Organization (1995): Regionalism and the World Trading System. Geneva. World Trade Organization (2005) webpage, http://www.wto.org/ (last visit: November 2007).

On Distance Effects in Gravity Models

163

Appendix Table A.1: Trade Agreements (in chronological order of date of entry into force) (to be continued) Date Type of agreement and related provisions Salvador-Nicaragua FTA 1951 Free Trade Agreement (GATT Art. XXIV) EC (Treaty of Rome) 1958 Customs Union (GATT Art. XXIV) EFTA (Stockholm Convention) 1960 Free Trade Agreement (GATT Art. XXIV) LAFTA (Latin American FTA) 1961- Free Trade Agreement 1979 (GATT Art. XXIV) CAFTA (Central American FTA) 1961- Free Trade Agreement 1975 (GATT Art. XXIV) FINEFTA 1961 Interim agreement for the formation of a FTA (GATT Art. XXIV) African Common Market 1963 Customs Union (GATT Art. XXIV) Arab Common Market 1965 Interim agreement for the formation of a FTA leading to a customs union (GATT Art. XXIV) Ireland-United Kingdom FTA 1966 Free Trade Agreement (GATT Art. XXIV) Trade Expansion and Cooperation 1968 Preferential Arrangement Agreement (TRIPARTITE) (Enabling Clause) EFTA-FINEFTA accession of 1970 Free Trade Agreement Iceland (GATT Art. XXIV) Protocol relating to Trade 1973 Preferential Arrangement Negotiations among developing (Enabling Clause) countries (PTN) EC-Accession of Denmark, 1973 Customs Union Ireland and United Kingdom (GATT Art. XXIV) EC-EFTA Free Trade Agreement 1973 Free Trade Agreement (GATT Art. XXIV) CARICOM (Caribbean 1973 Customs Union Community and Common (GATT Art. XXIV) Market) Bulgaria-Finland FTA 1975 Free Trade Agreement (GATT Art. XXIV)

164

Laura Márquez-Ramos and Inmaculada Martínez-Zarzoso

Table A.1 (continued): Trade Agreements

Bangkok Agreement PTA for Eastern and Southern African States LAIA (Latin American Integration Association) EC- Accession of Greece Israel-United States FTA ECO (Economic Cooperation Organization) EC- Accession of Portugal and Spain CUFTA (Canada-United States FTA) Andean Group (CAN)

Date Type of agreement and related provisions 1976 Preferential Arrangement (Enabling Clause) 1981 Preferential Arrangement (Enabling Clause) 1981 Preferential Arrangement (Enabling Clause) 1981 Customs Union (GATT Art. XXIV) 1985 Free Trade Agreement (GATT Art. XXIV) 1985 Preferential Arrangement (Enabling Clause) 1986 Customs Union (GATT Art. XXIV) 1988 Free Trade Agreement (GATT Art. XXIV) 1988 Preferential Arrangement (Enabling Clause) 1989 Preferential Arrangement (Enabling Clause)

General System of Trade Preferences among developing countries (GSTP) MERCOSUR (Southern Common 1991 Customs Union (Enabling Clause) Market) EFTA-Turkey FTA 1992 Free Trade Agreement (GATT Art. XXIV) Cross Border Initiative 1992 Common Policy Framework--PTA EFTA-Czech and Slovak 1992 Interim agreement for the Republic FTA formation of a FTA (GATT Art. XXIV) CACM (Central American 1993 Customs Union Common Market) (GATT Art. XXIV) EFTA-Israel FTA 1993 Free Trade Agreement (GATT Art. XXIV) EFTA-Poland FTA 1993 Interim agreement for the formation of a FTA Czech Republic-Slovak Republic 1993 Customs Union (GATT Art. Customs Union Agreement XXIV)

On Distance Effects in Gravity Models

165

Table A.1 (continued): Trade Agreements

CEFTA (Central Europe FTA) EFTA-Bulgaria FTA Single Market EU EU-Bulgaria FTA EU-EFTA EEA (European Economic Area) NAFTA (North American Free Trade Agreement) EU (Accession of Austria, Finland and Sweden) Andean Group (CAN) EU-Israel agreement MERCOSUR (Southern Common Market) SAPTA (South Asian Preferential Trade Arrangement) EU-Turkey Canada-Chile Canada-Israel Israel-Turkey CEFTA- Accesion of Bulgaria European Monetary Union (11 members) Chile-Mexico Bulgaria-Turkey

Date Type of agreement and related provisions 1993 Free Trade Agreement (GATT Art. XXIV) 1993 Free Trade Agreement (GATT Art. XXIV) 1993 Single Market 1994 Free Trade Agreement (GATT Art. XXIV) 1994 Single Market 1994 Free Trade Agreement (GATT Art. XXIV) 1995 Single Market (GATT Art. XXIV) 1995 Customs Union (Enabling Clause) 1995 Agreement on the implementation of a FTA 1995 Single Market (Enabling Clause) 1995 Preferential Arrangement (Enabling Clause) 1996 Customs Union (GATT Art. XXIV) 1997 Free Trade Agreement (GATT Art. XXIV) 1997 Free Trade Agreement (GATT Art. XXIV) 1997 Free Trade Agreement (GATT Art. XXIV) 1998 Free Trade Agreement (GATT Art. XXIV) 1999 Monetary Union 1999 Free Trade Agreement (GATT Art. XXIV) 1999 Free Trade Agreement (GATT Art. XXIV)

Source: WTO (2005), Regional Trade Agreements Notified to the GATT/WTO and in Force WTO (1995), Baier and Bergstrand (2005).

166

Laura Márquez-Ramos and Inmaculada Martínez-Zarzoso

Figure A1: Selected countries

Algeria Argentina Australia Austria Belgium-Luxembourg Bolivia Brazil Bulgaria Canada Chile China Colombia Costa Rica Croatia Cyprus Czech Republic Denmark Dominican Republic Ecuador Egypt, Arab Rep. El Salvador Finland France Germany

Ghana Greece Honduras Hong Kong, China Iceland India Ireland Israel Italy Jamaica Japan Kenya Korea, Rep. Mexico Mozambique Nepal Netherlands Nicaragua Norway Pakistan Panama Paraguay Peru Poland

Portugal Senegal Singapore Slovak Republic South Africa Spain Sudan Sweden Switzerland Syrian Arab Republic Tanzania Trinidad and Tobago Turkey United Kingdom United States Uruguay Venezuela

Developed countries: Argentina, Australia, Austria, Belgium-Luxembourg, Canada, Chile, Costa Rica, Cyprus, Czech Republic, Denmark, Finland, France, Germany, Greece, Hong Kong, Iceland, Ireland, Israel, Italy Japan, Netherlands, Norway, Poland, Portugal, Rep. Korea, Singapore, Slovak Republic, South Africa, Spain, Sweden, Switzerland, United Kingdom, United States, Uruguay. Developing countries: Algeria, Arab Rep. Egypt, Bolivia, Brazil, Bulgaria, China, Colombia, Croatia, Dominican Republic, Ecuador, El Salvador, Ghana, Honduras, India, Jamaica, Kenya, Mexico, Mozambique, Nepal, Nicaragua, Panama, Pakistan, Paraguay, Peru, Senegal, Syrian Arab Republic, Tanzania, Trinidad and Tobago, Turkey, Venezuela.

On Distance Effects in Gravity Models

167

Figure A2: Distance elasticity in developed countries Distance [0, 600)

0,00 -0,20

1980

1985

1990

1995

1999

-0,40

Distance [600, 1200)

-0,60

Distance [1200, 2400)

-0,80

Distance [2400, 4800)

-1,00

Distance [4800, 9600)

-1,20 -1,40

Distance [9600, max]

Figure A3: Distance elasticity in developing countries. Distance [0, 600) 0,00 -0,20 -0,40 -0,60 -0,80 -1,00 -1,20

1980

1985

1990

1995

1999

Distance [600, 1200) Distance [1200, 2400) Distance [2400, 4800) Distance [4800, 9600) Distance [9600, max]

Dierk Herzer

Trade, its Composition and Total Factor Productivity: Cointegration Evidence for Chile

Abstract Using cointegration analysis, this study examines the long-run impact of capital goods imports, exports of manufactured and primary goods on total factor productivity in Chile. The study finds productivity-enhancing effects of capital goods imports as well as manufactured exports, and productivity-limiting effects of primary exports. This finding indicates that aggregate trade measures, such as the ratio of imports plus exports to GDP, may mask important differences between different trade categories. I. Introduction Does international trade increase the productivity of an economy? This question has been the subject of considerable theoretical and empirical research in recent years. In theory, there are several arguments in favor of a long-run relationship between trade and aggregate productivity. As far as imports are concerned, it is argued that in particular capital goods imports may increase domestic productivity, since capital goods embody technological knowledge generated by research and development (R&D). The logic behind this is that foreign R&D activities spill over from one country to another through international trade. In this context, the literature usually suggests two basic mechanisms for trade-related R&D spillover effects (see, e.g., Grossman and Helpman 1991, RiveraBatiz and Romer 1991). First, knowledge spillovers may arise from studying blueprints of new capital goods, so that the recipient countries can imitate the technology. If the cost of obtaining knowledge is less than the cost of the corresponding invention, then a knowledge externality occurs. Second, if R&D expenditures by countries from abroad create new capital goods that are different or better than those that already exist, then the productivity of an im-

170

Dierk Herzer

porting country improves by employing a wider variety of capital inputs or by simply using better capital inputs in final production. As far as the productivity effects of exports are concerned, it is argued that an expansion in exports may promote specialization in sectors in which a country has comparative advantage, and lead to a reallocation of resources from the relatively inefficient non-trade sector to the more productive export sector. Moreover, the growth of exports can increase productivity by offering larger economies of scale (Helpman and Krugman 1985). And finally, increasing exports may affect aggregate productivity through dynamic spillover effects on the rest of the economy (Feder 1983). The possible sources of these externalities include productivity enhancements resulting from increased competitiveness, more efficient management styles, better forms of organization, labor training, and knowledge about technology and international markets. Chuang (1998), for example, argues that entering competitive international markets requires knowledge about foreign buyer's specifications, quality and delivery conditions. To satisfy these requirements, foreign purchasers help and teach local exporters to establish each stage of the production process and improve management and marketing practices. The development of efficient quality control procedures, management and marketing methods, product specifications and production guidelines is simultaneously fostered by the competitive pressure in the international markets. Consequently, knowledge is generated through a systematic learning process initiated by exports and spilling over to the rest of the economy. In this context, the literature usually distinguishes between manufactured and primary exports. Lucas (1993), among others, points out that the dynamic technological spillover effects are mainly associated with manufactured exports rather than with primary exports. Furthermore, several authors hypothesize that primary exports are actually an obstacle to greater productivity growth. The main arguments advanced in support of this hypothesis are: (i) Primary products offer no sustainable potential for knowledge spillovers, and an increase in primary exports can draw resources away from the externality-generating manufacturing sector (Matsuyama 1992, Sachs and Warner 1995). (ii) Primary exports are subject to extreme price and volume fluctuations. Increasing primary exports may therefore lead to increasing GDP variability and macroeconomic uncertainty. High instability and uncertainty may, in turn, hamper efforts at economic planning and

Trade, its Composition and Total Factor Productivity: Cointegration Evidence for Chile

171

reduce the quantity as well as the efficiency of investments (Dawe 1996). Consequently, the theoretical literature suggests that the effects of exports on economic productivity differ significantly between primary and manufactured products. Admittedly, the empirical literature on trade and total factor productivity generally uses aggregate measures of international trade, such as the ratio of imports plus exports to GDP, usually referred to as openness (see, e.g., Frankel and Romer 1999, Miller and Upadhyay 2000, Jonsson and Subramanian 2001, Alcalá and Ciccone 2004). Without exception, all studies appear to find a statistically significant positive relationship between trade and total factor or labor productivity. Most studies on this topic are based on cross-country data. Few studies apply panel-data regressions (see, e.g., Miller and Upadhyay 2000). To our knowledge, only Jonsson and Subramanian (2001) and Thangavelu and Rajaguru (2004) have investigated the relationship between trade and aggregate productivity using time series techniques. This study differs from previous works in two main ways: first, given that aggregate trade measures may mask important differences between different trade categories, we disaggregate trade into capital goods imports, manufactured exports, and primary exports. Second, given the problems inherent to cross-country studies, such as simultaneity bias and parameter heterogeneity (see, e.g., Levine and Renelt 1992, Ericsson et al. 2001), we apply time series analysis. More concretely, we apply unit root tests that allow for an unknown number of structural breaks. Moreover, we use cointegration techniques to examine the long-run impact of trade on total factor productivity. Finally, misspecification and structural stability tests are conducted for the estimated long-run relation between capital goods imports, primary and manufactured exports and total factor productivity. In order to investigate the impact of capital goods imports, primary and manufactured exports on productivity, we use Chilean time series data from 1960 - 2001. Chile is an interesting case study for several reasons. First, Chile experienced a pattern of high longrun growth, which, admittedly, was interrupted by the collapse of the Allende government in 1973, the 1975 recession, and the 1982 economic crisis. As shown by Chumacero and Fuentes (2005) total factor productivity has played an important role in this growth process, especially in the period 1975 - 1981 and after 1985. Second, Chilean exports and imports grew very rapidly after 1974,

172

Dierk Herzer

when trade liberalization was initiated. Admittedly, this growth was interrupted by the balance-of-payments crisis in 1982. Third, Chilean exports rely heavily on primary products, although the share of manufacturing exports in goods exports rose from about 7 percent in 1973 to 44 percent in 2001. And fourth, as found by Romaguera and Contreras (1995), Chile is extremely vulnerable to fluctuating commodity prices, especially copper prices. Today, copper still accounts for about 37 per cent of total exports of goods in Chile.1 The rest of the paper is organized as follows. Section II. presents the empirical model and the data. The econometric methodology is described in Section III. The estimation results are presented in Section IV. A final section summarizes the conclusions. II. Empirical model and data A. The Model Our objective is to investigate whether and how increasing capital goods imports, manufactured exports and primary exports affect total factor productivity. To this end, we assume that the level of total factor productivity, TFPt, at time t can be expressed as a CobbDouglas function of capital goods imports, CMt, manufactured exports, IXt, primary exports PXt, and other exogenous factors Ct:

TFPt

Ct CM tD IX tE PX G

(1)

where D , E , and G are the elasticities of TFPt with respect to CMt, IXt, and PXt . Taking natural logs (ln) of both sides of equation (1) gives the estimable linear function:

ln TFPt c  D ln CM t  E ln IX t  G ln PX t  et

(2)

in which all coefficients are elasticities, c is a constant parameter, and et is the usual error term with mean zero and finite variance V 2 , which reflects the influence of all other factors. Since equation (2) can be interpreted as the long-run (cointegration) relationship between the variables, we estimate D , E , and G using cointegration techniques. 1

In 1971 -1973 the share of copper represented almost 80 per cent of total exports of goods, and the share of minerals as a whole announced to almost 90 per cent. See, for example, Agosin (1999).

Trade, its Composition and Total Factor Productivity: Cointegration Evidence for Chile

173

Before estimation, we follow Miller and Upadhyay (2000) and Jonsson and Subramanian (2001) and calculate lnTFPt on the basis of a constant-returns-to-scale Cobb-Douglas production function,

Yt

TFPt K ta L(t1 a)

(3)

where Yt denotes the aggregate production of the economy, and TFPt, Kt, Lt stand for the level of total factor productivity, the capital stock, and labor, respectively. The key parameters necessary for the calculation are the factor-output elasticities a and (1 - a), respectively. From the growth accounting point of view, these parameters are given by the capital and labor shares from the national accounts. The average capital share in the period 1960 - 2001 was calculated to be a = 0.5, which is in line with the results of Chumacero and Fuentes (2005), who report an average value of a = 0.507 for the period 1960 - 2000. Therefore, we calculate the natural logarithm of total factor productivity according to2

ln TFPt

ln Yt  0.5 ln K t  0.5 ln Lt

(4)

B. Data The data used to calculate ln TFPt and to estimate equation (2) are annual from 1960 to 2001. They were gathered from the Indicadores económicos y sociales de Chile 1960-2000 and the Boletínes mensuales (various issues) published by the Chilean Central Bank. The output, Yt, is measured by real Chilean GDP. Kt is the Chilean capital stock in real terms, which was computed on the basis of accumulated capital expenditure using the perpetual inventory method. GDP, capital stock, capital goods imports, exports of manufactured products, and primary products are evaluated in Chilean pesos at constant 1996 prices. The labor variable, Lt, is represented by the total number of people employed during year t. Figure 1 shows the evolution of total factor productivity, capital goods imports, exports of manufactured products, and primary products in the period under consideration. (All variables are in logarithms).

2

In a preliminary study on this topic, we calculated the Chilean total factor productivity based on the conventional value of a = 1/3 (see Herzer, 2006). Our previous results (obtained by different estimation methods) are very similar to the present study.

174

Dierk Herzer

Figure 1: Time series used 15.0

4.1

14.5

4.0

14.0

3.9 13.5

3.8 13.0

3.7

12.5

3.6 1960 1965 1970 1975 1980 1985 1990 1995 2000

12.0 1960 1965 1970 1975 1980 1985 1990 1995 2000

logarithms of real capital goods imports, lnCM

logarithms of total factor productivity, lnTFP 16

16.0

15

15.5

14

15.0

13

14.5

12

14.0

11

13.5

10 1960 1965 1970 1975 1980 1985 1990 1995 2000

13.0 1960 1965 1970 1975 1980 1985 1990 1995 2000

logarithms of real manufactured exports, lnIX

III.

logarithms of real primary exports, lnPX

Methodology

A. Time series properties From Figure 1, it can be inferred that all variables are trending. Therefore, we test lnTFPt, lnCMt, lnIXt, and, lnPXt for unit roots to verify their order of integration. It is well known that standard unit root tests are biased in favor of identifying data as integrated if there are structural changes. For all the series there is indeed a strong likelihood that structural discontinuities are present (e.g., the socialist government of President Allende (1970-1973), the drastic trade liberalization initiated in 1974 by the military Regime of

Trade, its Composition and Total Factor Productivity: Cointegration Evidence for Chile

175

Pinochet, the 1975 economic crisis, and the deep 1982 recession). Therefore, we use the unit root test recently developed by Kapetanios (2005). The Kapetanios procedure permits a formal evaluation of the time series properties in the presence of structural breaks at unknown points in time. It allows the break dates to be identified endogenously through the testing procedure itself. In order to test the unit-root hypothesis against the alternative of one or more structural breaks, we estimate three models of the Dickey-Fuller type without any prior knowledge of any potential break dates:

y3 t

m

k

j 1

i 1

m

k

j 1

i 1

y1t

P1  b1t  a1 yt 1  ¦ G1 DU j ,t  ¦ c1i 'y t i  e1t

y2t

P2  b2t  a2 yt 1  ¦ G 2 DTj,t  ¦ c2i 'yt i  e2t m

m

k

j 1

i 1

i 1

(5) (6)

P3  b3t  a3 yt 1  ¦ G 3 DU j ,t  ¦ G 4 DTi ,t  ¦ c3i 'yt i  e3t (7)

where y n t , n = 1, 2, 3, is the variable considered (lnTFPt, lnCMt, lnIXt, lnPXt), k is the lag length, t is the time trend, m denotes the number of structural breaks, and DUt = 1 (t >TB), DTt = 1 (t > TB)(t -TB) are indicator dummy variables for the break at time TB  T (T = 42, 1 ” t ” 42). As can be seen, Model (5) allows for up to m changes in the intercept of the trend function, while Model (6) accounts for m breaks in the slope of the trend function without a change in the level. Model (7) allows for both effects to take place simultaneously. In the empirical analysis we consider the possibility that up to two break points occurred over the relevant period: m = 1, m = 2. The first break point is chosen by estimating the models for each possible break date in the data set, and TB is selected as the value which is associated with the minimum sum of squared residuals ( ¦ (e1t ) 2 , ¦ (e2t ) 2 , ¦ (e3t ) 2 ). If the corresponding minimum t-statistic of the hypothesis ai = 1 under model i = 1, 2, 3 does not exceed (in absolute value) the critical value reported by Kapetanios (2005), the unit root hypothesis is not rejected. Imposing the estimated break date on the sample, we start looking for the second break. Again, the second break point is chosen with the minimum sum of squared residuals in order to test the null hypothesis ai = 1.

176

Dierk Herzer

B. Cointegration The second step is an investigation of the cointegration properties of the time series. To this end, we apply the autoregressive distributed lag (ARDL) approach developed by Pesaran et al. (2001), which is also known as the bounds test procedure. This procedure is applicable irrespective of whether the explanatory variables are I(1) or I(0). It is based on an unrestricted error correction model, which in our case is given by

' ln TFPt D 0  G1 ln TFPt 1  G 2 ln CMt 1  G 3 ln IX t 1  G 4 ln PX t 1 k

k

k

k

i 1

i 0

i 0

i 0

¦ Ei ' ln TFPt i  ¦ J i ' ln CMt i  ¦ Oi ' ln IX t i  ¦Ti ' ln PX t i  Pt

(8)

In this model, which can be interpreted as an autoregressive distributed lag model, we test the absence of a long-run relationship between lnTFPt, lnCMt, lnIXt, and lnPXt by calculating the F-statistic for the null of no cointegration H 0 : G1 G 2 G 3 G 4 0 against the alternative H1 : G1 z G 2 z G 3 z G 4 z 0 . The distribution of the test statistic under the null depends on the order of integration of the variables. In the case where (a) all four variables are I(0) the asymptotic 1 per cent critical value is 4.29 (see Pesaran et al. 2001, p. 300, Table CI(iii)). If the calculated Fstatistic falls below this value, the null hypothesis cannot be rejected (at the 1 per cent significance level). In the case where (b) one or more series are I(0) and one or more series are I(1), the critical value falls in the interval [4.29, 5.61]. If the F-statistic falls within these bounds, the result is inconclusive. Thus, the order of integration must be known before any conclusion can be drawn. In the case where (c) lnTFPt, lnCMt, lnIXt, and lnPXt are I(1), the 1 per cent critical value is 5.61. If the F-statistic lies above 5.61, the null of no cointegration is rejected at the 1 per cent level. C. Long-run elasticities If lnTFPt, lnCMt, lnIXt, and lnPXt are found to be cointegrated, we use the Stock (1987) approach, which involves estimating equation (8) in order to obtain D , E , and G . Given the low frequency of the data (annual) and the small sample size, we set the lag length k equal to 2. Then, following Lütkepohl and Wolters (1998), the general-to-specific approach is used by removing insignificant variables step by step until there remain only coefficients significant at the 1 per cent level. Normalizing on the total factor produc-

Trade, its Composition and Total Factor Productivity: Cointegration Evidence for Chile

177

tivity variable ( G 1 ) yields the long run relationship between capital goods imports, manufacturing exports, primary exports and total factor productivity. However, if the explanatory variables are not weakly exogenous, the estimates may be biased and inefficient and t-tests based on the model parameters may be misleading. Therefore, we check the robustness of the long-run estimates by means of the Dynamic OLS (DOLS) procedure developed by Saikkonen (1991). This procedure is asymptotically equivalent to Johansen’s (1995) maximum likelihood estimator and is known to perform well in small samples. Moreover, DOLS has been shown to provide unbiased and asymptotically efficient estimates, even in the presence of endogenous regressors (Stock and Watson 1993). The DOLS regression in our case is given by equation (9) below:

ln TFPt

c  D ln CMt  E ln IXt  G ln PXt

i k

i k

i k

i k

i k

i k

 ¦ )1' ln CMt i  ¦ )2' ln IXt i  ¦ )3' ln PXt i Ht

(9)

where D , E , and G are the long-run elasticities, and )1 , ) 2 , and ) 3 are coefficients of lead and lag differences of the I(1) variables, which are treated as nuisance parameters. They serve to adjust for possible endogeneity, autocorrelation, and non-normal residuals and result in consistent estimates of D , E , and G . IV.

Empirical results

A. Unit root test results Table 1 reports the results of testing the unit root null against the alternative of m structural breaks. At first, it can be seen that the estimated break points coincide with the beginning of the government of Allende in 1970, the drastic trade liberalization initiated by the military government of Pinochet in 1974, the 1975 economic crisis, and the 1982 recession. Furthermore, the test statistics show that we generally cannot reject the unit root hypothesis in favor of broken trend-stationary at the 1 per cent level. Since for the first differences the unit root hypothesis can be rejected, we can conclude that lnTFPt, lnCMt, lnIXt, and lnPXt are integrated of order one, I(1).

178

Dierk Herzer

Table 1: Kapetanios (2005) unit root test (to be continued) Series

Model

Levels lnTFPt

(5)

lnCMt

(5)

lnIXt

(5)

lnPXt

(5)

lnTFPt

(6)

lnCMt

(6)

lnIXt

(6)

lnPXt

(6)

lnTFPt

(7)

lnCMt

(7)

lnIXt

(7)

lnPXt

(7)

1st differences ǻ(lnTFPt) ǻ(lnCMt) ǻ(lnIXt) ǻ(lnPXt)

(5) (5) (5) (5)

m=1 Dummy Test statistic variable (Break year) DU75 (1974) DU71 (1970) DU74 (1973) DU71 (1970) DT75 (1974) DT82 (1981) DT81 (1980) DT74 (1973) DU75 DT75 (1974) DU82 DT82 (1981) DU74 DT74 (1973) DU71 DT71 (1970) i75 i71 i74 i71

Critical value 5% (1%)

-3.39

-4.930 (-5.338)

-2.75

-4.930 (-5.338)

-3.94

-4.930 (-5.338)

-3.87

-4.930 (-5.338)

-2.45

-4.495 (-5.014)

-3.95

-4.495 (-5.014)

-3.34

-4.495 (-5.014)

-4.88

-4.495 (-5.014)

-3.47

-5.081 (-5.704)

-4.06

-5.081 (-5.704)

-5.24

-5.081 (-5.704)

-5.32

-5.081 (-5.704)

-6.95 -6.31 -5.19 -8.39

-3.53 (-4.23) -3.53 (-4.23) -3.53 (-4.23) -3.53 (-4.23)

Trade, its Composition and Total Factor Productivity: Cointegration Evidence for Chile

179

Table 1 (continued): Kapetanios (2005) unit root test m=2 Series

Model

Levels lnTFPt

(5)

lnCMt

(5)

lnIXt

(5)

lnPXt

(5)

lnTFPt

(6)

lnCMt

(6)

lnIXt

(6)

lnPXt

(6)

lnTFPt

(7)

lnCMt

(7)

lnIXt

(7)

lnPXt

(7)

1st differences ǻ(lnTFPt) ǻ(lnCMt) ǻ(lnIXt)

ǻ(lnPXt)

Dummy variables Test statistic (Break years)

DU75, DU82 (1974, 1981) DU71, DU82 (1970, 1981) DU74, DU82 (1973, 1981) DU71, DU81 (1970, 1980) DT75, DT82 (1974, 1981) DT82, DT75 (1981, 1974) DT81, DT74 (1980, 1973) DT74, DT82 (1973, 1981) DU75, DU82 DT75, DT82 (1974, 1981) DU82, DU75 DT82, DT75 (1981, 1974) DU74 DU71 DT74 DT71 (1973, 1970) DU71, DU81 DT71, DT81 (1970, 1980)

-4.41 -3.70 -3.95 -4.01 -2.41 -3.76 -3.95 -4.92 -5.89

Critical value 5% (1%) -5.685 (-6.162) -5.685 (-6.162) -5.685 (-6.162) -5.685 (-6.162) -5.096 (-5.616) -5.096 (-5.616) -5.096 (-5.616) -5.096 (-5.616) -6.113 (-6.587)

-4.58

-6.113 (-6.587)

-6.20

-6.113 (-6.587)

-6.43

-6.113 (-6.587)

(5) (5) (5)

i75 D82 i71 D82 i74 D82

-6.87 -6.83 -4.89

-3.53 (-4.23) -3.53 (-4.23) -3.53 (-4.23)

(5)

i71 D81

-8.37

-3.53 (-4.23)

Note: The lag length was chosen using the Akaike information criterion. The dummy variables are specified as follows: i71, i74, i75, i81, i82 are impulse dummy variables with zeros everywhere except for a one in 1971, 1974, 1974, 1981, 1982. DU71, DU74, DU75, DU81, DU82 are 1 from 1971, 1974, 1975, 1981, 1982 onwards and 0 otherwise. DT71 (DT71, DT75, DT81, DT82) is 0 before 1971 (1974, 1975, 1981, 1982) and t otherwise. Critical values for the levels are provided by Kapetanios (2005). Critical values for the first differences are from MacKinnon (1991). For the first differences, only impulse dummy variables were included in the regression. Impulse dummy variables, that is, those with no longrun effect, do not affect the distribution of the MacKinnon Test statistics.

180

Dierk Herzer

However, using model (7) with one or two structural breaks, the unit root null can be rejected at the 5 per cent level for the log of manufactured and primary exports. Accordingly, lnIXt, and lnPXt are possibly trend-stationary with at least one structural break. This introduces a degree of uncertainty to the analysis. Therefore, in the next step, we use the bounds test approach in order to test for the existence of a long-run relationship between capital goods imports, manufacturing exports, primary exports and total factor productivity. As already mentioned, this approach is applicable irrespective of whether the explanatory variables are I(1) or I(0). B. Cointegration test results In order to conduct the cointegration test developed by Pesaran et al. (2001) we need to establish the lag order, k, of the unrestricted error correction model (8). The lag order is selected using the Akaike (AIC) information criterion, the Hannan-Quinn (HQ) criterion, and the Schwarz (SC) criterion with a maximum of three lags. Table 2 shows the results of the lag selection procedure. As can be seen, both the Akaike information and the Schwarz criterion chose a lag length of two, whereas the HQ selects a lag order of one. Table 2: Lag length selection K AIC HQ SC 0 -0.074 0.266 0.048 1 -6.162 -5.139 -5.795 -6.836 -5.130 -6.224 2 3 -6.704 -4.315 -5.847 Note: k is the lag order of the underlying VAR model for the conditional error correction model. Bold indicates lag order selection by the criterion.

Therefore, we estimate the ARDL model with one and two lags. Furthermore, an impulse dummy for 1971, i71, is included in equation (8) to make the residuals normally distributed. One possible reason for i71 to be important are the drastic reforms of the Allende government in that year (the nationalization of the mining, banking and agricultural sectors along with a expansionary fiscal policy). The results of the F-tests based on the unrestricted error correction model with k = 1 and k = 2 are reported in Table 3. Since the calculated F-statistics exceed the upper bound of the critical value band, we reject the null hypothesis of no long-run relationship between capital goods imports, manufacturing exports, primary exports and total factor productivity at the 1 per cent significance level. This result has a further implication: It justifies considering

Trade, its Composition and Total Factor Productivity: Cointegration Evidence for Chile

181

lnIXt, and lnPXt as I(1) variables, because the F-statistics are higher than the 1 per cent critical value, if all series are I(1). Therefore, in the following we treat all variables as I(1)-Variables. Table 3: F-statistics for testing the existence of a long-run relationship k

F-statistic

1 per cent critical value bounds of the F-statistic I(0) I(1) 1 16.35 4.29 5.61 2 5.89 4.29 5.61 Note: k denotes the lag length. The critical value bounds are from Table CI(iii) in Pesaran, Shin, and Smith (2001) p. 300, with 3 regressors.

C. Estimation of the Long-Run Coefficients: Error Correction Model Results We use error correction model (8) to estimate the coefficients of the long-run relation between capital goods imports, manufacturing exports, primary exports, and the level of total factor productivity. That is, we regress ¨lnTFPt on lnTFPt-1, lnCMt-1, lnIXt-1, and lnPXt1, the differences of lnCMt, lnIXt, and lnPXt up to lag order two, the lagged difference of lnTFPt-1 also up to lag order two, an intercept term, and the impulse dummy i71. The following equation results when applying Hendry's general-to-specific approach, where successively the least significant variables are eliminated until there remain only coefficients significant at the 1 per cent-level (t-statistics are given in parenthesis beneath the estimated coefficients): ' lnTFPt 0.691lnTFPt 1  0.120ln CMt 1  0.038ln IXt 1  0.087ln PXt 1 (8.676) (8.404) (4.854) (4.399)  0.134' ln CMt  0.041' ln CMt 1  0.051' ln IXt 1  1.754  0.120i71 (10.874)  eˆt

(3.245)

(4.646)

(8.629) (6.786) (10)

R 2 0.869 SE 0.016 JB 1.501 (0.472) Arch(1) 0.461(0.501) Arch(2) 0.208(0.813) Arch(4) 0.497(0.738) LM (1) 0.001(0.973) LM (2) 0.068(0.934) LM (4) 1.969(0.128)

The numbers in parentheses behind the values of the diagnostic test statistics are the corresponding p-values. These test statistics suggest that the model is well specified: The assumption of normally distributed residuals cannot be rejected (JB) and the Lagrange multiplier (LM) tests for autocorrelation based on 1, 2 and

182

Dierk Herzer

4 lags, respectively, do not indicate any problems concerning autocorrelated residuals. The model also passes the LM tests for autoregressive conditional heteroscedasticity (ARCH) of order p = 1, 2, 4. Moreover, in Figure 2 CUSUM and CUSUM of square-tests are presented, which overall support a stable relation for the period of interest. Accordingly, the model does a good job even in the Chilean ‘breakdown periods’ (e.g., 1975, 1982). Figure 2: Stability Test 16

1.6

12

1.2 8 4

0.8

0

0.4

-4 -8

0.0 -12 -16

-0.4 1975

1980

1985

1990

1995

CUSUMs (—) and 5% significance bounds (---)

2000

1975

1980

1985

1990

1995

2000

CUSUM of squares (—) and 5% significance bounds (---)

A significant negative coefficient of the lagged dependent level variable indicates cointegration. Therefore, this coefficient (which we denote here as G 1 ) can be used to test for cointegration. Accordingly the null to be tested is G 1 = 0. Critical values are provided by Ericsson and MacKinnon (2002). For 40 included observations and three regressors, the 1 per cent finite sample critical value is -4.31. Since the absolute value of the estimated t-statistic of Gˆ 1 (8.404) is higher than the absolute value of the 1 per cent critical value (4.31), we conclude that the null of no cointegration can be decisively rejected. Furthermore, we can reject the null hypothesis of weak exogeneity of lnTFPt, because Gˆ 1 is highly significant. Moreover, the t-statistics of lnCMt-1, lnIXt-1, and lnPXt-1 indicate that none of the variables can be excluded from the log-run relation. Normalizing on Gˆ 1 finally yields the equation

Trade, its Composition and Total Factor Productivity: Cointegration Evidence for Chile

183

ln TFPt 0.174 ln CM t  0.056 ln IX t  0.125ln PX t (11) From this equation, it can be inferred that the level of total factor productivity increases by 0.174 per cent in response to a one percent increase capital goods imports. With an increase of manufactured exports by one percent, total factor productivity increases by 0.056 percent. In contrast, a one percent increase in primary exports leads to a 0.125 per cent decrease of total factor productivity. Accordingly, capital goods imports are the main productivity determinant. However, due to endogeneity problems, regression (10) can be biased and the conclusions faulty. Therefore, in the final step of the analysis we check the robustness of the cointegration estimates. D. Re-estimation of the long-run elasticities: dynamic OLS results

We re-estimate D , E , and G by means of the DOLS procedure since DOLS generates unbiased and asympotically efficient estimates for variables that cointegrate, even with endogenous regressors. According to equation (9), we regress lnTFPt on lnCMt, lnIXt, and lnPXt, the leads and lags of the differences of lnCMt, lnIXt, and lnPXt up to order two, an intercept term, and the impulse dummy i71. The results of the DOLS procedure are presented in Table 4 (tstatistics are given in parenthesis beneath the estimated coefficients). Table 4: DOLS procedure results Dˆ Eˆ

0.167 (12.203)

0.066 (3.894)

R 2 0.984 SE 0.018 JB 0.489 (0.783) Arch (1) 0.657 (0.423) Arch (2) 0.503 (0.609) LM (1) 0.003 (0.954)

Gˆ -0.132 (-3.163) Arch (4) 0.554 (0.649)

LM (2) 0.795 (0.468) LM (4) 2.282 (0.112)

The results are very similar to those obtained by the Stock (1987) approach. Again, the diagnostic test statistics underneath Table 4 do not indicate any problems with autocorrelation, heteroscedasticity or nonnormality. All p-values exceed the usual (5 per cent) significance levels. As in the error correction model estimation, the effects of capital goods imports and manufactured exports on total factor productivity are significantly positive (see Dˆ and Eˆ ). The effect of an increase of primary exports on the level of total

184

Dierk Herzer

factor is again found to be strong and significantly negative (see Gˆ ). The magnitude of the coefficients in Table 4 does not differ substantially from equation (11). From this, we conclude that the coefficient estimates are fairly robust to different estimation techniques. Both the DOLS and the error correction model results can be interpreted as evidence of productivity-enhancing effects of capital goods imports and manufactured exports and of productivity-limiting effects of primary exports. V. Conclusions

This paper has used cointegration techniques to examine the impact of increasing capital goods imports, exports of primary products, and exports of manufactured goods on total factor productivity in Chile. The results suggest that there exists a long-run relationship between capital goods imports, manufactured exports, primary exports and total factor productivity. However, primary-product exports were found to have a statistically negative impact, whereas manufactured-product exports and capital goods imports have a statistically positive impact on total factor productivity. This is consistent with the results of Herzer et al. (2006), who, in their analysis of the export-led growth hypothesis for Chile, demonstrated a negative effect of primary exports and a positive effect of manufactured exports and capital goods imports on Chilean non-export GDP (i.e., GDP net of exports). These negative effects may be due to the problem of fluctuating commodity export prices and earnings, especially copper prices, which is well known in the Chilean literature. Romaguera and Contreras (1995), for example, found that copper price volatility had negative effects on the Chilean economic development. Additionally, manufactured exports might offer greater potential for knowledge spillovers and other externalities than primary exports. Admittedly, the main productivity effects come from capital goods imports, suggesting that knowledge and technology is embodied in machinery and equipment and thereby transferred through international trade. Based on these findings, we conclude that aggregate trade measures, such as the ratio of imports plus exports to GDP, may mask important differences between different trade categories. Even if there is evidence of trade-induced productivity gains at the aggregate level, this finding may not hold for certain trade categories, and spurious conclusions may be drawn when disaggregated trade is not examined.

Trade, its Composition and Total Factor Productivity: Cointegration Evidence for Chile

185

References

Agosin, Manuel (1999), Trade and growth in Chile, Cepal Review 68, 79 - 100. Alcalá, Francisco, Ciccone, Antonio (2004), Trade and productivity, The Quarterly Journal of Economics 119, 613-646. Chuang, Yih-Chyi (1998), Learning by doing, the technology gap, and growth, International Economic Review 39, 697-721. Chumacero, Rómulo A., Fuentes, Rodrigo J. (2005), On the determinants of Chilean economic growth, in: Rómulo A. Chumacero, Klaus Schmidt-Hebbel (eds.), General equilibrium models for the Chilean economy, Santiago, Chile, Central Bank of Chile, 163-187. Dawe, David (1996), A new look at the effects of export instability on investment and growth, World Development 24, 1905-1914. Ericsson, Neil R., Irons, John S., Tryon, Ralph W. (2001), Output and inflation in the long run, Journal of Applied Econometrics 16, 241-251. Ericsson, Neil R., MacKinnon, James G. (2002), Distributions of error correction tests for cointegration, Econometrics Journal 5, 285-318. Feder, Gerhson (1983), On exports and economic growth, Journal of Development Economics 12, 59-73. Frankel, Jeffrey, Romer, David (1999), Does trade cause growth?, American Economic Review 89, 379-399. Helpman, Elhanan, Krugman, Paul (1985), Market structure and foreign trade, Cambridge: MIT Press. Herzer, Dierk (2006), How does trade composition affect productivity? Evidence for Chile, Applied Economics Letters 13, 981986. Herzer, Dierk, Nowak-Lehmann, D. Felicitas, Siliverstovs, Boriss (2006), Export-led growth: assessing the role of export compo-

186

Dierk Herzer

sition in productivity growth, The Developing Economies 44, 306-328. Grossman, Gene M., Helpman, Elhanan (1991), Innovation and growth in the global economy, Cambridge, Mass: MIT Press. Johansen, Sören (1995) Likelihood-based inference in cointegrated vector autoregessive models, New York: Oxford University Press. Jonsson, Gunnar, Subramanian, Arvind (2001), Dynamic gains from trade: Evidence from South Africa, IMF Staff Papers 48, 197-224. Kapetanios, George (2005), Unit root testing against the alternative hypothesis of up to m structural breaks, Journal of Time Series Analysis 26, 123-133. Levine, Ross, Renelt, David (1992), A sensitivity analysis of crosscountry growth regressions, American Economic Review 82, 942-963. Lucas, Robert E. (1993), Making a miracle, Econometrica 61, 251272. Lütkepohl, Helmut, Wolters, Jürgen (1998), A money demand system for German M3, Empirical Economics 23, 371-386. MacKinnon, James G. (1991), Critical values for cointegration tests, in: Engle, Robert E., Granger, Clive W. J. (eds.), Long-run economic relationships: Readings in cointegration, New York: Oxford University Press, 266-276. Matsuyama, Kiminori (1992), Agricultural productivity, comparative advantage, and economic growth, Journal of Economic Theory 58, 317-34. Miller, Stephen M., Upadhyay, Mukti P. (2000), The effects of openness, trade orientation, and human capital on total factor productivity, Journal of Development Economics 63, 399-423. Pesaran, M. Hashem, Shin, Yongcheol, Smith, Richard J. (2001), Bounds testing approaches to the analysis of level relationships, Journal of Applied Econometrics 16, 289-326.

Trade, its Composition and Total Factor Productivity: Cointegration Evidence for Chile

187

Rivera-Batiz, Luis A., Romer, Paul (1991), Economic integration and endogenous growth, Quarterly Journal of Economics 106, 531-550. Romaguera, Pilar, Contreras, Dante (1995), Impacto Macroeconomico de la Inestabilidad del Precio del Cobre en la Economía Chilena, in: Pascó-Font, Alberto (ed.), La administración de los ingresos por exportaciones mineras en Bolivia, Chile y Perú, GRADE/Lima, Perú, 171–196. Sachs, Jeffrey D., Warner, Andrew M. (1995), Natural resource abundance and economic growth, National Bureau of Economic Research Working Paper No. 5398. Saikkonen, Pentti (1991), Asymptotically efficient estimation of cointegrating regressions, Econometric Theory 7, 1-21. Stock, James H. (1987), Asymptotic properties of least squares estimators of cointegrating vectors, Econometrica 55, 10351056. Stock, James H., Watson, Mark W. (1993), Simple estimator of cointegrating vectors in higher order integrated systems, Econometrica 61, 783-820. Thangavelu, Shandre Mugan, Rajaguru, Gulasekaran (2004), Is there an export or import-led productivity growth in rapidly developing Asian countries? A multivariate VAR analysis, Applied Economics 36, 1083-1093.

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer and Inmaculada Martínez-Zarzoso

Chile’s Market Share in the EU Market: The Role of Price Competition in a Panel Analysis Setting Abstract The objective of this paper is to analyze how Chile’s share in the EU market developed between 1988 and 2002, using panel data to test for the impact of price competitiveness on market share. Price competitiveness is considered a decisive determinant of Chile’s market share given the relative homogeneity of the country’s most successful export products. The impact of price competitiveness on market shares is estimated with Dynamic Ordinary Least Squares (DOLS) in a semi-static model. The Three Stage Feasible Generalized Least Squares (3SFGLS) and the Generalized Method of Moments (GMM) technique are applied to a dynamic model (ARDL). I. Introduction On 3 October 2002, Chile signed a far-reaching FTA that will vastly improve its access to EU markets. Once fully implemented, the FTA will promote the interests of both Chile and the EU, providing important and wide-ranging benefits to both.1 With respect to trade, the EU expects a major expansion of its manufactured exports to the Chilean market, whereas Chile hopes to expand its agricultural and light manufactured exports to the EU. From Chile’s point of view, the agreement offers a means of maintaining and potentially strengthening its competitive position 1

In addition to facilitating trade through reduction and elimination of tariffs and introducing modern customs practices, the FTA promotes economic cooperation and technological innovation, environmental and resource protection, and support for governmental reform (European Commission, 2005).

190

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

in the EU market. In the short run, the FTA will reduce or even eliminate trade barriers, and the effect this will have on relative prices will improve Chile’s competitive position both within the EU and also with respect to third countries that do not have FTAs with the EU. In the medium to long run, however, the positive effects of the FTA will be eroded if the EU concludes FTAs with other countries such as the full members of MERCOSUR and some Asian countries. Given that Chile’s main export commodities—copper, fish, fruits, paper and pulp, and wine—are heavily natural-resourcebased, it already has numerous competitors.2 Some of them, including Norway, Russia, Indonesia, Malaysia, the Philippines and Thailand are also major exporters of timber and rubber. Furthermore, the Southeast Asian countries have succeeded in dramatically increasing their light manufactured exports to industrial countries in the last decade. In the Southern Hemisphere, South Africa, Australia, and New Zealand threaten Chile’s position as a successful fruit and wine exporter. Chile also faces stiff competition from the EU in agricultural products, with the UK, Ireland and Norway as Chile’s main competitors in fish exports worldwide. In addition, China’s low labor costs have enabled it to become a major exporter of machinery and equipment, textiles and clothing, footwear, toys and sporting goods, and mineral fuels, thus overtaking Latin America’s traditional dominance in textile, clothing and shoe exports. Based on 2003 data, the EU is Chile’s top trading partner worldwide: 25 per cent of Chile’s exports go to the EU and 19 per cent of its imports come from the EU. During the first half of 2003, mining (predominantly copper) still represented 46 per cent of total Chilean exports, while agriculture, farming, forestry and fishing products represented 13.02 per cent. Trade with Chile represents 0.45 per cent of total EU trade, placing Chile 41st in the ranking of the EU’s main trading partners. Between 1980 and 2002, EU imports from Chile increased from € 1.5 billion to € 4.8 billion, while EU exports to Chile increased from € 0.7 billion to € 3.1 billion (European Commission, 2005).

2

Nevertheless, Chile can still be considered the most competitive and the least corrupt economy in Latin America.

Chiles Market Share in the EU Market

191

The purpose of this paper is to analyze Chile’s share in the EU market on a sectoral level and to evaluate its relative competitiveness on the EU market in the period of 1988 to 2002 using panel time series techniques. Economic reasoning would suggest that market shares are determined by the relative prices of Chile and its main competitors in the EU countries, and by an unobserved variable—for example, strategic behavior. Price competitiveness is considered a decisive determinant of Chile’s market shares given the relative homogeneity of Chile’s most successful export products: fish, fruit, beverages, ores, copper, and wood. We conduct our empirical analysis of Chile’s market shares in two distinct approaches. In the first approach, we use panel unit root tests and panel cointegration tests. If cointegration of the series results, then we set up a Panel Dynamic OLS Model (DOLS) to deal with the problem of non-stationarity of the series and the endogenity problem. This step builds on the groundbreaking studies of panel unit root and panel cointegration techniques by Breitung and Pesaran (2005), Dreger and Reimers (2005), Westerlund (2005a, 2005b), Pedroni (2004), Pedroni (1999) and Banerjee (1999a, 1999b). In order to deal with cross-sectional correlation of disturbances, we then use the Seemingly Unrelated Regression (SUR) technique. A second method of analyzing market shares utilizes a dynamic model—the partial adjustment model— estimated both by Three Stage Least Squares (3SLS) and by the Generalized Method of Moments (GMM), in combination with Feasible Generalized Least Squares (FGLS). This enables us to avoid the problems of both endogenity and autocorrelation of residuals across cross-sections and over time. The study is set up as follows. Section II gives an overview of Chilean market shares in the EU market and develops a very simple model to explain sectoral market shares. Section III contains some general remarks on the panel unit root tests, panel cointegration tests, DOLS modeling and FGLS in a 3SLS and a GMM framework. In Section IV we present and discuss the results. Section V concludes with a more general comparison of results and approaches. II. Chile’s Market Share in the EU Market A. The Development of Chile’s Market Shares over Time In Table 1 we list Chile’s largest export sectors, its export shares, and its share in the EU market. In this table, the EU market is con-

192

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

sidered as one market. However, in the empirical analysis, we investigate Chile’s sectoral market shares in specific EU countries. Table 1: Chile’s seven most important export sectors and their competitive position HS code*

Sector

annual share in potential extra-EU percentage total competitor*** change of exports exports 2002 (‘88-‘02)

average market share in the EU** (‘88- ‘02)

Fish and crustaceans, molluscs Edible fruit and nuts Beverages, spirits, and vinegar Ores, slag, and ash

7.2.0 %

5.2 %

Norway

7.5.0 %

10.0 %

44.6 %

7.8 %

Australia, South Africa, 2.62 % New Zealand South Africa, Australia 0.77 %

11.9 %

9.1 %

44

Wood and wood products

12.4 %

47

Wood pulp

13.9 %

74

Copper and copper products

5.4 %

03

08 22

26

Brazil, China

1.22 %

Australia,

3.75 %

0.26 % 1.5.0 % Norway, Russia, Canada, Malaysia, Indonesia 6.6 % Norway, Canada, 2.89 % Russia 37.0 % South Africa, Canada 10.34%

Source: EUROSTAT (2003); COMEXT CD ROM, ‘Intra- and Extra-EU Trade, Annual data, Combined Nomenclature’, European Commission; authors’ calculations. HS is the Harmonized System of Classification of Trade. **Share of EU imports from Chile in total EU imports (both from other EU-countries and non-EU countries). ***According to TradeCAN (Database Software for a Competitiveness Analysis of Nations, The World Bank, 2002).

All seven sectors experienced remarkable export growth, beverages being the most dynamic sector. It should be noted, however, that in 1988 ‘beverages’ started from a lower level than the more traditional sectors such as fruit, wood, wood pulp, and copper. Copper held the biggest market share in EU imports with 10.34 per cent, followed by ores (3.75 per cent), pulp of wood (2.89 per cent) and fruit (2.62 per cent) in the period 1988 to 2002. Figure 1 depicts Chile’s market share position with respect to the EU countries (sheu), with respect to non-EU countries (shnoneu) or with respect to the world (shw), which comprises all EU and all non-EU countries.

Chiles Market Share in the EU Market

193

Figure 1: Chile’s market share in the EU with respect to EU and non-EU competitors in the period 1988 to 2002, in per cent (to be continued) fish imports

4 .0 3.6 3.2

2.8

2.4 2.0

1.6

1.2 0.8 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 SHEU03

SHNONEU03

SHW03

fruit imports 9 8 7 6 5 4 3 2

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

SHEU08

SHNONEU08

SHW08

194

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

Figure 1 (cont.): Chile’s market share in the EU with respect to EU and non-EU competitors in the period 1988 to 2002, in per cent beverage imports 12 10 8 6 4 2 0 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

SHNONEU22

SHW22

ores, slag and ash 1.6 1.4 1.2 1.0 0.8

0.6 0.4 0.2 0.0

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 SHEU44

SHNONEU44

SHW44

imports of wood 8 7

6 5

4 3

2

% 1 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 SHNONEU47

SHW47

Chiles Market Share in the EU Market

195

Figure 1 (cont.): Chile’s market share in the EU with respect to EU and non-E U competitors in the period 1988 to 2002, in per cent imports of pulp of wood 35

30 25

20 15

10 5 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 SHNONEU74

SHW74

imports of copper 8 7 6 5 4 3 2

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

SHW26

SHNONEU26

Source: Own calculations based on Eurostat, Comext.

According to Figure 1, Chile lost market shares in the fish sector with respect to both EU and non-EU countries between 1991 and 1996. It caught up after 1996, reaching again 1988 levels in 2002. Overall, competition in the 1988-2002 period was very stiff. Main competitors in fishery products were within Europe—the UK, Ireland, Portugal, Spain, and Italy—and Norway in the case of salmon. The competition for market shares was also very fierce in the fruit sector, from both within the EU and outside (Australia, New

196

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

Zealand, South Africa). While Chile increased its market shares in the late 1980s and early 1990s, Chilean fruit exporters had severe problems defending their market shares thereafter, and since 1993, Chile has clearly lost its competitive position with respect to the EU countries. A pre-test conducted to determine whether this was due to the appreciation of the real effective exchange rate found that exports reacted severely to the loss of price competitiveness. In the beverages sector, we can observe a steady increase in Chile’s market shares in relation to both non-EU and EU countries. For Chile, the most important export item in the beverages sector is wine. In a pre-test examining non-EU competition, Australia did not turn out to be a threat to Chile, but South Africa did. However, since 1994, Chile has clearly been gaining competitiveness against the non-EU countries in this product group. The ores sector reveals a number of strong fluctuations. When looking at the endpoints, Chile succeeded in improving its market share in this product group from 1988 to 2002. Competition with the EU countries was particularly intense in the 1988-91 period. Given that Brazil and Australia are the world’s main exporters of ores, we pre-tested the role of these non-EU competitors. However, their price competitiveness turned out to be irrelevant for Chilean export success. This could be due to the fact that Chile produces different qualities of ores and in different subsegments than Australia and Brazil. Chile faced strong competition in the wood sector from the EU (Sweden, Finland) and even lost market shares in the 1988-1996 period. Competition from non-EU countries such as Norway, Russia, and Canada was characterized by major up and downswings. Regarding its competitive position in the pulp of wood sector, Chile increased its overall market share, especially with respect to non-EU countries. With copper, Chile succeeded quite well in defending its competitive position in the EU market in the period 1988 to 2002. Chile is the world’s largest copper-producing country, followed by the United States, which is both a producer and a net importer of copper. Success in the copper industry depends on keeping production costs low compared to market prices. Major production costs include labor costs, energy costs, and environmental regulations, which play a greater role in industrialized countries. To sum up, the development of Chile’s market shares has been subject to strong fluctuations in most of its export sectors. Defend-

Chiles Market Share in the EU Market

197

ing its market share has been no easy task for Chile, except in the sectors ‘beverages’, ‘wood pulp’ and ‘copper’. B. Development and Determinants of Market Shares Following Sutton (2004), there are two contrasting views on the development of market shares over time: the first goes back to Alfred Chandler and asserts that market shares are robust over time and that market leadership tends to persist for a long period of time. The second view, propagated by Schumpeter, emphasizes the transience of leadership positions. Schumpeter labels those positions ‘temporary monopolies’ that are created by invention and innovation. However, there is no benchmark for long or short-term leadership positions (2002 Japan Conference, 2005). We will test the relevance of these hypotheses by means of panel unit root tests. If market shares turn out to be stationary, I (0), we will conclude that they are robust and persistent during the period 1988 to 2002. If they turn out to be non-stationary, we will conclude that the Schumpeter hypothesis cannot be rejected by the 1988-2002 data. There are also two approaches to modeling market shares: one which views market shares as basically stochastic, and the other which views them as influenced by hard economic factors such as prices, marketing expenditures, number and strength of competitors, etc. When modeling market shares, Sutton (2004) chooses an eclectic approach. He starts from the idea of building a stochastic model,3 enriching this model by including industry-specific features (e.g., a strategic representation of firms’ competitive responses to market share changes). However, he has to concede that strategic behavior is often intrinsically unobservable. In contrast to Sutton, we emphasize the stochastic nature of market shares less, stressing instead the role of sectoral real effective exchange rates, which can be treated as an industry-specific feature. We believe that exchange rates, cost differentials, tariffs, and subsidies are important ‘hard’ factors explaining market shares over time. Thus we consider price competitiveness as decisive for a country’s competitive position. Strategic behavior being difficult to model, we restrict our model by 3

It is obvious that equations (1) and (2) below only hold if market shares are determined mainly by observable economic fundamentals, e.g., the real effective exchange rates. They do not apply if market share dynamics are represented purely by a stochastic model.

198

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

allowing strategic behavior and sector-specific characteristics to be incorporated in the residuals of equations (1) and (2) below. Market shares in a specific sector (or sectors) are computed as the ratio of Chile’s sectoral exports (X in the numerator) and EU country i’s imports from the world M.i = MEU+Mnon-EU (in the denominator). Due to missing data, we consider only Chile’s market shares in France (FRA), the Netherlands (NDL), Germany (DEU), Italy (ITA), UK (GBR), and Spain (ESP). Market shares are computed for seven sectors based on the two-digit HS (Harmonized System) classification: fish (03), fruit (08), beverages (22), ores (26), wood (44), wood pulp (47) and copper (74).4 The period covered is 1988 to 2002. Thus, we obtain a maximum of 6 crosssections and 15 years, resulting in a maximum of 90 observations per sector. The number of observations varies depending on the sector. A log-log specification was chosen for Chile’s market share in the EU market. The market share of the country under investigation (Chile) in country i in sector s at time t is modelled as:

lshwist

D is  Ei lreerist  J i lreerist *  Pist

(1)

where i = 1, 2,…, 6; represents the cross-sections: FRA, NDL, DEU, ITA, GBR and ESP; t = 1988, 1989, …, 2002 are years (annual observations); and s = 03, 08, 22, 26, 44, 47 and 74 are the sectors (according to the two-digit HS classification). lshwist stands for Chile’s market share in EU country i in sector s at time t. lreerist is Chile’s real effective exchange rate, prevailing in country i and in sector s and lreerist * is Chile’s competitor’s (*) real effective exchange rate, prevailing in country i and in sector s. Equation (1) will be applied in Section IV.A. According to Cable (1997), the market shares can be best modeled by means of an autoregressive distributed lag model (ARDL) with lag length k.5 Cable uses a geometric lag model as 4 5

Sources of the data are outlined in the Appendix. There are two types of autoregressive distributed lag models: the geometric lag model and the transfer function model, also known as the ARMAX model (for an application, see Nowak-Lehmann D., 2004 and Greene, 2000).

Chiles Market Share in the EU Market

199

seen in Equation (2) to model the reaction of market shares in the short and in the long run.6 In this model, changes in the real effective exchange rate in the more distant past have a smaller impact on changes in market shares than exchange rate changes from the more recent past. k 0 k lshwist Dis  E0O0lreer ist  ... E0O lreer istk  J0O lreer ist * ... J0O lreer istk * Pjst (2) This model is often utilized in its Koyck lag transformation, which allows to save degrees of freedom and which is linear in its parameters (Greene, 2000): (2’) lshwist D is  E i lreerist  J i lreerist * Olshwist 1  P ist The Koyck lag transformation, called ARDL (Equation (2’), of the geometric lag model (Equation (2)) will be applied in Section IV.B. III.

Estimation Techniques for Non-Stationary Panel Data Controlling for Endogeneity

A. Unit Root Based Techniques Before turning to the econometric analysis, the time series properties of the data (all in natural logs) were tested. All series, i.e., market shares (lshw), Chile’s real effective exchange rate (lreer) and Chile’s competitors’ real effective exchange rates (lreer*) for all country pairs were subject to tests of non-stationarity (panel unit root tests) in a first step. This procedure had to be applied to all seven sectors under investigation. The possible existence of structural breaks in the series was disregarded for two reasons: first, consideration of structural breaks would further complicate the econometric analysis from a technical perspective at this point in time (Stock, 1994).7 Second, neither fundamental, abrupt changes in economic policy nor tremendous exogenous shocks were detected in the period 1988-2002. Under Aylwin, Frei and Lagos, the government continued the economic policy of Pinochet. Major shocks (the Tequila crisis of 1994, the spillover effects of the Asian crises of 1998, and the collapse of the currency board in Argentina in 2001/2002) seem to have been adequately reflected in the market

6 7

Geometric lag models are also known as partial adjustment models. Unit root tests considering structural breaks are discussed and applied extensively by Herzer and Nowak-Lehmann D. (2006).

200

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

share and real effective exchange rate variables in the period of 1988-2002. In the statistical analysis we allowed for different unit root processes in the panel, i.e., individual, cross-section-specific (countryspecific) unit roots. We applied the Im, Pesaran, and Shin (2003) panel unit root test on all series, thus considering the possibility of individual unit roots of our panel data. All variables (lshw, lreer, and lreer*) were non-stationary, integrated of order one, I(1), with a p-value of 0.00 (exception: lrpcopper with p= 0.02). As to market shares, this finding supports Schumpeter’s view on market shares further. According to Schumpeter, gains in market shares are of temporary value. Monopolistic positions have to be defended, otherwise they are quickly lost. This view seems to apply especially to the fish, fruit, beverages, ores, and copper sectors. In the wood sectors (44 and 47), market shares appeared more stable, but were still non-stationary according to the tests. Table 2 presents the results. Table 2: Results from the Im, Pesaran, and Shin (2003) panel unit root test stating t-bar values (to be continued) IPS Panel Unit Root Test Based on Individual Unit Roots H0: Series has a unit root (series is non-stationary)† Sector 03 Fish and crustaceans, mollusks lshw03 lreer03 lreer03*=lreer03nor level series -1.81 -1.58 -1.94 -4.36 -3.42 -3.47 ' series Sector 08 Edible fruit and nuts lshw08 lreer08 lreer08*=lreer08aus level series -1.68 -1.58 -2.53 -5.90 -3.42 -4.11 ' series Sector 22 Beverages, spirits, and vinegar lshw22 lreer22 lreer22*=lreer08saf level series -1.62 -1.58 -0.92 -4.25 -3.42 -3.34 ' series Sector 26 Ores, slag, and ash lshw26 lreer26 lreer26*=lreer26bra level series -1.29 -1.58 -2.26 -4.18 -3.42 -7.43 ' series Sector 44 Wood and wood products lshw44 lreer44 lreer44*=lreer44nor level series -1.83 -1.58 -1.94 -2.80 -3.42 -3.47 ' series

Chiles Market Share in the EU Market

201

Table 2 (cont.): Results from the Im, Pesaran, and Shin (2003) panel unit root test stating t-bar values Sector 47 level series ' series Sector 74 level series ' Series

Wood pulp lshw47 lreer47 -1.68 -1.58 -2.93 -3.42 Copper and copper products lshw74 lrpcopper†† -1.34 -1.58 -4.22 -3.42

lreer47*=lreer47nor -1.94 -3.47

----------

Note: †The critical value is -2.48 for D =1%. A trend and an intercept are included in the test equation whenever suggested by the series’ graphs. †† lrpcopper serves as an indicator of Chile’s real copper production costs. It is used instead of lreer in the market share analysis.

Given that the variables lshw, lreer and lreer* were all I(1), panel cointegration tests were performed, relying on a residualbased cointegration test.8 The idea of the residual-based cointegration test goes back to Engle and Granger (1987), who applied it to time series. As to regressions with time series, if the residual (ut) of a regression is built around variables with the same order p of integration (i.e., the variables ~ I(p)) are stationary, i.e. ut ~ I(0)), it is said that the I(p) variables are cointegrated, and therefore that a long-run relationship does exist. However, these tests not only tend to suffer from unacceptably low power when applied to series of only moderate length, but also have to use special critical values (e.g., Kapetanios’ critical values)9 if stationarity of the residuals is to be tested (Kapetanios, 1999). Pooling data across individual members of a panel when testing for cointegration is therefore advantageous. Pooling increases the power of the unit root test by making available considerable more information regarding the cointegration hypothesis10 (Pedroni,

8 9

10

See Breitung’s and Pesaran’s (2005) overview. MacKinnon’s critical values cannot be used when testing the nonstationarity of residuals. In this case adjustments for the number of regressors in the regression equation are necessary and different critical values result. H0: The variables of interest are not cointegrated for each member of the panel and H1: For each member of the panel, there exists a single

202

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

1999). But testing for cointegration in a panel setting is also more complicated since two types of cointegration can be present and must be taken into account: first, between series over time (the type prevailing in time series) and second, between cross-sections (the type potentially existing in a panel setting) (Breitung and Pesaran, 2005). We controlled for the second type of cointegration by building a system of equations around Equation (1). Then we applied the Seemingly Unrelated Regression (SUR) estimation method that took cross-sectional correlation of the residuals into account by weighting the matrix with the regressors (X’X). As in time series analysis, standard unit root tests on the residuals,11 which use inadequate test statistics (MacKinnon, 1991), cannot be utilized. First of all they do not account for the number of regressors in Equation (1). Second, they have not been adjusted for heterogeneous intercepts and heterogeneous deterministic trends, and are therefore too rough (Pedroni, 1999). Pedroni’s (1999) cointegration test statistic solves those problems. Following Pedroni’s panel cointegration test (1999), we allowed for a maximum of heterogeneity between countries and flexibility by formulating Equation (1) with cross-section-specific intercepts ( D i ) and cross-section-specific coefficients ( E i and J i ). Thus we are able to take country-specific cointegration vectors into account. Finally, we derived the residuals from this system, obtaining ui03t, ui08t, ui22t, ui26t, ui44t, ui47t and ui74t for the seven sectors under investigation. We applied and programmed Pedroni’s (1999) formulas for a residual-based panel unit-root test and computed the test statistics, which follow a standard normal distribution. Pedroni’s test revealed that the residuals of all sectors were station-

11

cointegrating vector, although this cointegrating vector need not be the same for each member (Pedroni, 1999). Out of curiosity we performed ‘invalid’ unit-root tests (pre-tests of non-stationarity assuming individual unit root processes) on the residuals of Equation (1) by utilizing both the ADF-Fisher Chi-square test and the PP-Fisher-Chi-square test. Both tests rejected the null hypothesis of individual unit root processes with p-values of 0.00 for all seven sectors. These pre-tests showed that the residuals were stationary and hinted at cointegration.

Chiles Market Share in the EU Market

203

ary and the variables lshw, lreer and lreer* were cointegrated (pvalue 0.00) and therefore in long-run equilibrium.12 Given that cointegration exists, the regression coefficients can be estimated by different methods. First, the regression coefficients can be estimated by the Johansen method, which is based on a Vector Error Correction Model (VECM). It applies Maximum Likelihood (ML) estimation and yields consistent estimates. By having only lagged first differences and the EC term on the righthand side of the VECM, this approach is also able to deal with endogenous variables (Johansen, 1988). Second, regression coefficients can be estimated in the error-correction (ECM) framework developed by Stock (1987), who utilizes Non-Linear Least Squares. If, however, regressors are endogenous, the estimates will be biased. The use of instrumental variables could solve this problem. Third, the long-run regression coefficients can be estimated with the Dynamic Ordinary Least Squares (DOLS) approach that was proposed by Stock and Watson (1993, 2003). This approach takes endogenity of the regressors into account and therefore yields consistent estimates. We follow this most recent approach for estimating the long-run regression coefficients. Before doing so, however, we set up panel error correction models (ECM) of the Stock type for all sectors.13 This procedure allows for another check of cointegration. We obtained coefficients belonging to the error correction term (EC term) that carried the correct (negative) sign and were significant at a p-value of 0.00 for all seven sectors. A significant and negative sign indicates the existence of a cointegrating relationship as we know from time series analysis (Banerjee, Dolado and Mestre, 1998; Ericsson and MacKinnon, 2002). We did not utilize the ECM estimates for further analysis due to correlation between the autocorrelated disturbances and the lagged endogenous variable, which would cause biased estimates, but apply DOLS instead. The DOLS approach led to Equation (1’): lshwit = a + b lreerit + c lreer*it + ¦kk 11 E k 'lreerit k + ¦kk

12 13

1 1J k 'lreer *it k

+ uit

The program and the results are available upon request. An application can also be found in (Hendershott et al. (2002).

(1’)

204

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

b and c represent the long-run coefficients and E k and J k represent adjustments of lshwit with respect to past, present, and future values of the change in lreer and lreer*. Corrections for autocorrelation were made whenever necessary. According to Stock and Watson (2003) statistical inferences about the parameters in Equation (1’) based on autocorrelation-consistent standard errors are valid. Furthermore, Equation (1’) was estimated with SUR, thus controlling for cross-sectional correlation of the disturbances. When utilizing DOLS, statements on the short- and medium-run relationship between the dependent variable lshw and the independent variables lreer and lreer* are not possible; only the long-run relationship can be identified. This failure can be adequately addressed when using a distributed lag model (ARDL, see Section III.B). Results obtained by means of an ARDL will be presented in Sections IV.B and IV.C. B. Feasible Generalized Least Square (FGLS) Approaches The cointegration approach is not the only one that can deal with non-stationary series and yields efficient, unbiased estimates. As is known from time series analysis, FGLS offers another possibility. FGLS can also be applied to panel data and works very well in dynamic models. In the present study, we take advantage of these features. We also stress the time series properties more than is usual in the dynamic panel analysis literature (Baltagi, 2005), which tends to abstract from autocorrelation of the disturbances in order to elaborate the characteristics of one-way error component models in which cross-section-specific random effects are present. We take a different route for several reasons: first, we choose a fixed effects model since our cross-sections were not drawn randomly but selected intentionally. If cross-section-specific disturbances Pi ~ IID (0; V P2 ) prove to exist, we think that the cross-section specification should be improved. Second, we try to account for time series properties because our time dimension exceeds our cross-section dimension and therefore time series problems should be given more weight. These considerations lead us to use an alternative method of dealing with non-stationary series in a panel regression framework, namely, to FGLS estimation. FGLS works similarly in a panel analysis setting to a time series setting. The idea remains the same:

Chiles Market Share in the EU Market

205

non-stationarity of the series in a regression equation is reflected in the autocorrelation U of the residuals over time14

u it =

¦

K k 1

U ik u it  k + eit

(3)

with eit ~ N(0; V ei ) and k = 1, 2,…K number of lags. Furthermore, FGLS has the tremendous advantage of working well in dynamic regression models such as autoregressive distributed lag models (ARDL models)—in our case, the Koyck lag model in Equation (2’). ARDL models are able to describe the reaction of the dependent variable of a regression very precisely over time (in the short, medium and long run), whereas Equation (1’) is basically a semi-static model. The FGLS method works as follows: first, the residuals of Equation (2’) are computed by means of SUR. Second, the order (first order, second order, or p-order) of autocorrelation is tested in Equation (3). Here we found first-order autocorrelation of the type uit = U i1 uit-1 + eit to be present and dominant. Third, the variables of Equation (2’) are transformed into lshwzit = lshwit - U i1 lshwit-1, lreerzit = lreerit- U i1 lreerit-1, lreerzit* = lreerit*- U i1 lreerit-1* and eit = uit- U i1 uit-1 thus generating variables in soft or quasi-first differences. Equation (2’) can then be estimated on basis of the transformed variables applying the Cochrane-Orcutt method (Stock and Watson, 2003). The endogenity problem of the lagged dependent variable (lshwit-1), which is caused by first-order correlation of the residuals, requires the use of either the Three-Stage Least Squares or the GMM (Generalized Method of Moments) technique. Modern computer programs (e.g., EViews 5.1) allow one to generate the variables in soft first differences directly in Equation (2’) by adding e.g., an AR(1) term for first-order autocorrelation and to simultaneously apply methods to control for the endogeneity of the regressors (see Sections IV.B and IV.C). IV.

Empirical Analysis of Market Shares

In the econometric part of this study we used EUROSTAT’s trade database COMEXT (Intra- and Extra-EU Trade, Supplement 2, 14

It is usually well below 1 so that first differencing is a very rough method to get rid of stationarity.

206

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

2003). The analysis had to be restricted to six EU countries: France, Germany, Italy, the Netherlands, Portugal, Spain and the UK. Incompleteness of the data led to the exclusion of nine EU-15 countries and all ten EU-1015 countries from the analysis. Data and computation of the variables are described in Appendix 1. In the following sections a fixed effects model was estimated, allowing for cross-section-specific intercepts. This model could still be enriched further by estimating cross-section-specific slope parameters for lreer and lreer*. However, since our focus at this stage is on comparing estimation techniques (DOLS, ARDL estimated by 3SLS, ARDL estimated by GMM), we capture country-specific effects only through cross-section-specific intercepts and try to save degrees of freedom by modeling common slope parameters. A. Estimating the Impact of Price Competition on Market Shares Using the Cointegration Approach Table 3 presents the results for the market share model (Equation (1’)) estimated by means of DOLS controlling for inter-temporal (inserting an AR(1) term) and cross-sectional correlation (estimating the DOLS by SUR). Sector results are shown in lines. An increase in price competition by Chilean exporters has the expected positive impact on Chile’s market shares in the fruit (08), ores (26), and wood (44) sectors. Increasing foreign price competition has the expected negative impact in the wood (44) and wood pulp (47) sectors. Rising Chilean real copper prices are bad for Chile’s market share, as expected. Interestingly, we get significant (but not the expected) signs for the beverages sector, which is dominated by wine exports. The opposite signs make economic sense if low prices are interpreted by consumers as an indicator of low quality (and vice versa). Therefore, we consider this result plausible and in line with economic expectations. This result is confirmed by the techniques utilized in Sections IV.B and IV.C.

15

The E-10 countries have not yet been integrated into the COMEXT trade statistics thus impeding their analysis.

Chiles Market Share in the EU Market

207

Table 3: Results for the market share model estimated by DOLS regression coefficients1 Equation (1’) results Sector

long-run impact lreer b

03

-0.63 (0.34) 1.75*** (0.00) -4.39*** (0.00) 2.69*** (0.00) 0.99*** (0.00) -1.51*** (0.00)

08 22 26

44 47 74

lrpcopper

goodness of fit measures

AR-term AR(1)

R2 adjusted

S.E. of regression

Durbin Watson stat.

-0.91 (0.25) -0.62 (0.38) 4.90*** (0.00) 0.29*** (0.00)

0.43*** (0.00) 0.60*** (0.00) 0.60*** (0.00) 0.64*** (0.00)

0.97

1.03

2.12

0.99

1.07

2.26

0.99

1.06

2.37

0.95

1.07

2.12

-5.74*** (0.00) -0.59*** (0.37) --------

---------

0.98

1.03

2.05

---------

0.97

1.09

1.98

0.74*** (0.00)

0.99

1.06

2.21

long-run impact lreer* c

-2.09*** (0.00)

Note: 1 p-values in parentheses

B. Estimating the Impact of Price Competition on Market Shares Applying the FGLS-Approach (Soft First Differences Approach) in a Dynamic Model In the dynamic model a new problem arises: when a lagged endogenous variable appears on the right-hand side of a regression equation (as in the geometric lag model in Equation (2)) and when the disturbances are autocorrelated (a phenomenon that goes hand in hand with non-stationary series), the lagged endogenous variable is automatically correlated with the disturbance term and thus becomes endogenous. Endogenity and cross-sectional correlation of the disturbances are controlled by instrumental variables in the framework of the Three-Stage Least Squares (3SLS) technique, which is the SUR version of Two-Stage Least Squares (see EViews 5: User’s Guide, 2004, p. 700) and autocorrelation is controlled by means of an AR term. In Table 4, the impact of price competitiveness on market shares in a dynamic model (ARDL model) is summarized. The long-run coefficients are obtained by dividing the short-run coefficients ( E , J ) by 1- O if 0< O <1 (Greene, 2000).

208

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

Table 4: Results for the dynamic market share model estimated by panel-3SLS regression coefficients1 Equation (2’)

goodness-of-fit measures2

lshwist D is  E i lreerist  J i lreerist * Olshwist 1  P ist sectorresults

impact impact adjustm. AR-term R2 S.E. Durbin lreer* coeff. adjusted regres- Watson lreer J sion stat. O E

03 short run 03 long run 08 short run 08 long run 22 short run 22 long run 26 short run 26 long run 44 short run 44 long run 47 short run 47 long run 74 short run 74 long run

0.82** -0.72 -0.19 0.68*** (0.02) (0.19) (0.20) (0.00) ---------- ---------- ----------- ---------

0.97

1.02

2.15

0.97

1.02

2.15

1.82** -0.14 -0.07 0.69*** (0.02) (0.85) (0.70) (0.00) ---------- ---------- ----------- ---------

0.99

1.05

1.99

0.99

1.05

1.99

-2.09*** 2.01*** 0.62*** (0.01) (0.01) (0.00) -6.96*** 6.04*** --------

-0.08 (0.64) --------

0.98

1.05

2.04

0.98

1.05

2.04

1.83*** 0.06 (0.00) (0.42) 6.10*** 0.20

0.70*** (0.00) ---------

-0.29* (0.07) --------

0.96

1.02

2.06

0.96

1.02

2.06

0.35 (0.76) 0.65

0.46*** 0.60*** (0.00) (0.00) ---------- ---------

0.94

1.06

2.36

0.94

1.06

2.36

0.37*** (0.00) ---------

0.01 (0.91) --------

0.99

1.07

1.87

0.99

1.07

1.87

0.80*** (0.00) ---------

-0.07 (0.66) ---------

0.99

1.04

2.16

0.99

1.04

2.16

-1.20*** (0.00) -1.90*** (0.00) -0.45*** (0.00) -2.25*** (0.00)

-2.35 (0.13) -4.37 -0.27 (0.42) -0.43 (0.21) ---------------

Note: 1 p-values in parentheses. 2 Taken from OLS estimation. With 3SLS the adjusted R2 is sometimes negative. It is also unclear how the goodness-of-fit measures of the different cross-sections are to be weighted in order to derive an overall goodness-of-fit measure.

We find a significant positive impact of increased Chilean price competition on market shares in the fish (03), fruit (08) and ores (26) sectors but no significant negative impact of foreign price competition on market shares in the seven sectors under study. As to beverages, we find a negative impact of competitive (low)

Chiles Market Share in the EU Market

209

Chilean prices and a positive impact of low foreign prices on market shares. This latter result was obtained in Section IV.A. as well. Adjustment to the long-run equilibrium was significant in the beverages (22), ores (26), wood (44), wood pulp (47) and copper (74) sectors, whereas no significant adjustment took place in the fish (03) and fruit (08) sectors. C. Estimating the Impact of Price Competition on Market Shares Utilizing the GMM Approach in a Dynamic Model Alternatively to 3SLS, we estimate the dynamic model by GMM. The Arellano and Bond (1991) estimator (see Baltagi, 2005) is not applicable in our case since the number of instruments created by the GMM technique exceeds the number of observations. Besides, the classical GMM technique does not allow one to control for the correlation between the lagged endogenous variable and the autocorrelated error terms. Judging from the way GMM works, this approach should have a comparative advantage over 3SLS in controlling for endogenity. However, efficiency is lost by creating a tremendous amount of moment conditions that have to be respected. In our case, we get 210 moment conditions, i.e., 210 restrictions, highlighting the computational burden of this approach (Schmidt et al., 1992). In Table 5, we see a positive relationship between an increase in Chilean price competitiveness and market share in the fruit sector (08) and a negative relationship between low Chilean wine prices (sector 22) and high Chilean copper prices (sector 74) and the respective market shares. Foreign relative prices have a significant and plausible impact on the fruit (03) and beverages (22) sectors. In the latter sector, the aspect of wine quality plays a dominant role. To sum up: all estimations (IV.A, IV.B and IV.C) have very respectable adjusted R2 measures, low standard errors and DurbinWatson (DW) statistics around 2. Although the DW must be adjusted in the presence of a lagged endogenous variable, the DW statistic is still able to roughly indicate problems of misspecification and autocorrelation of the disturbances. Price or quality competition is always relevant in the wine sector. We find in all estimations that low wine prices (standing for poor quality) are bad for Chile’s market share in the EU and that conversely, Chile can take advan-

210

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

Table 5: Results for the dynamic market share model estimated by panel GMM regression coefficients Equation (2’)

goodness-of-fit measures

D is  E i lreerist  J i lreerist * Olshwist 1  P ist lshwist

sectorresults

S.E. of adjustm. AR-term R2 adjusted regresCoeff. O J sion -0.78*** 0.64*** -0.24** 0.98 1.04 (0.00) (0.00) (0.02)

Durbin Watson stat. 2.11

-2.17*** ---------

------

0.98

1.04

2.11

-0.15 (0.90)

-0.15 (0.42)

0.69*** (0.00)

0.99

1.10

1.98

--------

---------

--------

0.99

1.10

1.98

2.29*** 0.58*** -0.13 (0.00) (0.00) (0.41)

0.98

1.06

2.08

5.45*** --------

0.98

1.06

2.08

0.89

1.04

2.04

impact impact lreer E lreer*

-0.20 03 short (0.24) run 03 long -0.55 run 2.29* 08 short (0.07) run 08 long --------run -2.53*** 22 short (0.00) run 22 long -6.02*** run 26 short 0.32 (0.52) run 26 long 1.10 run -1.22** 44 short (0.04) run 44 long -4.69** run -1.07** 47 short (0.05) run 47 long -1.78** (0.00) run -1.45** 74 short (0.02) run 74 long -2.30 run

--------

-0.17 (0.13) 0.24

0.71*** -0.28* (0.00) (0.06) --------- ---------

0.89

1.04

2.04

-0.98 (0.14)

0.74*** -0.37*** 0.90 (0.00) (0.00)

1.06

2.26

-3.77

---------- ---------

0.90

1.06

2.26

-0.31 (0.52)

0.40*** -0.05 (0.00) (0.80)

0.74

0.26

1.87

-0.52 (0.26) --------

---------- --------0.37*** 0.49*** (0.00) (0.03)

0.74

0.26

1.87

0.99

1.18

2.01

--------

--------

0.99

1.18

2.01

--------

Note: p-values in parentheses

tage of its competitors’ low-quality wine exports. The short-run price elasticity is around -2 in both the 3SLS and the GMM estimations, and the long-run price elasticity is very high: around -4 in the

Chiles Market Share in the EU Market

211

DOLS and -6 in the 3SLS and the GMM approach. Chilean relative prices significantly influence Chile’s market share in the fruit (08) and copper (74) sectors in all estimations. The impact of foreign price competitiveness is not significant in most sectors and also not robust when comparing different estimation techniques. The role of prices in the wood (44) and the wood pulp (47) sectors might be severely impeded by illegal logging and illegal imports of wood products. This phenomenon can be observed in the dynamic models that also contain the short and medium-run view. Illegal logging distorted official trade flows not only of all timber products (roundwood, sawnwood, veneer, plywood, boards, semi-finished and finished products, and furniture), but also of pulp, paper, printed products, and cellulose. Illegal logging is estimated to comprise up to 50 per cent of all logging activity in the key countries of Eastern Europe and Russia, up to 94 per cent in the key Asian countries, up to 80 per cent in the key African countries and up to 80 per cent in the key Latin American countries (WWF, 2005; FERN, 2004). V. Conclusions

In econometric terms, the DOLS approach using the usual semistatic model is inferior to the ARDL specification since it does not allow short-run inferences to be drawn. The ARDL specification solves the problem of having non-stationary series through intensive use of the FGLS technique. Applied to a system of equations, this technique transforms the variables in the regression equation through weighting the regressor matrix with a weighting matrix that can control for autocorrelation of the disturbances, for heteroscedasticity of the variance of the residuals and for cross-sectional correlation of the disturbances. The endogenity problem is solved by building in instrumental variables in either a 3SLS or GMM approach. Both techniques produce efficient and consistent estimates. In terms of good estimation properties, the DOLS estimator is a fine estimator as well. It delivers efficient estimates in large samples and valid statistical inferences when heteroscedasticityand autocorrelation-consistent (HAC) standard errors are used. In economic terms, we find that Chile´s market shares in the EU are subject to ups and downs and are therefore more of the Schumpeterian type. Market shares have to be permanently defended and entrepreneurs are under constant pressure to innovate and perform well. As to market shares in the wood (44) and wood pulp (47)

212

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

sectors, price competitiveness provides a poor explanation given to the worldwide problem of illegal logging. Product quality determines market shares in the wine exports sector (beverages 22), with customers demanding products of high or superior quality. The 3SLS approach, which we consider superior to the GMM approach, underlines the positive role of Chile’s price competitiveness for its market shares in the EU with respect to fish, fruit, ores, and copper. Estimation results obtained by DOLS or GMM methods were less conclusive.

References

Arellano, Manuel, Bond, Stephen (1991), Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations, Review of Economic Studies 58, 277297. Baltagi, Badi H. (2005), Econometric Analysis of Panel Data, 3rd Edition, John Wiley & Sons Ltd., Chichester. Banerjee, Anindya, Dolado, Juan J., Mestre, Ricardo (1998), ErrorCorrection Mechanism Tests for Cointegration in a SingleEquation Framework, Journal of Time Series Analysis 19, 267283. Banerjee, Anindya (1999a), Testing for Unit Roots and Cointegration Using Panel Data: Theory and Applications, Oxford Bulletin of Economics and Statistics 61(4), Special Issue. Banerjee, Anindya (1999b), Panel Data Unit Roots and Cointegration: An Overview, Oxford Bulletin of Economics and Statistics 61(4), Special Issue, 607-629. Breitung, Joerg, Pesaran, Hashem (2005), Unit Roots and Cointegration in Panels, : http://www.econ.cam.ac.uk/dae/repec/ cam/pdf/cwpe0535.pdf (last visit: 4 November 2005). Cable, John R. (1997), Market Share Behavior and Mobility: An Analysis and Time-series Application, The Review of Economics and Statistics Vol. LXXIX (1), 136-141. Dreger, Christian, Reimers, Hans-Eggert (2005), Health Care Expenditures in OECD Countries: A Panel Unit Root and Cointe-

Chiles Market Share in the EU Market

213

gration Analysis, : http://opus.zbw-kiel.de/volltexte/2005/ 3228/pdf/dp1469.pdf (last visit: 4 November 2005). Engle, Robert F., Granger, Clive W. J. (1987), Cointegration and Error Correction: Representation, Estimation and Testing, Econometrica 55(2), 251-276. Ericsson, Neil R., MacKinnon, James R. (2002), Distributions of Error Correction Tests for Cointegration, The Econometrics Journal 5 (2), 285-318. European Commission (2003), Intra- and extra-EU trade, Annual data, Combined Nomenclature, Supplement 2, EUROSTAT, CD ROM of COMEXT trade data base. European Commission (2005), The EU Relations with Chile, : http://europa.eu.int/comm/external_relations/chile/intro/ (last visit: 24 November 2005). EViews 5, User’s Guide (2004), Quantitative Micro Software, LLC, Irvine, CA. FERN, Greenpeace, WWF (2004), : http://www.panda.org/news_ facts/newsroom/news.cfm?uNewsID=17214 (last visit: 28 July 2005). Greene, William H. (2000), Econometric Analysis, London: Prentice Hall International (UK) Limited. Hendershott, Patric, MacGregor, Bryan, White, Michael (2002), Explaining Real Commercial Rents Using an Error Correction Model with Panel Data, The Journal of Real Estate Finance and Economics 24 (1/2), 59-87. Herzer, Dierk, Nowak-Lehmann, D. Felicitas (2006), What Can Export Diversification Do for Growth?, Applied Economics 38(15), 1825- 1838. 2002 Japan Conference: A Summary of the Papers, NBER Website, Friday, July 29, 2005; : http://www.nber.org/2002japanconf/ sutton.html (last visit: 29 July 2005). Johansen, Soren (1988), Statistical Analysis of Cointegration Vectors, Journal of Economic Dynamics and Control 12, 231254. Kapetanios, George (1999), A Test for m Structural Breaks under the Unit Root Hypothesis, Discussion Paper 152, National Institute of Economic and Social Research, London.

214

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

MacKinnon, James G. (1991), Critical Values for Cointegration Tests, in: Engle Robert F. , Granger Clive W. (eds.), Long-run Economic Relationships: Readings in Cointegration, Oxford, Oxford Press, 267-276. Nowak-Lehmann, D. Felicitas (2004), Different Approaches of Modeling Reaction Lags: How do Chilean Manufacturing Exports React to Movements of the Real Exchange Rate?, Applied Economics 36(14), 1547-1560. OECD (1997), The Uruguay Round Agreement on Agriculture and Processed Agricultural Products, OECD Publications, Paris. Pedroni, Peter (1999), Critical Values for Cointegration Tests in Heterogeneous Panels with Multiple Regressors, Oxford Bulletin of Economics and Statistics 61 (4), 653-670. Pedroni, Peter (2004), Panel Cointegration: Asymptotic and Finite Sample Properties of Pooled Time Series Tests: An Application to the PPP Hypothesis, Econometric Theory 20 (3), 597-625. Schmidt, Peter, Ahn, Seung C., Wyhowski, Donald (1992), Comment, Journal of Business and Economic Statistics 10, 1014. Stock, James H. (1987), Integrated Regressors and Test of the Permanent Income Hypothesis, NBER Working Paper no. 2359. Stock, James H. (1994), Unit Roots, Structural Breaks, and Trends, Chap. 46, in: Engle, Robert, McFadden, Daniel (eds.), Handbook of Econometrics, Vol. IV, Amsterdam: Elsevier. Stock, James H., Watson, Mark W. (1993), A Simple Estimator of Cointegrating Vectors in Higher Order Integrated Systems, Econometrica 61 (4), 783-820. Stock, James H., Watson, Mark W. (2003), Introduction to Econometrics, Boston: Addison Wesley. Supper, Erich (2001), Is there a Level Playing Field for Developing Countries Exports? UNCTAD, Policy Issues in International Trade and Commodities Study Series No.1, Geneva: UNCTAD. Sutton, John (2004), Market share dynamics and the ‘persistence of leadership’ debate, The Economics of Industry Group/ Suntory and Toyota International Centers for Economics and Related Disciplines, London School of Economics, 37.

Chiles Market Share in the EU Market

215

Westerlund, Joakim (2005a), New Simple Tests for Panel Cointegration, Econometric Reviews 24 (3), 297-316. Westerlund, Joakim (2005b), Testing for Error Correction in Panel Data, : http://swopec.hhs.se/lunewp/abs/lunewp2005_011.htm (last visit: 4 November 2005). World Bank (2002), TradeCAN (Database Software for a Competitiveness Analysis of Nations) 2002 CD-ROM, Washington, D.C. World Bank (2005), World Development Indicators, Data on CD ROM, Washington, D.C. WTO Trade Policy Review (1995, 1997, 2000), European Union, World Trade Organisation, Geneva. WWF (World Wildlife Fund) (2005), EU imports of wood based products 2002, : http://www.panda.org/about_wwf/wherewe work/europe/problems/illegal_logging/ (last visit: 29 April 2005).

216

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

Appendix 1

Description of Data In the following, the variables: sheu, shnoneu, shw, lreer, and lreer* will be described in original form (not in logs). All data run from 1988 to 2002. In our case, six cross-sections (six EU countries: Germany, Spain, France, UK, Italy, the Netherlands) had basically complete time series.16 (Ia) Chile’s market share in the EU with respect to the EU countries: sheu sheuist measures the share of Chilean exports (x) of sector s in EU country i at time t when competing against imports (m) from EU countries only: (A.1) Sheuist = xist/mEUist (Ib) Chile’s market share in the EU with respect to the non-EU countries: shnoneu shnoneuist measures the share of Chilean exports of sector s in EU country i at time t when competing against imports (m) from non-EU countries only: (A.2) shnoneuist = xist/mnon-EUist (Ic) Chile’s market share in the EU with respect to the world (EU and non-EU countries): shw shwist measures the share of Chilean exports of sector s in EU country i at time t when competing against imports (m) from EU and non-EU countries: (A.3) shwist = xist/mEUi+non-EUist (II) The Chilean real effective exchange rate: reer reer is the bilateral real effective exchange rate between Chile and the EU countries (price quotation system), taking Chile’s point of view. It consists of the real exchange rate (rer) and basic indicators of EU protection such as EU tariffs (t) and EU subsidies (s). 16

Due to missing data, Austria, Belgium, Finland, Luxemburg and Sweden were excluded from the analysis.

Chiles Market Share in the EU Market

217

It is computed (all data for ‘rer’ are taken from World Development Indicators CD ROM of 2005) as: (A.4) rer = e ˜ PEUi/PChile with rer = real bilateral exchange rate between Chile and relevant EU country e = nominal exchange rate (x Chilean Peso/1 EUR) between Chile and relevant EU country PEUi = GDP deflator of the EU country under consideration with 1995 as base year (1995 ˆ 100) PChile = GDP deflator of Chile with 1995 as base year (1995 ˆ 100) rer has been adjusted for EU tariff protection in terms of average EU tariff rate (t), and non-tariff protection in terms of EU subsidy rate (s). Tariff rates prevailing in the EU can be found in Trade Policy Review European Union, Volume 1, 2000, pp. 88-101 (WTO) and rough subsidy equivalents are based on qualitative information on non-tariff protection collected, explained and nicely put together for UNCTAD by Supper (2001). So we get: (A.5) reer = rer ˜ (1-s)/(1+t) For the simulations, we assume that the FTA between Chile and the EU brings tariffs down to zero. (III) Chile’s competitors’ (*) real effective exchange rates: reer* In analogy to (A.4 Appendix 1) the real effective exchange rates of Chile’s main competitors Norway, Australia, South Africa, Brazil are computed. Nominal exchange rates, Norway’s, Australia’s, South Africa’s, and Brazil’s GDP deflators are taken from World Development Indicators CD ROM 2005. Tariff and subsidy rates are borrowed from WTO and UNCTAD (see (A.4 Appendix 1)).

218

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso

Appendix 2

Development of the variables entering the market share model using the fish sector (03) as an example LSHW 03__DEU -5

LSHW 03__ESP 2.0

1.8

-6 1.6

-7

1.4 1.2

-8

1.0

-9 0.8

-10

0.6

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

LSHW 03__FRA 1.2

LSHW 03__GBR 0.4 0.0

0.8

-0.4

0.4

-0.8

0.0 -1.2

-0.4

-1.6

-0.8

-2.0

-1.2

-2.4

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

LSHW 03__ITA 0.0

LSHW 03__NDL -0.4

-0.4

-0.8

-0.8 -1.2

-1.2 -1.6

-1.6 -2.0

-2.0 -2.4

-2.4

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

Chiles Market Share in the EU Market

LREER03__DEU

LREER03__ESP

4.5

4.7

4.4

4.6

4.3

4.5

4.2

4.4

4.1

4.3

4.0

219

4.2

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

LREER03__FRA

LREER03__GBR

4.6

4.8

4.5

4.7

4.4

4.6

4.3

4.5

4.2

4.4

4.1

4.3

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

LREER03__ITA 4.9

LREER03__NDL 4.52 4.48

4.8 4.44

4.7

4.40

4.36

4.6 4.32

4.5

4.28 4.24

4.4 4.20

4.3

4.16

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

220

Felicitas Nowak-Lehmann Danzinger, Dierk Herzer, Sebastian Vollmer, Inmaculada Martínez-Zarzoso LREER03NO__DEU

4.5

LREER03NO__ESP 4.55

4.50

4.4

4.45

4.3 4.40

4.2 4.35

4.1

4.30

4.0

4.25

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

LREER03NO__FRA

LREER03NO__GBR

4.50

4.76

4.45

4.72

4.40

4.68 4.64

4.35

4.60

4.30 4.56

4.25

4.52

4.20

4.48

4.15

4.44

4.10

4.40

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

LREER03NO__ITA 4.65

LREER03NO__NDL 4.45

4.60

4.40

4.55 4.35

4.50 4.30

4.45 4.25

4.40

4.35

4.20

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

88 89 90 91 92 93 94 95 96 97 98 99 00 01 02

María Luisa Recalde and Marcelo Florensa

Mercosur: Trade Creation or Trade Diversion? An Application of the Gravity Model and Kalman Filter Abstract This paper aims to develop an econometric model which permits to assess the impact of MERCOSUR on Argentina’s international trade in manufactured products by specifying and estimating a gravity equation. First, we identify the starting year of the impact of the integration process and determine the variables that best explain bilateral trade flows. After determining that the structures of the monde and the antimonde are statistically different, we compute the impact of MERCOSUR. Our results show that MERCOSUR affected bilateral trade flows of manufactured products particularly since 1994. Trade creation, internal and external, in imports and exports is the main effect. I. Introduction In different empirical works, it was shown that trade flows respond to basic principles of the law of gravity. Two opposing forces determine the bilateral trade volume between countries: the economic activity level measured in terms of income and the magnitude of the obstacles against it. The latter are associated with transport cost, trade policies, cultural differences, consumer preferences, etc. Variables such as income, population and geographic distance were shown to be good forecasts of potential trade; consequently, the gravity models have been used extensively in articles meant to explain the determinants of international trade (Havrylyshin and Pritchett, 1991; Bayoumi and Eichengreen, 1995; Sanz, 2000; Martínez-Zarzozo and Nowak-Lehman, 2003).

We thank Iván Iturralde, member of the research team of the IEF, for his helpful collaboration.

222

Maria Luisa Recalde and Marcelo Florensa

In the integration process of a group of countries, two effects can be clearly seen: a) trade creation which manifests in new flows amongst member countries which replace domestic production. This effect is possible when the member countries reduce their tariffs on imports from third countries as part of the commitment to reach a common external tariff. b) Trade diversion consisting in import substitution from non member countries (lower cost products) to imports from member countries (higher cost products). Trade creation minus trade diversion gives the net effect of an agreement. The ex-post measurement of the integration effects on trade flows has led many authors to build and implement the so-called trade impact models. Although some of these studies were developed on the concepts of general or partial equilibrium models, most of them have used the so-called “gravity equation” which has demonstrated to be a very useful tool in modelling bilateral trade flows. When using these models, special attention should be paid to the estimation of trade flows which hypothetically would have taken place if integration had not been the case. The structure that generates the data in the post-integration period of a trade agreement is known as monde, while the antimonde is the structure that would generate the data in the same period in the absence of such agreement. The impact of the integration process is therefore defined as the gap between the predictions of both structures provided that they are statistically different. As can be expected, the main difficulty of this kind of models lies in the estimation of the antimonde, which will be dynamically performed in this paper. The integration process of MERCOSUR begins with the signature of the Argentinean-Brazilian Cooperation and Integration Act in 1986, within the frame of the ALADI. The subscription of the Agreement of Asuncion in 1991, benchmark of the MERCOSUR, constitutes a free trade area between Argentina, Brazil, Paraguay and Uruguay and establishes the goal to achieve a common market that would start working in 1995. The agreement sets an initial decrease of 40 per cent on tariffs between member countries starting in June 1991 with reductions every six months to reach a zero-tariff situation in 1995. A common external tariff would arrive as well. This paper is aimed at developing an econometric model which permits assessing the impact of MERCOSUR on Argentina’s international trade of manufactured products by specifying and estimat-

Mercosur: Trade Creation or Trade Diversion?

223

ing a gravity equation. To achieve this goal, the paper is divided into two parts. In the first one, we identify the year from which the integration process begins to impact on the international trade of Argentina. Likewise, we determine the variables that best explain bilateral trade flows. In the second part, and as the main contribution of this study, we estimate the antimonde. After determining that the structures of the monde and the antimonde are statistically different, we compute the impact of MERCOSUR on the trade flows of Argentina’s manufactured products. The structure of the antimonde is dynamically estimated by using the Kalman filter. This recursive algorithm uses all the available data on trade flows until the first year in which the agreement became effective, to estimate a different antimonde for each post-integration year. The Kalman filter allows to obtain an update of the vector of coefficients when new information becomes available. However, as the forecast horizon lengthens, the explanatory power of these regressions decreases because the forecast for the next period is based on a previous forecast and not on observable data. Nevertheless, this approach appears to be more realistic than the ones used in former articles, where the structure of trade in the antimonde was assumed to remain constant since integration. The paper is divided into five main sections. After the introduction, the following section describes the evolution of Argentina’s international trade of manufactures. The following section develops the adopted gravity model in detail. In section IV, results are commented and finally section V concludes. II. Argentina's International Trade in Manufactured Goods The Argentine international trade in manufactured goods in the period 1970-2004 had a very different development if looking at exports and imports.

224

Maria Luisa Recalde and Marcelo Florensa

Figure 1: Evolution of the Manufactures Exports and Imports in millions of US-$ 35000

Imports

30000

US-$

25000

Exports

20000 15000 10000 5000

2003

2000

1997

1994

1991

1988

1985

1982

1979

1976

1973

1970

0

years

Source: Based on CEPAL (2005).

If the first and the last year of the period are compared, exports multiplied by a factor of 19 while imports only by about 13. It is necessary to highlight the little fluctuation, but with an increasing trend of exports compared to the cyclic behaviour shown by imports. This resulted in the fact that only eight years of the period exhibited positive trade balance for the manufactured goods. The magnitude of the deficit became remarcable in the 1990s due to

Manuf M/ Total M Manuf (X+M)/Total (X+M) Manuf X / Total X

Source: Based on CEPAL (2005).

2003

2000

1997

1994

1991

1988

1985

1982

1979

1976

1973

100% 80% 60% 40% 20% 0%

1970

%

Figure 2: Exports and Imports shares of Argentina´s international trade in manufactures

years

Mercosur: Trade Creation or Trade Diversion?

225

currency overvaluation and openness of trade policy. In the Argentine case, fluctuations of international trade of manufactured goods have been determined not only by the exchange rate policy implemented in the country but also by the cycle of domestic economic activity. The share of manufactured goods in global exports shows a clear growing trend starting with a 38 per cent in the beginning of the period and reaching 60 per cent in 2004. We observe a peak of 70 per cent in 1989 (Figure 2). The behaviour of imports was much more stable varying between 80 per cent and slightly above 90 per cent. Table 1: Geographical Distribution of Argentina´s exports of manufactures (in per cent) (to be continued) 1970 1974 1978 1982 1987 1991 1995 1999 2004 Germany

5.97 4.59 6.45 4.52 3.80 3.26 1.80 1.78 2.11

Spain

0.97 1.93 2.05 1.04 1.26 2.56 2.09 3.14 4.42

France

2.62 2.24 3.04 1.99 1.38 1.75 1.32 1.44 1.39

Italy

5.54 4.46 4.71 3.60 3.11 4.21 3.98 3.53 3.96

Netherlands

19.55 8.05 11.05 11.75 10.59 10.57 5.16 4.14 5.25

TOTAL EU

47.23 28.73 35.25 29.07 26.33 27.49 19.87 19.01 21.93

United States and Canada

22.02 17.64 18.28 26.74 22.80 15.94 11.45 13.60 14.18

Brazil

7.75 12.91 9.57 7.43 8.02 11.86 26.85 26.89 18.13

TOTAL MERCOSUR Chile Rest of Latin America

13.45 18.34 16.96 13.72 13.52 17.78 35.23 35.79 23.15 4.51 7.62 4.85 3.61 3.38 4.92 5.17 5.80 8.96 9.06 15.27 15.88 11.13 10.24 11.13 9.04 8.03 10.35

Arab Countries 0.47 6.89 2.83 3.89 3.77 6.65 6.03 4.79 4.44 China

0.10 0.10 0.44 1.43 3.20 2.77 0.86 2.15 7.06

Rest of the World TOTAL

1.91 4.74 3.82 6.02 7.71 10.16 11.58 10.05 9.44 100

100

100

100

100

100

100

100

100

226

Maria Luisa Recalde and Marcelo Florensa

As regards geographical distribution of trade, Tables 1 and 2 show the main origins and destinations of manufactures trade of Argentina. There are three facts regarding exports that should be pointed out: the relative decrease of the European Union, United States and Canada as destinations; the increase of the relative importance of Brazil, Chile, Arab countries and the rest of the world; and the great increase and ulterior diminishment of MERCOSUR in a similar way as that of Brazil. The picture is similar when imports are analysed; the relative importance of the European Union, United States and Canada has decreased, whilst Mercosur and Brazil have gained position. The difference with respect to the behaviour of exports is that imports coming from Mercosur and Brazil have had a steady growth in the relative share with no slowdown as in the exports case. Table 2: Geographical Distribution of Argentina´s imports of manufactures (in per cent) 1970 1974 1978 1982 1987 1991 1995 1999 2004 Germany

12.56 13.72 14.51 10.98 13.15 11.38 6.79 5.95 6.19

Spain

2.00 1.19 3.74 4.04 2.49 3.19 4.26 4.11 3.13

France

4.21 4.05 4.68 4.50 5.86 5.21 5.37 6.34 2.53

Italy

8.15 7.05 9.40 5.29 6.05 6.41 6.98 5.64 3.50

Netherlands

2.10 2.47 1.87 2.21 2.43 1.44 1.75 1.14 0.84

TOTAL EU

43.04 41.67 48.94 37.20 38.77 35.92 32.85 31.13 23.01

United States 45.14 44.14 50.80 39.41 41.20 37.36 34.60 32.27 23.84 and Canada Brazil

10.28 8.26 7.65 13.20 14.27 16.85 20.22 22.31 34.89

TOTAL 11.28 10.08 9.72 16.00 17.04 19.77 24.38 24.85 37.92 MERCOSUR Chile

4.70 6.03 4.08 3.23 3.32 2.33 3.62 2.31 2.05

Rest of Latin America

3.76 4.20 2.65 3.06 6.09 4.58 2.81 2.67 2.37

Arab Countries 0.03 0.23 0.00 0.00 0.00 0.13 0.22 0.21 0.19 China

0.06 0.03 0.07 0.15 0.16 0.87 1.07 4.19 5.75

Rest of the World

1.55 2.58 2.82 4.11 6.01 7.74 6.98 8.29 7.20

TOTAL

100

100

100

100

100

100

100

100

100

Mercosur: Trade Creation or Trade Diversion?

III.

227

The Gravity Equation

A. Precedents The gravity models were used for the first time in the 1970s by Jan Tinbergen and Pentti Pöyhönen to explain the trade flows and ever since then they have been widely used when it comes to assessing the effects of integration in trade agreements. Sanz (2000, 2001) points out that the impact of trade models that use the gravity equation are divided into two groups: those (dummy-variable approach) which estimate the coefficients of the dummy preference variable to compute the effect of trade creation amongst the member countries; and those (projection approach) that use the statistic significance of the dummy variable to determine the first year from which integration generates an effect. An extension of the last ones will be applied in this paper. Aitken (1973) has developed both kinds of models to assess the effects of integration for the member countries of the European Economic Community (EEC) and the European Free Trade Agreement (EFTA) during the period 1959-1967. Using the first model type, he estimated an equation for each year of the period for testing existence and magnitude of the integration effects. Estimating this equation for several years before the integration period and using dummy variables, Aitken could determine the first year from which integration effects took place. This is achieved by the observation of the significance levels of the preference dummy variables. Using the second type of model, he eliminated the estimated coefficient of the dummy variable of the last year, in which integration had caused no effect, from the equation (crosssection of year 1985) and that structure was projected for the postintegration period. Hence, trade creation and trade diversion could be directly computed. Similar methodologies to those of Aitken were applied in later articles, amongst them: Aitken and Obutelewicz (1976), Sapir (1981), and Brada and Méndez (1985) for other trade agreements. Pelzman (1977) developed a similar but more sophisticated method, differing in methodological aspects as well. Firstly, he detected whether there was a structural break during the years after the constitution of the Council for Mutual Economic Assistance (CMEA, 1954-70) for a group of Eastern Europe countries; this was done by using a statistic procedure somehow more complex than the simple analysis of significance of the dummy variable. A

228

Maria Luisa Recalde and Marcelo Florensa

second methodological aspect was introduced by Pelzman (1977) in response to the probability that the performed estimates based on the structure of just one year may bias the results due to the absence of prices in the gravity equation. To eliminate the bias raised from changes of prices in the short run, he proposed to add cross-sections for five more years. More recent papers which applied the gravity equation to assess trade flows made use of advanced of panel data techniques, for example, Martínez-Zarzoso and Nowak-Lehmann (2002), Kamil and Ons (2001), Carrère (2004), Egger (1999), Filippini and Molini (2003) and De Groot et al. (2003). In this vast literature, gravity equations possess a common design which can be adapted for different purposes: i) They are bilateral-type. They explain the behaviour of a dependent variable related to foreign trade (for instance, exports) through a set of macroeconomic variables (income, exchange rate, prices, etc.) between two countries; ii) They can be used to estimate the determinants of the trade flows volume and nature; in the latter case, an intra-industrial trade index can be used as dependent variable. iii) They permit to use basic indicators and consequently the integration of a great number of countries in the sample. iv) They are estimated by using aggregated data. v) They do not respond to a specific international trade model.1 B. The adopted model In spite of the advances introduced by Pelzman (1977) using the cross-section-type technique and of the alternative methodologies which use panel data, it is necessary to highlight that the projection of the structure of the gravity equation is basically static. It does not take the development of trade flows over time into account. This means to consider that, if there were no integration (in the antimonde), factors such as the penetration of exports in foreign markets, the efficiency level of firms, the appearance of economies of scale in production, the evolution of non-tariff barriers, economic 1

The discussion on the theoretical foundations of the gravity equation can be found in Anderson (1979) and Bergstrand (1985, 1989). Nevertheless, the success of the gravity equation is based on its flexibility and the fact that its results do not respond to a specific international trade model. This makes them a powerful tool (Rose, 2000).

Mercosur: Trade Creation or Trade Diversion?

229

cycles, etc. would remain unchanged or their effects on international trade would be exactly compensated (Mayes, 1978). This is a highly unlikely situation, especially when considering the Argentinean economy, prone to a great instability during the analysed period and the application of the Convertibility Plan (currency board) between 1991 and 2001 with a large number of measures aiming to achieve more economic openness. In this respect, Sanz (2000) and Sanz and Gil (2001) managed to solve the problems associated with the static presentation derived from the use of the gravity equation applying space-state models and the Kalman filter. We follow this approach and use these econometric techniques in our paper. The applied methodology is based on the estimation of the gravity equation that responds to the following specification:

M ij

AYi E1Y jE2 LEi 3 LEj 4 DijE5 e

uij

(1)

where

M ij is the current value of the sales from country i to country j,

A is a constant, Y is income, L is population, Dij is distance between countries i and j,

uij is the error term. Although earlier empirical papers which used the gravity equation to model bilateral trade flows, among them Aitken (1973), used the log-linear specification, Sanso, Cuairan and Sanz (1993) showed that the specification of the gravity equation in logarithmic terms is not necessarily the best functional form. Since the model in logarithms represents a particular case of the Box-Cox transformation, an appropriate analysis of the best functional form would consist on estimating the optimal parameters of such transformation. Therefore, the following model has been estimated for each year:2 M ijO0 E 0  E1 yiO1  E 2 y Oj 2  E 3 LOi 3  E 4 LOj4  (2)  E 5 DijO5  E 6 Merco  E 7 Languaje ge  P ij 2

The procedure by which the values of the parameters of the Box-Cox transformation are obtained is shown in detail in the Appendix 1.

230

Maria Luisa Recalde and Marcelo Florensa

where M, L and D are as defined previously, y is per capita income3 (the subscript i is for the exporter country and j for the importer ), Merco is a dummy variable which equals 1 if Argentina´s trading partner is a MERCORSUR member, Language is a dummy variable which takes 1 if Argentina trades with a Spanish-speaking country, Oi is the parameter of the Box-Cox transformation associated with the corresponding variable. The measurement of the effects of integration on international trade requires that the coefficients of the gravity model, the E i s, vary over time. The regression models with coefficients that vary over time can be conveniently implemented by using what is denominated as the Space-State Form (SSF henceforth). Generally speaking, the SSF includes two equations. The first, called measurement equation, relates the vector of observable variables, Yt , to a vector of non observable variables, D t , called state vector, in the following way:4 Yt Z t D t  d t  H t , where (3) t 1,..., T , Yt is a vector of n elements, D t is a vector of m elements, Z t is an n x m matrix, d t is a vector of n elements and finally, H t is a vector of n elements of uncorrelated disturbances with zero mean and a matrix of variances and covariances H. Although the elements of the state vector are non-observable, it is known that they are generated by a first order Markov process, which has the following structure: D t Tt D t 1  ct  RtK t , where (4) t 1,..., T , D t and D t 1 are vectors of m elements, 3

4

Sanso et al. (1993) show that the gravity equation in which income is defined in per capita terms performs better than with GDP or GNP as indicator for income. The notation follows that adopted by Harvey (1989).

Mercosur: Trade Creation or Trade Diversion?

231

Tt is an m x n matrix,

ct is a vector of m elements, Rt is an m x j matrix and K t is a vector of j elements of uncorrelated disturbances of zeromeans and matrix of variances and covariances Q. The last equation is called transition equation. Once the model has been set in the SSF, the technique known as Kalman filter can be applied. The advantage of using the SSF and the Kalman filter is that the latter provides a statistically optimal estimator, since it minimises the mean square error of the estimation of the state vector at the moment t, t = 1,...T ; giving a solution to the dynamic regression framework as in this paper. We will use the Kalman filter to estimate the antimonde in an essentially dynamic way, applying it over the gravity equation. Using the above notation, the model can be specified as follows in matrix terms: (5) Mt Xt Et  Pt t 1,,T

Et

C E t 1  Xt

(6) where Ms is an N x 1 vector of trade flows during period t Xt is the N x k matrix of exogenous variables of the gravity equation E t is the state vector of dimension k x 1 that defines the structure during period t Pt is an N x 1 vector of random errors which is distributed as Pt ~ N ( 0 , H t ) with Ht being an N x N scalar matrix.

Xt is a k x1 vector of random errors distributed N a (0, Q) Equation (5) is the measurement equation and represents a cross section version of the gravity equation. Equation (6) is the transition equation, which includes the dynamic behaviour of the parameters of the gravity equation. C is a k x k matrix which establishes the relationship between the structures of two consecutive periods. The model which independently generates each E t is given by E it ci E it 1  X it i 1, 2, , k t 1, , T (7) which is the same as saying that C is diagonal in equation (6). The errors vector Xt has a normal distribution with a zero means vector and a matrix of variances and covariances Q, diagonal k x k.

232

Maria Luisa Recalde and Marcelo Florensa

Besides, the initial state vector E 0 is assumed to be uncorrelated with neither Pt nor Xt , which are not correlated with each other as well. Having defined the model in equations (5), (6) and (7), the Kalman filter is a set of equations that solves the problem of estimating Et using the available information up through period Z(It); E( Et /Iz) = Et/z denotes the conditional expectation of Et given Iz . The evaluation of Et/z is called “prediction” when t ! z and “updating” when t = z. We define bt/z to be the estimate of Et/z with covariance matrix Pt/z. Therefore, the form of the Kalman filter for the present model is the following (Harvey, 1989): a) Prediction equations:

bt / t 1 Cbt 1 Pt / t 1 CPt 1C c  Q

(8)

b) Updating equations:

bt / t Pt / t where

bt / t 1  Pt / t 1 X tcFt 1  M t  X t bt / t 1

(9)

1

Pt / t 1  Pt / t 1 X tcFt X t Pt / t 1 Ft X t Pt / t 1 X tc  H t

This algorithm will use all available trade data for the pre-integration period but no data from the post-integration period, defining a different antimonde for each post-integration year. IV.

Empirical Application

A. Data We decided to model manufacturing trade instead of total trade since Argentina´s international trade in manufactured goods represents about 80 per cent of total trade and has steadily increased.5 Using data on Argentina´s bilateral trade flows, we have 34 cross sections covering the period from 1970 to 2003. Trade with the following 54 countries is included: Germany, Saudi Arabia, Algeria, Australia, Austria, Belgium-Luxemburg, Bolivia, Brazil, Canada, Chile, China, Colombia, South Korea, Cuba, Denmark, Equator, Egypt, Spain, United States, Finland, France, Greece, 5

In Recalde and Florensa (2007), the gravity equation was estimated for agricultural products.

Mercosur: Trade Creation or Trade Diversion?

233

Haiti, India, Indonesia, Iran, Ireland, Israel, Italy, Japan, Jordan, Libya, Malaysia, Morocco, Mexico, Nigeria, Netherlands, Panama, Paraguay, Peru, Portugal, United Kingdom, Singapore, Syria, South Africa, Sweden, Switzerland, Thailand, Taiwan, Tunisia, Turkey, the former Soviet Union countries, Uruguay and Venezuela. Consequently, we have 108 observations per year, which represent over 90 in per cent of the international trade of the manufacturing sector of Argentina. Data on trade flows is taken from the Base de Datos de Comercio Exterior (BADECEL) of CEPAL. The value of exports and imports is taken in current US dollars (US-$). Data on income and population are from International Financial Statistics Online, published by the International Monetary Fund in its web site. The Data base of Angus Maddison (from the book “The World Economy”) and from SESRTCIC (Statistical, Economic and Social Research and Training Centre for Islamic Countries) were used for certain countries for which information is not complete.6 Data corresponding to distances is from the data base on international trade available on Jon Haveman’s web site. B. Monde, Antimonde and Structural Break. Before explaining the Kalman filter, we first have to determine the optimal functional form of the gravity equation, that is to say, the set of O s that fits best to the data. This is performed using the maximum likelihood method. The results for each year are given in Table 3. Although the obtained estimates are close to zero, the likelihood ratio test LRT given in the last column indicates in each year to reject the null hypothesis of log-linearity. Once the optimal functional form has been obtained, the transformed variables are used as inputs to obtain the E i , through Ordinary Least Squares (OLS). The obtained estimates with the R 2 are shown in Table 4. The signs of the coefficients for per capita income and population are positive as predicted by the gravity equation. The coefficient of distance has the expected sign and is negative for each year except 2002, where it is not significant.

6

These countries are: Saudi Arabia, Algeria, Cuba, Egypt, Indonesia, Iran, Jordan, Libya, Syria, Tunisia and former Soviet Union countries.

234

Maria Luisa Recalde and Marcelo Florensa

Table 3: Maximum Likelihood Estimates of the Box-Cox Parameters (lambdas) Year 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003

Ȝ0 0.12 0.12 0.13 0.16 0.16 0.16 0.18 0.17 0.19 0.2 0.17 0.19 0.19 0.19 0.18 0.19 0.19 0.19 0.2 0.2 0.19 0.19 0.18 0.16 0.16 0.17 0.16 0.14 0.17 0.2 0.18 0.18 0.2 0.22

Notes:

+

Ȝ1 -0.19 -0.16 -0.13 -0.16 -0.19 -0.01 -0.17 -0.17 -0.12 -0.07 -0.06 0.00 -0.05 -0.03 -0.11 -0.13 -0.04 -0.16 -0.11 -0.25 -0.17 -0.22 -0.01 -0.05 0.03 -0.14 -0.03 -0.18 -0.01 0.01 -0.13 -0.21 -0.32 -0.16

Ȝ2 -0.21 -0.16 -0.28 -0.28 -0.53 -0.31 0.07 -0.05 0.61 0.48 0.03 0.17 0.54 1.19 0.75 1.75 0.43 -0.08 0.35 0.23 0.59 0.35 -0.26 -0.21 -0.38 -0.67 -0.94 -0.97 -0.73 -0.41 -0.39 -0.8 0.66 0.88

Ȝ3 0.22 0.21 0.24 0.07 0.23 0.08 0.15 0.13 0.12 0.16 0.2 0.19 0.15 0.12 0.17 0.09 0.09 0.15 0.07 0.03 -0.09 0.24 0.17 0.21 0.24 0.24 0.13 0.26 0.18 0.1 0.18 0.16 -0.03 0.02

Ȝ4 0.01 -0.21 0.06 -0.1 -0.18 -0.09 0.15 0.3 0.09 0.07 0.08 0.17 0.14 -0.16 0.25 0.11 0.12 0.2 0.21 0.29 0.03 -0.02 -0.09 -0.1 -0.14 -0.03 -0.02 -0.06 -0.12 -0.12 -0.12 -0.07 0.2 0.27

Ȝ5 0.39 0.37 0.39 0.43 0.42 0.28 0.36 0.4 0.35 0.34 0.22 0.27 0.24 0.19 0.27 0.29 0.24 0.24 0.28 0.25 0.1 0.19 0.19 0.17 0.22 0.29 0.3 0.21 0.19 0.18 0.26 0.26 0.29 0.19

LRT 27.3* 26.67* 32.62* 48.01* 48.87* 42.68* 62.52* 58.8* 72.16* 81.05* 60.2* 70.99* 71.62* 63.76* 67.92* 73.12* 66.77* 84.35* 83.94* 78.92* 77.21* 81.2* 68.48* 68.56* 69.49* 79.8* 59.17* 61.61* 69.72* 94.23* 77.78* 76.48* 80.76* 100.26*

Likelihood Ratio Test; * indicates that the variable is significant at the 5 per cent level.

Regarding the statistical significance of the coefficients at the 5 per cent level, we verified that yi O , y j O , and L j O are significant in 1

2

4

Mercosur: Trade Creation or Trade Diversion?

235

each year of the period. The variable Dij is significant in each year of the period except 2002 and 2003. Concerning the dummy variables, Language is only significant in 1993 and 1994, therefore we can conclude that Argentina does not trade more, on average, with Spanish-speaking countries. Finally, the dichotomic variable Merco becomes statistically significant since 1994, the period in which the formation of Mercosur has already advanced. Hence, we find evidence that 1994 is the first year in which integration produces effects on the bilateral trade flows. Accordingly, we will consider 1970-1993 as the pre-integration period and 1994-2003 as the post-integration period. The estimates obtained for the period 1994-2003, shown in Table 3 and 4, represent the monde; that is to say, the structure which fits the data best for the period after the signing of the integration agreement. Table 4: OLS estimates of the position parameters (ßi) (to be cont.) Year (D.S)

ȕ0 (C)

1970 -132.6817* (16.7032) 1971 -406.5610* (77.465) 1972 -150.1939* (18.2363) 1973 -318.2210* (37.9724) 1974 -650.9428* (93.0698) 1975 -264.5851* (42.0774) 1976 -159.5016* (19.9838) 1977 -183.6036* (21.093) 1978 -158.7623* (20.0198) 1979 -168.7048* (20.8231) 1980 -162.2312* (22.3155) 1981 -142.6403* (20.3807) 1982 -140.2085* (21.0208)

ȕ4 ȕ7 ȕ3 ȕ1 ȕ6 ȕ2 ȕ5 (langu(popula- (popula(distance) (Merco) (pcgdpi) pcgdpj) tion ) i tionj) age) 17.2135* (2.1287) 14.0262* (1.6825) 14.4915* (1.6146) 21.9668* (2.2997) 28.9064* (2.9524) 7.7886* (0.9087) 32.4929* (3.575) 34.8909* (3.6973) 25.2916* (2.7547) 20.2808* (2.1094) 16.5798* (1.9135) 12.6584* (1.4312) 18.7991* (2.0673)

14.4036* 0.0599* (2.3684) (0.0116) 8.5080* 0.0744* (1.6054) (0.0131) 24.7955* 0.0479* (4.2516) (0.0084) 24.5559* 1.1616* (5.1674) (0.1935) 126.1680* 0.0781* (28.42779) (0.0124)

32.0425* (7.8237) 2.4741* (0.5763) 6.7085* (1.4970) 0.0399* (0.0084) 0.1349* (0.0265) 3.9634* (0.934) 1.8434* (0.3706) 0.0777* (0.0161)

1.1565* (0.1972) 0.4326* (0.0682) 0.4999* (0.0753) 0.7442* (0.1164) 0.4654* (0.0705) 0.1865* (0.0307) 0.2753* (0.0448) 0.5043* (0.0825)

1.5322* (0.4529) 70.6572* (16.537) 0.8858* (0.1923) 18.0873* (3.4268) 60.6272* (12.5697) 14.1337* (3.4826) 0.2349* (0.063) 0.0175* (0.0042) 0.9048* (0.2107) 1.6777* (0.3439) 1.2951* (0.2466) 0.3043* (0.0589) 0.4173* (0.0884)

-0.0097* (0.0031) -0.0119* (0.0039) -0.0120* (0.0031) -0.0094* (0.0022) -0.0106* (0.0026) -0.0822* (0.0236) -0.0370* (0.0086) -0.0152* (0.0036) -0.0368* (0.0104) -0.0432* (0.0128) -0.1756* (0.083) -0.1056* (0.0462) -0.2047* (0.0721)

2.9198 (3.0699) 2.6281 (2.9041) 1.4083 (3.2797) 1.2947 (3.793) 0.6949 (3.79649)

0.5583 (4.3935) -0.6737 (4.8878) -1.0403 (4.6873) 3.1522 (5.5315) 5.4654 (6.1823) 7.0275 (5.949) 7.1277 (6.9876) 5.2957 (6.5188)

0.5872 (1.939) 1.267 (1.8337) 0.9186 (2.0743) 0.8261 (2.4355) 2.2765 (2.4266) 4.1677 (2.7111) 3.7229 (3.0778) 5.3507 (2.9879) 6.34 (3.4946) 8.0525 (3.8962) 8.0016 (3.5816) 7.333 (4.2627) 6.4989 (3.9464)

R2 0.57 0.58 0.61 0.62 0.61 0.57 0.6 0.61 0.61 0.63 0.58 0.59 0.59

236

Maria Luisa Recalde and Marcelo Florensa

Table 4 (cont.): OLS estimates of the position parameters (ßi) Year (D.S)

ȕ0 (C)

1983 -447.2396* (92.3544) 1984 -136.2011* (21.0607) 1985 -193.7770*

ȕ7 ȕ4 ȕ3 ȕ2 ȕ1 ȕ6 ȕ5 (popula- (popula(langu(distance) (Merco) (pcgdpi) pcgdpj) tion ) i tionj) age) 14.0295* (1.7139) 26.1805* (3.0525) 32.3098*

0.0002* (0.0001) 0.0108* (0.0028) 0.0000*

0.7549* 60.5758* -0.3486* 6.8923

6.681

R2 0.52

(0.145) (15.0458) (0.1482) (6.7606) (4.0625) 0.2758* 0.0550*

-0.1100* 6.3952

6.3798

0.55

(0.0524) (0.01269 (0.0409) (6.3218) (3.8691) 1.1975* 0.8246*

-0.1023* 5.2085

4.6829

(3.6278) (0.0000) (0.2119) (0.1519) (0.031) (6.3076) (3.9072) 1986 -151.8866* 15.3092* 0.1493* 1.5209* 0.7699* -0.2137* 3.9262 7.4625 0.0617 6.0007 3.6483 19.8247 1.5455 0.0355 0.2304 0.1301

0.59

(24.5841)

0.62

1987 -263.4316* 41.7482* 12.2118* 0.5047* 0.2387* -0.2484* 2.5993 5.9426 (23.8749) (3.3907) (1.9417) (0.0648) (0.0296) (0.0567) (5.3425) (3.219)

0.71

1988 -192.2790* 27.3348* 0.2966* 2.1336* 0.2030* -0.1396* 4.9514 6.9741 (23.108) (2.7847) (0.0603) (0.3399) (0.029) (0.0391) (6.4439) (3.9496)

0.64

1989 -317.4777* 72.3904* 1.0364* 4.1747* 0.0443* -0.1818* 4.7258 6.9288 (36.4435) (8.3483) (0.1602) (0.6537) (0.0065) (0.0564) (6.3168) (3.8147)

0.62

1990 -495.5508* 38.3480* 0.0302* 36.9838* 3.6913* -1.2907* 7.9811 7.0759 (49.0292) (3.8899) (0.0057) (4.5043) (0.5912) (0.4793) (5.9013) (3.4024)

0.66

1991 -355.1760* 66.2441* 0.2173* 0.1144* 9.1425* -0.3809* 6.1524 8.7014 (32.6715) (5.7093) (0.0495) (0.0141) (1.3648) (0.1215) (5.7849) (3.4248)

0.68

1992 -397.6552* 10.7683* 30.9049* 0.3763* 26.6374* -0.3837* 7.5768 6.3617 (49.7237) (0.9971) (7.105) (0.0444) (4.1296) (0.122) (5.4808) (3.2383)

0.67

1993 -357.1658* 14.5354* 18.8711* 0.1497* 26.9967* -0.4040* 8.8715 5.9743* (43.4853) (1.1936) (4.2369) (0.0165) (4.3431) (0.1322) (4.6756) (2.7347)

0.7

1994 -539.4586* 6.8021* 66.3604* 0.0823* 51.9534* -0.1685* 9.1538* 5.6837* (63.7304) (0.549) (13.2472) (0.0097) (7.3846) (0.0574) (4.0901) (2.4349)

0.72

1995

0.8

-1440.1*

35.1689* 779.7985* 0.1306* 9.8411* -0.0619* 10.8932* 4.0055

(189.7955) (2.2103) (123.8273) (0.0106) (1.1624) (0.0181) (3.9672) (2.4093) 1996 -3586.62* 11.34* 3190.69* 0.7121* 7.4641* -0.0411* 11.4457* 3.7139

0.77

(879.3069) (0.8343) (822.7003) (0.0644) (0.9104) (0.0145) (3.5609) (2.1769) 1997 -4327.41* 34.03* 3942.59* 0.0565* 13.8263* -0.1279* 10.4585* 2.7066

0.76

(981.7737) (2.4673) (949.2094) (0.0053) (1.6569) (0.0525) (3.2835) (1.9422) 1998 -1665.32* 11.0357* 894.7554* 0.3181* 50.3706* -0.2466* 16.4386* 4.5534

0.77

(296.0559)

(0.805) (205.5906) (0.0283) (6.0595) (0.1045) (4.4348) (2.6159) 1999 -823.9210* 12.8435* 99.4086* 2.1544* 63.5893* -0.4063* 22.6739* 5.0946 (0.9788) (27.2515) (0.1868) (8.0451) (0.1605) (5.9534) (3.5025) 2000 -724.8783* 35.9902* 74.2856* 0.3900* 50.9623* -0.1027* 20.6168* 3.0701 (84.0422) (2.8763) (20.7426) (0.03589) (7.0215) (0.0394) (5.1593) (3.1239)

0.76

(97.6567)

2001 -2166.48* (592.4769) 2002 -675.9916* (71.3084) 2003 -400.6119* (46.5904)

62.0071* 1400.56* (5.0946) (468.9191) 154.0557* 0.01606* (19.6116) (0.0032) 49.8358* 0.0022* (6.51) (0.0005)

0.74

0.5721* 18.3567* -0.1065* 18.5750* 2.4183 (0.0521) (82.6838) (0.0402) (4.9371) (2.9918)

0.73

20.6115* 0.1949*

0.63

-0.0689 25.8144* 5.8240

(2.4549) (0.0309) (0.0346) (7.2025) (4.4342) 9.9439* 0.0842* -0.2517 37.8154* 11.8614 (1.2251) (0.014) (0.2229) (9.9846) (5.935)

0.62

Notes: * indicates that the variable is significant at the 5 per cent level. Standard deviations are in parentheses.

Mercosur: Trade Creation or Trade Diversion?

237

Given these results, the next is to define the antimonde, for which it is necessary to apply the Kalman filter to the E s estimated for the pre-integration period. However, from the observation of the evolution of the estimates over time it appears that they have a high variability, which makes the application of the filter difficult. A strategy to solve this situation is to test the hypothesis of a single functional form for the whole pre-integration period, and in case of not rejecting it, to apply the Kalman filter to the data of the period 1970-1993, after these were conveniently transformed. This functional form, which will be symbolised as O7093 is the average of the values of the Oi obtained in the period 1970 to 1993:

O7093 = (0.175 -0.115 0.214 0.142 0.055 0.287) To determine the validity of this functional form, we perform a likelihood ratio test for each year of the period 1970-1993, which is symbolised as LR1, where the null hypothesis is O7093 versus the optimal functional form of the corresponding year, which was computed by maximum likelihood (Table 3). The average functional form is rejected only for the year 1971, as shown in the first column of Table 5. Thus, it can be considered that O7093 adequately represents the pre-integration period. Table 5: Likelihood Ratio Test – Pre-integration Period Year LR1 LR2 Year LR1 1970 10.51 10.51 1982 2.22 1971 13.17* 13.50* 1983 1.61 1972 9.53 10.31 1984 3.33 1973 5.60 6.89 1985 7.67 1974 8.70 9.81 1986 1.28 1975 5.69 6.77 1987 1.91 1976 0.44 1.42 1988 4.51 1977 1.43 3.48 1989 8.99 1978 2.47 3.70 1990 8.78 1979 1.77 4.37 1991 1.77 1980 1.07 5.44 1992 7.15 1981 1.66 5.15 1993 7.45 Notes: * indicates that the null hypothesis is rejected at the 5 level.

LR2 3.05 2.18 3.93 9.15 2.17 3.23 5.07 10.17 10.90 4.14 8.98 9.86 per cent

The OLS estimates of the position parameters, taking O7093 as the functional form, produce the following results: The explanatory variables (per-capita GDP, population and distance) show the

238

Maria Luisa Recalde and Marcelo Florensa

expected sign, are significant for each year of the period 1970-1993 and the R 2 varies between 0.55 and 0.70. Nevertheless, since the Ei s are deduced from the same Box-Cox transformation, they show an evolution which is more predictable over time, so the Kalman filter can be applied. The Filter is initialised using the estimates vector corresponding to 1970, as bt 1 and the matrix of variances-covariances of these, as Pt 1 . The successive application of equations (8) and (9) allows to obtain, in a recursive way, the estimated values until 1993.7 Finally, the application of the prediction equation: (10) bt /1993 Cbt 1/1993 t 1994, 1995, , 2003 , Gives the predicted values of the Ei s for 1994-2003, which jointly with O7093 form the structure corresponding to the antimonde. To obtain the predicted values, which are shown in Table 6, only information for the period 1970-1993 has been used, therefore, at least from a conceptual point of view, they contain the hypothetical evolution of Argentina’s trade flows if MERCOSUR had not started. The projection of these E s, using data of the post-integration period, transformed with O7093 , permits to obtain the absolute values of the trade flows associated with the antimonde. Table 6: predicted E’s obtained from the Kalman Filter ȕ0 (C)

E94/93 E95/93 E96/93 E97/93 E98/93 E99/93 E00/93 E01/93 E02/93 E03/93 7

-177.34 -172.76 -168.29 -163.93 -159.69 -155.56 -151.53 -147.61 -143.79 -140.07

ȕ1 pcgdpi

23.17 22.72 22.28 21.85 21.42 21.01 20.60 20.20 19.81 19.42

ȕ2

ȕ3

pcgdpj

populationi

0.65 0.67 0.68 0.69 0.71 0.72 0.74 0.75 0.77 0.78

0.56 0.54 0.53 0.51 0.50 0.48 0.47 0.45 0.44 0.43

ȕ4

ȕ5

populationj Distance

1.88 1.84 1.79 1.74 1.70 1.65 1.61 1.57 1.53 1.49

-0.09 -0.09 -0.09 -0.09 -0.09 -0.08 -0.08 -0.08 -0.08 -0.08

ȕ6

ȕ7

lanMerco guage

5.16 4.33 3.63 3.05 2.56 2.15 1.80 1.51 1.27 1.06

5.34 5.00 4.69 4.40 4.12 3.87 3.63 3.40 3.19 2.99

For a discussion of the methodology to obtain the matrices C, Q and H that appear in equation 5 see Sanz and Gil (2001).

Mercosur: Trade Creation or Trade Diversion?

239

Nevertheless, before obtaining the values corresponding to the commercial impact of MERCOSUR, it is necessary to determine if the antimonde obtained from the filter is appropriate. The problem with the antimonde is that it is counterfactual (there is no data). However, if the O7093 with the Es predicted by the filter fit well to the data of the period 1970-1993, then, they can be applied for the period 1994-2003. This can be determined by a Likelihood Ratio Test, symbolised by LR2, for which the null hypothesis is composed by the Es predicted by the Kalman filter and O7093 , versus the alternative, for which the parameters of the Box-Cox transformation and the position parameters are estimated through maximum likelihood. The results, which are shown in Table 5, explain that the null hypothesis is rejected only in 1971. Hence, from 1970 to 1993, the dynamic estimation procedure represents appropriately the pre-integration structure and its projection to the post-integration period makes sense. A further contribution of the LR1 and LR2 tests is that, when applied to the period 1994-2003, they allow to identify the presence of a structural break. The results of these tests, which are shown in Table 7, show that the null hypothesis is rejected in 1994, 1995, 1997, 1998, 2002 and 2003 in the case of LR1 and in the period 1995-2003 for LR2; confirming, to some extent, what was obtained from the significance analysis of the dummy variable MERCO. Table 7: Likelihood Ratio Test. Post-integration period Year LR1 LR2 LR3 1994 15.94* 20.75 6.55 1995 16.59* 41.64* 2.88 1996 9.56 40.62* 3.04 1997 14.44* 53.82* 5.08 1998 15.83* 58.53* 4.84 1999 8.18 53.17* 1.90 2000 5.74 50.24* 0.78 2001 3.46 51.48* 0.86 2002 14.16* 51.46* 29.87* 2003 13.44* 46.11* 26.40* Notes: * indicates that the null hypothesis is rejected at the 5 per cent level.

Once a single structure in the pre-integration period and the presence of structural break are determined, the issue remains:

240

Maria Luisa Recalde and Marcelo Florensa

whether there exists more than one structure for the post-integration period. To perform this analysis, we obtain the average functional form for the post-integration period: O9403 = (0.178 -0.115 -0.374 0.146 -0.021 0.238) The likelihood ratio test LR3 with the null hypothesis O9403 and the optimal functional form for each year of the period 1994-2003 as the alternative is shown in the last column of Table 7. It shows that the null hypothesis is only rejected in 2002 and 2003. This indicates the existence of two structures for the post-integration period; one for the period 1994-2001 and another one since 2002. C. The Impact of MERCOSUR Once the structures of the monde and the antimonde have been estimated, which are different in statistical terms, we can obtain the trade flows under both structures and the values of the impact of the trade agreement. As the gravity equation includes in its specification variables like income, population, distance and dummies, “so the divergence between both structures may be considered as a reasonable approximation to the true impact”.8 Let mo~nit and ant~i monit be, respectively, the trade flow esti-

mates for the monde and antimonde in period t for the flow maintained with the ith country. The corresponding impact I it is given by: I = mo~n  ant~i mon t 1994, , 2003 it

it

it

i 1, ,108

A positive value of I it reveals trade creation with the country i in period t, whilst a negative value implies trade diversion. If trade creation takes place with a member country, we speak of internal creation; whereas it is called external creation if it takes place with a non-member country. In Table 8 we show the figures for the global impact, separating between impact on exports and on imports. First, it can be seen that the figures, on the import side, are increasing until 1998, then they decrease to a minimum in 2002, improving again in 2003. On the exports side, the values show an increase until 1998, they decrease afterward and remain constant

8

See Sanz (2000).

Mercosur: Trade Creation or Trade Diversion?

241

during 1999-2001, to experience an important growth in 20022003. Table 8: Effects of Mercosur on trade in manufactures (millions of US-$ in current prices) 1994 1995 1996 1997 Impact on Imports (1) Impact on Exports (2) (1) x 100/ current imports (2) x 100/ current exports

1998

1999

2000 2001 2002 2003

5488 8403 8454 12038 13866 10845 10465 8570 3714 4911 478 27.69

2106 3112 4904 46

5130

4160

4390 4221 5867 6496

39.43 43.86 48.25 46.25 45.92 46.98 46.73 39.97

5.19 17.23 23.45 30.93 32.58 29.89 28.55 27.42 39.26 38.95

The steady increase of exports as well as of imports, is a typical result of the impact models and reveals the cumulative character of any integration process. This is a consequence of both, the adoption of a common tariff and trade policy with foreign countries over several years. In contrast, the behaviour of imports and exports since 1999 may have another explanation. The impact is the difference between two estimated values, the first one is based on the observed data (monde) and the second is a forecast based on the data of the period 1970-1993 (antimonde). Therefore, while the first reflects the effects of economic policies and exogenous shocks in different countries, those do not effect the values obtained for the antimonde. Hence, the decrease of imports can be explained by the recession of Argentina by the end of 1998 and the exit from the currency board in 2002. In the same way, the marked improvement of the real exchange rate since 2002 can explain the substantial increase of industrial sector exports. Secondly, we see that there was an important trade creation both in imports and exports. However, the impact on imports has been greater than that on exports in absolute terms as in percentages; reaching an increase of almost 50 per cent in 1998 and remaining above 39 per cent in the period 1995-2003. Finally, if we disaggregate the values of the impact arranging the countries according to whether they are members of the MERCOSUR or not, the observations above are not significantly modified.

242

Maria Luisa Recalde and Marcelo Florensa

Table 9: Effects on manufacturing trade, disaggregation according to MERCOSUR membership (millions of US-$ in current prices) year

1994

imports exports

1117.03 1001.34

imports exports

4370.75 -523.84

year

1999

imports exports

2835.1 3242.14

imports exports

8009.79 917.781

1995 1996 MERCOSUR 1694.22 1826.57 1616.12 2361.85 Non-MERCOSUR 6708.68 6627.77 489.736 749.923 2000 2001 MERCOSUR 3275.03 2654.2 3699.98 3646.31 Non-MERCOSUR 7190.18 5915.68 690.455 574.322

1997

1998

3287.2 3439.21

3547.3 3835.33

8751.21 1465.16

10318.5 1294.98

2002

2003

1752.96 2372.56

1947 2983.99

1961.15 3494.1

2964.38 3512.32

From Table 9 we see that trade creation has been the exclusive effect, with the only exception of exports to non-member countries in 1994. Furthermore, the increase in imports from non-member countries is much higher; whereas of exports to member countries increased much more. V. Conclusions

The goal of this paper was to develop an econometric model that allows to assess the impact of Mercosur on the international trade of Argentina in manufacturing products. We follow the approach of Sanz (2000). Our main contribution consists in dynamising the gravity equation by using the Kalman filter. A first conclusion is, that the gravity model reasonably explains bilateral trade flows of the manufactured products of Argentina during the whole period 1970-2003. Per-capita income, population and distance are the most relevant explanatory variables. Language, however, is not significant. The results also indicate that MERCOSUR had an impact on bilateral trade flows of manufactured products, in particular since 1994. Therefore we consider 1994 as the first year of the post-integration period instead of 1991, the year when the Agreement of Asuncion was signed.

Mercosur: Trade Creation or Trade Diversion?

243

Models of trade effects traditionally use static antimondes, which implies to assume that the structure of the pre-integration period remains unchanged in the subsequent years. In contrast, using the Kalman filter permits us to use all the information of the pre-integration period to predict the coefficients of the gravity equation for the post-integration period (which begins in 1994). These coefficients differ year by year but since they are estimated from the behaviour and evolution of the pre-integration period, they constitute a continuance of it. The results also show that the structures of the monde and the antimonde are statistically different, although there is not a single structure for the post-integration period. Probably the change in the monetary-exchange regime of Argentina since 2002 is one cause. Trade creation, be it with the MERCOSUR or with countries outside dominate both on the import and export side. With exports, trade creation largely took place with MERCOSUR members, whereas imports from outside increased much more. These results are similar to those obtained by other authors, as for example Nagarajan (1998), Tigre et al. (1999), Hasenclever et al. (1999), Chudnovsky et al. (1999), and Bartholomew (2002), who applied different methodologies. However, most studies apply panel data techniques which do not assess the impact but only detect the existence of trade diversion or trade creation without quantifying them. This does not permit to compare the magnitudes that at a first glance would seem to be overestimated. We have to keep in mind that Argentina is a country that features great economic instability and that actions of monetary, exchange rate, tariff and deregulation policies, etc. implemented together with the Convertibility Plan of 1991 affected its international trade. Therefore it is difficult to state that the magnitude of the impact on trade flows can be exclusively attributed to the creation of the Mercosur. The use of the gravity equation, the Box-Cox transformation and the Kalman filter are useful statistic tools, appropriate to model Argentina’s international trade in manufactures. Future work should consider two further aspects. First, all the variables included in the gravity equation are exogenous to the integration process except for income. When we define the antimonde, the effect of income should be excluded and, in consequence, a specific treatment of the dynamic effects of integration on the income level of Argentina is be required.

244

Maria Luisa Recalde and Marcelo Florensa

Second, the information we used could be taken using the panel data technique to verify the existence (inexistence) of the impact although it would not be possible to quantify it. In that sense, the Kalman filter results have to be considered as the superior technique.

References

Aitken, Norman D. (1973), The Effect of the EEC and EFTA on European Trade: A Temporal Cross-Section Analysis, American Economic Review 63 (5), 881-92. Aitken, Norman D., Obutelewicz, Robert S. (1976), A CrossSectional Study of EEC Trade with the Association of African Countries, Review of Economics and Statistics 58 (4), 425-33. Anderson, James E. (1979), A Theoretical Foundation for the Gravity Equation, American Economic Review 69(1), 106-116. Bartholomew, Ann (2002), Trade Creation and Trede Diversion: The Welfare Impact of Mercosur on Argentina and Brazil, University of Oxford Centre for Brazilian Studies, Working Paper Series CBS-25. Bayoumi, Tamim, Eichengreen, Barry (1995), Is Regionalism Simply a Diversion? Evidence from the Evolution of the EC and EFTA, IMF Discussion Paper, no.109. Bergstrand, Jeffrey H. (1985), The Gravity Equation in International Trade: some Microeconomic Foundations and Empirical Evidence, Review of Economics and Statistics 67 (3), 474-481. Bergstrand, Jeffrey H. (1989), The Generalized Gravity Equation Monopolistic Competition, and the Factor-Proportions Theory in International Trade, Review of Economics and Statistics 71 (1), 143-153. Brada, Josef, Méndez, José A. (1985), Economic Integration among Developed, Developing and Centrally Planned Economies: A Comparative Analysis, Review of Economics and Statistics 67 (4), 549-56.

Mercosur: Trade Creation or Trade Diversion?

245

Carrère, Cèline (2004), Revisiting the effects of regional trade agreements on trade flows with proper specification of the gravity model, European Economic Review 50 (2), 223-247. Chudnovsky, Daniel, Erber, Fabio (1999), MERCOSUR's Impact on the development of the Machines Tools Sector, Integration and Trade 3 (7/8), Inter-American Development Bank, Washington D.C. De Groot, Henry L., Linders, Gert-Jan, Rietveld, Piet, Subramanian, Uma (2003), The Institutional Determinants of Bilateral Trade Patterns, Tinbergen Institute Discussion Paper, IT 2003044/3. Egger, Peter (1999), A note on the proper econometric specification of gravity equations, Economic Letters 66 (1), 25-31. Filippini, Carlo, Molini, Vasco (2003), The determinants of East Asian trade flows: a gravity equation approach, Journal of Asian Economics 14 (5), 695-711. IMF (2001), International Financial Statistics, version 11.54 in CDROM. Harvey, Andrew C. (1989), Forecasting, Structural Time Series Models and the Kalman Filter, Cambridge, Cambridge University Press. Havrylyshin, Oleh, Pritchett, Lant (1991), European Trade Patterns after the Transitions, World Bank Policy Research Working Paper Series no. 748. : Haveman, Jon, Base de datos de distancias, http://www.macalester.edu/research/economics/PAGE/HAVEM AN/Trade.Resources/TradeData.html, (last visit: 20.09.2007).

Hasenclever, Lia, Lopez, Andrès, Clemente de Oliveira, Josè (1999), The Impact of MERCOSUR on the Development of the Petrochemical Sector, Integration and Trade 3 (7/8), InterAmerican Development Bank. Kamil, Herman, Ons, Alvaro (2001), Los flujos de comercio de los países del MERCOSUR en los noventa: el rol de las preferencias comerciales intrabloque, Mimeo. Martinez-Zarzoso, Immaculada, Nowak-Lehman, Felicitas (2002), Explaining MERCOSUR sectoral exports to the EU: The role of

246

Maria Luisa Recalde and Marcelo Florensa

economic and geographical distance, Ibero America Institute for Economic Research no 85. Martinez-Zarzoso, Immaculada, Nowak-Lehman, Felicitas (2003), Augmented Gravity Model: An Empirical Aplication to Mercosur-European Union Trade Flows, Journal of Applied Economics 6 (2), 291-316. Mayes, David G. (1978), The Effects of Economic Integration on Trade, Journal of Common Market Studies 17 (1), 97-121. Nagarajan, Nigel (1998), La Evidencia sobre el Desvío de comercio en el Mercosur, Integración y Comercio no 6, INTAL, IADB. Pelzman, Joseph (1977), Trade Creation and Trade Diversion: in the Council of Mutual Economic Assistance: 1954-70, American Economic Review 67 (4), 713-22. Recalde, Maria Luisa, Florensa, Marcelo (2007), Effects of Mercosur on the Argentine Agricultural International Trade: an Application of a Gravity Equation, in: Eisen, Roland, Díaz Cafferata, Alberto M., Neder Ángel Enrique, Recalde María Luisa (eds.), Trade, Integration and Institutional Reforms in Latin America and EU, Papers and Proceedings of the Arnoldshain VI Seminar Cordoba,: Peter Lang, Frankfurt. Rose, Andrew, Engel, Charles (2000), Currency Unions and International Integration, NBER Working Paper 7872. Sanso, Marcos B., Cuairán, Rogelio, Sanz, Fernando (1993), Bilaterial Trade Flows, the Gravity Equation, and Functional Form, Review of Economics and Statistics 75 (2), 266-275. Sanz, Fernando, M. Gil, José (2001), An Assessment of the Agricultural Trade Impact of Spain´s Integration into the EU, Canadian Journal of Agricultural Economics 49(1), 53-69. Sanz, Fernando (2000), A Kalman Filter-Gravity Equation Approach to Assess the Trade Impact of Economic Integration: the Case of Spain, 1986-1992, Weltwirtschaftliches Archiv 136 (1), 84-109. Sapir, André (1981), Trade Benefits under the EEC Generalized System of Preferences, European Economic Review 15 (3), 33955.

Mercosur: Trade Creation or Trade Diversion?

247

Tigre, Paulo Laplane, Lugones, Mariano, Porta, Fernando Gustavo (1999), Technological Change and Modernization in the MERCOSUR Automotive Industry, Integration and Trade 3 (7/8), Inter-American Development Bank, Washington D.C.

Appendix

The Box-Cox transformation for a variable X (in our case M ij , yi , y j , Li , L j , Dij ) is defined as ­° ( X Or 1) / Or if Or z 0 (A1) r 0,1....,5 ® log X Or 0 if °¯ Different values of Or will define different functional forms.

X (Or )

In equation (A2), not only the S n ( n 0,1,..., 7 ) but also the Or must be estimated. Precisely, the values of these define the optimal functional form. The estimation of the parameters of the gravity equation and the functional form must be solved simultaneously. This can be achieved applying the maximum likelihood method. Assume, in general terms, that there is a cross section constituted by N observations for each of the T available years and the observations of the different years are independent. In this case, each of the years can be estimated separately. The way to proceed is as follows (where subscript t 1,..., T has been omitted for the sake of simplicity). Let y (O ) be an Nx1 observations vector of the dependent variables, for a certain year, and X (O ) an N x k observations matrix of the exogenous variables for the same year (in this case k 8 ). If the Nx1 error terms vector is represented with U, we can write:

y ( O ) X ( O )S  P P ~ N (0, V 2 I ) (A2) where S is a S n vector of parameters. The likelihood function for the sample is, under the specified assumptions

248

Maria Luisa Recalde and Marcelo Florensa





§ (O ) O t O O ¨ y X S y X S L(S , O; y, X ) constant- (N/2)1V 2  ¨ 2 ¨ 2V ©

·¸  O0 1 ¦N log yi ¸ ¸ ¹

(A3)

i 1

If the O vector is known, the usual methods of estimation work (with the variables appropriately transformed). If it is unknown but all of its elements are equal to OC , it would be necessary to choose

OˆC in order to maximise the support function. The search for the optimum value of OC can be done, because the logarithm of the concentrated likelihood function depends only on OC and it is enough to go through the search algorithm to estimate such value. For the case in which each Or is different, the logarithm of the likelihood function generalised for each value of O results to be L (O )

N

(A4)

constant  ( N 2) ln V 2 (O )  (O0  1)¦ ln yi i 1

where V~ 2 (O ) is the maximum-likelihood estimator of V 2 . In this case, the application of a search procedure is not possible. Nevertheless it is possible to apply the non-linear least squares. The algorithm consists in taking O0 as a guide parameter in the optimisation process. Once O0 is determined, the other parameters are estimated using non-linear least squares. In this way, for each value of O0 , the estimators of the parameters are obtained with the purpose of giving the support function its maximum value, since the likelihood finally depends on this parameter. Thus, it is enough to select the value of O0 (symbolised with O ) that maximises 0

L (O0 )

N

(A5)

constant  ( N 2) ln V 2 (O0 )  (O0  1)¦ ln yi i 1

with ª y ( O0 )  X ( O ( O0 ))S ( O ) T y ( O0 )  X ( O ( O0 ))S ( O ) º N (A6) 0 0 «¬ »¼ where O (O0 ) and S (O0 ) are the vectors that maximise the support function for the given value of O . Once O has been found, 2

V ( O 0 )







0

0

the remaining maximum-likelihood estimates, O (O0 ) S (O0 ) and V 2 (O ) are directly obtained. 0

Rinaldo Antonio Colomé and Fernando M. Giuliano

Agricultural Policies and Trade in the European Union and Selected Latin American Countries Abstract The objective of this paper is to analyze agricultural policies of selected countries: A high subsidizing group (EU), a moderate subsidizing country (Mexico), one with almost no agricultural support (Brazil), and one with high negative support (Argentina). Nevertheless, the way support is implemented is very important as well. Support in Brazil is very low, though distortive. Support in the EU and Mexico is not only distortive, but spread among different commodities. This fact has negative policy implications: further decreases in absolute and distortive support requires reforms in policy measures affecting several products, having to deal with resistance from different interest groups. I. Introduction Agricultural policies in the world have started to change since the Uruguay Round of the General Agreement on Trade and Tariffs (GATT) from highly distortive support measures to less distortive ones. The Uruguay Round was the sixth GATT round of negotiations, but the first to include agricultural trade. It was the longest GATT round in getting to an ending, due to the difficulties in achieving an agreement on agricultural and trade policies. Argentina and any other agricultural exporting country -that could be considered a “small country” - suffers the consequences of the agricultural and trade policies of developed countries, like the European Union (EU), and the United States of America (USA). For these reasons, Argentina and other “efficient” countries have constituted the “Cairns Group”. This Group started acting preparing the documents to maintain an agricultural and trade proposal to be formally discussed, for the first time, on GATT’s Uruguay Round. This proposal claimed for international agricultural policies with lower subsidy levels and more liberalized agricultural markets. This Group is still active within the World Trade Organization (WTO).

250

Rinaldo Antonio Colomé and Fernando M. Giuliano

The main objectives of this paper are to analyze the agricultural policies of the European Union (EU) –the Common Agricultural Policy (CAP)- in comparison with those of Argentina, Brazil, and Mexico, and how they have changed, especially after the Uruguay Round. To be more precise, the objectives are to analyze agricultural support levels in the European Union and in these other countries, and the instruments used -classifying them according to their degree of distortion- in order to examine the possibilities to achieve less distortive policies. For that purpose, the next section gives a historical summary review of agricultural and trade policies in the European Union and Argentina. Since the history of the CAP is well known, emphasis is posed on Argentina. The specific instruments of the agricultural policies of the EU, Mexico and Brazil will appear during the analysis. In section III the methodology is explained, and the analysis is carried out. In section IV main conclusions are presented. II. A historical summary review of agricultural and trade policies in the European Union and Argentina Argentina and the European Union show a quite different history concerning agricultural policies. While the EU is known as one of the largest subsidizing group of countries in the world, Argentina is probably the one which greater detraction applied on agricultural producers income. The history and current situation of the CAP is well known through a great number of official documents published by the EU, the OECD, books, and papers submitted in academic meetings. For this reason it will not be discussed in detail here, and only the necessary features for the purpose of this paper will be mentioned. One significant question is to remember that the CAP was established in Stresa, in 1958. As its name suggests, the Common Agricultural Policy substituted prior national agricultural policies which, according with Hitiris and Vallés (1999), were of a heavy intervention in each country of the original members of the European Economic Community (EEC). That is to say, prior to the “Common” Agricultural Policy there were agricultural policies in each of these European countries. Another important question to be mentioned is that agricultural policy in the EU (also in other developed countries in the world) have started to change since the GATT’s Uruguay

Agricultural Policies and Trade in the European Union and Selected Latin American Countries

251

Round from highly distortive support measures to less distortive ones. Argentina started its agricultural policy in November 1933, following the United States of America, which had started in 1932 with the “Agricultural Marketing Act”, which created the “Federal Farm Board”. However, a year later this Act was substituted by the “Agricultural Adjustment Act”. This Act established the policies which -“mutatis mutandi”- have continued until now -that is to say, for more than seventy years- and which were the model for Argentina. International prices for the principal agricultural commodities started to go down in the beginnings of the thirties, and were down dramatically in 1933. Under these circumstances the Argentine government decided, in November 1933, to support agricultural producers through two main measures: (i) To fix market price support (MPS) for the three main commodities produced in Argentina at that time: wheat, corn, and linseed; (ii) To control the exchange rate. To accomplish the first one, the National Grains Board (NGB) was founded; for the second one, the Exchange Rate Control Office was created. To act in defense of the livestock production, the National Meats Board was founded, and in 1934 –as main instrument of this law – the Argentine’s Meat Producers Corporation was established. The National Grains Board had the obligation to buy all grains offered by producers. Producers had the faculty to sell in the free market (if prices were over the MPS), or to the NGB in the case the market price for each grain were lower. The Bank of the Argentine Nation (Banco de la Nación Argentina) was the financial agent of the NGB. Due to the fact that the international market started to recover after the crisis, the NGB had to act only during the years 1933, 1934 and 1935. However, the NGB was preserved (just in case of new bad times). Bad times came again in 1938, and MPS was established for the same commodities, and the NGB started to act again. It occurred for the whole time of World War II. With respect to the second measure, the Exchange Rate Control Office was entitled with the faculty to buy and sell all the foreign currencies, in order to match the demand with the supply. As selling prices of the foreign currencies were (according with the demand and supply) higher than the buying ones, these differences were used to constitute the Exchange Rate Fund. This Fund ought to be

252

Rinaldo Antonio Colomé and Fernando M. Giuliano

used to pay the differences between the MPS and the market prices for each grain. In short, these two measures – MPS for these grains, and the exchange rate control (the exchange rate policy) – acted in favor of agricultural producers in the short and medium term, in spite of the fact that the “differences of exchange” between the buying and selling rate returned only partially in benefit to the producers, due to the rapid recovery of prices. It is very important to notice that from that time on, this last instrument (the exchange rate policy) was used to detract income from agricultural producers, we shall see later. As in 1946 an economic model characterized as a state managed and closed economy was established in Argentina, a state agency to run the foreign agricultural trade was founded. It was named as Instituto Argentino de Promoción del Intercambio -IAPI- (Argentine Institute to Promote the Inter-exchange). This agency started to be the official monopoly in buying and exporting the whole of agricultural products including, naturally, all grains and by products. Its financial agent were the Bank of the Argentine Nation, and the Bank of Industrial Credit (Banco de Crédito Industrial). But the most important question concerning agricultural policy is that the IAPI was entitled with the faculty to fix the “official price” for the main agricultural commodities (wheat, corn, and linseed). Notice that this official price was not a support price, but a fixed discretional price under the value of the international market price. The prices of these commodities were dramatically lower than the international market prices for each grain and -according to Díaz Alejandro (1975), Colomé (1966), and Colomé (1976) - it was the principal cause for the agricultural sector stagnation. This model lasted for ten years. By the end of 1955, the new government –with a market oriented economic philosophy- started to establish new economic rules (which were, indeed, the rules and institutions previous to this ten years period). As a consequence, the IAPI was dissolved, the NGB and the NMB were reestablished, as well as the market price support (MPS) for the principal grains. As the argentine peso was devaluated by the end of 1956, an export tax was created to be applied to main agricultural products (main grains and beef) at the moment to be exported, with the main purpose to maintain domestic prices lower than the international ones. This export tax is known as “Retentions” (“Retenciones”), since it is applied as a percentage of

Agricultural Policies and Trade in the European Union and Selected Latin American Countries

253

agricultural FOB prices. This tax was assumed to be a temporary measure, that should last until the equilibrium of the real exchange rate was achieved again. However, this tax was used “off and on” any time a devaluation of the peso occurred. So, since then, this “trade policy” (this export tax is qualified in this way), is one of the two basic instruments of the Argentine agricultural policy. The other one – as has been said above – is the “exchange rate policy”. These two instruments were used –with very short exceptional periods- until now to detract agricultural incomes to benefit the rest of the economy. The effects of these two instruments were studied by Reca (1980) and, especially, by Sturzenegger and Martínez Mosquera (1986), who found that the average of the total nominal negative protection rate was -36.26 per cent in the period 1976-78, and -52.20 per cent in the period 1979-81. From the second semester of 1989 on, a rapid process of elimination of the “Retentions” (export taxes) started again, and they were totally eliminated at the beginning of the nineties. On April 1st, 1990 the so called “convertibility law” was passed. It established convertibility of one peso with one US-Dollar, eliminating the possible discrimination to agriculture through the exchange rate. Notice that with these two measures the two instruments of discrimination of the agricultural sector: exchange rate control (the indirect discrimination), and “retentions” (the direct discrimination) were eliminated. However, diverse protection policies in favor of several manufacture industries remain, which still determine a negative protection for agriculture of about 7 per cent (Medina and Soto, 2000). At the beginning of the year 2002 –after the 2001 crisis- the convertibility law was derogated. The rate of exchange jumped first to four pesos for one dollar, getting shortly a stabilized price of three pesos for one dollar. As a consequence, the export taxes (“retentions”) were established again. So, the agricultural policy was reestablished with one of the classic instruments of negative protection. It is a paradox that Argentina requests the reduction of agricultural subsidies of the developed countries in international organizations –like the WTO- while –at the same time- applying a heavy domestic negative protection. III. Evaluation of the effects of agricultural and trade policies In this section agricultural and trade policies of the European Union are analyzed, comparing them with those of other Organization for

254

Rinaldo Antonio Colomé and Fernando M. Giuliano

Economic Co-operation and Development (OECD) countries (which include Mexico) and Brazil, focusing on their effects on agricultural markets. A. Methodology To compare their impact we use the so-called “PSE’s”. “The use of the Producer Subsidy Equivalent (PSE) method to estimate assistance to agriculture was initially developed by Professor Tim Josling for the Food and Agriculture Organization of the UN in the early 1970’s, although the theoretical basis may be found in the work of, in particular, Max Corden. It was adopted by the OECD in implementing the 1982 Ministerial Trade Mandate.” (Cahill and Legg, 1989-90) The method estimates the monetary value of transfers associated with all policy measures affecting agriculture, grouped into four main categories: (i) Market Price Support, (ii) Direct Payments, (iii) Reduction of Input Costs, and (iv) General Services. The “subsidy equivalent” was initially defined as “the monetary value that would be required to compensate farmers or consumers for the loss of income resulting from the removal of a given policy measure”. However, the current OECD indicator corresponds to a broader definition. It measures more than the “subsidy element”, since it includes implicit as well as explicit payments. Therefore, in order to make the names of the indicators reflect as closely as possible the underlining definitions and to make them consistent with one another, OECD countries agreed to replace “subsidy equivalent” by “support estimate”.1 Thus the abbreviation PSE now stands for “Producer Support Estimate”. More precisely, “Producer Support Estimate” is “an indicator of the annual monetary value of gross transfers from consumers and taxpayers to agricultural producers, measured at the farm-gate level, arising from policy measures that support agriculture, regardless of their nature, objectives or impacts on farm production or income” (OECD, 2005a). The main components of the PSE are: Market Price Support (MPS) Payments based on output (PBO) Payments based on area planted/animal numbers (PBAP) Payments based on historical entitlements (PBHE) Payments based on input use (PBIU) 1

See OECD (2005a).

Agricultural Policies and Trade in the European Union and Selected Latin American Countries

255

Payments based on input constraints (PBIC) Payments based on overall farming income (PBOFI) Miscellaneous payments2 (MP) Under GATT´s Uruguay Round Agreement on Agriculture,3 the OECD is responsible for estimating agricultural support for trade negotiations. The OECD measures and publishes the PSE in this context. As it is usual with an aggregated indicator such as the PSE, this measure has been subject to criticism in economic literature, and more recently in the political debate over world trade. The three central questions raised by the critics are:4 i) The PSE does not reflect properly changes in agricultural policies and reform efforts; ii) World market conditions distort the PSE; iii) Actual world market prices are not a proper reference point for estimating the PSE. These three issues are addressed as factors that in one way or another make the PSE deliver misleading information.5 In an extensive, yet simple manner, Tangermann rejects these critics. He basically argues that the PSE is a good indicator to monitor the nature and evolution of agricultural policies, though stressing the importance of not only analyzing its level, but also its composition. This paper tries to follow this recommendation. For cross country comparisons, percentage PSE (per cent PSE) is used. It is “the PSE expressed as the ratio of the value of total

2

3

4 5

For more detailed information regarding this concepts, see OECD (2005a). The Uruguay Round was the eighth round of GATT formal multilateral trade negotiations. It started in Punta del Este, Uruguay, on September 1986. The Final Act of the Uruguay Round, as well as the Agreement establishing the World Trade Organization, incorporating all detailed results of the negotiation, was signed at Marrakech, Morocco, in April 1994 (see Sumner and Tangermann, 1999). For an overview, see Tangermann (2005). These are not the only criticisms. Interesting approaches over the limitations of the PSE for reflecting true support to producers in transition economies (former communist republics), can be found in Strokov and Meyers (1996).

256

Rinaldo Antonio Colomé and Fernando M. Giuliano

gross farm receipts,6 measured by the value of total production (at farm-gate prices), plus budgetary support”(OECD, 2005a). Again, the interpretation of this index should not rest only upon its level, but also (and mainly) on its composition. The level of producer support can also be measured by the producer Nominal Assistance Coefficient (NAC), which expresses “the monetary value of total gross farm receipts (including support), relative to production valued at world market prices without support” (OECD, 2005a). Another useful indicator is the producer Nominal Protection Coefficient (NPC). It “measures the ratio between the average price received by producers (at farm gate) and the border price (at farm gate) serving as an indicator of the most distortive way of support (eg. price support)”. The OECD estimates producer NAC for each product and as an aggregate for each country, whereas NPC is estimated on commodity basis. In what follows two different PSE databases are used. The OECD publishes PSE for OECD countries, taking the European Union as a whole, and more recently for Brazil, China, and South Africa as well. Hence, no figures for Argentina are available. However, USDA (1995) published figures for Argentina during the eighties and beginnings of the nineties for “Producer Subsidy Equivalent”, the predecessor of Producer Support Estimate. Those estimations are helpful for the paper’s objective, thus will not be dismissed from the analysis. B. Effects of Agricultural Support Measures To gain insight on the magnitude of agricultural support measures around the world, it is worth noting that around 279 billion U$dollars were spent by OECD countries7 in 2004 for that purpose. That amount almost doubles Argentina’s GDP of that year. It should not be surprising then, that the primary sector represents the most conflict area of free trade negotiations within the World Trade Organization (WTO). The discussions about agricultural subsidies have been the most important cause of the recent failure of Doha´s 6

7

Gross farm receipts are not the same as farm income, which is farm receipts less farm costs. Australia, Austria, Belgium, Canada, the Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Japan, Korea, Luxembourg, Mexico, New Zealand, the Netherlands, Norway, Poland, Portugal, the Slovak Republic, Spain, Sweden, Switzerland, Turkey, the United Kingdom and the United States.

Agricultural Policies and Trade in the European Union and Selected Latin American Countries

257

Round of Negotiations. Studying agricultural support measures is particularly relevant in this context. Figure 1 shows a comparison of the evolution of percentage PSE for OECD members, since 1986.

1986-88

Norway

Switzerland

Korea

Iceland

Hungary European Union Japan

Canada Slovak Republic Turkey Czech Republic OECD

Poland United States Mexico

Australia

80 60 40 20 0

New Zealand Brazil

%

Figure 1: per cent PSE*

2002-04

*Series for Brazil start in 1995. Average values for the 1986-1988 and 2002-2004 sub periods are shown for OECD members (showing the European Union as a whole), in an ascending way based on the most recent sub period from left to right. It can be seen that OECD’s average of per cent PSE was about 30 per cent in 2002-04, 7 points below its figures in the first sub period. All countries decreased its per cent PSE, except for Brazil, Mexico, and Turkey. In the last sub period, countries showing the least agricultural support were New Zealand, Brazil, and Australia, and the ones with highest levels of support were Switzerland and Norway. The position of New Zealand, Brazil, and Australia should not be surprising, due to their membership in the Cairns Group.8 Moreover, New Zealand and Australia not only show a low per cent PSE (3 per cent and 4 per cent respectively), but they have also shown a remarkable decrease in this index since the first sub period. New Zealand’s per cent PSE reduction from 11 per cent in 1986-1988 to 3 per cent in 2002-04 represents the 8

A group formed in 1986 in Cairns, Australia, that seeks the removal of trade barriers and substantial reductions in subsidies affecting agricultural trade. The group includes Argentina, Australia, Bolivia, Brazil, Canada, Chile, Colombia, Costa Rica, Guatemala, Indonesia, Malaysia, New Zealand, Pakistan, Paraguay, the Philippines, Thailand, South Africa, and Uruguay. The Cairns Group was a strong coalition in the Uruguay Round of multilateral trade negotiations.

258

Rinaldo Antonio Colomé and Fernando M. Giuliano

greatest percentage reduction of all the analyzed countries (an 80 per cent reduction). Mexico was the country that increased its per cent PSE the most, going from 3 per cent in the first sub period to a 21 per cent in 2002-04 (although it is worth mentioning that Mexico had the lowest per cent PSE in 1986-88). As it has been said, the other country that increased its per cent PSE in the second sub period was Turkey, a country willing to join the European Union. The low support levels that Australia and New Zealand show contrast with high support levels in Iceland, Switzerland and Norway, to name a few. All of them present a per cent PSE of about 71 per cent in the last sub period, although only Norway has not decreased it since 1986-88. The unavailability of data for Argentina makes it impossible to compare current per cent PSE values for OECD countries with those of Argentina. However, estimates of “Producer Subsidy Equivalent” by the United States Department of Agriculture for 1982-1992, place Argentina’s Percentage Producer Subsidy Equivalent in an average of –34 per cent. This negative value is a sign of negative support for Argentina’s agricultural sector, which contrasts with the support enjoyed by farmers in OECD countries, and also with its neighbour and Mercosur partner, Brazil. For the motives already discussed above, the analysis would be incomplete if it were a mere description of per cent PSE in different countries. To go deeper, Figure 2 shows the evolution of NAC for the same countries, through the same period. Again, countries are displayed in an ascending way based on the most recent sub period from left to right. Producer NAC can be seen as a sign of market orientation, that is, the degree of influence of market signals on the orientation of agricultural production. As in Figure 1, New Zealand, Brazil and Australia rank first. In these countries gross farm receipts come from the market almost without any support.

Agricultural Policies and Trade in the European Union and Selected Latin American Countries

259

Figure 2: Producer NAC

N EW

ZE A. .. AU BR ST AZ R IL PO ALI LA A N D M US EX A C IC A O SL NA O DA VA TU KI R A K C EY H ZE U C N H G AR Y E O U EC JA D P KO AN I SW C RE E A IT LA ZE N R D N ... O R W AY

5 4 3 2 1 0

86-88

2002-2004

In the European Union, receipts are on average 52 per cent higher than what they would be if valued at international prices (a 1.52 NAC). It is a high percentage, but lower than the average for the 1986-89 period (1.71). The highest NAC values are those for Norway, where gross farm receipts, are on average three and a half times higher than what they would be valued at international prices. Not only that, but Norway is one of the countries – together with Mexico and Turkey – were NAC has increased. Mexico’s current NAC is 1.26, higher than its average NAC in the sub period 1986-88 (1.03) C. A closer look on the PSE for Brazil, the EU, and Mexico Let’s analyze more in detail the per cent PSE for Brazil, the European Union, and Mexico. Figure 3 shows the behavior of the series for these countries. Figure 3: Development of per cent PSE 50 40

%

30 20 10 0 -10 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 EU MEXICO BRAZIL

Brazil is, by far, the country that subsidizes its agriculture the least. Since 1995 – first year of available data – Brazil’s per cent PSE has fluctuated between – 1 per cent (1995) and 6 per cent (1998), and has been fairly stable since 2000. European Union’s per

260

Rinaldo Antonio Colomé and Fernando M. Giuliano

cent PSE reached a minimum in 1989 (29 per cent), and has fluctuated between that value and 40 per cent since then. Mexico shows a more volatile behavior. Its per cent PSE moved between –5 per cent in 1995 (Tequila crisis) and 30 per cent on 1993. Although it shows higher values in 2004 than in the beginnings of the series, it has declined since 1994 (Marrakech agreement). But, as said before, PSE’s composition should be addressed to find whether significant policy reform efforts have been made. The impacts of a policy measure on production and trade depend on both the degree to which extra resources are attracted to produce a commodity, and the degree it affects consumption of that commodity. In general, the more a policy measure provides specific support to a commodity, the greater the impact on production and trade of that commodity, although restrictions or constraints on providing support will generally limit these impacts. All other things being equal, the main categories of PSE can be ranked according to their relative impact on production and trade as follows (OECD): (i) Market Price Support, (ii) Payments based on outputs, (iii) Payments based on input use, (iv) Payments based on area planted/animal numbers, (v) Payments based on historical entitlements, (vi) Payments based on input constraints, (vii) Payments based on overall farming income. Figure 4 displays the combined share of the three most distortive support measures in total PSE. Brazil has very low support levels (Figure 3), but those low levels of support are sustained by heavily distortive support measures. All of its support is explained by the three most distortive categories of PSE. Mexico’s support measures were 100 per cent distortive as well before the 1995 crisis, but its share has decreased, representing in 2004 a 74 per cent of total PSE. The European Union also showed 100 per cent distortive support measures in the mid eighties, but they have been decreasing steadily, representing 64 per cent of total PSE in 2004. That is, the European Union shows both diminishing % PSE levels and a shift towards less distortive support measures, implying a noticeable effort of policy reform.

Agricultural Policies and Trade in the European Union and Selected Latin American Countries

261

Figure 4: Share of the Most Distortive Support Measures in Total PSE 200% 150% 100% 50% 0% 1986

1988

1990

1992

1994

BRAZIL

1996 EU

1998

2000

2002

2004

MEXICO

Figure 5 reinforces the results shown in Figure 4. It displays the combined share of the two less distortive support measures in total PSE, showing the European Union a larger share of this support measures, but still at low levels (5 per cent). These “healthy” measures are absent from Brazilian and Mexican policies. Figure 5: Share of the Less Distortive Support Measures on Total PSE 6% 5%

%

4% 3% 2% 1% 0% 1986

1989

1992

Brazil

1995

1998

EU

Mexico

2001

2004

D. PSE values for Brazil, the EU, and Mexico by commodities So far we have exposed aggregated PSE values for Brazil, the EU, and Mexico. The OECD also estimates disaggregated PSE values, analyzing different agricultural products for each country.9 An 9

The analyzed products are: wheat, maize, sorghum, milk, eggs, other grains, poultry, beef and veal, sheep meta, barley, soybeans, oilseeds, pigmeat, sugar, rice, wool, oats, potato, tomato, wine, plants and flowers, apple, cabbage, cucumber, grapes, tangerines, pears, spinach, strawberry, onions, garlic, peppers, coffee, rye, cotton, and tobacco.

262

Rinaldo Antonio Colomé and Fernando M. Giuliano

analysis that considers differences between products within each country would be important. 1. Brazil Two three-year average sub-periods are compared: 1995-97 and 2002-04. Rice is the commodity with the highest per cent PSE in Brazil in both sub-periods, reaching a 17 per cent average in 200204. The other commodity exceeding a 10 per cent PSE is cotton (12 per cent in 2002-04), as Figure 6 shows. Figure 6: Brazil-Commodity per cent PSE Pigmeat Poultry Sugar Cane Beef Milk Soybeans Coffee Maize Wheat Cotton Rice -50

-40

-30

-20

-10

0

10

20

% 1995-1997 2002-04

It is worth noting the –48 per cent PSE value for sugar cane in the earlier period, due to a strong negative market price support, support that has nowadays disappeared, enjoying sugar cane producers a low – but positive – support that allows Brazil to be a worldwide leader in bioethanol production. It is also worth to notice the great decrease of per cent PSE for milk. Rice is not only the commodity with the highest per cent PSE in Brazil, but it also has the highest PSE in absolute terms. Besides that, rice protection is of the most harmful kind: market price support (82 per cent).

Agricultural Policies and Trade in the European Union and Selected Latin American Countries

263

Figure 7: %razil: Rice PSE Composition (2004) 

B. Payments based on output 1%

E. Payments based on input use 17%

A. Market pric e support 82%

Being rice the product with the highest percentage and absolute PSE in Brazil, and being that its support is highly distortive, we wonder how distortive support would have decrease in Brazil in the hypothetical case where rice had no support at all. That is, to analyze how Brazilian PSE composition changes when rice is excluded from PSE calculations. Given that all of Brazil’s agricultural policies are either MPS, PBO, or PBIU, the combined share of these three support measures in total PSE would not change if rice producers received not support at all. However, given that rice support is highly concentrated in MPS measures, the removal of rice from PSE estimations decreases the importance of MPS (the most distortive support) in total PSE. Figure 8 shows the results obtained from this exercise for the last five years.10 The decrease is more noticeable since 2002. The share of MPS in total PSE drops in 2004 from 28 per cent to 10 per cent when rice is not taken into account in PSE estimations. Actually, the increase in the participation of MPS in 2004 is totally explained by rice’s MPS: if rice were not supported at all, the share of MPS would have diminished in 2004, instead of increasing, as it actually did that year.

10

Years when both PSE and MPS had positive values.

264

Rinaldo Antonio Colomé and Fernando M. Giuliano

Figure 8: Brazil: Share of MPS in total PSE 40% 20% 0% 2000

2001

2002

WithRice

2003

2004

Without Rice

2. Mexico As noted earlier, Mexico has higher levels of support than Brazil, and that can be also seen on a commodity basis. The differences are impressive: we said that rice and cotton were the only two commodities showing per cent PSE above 10 per cent in the case of Brazil; for Mexico, only eggs and beef show per cent PSE below 10 per cent (Figure 9). Commodities with the highest per cent PSE in Mexico are: oilseeds (48,2 per cent), sugar (41,5 per cent) and maize (34,7 per cent). All other products show values over 20 per cent, except for pigmeat. Figure 9: Mexico - Commodity per cent PSE Eggs Beef Pigmeat Poultry Barley Soybeans Wheat Other Grains Rice Milk Sorghum Maize Sugar Oilseeds -60,0

-40,0

-20,0

0,0

%

1986-88

20,0

40,0

60,0

80,0

2002-04

It is tempting to perform a similar experiment for Mexico than the one performed for Brazil, that is, identify the commodity whose removal from PSE estimations diminishes distortive support the most. Milk is the chosen product. It used to have the highest per cent PSE in the first sub-period, but not in the second, showing a remarkable reduction (from 67 per cent to 31,2 per cent). Although

Agricultural Policies and Trade in the European Union and Selected Latin American Countries

265

it does not have in 2002-04 the highest per cent PSE, it represents 13 per cent of total PSE, following maize with 18 per cent. However – unlike maize – milk’s support is heavily distortive: Figure 10: Mexico: Milk PSE Composition (2004) C . P a ym en ts ba se d o n a re a p lan te d /a n im al n u m be rs 1%

E . P aym e n ts b ase d o n in pu t u se 8%

A . Ma rke t p rice sup p o rt 9 1%

Not only MPS represents 91 per cent of milk’s PSE, but PBIU represents an extra 8 per cent. That is, 99 per cent of milk’s support is highly distortive. There’s nothing left but to analyze the composition of Mexican PSE with and without milk, exercise shown in Figure 11. Figure 11: Mexico: Share of Most Distortive Support in Total PSE 120% 100% 80% 60% 40%

With Milk

2004

2003

2002

2001

2000

1999

1998

1997

1996

1994

1993

1992

1991

1990

1989

1988

1987

1986

20%

Without Milk

Data for 1995 were left out of the analysis due to the distortions in the estimations caused by the macroeconomic fluctuations during the Tequila crisis. The removal of milk from PSE estimations does not alter significantly the share of the three most distortive support measures in total PSE, only 2 to 4 points below the regular series for (2002-04).

266

Rinaldo Antonio Colomé and Fernando M. Giuliano

3. European Union As has been shown in Figure 3, the European Union shows a higher per cent PSE than Mexico and Brazil for the whole period. Analyzing now the per cent PSE for each product, it can be seen that beef, durum wheat, and sugar show per cent PSE over 60 per cent. Beef is the product with the highest protection, over 70 per cent, followed by durum wheat (over 60 per cent), sugar and oats (60 per cent). It is also worth to note that these are the only products that have increased their per cent PSE besides poultry, and pigmeat. At the other end, eggs have almost no protection: 2 per cent. Figure 12: EU: Commodity per cent PSE Eggs Pigmeat Rice Rapeseed Oilseeds Sunflower Common Maize Poultry Milk Soybeans Wheat Coarse Barley Other Sheep Oats Sugar Durum Beef 0

10

20

30

40 % 1986-88

50

60

70

80

2002-04

For the European Union, although milk ranks eleventh according to its per cent PSE, it is the product whose support distorts agricultural markets the most. It represents 12 per cent of total PSE, and its composition looks as follows: Figure 13: EU: Milk PSE Composition (2004) p E. Payments based on input use 10% D. Payments based on historical entitlements 4%

(

)

Others 5%

A. Market price support 81%

Agricultural Policies and Trade in the European Union and Selected Latin American Countries

267

Thus milk, as in the case of Mexico, will be the product excluded from PSE in order to see how distortive support to this product is. The results are shown on Figure 14. Figure 14: EU: Share of the Most Distortive in total PSE 100% 90% 80% 70% 60% 50% 1986

1988

1990

1992

1994 1996 years

Without Milk

1998

2000

2002

2004

With Milk

Distortive support in the European Union slightly decreases when milk is not taken into account in PSE. The average in the last years for the series without milk is 66 per cent, only 8 per cent below the original series, which show an average value of 61 per cent for the last years. Finally, Figure 15 shows the results of removing the most distortively supported commodity from PSE for the European Union, Brazil, and Mexico. As said before, products excluded are rice for Brazil, and milk for both Mexico and the European Union. Figure 15: Share of the Most Distortive Support Measures in Total PSE Exluding the Most Distortively Supported Commodity for Each Country 200% 150% 100% 50% 0% 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 EU

Brazil

Mexico

268

Rinaldo Antonio Colomé and Fernando M. Giuliano

Series in Figure 15 do not differ significantly from those plotted in Figure 4, a not surprising result given the null or slight decreases in the series profile when removing the most distortive commodity for each country. Distortive support remains high (about 60 per cent) for the European Union and Mexico, and represents all of agricultural support in Brazil. IV.

Conclusions

Debates over agricultural subsidies are the most controversial ones in the WTO negotiation rounds. In general, low subsidizing developing countries pressure highly subsidizing developed countries to reduce their support levels for their products to be competitive in world markets. This characterization is not completely accurate, though. Some developing countries as Mexico and Turkey do subsidize their agricultural production, and some developed countries, as Australia and New Zealand do it only slightly. The paper gave an extensive analysis of the level and composition of agricultural support in the European Union, Mexico, and Brazil. Unfortunately, this analysis could not be done for Argentina, due to the lack of data. Countries analyzed include a high subsidizing group (European Union), a moderate subsidizing country (Mexico), one with almost no agricultural support at all (Brazil), and one with high negative support (Argentina). Nevertheless, the level of support is not but only one issue that should be addressed in a serious agricultural subsidy’s debate. The way that support is implemented is very important as well. In that sense, we can find “healthy” agricultural support and “harmful” agricultural support. A healthy support measure is one that achieves its goal of improving farm receipts without significantly affecting production and trade. On the other hand, harmful support is not very effective at increasing farm receipts and distorts agricultural markets the most. When analyzing PSE composition, the country order changes: The European Union is the most efficient subsidizing group, followed by Mexico and then Brazil, which implements all of its agricultural support policies through heavily distortive measures. Not only that, but the European Union is the only region that shows diminishing per cent PSE figures together with a decreasing share of the three most distortive support measures, a sign of significant policy reform efforts. Of course, being European farmers heavily

Agricultural Policies and Trade in the European Union and Selected Latin American Countries

269

subsidized, there was not much room to increase agricultural support, and there was a lot of pressure to decrease it. However, it is worth noting the policy reform efforts the European Union has taken. Still, distortive agricultural support is far from being negligible. Support in the selected countries is not only distortive, but spread among different commodities as well. We have seen that distortive support levels do not significantly diminish after the most distortively supported commodity is withdrawn from PSE. This fact has negative policy implications: further decreases in absolute and distortive support requires reforms in policy measures affecting several products, having to deal with resistance from different groups. As said in section II, it is a paradox that Argentina requests the reduction of agricultural subsidies of the developed countries in international organizations while –at the same time- it applies a heavy domestic negative protection.

References

Cahill, Carmel, Legg, Wilfrid (1989-90), Estimation of Agricultural Assistance Using Producer and Consumer Subsidy Equivalents: Theory and Practice, OECD Economic Studies 13, Paris, OECD. Colomé, Rinaldo A. (1966), La Oferta Agropecuaria de la Región Pampeana, Tesis Doctoral, Facultad de Ciencias Económicas, Universidad Nacional de Córdoba, Córdoba, Argentina. Colomé, Rinaldo A. (1976), Importancia estratégica de la Agricultura Pampeana en el Desarrollo Económico Argentino, in Instituto de Economía, Federación Argentina de Colegio de Graduados en Ciencias Económicas, La Producción de Cereales en la Argentina Medios para lograr su Desarrollo, Editorial El Coloquio, Buenos Aires, and Revista de Economía y Estadística, Nueva Serie, Primero, Segundo, Tercero y Cuarto Trimestre de 1977-78, Vol. XXI, No. 1, 2, 3 and 4, 39-65. Díaz, Alejandro C.F. (1975), Ensayos sobre la Historia Económica Argentina, Amorrortu Editores, Buenos Aires.

270

Rinaldo Antonio Colomé and Fernando M. Giuliano

Hitiris, Theo, Vallés, José (1998), Economía de la Unión Europea, Prentice Hall, Madrid, 4th edition, chapters 7 and 8. Medina, Juan J., Soto, Luis A. (2000), Política Comercial y Fiscal: Efectos Sobre el Sector Agropecuario Argentino, Anales, de la Asociación Argentina de Economía Política, XXV Reunión Anual, Córdoba. OECD (2005a), Producer and Consumer Support Estimates, OECD Database 1986 – 2004. Explanatory Notes, OECD Publications, Paris. Reca, Lucio G. (1980), Argentina: Country Case Study of Agricultural Prices and Subsidies, World Bank Staff Working Paper No. 386, Washington, D.C., USA. Strokov, Sergei, Meyers, William H. (1996), Producer Subsidy Equivalents and Evaluation of Support to Russian Agricultural Producers, Center for Agricultural and Rural Development, Working Paper 96-WP 168, Iowa State University, Iowa. Sturzenegger, Adolfo C., Martínez Mosquera, Beatriz (1986), Incidencia de las Políticas Comercial y Cambiaria sobre Precios Agrícolas: Argentina 1960-85, Anales, de la Asociación Argentina de Economía Política, XXI Reunión Anual, Salta, 1049-1085. Sumner, Daniel A., Tangermann, Stefan (1999), International Trade Policy and Negotiations, in: Bruce Gardner / Gordon Rausser (eds.) Handbook of Agricultural Economics, North Holland Press. Tangermann, Stefan (2005), Is the Concept of the Producer Support Estimate in Need of Revision?, OECD Food, Agriculture and Fisheries Working Papers No. 1, OECD Publishing. USDA (1995), Producer and Consumer Subsidy Equivalents, Economic Research Services, USDA.

Matteo Grazzi and Antonella Mori

The Regulation of FDI in Latin America and the Caribbean: What Impact on Host Countries? Abstract Foreign direct investment (FDI) inflows have been an important capital source in Latin America over the past decades. This paper analyzes different regulation models for FDI in the region and discusses the factors which affect the choice of the level of regulation, e.g. the quality of the host country’s institutions, features of sectors which are destinations of FDI (natural resources, services or manufacturing) and motivations for FDI. Moreover, we address the issue of identifying a first-best regulatory framework in order to maximize benefits for the host country, in terms of attracting flows that contribute to development. I. Introduction Foreign direct investment (FDI) inflows have been an important capital source in Latin America over the past decades: initially, foreign companies were mainly attracted by the import substitution industrialization policies and the natural resources richness of the region, while thereafter trade and capital liberalization as well as privatization generated a new increase in FDI inflows. The past 15 years have seen not only a wave of FDI inflows but also a considerable increase in the number of investment treaties at bilateral level as well as at regional level. Little progress has been made at the multilateral level, while few countries have decided to foster FDI inflows only by granting national treatment to foreign companies. At present, different regulatory framework levels coexist in Latin America: national, bilateral (BITs and Double Taxation Treaties), regional (e.g. Decisions 291 and 292 of CAN) and multilateral (e.g. TRIMs and the ICSID Convention). The aim of the paper is to identify different regulation models in the region by studying four countries: Argentina, Chile, Costa Rica and Trinidad and Tobago. We discuss the factors which affect the choice of the level of regulation, e.g. the quality of the host

272

Matteo Grazzi and Antonella Mori

country’s institutions, features of sectors which are destinations of FDI (natural resources, services or manufacturing) and motivations for FDI. Finally, we address the issue of identifying a first-best regulatory framework in order to maximize benefits for the host country, in terms of attracting flows that contribute to development. The paper is organized as follows. Section II presents the regulatory framework in the four countries selected. Each country brief outlines the major regulations at all levels: national (by dividing the measures into three different categories investment facilitation, investment protection, and investment promotion and incentives), bilateral, regional, and multilateral. Section III describes the performance of FDI flows to the four Latin American and Caribbean countries over the past years. Finally, in section IV we compare and contrast the different regulatory models and FDI motivations in the four case studies, the aim being to investigate whether different regimes give rise to different performances in attracting FDI. The section ends with a discussion of some general policy implications regarding the choice of the regulatory framework for treatment of foreign companies which maximizes benefits for the host countries. II. The selected case studies The four Latin American and Caribbean countries analysed (Argentina, Chile, Costa Rica and Trinidad and Tobago) have been selected in order to represent countries which are small and medium-sized in terms of their populations and gross domestic products, as well as to cover South and Central America and the Caribbean geographically. As regards the period considered, the economic and regulatory evolution of the countries has been studied up until 2004-05. The four countries are all open to FDI, but they differ in their regulatory frameworks and abilities to attract FDI. Table 1 shows the amount of FDI inflows in per capita terms that these Latin American and Caribbean countries received in 2004. Trinidad and Tobago, the smallest economy, attracted the highest value (US$ 919.5), followed by Chile (US$ 475.8), while Argentina ranked last with US$ 107.6. Although not analysed in this paper, Table 1 includes also Brazil and Mexico, which are the two major FDI destination countries in Latin America.

The Regulation of FDI in Latin America and the Caribbean

273

Table 1: Per capita FDI inflows in 2004 (US dollars) Argentina Brazil Chile Costa Rica Mexico Trinidad and Tobago Source: Authors’ calculation on UNCTAD data.

107.6 97.6 475.8 153.9 156.3 919.5

Each country presentation is organised with a similar format. The first part describes the relevant regulation at national level, which is analysed in terms of three different features: measures to facilitate investment (e.g. national treatment, stability pact), measures to protect investment (e.g. expropriation law), and measures to promote investment and incentives (e.g. tax incentives). The rest of each country presentation is devoted to agreements at bilateral level, like the BITs, agreements at regional as well as multilateral level. A. Argentina In the 1990s, economic liberalization in Argentina, combined with a large and extensive privatization process fully open to foreign capital, induced a large FDI inflow into the country. National treatment (guaranteed both in the pre-establishment and in the postestablishment phase), together with macroeconomic stability and the investment protection provided by the several BITs signed by the country, helped it to pass from an FDI inflow of 678 million dollars in 1980 to 1,836 million in 1990 and to a peak of 23,986 million in 1999. The fall of FDI inflows recorded in the early 2000s was obviously due to the severe economic crisis that hit the country in those years. In 2003 only 1,887 million dollars entered the country, which was the level of 13 years before. However, in 2004, FDI inflows revived because of the favourable exchange rate of the peso and amounted to 4,274 million, which helped the Argentine economy to recover. 1. National Level The regulatory framework pertaining to foreign investment at the national level is based primarily on: - The 1994 Argentine Constitution - The Foreign Investment Act N.21382 (enacted by Executive Order 1853/93) - The Investment Promotion Law n.25924

274

Matteo Grazzi and Antonella Mori

i) Investment Facilitation The 1994 Argentine Constitution (art.20 and art.25) and the Foreign Investment Act N.21382 grant full national treatment to foreign investors. Foreign companies may invest without prior approval or any registration requirements; they have the same access to incentive programs as local investors and unrestricted access to all economic activities. An exception to this general rule is constituted by the legislation enacted in 2003 on “cultural goods” (including media and internet companies), which restricts foreign ownership to 30 per cent. Moreover, after the extensive privatization process enacted in the country, most state enterprises and public services have passed into private ownership, often involving foreign investment. Foreign investors are entitled to utilize any of the corporate structures allowed by the Argentine business law and may remit abroad any liquid earnings and profits realized as a result of their investment, and they may also repatriate their capital without any restrictions. There are no nationality requirements to work in Argentina, and an employment contract with a company located in Argentina is sufficient to obtain a visa authorizing work in the country. ii) Investment Protection Full access to Argentine courts is guaranteed to foreign investors in accordance with the national treatment principle. But public confidence in the Argentine judiciary system is traditionally weak, and today, after the crisis, it is even weaker. The Transparency International Corruption Perception Index for Argentina was 2.8 in 2005, which placed the country 97th in the World Ranking, at the same level as Algeria and Madagascar. In 2000, the index was 3.5 and Argentina occupied 52nd place in the World Ranking. The Fraser Institute Index for Legal Structure and Security of Property Rights fell from a score of 5.4 in 2000 to 3.8 in 2004. Inspection of the World Bank Governance Indicators shows that Argentina’s percentile rank1 in Rule of Law diminished from 64.9 in 1998 to 28.5 in 2004. Moreover, according to World Bank Doing Business, 33 procedural steps and 520 days are necessary to enforce a contract in 1

Percentile rank indicates the percentage of countries worldwide that rate below the selected country.

The Regulation of FDI in Latin America and the Caribbean

275

Argentina, with a cost amounting to 15 per cent of the debt. This signifies that an important issue for investors is finding a way to evade domestic jurisdiction so as to gain access to arbitration. The government of Argentina accepts the principle of international arbitration under the procedures set out in the text of the many BITs signed by the country, although domestic justice is an alternative to international arbitration. Expropriation is regulated in domestic legislation by the Expropriation Law N.21499 of 1977, which establishes that the State may expropriate only for a public purpose and under due process of law. The law provides for an agreement between the State and the private party which establishes the amount of compensation to be paid. If this agreement cannot be reached, a special procedure is followed, and compensation is calculated on consequential damage plus up to 10 per cent. However, the large number of BITs signed by Argentina means that the classic “prompt, effective and adequate” formula applies to most of the foreign investment stock. Recently, several foreign investors have resorted to international arbitration, claiming that the pesoification of dollar-denominated contracts of January 2002 has constituted an effective indirect expropriation of their investment, or protesting against the price freeze on public services imposed by the Government. Thirty-six of the 104 cases pending in December 2005 at ICSID involved Argentina. In an attempt to avoid a generalized claim for international arbitration which might undermine the fragile Argentine recovery, in 2003 two new agencies (the “Amicable Negotiations Federal Council” and the “Amicable Negotiations Proceedings Body”) were created by the Argentine Government to conduct amicable negotiations under BITs provisions and to conclude agreements without resorting to international arbitration. iii) Investment promotion and incentives The current regime of incentives in Argentina is designed to promote investment by both domestic and foreign investors in compliance with the national treatment principle. It is possible to identify three broad categories of incentives: Horizontal (i.e. incentives which can be applied in any region or sector), Regional and Sectoral. In September 2004, the Investment Promotion Law N.25924 created a new mechanism to attract investment which provides for the accelerated reimbursement of VAT relative to the investment, accelerated depreciation for machinery, equipment and

276

Matteo Grazzi and Antonella Mori

infrastructures destined for investment projects, and the refund of VAT paid on goods purchased for investment projects. On a regional basis, most Argentine Provinces offer their own investment incentives, which usually include tax and duty exemption or reduction, and support for infrastructure and equipment projects. Furthermore, some Free Trade Zones, where goods can be imported and exported without duties and duty controls, and a special Foreign Trade Area in Tierra del Fuego have been created.2 On a sectoral basis, Argentine legislation provides a large set of incentives for investment in mining, forestry, software and tourism. The Mining Investment Regime guarantees double fiscal deduction and VAT reimbursement for exploration expenses. Moreover, investment in extraction and production activities is granted fiscal stability for up to 30 years, duty exemption on imports of capital goods, accelerated depreciation of investments made in equipment, construction and infrastructure, and VAT refunds on imports of new capital goods used in the mining production process. The Forest Investment Law N.25080 complements the Promotion Regime for Forest Plantations (1992) and provides fiscal stability, a special regime of deductions for Income Tax, VAT refunds and some non-refundable financial aid for forestry.3 Law N.25922 promotes the creation, design, development, production, implementation and conditioning of software systems. A number of benefits are offered for investment in this sector, including fiscal stability for a ten-year term, non-refundable and nontransferable credits for employer social security contributions, a 60 per cent reduction of income tax liability, and exemption from import duties on hardware and IT components. Finally, investment in tourism sector, for example the construction of new hotels, is promoted by an accelerated refund of VAT on the investment or with an accelerated depreciation regime.

2

3

Nine Tax Free Zones are currently operating in the country, in the Provinces of Buenos Aires, Cordoba, Chubut, La Pampa, Mendoza, Misiones, Salta, San Luis and Tucuman. The non-refundable financial aid consists of a fixed sum per hectare which varies according to the area, the tree species and the forestry work.

The Regulation of FDI in Latin America and the Caribbean

277

2. Bilateral level Over the past 15 years, Argentina has signed a large number of BITs which have played an important role in attracting foreign investment into the country. These treaties occupy a rank higher than domestic laws in the juridical hierarchy, so that in case of conflict of laws, they should prevail. Fifty-four of these agreements have been ratified by the Argentinean Congress and are currently in force, protecting virtually all the foreign investment stock in the country. Investment treaties have been signed and ratified with 14 countries of the EU-15 (except Ireland) and 17 of the EU-25.4 Argentina has also signed several Double Taxation Treaties.5 3. Regional level In January 1994, the MERCOSUR Member States signed a reciprocal investment promotion and protection agreement (entitled the ‘Colonia Protocol’) which guarantees national treatment and mostfavoured nation treatment to intra-regional investors, both in the pre-establishment and post-establishment phases. The agreement does not generally stipulate performance requirements, but both Argentina and Brazil have maintained the right to impose some such requirements in the automobile sector. Moreover, The Colonia Protocol contains provisions preventing expropriation not motivated by the public interest, granting due process and prompt and fair compensation. Restrictions on capital repatriation and profits in convertible currency are not allowed. Argentina has made exempt from the agreement border real estate, air transportation, shipbuilding, nuclear power generation, uranium mining, insurance and fishing. To date, however, the treaty has not entered into force because it has not been ratified, with the paradoxical result that in-

4

5

With Austria (1992), Belgium and Luxemburg (1990), Czech Republic (1996), Denmark (1992), Finland (1993), France (1991), Germany (1991), Greece (1999), Hungary (1993), Italy (1990), Lithuania (1996), Netherlands (1992), Portugal (1994), Spain (1991), Sweden (1991), U.K. (1990). With Australia (1999), Austria (1992), Belgium (1996), Bolivia (1976), Brazil (1980), Chile (1985), Canada (1993), Denmark (1995), Finland (1994), France (1980), Germany (1979), Italy (1983), Mexico (1997), Netherlands (1996), Norway (1997), Spain (1992), Sweden (1995), Switzerland (2000).

278

Matteo Grazzi and Antonella Mori

vestments from MERCOSUR countries in Argentina are covered by neither regional nor bilateral regulation. In 1994 The Member States also signed an agreement which regulates extra-regional investment, i.e. investment from third states in the region: the so-called Buenos Aires Protocol. Prompted by the need to harmonize the investment regimes of the Member States in order to prevent investment inflows’ distortion, the Protocol provides a model for investment regulation. Its sphere of application is only the post-establishment phase, so that admission is regulated by national laws. The Buenos Aires Protocol is not yet in force either. Both Protocols provide mechanisms for dispute settlement between investors and contracting parties. 4. Multilateral level Argentina is a member of the World Bank’s Multilateral Investment Guarantee Agency (MIGA), of the Overseas Private Investment Committee (OPIC), and, since 1994, of the International Centre for Settlement of Investment Disputes (ICSID). Argentina also belongs to the World Intellectual Property Organization (WIPO) and to the World Trade Organization, the Argentine Congress having ratified the Uruguay Round Agreements in Law 24425 on 5 January 1995. B. Chile Since the 1970s Chile has based its national development strategy on openness to foreign investment. The Chilean Constitution grants no discrimination against foreigners, and national treatment is generally guaranteed in the legislation. Nevertheless, it is only since the return of full democracy in the country, in 1990, that investment flows have markedly increased. The country’s business-friendly environment based on certainty of law and transparency, together with political stability and the signing of numerous investment treaties, were effective in attracting a large amount of foreign capital during the 1990s. The FDI inflow was 287 million dollars in 1980, 661 in 1990 and 7,603 in 2004. Today, with 475.76 dollars of FDI per capita, Chile is the third largest recipient of foreign investment in Latin America (after Brazil and Mexico) and the second after Trinidad and Tobago. 1. National Level The regulatory framework pertaining to foreign investment at the national level is based primarily on: - The Chilean Constitution

The Regulation of FDI in Latin America and the Caribbean

279

- Chapter XIV of the Central Bank’s Compendium of Foreign Exchange Regulations - The Foreign Investment Statute (Decree Law 600). i) Investment Facilitation Foreign investment is regulated in Chile by two different mechanisms: Chapter XIV of the Central Bank’s Compendium of Foreign Exchange Regulations (CFER) and The Foreign Investment Statute (Decree Law 600). Since 1974, when The Foreign Investment Statute came into force, the large majority of investments have entered the country under this mechanism. Indeed, between 1974 and 2004, 78.6 per cent of total foreign investment entered Chile under Decree Law 600, against 16.6 per cent under Chapter XIV,6 although recently, given a set of reforms of the CFER enacted in 2000-20027, the situation seems more unstable. In 2003 foreign investment under Chapter XIV (50.9 per cent) for the first time exceeded that under Decree Law 600 (49.1 per cent), but in 2004 the percentages were once again favourable to Decree Law 600 (70.0 per cent against 30.0 per cent). Table 2: Chile: FDI inflows by mechanism, 1974-2004 (million US $ - percentages) Investment Mechanism Foreign Investment Statute (D.L.600)

1974-2004

2001

2002

2003

2004

58,672 78.6 5,017 82.9 3,381 67.2 1,286 49.1 5,004 70.0

Chapter XIV 12,351 16.6 1,037 17.1 1,654 32.8 1,334 50.9 2,144 30.0 CFER Chapter XIX 3,600 CFER Total

4.8

0

0

0

0

0

0

0

0

74,623 100 6,054 100 5,035 100 2,620 100 7,148 100

Source: Foreign Investment Committee, Central Bank of Chile 6

7

The residual 4.8 per cent entered the country under Chapter XIX of the CFER, a debt conversion mechanism which played an important role between 1985 and 1991 but which is no longer in place. In May 2000, for example, the Central bank lifted the one-year withholding period requirement for capital entering the country under Chapter XIV.

280

Matteo Grazzi and Antonella Mori

Chapter XIV establishes regulations applicable to credits, deposits, investments and capital contributions from abroad, and requires the foreign investor only to register. However, it does not provide all the guarantees included in Decree Law 600, which imposes the principles of non-discrimination and non-discretion. Under the Decree Law 600 mechanism, the foreign investor concludes a binding contract with the Chilean State, which cannot be modified unilaterally by the State or by changes in the law. The intending investor must submit an application to the Foreign Investment Committee,8 which establishes the terms and conditions of the investment. In particular, the Committee defines the period in which the investor must transfer the capital.9 Foreign investors have the right to remit profits freely, but they may repatriate capital only after at least one year of operation in the country. This repatriation is devoid of tax, duty or charges up to the amount of the original investment. Moreover, after the payment of the relevant taxes, investors have access to freely convertible foreign currency without limits for both capital and profits remittances. Although Chilean legislation is generally “foreign investor friendly”, some sectors are still subject to exceptions from the national treatment. In particular, foreigners are not allowed to invest in media or fishing companies unless a reciprocity agreement exists between Chile and their country. In other sectors, besides the approval of the Foreign Investment Committee, some investments require additional authorizations from other institutions. For example, investment projects in the particularly important mining sector are subjected to the scrutiny of the Chilean Copper Commission, and the Undersecretariat of Fishing reports on projects in this field. Operations in the banking sector are approved by the Banks and Financial Institutions Regulatory Agency, while activities in the insurance and investment fund field are evaluated by the Securities 8

9

The Foreign Investment Committee consists of the President of the Central Bank and the Ministers of Economy, Finance, Foreign Affairs, and Planning, and the relevant Minister in the case of applications filed with Ministries not otherwise represented in the Committee. A maximum period of three years may be requested by the investor to transfer the capital. Investments of more than 50 million dollars can obtain a time limit of up to eight years. In case of mining projects the limit is also eight years and, if a previous exploration is required, the Foreign Investment Committee can extend it to twelve years.

The Regulation of FDI in Latin America and the Caribbean

281

and Exchange Commission. Finally, operating in the telecommunications sector requires a license, and the number of licenses is often limited. Temporary entry for foreign workers is defined as a stay in the country for up to 90 days. Visas are easy to obtain. Necessary for longer periods is a temporary residence permit (one year renewable) which requires a more complex procedure. However, the Chilean Labor Code provides that at least 85 per cent of the workforces of companies with more than 25 employees must be Chileans.10 ii) Investment protection Foreign investors have full access to the Chilean judiciary system, and under article 9 of the Foreign Investment Statute they must not be discriminated against either directly or indirectly. If any juridical rule is deemed discriminatory against foreign investors, the latter are entitled to submit a complaint to the Foreign Investment Committee. However, the judicial environment in Chile is generally transparent and efficient, and it constitutes an important factor in attracting foreign investment into the country. According to World Bank Doing Business, 28 steps and 305 days are necessary to enforce a contract in Chile, at a cost of 10.4 per cent of the debt. Chile’s performance is therefore considerably better than the regional average (35.5 steps and 461.3 days with a cost of 23.3 per cent of debt). Transparency International’s Corruption Perception Index for Chile was 7.3 in 2005, which placed it 21st in the World Ranking, with the same score as Japan and above Spain and Italy. The 1980 Political Constitution and the Expropriatory Procedures Law (Decree Law N.2186) permits the Government to expropriate private property only for the public and national interest, on a non-discriminatory basis and under due process of law. Compensation must be provided without delay and at market value, any applicable interest having been considered. Moreover, practically all foreign investment is also protected by the several BITs signed by Chile, which generally guarantee that expropriation can only be enacted in accordance with a law based on public or national interest, on a non-discriminatory basis and with prompt, adequate and effective compensation. However, no nationalizations have been enacted in Chile since 1973. 10

This limit does not apply to high-specialized workers which cannot be replaced by Chilean staff.

282

Matteo Grazzi and Antonella Mori

iii) Investment promotion and incentives The Chilean regulation model for foreign investment does not provide a large set of incentives, this being in line with the country’s commitment to free-market oriented economic policies. Under Decree Law 600, no “tax break” or “tax holidays” are provided to investors, although article 8 allows them to opt for a mechanism which provides a stable tax horizon and basically guarantees the invariability of the regime, of both direct and indirect taxation, prevailing at the time when the investment is made. Besides this article, some benefits, such as tax rebates and grants, are provided by the government only for investments, both foreign and domestic, in isolated geographical zones and in the information technology sector. However, recent legislation11 provides certain benefits, such as taxfree status on earnings from international operations, for multinational enterprises which choose to use Chile as their regional bases. 2. Bilateral level By October 2005 Chile had signed 52 BITs, and 38 of them are currently in force. Classifying them on a geographical basis, there are 14 agreements with EU countries,12 as well as treaties with the Netherlands (1999) and Hungary (1997) which have been signed but not yet ratified by both parties. On considering the EU-15, one notes that Chile has not signed an investment treaty only with Ireland. Moving to Latin America, Chile has signed 14 agreements currently in force.13 Treaties with Brazil (1994), Colombia (2000) and Dominican Republic (2000) have been signed but are not yet in force. Furthermore, by virtue of Chile’s membership of the APEC, several BITs have been signed with countries in the Pacific area: 11 12

13

See 2002 Investment Platform Law (Law N. 19.840) With Austria (1997), Belgium (1992), Czech Republic (1995), Denmark (1993), Finland (1993), France (1992), Germany (1977), Greece (1996), Italy (1993), Poland (1995), Portugal (1995), Spain (1991), the United Kingdom (1996), Sweden (1993). Argentina (1991), Bolivia (1995), Costa Rica (1996), Cuba (1996), Ecuador (1994), El Salvador (1996), Guatemala (1996), Honduras (1996), Nicaragua (1996), Panama (1996), Paraguay (1995), Peru (2002), Uruguay (1995), Venezuela (1993).

The Regulation of FDI in Latin America and the Caribbean

283

Australia (1996), China (1994), Indonesia (1999), Malaysia (1992), New Zealand (1999), Philippines (1995), Republic of Korea (1996) and Vietnam (1999). The treaties with Indonesia, New Zealand and Vietnam are not yet in force. Finally, also in force are investment agreements between Chile and Croatia (1994), Norway (1993), Romania (1995), Switzerland (1999) Ukraine (1995). Treaties with Egypt (1999), Lebanon (1999), Iceland (2003), South Africa (1998), Tunisia (1998) and Turkey (1998) have been signed but not yet ratified. In the ten years after 1991, when Chile became a signatory to the Washington Convention of 1965, 49 BITs were signed to cover all the foreign investment stock in the country. Moreover, many issues normally covered by BITs are regulated by the Free Trade Agreements that Chile has signed with important investor countries like Canada, Mexico and the U.S. Chapter 10 of the US-Chile Free Trade Agreement (which entered into force in 2004) is modelled on the NAFTA investment chapter and guarantees the following to foreign investors: a non discriminatory treatment through national treatment and MFN treatment; a minimum standard of treatment; no performance requirements; free transfers of funds related to the investment; expropriation only in accordance with customary international law; permission to hire key personnel without regard to nationality. Section B of Chapter 10 also provides a mechanism for the settlement of investor-State disputes based on international arbitration. Chile has also signed 21 double taxation treaties. 3. The EU-Chile Association Agreement On 18 November 2002, Chile and the EU signed an Association Agreement which entered into force in March 2005. It consists of a complex body addressing political, economic and cooperation issues. As regards foreign investment regulation, Part IV, on Trade and Trade-related Matters, contains some provisions on establishment. Article 132 of the agreement grants national treatment to investors of both parties, although article 135 lists numerous possible exceptions to this general principle. Some exceptions allow the host country to adopt or to enforce measures necessary to protect public health and security, the environment, and the artistic and historic heritage in order to avoid problems arising from Chapter XI of NAFTA. Together with this treaty the EU also signed, on behalf of the Member States, an investment agreement which, for the moment, deals only with the admission

284

Matteo Grazzi and Antonella Mori

phase but which may be an important step towards a European BIT. 4. Multilateral level Chile is a member of the Multilateral Investment Guarantee Agency (MIGA) of the World Bank and, since 1991, of the International Centre for the Settlement of Investment Disputes (ICSID) as well as of the WTO (since 1995) and the WIPO. Moreover, Chile has signed and ratified the New York Convention on the Recognition and Enforcement of Foreign Arbitral awards, and the Inter-American Convention on International Commercial Arbitration. C. Costa Rica Foreign direct investment in Costa Rica has grown considerably during the past 20 years. The structural policies implemented after the economic crisis of the early 1980s aided an economic recovery which, together with the establishment of Free Trade Zones, attracted new foreign investors. Many foreign companies, mainly in manufacturing (textiles and electronics), used Costa Rica as a platform for the exporting of goods and services around the world. Over time, other sectors, notably food products, financial services and tourism, have attracted foreign investments. The recent high FDI inflows to Costa Rica are a consequence of the country’s explicit strategy to attract investments in priority sectors, these being high technology, pharmaceuticals, tourism and environment. The strategy is based on three schemes of FDI incentives - the Free Trade Zone System, the Active Finishing Regime and a duty drawback procedure - with various degrees of tax exemption. Since the mid-90s an active policy to encourage FDI in high technology sectors has been implemented by developing science and technology, increasing productivity and quality, and strengthening education. The country has a generally open international trade and investment regime, with the exception of certain sectors reserved for state companies. The Commercial Code details all business requirements that must be fulfilled to operate in Costa Rica. The investment requirements for foreign and national persons and companies are identical.

The Regulation of FDI in Latin America and the Caribbean

285

1. National Level Costa Rica has no specific legislation targeted exclusively on foreign investment, and there are no specific limitations applicable to foreign investments. i) Investment Facilitation There is no registration requirement for FDI; nor is any FDI project subject to screening or approval. According to the jurisprudence of the Constitutional Tribunal established in 1989 by constitutional mandate, foreigners and nationals are considered equal and are granted the same rights and obligations. For those sectors in which limitations to private participation exist, the same rules are applied to both nationals and foreigners. Under DR-CAFTA,14 Costa Rica has committed itself to gradually opening up parts of the telecommunications and insurance sectors. Some sectors require the participation of at least a certain percentage of Costa Rican citizens or residents (electrical power, broadcasting, professional services, and wholesale distribution). There are no restrictions on receiving, holding or transferring foreign exchange. No restrictions are imposed on reinvestments or on the repatriation of earnings, royalties, or capital except when stipulated in contractual agreements with the government of Costa Rica. Foreign investors are able to borrow in the local market, but they are also free to borrow abroad. Costa Rica is signatory to many international major international agreements and conventions regarding intellectual property (see section II.C.4). Although the legal framework governing intellectual property is in place, however, its enforcement is often ineffective. Costa Rican laws, regulations and practices are generally transparent and pro-competitive, except in the state monopoly sector, where competition is excluded. Although corruption has recently become a major concern – in 2004 two former Presidents were detained on corruption charges stemming from state contracts – Costa Rica has a low level of 14

At the time of writing, DR-CAFTA had been ratified by Guatemala, El Salvador, Honduras, Dominican Republic, Nicaragua and the United States, and still needed to be ratified by Costa Rica before it could enter into force.

286

Matteo Grazzi and Antonella Mori

perceived corruption compared to the rest of Latin America. According to the 2005 Corruption Perception Index drawn up by Transparency International, Costa Rica ranks 51st worldwide, and only Chile and Uruguay have better scores in Latin America and the Caribbean. Bureaucratic procedures are frequently lengthy and discourage new investment. ii) Investment protection The 1949 Political Constitution of Costa Rica states that no property can be expropriated from a Costa Rican or foreigner without prior payment and demonstrable proof of public interest (article 45). Foreigners and Costa Ricans receive equal treatment. The 1995 Expropriation Law 7495 and its 1998 amendment further stipulate that expropriations can take place only after full payment has been made and certain procedures have been expedited, particularly those necessary for acquiring land for the construction of new roads. Invasions and occupations of private land by squatters are rather common. Squatters seek to take advantage of adverse possession devices in law which permit occupiers to receive title to unused farmland. The Civil and Commercial Codes govern commercial transactions. The courts are independent and their authority is respected. Judgements of foreign courts are generally accepted and enforced. The main problem in dispute settlement is that litigation may be lengthy and costly. The 1998 Law of Alternative Resolution of Conflicts and Promotion of Social Peace (Law 7727) seeks to encourage arbitration and to simplify the procedures under which it takes place. iii) Investment promotion and incentives Three investment incentive programs operate in Costa Rica: the Free Trade Zone System, the Active Finishing Regime and a duty drawback procedure. These incentives are available equally to foreign and domestic investors. After the 1981 Export Processing Law, in 1982 the first free trade zone was created, and it attracted mainly textiles companies. Free trade zones consist of primary extraterritorial customs and fiscal areas that allow the establishment of companies whose main purpose is to use Costa Rica as an exporting platform. Free trade zones also confer other incentives, such as tariff exemptions for raw materials and capital goods, income tax payment, added value, asset, and municipal tax exemptions, and the possibility of selling up to 25 per cent of the product in the local market (50 per cent for

The Regulation of FDI in Latin America and the Caribbean

287

service companies). Currently, there are nine different private free trade zones, the majority of them located near San José, as well as the two important main ports of entry, the Juan Santamaria Airport, the key highways and container terminals. The tax holidays provided for investment in Free Trade Zones are scheduled to be phased out in accordance with WTO Agreements. In addition, the tourism sector has a specific regime which is granted through the Tourism Interest Declaratory. The Active Finishing Regime, created by decree in August 1997, suspends taxes on imported inputs when they enter the national customs territory in order to be subsequently exported. The regime also facilitates suspension of taxes on capital goods used to manufacture exported goods. To enjoy the benefits of this regime, it is not necessary to operate within a specific industrial park or Free Trade Zone. The drawback procedure provides for rebates of duties or other taxes that have been paid by an importer for goods subsequently incorporated into an exported good. 2. Bilateral level Costa Rica has currently 14 BITs, with: Argentina (2001), Canada (1999), Chile (1998), the Czech Republic (2001), France (1997), Germany (1997), Great Britain (1997), Korea (2002), the Netherlands (2001), Paraguay (2001), Spain (1999), Switzerland (2002), Taiwan (2000) and Venezuela (2001). Awaiting ratification in the Legislative Assembly are investment treaties with Belgium, Bolivia, Ecuador, Finland and Luxembourg. 3. Regional level Within the framework of the Central American Common Market (CACM), member countries had agreed on a Central American Treaty on Investment and Services; however, they had not finished negotiating the annexes when the CAFTA negotiation started. CACM member countries decided that the investment provisions within DR-CAFTA would regulate investment relations among them, as well as between the US and Central American countries. The investment provisions in the DR-CAFTA are set forth in Chapter 10, which is divided into three sections, and in the annexes to the Chapter. Section A contains the substantive obligations of the Chapter, whereas Section B includes the investor-state dispute settlement procedures and Section C the Chapter’s definitions.

288

Matteo Grazzi and Antonella Mori

Section A contains fourteen Articles: scope and coverage, national treatment, most-favoured-nation (MFN) treatment, minimum standard of treatment, performance requirements, senior management and boards of directors, non-conforming measures, transfers, expropriation and consultation, special formalities and information requirements, denial of benefits, and investment and environment, relation to other chapters, treatment in case of strife. Section B contains thirteen Articles detailing the investor-state dispute settlement procedures. These Articles cover the following issues: consultation and negotiation, submission of a claim to arbitration, consent of each party to arbitration, conditions and limitations on consent of each party, selection of arbitrators, conduct of the arbitration, transparency of arbitral proceedings, governing law, interpretation of annexes, expert reports, consolidation, awards and service of documents. Section C on definitions covers the following terms: Centre, claimant, disputing parties, disputing party, enterprise, enterprise of a party, freely usable currency, ICSID Additional Facility Rules, ICSID Convention, Inter-American Convention, investment, investment authorization, investor of a non-party, investor of a party, New York Convention, non-disputing party, respondent, SecretaryGeneral, tribunal, and UNCITRAL Arbitration Rules and protected information. The DR-CAFTA has seven annexes: customary international law, public debt, expropriation, submission of a claim to arbitration, service of documents on a party under Section B, treatment in case of strife for Guatemala. 4. Multilateral level In 1996, Costa Rica signed the Convention Establishing the Multilateral Investment Guarantee Agency (MIGA) of the World Bank. In 1994 the country signed the Agreement on Trade related Aspects of Intellectual Property Rights and the General Agreement on Trade in Services, which entered into force in 1995. Differences between investors and the State on investment matters may be referred to the Arbitration Tribunal of the Convention on the Settlement of Investment Disputes – ICSID since 1993. D. Trinidad and Tobago An important element of the development policy of Trinidad and Tobago in the 1990s was the attraction of foreign investment, mainly in the hydrocarbons sector and in related industries.

The Regulation of FDI in Latin America and the Caribbean

289

In 1990, Trinidad and Tobago introduced a new foreign investment regime which has been effective in attracting FDI. The average annual foreign investment flow during the 1980s was 101.3 million dollars, while in the 1990s it was 491.5 million dollars. FDI inflows to Trinidad and Tobago were 185 million dollars in 1980, 109 million in 1990, but they rose to 1,001 millions in 2004. In 2004, the country received a per-capita FDI amount of 919.49 dollars, the highest level in Latin America and in the Caribbean. 1. National Level The regulatory framework pertaining to foreign investment at the national level is based primarily on: - The Foreign Investment Act N.16 of 1990 - The Companies Act of 1995 (as amended by the Companies Amendment Act of 1997). i) Investment Facilitation The foreign investment regime in Trinidad and Tobago is governed by the Foreign Investment Act N. 16 of 1990, which has reformed a previous, more restrictive, legislation. This Act guarantees national treatment to CARICOM investors, but it does not extend the same right to other foreign citizens. Foreign investors are allowed to hold 100 per cent of the share capital in a private local company, but they are required to notify the Ministry of Finance before making the investment. Furthermore, in order to own more than 30 per cent of the share capital of a local public company, foreign investors must obtain a licence. Current legislation also requires Government approval of the acquisition by foreign investors of commercial and residential land amounting to more than five acres for business purposes and over one acre for residential purposes. To date, however, licences on corporate ownership and land acquisition have been generally granted: according to the WTO Trade Policy Review, in the period 1998-2004, 294 applications for land ownership were submitted and 234 were granted. The repatriation of capital, profits, dividends, interests or other gains on investment is transacted without limits and there are no performance requirements for foreign investors. No sectors are entirely closed to foreign investment, but water production and distribution is reserved exclusively for the State

290

Matteo Grazzi and Antonella Mori

Foreign workers are allowed to enter Trinidad and Tobago without a work permit for a single period not exceeding 30 days in every 12 consecutive months. For any longer period, a work permit must be obtained from the Ministry of National Security. This system is widely seen as an obstacle by companies operating in the energy sector, which often need technicians for long periods. ii) Investment protection The High Court of Justice has jurisdiction over all matters involving sums in excess of TT$ 15,000 (about 2,000 Euro at current Exchange Rate) and foreign investors have free access to it. Corruption is traditionally moderate and does not undermine government and business operations. The Corruption Perceptions Index calculated by Transparency International for Trinidad and Tobago in 2005 was 3.8, which placed the country 59th in the World ranking, at the same level as Cuba and Thailand. However, an alternative international dispute settlement mechanism is provided by Trinidad and Tobago’s various BITs. Inspection of FDI flows into Trinidad and Tobago by country of origin in 2003, shows that 84.7 per cent of the investment was covered by a BIT. Expropriation is possible only for the purpose of public interest, under due process of law, on a non-discriminatory basis, and against prompt, adequate and effective compensation. Since the mid-1980s, no expropriations have been undertaken by the Government, and all prior expropriations were adequately compensated. iii) Investment promotion and incentives The majority of incentives to invest in Trinidad and Tobago are available under the Fiscal Incentives Act of 1979, the Trinidad and Tobago Free Zones Act of 1988 (last amended in 1997) and the Hotel Development Act of 2000 (amended in 2004). The Fiscal Incentives Act allows a tax holiday (or partial holiday) for periods of up to ten years for the manufacture of approved products by highly capital intensive enterprises investing in excess of TT$50 million (US$8.3million) and enterprises using a significant portion of local inputs. Approved enterprises also obtain exemption from customs duties and VAT on the construction of the approved project. Free Zone incentives have been developed in Trinidad and Tobago to promote the establishment of companies in Trinidad and Tobago that export the majority of their goods and services. A free zone enterprise under the Free Zones Act of 1988 is totally exempt from fiscal and administrative weights charges. At present, activi-

The Regulation of FDI in Latin America and the Caribbean

291

ties involving investments in excess of US$ 50 million, as well as production activities involving petroleum, natural gas or petrochemicals, are excluded from the program. Moreover, in order to develop tourism activities, the Hotel Development Act provides for tax holidays of up to ten years and duty free concessions for hotel developments. In addition, accelerated depreciation and capital allowances on the construction cost of the hotel may be granted after the holiday period. No depreciation is charged during the holiday period and an allowance at 20 per cent per annum is allowed for any five of the following eight years after the holiday period. Tax exempt profits may be distributed tax free by way of dividend. The incentive program in Trinidad and Tobago seems unable to attract a significant amount of investment because of its complexity, and because of Government discretion in granting concessions. In 2002 the former Tourism and Industrial Development Company of Trinidad and Tobago (TIDCO 2004)15 carried out a Survey on Investor Perceptions in Trinidad and Tobago16 which reported that only 36.1 per cent of the enterprises interviewed evaluated the Government’s incentives to invest as having a positive effect (limited or strong) on investment decisions. The final conclusion and recommendations of the TIDCO Survey advise the Government to investigate the real effectiveness of the incentive program and to consider the possibility of readdressing it or eliminating it in favour of a uniform lower tax rate. 2. Bilateral level Trinidad and Tobago has signed Bilateral Investment Treaties with Canada (1995), China, Cuba (1999), France (1993), South Korea (2002), Spain, The United Kingdom (1993) and the United States (1994). All these agreements have been ratified and are currently in force. Double taxation treaties have been signed with Canada, Venezuela and the United States.

15 16

TIDCO ceased operations in October 2005. TIDCO (2004), Survey of Investor Perceptions, http://www.investtnt.com.

292

Matteo Grazzi and Antonella Mori

3. Regional level Trinidad and Tobago is the largest economy of the 15 States belonging to the Caribbean Community and Common Market (CARICOM)17 and accounts for about 30 per cent of its GDP. The CARICOM was created by the Treaty of Chaguaramas in 1973 with the main economic goal of instituting a Common Market in the region. In 1989, the Heads of Government agreed to transform the Common Market into a single market and economy and decided that, for the transformation to take place, the Treaty would have to be revised. Between 1993 and 2000, nine Protocols18 were produced by an inter-governmental task force in order to amend the treaty. In 2001 these protocols were combined to create a new version of the Treaty formally entitled “The Revised Treaty of Chaguaramas Establishing the Caribbean Community Including the CARICOM Single Market and Economy.” By signing this new version of the original treaty, member states agreed to the establishment of a community investment policy which “shall include sound national macro-economic policies, a harmonised system of investment incentives, stable industrial relations, appropriate financial institutions and arrangements, supportive legal and social infrastructure, and modernisation of the role of public authorities” (art.68 of the Revised Treaty). Three draft intergovernmental instruments have been proposed in order to comply with this goal. The draft CARICOM Financial Services Agreement is designed to facilitate financial operations and the delivery of banking, insurance and securities services. The draft CARICOM Agreement on Investment (CAI) deals with intra-CARICOM investment, and it is the main instrument with which to achieve the objective of free movement of production factors in the Community. Finally, the draft CARICOM Investment Code is a harmonised regime for the treatment of investment from extra-Regional sources. With its 17

18

Other members of CARICOM are: Antigua and Barbuda, Bahamas, Barbados, Belize, Dominica, Grenada, Guyana, Haiti, Jamaica, Montserrat, St. Kitts and Nevis, St. Lucia, St. Vincent and the Grenadines, and Suriname. The nine protocols regard: New Institutional Structures of the Community; Rights of Establishment, Provision of Services and Movement of Capital; Industrial Policy; Trade Policy; Agricultural Policy; Transport Policy; Disadvantaged Countries, Regions and Sectors; Competition Policy; Disputes Settlement.

The Regulation of FDI in Latin America and the Caribbean

293

adoption, foreign investors will be subject to just one common Regional Investment Regime, rather than rules pertaining to fifteen different investment regulations. The schedule of completion agreed at the last Heads of Government Conference requires all these instruments to be finalised no later than 2006, but yet it is already possible to identify some steps towards a common Caribbean Investment Space. It is of particular interest to analyze the evolution of trade and investment agreements signed by CARICOM with third states. The first of these agreements was signed in 1993 with Venezuela. Its Treatment of Investments section envisages “...an encouragement to the promotion and protection of investment through the conclusion of bilateral investment treaties between the individual Member States and Venezuela...” (Art.9.1) with a general invitation to facilitate by means of these BITs movement of capital, right of establishment, joint ventures and repatriation of profits (Art 9.2). Similar encouragement is provided for Double Taxation Treaties (Art.10). One year later, the investment chapter had disappeared from the CARICOM-Colombia Agreement signed on 24 July 1994 on Trade, Economic and Technical Cooperation. In 1996 Heads of Government of CARICOM agreed to give priority to negotiating trade and investment agreements with selected countries in the greater Caribbean. In 1998, the CARICOM negotiated and signed on behalf of its Member States the first investment agreement, with specific provisions, including admissions clauses, investment treatment and dispute settlement rules, which has formed the Annex III to the Agreement establishing a Free Trade Area with Dominican Republic. Similar investment chapters have been subsequently signed: in 2000 with Cuba (Appendix A to the CARICOM-Cuba Trade and Economic Co-operation Agreement) and in 2004 with Costa Rica (Chapter X of the Agreement establishing the Free Trade Area between the Caribbean Community and Costa Rica). 4. Multilateral level Trinidad and Tobago has been a member of the International Centre for the Settlement of Investment Disputes (ICSID) since 1967 and is a signatory to the World Bank Multilateral Investment Guarantee Agency (MIGA). Moreover Trinidad and Tobago has ratified the New York Convention on the Recognition of Enforcement of Foreign Arbitral Awards.

294

Matteo Grazzi and Antonella Mori

Trinidad and Tobago has been a member of WTO since March 1995, and through ratification of the Marrakech Agreement it has incorporated into its legislation the General Agreement on Trade in Services and the agreement on Trade-Related Investment Measures (TRIMs). III.

Foreign direct investment performance

Over the past two decades FDI flows have increased considerably to both industrialized and developing economies. Comparing 2004 to 1980, Latin America and the Caribbean have been able to attract a growing amount of total FDI flows, bringing about a small increase in their share of FDI inward stock. In 2004 one fourth of FDI inward stock was in developing countries, about 2,200 billion US dollars. Since 1980, the FDI inward stock in developing countries has increased sixteen times, an increase similar to that recorded worldwide. In 2004, nearly one third of the FDI stock in developing countries was in Latin America and the Caribbean, a share slightly higher than in 1980. The four countries examined accounted for about 17 per cent of total FDI inward stock in Latin America and the Caribbean in 2004, a lower amount than in 1980. In fact, all countries except Chile had a lower share of FDI stock in 2004 compared to 1980.19 Table 3: FDI inward stock (Millions of U.S. Dollars) World of which (%) Developing Economies of which (%) Latin America & the Caribbean of which (%) Argentina Costa Rica Chile Trinidad and Tobago Source: UNCTAD.

19

1980 530 244

1990 2000 2004 1 768 589 5 786 029 8 902 153

24.9

20.6

30.1

25.1

30.2

32.4

29.9

32.7

13.4 1.2 2.2 2.4

7.4 1.1 8.5 1.8

13.0 0.5 8.8 1.3

7.3 0.7 7.5 1.4

Data on FDI stocks over long period of time have various shortcomings, such as the evaluation of existing assets and exchange rate volatility.

The Regulation of FDI in Latin America and the Caribbean

295

The rest of this section analyses the FDI performances of Argentina, Chile, Costa Rica and Trinidad and Tobago. A. Argentina A considerable amount of FDI arrived in Argentina in the 1990s, being attracted by the economic reforms enacted in the country, the intense privatization process and the favourable international context. Analysis of the data shows a positive trend until 1999, when a peak FDI inflow of 23,986 million dollars was recorded as a consequence of the purchase of Yacimientos Petrolìferos Fiscales (YPF), the former national oil company. Thereafter, because of the economic and political crisis of the country, the end of the privatization process and the changed international business climate, the trend went into reverse and FDI inflow fell to a level of 1,652 million dollars in 2003. In 2004, after a four-year decrease, FDI inflows rose to 4,274 million dollars, 159 per cent more than in 2003. On a sectoral basis, before 1990 most FDI inflows were mainly directed to the primary sector, and to manufacturing industries, in order to penetrate the protected domestic market. In the period 1990-1993, the privatization process started and high investment inflows were directed to Electricity, Gas and Water and Transport and Communications industries, and, after 1994, also to the banking sector, which attracted particularly large inflows in 1996-1998. In percentages, petroleum and mining accounted for 37 per cent between 1993 and 2004 and for 32 per cent between 2000 and 2004. The transport and communications sector recorded the most significant increase, from 8 per cent (1993-2004) to 19 per cent (2000-2004) as FDI recipient. As to the geographical origin of FDI in Argentina, the National Bureau of International Accounts of Argentina (DNCI) provides data on FDI inflows based on two different criteria: first and second ownership. The former criterion identifies the investment origin as the residence country of the first owner, often a subsidiary of the real investor used as an investment bridge, while the latter concerns the nationality of the head offices. In regard to first ownership, Europe is the main investor in Argentina with a share of about 58 per cent of FDI made in the country between 1992 and 2004, while North America accounts for 20 per cent, Central America for 10 per cent, South America for 8 per cent and other regions for 4 per cent. In regard to second ownership, the total percentage of FDI inflows

296

Matteo Grazzi and Antonella Mori

from Europe does not change, but the shares of countries with more favourable fiscal regimes (like the United Kingdom and the Netherlands), not surprisingly, decrease. Investment from Central America and the Caribbean, given the importance of off-shore financial centres in the area, also falls (from 10 per cent to 1 per cent), being off-set by the growth of the North American percentage (24 per cent from 20 per cent) and that of other regions (10 per cent from 4 per cent). Finally it is interesting to note the substantial presence of investment flows from other South American countries, in particular Brazil. B. Chile From 1974 to 2004, 74,623 million dollars in foreign direct investment entered Chile and represented an important element in the country’s development strategy. The Foreign Investment Committee of Chile provides detailed data on FDI inflows, although only the investments made in the country through the D.L.600 mechanism are considered. Moreover, the analysis tends to be biased. Investment in sectors (e.g. mining) where utilizing the D.L. 600 is particularly favourable are overestimated, while other sectors (e.g. some light manufacturing branches), where a large part of the inflows arrive also through Chapter XIV of CFER, are underestimated. Analysis of FDI gross inflows by sector through the D.L.600 mechanism between 1974 and 2004 shows that Mining accounted for 32.6 per cent; Electricity, Gas and Water Industries for 19.6 per cent; Services for 19.6 per cent; Manufacturing for 12.9 per cent; Transport and Communication for 11.5 per cent; Construction for 2.4 per cent and Agriculture, Forestry and Fishing for 0.5 per cent each. However, on evaluating the variation of percentages over time, it is interesting to note that the mining sector represented 47 per cent of total inflows until 1990 and progressively lost importance thereafter. In the 2000-2004 period its share fell to 23.2 per cent, and in 2004 FDI gross inflows directed to the mining sector amounted to just 7 per cent of the total. Given the increase in investment entering the country through Chapter XIV in recent years, the loss of relative importance of the mining sector may be even greater. This decrease has been off-set by higher inflows into the industries involved in the privatization process: the transport and communication sector (25.3 per cent in the 2000-2004 period and 35.4 per cent in 2004) and the electricity, gas and water sector (25.9 per

The Regulation of FDI in Latin America and the Caribbean

297

cent in the 2000-2004 period and 43.8 per cent in 2004). In the period between 1974 and 2004, materialized FDI through the D.L.600 mechanism mainly originated from the European Union (42.5 per cent), the United States (29.2 per cent), Canada (14.2 per cent), Australia (3.8 per cent) and Japan (3 per cent). But data on more recent periods show that the share of investment originated from the EU is growing, while investment from the United States is decreasing. Between 2000 and 2004, EU-15 accounted for 53.2 per cent and the United States for only 20.6 per cent. The main investor from the European Union is Spain (23.4 per cent between 1974 and 2004 and 30.7 per cent between 2000 and 2004), followed by the United Kingdom (8.9 per cent between 1974 and 2004 and 13.4 per cent between 2000 and 2004). Investment originating from other South American countries remains very low, less than 1 per cent of total FDI inflows between 2000 and 2004. C. Costa Rica Costa Rica’s political and economic stability, educated workforce, and attractive incentives have led to steady FDI flows, mainly from the US, in areas such as high-tech electronics manufacturing, medical devices, services, pharmaceuticals and tourism. Expectations are that inflows will be 609 million dollars in 2005 (Grupo Interinstitucional de Inversion Extranjera Directa, Julio 2005). The country has enjoyed more than 100 years of democracy, and more than 20 years of economic growth and financial stability. The legal framework grants foreign companies the same rights as local companies. There are no restrictions on the foreign ownership of assets, profit or capital repatriation. The government offers fiscal incentives, employee training benefits and expedited customs procedures. The Grupo Interinstitucional de Inversion Extranjera Directa, created in 2000 and composed of the Central Bank (BCCR), the Institute of Tourism (ICT), the Ministry of Foreign Trade (COMEX), the Trade Promotion Agency (PROCOMER) and the Investment and Development Board (CINDE), compiles FDI statistics. The internationally accepted definition of FDI applies. The Grupo Interinstitucional started the compilation of data with the current methodology in 1997; hence figures prior to 1997 are not directly comparable. There is no obligation to register FDI in Costa Rica and there is no register of FDI enterprises. All businesses must be registered in the National Registry, thereby becoming national companies that may have national or foreign owners. Contrary to other Latin

298

Matteo Grazzi and Antonella Mori

American countries, e.g. Argentina and Peru, the surge of FDI inflows into Costa Rica cannot be explained by privatization programs, because most of the State monopolies in Costa Rica have not been dismantled. The surge of FDI inflows into Costa Rica during the 1990s was the result of a clearly defined export promotion and diversification strategy. As regards the sectoral distribution of foreign investments in Costa Rica, the manufacturing industry accounted for nearly three fourths of total FDI in 2004, followed by the services and tourism industries, which have also been growing in recent years. The distribution of foreign investment according to the special promotion regimes in place shows that more than half of the total is established under the free zone regime, which is followed in importance by companies that do not use special regimes and tourismrelated FDI under the Tourism Declaration regime. The United States accounts for more than half of the foreign direct investments in Costa Rica. Nevertheless, direct investment from Europe is growing in importance, mainly due to the large investments undertaken by Dutch and German companies in the tourism sector and in the food and beverage industry at the beginning of this decade. D. Trinidad and Tobago In the 1990s, Trinidad and Tobago experienced an important shift towards liberalization of the economy and the privatization of key sectors. In 1990 the Government reformed the country’s Foreign Investment Regime and encouraged foreign investment also in several state enterprises. This trend has been reflected in a considerable increase of foreign investment inflows. The presence of large oil and gas deposits has traditionally attracted important oil companies like Chevron-Texaco, EOG Resources and Exxon Mobil. According to data available by sector, between 1992 and 2004 the petroleum industry accounted for 87.5 per cent of gross inflows of FDI. Despite the Government’s effort to diversify FDI sectors of destination, the bulk of foreign investment in Trinidad and Tobago is still made in the country’s energy sector. In 2004, 913.4 million dollars out of a total of 998.1 millions was directed to petroleum industries. The data aggregation provided by the Central Bank does not allow complete analysis of sectors of destination other than the energy sector; however, among the “other sectors”, there is

The Regulation of FDI in Latin America and the Caribbean

299

evidence that the tourism sector and information technology are growing in importance. The United States are traditionally the largest foreign investors in Trinidad and Tobago. As for Europe, given Trinidad and Tobago’s membership of the Commonwealth, the United Kingdom is the most important European investor country. In 2004, 697.5 million dollars arrived in the country from the U.S., while 169.9 millions arrived from the U.K. IV.

Regulatory models and impact on host countries

All four of the countries examined are open to foreign investment, but with rather different FDI policies and regulatory models. Analysis shows that a more favourable regulation of FDI does not necessarily generate higher FDI inflows: other variables, such as political and economic stability and business climate, are also important. On the other hand, too many incentives to foreign investors may be detrimental to the host country’s economy in terms of loss of potential tax revenues, environmental damage, or additional costs in the case of a domestic economic crisis which requires important economic policy changes. It emerges from the analysis in the previous section that the main factors that determine FDI inflows in the four countries are the search for natural resources and the search for markets. Table 4: Strategies of foreign companies FDI strategy

Recipient countries

Natural resource seeking

Argentina Chile Trinidad and Tobago Local market seeking Argentina (goods and services) Chile (services) Efficiency seeking (to export) Costa Rica Source: Authors’ elaboration based on Dunning’s classification on the motives of FDI.

Natural resource-seeking foreign companies are active in Argentina (hydrocarbons and mining), Chile (mining), and Trinidad and Tobago (oil and gas). Local market seeking strategies have been important in Argentina and Chile, being nearly exclusively triggered by the 1990 privatization programs in the case of Chile. In the case of Costa Rica, foreign companies have sought efficiency in

300

Matteo Grazzi and Antonella Mori

order to capture export markets. Table 5 summarizes the main features of the regulatory systems in the four Latin American and Caribbean countries described in section II. As far as investment facilitation is concerned, only Trinidad and Tobago has a rather different model, in that it does not grant national treatment and imposes more registration requirements and screening procedures. This is the consequence of the fact that the country attracts mainly natural resource-seeking FDI. In regard to investment protection, although the four countries are similar in terms of regulation, they Table 5: The regulatory models in the countries examined: main features in force (November 2005) Argentina

Chile

Costa Rica

Trinidad& Tobago

National treatment Registration requirements Performance requirements

YES

YES

YES

NO

NO

NO*

NO

YES

NO

NO

NO

NO

Capital controls

NO

NO*

NO

NO

Legal stability

NO

YES*

NO

NO

Expropriation protection

YES

YES

YES

YES

ICSID member

YES

YES

YES

YES

Free Trade Zones

YES

NO

YES

YES

Fiscal Incentives

NATIONAL LEVEL Investment Facilitation

Investment Protection

Investment Promotion YES

NO*

YES

YES

BILATERAL LEVEL BITs

54

38

14

8

Agreements with the EU

NO

YES

NO

NO

REGIONAL LEVEL

NO

NO MIGA TRIPS TRIMS GATS ICSID

NO MIGA TRIPS TRIMS GATS ICSID

CARICOM MIGA TRIPS TRIMS GATS ICSID

MULTILATERAL LEVEL

MIGA TRIPS TRIMS GATS ICSID

* with few exceptions or conditions Source: Authors’ elaboration.

The Regulation of FDI in Latin America and the Caribbean

301

are rather different in terms of actual enforcement of the regulation, as described in section II. In Argentina there are several limitations due to corruption or to the poor quality of the legal system (see also Table 6). It should be noted that Chile, which has been the most successful country in attracting FDI, is the one with the lowest investment promotion (see also II.B.1.). In order to gain a more complete picture, consideration should also be made of the actual enforcement and functioning of the FDI regulatory system. Table 6 presents a qualitative evaluation of the different models. While only few differences are apparent in Table 5, a more diversified panorama emerges from the qualitative scoreboard of the current FDI policy and regulatory framework. Categories have been evaluated by addressing the following issues: i) Investment facilitation: Is the national treatment really enforced and guaranteed? How binding are registration and screening requirements? Are capital controls a real constraint? ii) Investment protection: Are property rights protected? Are contracts enforced? Is the legal system working effectively? iii) Investment promotion: How active and effective is the investment promotion policy? iv) (4) Bilateral level: Does the country have BITs with its major investors? I.e. what percentage of total FDI is covered by BITs? Table 6: The regulatory models in the selected countries: authors’ evaluations Argentina NATIONAL LEVEL Investment FaciliHigh tation (1) Investment Protec- Medium tion (2) Investment PromoHigh tion (3) High BILATERAL LEVEL (4) Zero REGIONAL LEVEL (5) MULTILATERAL High LEVEL (6) Source: Authors’ elaboration.

Chile

Costa Rica

Trinidad& Tobago

High

High

Medium

High

High

High

Low

High

Low

High

Low

High

Zero

Zero

Low

High

High

High

302

Matteo Grazzi and Antonella Mori

v) (5) Regional Level: If the country is a party to Regional Trade Agreements, do the Agreements regulate investment at the regional level? vi) (6) Multilateral level: Has the country signed the major multilateral agreements? On considering the overall investment facilitation quality, Trinidad and Tobago scores ‘medium’ - the lowest score - owing to the absence of the national treatment and the existence of pre-investment requirements. Argentina’s investment protection scores are lower than the other three countries because of the poor quality of its legal system and corruption, as noted earlier. In regard to investment promotion, the situation depicted by Table 6 is even more different from the one presented in Table 5. Argentina and Costa Rica have the most active and generous investment promotion policies, which in the case of Costa Rica applies to both national and foreign investors. The investment promotion policies of Trinidad and Tobago and Chile are both evaluated at a low level, but for different reasons. Chile does not provide a large set of incentives because of a specific policy choice, in that foreign investment promotion is considered to be a distortion of the national treatment principle. On the other hand, Trinidad and Tobago offers important incentives, but because of the complexity of the bureaucracy, they have little effectiveness in actually attracting foreign investment. Finally, although Costa Rica has a higher number of BITs than Trinidad and Tobago, its score is low because Costa Rica does not have a BIT with the USA, the major source of FDI for a country. When the CAFTA-DR enters into force, FDI flows from U.S. companies will be regulated at the regional level. By contrast, Trinidad and Tobago has BITs in force with all its major investor countries. A. Risks for host countries: an important element to consider Sometimes, an excess of protection or a lack of regulation induce a potential loss of benefit for the host country in terms of additional costs in enforcing provisions taken in emergency situations, lower environment protection, or loss of potential tax revenues. This loss of benefit may exceed the beneficial effects of FDI inflows, and the net result may be negative. The experience of Argentina, Ecuador and Chile shows that these risks are not just theoretical but should be carefully evaluated by Latin-American policymakers.

The Regulation of FDI in Latin America and the Caribbean

303

The privatization program enacted in Argentina in the 1990s concerned the majority of public utility firms. With the aim of attracting investment in these sectors, a large set of incentives was introduced, and in order to provide stability and security to investors, utility rates were denominated in dollars and linked with the United States inflation index. In consequence of the explosion of the economic crisis, the “ley de covertibilidad” (the law that fixed at par the exchange rate between peso and dollar) was abrogated. The subsequent strong appreciation of the dollar against the local currency induced the Argentine Government to transform all the contracts denominated in dollars into contracts denominated in pesos. Moreover, the indexation of public utility tariffs to a foreign inflation index was eliminated. These provisions were seen by the majority of foreign investors as an indirect expropriation of their investment or as a violation of the “fair and equitable treatment” guaranteed by Argentina in the various BITs signed. Then, as allowed by such treaties, several foreign investors resorted to international arbitration. At the time of writing, 36 cases against the Argentine Government are pending in ICSID tribunals for a total claim of about 20 billion dollars (about 4 per cent of Argentine GDP in 2004). In a first award rendered unanimously by the tribunal in May 2005, Argentina was ordered to pay to the US-based CSM Gas Transmission Company (CSM) 133.5 million dollars plus interest in compensation for a violation of the BIT between Argentina and the U.S.A., rejecting Argentina’s argument based on state of necessity.20 If this decision is confirmed,21 it will constitute an important precedent for the other cases pending. In this scenario, a large bill will be presented to the Argentine Government for an emergency regulation which has been applicable to domestic and foreign investors without distinction. Treaties meant to guarantee real national treatment to foreign investment may be transformed into instruments which actually remove FDI from the sovereignty of the host country. Sometimes, a lack of environmental regulation may lead to higher FDI inflows, but it may also permit foreign companies to cause considerable damage which can be thoroughly evaluated only in the future. A clear example of these risks is provided by the ac20 21

UNCTAD (2005), World Investment Report 2005. Argentina moved to annul this award under the procedure established in Section 5 of Chapter IV of the ICSID Convention.

304

Matteo Grazzi and Antonella Mori

cusations brought by indigenous communities in Ecuador against the U.S. transnational oil company Texaco22, which left the country after 23 years of extractive activity. In those years Texaco dumped tons of waste and crude oil in the area where it was based, causing a high level of contamination: toxic and hazardous waste dumps were abandoned in the zone, waters were polluted and many local inhabitants started to suffer from chronic diseases as a result of environmental pollution. Texaco acknowledges the environmental and human damage caused by its activity, but it claims to have fully adhered to the industrial standards established by Ecuadorian Law at the time. Finally, in some cases, preferential fiscal regimes attract foreign investment but at the same time allow the removal of profits from the normal fiscal system, which heavily affects the host country’s state finances. For example, enterprises operating in the mining sector in Chile are generally subject to income tax and tax on dividends at 35 per cent, but Chilean law comprises mechanisms which allow the avoidance of tax payments through intra-group operations. In particular, it is possible to make remittances abroad without reporting them as profits and thus avoiding dividend taxation. Furthermore, because interest tax is only 4 per cent, firms record their contributions to subsidiaries as intra-group loans, and thus avoid paying dividend taxes at 35 per cent. This means that transnational mining enterprises are able to reduce the fiscal weight to a minimum, thus acquiring competitive advantages over national enterprises and reducing sectoral fiscal revenues. Exxon Mobil, for instance, has always declared losses in its 20 years of activity in Chile, avoiding payment of any income or dividend tax. B. Some policy implications and conclusions Policy towards FDI should focus on ensuring that investments are conducive to the country’s development goals. It is not a matter of attracting more investments, but of exploiting their potential net benefits (by increasing the benefits and decreasing the costs). The potential benefits for the host country may arise from technology transfer, productive linkages, human capital enhancement, local business development, employment, international competitiveness, 22

CEPAL (2004), Foreign Investment in Latin America and the Caribbean 2003.

The Regulation of FDI in Latin America and the Caribbean

305

and fiscal revenues. However, these impacts may be very low or even negative, e.g. the crowding out of local firms and an increase in imports higher than that of exports; there are other possible costs as well, related to environmental pollution and disputes settlements. If the policy and regulation are too favourable to foreign companies (e.g. no taxes, high incentives, no competition policy, no environmental regulation), FDI will come about, but the outcome in terms of benefits for the host country may be negligible. At the other extreme, if the policy and regulation are totally hostile to foreigners, FDI inflows will be zero; hence net benefits for the host economy will be zero. Between the two extreme cases of “treatment of foreign investment”, net benefits will be positive, with behaviour possibly taking the form of an inverted U-shape, as shown in Figure 1. As a consequence, depending on the initial host country’s position in terms of “treatment of foreign investment”, it will be convenient to change FDI policy and regulation in order either to improve the treatment of foreign companies (as in point A) or to decrease the advantages given to FDI (as in point B). Figure 1: The inverted U-shape relation between FDI treatment and net benefit of the host country Host country’s net benefits

A

0

B

1 Treatment of foreign investment

306

Matteo Grazzi and Antonella Mori

References

APEC (2004), Guide to the Investment Regimes of the APEC Member Economies, Forth Edition, Apec, Singapore. Blonigen, Bruce A. (2005), A Review of the Empirical Literature of FDI Determinants, NBER Working Paper 11299, NBER, New York. CEPAL, Various Issues, La Inversión extranjera en América Latina y el Caribe, Santiago de Chile. DNCI, Dirección Nacional de Cuentas Internacionales. Ministerio de Economía y Producción de Argentina (2003), La Inversión Extranjera Directa en Argentina, 1992-2002, Buenos Aires. Fraser Institute, Various years, Economic Freedom Index, : www.fraserinstitute.org (last visit: November 2007). TIDCO, Tourism and Industrial Development Company of Trinidad and Tobago (2004), Survey of Investor perceptions. TIDCO (2003), A Guide to Investing in Trinidad and Tobago. Transparency International, Various years, Corruption Perception Index, : www.transparency.de (last visit: November 2007). UNCTAD, Various Issues. World Investment Report, Geneva. UNCTAD (2005), Recent Developments in International Investment Agreements, Research Note, UNCTAD/WEB/ITE/IIT/ 2005/1. UNCTAD (2004), International Investments Agreements: Key Issues, (Vol. I, II, III), UNCTAD/ITE/IIT/2004/10. World Bank, Doing Business on-line database.

Ángel Enrique Neder, Julieta Schiro and Jonatan Saúl

Financial Integration in Some Countries of South America - The Use of Interest Parity Conditions as Indicators Abstract The objective of this paper is to measure the degree of financial integration in some countries of South America. A simple way of observing how economies are financially integrated is taking into account specific variables that reflect the properties of each financial system. It is known that, as markets become more open and unified, differences in rates of return should only show fundamental factors such as asset quality, risk, and the like. As a measure of financial integration we use the uncovered interest rate parity. Additionally, we prove the co-integration between the variables involved as a measure of a long run financial integration. I. Introduction Financial integration consists not only in an increase in the financial transactions taken by countries but also in a tendency for prices to converge in common currency terms. The integration of financial markets may also require more capital flows and may induce some changes in the economic structure and in the business and agents’ behavior, as well. At the same time, it gives support to financial deepening, and this promotes economic growth and an increase in social welfare. Financial integration could emerge from formal agreements or from de facto situations. In the first case, we can highlight the elimination of restrictions to cross border financial operations, the harmonization of tax policy, regulations and, the equal treatment all the participants receive when they operate in the markets involved in the agreement. On the other hand, de facto financial integration can be observed when foreign banks start having a strong presence in domestic markets, or when domestic firms have open access to international financial markets. This kind of financial integration in Latin America was more relevant than the formal one, particularly

308

Ángel Enrique Neder, Julieta Schiro, Jonatan Saúl

during the last twenty years. Besides, the integration in financial markets has been linked to the integration in trade and investment. Not only large international banks entered into the Latin American markets (and Argentina is a good example of this, particularly during the 1990s) but also local firms started widening their sources of funding, having access to international financial markets (sometimes also becoming a branch or a subsidiary of foreign firms). This also strengthened market forces and increased competition. The greater the financial integration, the larger the number of participants in that market and, as a consequence, the greater the efficiency (at least, from a theoretical point of view). In contrast, financial integration also presents some disadvantages (or problems). When financial markets are integrated, a sort of transmission from events that take place in one country to the financial markets of others would be present. Particularly, the effects of economic crises may be transmitted from one country to another which is financially related to. Moreover, asset price differentials would increase international capital flows and hence accelerate transmission of economic fluctuations. This appears as a main issue in financial market integration. Nevertheless, economic crises are not the only factor that could be transmitted. There are many other positive aspects that could be mentioned. For example, development and modernization of the financial system could be enhanced and the possibility of a greater and more sustainable growth could emerge. That is to say, as pointed out by Trichet (2006), financial integration enables economies of scale and enlarges the supply of funds for investment opportunities. And in order to reduce the transmission of crises, a good regulation and supervision in the financial market must be present. Furthermore, financial integration can be boosted by an increase in financial intermediation and bankarisation. Unfortunately, Latin America is not the best example for this (perhaps with the exception of Chile). As pointed out by González-Páramo (2005), the level of private domestic credit and deposits, in terms of GDP, is lower in Latin America than in developed countries, and in other emergent countries, as well. And, on top of all, institutions have not developed a proper role in order to favor the efficiency in the financial sector. A well known case is the “corralito” experienced in Argentina during the crisis suffered in 2001. The integration of financial markets could also be tested using the so called parity between interest rates. In this way, this paper is

Financial Integration in Some Countries of South America

309

aimed at analyzing the financial integration, focusing on some Latin American countries and using some tests related to that parity between interest rates. The remainder of the paper is organized as follows. The next section proposes different ways of measuring the interest rate parity as a proxy for financial integration. In section III, we present a model to evaluate financial integration using the uncovered interest rate parity. A description of the data used in the estimation of the model for each country is provided in section IV. Section V discusses the results, and concluding remarks are presented in section VI. II. Measuring interest parity There are some ways, among others, of measuring financial integration using indicators of “interest rate parity condition”. The main indicators of this type are the closed interest rate parity, the covered interest rate parity, and the uncovered interest rate parity. There are many other indicators that could be considered but we only pay attention to those connected to the interest rate parity.1 Closed interest parity holds if the rates of return are the same on financial instruments which are denominated in the same currency and are otherwise identical except that they are traded in different jurisdictions. This means that

it , k

itoffshore ,k

(1) where t is the current time and k the period of the instrument (monthly, quarterly, annual, etc.). Given market efficiency, closed interest parity fails when exchange rate controls are in place or when there are differences in country risk premium. Because these two types of restrictions were present in Latin America during the period of our analysis, we discard this indicator. Covered interest parity holds when the domestic rate of return ( it ,k ) equals the foreign one ( it*,k ), plus the forward discount on the home currency ( f t , t  k ). This forward discount measures the probability of depreciation of the spot rate in a specific term (i.e. three months future dollar). Consequently, the domestic interest rate would be

1

For a good explanation of this topics, see De Brouwer (1999).

310

Ángel Enrique Neder, Julieta Schiro, Jonatan Saúl

it ,k

it*,k  f t , t  k

(2)

The above definition is the baseline for assessing the international integration of traded financial asset markets. Because there are no developed market for forwards in Latin America, we also discard this method. Uncovered interest rate parity (UIRP) refers to the expectations of agents about the behavior of the spot rate (and inflation if we consider the real rate of interest instead of the nominal one). The UIRP states that the interest differential is, on average, equal to the ex post exchange rate change (Flood and Rose, 2001). Thus, we focus on this indicator and check the uncovered interest rate parity against the interest rate of the United States as a reference interest rate. In this way, our methodology involves the following: first, we perform a regression taking as the dependent variable the expected depreciation of the domestic currency2 ('sj) versus the difference between the domestic interest rate and the United States interest rate.3 We check the obtained coefficients and stationarity of residuals. The obtained coefficient should ideally be one. In case of not obtaining the expected results for the parameters (particularly valid for the short run) we consider the hypothesis of co-integration between the variables involved in order to find a long-run relationship. III. The model The equation to be estimated is:

's tj

D  E (it , j  itUS )  H t

(3)

where it , j is the domestic interest rate of the country j for the period t, and itUS is the same for the United States, and H t is an error term. The null hypothesis of UIRP is E = 1. The constant term Į could be interpreted as an exchange risk premium. In case we do not obtain the expected results for E , we test for unit roots in the residuals, and apply the Engel and Granger (1987) method, to prove co-integration between the involved variables. This method consists in estimating the original regression, and ap2 3

Calculated in a simple way as the rate of change between periods. The integration could have also been checked with European interest rates.

Financial Integration in Some Countries of South America

311

plying the Augmented Dickey and Fuller (ADF) Test to the residuals. The methodology described above gives us the result for the analysis of financial integration between the country under consideration (in our case Argentina, Brazil, and Chile) and the United States. Additionally, we test the integration between Latin American countries. If expectations were formed rationally and investors were risk neutral, the forward discount should equal the expected depreciation of the spot rate ( ' S te, t  k ). Given the measure of the country risk (i.e. Embi+, the risk indicator elaborated by the J. P. Morgan Bank) in the considered countries, we set the domestic interest rate (i.e. for Argentina) as follows:

it , Arg

itUS  V t , Arg

(4)

and in the other countries we have

it , j

itUS  V t , j

(5)

where j is the country considered. Then, the difference between country’s domestic interest rates will equal the difference in risks. it , Arg  it , j V t , Arg  V t , j (4) – (5) Finally, we can get a sort of uncovered interest rate parity considering the risk premium ( V t ), which takes into account the expected depreciation of the home currency and the fundamentals of the corresponding country: (6) it , Arg it , j  V t , Arg  V t , j it , j  ' V t

>

@

In this way, we parameterize Equation (6), and we should observe that the difference between interest rates in different countries should equal their difference in risk. Measuring financial integration, we concentrate on the uncovered interest rate measure because in one of its parameters we will take into account not only the expectation of domestic money depreciation (or devaluation) but also the default risk and the fundamentals observed in the country involved. The equation to be estimated is the following:

'it

D  E ( 'V t )  H t

(7) To confirm the uncovered interest rate parity (or a complete financial integration) we need to verify, as it was mentioned before,

312

Ángel Enrique Neder, Julieta Schiro, Jonatan Saúl

that E is equal to one. Nevertheless, we may get the results of weakly integrated financial markets due to risk, although being operatively well integrated. IV. The data The countries included in our estimation are Argentina, Brazil, and Chile. In a first step, a number of interest rates were considered, but finally we chose a set of short-run interest rates. In all cases, the reference interest rate was the CDs one month for the United States. Thus, the interest rate differential, for each considered country, is the difference between the local interest rate and that of the United States. For Argentina we used the nominal interest rate for term deposits (30 to 59 days). This variable is a monthly weighted average of a sample of banks that operate in the City of Buenos Aires. This time series is provided by the Ministry for Economic Affairs. In order to account for the expected depreciation (or devaluation), we used data from the IMF calculating the change in the exchange rate. For Brazil, the local interest rate used was that of the interbanking rate of certificates of deposits on the secondary market. And for the expected depreciation we used the change of the free exchange rate. The series used for Chile were obtained from the Central Bank of Argentina and from the Banco de Chile. The interest rate used was that of the term deposits (one month) and the expected depreciation was built in the same way as with the other two countries. V. Results For all of the three countries, comparisons of financial integration and the uncovered interest rate parity with respect to the United States of America were evaluated. Furthermore, not only the UIRP for each country compared with the United States was examined, but also the UIRP with intra Latin American countries was evaluated. However, we only present the results of the UIRP related to the United States. The intra Latin American tests did not deliver the expected results.4

4

One explanation of these unexpected results could be the difference in the policies adopted by each country: capital controls in Chile, the convertibility in Argentina, huge devaluations in Brazil.

Financial Integration in Some Countries of South America

313

A. The Case of Argentina Analyzing the case of Argentina, we first compare the development of the interest rate differential (DIFARGUS)5 and the expected devaluation (SARG)6. As we can see in Figure 1, from mid 2001 until mid 2002, the data reveals a lot of problems. Figure 1: Argentina:. Interest rate differential with US and expected devaluation 1.0

Basis points

0.8 0.6

SARG DIFARGUS

0.4 0.2 0.0 -0.2

96

97

98

99

00

01

02

03

04

05

Years

During the Convertibility period, we observe very little difference between interest rates (due to the confidence in the Convertibility Law, and despite the international crises in Asia, Russia, and Brazil). This situation started to change by the end of 2000. This was the consequence of the deep crisis that took place in Argentina and which was not only an economic crisis but also an institutional, political, and social crisis. Let us now turn to the regression analysis on the UIRP according to Equation (7). In order to eliminate the influence of outliers, we used dummy variables. Since January (when a big devaluation took place and the Convertibility Law was abolished), February, and March 2002 showed the biggest fluctuations, we included them as dummies.

5

(it , Arg  itUS ) in Equation (3).

6

'stArg in Equation (3).

314

Ángel Enrique Neder, Julieta Schiro, Jonatan Saúl

The fit of the regression turns out to be very good with an adjusted R² equal to 0.93. The constant is not significantly different from zero, which is in line with theory, but the coefficient of the variable DIFARGUS, although significant at the 1 per cent level, is not close to one, as expected. This is due to other aspects that should be taken into account, namely, constraints to capital movements, institutional reforms, financial restrictions (the socalled “corralito”), the asymmetric “peso-ification” and its negative influence on deposits and loans, etc. However, if we look at the residuals of this regression, we can see that the series involved still show heteroscedasticity problems. In order to correct for this problem, we run the same regression but with the Newey-West heteroscedasticity and autocorrelation consistent covariance, getting the following results: Table 1: Regression results UIRP Argentina – US (with Newey West heteroscedasticity and autocorrelation consistent co-variance) Dependent Variable: SARG Method: Least Squares Sample: 1996M01 2006M03 Included observations: 123 Newey-West HAC Standard Errors & Covariance (lag truncation=4) Variable Coefficient Std. Error t-Statistic Prob. C -0.002289 0.002068 -1.106848 0.2706 DIFARGUS 0.043506 0.046650 0.932600 0.3529 DJAN_02 0.940646 0.001775 529.9041 0.0000 DFEB_02 0.085997 0.002289 37.56616 0.0000 DMAR_02 0.705490 0.009458 74.59400 0.0000 R-squared 0.932715 F-statistic 408.9358 Adjusted R-squared 0.930435 Prob(F-statistic) 0.000000 Durbin-Watson stat 1.770091

As we can see in Table 1, the results are very similar to those obtained before, except for the coefficient of the interest rate differential which is now not significantly different from zero. This is supported by the constraints mentioned above. In addition, the remainder of the year 2002 showed big residuals (particularly in May and December) but stabilized since mid 2003.

Financial Integration in Some Countries of South America

315

So far, we could not affirm the existence of financial integration between Argentina and the US in the traditional way. Actually, we should talk about no relationship, at least in the short run, between the variables due to the aforementioned restrictions. However, we could check a long-run or equilibrium relationship between the expected depreciation of the domestic currency and the difference between the domestic interest rate and that of the United States, using co-integration analysis. The method used is that proposed by Engle and Granger and it consists in estimating the original regression and applying the Augmented Dickey and Fuller (ADF) Test to the residuals. As Gujarati (1995) points out, caution is required with the EngelGranger Test. Engle and Granger recalculated the critical significance values because they considered that the Dickey-Fuller’s criticals are not appropriate. The 1 per cent, 5 per cent, and 10 per cent critical values reported for two variables in the relationship and for a sample of 100 observations are –4.07, -3.37, and –3.03, respectively.7 In our case the result of the ADF test is –9.759710. We see that the null hypothesis is rejected and that the residuals are stationary. This confirms the existence of a long-run relationship between the variables involved. Thus, despite the fact of the presence of some constraints, we can confirm that there is financial integration between markets in Argentina and the United States. However, this integration is weak in the short run due to the above mentioned restrictions and due to the severity of the crisis in Argentina in 2001-2002. Coming back to the UIRP, in order to prove the analyzed relationship in a period of relative stability, we redefined the sample and included only the data from March 2003 to March 2006. The results are presented in Table 2:

7

See Urbisaia and Brufman (2001).

316

Ángel Enrique Neder, Julieta Schiro, Jonatan Saúl

Table 2: Regression results UIRP Argentina – US 03/2003 – 03/2006 Dependent Variable: SARG Method: Least Squares Sample: 2003M03 2006M03 Included observations: 37 Variable Coefficient Std. Error t-Statistic C 0.006630 0.004003 1.656185 DIFARGUS -0.277539 0.078015 -3.557495 R-squared 0.265566 F-statistic Adjusted R-squared 0.244583 Prob(F-statistic) Durbin-Watson stat 2.035655

Prob. 0.1066 0.0011 12.65577 0.001099

With fewer observations we obtained a smaller adjusted R² and, again, a constant coefficient non-significantly different from zero. However, the result of the coefficient of the interest rate differential is striking. The coefficient is statistically significant but has not the expected sign. One could argue that this indicates that a devaluation was expected. This would be in line with the behavior of the Central Bank (which consisted in issuing money to buy international reserves) in order to maintain the exchange rate on a relatively high level. The main policy adopted by the Central Bank was to sell bills and notes to cushion inflationary pressures.

B. The Case of Brazil Let us now come to the estimation of the UIRP for Brazil. As mentioned before, we use as domestic interest rate the interbanking rate of certificates of deposits on the secondary market and the CDs one month interest rate of the United States as the international interest rate of reference. The series is named DIFBRAUS. For the expected depreciation in the local money (dependent variable SBRA) we used the rate of change of the free exchange rate. Both variables are shown in Figure 3.8

8

The sample is considerably longer than that of Argentina because of the better quality of the data, particularly related to the exchange rate.

Financial Integration in Some Countries of South America

317

Figure 3: Brazil: Interest rate differential with US and expected currency depreciation .8

Basis points

.6

.4 D IF B RA US SB RA

.2

.0

-.2

1990 1992 1994 1996 1998 2000 2002 2004 Ye a rs

Both time series follow a similar trend. However, there appear some problems in the years where crises took place, particularly the devaluation in 1999, and the year 2002. The huge depreciation in January 1999 and the adoption of a free exchange rate regime, both provoked a deep modification in the management of monetary aggregates and in the domestic interest rate. The UIRP was examined using data from the third quarter of 1989 to the first quarter of 2006. Since the results showed the presence of autocorrelation we used the expected change in the exchange rate lagged one period as an additional explanatory variable. Table 3: Regression results UIRP Brazil - US Dependent Variable: SBRA Method: Least Squares Sample: 1989M08 2006M03 Included observations: 200 Variable Coefficient C 0.026486 DIFBRAUS 0.623087 SBRA(-1) 0.153309 R-squared 0.707245 Adjusted R-squared 0.704273 Durbin-Watson stat 1.917226

Std. Error t-Statistic 0.005956 4.447232 0.074246 8.392253 0.082570 1.856705 F-statistic Prob(F-statistic)

Prob. 0.0000 0.0000 0.0648 237.9588 0.000000

318

Ángel Enrique Neder, Julieta Schiro, Jonatan Saúl

All coefficients were statistically significant and the high R² points to a good fit. However, the presence of other aspects that explain the UIRP is evident; particularly, because the null hypothesis that the coefficient of the variable DIFBRAUS is equal to one is rejected. Thus, some restrictions connected with capital movements, the financial system, etc., are operating. The uncertainty that emerged after the presidential elections in 2002, with capital outflows and a decrease in the rate of economic growth may be an explanation. The most important financial effect was an increase in the domestic interest rate and this provoked an increase in the public debt. However, from the very beginning of 2003 the confidence of international markets was recovered. In 2004 the appreciation of the domestic money compensated (only in part) the devaluation that occurred in 2002. In order to confirm the existence of financial integration, we check the presence of co-integration between the expected depreciation of the domestic currency and the difference between the domestic interest rate and that of the United States. Again, we use the method proposed by Engle and Granger. The value obtained for the ADF test is –13.47818. Since the null hypothesis is rejected, we prove that the residuals are stationary, confirming the existence of a long-run relationship between the variables involved. Thus, despite the fact of the presence of some constraints, we can also confirm for the case of Brazil that there is financial integration between the domestic market and the United States.

C. The Case of Chile Our last Latin American country analyzed is Chile. The macroeconomic policy in Chile during the 1990’s had two main objectives, namely, to reduce the inflation rate and to limit the current account deficit. In order to reach these objectives, the monetary policy followed an inflation target.9 The main instrument used by the Central Bank was the open market operation based on short-run bills. Concerning the exchange rate policy, the main objective was to be a complement of the fiscal and monetary policy to reach external medium term equilibrium. More precisely, to have a real exchange rate consistent with a sustainable current account deficit. The exchange rate policy followed a “dirty-floating rate” and the 9

See Zahler (1997) and Rojas (2000).

Financial Integration in Some Countries of South America

319

Central Bank played a very important role in accumulating international reserves. This process of international reserves accumulation was favored by the growth in exports, and the increase in investments using profits, among others. In the Chilean case, we find a very weak, but positive and significant, short-run relationship between the expected exchange rate change and the difference between the domestic interest rate and the interest rate of the United States. Comparing the evolution of both variables (see Figure 4), they show the same trend. However, the equation fit is very poor. This could be explained by the rigid constraints for capital movements and by the policy followed by the authorities. Table 4: Regression results UIRP Chile - US Dependent Variable: SCHI Method: Least Squares Sample (adjusted): 1995M05 2006M01 Included observations: 129 after adjustments Variable Coefficient Std. Error t-Statistic C -0.001589 0.001589 -1.000185 DIFCHIUS 0.132981 0.076409 1.740390 R-squared 0.023294 Mean dependent var Adjusted R-squared 0.015604 S.D. dependent var S.E. of regression 0.017825 Akaike info criterion Sum squared resid 0.040351 Schwarz criterion Log likelihood 337.4686 F-statistic Durbin-Watson stat 1.573144 Prob(F-statistic)

Prob. 0.3191 0.0842 -0.001159 0.017966 -5.201063 -5.156725 3.028957 0.084213

The objective of maintaining the inflation rate and the external equilibrium required the intervention in the market by the Central Bank, using open market operations. The spread between interest rates generated a pressure on the exchange rate. Thus, the Central Bank had to constrain the inflow of capital (particularly short run capital). This explains why the UIRP in the short run is not verified.

320

Ángel Enrique Neder, Julieta Schiro, Jonatan Saúl

Figure 4: Chile: Interest Rate Differential with the US and expected currency depreciation .12

Basis points

.08

.04 SCHI DIFCHIUS

.00

-.04

-.08 95

96

97

98

99

00

01

02

03

04

05

Years

To verify the long run financial integration, we used the analysis of co-integration. The result for the ADF test is –9.155379. The null hypothesis is also rejected, indicating that the residuals are stationary, and confirming the existence of a long-run relationship between the variables involved. Thus, there is a financial integration between the Chilean domestic market and the United States. VI. Concluding Remarks The financial sector is important for a modern economy. In particular, the integration of financial markets plays an important role as an “engine” to promote economic growth. Generally speaking, we can observe two forms of financial integration: one arising from formal agreements and another from the de facto situation. The latter has been mostly prevailing in Latin America. Furthermore, integration in financial markets clearly mirrors integration in trade and investment. Financial integration can be tested in different ways. In this paper we used the UIRP to show the financial integration of three Latin American countries (Argentina, Brazil, and Chile) with the United States. A short-run financial integration could not be found. This was revealed in the values of the estimated parameters which had not the expected results. However, applying a co-integration analysis,

Financial Integration in Some Countries of South America

321

we could find long-run financial integration and we explain this situation by the restrictions in capital movements, financial restrictions and exchange rate policies adopted in Latin America.

References

De Brouwer, Gordon (1999), Financial Integration in East Asia, Cambridge University Press, United Kingdom. Engle, Robert F., Granger, Clive W. J. (1987) Co-integration and Error Correction: Representation, Estimation, and Testing, Econometrica 55 (2), 251-276. Flood, Robert P., Rose Andrew K. (2001), Uncovered Interest Parity in Crisis: The Interest Rate Defense in the 1990’s,  http://faculty.haas.berkeley.edu/arose/UIPC.pdf (last visit: June 2007). González-Páramo, José Manuel (2005), Integración financiera y crecimiento económico. Algunas lecciones desde la experiencia  http://www.ecb.int/press/key/date/2005/html/ europea, sp050128.es.html (last visit: June 2007). Gujarati, Damodar N. (1995), Basic Econometrics, McGraw-Hill, USA. Rojas, Patricio (2000), Política Monetaria y Cambiaria en Chile durante los Noventa, Estudios Públicos 78, Centro de Estudios Públicos,  www.cepchile.cl/dms/archivo_1911_59/ 229ximena.pdf (last visit: June 2007). Trichet, Jean-Cleaude (2006), The process of European financial integration,  http://www.ecb.int/press/key/date/2006/html/ sp060511_1.en.html (last visit: June 2007). Urbisaia, Heriberto L., Brufman Juana Z. (2001), Análisis de Series de Tiempo, Univariadas y Multivariadas, 2nd edition, Ediciones Cooperativas, Buenos Aires, Argentina. Zahler, Roberto (1997), La política macroeconómica de Chile en los años noventa: la visión del Banco Central,  http://www.eclac.org/cgi-bin/getProd.asp?xml=/publicaciones/ xml/3/4233/P4233.xml&xsl=/comercio/tpl/p9f.xsl&base=/comer cio/tpl/top-bottom.xsl (last visit: July 2007).

Part 4: Regional Issues

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata1,2

Analysis of Regional Income Mobility in Argentina3 Abstract We study regional income mobility in Argentina using annual data for the 1980 – 1998 period. The Hodrick-Prescott filter allows us to disentangle the cyclical from the trend components in the series. Annual transition matrices are estimated but other time-spans are also taken into account: 2, 3, 6, 9, and 18 years. An ergodic vector is obtained which points to an important degree of polarisation in the long-run. For the 1986 – 1998 period we obtain significantly different results because the long-run outcome now crucially depends on the initial conditions and, consequently, points to a much higher degree of polarisation. I. Introduction Figueras et al. (2003) presented empirical evidence that led to the rejection of the absolute beta convergence for the Argentine regions. In contrast, the hypothesis of conditional convergence was not rejected indicating that each territorial unit converges to its own long-run equilibrium. In Figueras et al. (2004) the Hodrick- Prescott filter was applied to eliminate cyclical fluctuations that were present in the original data series. The new empirical findings were much in line with those outlined above, that is, absolute convergence was rejected 1 2

3

We thank María Soledad Puechagut for her research assistance. We thank Ernesto Rezk, Gastón E. Utrera, Jorge M. Oviedo, Hugo Balacco, Elena Bárcena Martín and Roland Eisen for helpful comments and insight. The usual disclaimer applies. This is a revised version of “Análisis de la Movilidad Regional en Argentina: Un Enfoque basado en las Cadenas de Markov” by Arrufat et al., presented at the XLI AAEP Annual Meeting, La Plata, Argentina, November 2005.

326

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata

whereas conditional convergence at the regional level was not rejected. It has been repeatedly argued in the literature that the study of the dynamic characteristics of an economic system is seriously flawed if one resorts to the by now traditional approach to convergence. He pointed out that convergence to two very different income levels may arise whereby a bimodal (twin peaks) income distribution may arise in the long-run. It is precisely to the type of analysis put forward by Quah (1993) that we turn our attention in the present paper analysing annual regional data for the Argentine economy in the period 1980 – 1998. To make this approach operative, we proceed as follows: x First, sort the regions in ascending order according to per capita income levels for the given time-span adopted for the analysis. x Second, calculate the probabilities that govern the transitions of the territorial units between periods by means of Markov chains. There are two interesting papers that approached this issue before in Argentina: Utrera and Koroch (2000) (“Regional Convergence in Argentina: Empirical Evidence”), which analyzes the evolution of geographycal gross product for Argentine regions by using annual data for the period 1961 – 1994, and Garrido, et al. (2002) (“Dinámica de la distribución del producto a través de las provincias Argentinas”), which considers the period 1970-1995. Our approach in the present paper has two main features: i) By concentrating on the 1980 – 1998 period we are in a better position to address the significant structural and technological changes which affected Argentina during the nineties. ii) As stated above, our use of the Hodrick- Prescott filter should enable us to work with data which offers a better perspective to deal with long-run issues, that is, data not so much affected by the extremely common fluctuations that are a permanent feature of the statistical series usually found in Argentina. The rest of the present paper is organised as follows. In Section II we present an empirical overview which covers the 1980 – 1998 period and we also describe three different methods to estimate transition probabilities. In Section III we focus on the study of mobility for the 1986 – 1998 period which is more likely to provide a useful benchmark on which to base extrapolations for the distant future evolution of the Argentina economy. A brief sensitivity analysis is also considered. In Section IV we present our conclu-

Analysis of Regional Income Mobility in Argentina

327

sions. Appendix 1 is devoted to methodological issues, Appendix 2 to the gambler’s ruin problem and Appendix 3 is concerned with the diagonalisation of Markov matrices and the calculation of longrun equilibria. II. Empirical Overview 1980 – 1998 In all cases the relevant variable is GDP per capita relative to the national average. Cyclical fluctuations were eliminated using the Hodrick-Prescott filter. Table 1: Per capita GDP as a percentage of national GDP Lower bound

Upper bound

First group

27.621

52.639

Second group

52.639

73.157

Third group

73.157

116.598

Fourth group

116.598

422.553

It is straightforward that very sharp regional disparities are present in our data base. For example, the territorial units in the first group register a relative per capita income of between 27.62 per cent and 52.64 per cent of the national average whereas those in the fourth group have relative income ranging between 116.60 per cent and 422.55 per cent. The ratio between this latter figure and the 27.62 per cent of the lower bound of the first group is about 15. x Method I This first approach considers just an eighteen-year period to obtain the relevant Markov Matrix to model transitions between states. The transitional matrix in Table 2 can be interpreted as follows: In 1980, with respect to their Gross Geographical Product (relative to the national average), Catamarca, Chaco, Formosa, Santiago del Estero and San Luis were included in the first group. The lower and upper boundaries of this group are 27.621 – 52.639. At the end of the analysed period Catamarca, Chaco, Formosa and Santiago del Estero remained in the first group. Meanwhile, San Luis which had an extremely high growth rate, ended up in the fourth group the boundaries of which are 116.598 – 422.553. Analogous interpretations apply to the remaining cells in Table 2.

328

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata

Table 2:

Transition matrix, Analysis of Argentine regions over the 1980 – 1998 period Final State

Initial State

I

II

III

IV

I

Catamarca, Chaco, Formosa, and Santiago del Estero

None

None

San Luis

II

Jujuy and Tucumán

Entre Ríos, Misiones, Salta, and San Juan

La Rioja

None

III

Corrientes

Mendoza

Buenos Aires, Córdoba, andy Santa Fe

Neuquén

IV

None

None

Río Negro

Chubut, CABA, La Pampa, Santa Cruz, and T. del Fuego

Source: Own elaboration. CABA stands for Ciudad Autónoma de Buenos Aires, that is the City of Buenos Aires.

The Markov matrix for Table 2 is given as follows:

M=

0.800

0.000

0.000

0.200

0.286

0.571

0.143

0.000

0.167

0.167

0.500

0.166

0.000

0.000

0.167

0.833

It can be shown that this is a regular matrix with eigenvalues 1, 0.684 + 0.132i, 0.684 – 0.132i and 0.337. It is apparent that it has a single unit root and can be shown to be a regular matrix. The ergodic vector corresponding to the M matrix turns out to be equal to the following row vector:

V =

0.255 0.071 0.184 0.490

Regardless of the initial group in which a province is found, 25.5 per cent of the territorial units will converge to the first group, 7.1 per cent to the second one, 18.4 per cent to the third one, and finally 49 per cent to the fourth group. As the percentages associated with the first and fourth groups are substantially greater than the others, a very high degree of polarisation is an important feature of the long-run solution.

Analysis of Regional Income Mobility in Argentina

329

x Method II The relevant Markov Matrix is estimated as the product of eighteen one- year transition matrices which is given by:

M=

0.694

0.260

0.031

0.015

0.408

0.394

0.131

0.067

0.079

0.154

0.473

0.294

0.010

0.026

0.182

0.782

Here every state is accessible, so this matrix is not tridiagonal. It can also be shown to be regular and, as a consequence, it has only one eigenvalue equal to unity. The resulting ergodic vector is 0.313, 0.193, 0.176, 0.318. Therefore 31.3 per cent of the jurisdictions will converge to the first group (or state), 19.3 per cent to the second one, while 17.6 per cent and 31.8 per cent will do so to the third and fourth, respectively. x Method III This method was used by Quah (1996b), and estimates the Markov matrix as an average of 18 one-year transition matrices. It gives rise to the following transition matrix: 0.9717 0.0283 0.0000 0.0000

M=

0.0459 0.9358 0.0183 0.0000 0.0000 0.0183 0.9541 0.0276 0.0000 0.0000 0.0185 0.9815

This is a tridiagonal matrix showing a single period transition which for that reason is not comparable to the previous ones. To make it comparable it has to be raised to the 18th power which gives: 0.7038 0.2502 0.0394 0.0066 18

M =

0.4055 0.4118 0.1421 0.0406 0.0639 0.1421 0.4964 0.2976 0.0072 0.0273 0.2003 0.7652

The ergodic vector associated with it indicates that in the longrun equilibrium 31.74 per cent, 19.58 per cent, 19.58 per cent and 29.10 per cent of the jurisdictions will converge to the states I, II, III, and IV respectively. Once again there is strong evidence poin-

330

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata

ting to a significant degree of polarisation, although the effects are now somewhat weaker. The Shorrocks index value (See Shorrocks (1978)) associated to the M matrix is 0.43179. To summarise, methods I and II point to a strong degree of polarisation of the territorial units in the first and fourth states. In contrast, the calculations based on the Markov transition matrix obtained by method III show a much weaker degree of polarisation than that obtained by the other two methods. Which of these three alternative methods provides the best answer to the proper estimation of the probabilities of transition over time? If we now turn our attention to the main diagonals of the transition matrices obtained by using the first two methods, it is apparent that the elements contained in the second matrix are significantly smaller than their counterparts in the first. These results are in line with the empirical findings usually reported in the literature (Shorrocks,1976; Cantó, 2000; Bárcena Martín et al., 2004, for example) and may be taken as strong evidence that the underlying assumptions on which the one-period transitions are based (homogeneity, first-order Markov property, absence of state dependence, stationary probabilities of transition) are not strictly valid. Consequently, following the suggestions usually found in the literature, we chose to work with the matrix obtained by method I. It should also be stressed that we followed the advice given in Kremer et al. (2000) who propose to use different time frames for the computation of transition matrices for convenience. More specifically, for our purpose this calls for the factorisation of the eighteen period transition matrices as follows: only one matrix based on 18-year transitions, which is our benchmark case, just 2 matrices, each based on nine-year transitions, 3 matrices based on 6-year transitions, 6 matrices each based on 3-year transitions, 9 transition matrices each based on 2-year transitions, and, finally, 18 matrices each of which is based on 1-year transitions, which is given by our method II. None of these alternatives proved superior to our benchmark case based on only one matrix based on 18-year transitions. If we now compare the relative performance of methods I and III, we conclude that method I provides the best alternative. By inspection of the relative sizes of the transition probabilities located on the main diagonal we are in a position to state that those transition probabilities based on method III are systematically higher and,

Analysis of Regional Income Mobility in Argentina

331

therefore, point to a very low degree of mobility which is not present in the data. A careful analysis of the data for the period 1980 to 1998 reveals the presence of some striking features such as the case of San Luis. The available data for this province shows a remarkable upward trend, going from state one at the beginning to state four in the final year. It is reasonable to assume that such an extraordinary change is not likely to be repeated in the future. Quite on the contrary, we are more inclined to believe that behind that impressive performance which took San Luis’s gross geographical product from a mere 49.18 per cent in 1980 to a value as high as 128.66 per cent in 1985, and even more significant 195.33 per cent in 1998, all in terms of the national average, there were very powerful, hardly to be repeated in the future, circumstances which were working during this historical period. Equally striking is the fact that the Province of Mendoza ended up in state two in 1998, having started in state three in 1980. It is not our purpose here to provide a full explanation for this divergent performance of the two neighbouring provinces of San Luis and Mendoza. Dagnino Pastore, Costa & Asociados (1999) analysed the economic impact on Mendoza stemming from certain industry promotion fiscal regimes which were aimed to benefit the provinces of La Rioja, Catamarca, San Luis and San Juan. These researchers provide some evidence that the benefits arising to these territorial units could have had a significantly negative impact on Mendoza and might, at least, partly explain their relative performances during the period. To summarize, observed transitions during 1980 to 1998 which affect the Markov mobility matrices calculated in this section, may be regarded, to a great extent, as exceptional. It does not appear to be sensible to rely on those transitions in order to make long-run predictions as it would be tantamount to admitting that such conditions are likely to prevail in the future. To try to remedy this point, two approaches were taken. First, we decided to drop San Luis from our sample, given that it is a relatively small province with a population of approximately 1 per cent of the total population of Argentina. The results we obtained to remove a possible bias were extremely frustrating because the ergodic vector was calculated to be equal to: 1

0

0

0

332

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata

This result amounts to saying that 100 per cent of the remaining territorial units should be found in state I in the long-run. Needless to say this is absurd because there is no satisfactory explanation which could provide a rationale for all jurisdictions having a gross geographical product which amounts to a small fraction of the national, country-wide average. Second, given the results just quoted we decided to analyse the period 1986 to 1998. The rationale for this is that San Luis had reached the fourth state in 1985 and, therefore, what we have argued was unlikely to be repeated in the future so this province’s performance was no longer a source of concern to us. In the next section we turn our attention to the results obtained following this approach. III. Analysis of the 1986 – 1998 period In this section we restrict our analysis to the 1986 – 1998 period. This matrix has five special features worth noting: i) It is not tridiagonal as is the case in most of the examples analyzed in the literature. ii) It is block-diagonal, given that the first and second group form a block whereas the third and fourth belong to a different one. iii) In each block there is an absorbing state, being I for the first block and IV for the last one. iv) The states corresponding to the groups II and III are transitory. v) Its trace is equal to 3.5476. As a consequence the Shorrocks index attains an approximate value of 0.15 which indicates very low mobility.

Analysis of Regional Income Mobility in Argentina

333

Table 3: Transition matrix of regions 1986 - 1998

Final State – 1998 Initial State 1986

I

II

III

IV

I

Catamarca, Chaco, Formosa, Jujuy, Santiago del Estero and Tucumán

None

None

None

II

Corrientes

Entre Ríos, Mendoza Misiones, Salta and San Juan

None

None

La Pampa and Neuquén

Chubut, CABA, San Luis, Santa Cruz and T. del Fuego

III

None

None

Buenos Aires, Córdoba, La Rioja, Río Negro and Santa Fe

IV

None

None

None

Source: Own elaboration.

The new transition matrix M is as follows:

§ ¨ ¨ ¨ ¨ M  ¨ ¨ ¨ ¨ ¨ ©

6 6

0.0

1

5

6

6

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

5

2

7

7

0.0

5 5

¸· ¸ ¸ ¸ ¸ ¸ ¸ ¸ ¸ ¹

334

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata

The eigenvalues of M are shown in the following column vector: vv = [ 1, 0.83333, 1, 0.71428571 ]. Two points are worth emphasising. First, all four eigenvalues are real, in contrast with the previous case in which a complex conjugate pair was obtained. Second, and more important, the M matrix has now two eigenvalues equal to 1 while the remaining ones have modules strictly less than one. The successive powers of M do not converge to a matrix with identical rows. In other words, there is not just a single long-run equilibrium vector. For instance, the M matrix raised to 1500 turns out to be equal to:

M

1500

§1 ¨ ¨1 ¨0 ¨ ¨0 ©

0 0 0 0

0 0 0 0

0· ¸ 0¸ 1¸ ¸ 1 ¸¹

This calculation shows that in order to predict the probability that a specific territorial unit will be found in a given group in the long run, its initial state has to be taken into account. For instance, a territorial unit which was initially (in 1986) found in either group I or II, has a probability equal to 1 of finding itself trapped within group I in the long-run. By way of contrast, starting in III or IV leads a territorial unit to end up in group IV in the long-run. It is easy to show that given that 6 territorial units belonged to the first group, 6 to the second, 7 and 5 to the third and fourth states in the initial (1986) distribution, the following long-run distribution results: an equal number of 12 territorial units will be found in groups I and IV which is a very strong indication of a remarkably high degree of polarization. Given that transition probabilities that lie at the heart of the Markov matrix we are using are estimated, it appears to be reasonable to conduct a simulation exercise. By this we intend to answer the following question: What would happen if matrix M in this section were to be subject to a small perturbation which might arise, for example, from the presence of measurement errors in the gross geographical product? In the following we consider two alternative transition probability matrices, as follows:

Analysis of Regional Income Mobility in Argentina

335

Case 1: We assume that the relevant Markov matrix is the following:

§ 5 ¨ ¨ 6 ¨ 1 ¨ 6 M ¨ ¨ 0.0 ¨ ¨ ¨ 0.0 ©

1 6 5 6

0.0 0.0 ·¸ 0.0

0.0

5 7

0.0 0.0

¸ ¸ 0.0 ¸ ¸ 2 ¸ 7 ¸ 5 ¸ ¸ 5 ¹

The resulting new long-run equilibrium matrix is as follows:

M

1500

§ 0.5 0.5 0 0 · ¨ ¸ ¨ 0.5 0.5 0 0 ¸ ¨ 0 0 0 1¸ ¨ ¸ ¨ 0 0 0 1 ¸¹ ©

It is straight forward that: x All territorial units starting in either state 1 or 2 have a 50 per cent chance of ending up in state 1 or 2. x All units starting in either state 3 or 4 end up in state 4 Case 2: We assume that the Markov matrix is equal to:

§ 6 ¨ ¨ 6 ¨ 1 ¨ 6 M ¨ ¨ 0.0 ¨ ¨ ¨ 0.0 ©

0.0 0.0 0.0 ·¸ 4

1

6

6

0.0

5 7

0.0 0.0

¸ ¸ 0.0 ¸ ¸ 2 ¸ 7 ¸ 5 ¸ ¸ 5 ¹

It can be shown that the resulting long-run equilibrium matrix is equal to: § 1 0 0 0 · ¨ ¸ 1500 ¨ 0.5 0 0 0.5 ¸ M ¨ 0 0 0 1 ¸ ¸ ¨ ¨ 0 0 0 1 ¸ © ¹

336

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata

This indicates that: x Units starting in state 1 end up in 1 (absorbing state) x Units starting in state 2 may either end up in 1 or 4 with a 50 per cent chance x Units starting in either 3 or 4 end up in 4 It is interesting to measure the degree of polarisation which results from these two alternative transitions matrices. In Table 4 we have proxied polarisation by measuring the coefficient of variation for the long-run solution vector. Table 4: Long-run variation of geographical gross product 19361998 with different transition matrices Matrix original case 1 case 2

Coef. Variation 1.3544 1.2729 1.5135

IV. Conclusions and closing remarks

Our empirical findings suggest that not every defined group turns to be an accessible or communicated state because it is also the case that absorbing and transitory states arise. Focussing on long-run equilibria, two major results are found. When the complete time period (1980 – 1998) was, considered, one ergodic vector appeared as the solution indicating that we are in the presence of an ergodic Markov chain. In contrast, when the analysis was restricted to the period 1986 – 1998, a very different result was obtained. The initial state matters then in order to predict the longrun outcome. As such, this latter result points to a very severe degree of polarization rather than to convergence.

References

Arulampalam, Wiji, Booth, Alison L., Taylor, Mark P. (2000), Unemployment Persistence, Oxford Economic Papers 52 (1), 24-50. Arrufat, Josè Luis, Díaz Cafferata, Alberto M., Figueras, Alberto J. (2001), Apertura, integración y tendencias regionales de la

Analysis of Regional Income Mobility in Argentina

337

desocupación en Argentina, in: Mancha, Navarro, Sotelsek, Salem (eds.), Convergencia económica e integración, Ediciones Pirámide, Madrid, 313-348. Arrufat, José Luis, Díaz Cafferata, Alberto M., Figueras, Alberto J. (2000), Regional labor markets, The rate of participation, Argentina 1980- 1998, Annual Meeting of the Economic Association of Argentina, Universidad Nacional de Cordoba. Bárcena, Martín Elena, Fernández, Morales A., Lacomba Arias B., Martín, Reyes G. (2004), Dinámica de la pobreza a corto plazo en España y Reino Unido a través del Panel de Hogares Europeo, Estadística Española 46 (157), 461-488. Barro, Robert, Sala-i-Martín, Xavier (1992), Convergence, Journal of Political Economy 100, 223-251. Cantó, Olga (2000), Climbing out of Poverty, Falling Back in: Low Income’s Stability in Spain, Applied Economics, 34 (15), 1903 – 1916. Cappellari, Lorenzo, Jenkins, Stephen (2003), Transitions between unemployment and low pay, May, mimeo. Cuadrado Roura, Juan R., Parellada, Marti (2002), Regional Convergence in the European Union, Springer, Heidelberg. Cuadrado Roura, Juan R., García-Greciano, Begoña, Raymond, José Luis (1999), Regional productivity and productive structure: the Spanish case, International Regional Science Review 22 (1), 35- 53. Cuadrado, Roura Juan R., Mancha Navarro, Tomas (1999), Política regional y de cohesión, in: Jordan Galduf, Josep Maria, Antuñao, Maruri Isidro (eds.), Economía de la Unión Europea, third edition, Civitas Ediciones, 429-471. Cuadrado, Roura Juan, Mancha Navarro, Tomas, Garrido, Yserte Rubèn (1998), Convergencia regional en España, Fundación Argentaria, Madrid. Dagnino, Pastore, Costa & Asoc. (1999), Impacto económico de los regímenes de promoción industrial en La Rioja, Catamarca, San Luis y San Juan, Buenos Aires. Díaz Cafferata, Alberto M., Figueras, Alberto J., Arrufat, José L. (2000), Integration and regional restructuring, Fourth Arnoldshain Seminar, Frankfurt, Mimeo.

338

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata

Figueras, Alberto J., Arrufat, José L., Descalzi, Ricardo L., Rubio, Alberto A. (2002), Regional Analysis Based on Indicators, Fourth Arnoldshain Seminar, Frankfurt. Figueras, Alberto J., Arrufat, José L., Regis, Paulo J. (2003), El fenómeno de convergencia regional: una contribución, XXXVIII Asociación Argentina de Economía Política, Mendoza. Figueras, Alberto J., Arrufat, José L., De la Mata, M. Dolores, Álvarez, Sebastián (2004), Convergencia Regional: une estudio sobre indicadores de tendencia, XXXIX Annual Meeting of the Economic Association of Argentina, Buenos Aires. Friedman, Milton (1992), Do old fallacies ever die?, Journal of Economic Literature 30 (4), 2129-32. Garrido, Yserte Rubén (2002), Cambio Estructural y desarrollo regional en España, Pirámide, Madrid. Garrido, Yserte Rubén, Marina, Adriana, Sotelsek, Daniel (2002), Dinámica de la distribución del producto a través de las provincias argentinas (1970-1995), XXXVII Annual Meeting of the Economic Association of Argentina, Tucumán. Hillier, Frederick, Lieberman, Gerald (1997), Introducción a la investigación de operaciones, Mc.Graw Hill., Mexico. Kangasharju, Aki (1999), Relative economic performance in Finland: regional convergence, Regional Studies 33, 207-217. Kremer, Michael, Onatski, Alexei, Stock, James (2000), Twin Peaks or Lake Wobegon?, Testing Hypotheses On the Dynamics of World Income Distribution, Harvard University, mimeo. Krugman, Paul (1991), Increasing returns and economic geography, Journal of Political Economy 99 (3), 183-199. Krugman, Paul (1995), Development, geography and economic theory, MIT Press. Marina, Adriana (2001), Convergencia económica en Argentina, in: Mancha, Novaro, Sotelsek, Daniel (eds.), Convergencia economica e integración: La experiencia en Europa y América Latiná, Madrid, Pirámide. Moreno, Rosia, Vayá, Esther (2002), Econometría espacial: nuevas técnicas, Investigaciones Regionales, Asociación Española de Centros Regionales, Universidad de Alcalá, August.

Analysis of Regional Income Mobility in Argentina

339

Quah, Danny T. (1993), Galton´s fallacy and tests of the convergence hypothesis, Scandinavian Journal of Economics, (95) 4, 445-66. Quah, Danny T. (1995a), Convergence empirics across economies with (some) capital mobility, Centre for Economic Performance Discussion Paper No. 257, London. Quah, Danny T. (1995b), Regional Convergence clusters across Europe, Centre for Economic Performance Discussion Paper No. 274, London. Quah, Danny T. (1996a), Empirics of economic growth and convergence, European Economic Review 40 (6), 1353-1375. Quah, Danny T. (1996b), Twin peaks: growth and convergence in models of distribution dynamics, The Economic Journal 106 (437), 1045-55. Quah, Danny T. (1997a), Empirics for growth and distribution: stratification, polarization, and convergence clubs, Centre for Economic Performance, London, Discussion Paper No. 324, January. Quah Danny T. (1997b), Regional cohesion from local isolated actions: I. Historical outcomes, Centre for Economic Performance, Discussion Paper No. 378, London, December. Quah, Danny T. (1997c), Regional cohesion from local isolated actions: II. Conditioning, Centre for Economic Performance Discussion Paper No. 379, London. Ray, Debraj (1998), Economía del desarrollo, Bosch, Barcelona. Redwood, John (1991), Reversión de polarización, ciudades secundarias y eficiencia en el desarrollo: una visión aplicada al Brasil, Revista Eure 32. Sala-i-Martín, Xavier (1999), Apuntes de Crecimiento Económico, Bosch, Barcelona. Shorrocks, Anthony F. (1976), Income mobility and the Markov assumption, The Economic Journal 86, 566-578. Shorrocks, Anthony F. (1978), The measurement of mobility, Econometrica 46, 1013-1024. Simon, Carl P., Blume, Lawrence (1994), Mathematics for Economists, W.W. Norton, New York.

340

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata

Stewart, Mark B., Swaffield, Joanna K. (1999), Low Pay Dynamics and transition Probabilities, Economica, 66 (261), 23-42. Utrera, Gastón E., Koroch, Javier A. (1998), Convergencia: evidencia para provincias argentinas, XXXII Annual Meeting of the Economic Association of Argentina, Mendoza. Utrera, Gastón E. (1999), El crecimiento económico en Latinoamérica, XXXIV Annual Meeting of the Economic Association of Argentina, Rosario. Utrera, Gastón E., Koroch, Javier A. (2000), Regional Convergence in Argentina: Empirical Evidence, XXXV Annual Meeting of the Economic Association of Argentina, Córdoba. Willington, Manuel (1998), Un análisis empírico del crecimiento económico regional en Argentina, Estudios IERAL XXI(84).

Appendix 1: Methodology In what follows we briefly summarise some of the most salient features of the methodology employed in the empirical literature in order to tackle the problems arising from transition probabilities which are affected by path dependence, individual hereterogeneity and also the need to control for initial conditions. We first consider the important paper of Cappellari et al. (2003) who argue that there is substantial interest in the United Kingdom to quantify the likely existence of a cycle which involves transitions between low pay and no pay (“low pay / no pay cycle”). In other words, workers alternate between periods in which they earn low wages with other periods in which they earn no income. To what extent is it the case that those who earn low wages face a higher probability of experiencing a spell of unemployment vis-à-vis those other workers who earn high wages? Conversely, have the unemployed a greater probability of earning of earning a low wage rather than a high wage if they get a job? The authors propose and estimate an econometric model to explain the annual transitions between unemployment, low pay and high-wage employment. They resort to a first order Markov chain model which takes into account the endogeneity of initial conditions, the likely presence of selection bias, as well as sample attrition. They used annual data on

Analysis of Regional Income Mobility in Argentina

341

males taken from The Household Panel Survey which covers the period 1991-2000. The most important empirical results point to the need to take into account the three likely selection bias sources already mentioned if one is to model successfully the transitions observed in the labour market. Low pay males face a greater probability of experiencing a spell of unemployment than do their high pay counterparts. They also find that the unemployed have a greater likelihood of getting low wage jobs vis-a-vis high wage workers. Observed transitions between unemployment and low pay are strongly related to poor educational background. The authors also found empirical evidence which suggest that previous experience of people subject to unemployment or low pay greatly increased the probability of getting trapped in those states, which is an additional effect which arises even after taking individual heterogeneity into account. In other words their empirical findings point to the existence of significant individual heterogeneity and also of state dependence. This latter effect points to the fact that previous unfavourable spells in the labour market significantly increase the likelihood of facing similar problems in the future, even after taking individual heterogeneity into account. We now turn our attention to the interesting paper by Arulampalam et al. (2000). They resort to panel data econometric techniques to estimate the incidence of unemployment on males using data provided by the British Household Panel Survey. Some important econometric features of their research are worth mentioning. First, non-observable individual heterogeneity. Second, true state dependence. Third, the problems arising with the initial conditions. All three aspects are dealt with by the authors who were able to find strong empirical evidence of state dependence which is in line with the predictions of the scarring theory of unemployment and argue that previous unemployment spells have negative implications for future labour market perspectives. Accounting for these empirical findings the authors go on to suggest policies to lower short term unemployment because this will, in turn, bring about a reduction in the natural rate of unemployment.

342

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata

Appendix 2: The gambler’s ruin We present a deliberately simple example of a first order Markov chain. Consider two players (A and B) whose combined capital is $4. If player A starts with capital equal to zero, he is trapped into an absorbing state because he is not allowed to bet when he has no capital. If, on the contrary, player A starts with $1, he faces a 50-50 probability of ending up with $0 or $2, provided a fair coin is used. Likewise, if player A starts with $2, he may end up with either $1 or $3, with equal probability. If we repeat this line of reasoning, he may end up with either $2 or $4 if the starts to play with an initial capital of $3. In contrast, if he starts with $4, this means his opponent (player B) has no capital and therefore is not in a position to bet. Consequently, player A ends up with capital equal to $4, the capital with which he started. The transition probabilities may be summarised as follows: 1

M=

0

0

0

0

0.5 0 0.5

0

0

0

0.5 0

0.5

0

0

0

0.5

0

0.5

0

0

0

0

1

whose eigenvalues are given by the following column vector: -0.707 0 0.707 1 1 It is easy to see that the matrix M is not a regular Markov chain and that it has two eigenvalues equal to unity.

Analysis of Regional Income Mobility in Argentina

343

We now show the second power of matrix M which shows the results one obtains when one takes into account the results after two successive bets: 1 0 0 0 0 2

M=

0.5

0.25 0

0.25

0

0

0.5

0.25 0

0

0

0

0.25

0

0

0.25

0.25

0.5

0

1

By calculating the successive powers of M one may easily show that M is not a regular Markov chain because there is no power of it for which all entries are strictly positive. The calculation of the long-run equilibrium solution for the outcomes of this game necessitates the evaluation of successive powers of M until convergence is achieved. This is the case when one obtains a matrix in which entries remain fixed between two successive iterations. In our case, we get the following outcome when raising M to power 100: 1 0 0 0 0

100

M =

0.75 0

0

0

0.50

0

0

0 0.50

0.25 0

0

0

0.75

0

0

0

1

0

0.25

The following comments are in order: i) If player’s A initial capital is $0, it is obviously certain that he will end up bankrupt because, as said before, he who owns no capital is not allowed to bet. This piece of information is displayed in the first row of the matrix. ii) If player’s A initial capital is $1 he faces a probability of ¾ of going bankrupt because his opponent, player B, has a capital of $3 and a probability amounting to only ¼ of A’s winning and ending up with $4. This is the information contained in the second row of the matrix.

344

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata

iii) If A’s initial capital is $2 there is a 50-50 per cent change that either A or B go bankrupt. The relevant probabilities are displayed in the third row of the matrix. iv) If A’s initial capital is equal to $3 is tantamount to stating that B’s capital is only $1. There is now a probability of ¼ that A goes bankrupt and consequently player B face a ¾ probability of going bankrupt. This information is shown in the fourth row of the matrix above. v) Finally, if player A were endowed with a capital of $4, that of B would amount to $0 given our assumption of total capital being equal to $4. Since it is B who is not allowed to bet, A is certain to keep $4, which is the information contained in the fifth row of the matrix. The key point to highlight is that the successive powers of the Markov matrix M for the present problem do not give rise to a new matrix with identical rows. In other words, the long run equilibrium does depend on the initial state. All the remaining features of the mathematical model employed here are perfectly appropriate. In particular, no heterogeneity problem is present here because there is simply no room in the analysis for the transition probabilities to change from one bet to the next. It should also be noted that the no error-measurement problem arises either as the capitals of player A and B are correctly measured. Finally, the assumption that for each coin toss there is a 50-50 per cent chance of winning or losing is appropriate on account of the fair coin assumption adopted at the outset. There is therefore no room, or need for that matter, to consider alternative models like, for example, a Markov matrix of order greater than 1 or the introduction of state dependence of any sort.

Analysis of Regional Income Mobility in Argentina

345

Appendix 3: The Diagonalisation of Markov Matrices and the Calculation of Long-run equilibria We take the following transition matrix as our starting point:

§ 4 ¨ ¨ 5 ¨ 2 ¨ 7 M ¨ ¨ 1 ¨ 6 ¨ ¨ 0.0 ©

0.0 0.0 4

1

7

7

1

3

6

6

0.0

1 6

· ¸ 5 ¸ ¸ 0 ¸ ¸ 1 ¸ 6¸ 5¸ ¸ 6¹ 1

This Markov matrix shows the transition probabilities among states I to IV between the endpoints of the period 1980 to 1998. It can be easily seen that the row sums of this matrix are equal to 1, a typical feature of all Markov matrices. Even though the matrix is no symmetric, we will show that it can be diagonalised. Its eigenvalues turn out to be equal to: 1, 0.6838275 + 0.13177917i, 0.6838275 - 0.13177917i and 0.33710691, which are stored in the column vector w. This vector´s four elements occupy positions (0,0), (1,0), (2,0) and (3,0), respectively. Note that here we are using the convention of denoting the first row and column by 0, instead of 1. The matrix M above is regular and therefore has a single eigenvalue equal to 1. All the remaining eigenvalues have modulus less than 1. Next, we define a diagonal matrix D which contains the corresponding eigenvalues of M in the order already shown. Let us define matrix P, whose columns contain the eigenvectors of the transpose of M, as follows:

346

P

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata 0.74384961 0.74384961 0.06152676 · § 0.43506232 ¨ ¸ ¨ 0.12181745  0.12217  0.28666i  0.12217  0.28666i  0.55417119 ¸ ¨ 0.31324487  0.30905  0.09672i  0.30905  0.09672i 0.77912588 ¸ ¨ ¸ ¨ 0.83531966  0.31263  0.38338i  0.31263  0.38338i  0.28648145 ¸ © ¹

Using the ordinary procedure M’ (transpose of M) can be factored as follows: P*D*inv(P) = M’ , or, alternatively, by transposing this result one obtains: MM = (P*D*inv(P))’, which obviously reproduces the original matrix M. An important advantage of expressing M in factor form as in MM lies in the fact that it is a fairly straightforward exercise to calculate powers of M, including matrix square roots, cubic roots, etc. We calculate Dmedio (D raised to one half), which requires raising M’s eigenvalues to one half: ª  vv 0  0 0.5 º 0 0 0 « » « » 0 0 0 vv 1  0 0.5 Dmedio  « » 0.5 0 0 0 vv 2  0 « » « 0.5 » 0 0 0 vv 3  0 ¼ ¬

and, after that, we need to perform the following calculation: MMedio = (P*Dmedio*inv(P))’ The matrix MMedio is equal to: 1.02220e  03  7.41192e  03 0.11137974 · § 0.89500999 ¸ ¨  0.01737455 ¸ 0 . 16811925 0.74789751 0.10135778 ¨ MMedio ¨ 0.0930166 0.11620635 0.69177216 0.09900489 ¸ ¸ ¨ ¨  4.71632e  03  7.34483e  03 0.10466447 0.90739668 ¸¹ ©

It should be noted that the procedure we adopted here suffers from some minor numerical anomalies. This matrix square root should be a proper Markov matrix, and as such, all its entries should be non-negative. In the calculation we are reporting, 4 out of 16 elements, that is one quarter of the entries in MMedio are negative albeit very small, in the order of the thousandths. In future

Analysis of Regional Income Mobility in Argentina

347

research we intend to approach these calculations in a more satisfactory manner. Obviously, the matrices MMedio and MM have the same eigenvectors. MMedio*MMedio = (P*Dmedio*inv(P))’*(P*Dmedio*inv(P))’ Expansion of the previous formula gives rise to: MMedio*MMedio = inv(P’)*Dmedio*P’*inv(P’)*Dmedio*P’ In which we have taken into account that Dmedio is a diagonal and consequently also symmetric matrix. MMedio*MMedio = inv(P’)*Dmedio*Dmedio*P’ Which turns out to be equal to: MMedio*MMedio = inv(P’)*D*P’ Given that, by definition, Dmedio*Dmedio equals D. Bear in mind that: MM = (P*D*inv(P))’ After some manipulations, MM = inv(P’)*D*P’ And, finally, one obtains MMedio*MMedio = MM. It is evident that the same procedure outlined here may be adapted to calculate a “cubic root” or the “nth root” for that matter of a given matrix. In order to obtain the 18th root of MM, denoted by MM18avo, for example, we can proceed as follows: MM18avo = (P*D18avo*inv(P))’ The P matrix has already been defined whereas D18avo is as follows: 1 ª «  vv 0  0 18 0 0 0 « 1 « « 0 0 0  vv 1  0 18 D18avo  « 1 « 0 0 0  vv 2  0 18 « « 1 « 0 0 0  vv 3  0 18 ¬

º » » » » » » » » » ¼

348

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata

It is straightforward to show that: (MM18avo)^18 = ((P*D18avo*inv(P))’)^18 Number of matrices

Number of periods used in the calculation of each transition matrix

Power of D

1 2

18 9

1 ½

3 6

6 3

1/3 1/6

9

2

1/9

18

1

1/18

which in turn gives rise to: (MM18avo)^18 = (inv(P’)*D*P’) and, finally to: (MM18avo)^18 = MM The “matrix roots” so obtained refer to: one half, one third, one sixth, one ninth and one eighteenth because when one examines transition matrices for the period 1980 to 1998, only a maximum of eighteen annual transitions can be calculated. It is therefore possible to break those eighteen transitions as 2 9-year transitions, 3 6-year transitions, 6 3-year transitions, 9 2-year transitions or, finally, 18 one-year transitions. All ergodic vectors associated with these matrices are identical. In our numerical example, the ergodic row vector one obtains is the following: [ 0.2551 0.0714 0.1837 0.4898 ] This vector can be interpreted as follows. Regardless of the initial state in 1980, and on the assumption that M remains fixed throughout, 25.51 per cent of the territorial units will converge to state I, whereas 7.14 per cent, 18.37 per cent, and, 48.98 per cent, respectively, will end up in state II, III, and, IV. A slight difference emerges when we turn our attention to the period 1986 to 1998 because now only 13 years are taken into account. Consequently, one can only calculate one matrix for a 12year-period transition, 2 matrices based on 6-year-transitions, and so forth. The following table provides the details.

Analysis of Regional Income Mobility in Argentina

Number of matrices 1 2 3 4 6 12

349

Number of periods used for the calculation of each matrix 12 6 4 3 2 1

In each of these cases, the power to which the D matrix must be raised is found by taking the reciprocal of the number of matrices on which each particular breakdown is based. The Markov transition matrix for the period 1986 to 1998, based on 12-year transitions, calculated on the basis of the same states as before, is the following:

M

0 0 · §6/ 6 0 ¸ ¨ 0 ¸ ¨ 1/ 6 5 / 6 0 ¨ 0 0 5/ 7 2/ 7¸ ¸ ¨ ¨ 0 0 0 5 / 5 ¸¹ ©

Its eigenvalues are stored in the following column vector: vv = [ 1, 0.83333, 1, 0.71428571 ]. Two points are worth mentioning. First, all the eigenvalues are real as opposed to the previous case in which a pair of complex conjugate values were present. Second, the new M matrix has now not one but two eigenvalues equal to unity. It can be checked that the M matrix is not a regular Markov matrix. The appropriate calculations show that the matrix P is now:

P

§ 1 0.70710678 ¨ ¨ 0  0.70710678 ¨0 0 ¨ ¨0 0 ©

0 0 0

0 0 0.70710678

· ¸ ¸ ¸ ¸ 1  0.70710678 ¸¹

Note that this matrix is block-diagonal. The diagonal matrix D, as before, has the eigenvalues stored on the main diagonal in the order indicated above.

350

Jòse Luis Arrufat, Alberto J. Figueras, Valeria J. Blanco and M. Dolores De La Mata 0 0 · § vv 0  0 0 ¨ ¸ vv 1  0 0 0 0 ¨ ¸ D ¨ 0 vv 2  0 0 ¸ 0 ¨ ¸ vv 3  0 ¹ 0 0 © 0

Once again, it is possible to factorize the matrix M by expressing MM as follows: MM = (P*D*inv(P))’ Once MM (equal to M) has been factorised out in this fashion, all relevant powers or roots can be easily calculated as shown above. The calculations of successive powers of M give rise to a new matrix whose rows are not all identical. In other words, for this new example, one is not able to find an ergodic vector showing the probability of convergence to each state regardless of the initial state. For this case the opposite prevails: such probabilities are path dependent in the sense that they crucially depend on which initial state a territorial unit is found to belong.

Jorge Alberto Fornero1

Do Mercosur Countries Converge in Per Capita GDP and Productivity? Abstract This paper tests the hypothesis of income convergence among a group of Latin American countries. In particular, we ask whether the deepening of the integration process, which began at the end of the 1980s, enhanced convergence. The focus is placed on the empirical performance of GDP per capita, GDP per capita adjusted by terms of trade and GDP per worker of the Mercosur Countries and its associates. The speed of convergence is estimated with panel data analysis and the influence of alternative initial conditioning variables is explored. Time series and clusters analyses provide complementary evidence of income convergence clubs. I. Introduction Why are some countries richer than others? Why do some countries grow and develop more rapidly? These important questions were first addressed by the pioneering work of Kaldor (1963), who synthesized the “stylized facts” of economic growth. Essentially, this is a well-founded description of the United States steady-state income, which identifies empirical regularities using long data records. As soon as more countries were included in international databases, the literature on international growth comparisons flourished. This literature aimed to test one of the main theoretical implications of the standard neoclassical model: the so-called convergence hypothesis. In words, it states that given two countries with similar technology and savings rate, the one that is farthest away from the steady state will grow faster than the other and, eventually, they will converge to a common steady-state path. Although this hypothesis has been 1

I thank the European Union Alfa Program for financial support; project AML/B7-311-97/0666/II-0058-FA, Eulalia. I also thank Gabriele Tondl for helpful suggestions that substantially improved this paper. The usual disclaimer applies.

352

Jorge Alberto Fornero

widely accepted in groups of countries with similar per capita incomes, it has been rejected in samples that include heterogeneous countries. In particular, the majority of studies report that developing countries’ growth rates performed modestly compared with developed countries. In order to reconcile this evidence with the theory, a revision of neoclassical models and their assumptions took place. Several studies argued that if rivalry in technology were observed, income convergence would not be verified. Many features can potentially prevent technology from being non-rival in the models; for example, the introduction of imperfect markets and information would generate monopolistic power and supernormal profits and would increase the incentives to invest in purposeful research and development. Endogenous research and development investment will lead to endogenous technological progress and, as a result, convergence may not be observed.2 The neoclassical and endogenous growth models were broadly investigated in the empirical growth literature. The econometric techniques applied can be classified in four strands. First, in the early 1990s, the methodology used comprised cross-section growth regressions, where income growth is explained by initial income and technology, labor and other inputs. In this framework, a negative (zero) correlation between the initial income and income growth is interpreted as evidence of convergence (non-convergence). These results proved to be weak for at least three reasons: (i) the assumption of homogeneous coefficients in the regressions turned out to be too restrictive when the number of countries in the sample is large; (ii) the unobserved initial technology levels were omitted in the regressions, making their interpretation troublesome; and (iii) endogeneity of independent variables with the error term. Second, a substantial improvement resulted from the addition of the time dimension to the cross-section regressions in order to build static and dynamic panel data models. Panel data techniques successfully estimate the omitted initial level of technology. Third, regardless of what variables cause growth, time series analysis investigates whether convergence occurs in the relative income of: (i) two countries (to test leadership), or (ii) a country relative to the average income of a group (to test convergence clubs). In effect, evidence of unit roots in relative income series is consistent with 2

Pioneering studies are Romer (1990) and Aghion and Howitt (1992).

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

353

non-convergence of income. Formerly, this literature implemented country by country (or univariate) unit roots test and soon after these tests were generalized to a multivariate setting. Fourth, Quah (1996b) constructs a Markov switching model to explain the evolution of the income pattern described in Quah (1996a).3 In his proposal, the current state of the income distribution predicts the next period state with an estimated probability. The probabilities constitute entries of the transition probability matrix, which is crucial for the Markov chain updating process.4 Quah’s (1996b) study did not find any evidence of convergence in the cross-section distribution, challenging the implications of the neoclassical model. All these econometric methods were intensively applied in empirical research to provide an answer to the validity of the convergence hypothesis and, indirectly, which of the alternative growth models had more empirical support. Bearing in mind the theoretical aspects regarding economic growth and the appropriate empirical methods, this paper has a twofold objective. First, we briefly review and summarize relevant concepts and the major econometric methods that have been adapted to test income convergence. Second, we ask whether there is evidence of income and productivity convergence, with special focus on the speed of convergence of the Mercosur countries and its associates.5 This paper investigates the convergence idea by employing: (i) panel data analyses to estimate the so-called conditional Econvergence; and (ii) time series analysis to shed light on the existence of clubs of income convergence, asking whether any of Mercosur’s partners acts as a growth leader within the bloc. In 3

4

5

From his empirical analyses Quah (1996a) concludes that per capita incomes tend to cluster around high and low levels, leading to a clear rejection of the convergence hypothesis. The cross-section of income levels is rationalized by a (timedependent) bimodal distribution. Quah assumes that countries’ initial income level can be classified into two regimes, corresponding to wealthy or poor countries. Mercosur (Southern Common Market) formally began with the Asunción Treaty signed by Argentina, Brazil, Paraguay and Uruguay in 1991 to form a free trade area —and, ultimately to create a common market. Venezuela joined in 2007. Bolivia and Chile as associated countries negotiated tariff reductions in 1996. Peru and Mexico have also announced their wish to become members.

354

Jorge Alberto Fornero

particular, we report results on three alternative series that may serve as income proxies, namely: real per capita gross domestic product (GDPC), real per capita gross domestic product adjusted by terms of trade (GDPCtt) and real gross domestic product per worker (GDPW). The rest of the paper is organized as follows: Section II briefly reviews founding relationships and concepts from the neoclassical growth model. In section III we appraise available methods to test convergence and section IV summarizes our empirical results using data from Mercosur countries and its associates. Finally, section V concludes. II. A theoretical review of neoclassical models and convergence In this section we present the implications of so-called neoclassical growth models suggested by Solow (1956) and extended by Cass (1965) and Koopmans (1965).6 Following Durlauf and Quah (1999), we consider an economy with an aggregate production function that relates total output, Yt, with the following inputs: labor, Nt, a measure of the stock of human capital, Ht, embodied in the labor force and the stock of physical capital. Thus, the effective labor input can be defined as Ne,tŁNtHt and different types of capital stock are represented by the vector Kt=(K1,t, K2,t, …). In addition, we assume a labor augmenting aggregate technological process, At, that reflects the “state of the art” of the economy. Formally, Yt { Fe (K t , N e,t At ). (1) The usual concavity assumptions allow us to rescale all the variables in terms of effective workers, i.e., ye,t=(Ne,tAt)–1Yt and ke,t=(Ne,tAt)–1Kt.7 To turn the variables into observables, we multiply them by HtAt, which allows us to denote per capita GDP and per capita capital as yt and kt, respectively. It can be shown that the growth rate of output per worker is given by: 6

7

Solow focused only on the supply side of the model, while Cass and Koopmans found that, in addition, dynamic responses depend on consumers’ preferences. Strictly, we assume that Fe is twice differentiable, linear homogeneous, increasing, jointly concave and strictly concave in all of its arguments, and fulfills the Inada conditions.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

yt yt

§ k § H t At · H A ·  ¸  ¦ sl (k e,t ) ¨ l ,t  t  t ¸, ¨ ¨k ¸ © H t At ¹ l © l ,t H t At ¹

ke ,l ,t wf (k e,t ) / wke ,l ,t

sl (k e ,t )

f ( ke ,l ,t )

355

(2)

,

where sl (ke,t) is the factor share of total output paid to owners of the l-th physical capital good, as long as input markets are competitive and wf (k e ,t ) / wke ,l ,t fully describes the factor l’s contribution.8 In practice, if the researcher is looking at the determinants of the longrun growth, like Kaldor (1963), one concentrates on the economic forces driving the first bracketed term of Equation (2) and to disregard the second one. Conversely, if the goal is to investigate the dynamics around the balanced growth path, growth rates of human capital and technology are regarded as given. This dynamic analysis comprises what is known as conditional E-convergence. We define the balanced growth path in this deterministic model as the collection of paths, (yt, kt), such that for all l, yt kl ,t a, (3) yt kl ,t where a is a constant. It can be shown, following Section 4 and Appendix A in Durlauf and Quah (1999), that the same results in terms of balanced-growth predictions are derived from the models of Solow (1956), Cass (1965) and Koopmans (1965). Durlauf and Quah (1999) show that there always exists a unique long-run growth equilibrium and that the solution is dynamically stable. Given the nonlinear solution, we approximate it by a Taylor series expansion of log ke,t around its steady state, ke*, to find out the conditions that lead to convergence.9 The steady state income, ye*, is a 8

If technology is assumed Hicks-neutral, the production function (1) becomes Yt { At Fe (K t , N e , t ), and the rate of per capita GDP in Equation § kl , t H t · § H t At · ( k )   s A  ¸. ¨ ¸ ¦ l e,t t ¨ yt © H t At ¹ l © kl , t H t ¹ Smooth convergence results if the absolute value of the root of the differential Equation (4) is lower than one, i.e., _O_. In the CassKoopmans model, the same condition applies: (i) the eigenvalues of the matrix M are real and of opposite sign, and (ii) trace (M) > 0.

(2) is 9

y t

356

Jorge Alberto Fornero

function, g(.), that depends positively on the rate of savings, W, and negatively on the rate of depreciation, G, the rate of growth of population, Q, and the rate of technological change, [. Thus, the growth path is given by:10 W § · log y (t ) log g ¨ ¸  log A(0)  [ t  © G Q  [ ¹ (4) ª º Ot W § · « log ye (0)  log g ¨ G  Q  [ ¸  log A(0) » e . © ¹ ¬ ¼

Differentiating Equation (4) with respect to time, the growth rate of yt becomes: d > log y (t ) @ ª º W § ·  log A(0) » eOt [  O «log ye (0)  log g ¨ (5) ¸ dt © G Q  [ ¹ ¬ ¼ Equation (5) shows that the deterministic balanced growth path of per capita income depends on the growth rate of technology, an initial gap of per capita income with respect to its steady state and the initial technology level. The applied literature that flourished after the seminal work of Baumol (1986) aimed to estimate Equation (5) using cross-country regressions. As we mentioned in the introductory section, a negative correlation of the initial income is interpreted as evidence in favor of convergence, but the magnitude of this correlation will differ as long as the set of initial independent variables varies. After estimating Equation (5), we can calculate (see below) an intuitive measure of the speed of convergence (in percentages), called the Econvergence rate, which is easy to interpret and allows us, e.g., to calculate the half-life, the amount of time it takes for half of the countries’ income in a sample to decay to the steady state. If the set of conditional factors includes no other variables than the initial income and technology, we can calculate the absolute E-convergence rate. On the contrary, expanding the conditioning set to include a wide range of factors such as the political and cultural system in which the economy operates will allow us to interpret the estimated E-convergence rate as a conditional measure. Sala-iMartin (1996) pointed out that absolute E-convergence implies that poor economies simply grow faster than wealthy ones and eventually, in the long-run, will reach the same steady state. When we 10

Log refers to logarithms using as base e.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

357

refer to absolute convergence, then, we mean a specific case of conditional E-convergence. Another measure of convergence proposed is Vconvergence, which informs us about the dispersion of per capita income in the cross-section at a specific point in time. Intuitively, countries verify Vconvergence when the standard deviations of income growth in the cross-section decrease over time. Relating both convergence measures, however, may lead to the mistaken implication that Econvergence implies Vconvergence —known as Galton’s fallacy (Quah, 1993). III.

Methods to test convergence: A review

A. Cross-section and panel data analyses to compute the speed of convergence This subsection reviews alternative specifications to test the conditional speed of convergence that are, in general, derived from Equation (5). 1. Estimating convergence with cross-section equations In general, the cross-section equation was specified as follows: log yi ,t0 T  log yi ,t0 T 1 a  b log yi ,t0  G Xi  ui , i=1,...,N, (6)





where b {  1  e  E T T 1 . In words, Equation (6) states that the period average per capita GDP growth is explained by a constant term that captures the initial, t0, state of technology and the initial per capita GDP level. Moreover, G is a 1x k vector of coefficients that measure the sensitivity of income growth to conditioning (period averages) variables collected in a k x 1 vector Xi. The error term ui is assumed to be iid with E(ui)=0 and Var(ui)=Vu2 and represents averages of ui,t. The speed of convergence is calculated as in Barro and Lee (1993) and Barro and Sala-i-Martin (2004) from the above definition of b and bˆ it follows that E= –log(1+b)/T. The literature reports speeds of convergence ranging from 3 to 30 per cent using panel data and around 2 per cent using cross-section regressions.11

11

See Badinger et al. (2004), who calculate speeds of convergence after controlling for spatial dependence across regions with a SAR model.

358

Jorge Alberto Fornero

As regards control variables, since Xi may contain any number of them, adding too few/many will bias the estimates. In a review of the specifications utilized in the literature, Durlauf and Quah (1999) identified at least 36 categories of variables with 87 significant examples. They concluded that even though a new control variable likely increases the R2 of the regression, it is also expected that b will differ from the true E. Levine and Renelt (1992) and Sala-iMartin (1997) tried to overcome this specification problem proposing an extreme bounds algorithm to choose the most relevant variables from a large set of controls. However, Durlauf and Quah (1999, pp. 282-3) criticized the selection procedure of controls because such algorithms analyze with statistical criteria a large number (combinations) of dependent variables, whereas the optimum selections are obtained using criteria motivated by economic theory, which is more restrictive. The underlying inconsistency points to sets of controls which are not bounded enough, leading to controversial estimation results due to a lack of robustness. An implicit assumption of Equation (6) is that the steady state path is homogeneous in the cross sections. Lee et al. (1997, p. 359) remarked that cross-section regressions (or regressions using observations based on data averaged over long periods) disregard “either the complex dynamic adjustments involved in the countries’ output processes or the heterogeneity of growth rates across countries.” An important empirical aspect regards the validity of estimates of Equation (6). In particular, reasonable good estimates of variances to perform t-tests ʊi.e., var ( aˆ ) , var (bˆ) and var (Gˆ ) ʊ require a large number of countries. The seminal paper by Mankiw et al. (1992) tested the convergence hypothesis with a sample of 98 countries and confirmed the findings of Barro (1991). However, Temple (1998) argues that the results reported by Mankiw et al. (1992) maybe flawed due to: (i) a measurement error in the initial income; and (ii) the fact that some countries can be classified as outliers and that after removing them, the evidence favouring convergence vanishes. Other methodological objections were raised, e.g., by Islam (1995), Mankiw (1995) and Barro (1997), which we examine in the following subsection about panel data techniques. In short, the common concern revolves around the relevance and validity of estimates from cross-section regressions.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

359

2. Estimating convergence with panel data methods A major reason why panel specifications perform better than crosssection regressions is that at least one unobservable variable with permanent effects is omitted: the initial country-specific level of technology. A panel is a convenient way to deal with this misspecification problem because the unobserved effect can be estimated with a country-specific intercept (Islam, 1995). Going through similar derivations as in Section II, Islam (1995) proposes the following panel data specification: i=1,...,N; log yi ,t  log yi ,t 1 b log yi ,t 1  G Xi ,t  ui ,t , (6’) t=1,…,T where ui,t=Pi+Ji,t comprises the sum of: (i) Pi, a country-specific unobserved fixed effects (FE) that captures, inter alia, the level of technology —with the following binding restriction

¦

N i

Pi

0 ʊ,

2

and (ii) an iid error term Ji,t with E(Ji,t)=0 and Var(Ji,t) = VJ . If Pi = 0 for all i, fitting Equation (6’) with ordinary least squares (OLS) gives the best linear unbiased estimators (BLUE), see Greene (2003); but, if it exists at least an i such that Pi > 0, BLUE will result only by applying least squared dummy variables (LSDV), see Hsiao (1986) and Baltagi (2001). How do we discern whether we are right in assuming country-specific FE in our specification? It is possible to test whether the FE LSDV specification is more explicative than pooled OLS by performing the following F-test that involves R2 of these models: 2 2 ( RFE  R pooled ) /( N  1) 2 (1  RFE ) /( NT  N  k  1)

~ FN 1, NT  k 1 , under the null pooled.

(7)

The heterogeneity of cross-section regressions may turn unreasonable the assumption of scalar variance-covariance matrix for the residuals, i.e., Var(J) = VJ2I. To take into account heteroskedastic residuals of unknown form, we assume that Var(J) = 6 (symmetric and positive defined). There are two standard procedures to obtain consistent estimators: (i) to transform variables in order to make the variance-covariance matrix an identity matrix (e.g., through the Cholesky decomposition; see Greene, 2003) and apply OLS to the transformed variables; or (ii) to run Generalized Least Squares (GLS) on the original model of Equation (6’). However, since the

360

Jorge Alberto Fornero

entries of 6 are unknown, feasible GLS (FGLS) needs to be applied in order to obtain asymptotically BLUE.12 Durlauf and Quah (1999, pp. 285-7) pointed out two drawbacks arising from the panel methodology: (i) although we can appropriately estimate the unobserved technological level Pi in (6’), it is not possible to disentangle whether catching-up between poor and rich countries occurs; and (ii) when individual effects Pi are correlated with right hand side variables, a correlation of error terms arises with the common estimation procedure of FE models.13 Calculating a within estimator that can overcome this inconsistency implies to demean all variables, annihilating the FE. Hence, the validity of estimates’ interpretation is within the business cycle horizon.14 Endogeneity can be better treated either with instrumental variables (IV) or generalized method of moments (GMM); see, e.g., Hsiao (1986, pp. 73-5), Islam (1995, pp. 1137-8), Pesaran and Smith (1995) and Baltagi (2001). Durlauf and Quah (1999) pointed out two advantages of panel data models as well. First, these models would force the researcher to specify more explicitly the estimable equation. Second, panel regressions allow us to profit from greater flexibility in modeling “unobserved” or omitted variables. 3. Dynamic panel data methods with endogenous variables To take into account the possible correlation of the individual effects with independent variables, we rewrite Equation (6’) as standard dynamic panel of the production function with an autoregressive (AR) structure of order one, AR(1): log yi ,t Pi  c log yi ,t 1  G Xi ,t  J i ,t , i=1,...,N; t=1,…,T, (6”)

12

13

14

ˆ and (ii) Ȉˆ FGLS involves two stages: (i) the estimation of Ȉ substitutes the unknown 6 then, standard GLS is run to get BLUE. This problem arises when regressors are: the lagged dependent variable or variables simultaneously determined with growth (i.e., with bidirectional causal relation) such as investment. This trade off was put forward by Barro and Sala-i-Martin (2004). With a sample of fixed length, to produce more observations, a common strategy is to measure averages of income growth and RHS variables over shorter periods of time. However, more data points will bias model conclusions towards a short-run horizon.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

361

where cŁ1+b. To mitigate the endogeneity effect of dependent variables, Equation (6’’) can be estimated using IV and GMM techniques. The first IV estimator that remedies endogeneity was proposed by Anderson and Hsiao (1981). Arellano and Bond (1991) showed that the Anderson-Hsiao estimator is not necessarily efficient due to the fact that it does not make use of all information available in moment conditions and proposed a GMM estimator which includes all available lags as instruments.15 Several studies applied the GMM approach, e.g., Caselli et al. (1996). Moreover, as it was mentioned, control variables are likely to be endogenous in (6’’). In such an instance, an appropriate instrumentation strategy requires that each (lagged) variable used as instrument, say column j of Xi, is uncorrelated with 'Ji,t, i.e., E(Xij,t-1,'Ji,t) = 0, E(Xij,t-2,'Ji,t) = 0, etc. Arellano and Bover (1995) and Blundell and Bond (1998) suggested a system GMM estimator that exploits the (individual) initial condition E ( yi ,t Pi ) Pi (1  c) , which is sufficient to



assure E  Pi J i1 0 . In other words, to make sure that E yi ,t Pi

is

time invariant it is required that E ª¬ Pi  yi ,1  Pi 1  c º¼ = 0. These initial conditions along with Equation (6’’) lead to T-2 further linear moment conditions: E  Pi 'yi ,t 1 0 for i =1,2,…,N and t = 3, …,T. As before, a control variable j can be instrumented with their own lags, provided E('Xij,t-1,Ji,t) = 0, E('Xij,t-2,Ji,t) = 0, etc., hold. Consequently, system GMM combines derived moment conditions of the difference GMM estimator —suitable instruments being lagged levels— with an additional set of equations in levels and instruments in differences. Blundell and Bond (1998) showed how to construct the set of moment conditions, calculated the system GMM estimator and, also, offered supportive evidence for the superior performance of the system GMM estimator versus the difference GMM estimator.

15

Arellano and Bond (1991) propose to exploit all additional information: from available orthogonality conditions E(yi,t-2,'Ji,t) = 0, E(yi,t3,'Ji,t) = 0, etc., meaning that if the lagged observation yi,t-2 is not correlated with the error 'Ji,t, it will not be with further lags yi,t-3, etc. See Hsiao (1986) and Baltagi (2001) for a comprehensive treatment.

362

Jorge Alberto Fornero

Bond et al. (2001), in an influential review of the methodology applied in panel growth regressions, show that the necessary requirements of Arellano and Bond’s difference GMM estimators are unlikely to be satisfied. Typically, income series are very persistent and the number of observations is small because of the common practice of averaging (to avoid short-run fluctuations from contaminating the sample). Under these circumstances, the moment conditions suggested by Arellano and Bond (1991) do not hold and, consequently, lagged levels of the variables are weak instruments for their respective first differences. These results lead us to estimate panel growth regressions using system GMM. B. Time series approach to convergence Time series analysis is the other methodology used in the convergence literature, which is primarily aimed at testing unit roots (UR) and cointegration in the income data generating process. Bernard and Durlauf (1996) have reinterpreted the convergence concept and proposed a novel testing procedure.16 Defining the difference between countries i and j income (in logs) as d(i,j),t Ł (yi,t í yj,t), there exists convergence if the expected value of d(i,j),t tends to zero as time goes to infinity: Lim kĺ’ E(d(i,j),t+k | It) = 0,

for k > 0,

(8)

where It, is the information set at time t. Recall the intuition of Econvergence, which refers to an expected reduction in contemporary differences, whereas (8) implies the expectation of their eventual disappearance. In this framework, evidence of a UR or a deterministic component in the series d(i,j),t supports the null of nonconvergence. Equation (8) is very close to the concept of cointegration between the series that define d(i,j),t. In effect, countries i and j converge in income when both series are cointegrated with cointegrating vector [1 -1]. If both countries’ income series evolve showing a common trend, forecasts of income are proportional for any t. Alternatively, if d(i,j),t is a stochastic variable that follows a time trend with negative slope, i.e., with cointegrating vector [1,D], the differences in income will tend to decrease but not to disappear. This is interpreted as catching-up.

16

Bernard and Durlauf (1995) is the first study that implemented it.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

363

To estimate the rank of the cointegrating matrix in a N-variables framework, the output vector process is written in the following VAR(k) representation: 'yt *( L)'yt  3yt 1  P  H t , t: 1,…,T, (9) where 3 represents the long-run relationships of the cointegrating vectors, P is the mean of the stochastic process yt, *(L) (a polynomial of order k – 1) captures the short-run dynamics of the system, and Ht are independent Gaussian errors with E(Ht) and Var(Ht) = :. The reduced rank (0 < rank(3) = r < N) of the long-run impact matrix is formulated as 3 DEcwhere Eҏ is the matrix of cointegrating vectors (linear combinations Ecyt-1 present the r cointegrating relationships) and *( L) ҏ is the matrix of adjustment coefficients. Reichlin (1999) suggests using cointegration analysis to find evidence of convergence clubs aimed at searching for cointegration clusters. If we expect to find evidence of a convergence club in a group of countries, the number of cointegration relationships (CE) must be N – r = 1. In such a case, the demeaned variable y it { yit  yt is stationary, where yt {

¦

N i 1



yit N 1 is the average

income of the group. When N – r > 1, it is not clear which countries belong to different convergence clubs. A search procedure must be applied to classify the countries into the appropriate subgroups, each of them associated with particular convergence clubs.17 The search procedure checks all the permutations without repetition in a systematic manner and sorts the countries accordingly. Hobijn and Frances (2000) developed such an algorithm, which we will briefly explain in Subsection III.D. Finally, there are two criticisms of this methodology which are particularly applicable to developing countries. First, the likelihood that the mean series of the club can misrepresent the steady state of the group increases for countries that are far away from their steady 17

Reichlin (1999) suggests two conditions to be met for the case of two convergence clubs (G1 and G2): (i) G1 and G2 are each cointegrated within its own members, and (ii) the mean incomes of the two groups, y t and y t , are not cointegrated. If there are two convergence clubs, conditioning each group with respect to its own mean will imply that y ( i  G1 ) and y ( L  G 2 ) are sets of stationary variables, and each set is a convergence club. G1

G2

G1

G2

it

it

364

Jorge Alberto Fornero

state. Second, structural breaks make UR tests difficult to implement, suggesting a word of caution in the interpretation of the results (Camarero et al., 2002). This remark is especially applicable to our Latin American countries series. C. Univariate and multivariate UR tests to study convergence In this subsection we review the theory that gives support to univariate and multivariate tests that have been applied to examine income convergence. The idea is to consider differences in per capita GDP and productivity across countries and allow this difference to vary over time. Evidence of URs will be interpreted as evidence of non-convergence because the income differences will not vanish. There are two ways to construct the null for tests of URs either stating: (i) the series are I(0) as in Kwiatkowski et al. (1992), hereafter the KPSS test; or (ii) series are I(1) as in Dickey–Fuller (DF) type tests. With the latter approach, we will accept the stationarity hypothesis if we reject the null of a UR. However, most UR tests have low power against stationary, but highly autoregressive, alternatives. As a result, the standard DF approach does not find stationarity often. Hobijn et al. (2004), remark that stating the null of stationarity is not easy to deal with since it is also difficult to control the size of the test under a stationary process that is close to UR. They conclude that the best test for stationarity is the KPSS test. In addition, they propose a generalization that remedies the size problem and at the same time guarantees consistency.18 First, focusing on a null that assumes stationary series such as the KPSS test, univariate UR tests are presented.19 The KPSS test has the following model structure: d ( i , j ),t \ t  rt  I d ( i , j ),t 1  H t , t 1,,T, (10) rt rt 1  ut ,

18

19

In a Monte Carlo experiment, Hobijn et al. (2004) compare their generalized KPSS test with the Choi (1994) and Leybourne and McCabe (1994) tests and find that the first converges at a lower consistency rate and the second is inconsistent. The augmented DF test is implemented in a multivariate set up, see below. See Lim and McAleer (2004) for an application of this test to a sample with 5 countries of the Association of South-East Asian Nations.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

365

where Ht and ut are iid with mean 0, VH =1, Vu finite and _I_. Recall that the definition of d (i , j ),t is consistent with country j acting as a leader. If we restrict Vu = 0 in Equation (10) with the assumption that the initial value r0 is fixed, the process d ( i , j ),t is trend (mean) stationary if \= () 0 For Vu > 0, the d ( i , j ),t process has a UR, which means, intuitively, that country i diverges from country j. If d ( i , j ),t is trend or mean stationary, it will imply that the incomes of those countries are converging and are narrowing (catching up), respectively. Defining K k ,t >H 't u 't @ as a covariance stationary process, the KPSS statistic is:

Ts (l ) ¦ 1

T

(11) k:PW where St is the partial sum of residuals et, which can be computed from a regression of: (i) yt on an intercept ( KˆP ), or (ii) yt on an interKˆk

2

T

t 1

2

St ,

cept and a linear trend ( KˆW ), according to whether one is testing for stationary level or trend, respectively. sT2 (l ) is an estimate of Var(Ht) from the appropriate regression: T

l

T

sT2 (l ) T 1 ¦ t 1 et2  2T 1 ¦ s 1 w( s, l )¦ t

ee

s 1 t t  s

,

(12)

where w(s,l) is the weighting function that corresponds to the choice of the spectral window. KPSS use the Bartlett window: w(s,l) Ł 1-s/(l+1), for s ” l, and 0 otherwise, where l is the lag truncation parameter.20 Second, we consider multivariate tests for convergence following the seminal work by Abuaf and Jorion (1990). The idea is to build up a system of N AR equations taking into account contemporaneous correlations between the series in order to enlarge the power of univariate UR tests, which is notably low when true parameters are nearly one. Taylor and Sarno (1998) and Sarno and Taylor (1998) proposed a generalization to a multivariate set up of Abuaf and Jorion’s (1990) test, which corrects for serial correlation. We refer to this method as MADF, where ADF stands for augmented Dickey-Fuller test. In these approaches the null hy20

The quadratic spectral kernel produces more accurate estimates of V2 in small samples (see Hobijn et al., 2004).

366

Jorge Alberto Fornero

pothesis states that each and every series has a UR.21 Formally, they propose that the stochastic process {yit; i:1,...,N; t: 1,…,T} is generated as: k

ai  ¦ Uih d ( i , j ),t  h  H i ,t ,

d ( i , j ),t

(13)

h 1

where ai allows for the possibility of a nonzero mean under the alternative that a given series is stationary, H i ,t is a Gaussian error with E( H i ,t ) and Var( H i ,t ) = VH2 and k denotes the number of relevant lags for each series. The null can be stated as: k

H 0 : ¦ Uih  1 0

i 1,..., N .

(14)

h 1

In words, under the null it is assumed that all the series considered are realizations of I(1) processes. Moreover, note that Equation (14) is general in the sense that we obtain: (i) the univariate ADF test if we estimate each equation separately (using pooled OLS), and (ii) Abuaf and Jorion’s (1990) method if the researcher fixes k = 1 for all series. The critical values for this test must be estimated by simulation methods because the distribution of the statistic is unknown under the null. There are several steps to calculate the critical values. First, the model needs to be estimated using the seemingly unrelated regression (SUR) method, the residuals, eit, and the covariance matrix, E(ei,t’ei,t) saved. Second, under the assumption that ei,t ~ MN(0, E(ei,t’ei,t)), the Cholesky decomposition E(ei,t’ei,t)=PP’ is applied to recover residuals from a 1(0, IN). Then, samples of size T are constructed and further, replicated a large number of times. Under the null, the sample distribution is approximated by building up the statistic’s histogram. Finally, the estimates calculated are valid as long as the statistic accumulates at least the desired D per cent significance level of the distribution. Thus, critical values depend on how many series the panel accommodates and on the number of observations included in the sample. 21

The motivation behind the decision to apply the MADF method is to take into account the correlation between series. Other widely applied studies that focus on URs in panels, such as Im et al. (2003) and Levin et al. (2002), consider the individual cross sections of the panel as independent, neglecting information of cross-correlations (e.g., similar swings in income of neighbour countries).

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

367

If either the number of observations increases or the off-diagonal elements of the matrix E(ei,t’ei,t) are nearly zero, the critical values of the MADF test approach the corresponding ones from the ADF test. In general, the effect of adding new series to the panel is ambiguous. Breuer et al. (2002) show that even though the addition of new information is good if newer series do not have URs, the probability of rejecting the null increases if the marginal incoming series has a UR. Because of this trade-off, they suggest expanding the number of cross-sections in the panel only with caution. The generalized KPSS test by Hobijn et al. (2004) was extended to a multivariate setting by Hobijn and Frances (2000). The next subsection briefly introduces this method. D. Convergence of countries that form clusters The methods described so far are not precisely asking to what extent the countries converge in per capita GDP or productivity towards the countries’ club path. Hobijn and Frances (2000) defined, following Bernard and Durlauf (1995), two types of convergence, namely asymptotically perfect —which corresponds to that proposed by Bernard and Durlauf (1995)— and asymptotically relative convergence, as follows: i) n* countries are converging asymptotically perfectly if d(i,j),t is zero mean stationary; ii) n* countries are converging asymptotically relatively if d(i,j),t is level stationary; where n* is the subsample size. Intuitively, the weaker version i) is consistent with pairs of countries that are growing at constant rates (asymptotically the same in the long run) not necessarily having a common path but never reaching the same levels.22 Hobijn and Frances (2000) suggest testing these two definitions by implementing an algorithm that minimizes endogenously the probability of clustering together countries’ incomes that actually do not converge. The procedure uses the farthest distance between cluster partners as a measure, being very conservative. By construction, the endogenous algorithm of Hobijn and Frances has the following properties: (i) it tests for necessary and sufficient conditions of convergence, meaning that one can find out the number of convergence clubs by inference, whereas in cross-section studies 22

Calling this gap H Equation (8) becomes Lim kĺ’ |E(d(i,j),t+k | It)| ” H for k > 0.

368

Jorge Alberto Fornero

one is able to identify countries that might possibly converge; (ii) the cluster algorithm’s solution is consistent in the statistical sense; and (iii) the clusters’ solution is unaffected if the order of series is changed. Although a detailed analytic development of the algorithm is beyond the scope of the paper, the intuitive idea of the algorithm comprises several steps. First, it analyzes the estimated variance of different samples under different hypothesized clusters, then it calculates Equation (11) in a multivariate set up (Hobijn et al., 2004), and the UR test is performed systematically to form the clusters. The sensitivity of the algorithm depends on two parameters under control of the researcher: (i) the p-value indicating excess of probability, which can be loosely interpreted as a “significance level”, meaning that the smaller the p-value is, the less likely the rejection of the null hypothesis of convergence is and the larger the dimension of the convergence clubs that we find; and (ii) a lag that gives the width of the window of the Newey and West estimator (an extension of Equation (12) to a multivariate set up), which enters in the calculation of the KPSS test. This parameter determines the value of sT2 (l ) for all combinations of countries and clusters; therefore it is advisable to conduct sensitivity and robustness analyses. IV.

Empirical Results for the Mercosur

The aim of this section is to test the convergence hypothesis for Mercosur countries and its associates, using the methodologies described in Section III. A. Data The main source of the dataset is the 6.1 version of the Penn World Table (PWT) compiled by Heston, Summers and Aten (2002). Our sample comprises annual data that run from 1961 to 2000, with 40 observations for 7 LA countries Argentina, Brazil, Bolivia, Chile, Paraguay, Peru and Uruguay.23 To set the time dimension of the panel, we need to disregard “business cycle” fluctuations since we are interested in a long-run horizon. Following Barro (1997), we calculate five-year averages for all countries’ variables during a period of 40 years, T=40, so that we have in total

23

Venezuela is neglected since it became a member of Mercosur in 2007.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

369

56 observations in the panel, with i=7 (countries) and t=8 (periods). The variables are defined in Table 1. Table 1: Definition of variables and sources Variable GDPCit GDPCttit GPDWit INFit

Iit Git MERCit OPENit POSTSEC_25it

SEC_25it

Definition real GDP per capita real gross domestic income per capita (adj.by terms of trade) real GDP per worker percentage change of consumer price index. Source: World Bank (2001), WDI CD-Rom investment, share of GDP government expenditure, share of GDP dummy variable for Mercosur membership: takes 1 if the country is member of or associated, 0 otherwise exports plus imports as a share of GDP secondary education attainment, share of population older than 25 years old; Barro and Lee (2000) post-secondary education attainment, share of population older than 25 years old; Barro and Lee (2000)

To produce valid inferences from time series methods instead of using 5-year average data, we examine yearly income data. Hence, the sample runs from 1951 to 2000 for 7 LA countries; in total, the number of observations is 350. B. Estimation with panel data techniques This subsection aims to present a panel specification that overcomes the two main criticisms of cross-section regressions, namely (i) the improper treatment of the country-specific initial effect and (ii) the identification of which “flow” variables entering in the elements Xi are simultaneously determined with income and thus endogenous. The inclusion of the time dimension enables us to address more interesting research questions such as whether there are differences in the observed conditional speed of convergence of LA countries after the creation of Mercosur. Technically, to answer that question requires introducing a dummy variable capable of accounting for the change in convergence speed due to the establishment of Mercosur in 1991. Hence, MERCt indicates to which extent the membership in a larger market and the benefits of a new and common trade policy impact the speed of convergence.24 24

As in Barro (1991), Equation (6) is estimated with OLS to gain first insight into coefficients signs and the speed of convergence, with Xi

370

Jorge Alberto Fornero

1. Panel data regressions to estimate individual effects First, we estimate the panel specification of Equation (6’) with OLS and we find evidence of serial correlation and heteroskedasticity that invalidate estimates (results not shown), so we correct for these shortcomings by running FGLS (see Barro and Lee, 1993). Two versions are estimated: firstly disregarding individual FE (pooled FGLS) and secondly taking them into account explicitly (FE FGLS or LSDV FGLS). Results are reported in Table 2, where each re25 gression is displayed in a column. Treating explicitly the initial level of technology as FE in the panel enlarges estimates of the conditional speed of convergence, as other studies report. Our estimates of conditional–E speed are: (i) 1.21 per cent and 5.72 per cent for GDPC, (ii) 1.36 per cent and 6.3 per cent for GDPCtt and (iii) 1.16 per cent and 4.12 per cent for GDPW, with the first figure obtained from the model estimated using pooled FGLS and the second using FE FGLS, respectively. The dummy MERC always has a positive impact on growth and convergence; however, results are insignificant when estimating Equation (6’) with pooled FGLS, see columns (i) to (iii). Accounting for FE in the panel (FE FGLS), we also find positive and significant impact of MERC on GDPCtt at usual levels, while the impacts on GDPC and GDPW are positive and significant at the 10 per cent level. Overall, the results suggest a positive influence by the trading block on the speed of convergence across countries. When including FE in the regression, the estimated conditional comprising 40-period averages of INFi, Gi, Ii and POSTSEC_25i. The results is: log  yi ,2000 yi ,1961 40

0.27  0.079 log yi ,1961  0.0012 INFi  (0.05)

(0.02)

(0.003)

R2=0.99

0.038 I i  0.042 Gi  0.04 POSTSEC _ 25i  ui , (0.005)

(0.003)

(0.016)

All coefficients are significant ʊp-values are in bracketsʊ and of expected sign, except for POSTSEC_25. The implied speed of convergence is 0.2 per cent. OPEN and SEC_25 harm the results. 25

Two analyses are possible: (a) comparing the same income measures but different estimation methodology, i.e., by comparing columns (i) with (iv), (ii) with (v) and (iii) with (vi); or (b) different income measures but the same estimation technique.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

371

speed of convergence increases at least four times for our three income versions. In addition, we confirm the expected positive (negative) effect of private investment (public expenditure) on all income versions. However, the estimated effect of G is insignificant at the usual level. Although inflation has the expected negative sign, it is only significant at usual levels in columns (i) and (ii), while in column (iii) its inclusion drops the R2. In columns (iv) to (vi) the inflation coefficient is not significant. Regarding human capital, adding a proxy of high-quality human capital in (6’), POSTSEC_25, does not capture the expected effect of education on income and productivity: there is a negative and significant coefficient in the GDPW specification ʊat odds with the theoryʊ, while for GDPC and GDPCtt specifications these estimators are negative and insignificant. A measure of average-qualified human capital such as SEC_25 could not produce better results than POSTSEC_25. The sign of the estimates in all income versions were negative and not significant at 10 per cent level. Finally, trade openness (OPEN) seems to have an ambiguous sign in all growth regressions and estimated coefficients are insignificant.26 The estimation of individual FE in the panel regressions allows us to recover the countries’ steady-state income levels through dummy variables that correct the constant in the panel regression. Steady-state income levels are directly calculated using information from columns (iv), (v) and (vi) of Table 2, where we report deviations from an overall mean (common intercept, c) individual FE, labelled for convenience as “FE j –c” (j refers to the 7 countries included in the sample). Given our different measures of income, we calculate three rankings of steady-state income levels. When measuring income by GDPC and GDPW, identical rankings emerge (ordered from the highest to the lowest initial steady-state income): Argentina, Uruguay, Chile, Brazil, Paraguay, Peru and Bolivia. The ranking changes slightly when we proxy income with GDPCtt resulting in a permutation of Peru and Paraguay. However, as Caselli et al. (1996) remark, a country with a high observed income is not

26

Since we did not correct for bidirectional causality in estimates of Table 2 —recall the discussion of Section III—, it is advisable to conduct a prudent interpretation.

372

Jorge Alberto Fornero

necessarily closer to its steady-state than a country with a relatively low observed income.27 To assess the relevance of the regression that includes FE, we conducted the F-test proposed in Equation (7) at the bottom of Table 2, columns (iv), (v) and (vi). The evidence suggests that the null is rejected in all cases, which means that all the models estimated with FE FGLS have more explanatory power than the respective pooled FGLS and, therefore, preferred. FE FGLS estimates are preferred because the data seems to support country specific initial technology levels in contrast to the common initial technology level implied by pooled FGLS that turns to be too strong. Table 2: Panel data regressions of convergence in GDPC, GDPCtt and GDPW in Mercosur members and associates, 19612000 (to be continued) Pooled FGLS (i) (ii) (iii) GDPC GDPCtt GDPW c 0.4794 0.5030* 0.4180* (0.093) (0.048) (0.146) dep..variable 1960 -0.0658** -0.0563* -0.0585** (0.038) (0.021) (0.084) INF -3.7E-05** -4.9E-05*** (0.038) (0.007) I 0.0070*** 0.0068*** 0.0081*** (0.001) (0.002) (0.000) G 0.0014 0.0009 7.95E-06 (0.489) (0.685) (0.997) MERC 0.0353 0.0444 0.0534* (0.253) (0.152) (0.091) POSTSEC 25 -0.0003 -0.0023 -0.0100*** (0.944) (0.513) (0.005) OPEN -0.0003 -0.0003 0.0001 (0.686) (0.691) (0.900) Robust (weighted) statistics adjusted R2 0.2670 0.3440 0.3210 DW 2.0700 2.0570 2.0800 speed of convergence 0.0121 0.0136 0.0116 variable

27

One possible explanation of why problems of robustness appeared in GDPC and GDPCtt is because the LA countries included in our sample are developing countries. They are subject to strong shocks such as large fluctuations in the terms of trade, which give rise to greater instability tracked in GDPC, but not in GDPCtt.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

373

Table 2 (cont.): Panel data regressions of convergence in GDPC, GDPCtt and GDPW in Mercosur members and associates, 1961-2000 FE FGLS (iv) (v) GDPC GDPCtt c 2.289*** 2.0580*** (0.000) (0.000) dep.variable 1960 -0.2701*** -0.2487*** (0.000) (0.000) INF -3.8E-06 -1.6 E-05 (0.831) (0.349) I 0.0052 0.0072** (0.024) (0.107) G -0.0043 -0.0054 (0.259) (0.173) MERC 0.0657** 0.0544* (0.085) (0.045) POSTSEC 25 0.0032 0.0002 (0.358) (0.950) OPEN 0.0013 0.0019 (0.237) (0.132) FE Argentina – c 0.1305 0.1540 FE Bolivia – c -0.1741 -0.2010 FE Brazil – c 0.0410 0.0678 FE Chile – c 0.0614 0.0699 FE Paraguay – c -0.0675 -0.1140 FE Peru – c -0.0879 -0.0770 FE Uruguay – c 0.0965 0.1000 Robust (weighted) statistics adjusted R2 0.4250 0.4870 DW 2.1600 2.0700 speed of convergence 0.0572 0.0630 statistic eq. (7) 3.1570*** 3.1340*** p-value (0.01) (0.01) variables

(vi) GDPW 1.7210*** (0.006) -0.1860*** (0.003) 1.57E-05 (0.474) 0.0091*** (0.012) -0.0029 (0.495) 0.0622* (0.065) -0.0083** (0.044) 0.0009 (0.493) 0.0826 -0.1152 -0.0011 0.0650 -0.0444 -0.0544 0.0675

0.3920 2.3600 0.0412 3.1610*** (0.01)

Note: 56 observations, 7 cross sections, estimated with FGLS with one step ahead weighting matrix; p-values are in brackets. Individual FE estimates sum to zero and should be interpreted as deviations from an overall mean. DW stat means Durbin-Watson statistic. The speed of convergence is calculated according to: E  log(1  b ) / 5 . *(**)[***] denotes rejection of the null hypothesis at the 10%(5%)[1%] level.

2. Panel data regressions with FE and controlling for endogeneity To wipe out individual effects Pi of Equation (6’’), following Hsiao

374

Jorge Alberto Fornero

(1986), we take first differences to obtain:28 i=1,...,N; (16) t= 1,…,T. Equation (16) cannot be estimated directly by a FE GLS as in the previous subsection for two reasons: (i) we presume that “flow” variables in 'Xi,t are endogenous and (ii) the lagged-dependent variable ' log yi ,t is correlated with the composed error term, 'J i ,t , through contemporaneous terms in period t. As we described in Section III.A.3, the best suited estimation technique to investigate convergence is the so-called System GMM. To implement it, we need to carefully set the instrument set to address the problem of endogeneity. For that purpose, we performed the Durbin-WuHausman test and found that investment and public expenditures are endogenous in all income measures for our panel regressions. Therefore, we instrumented these endogenous variables with lagged values in order to get consistent estimates. The estimated results of Equation (16) and the constructed instruments set are reported in Table 3, where each column refers to a different income measure. For the specification where GDPCt is the dependent variable, the instruments’ set includes GDPCt-2 to GDPCt-6, It-1 to It-6, Gt-1 to Gt-6, while INFt, MERCt and POST_SEC25t are assumed totally exogenous serving themselves as instruments. The same instruments set is used in the income regressions of GDPCttt and GDPWt, with the exception of the instruments GDPCttt-2 to GDPCttt-6 and GDPWt-2 to GDPWt-6, which replace the corresponding instruments for GDPCt-1. The augmented instrument matrix is built as in Blundell and Bond (1998) and comprises both levels and first differences of the dependent and independent variables. To eliminate the individual FE, the system is differentiated. The summary of our findings is reported in Table 3. In all regressions the coefficient of the lagged income is large as the estimate remains in range 0.65 to 0.75 at the usual levels of significance. These estimates will be used to construct the estimated speed of convergence, see below. In comparison with the estimates of inflation reported in Table 2, on the contrary, we find a positive influence of inflation on income - though statistically insignificant ' log yi ,t

28

c' log yi ,t 1  G'Xi ,t  'J i ,t ,

If FE are not eliminated, cˆ presents a downward bias since E(Pi yi,t)0.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

375

for all income panel regressions. The negative influence of inflation is found in several studies using cross-section and panel regressions such as Barro (1997); however, a word of caution in interpreting these estimates was put forward by Levine and Renelt (1992) who showed that inflation estimates are fragile in Leamer’s (1985) sense. Estimates of private investment have expected positive signs and are significant in all regressions; these estimates are larger that those reported in Table 2 because of the correction of endogeneity. We find mixed evidence from estimates of public expenditures and in all regressions these are insignificant. The evidence in the literature suggests that public expenditures harm growth; see Barro and Sala-i-Martin (2004). Table 3: System GMM dynamic panel data regressions of convergence in GDPC, GDPCtt and GDPW in Mercosur members and associates, 1961-2000 dependent variables

variables c lagged dep.var. INF I G MERC POSTSEC_25 speed of convergence instruments AB AR(1) test 1st differences AB AR(2) test 1st differences

GDPC GDPCtt 2.194** (0.028) 2.839** (0.031)

GDPW 2.415* (0.077)

0.7256*** 6.9E-06 0.0130** 0.0016 0.1423** -0.0095

0.7506*** 1.3E-05 0.0112** -0.0020 0.0957* -0.0239**

(0.000) (0.696) (0.012) (0.804) (0.014) (0.213)

0.6476*** 2.7E-05 0.0117** 0.0028 0.1494** -0.0084

(0.002) (0.369) (0.049) (0.621) (0.027) (0.33)

(0.001) (0.449) (0.011) (0.718) (0.063) (0.029)

0.112 L(2,6)GDPC, INF, L(1,6)I, L(1,6)G, MERC, POSTSEC_25

0.091 L(2,6)GDPCtt, INF, L(1,6)I, L(1,6)G, MERC, POSTSEC_25

0.121 L(2,6)GDPW, INF, L(1,6)I, L(1,6)G,MERC, POSTSEC_25

-2.01** (0.045)

-1.9* (0.057)

-1.99** (0.046)

0.64 (0.506)

1.12

0.43

(0.67)

(0.262)

Note: L is a (sequence) lag operator, L(a,b)xt means that xt is lagged from a to b, with instruments’ timing is from t–a to t–b,(b>a). The augmented instrument matrix is built as in Blundell and Bond (1998). p-values are in brackets. The speed of convergence is calculated from: E  log(1  c) / 5 . “AB” means Arellano-Bond. *(**)[***] denotes rejection of H0 at the 10%(5%)[1%] level.

376

Jorge Alberto Fornero

The dummy variable MERC captures the effects on growth of joining the Mercosur. In all regressions the estimates are positive and significant (marginally significant for productivity) at usual levels and are substantially larger than those reported in Table 2. Human capital variables have the wrong sign in all regressions but are only significant at usual levels for productivity. This finding can be explained by the fact that the homogeneity assumption of the estimates is too strong; however, we cannot investigate this issue properly since we are constrained by a small number of cross sections. Finally, the liberalization process measured by the trade openness is discarded from the regressions because it harmed results. A further extension could be to separate exports and imports to better identify the influence on growth. In general, the fit of the model is good in all three versions of the model. The residuals pass usual normality tests as indicated by the Arellano-Bond AR(1) test for first differences and the set of instruments passes the Hansen overindentification test. Table 3 also reports the estimated speed of convergence. We find a higher speed of convergence once endogeneity is corrected. For productivity, the conditional E–convergence estimate increases to 12.1 per cent annually, while the corresponding estimates for GDPC and GDPCtt, slightly smaller, amount to 11.2 per cent and 9.1 per cent, respectively. Similar differences were reported by Canova and Marcet (1995) after endogeneity is taken into account. C. Time series approach In this subsection, we shed light on convergence leadership by considering Mercosur countries and associates and by specifying a VAR model with k lags. In general, preferred models are: VAR(2) for GDPC, VAR(1) for GDPCtt, and VAR(3) for GDPW.29 Table 4 reports the trace and maximal eigenvalue statistics of the stochastic matrix that determine the number of cointegrating vectors (r). First, we consider all the countries, so we have N = 7 variables. Once the relevant VAR(k) is determined, the next step is to test for CE. Two possible extreme results of testing rank(3) can be that: (i) the rank is equal to 7, so all series are stationary; and (ii) it is equal to 0 meaning that all the elements in yt are I(1) and no CE exists. In 29

Akaike Information Criterion (AIC) and Schwarz Bayesian Criterion (SBC) determined the lag order of the best models. In all specifications, residuals have passed normality tests.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

377

Table 4 we report the results of the rank(3) test. Beginning with GDPC, it is not possible to reject the null that states at most 3 CE are present against the alternative r = 4 at the 95 per cent level. Considering the maximal eigenvalue test we cannot reject the null that states at most 2 CE against the alternative r = 3. This result differs from the one reported in Camarero et al. (2002) which reports 5 CE in GDPC for the same countries. Considering GDPCtt, we cannot reject the null of at most 2 CE against the alternative of r = 3. This decision is made at the 99 per cent level of confidence with trace a test and at the 95 per cent level with maximal eigenvalue test. Regarding GDPW, we cannot reject the null of at most 5 CE at 95 per cent with a trace test against the alternative of r = 6. One CE less is found using maximal eigenvalue test since we cannot reject 4 CE at 99 per cent level against the alternative of r = 5 CE. However, if we relax the level of significance to 90 per cent we find 6 CE with both LR tests, implying that one common trend process is I(1). Intuitively, if the group of 7 LA countries belongs to the same convergence club, this common trend should be non-stationary. Following the strategy of Reichlin (1999), the next step is to test whether the average of GDPW has an UR with the KPSS test. Recalling that the KPSS test assumes that the series is stationary under the null, its rejection will show evidence of convergence clubs for the 7 LA countries. Since, at the one per cent of significance level we reject the null, we conclude that the average of GDPW has a UR ʊi.e., a common trend drives the system.30 To sum up the results for the group of Mercosur countries and its associates, the VAR that includes GDPC is driven by 4 or 3 common shocks depending on the rank test. A conservative criterion suggests using the maximal eigenvalue test, thus 3 common shocks. For the VAR that contains GDPCtt series there are 5 common shocks driving the system and, as we observed, this variable

30

The rejection of H0 results from testing both KPSS models from Equation (11). H0 for the model with intercept is rejected since the observed statistic’s value is 0.79 and the critical value at 1 per cent is 0.739 (upper tail asymptotic critical values from Kwiatkowski et al., 1992, Table 1). Also, H0 is rejected for the trend model since the observed statistic’s value is 0.222, while the critical at 1 per cent is 0.216.

378

Jorge Alberto Fornero

seems to be more unstable. Finally, productivity has one common trend that is driving the whole VAR system. Table 4: Maximal eigenvalue and trace statistics to test convergence in GDPC, GDPCtt and GDPW in Mercosur members and associates, 1951-2000 GDPC H0:# CE

trace stat

maxeigenvalue

GDPCtt trace stat

maxeigenvalue

GDPW trace stat

None 188.45*** 75.38*** 164.90*** 58.11*** 307.37*** At most 1 113.06*** 43.43** 106.80*** 39.86** 194.41*** At most 2 69.63** 27.68 66.94 31.09 120.28*** At most 3 41.95 24.45 35.85 14.84 72.22*** At most 4 17.5 10.71 21.02 12.48 38.19** At most 5 6.79 5.85 8.54 8.01 18.44 At most 6 0.94 0.94 0.53 0.53 7.18 Note: ** (***) denotes rejection of H0 at the 5(1) per cent level.

maxeigenvalue

112.96*** 74.13*** 48.06*** 34.03*** 19.76 11.26 7.18

Johansen tests can also shed light on the question of whether a particular country, say country j, acts as the leader of a group. If, e.g., Argentina acts as the leader of the Mercosur group, rank tests for the VAR that include variables d(i,Arg),t will indicate that the number of CE decreases exactly by one. We explored and estimated two VARs under the hypothesis that both Argentina and Brazil are leaders, motivated by the fact that they are the largest Mercosur countries in our sample. The results seem to suggest that Argentina and Brazil belong to different convergence clubs for all our income measures except GDPW. Because of the number of CE does not drop in the case of GDPW, the observed evolution of the productivity measure for Argentina and Brazil suggest that they are not leaders, instead both countries converge to a (regional) club. For GDPC and GDPCtt income versions, the number of CE reduces by one; however, it still implies that there are 3 shocks driving the system (N – r – 1 = 6 – 2 – 1 = 3 CE).31 Consequently, for GDPC and GDPCtt, it is necessary to check systematically all the possibilities to uncover robust convergence club configurations.

31

Recall that Table 4 suggests more than one convergence club for GDPC and GDPCtt.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

379

D. Univariate and multivariate UR tests applied to convergence Table 5 reports the results of testing Bernard and Durlauf’s (1995) third proposition rewritten here as Equation (8). Recall that in the KPSS test the null states that the variable under study is trend (or level) stationary. Beginning with GDPC, the null hypothesis of stationarity in levels cannot be rejected: either the leader is Argentina or Brazil (Argentina-Peru and Argentina-Uruguay at 90 per cent). But the null of stationarity around the trend can be rejected for Brazil acting as follower of Argentina and Chile follower of Brazil. Table 5: KPSS tests: GDPC, GDPCtt and GDPW, Argentina and Brazil regarded as leaders. Hypothesized followers: Bolivia, Chile, Paraguay, Peru and Uruguay, 1951-2000. leader follower

Arg.

Bra.

Brazil Bolivia Chile Paraguay Peru Uruguay Argentina Bolivia Chile Paraguay Peru Uruguay

GDPC KW KP 0.115 1.247*** 0.277*** 1.806*** 0.383*** 1.258*** 0.235*** 1.336*** 0.371*** 0.372* 0.389*** 0.433* 0.243*** 1.534*** 0.252*** 1.342*** 0.094 1.471*** 0.126* 1.206*** 0.277*** 1.419*** 0.152** 1.456***

GDPCtt KW KP 0.208** 0.408* 0.244*** 1.537*** 0.266*** 0.350* 0.238*** 1.170*** 0.156 0.125* 0.164** 0.176 0.150** 1.504*** 0.280*** 1.424*** 0.089 1.038*** 0.090 1.539*** 0.257*** 1.330*** 0.244*** 1.537***

GDPW KW KP 0.249*** 0.640** 0.265*** 1.447*** 0.247*** 0.244 0.150** 0.979*** 0.199** 0.346 0.113 0.250 0.265*** 1.447*** 0.096 1.579*** 0.172** 0.987*** 0.279*** 1.132*** 0.164** 1.363*** 0.272*** 1.171***

Note: Entries are observed KPSS statistics, see Equation (11). *(**)[***] means re*Ҙ jection of H0 at 10%(5%)[1%] level with critical values for K W: 0.119, *Ҙ 0.146, 0.216 and for K P: 0.347, 0.463, 0.739, respectively; see Kwiatkowski et al. (1992). A quadratic spectral kernel is used with l=2.

Concerning the GDPCtt, there is no evidence of convergence in levels for Argentina as leader of Peru and Argentina as leader of Uruguay. If Brazil acts as leader, Table 5 shows that there is no evidence of trend convergence when Chile and Paraguay are followers. Finally, for productivity there is no evidence of convergence in trend for Uruguay as a follower of Argentina. Also, note that Bolivia does not converge in levels to Brazil. Table 6 reports the results of MADF proposed by Taylor and Sarno (1998); note that if we use a lag order of 2 (the same used in KPSS calculated by quadratic spectral kernel) we reject at the 95 per cent level the null of UR of all series for GDPC and GDPCtt

380

Jorge Alberto Fornero

when Argentina is the leader country. It means that there exists weak evidence that not all the series are UR, and that there is consequently evidence of convergence. No support of convergence in all the series in GDPW is found when Argentina is taken as leader as well as when Brazil acts as a leader. Table 6: Taylor and Sarno (1998) tests on GDPC, GDPCtt and GDPW. Leaders: Argentina and Brazil. Hypothesized followers: Bolivia, Chile, Paraguay, Peru and Uruguay, 1951-2000 leaders obs. MADF(l=2) obs. MADF(l =1) Argentina 26.074 R 25.033 R Brazil 18.265 24.158 R Argentina 24.706 R 26.806 R GDPCtt Brazil 20.207 22.989 R Argentina 17.542 19.424 GDPW Brazil 15.481 22.845 R Note: Lags used in windows were 1 and 2. R means that the null, which states that all the series are I(1), is rejected. Critical values for l=1 and l=2 at 95 per cent level are 21.41 and 21.58, respectively; provided by Stata. series GDPC

Table 6 also shows results taking one lag (l = 1) as indicative. If Brazil is the leader of the group there is weak support for convergence for every series. Concerning Argentina as a leader, there is no support for convergence in productivity. Conversely, evidence of convergence in GDPC and GDPCtt is confirmed. For l > 2 convergence is not supported any more (results not shown). E. Clustering Analysis Is there a unique equilibrium? The last subsection suggests an affirmative answer for productivity. What the facts are telling us is that there are clubs in which a country acts as a leader. Here we discuss the results reported in Table 7 by running the cluster procedure by Hobijn and Frances (2000). The hypothesis of asymptotic perfect convergence, as tested here, is always more robust to the choice of lag length than asymptotic relative convergence. We only find evidence of convergence in GDPC, GDPCtt and GDPW for the cluster Chile-Peru when considering l = 5, but if we consider l • 5 the cluster in productivity changes to Peru-Uruguay, whereas the remaining countries follow their own balanced path.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

381

Table 7: Cluster test for income growth: GDPC, GDPCtt and GDPW. Argentina, Bolivia, Chile, Paraguay, Peru and Uruguay, 1951-2000. type of convergence

GDPC

Chile-Peru Argentina asymmetric Bolivia perfect Brazil convergence Paraguay Uruguay Argentina-Uruguay asymmetric Chile-Peru relative Bolivia convergence Brazil Paraguay

variables GDPCtt

GDPW

Chile-Peru Argentina Bolivia Brazil Paraguay Uruguay Bolivia-Uruguay Chile-Peru Brazil-Paraguay Argentina

Peru-Uruguay Argentina Bolivia Brazil Chile Paraguay Chile-Peru-Uruguay Argentina-Paraguay Bolivia Brazil

Note: Results obtained using Hobijn and Frances (2000) code. Lags for windows of 5, 6 and 6 for GDPC, GDPCtt and GDPW, respectively; p=0.05.

Let us now consider asymptotic relative convergence. As we expect, this weaker version is more unstable to changes of lag order, GDPCtt being the most unreliable to cluster. In particular, for GDPC there are two clusters integrated by Argentina-Uruguay and Peru-Chile and the rest follow their own pattern. For GDPCtt there are three clubs: Bolivia-Uruguay, Chile-Peru, Brazil-Paraguay while Argentina remains alone. Finally, a cluster of three countries is the outcome considering productivity: Chile-Peru-Uruguay and Paraguay-Argentina; Bolivia and Brazil are singletons ʊi.e., sets with only one element. To summarize the results shown, the cluster Chile-Peru seems to be a club of convergence of GDPC and GDPCtt and GDPW (relatively). No other cluster appears to be supported by the evidence, taking into account the stronger concept of convergence, i.e., in levels. Relaxing the assumption of convergence in levels to relative convergence leads to clusters with more countries. This suggests that larger members like Argentina and Brazil are never in the same club. F. Main findings from the time series analyses The results found from the time series analyses are at odds with conclusions from Camarero et al. (2002). For GDPC and GDPCtt we find that the number of CE is 2 or 3, while increasing to 6 for

382

Jorge Alberto Fornero

GDPW. The main finding shows that larger countries do not converge with each other; instead we found clusters of small-small and small-large countries. Chile and Peru is the cluster that remains whatever series is considered. In particular, we find a cluster for GDPC: Chile-Peru verify asymptotically perfect convergence. Evidence of two clusters is found if we allow for convergence in rates: Argentina-Uruguay and Chile-Peru. This is corroborated by univariate KPSS, which suggests that in levels neither Argentina nor Brazil, acts as a leader and also partly around trend because it suggests convergence between Argentina-Bolivia, for instance. However, conflicting evidence appears between MADF and the former. Whereas MADF supports the evidence whatever the leader is, KPSS states the opposite. For GDPCtt we find that Chile-Peru is a cluster for asymptotic perfect convergence and three of them are found in relative terms: Bolivia-Uruguay, Chile-Peru and Brazil-Paraguay. KPSS in trend is not consistent because it does not support convergence between Brazil and Paraguay; also it does not in levels because the test would suggest that Peru follows Argentina, and if Chile and Peru is a cluster, then it must be the case that Chile follows Argentina. The MADF test finds convergence no matter if the leader is Argentina or Brazil, which is inconsistent with the rejection of the null in KPSS tests. Finally, relatively perfect convergence in productivity is found in two clusters: Chile-Peru-Uruguay and Paraguay-Argentina. Only the cluster Chile-Peru is sustainable when the convergence concept is stronger. Here KPSS is consistent with the clusters found because all of them imply a null of KPSS rejected at usual levels. But here the conflicting evidence is MADF, which states no convergence of UR in all the series. V. Conclusions

The income convergence prediction is a testable hypothesis that we would expect to verify in the deepening of any integration processes. In particular, this empirical contribution examined whether income convergence is verified in Mercosur countries. First, in the framework of the neoclassical model, a panel specification estimated with FE FGLS suggests a speed of Econvergence that range from 4 to 6.3 per cent. When endogeneity is taken into account with system GMM, higher speeds of convergence ranging from 9.1 to 12.1 per cent are found, depending on the

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

383

income measure we consider. These point estimates are in accordance with previous studies. The most interesting result is that the speed of convergence was positively affected by the Mercosur regional integration process using panel data models with FE, estimated both with feasible generalized least squares and system GMM. Once we account for endogeneity of investment and public expenditure, we confirm results from panel data with FE, with larger estimates of Econvergence. Thus, the evidence suggests that the larger market that resulted from the constitution of Mercosur is one of the main sources of income per capita growth. Its positive impact on aggregate productivity is, however, more moderate. In addition, we find firm and robust evidence that private investment accelerates convergence. Moreover, estimates of public expenditure lead us to conclude that this aggregate demand component has no significant effect on income growth. Further, both human capital and openness seem irrelevant to explain income growth. Indeed, human capital even works in the wrong direction. One possible explanation could be that the homogeneity assumption is too restrictive. Openness, in turn, harmed estimates of our system GMM panel regressions. Second, using a time series approach we tested the presence of convergence between Mercosur countries. There is no a unique income per capita club in the LA countries considered: the number of CE is 2 and 3 for GDPC and GDPCtt, respectively. Using the Hobijn and Franses algorithm we confirm these results and also suggest that LA countries form non-homogeneous clusters, i.e., small-small and small-large countries. For productivity, we find 6 CE implying that there is one common trend process which is nonstationary. Therefore, the average GDPW (of 7 LA countries) follows a non-stationary path, which drives the whole system and all countries belong to the same convergence club. However, these results could not be confirmed by the Hobijn and Franses algorithm, which would suggest that the number of clusters exceeds one. Finally, no matter the income series, the club that comprises Chile and Peru is a robust cluster. To sum up, unambiguously, we find that the combination of investment and integration into Mercosur are the most important variables driving the speed of convergence among members, with levels of income and productivity quite persistent. The interesting question, however, concerns convergence towards which path or paths. The evidence does not point to a definitive answer; it

384

Jorge Alberto Fornero

depends on the method and the income measure. We conclude that Mercosur countries converge in productivity in a robust way. Notwithstanding, some doubts emerged from the clustering analysis which, in our view, deserves more attention. This would suggest that a possibly promising area of future research could be to take explicitly into account structural breaks in the specification.

References

Abuaf, Niso, Jorion, Philippe (1990), Purchasing Power Parity in the Long Run, Journal of Finance XLV(1), 157-174. Aghion, Philippe, Howitt, Peter (1992), A Model of Growth Through Creative Destruction, Econometrica 60(2), 323-351. Anderson, T.W., Hsiao, Cheng (1981), Estimation of dynamic models with error components, Journal of the American Statistical Association 76, 589-606. Arellano, Manuel, Bond, Stephen (1991), Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations, Review of Economic Studies 58(2), 277-97. Arellano, Manuel, Bover, Olympia (1995), Another look at the instrumental variable estimation of error-components models, Journal of Econometrics 68, 29-51. Badinger, Harald, Müller, Werner, Tondl, Gabriele (2004), Regional Convergence in the European Union, 1985- 1999: A Spatial Dynamic Panel Analysis, Regional Studies 38(3), 241253. Baltagi, Jadi (2001), Econometric Analysis of Panel Data, 2nd ed. John Wiley & Sons, Canada. Barro, Robert (1991), Economic growth in a cross-section of countries, The Quarterly Journal of Economics 106, 407-443. Barro, Robert (1997), Determinants of Economic Growth: A crosscountry Empirical Study, MIT Press, Cambridge, MA.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

385

Barro, Robert, Sala-I-Martin, Xavier (2004), Economic Growth, 2nd ed. MIT Press, Cambridge, Mass. Barro, Robert, Lee, Jong-Wha (1993), Losers and Winners in Economic Growth, NBER Working Paper No. 4341. Barro, Robert, Lee, Jong-Wha (2000), International Data on Educational Attainment: Updates and Implications, CID Working Paper No. 42. Baumol, William (1986), Productivity growth, convergence, and welfare, American Economic Review 76(5), 1072-1085. Bernard, Andrew, Durlauf, Stephen (1995), Convergence in international output, Journal of Applied Econometrics 10, 97108. Bernard, Andrew, Durlauf, Stephen (1996), Interpreting tests of the convergence hypothesis, Journal of Econometrics 71, 161-173. Blundell, Richard, Bond, Stephen (1998), Initial conditions and moment restrictions in dynamic panel data models, Journal of Econometrics 87, 115-143. Breuer, Janice, McNown, Robert, Wallace, Myles (2002), Seriesspecific Unit Root Tests with Panel Data, Oxford Bulletin of Economics and Statistics 64(5), 527-546. Canova, Fabio, Marcet, Albert (1995), The Poor Stay Poor: NonConvergence Across Countries, Regions, CEPR Discussion Papers 1265. Caselli, Francesco, Esquivel, Guillermo, Lefort, Fernando (1996), Reopening the convergence debate: a new look at cross-country growth empirics, Journal of Economic Growth 1, 363-389. Cass, David (1965), Optimum growth in an aggregate model of capital accumulation, Review of Economic Studies 32, 233-240. Camarero, Mariam, Flôres, Renato, Tamarit, Cecilio (2002), Multivariate time series evidence of international output convergence in Mercosur, Mimeo. Choi, In (1994), Residual based tests for the null of stationarity with applications to U.S. macroeconomic time series, Econometric Theory 10, 720–746.

386

Jorge Alberto Fornero

Durlauf, Stephen, Quah, Danny (1999), The new empirics of economic growth, in: Taylor, J.B. and Woodford M. (eds.), Handbook of Macroeconomics, Volume 1A, North-Holland, Amsterdam, 235-308. Flôres, Renato, Preumont, Philippe, Szafarz, Arian (1996), Multivariate Unit Root Test, mimeo, Université Libre de Bruxelles. Greene, William (2003), Econometric Analysis, Prentice – Hall Inc., New York, 5th edition. Heston, Alan, Summers, Robert, Aten, Betina (2002), Penn World Table v 6.1 available on internet. Hobijn, Bart, Frances, Philip (2000), Asymptotically perfect and relative convergence of productivity, Journal of Applied Econometrics 15, 59-81. Hobijn, Bart, Franses, Philip, Ooms, Marius, (2004), Generalizations of the KPSS-test for stationarity, Statistica Neerlandica 58(4), 483-502. Hsiao, Cheng (1986), Analysis of Panel Data, Cambridge University Press, Cambridge, UK. Im, Kyung, Pesaran, M. Hashem, Shin, Yongcheol (2003), Testing for unit roots in heterogeneous panels, Journal of Econometrics 115, 53-74. Islam, Nazrul (1995), Growth empirics: A panel data approach, The Quarterly Journal of Economics 110(4), 1127-1170. Kwiatkowski, Denis, Phillips, Peter, Schmidt, Peter, Shin, Yongcheol (1992). Testing the Null of Stationarity Against the Alternative of a Unit Root, Journal of Econometrics 54, 159178. Koopmans, Tjalling (1965), On the concept of optimal economic growth, in McNally R (ed), Pontificiae Academiae Scientiarum Scripta Varia 28. In Study Week on the Econometric Approach to Development Planning, pp. 225–287. Leamer, Edward (1985), Sensitivity Analyses Would Help, American Economic Review 75(3), 308-313.

Do Mercosur Countries Converge in Per Capita GDP and Productivity?

387

Lee, Kevin, Pesaran, M. Hashem, Smith, Ron (1997), Growth and convergence in a multi-country empirical stochastic Solow model, Journal of Applied Econometrics 12, 357-392. Levin, Andrew, Lin, Chien-Fu, James Chu, Chia-Shang (2002), Unit Root Tests in Panel Data: Asymptotic and Finite-Sample Properties, Journal of Econometrics 108, 1-24. Levine, Ross, Renelt, David (1992), A sensitivity analysis of crosscountry growth regressions, American Economic Review 82(4), 942-963. Leybourne, Stephen, McCabe, Brendan (1994), A Consistent test for a unit root, Journal of Business and Economic Statistics 12, 157-166. Lim, Kian, McAleer, Michael (2004), Convergence and catching-up in ASEAN: a comparative analysis, Applied Economics 36, 137153. Mankiw, N. Gregory (1995), The Growth of Nations, Brookings Papers on Economic Activity 1, 275-310. Mankiw, N. Gregory, Romer, David, Weil, David (1992), A contribution to the empirics of economic growth, The Quarterly Journal of Economics 107, 407-437. Pesaran, M. Hashem, Smith, Ron (1995), Estimating long-run relationships from dynamic heterogeneous panels, Journal of Econometrics 68, 79-113. Quah, Danny (1993), Galton’s Fallacy and Tests of the Convergence Hypothesis, The Scandinavian Journal of Economics 95 (4), 427-443. Quah, Danny (1996a), Convergence empirics across economies with (some) capital mobility, Journal of Economic Growth 1, 95-124. Quah, Danny (1996b), Twin peaks: Growth and convergence in models of distribution dynamics, The Economic Journal 106, 1045-1055. Reichlin, Lucrezia (1999). Discussion of convergence as distribution dynamics, in: Baldwin RD, Cohen D, Sapir A, Venables A (eds), Market Integration, Regionalism, and the

388

Jorge Alberto Fornero

Global Economy, Cambridge University Press and CEPR, Cambridge. Romer Paul (1990), Endogenous Technological Change, The Journal of Political Economy 98(5), 71-102. Sala-i-Martin, Xavier (1996), Regional Cohesion: Evidence and Theories of Regional Growth and Convergence, European Economic Review 40, 1325-1352. Sala-i-Martin, Xavier (1997), I ran just two million regressions. American Economic Association Papers and Proceedings 87(2), 178-183. Sarno, Lucio, Taylor, Mark (1998), Real exchange rates under the recent float: unequivocal evidence of mean reversion, Economics Letters 60, 131-137. Solow, Robert (1956), A contribution to the theory of economic growth, Quarterly Journal of Economics 70, 65-94. Temple, Jonathan (1998), Robustness Tests of the Augmented Solow Model, Journal of Applied Econometrics 13(4), 361-375. Taylor, Mark, Sarno, Lucio (1998), The behavior of real exchange rates during the post-Bretton Woods period, Journal of International Economics 46, 281-312.

Basilia Aguirre and Guilherme Dias

Fiscal Reform and Federal Relations Abstract Brazilian fiscal reform became deadlocked because reform proposals tried to avoid considering federal relations. There are two main types of federal relations: the competitive and the cooperative. With both types coordination mechanisms can be observed. Brazilian federalism is a mixture of both types which makes reforms difficult. We argue that it is more important to find mechanisms to facilitate cooperation than to discuss the qualities of any alternative fiscal structure. Moreover, since fiscal reforms entail a great deal of uncertainty, it is important to discuss the timing of reforms and the compensation mechanisms beforehand. I. Introduction The Brazilian tax burden is amazingly high even when compared to developed countries. This is contrast to other developing countries. In the beginning of the 1990s, the tax burden in Brazil accounted already for 25 per cent of GDP and in 2004 it reached 32.8 per cent of GDP1. In principle, this does not yet indicate a problematic fiscal structure. However, the way in which this high burden was generated is considered to be really harmful for the Brazilian productive structure and it is also not favorable for the cooperation between the different levels of government. Economists agree that a major fiscal reform is long needed. However, after the stabilization in 1994 all attempts to implement a broad reform failed. The fiscal reform debate avoided to discuss the problems of federal relations fearing that this could increase the resistance to changes. However, in a federation it is almost impossible to discuss fiscal reforms without dealing with the conflicting interests of different federal levels. In our view the fiscal reform debate in Brazil has taken a route that will not lead to a solution. Since the 1988 Constitution, the Brazilian fiscal structure has been changed in a way that may turn 1

This is the official figure after the revision made by IBGE in 2005.

390

Basilia Aguirre and Guilherme Dias

out problematic for the Brazilian economic structure. In this paper we will argue that a better approach to fiscal problems would require understanding the reasons of deadlocks and poor cooperation among actors involved. Institutional economics, in particular the approach of North (1990; 2005) and the extensions offered by Dixit (1996) and Acemoglu et al. (2004), provides an adequate framework to address this issue. The rest of the paper is organized as follows. Section II discusses the Brazilian fiscal problems and the failure of the reform proposals that where advanced to solve them. In Section III we will discuss how intergovernmental problems can be understood as a problem of allocation of property rights to different government levels in the presence of distributional conflict. In Section IV we identify the main conflicts that can be observed in inter-governmental relations and finally, in Section V, we show which are the main reasons of the observed paralysis in fiscal reforms and propose alternative approaches to reform. II. Fiscal Structure Problems in Brazil and Initiated Reforms The deficiencies of Brazilian federal fiscal structure were pointed out in numerous studies.2 They consider the following problems as most important: the inadequacy of the fiscal structure after having macroeconomic reforms and trade liberalisation, an excessive tax burden for firms, fiscal evasion, an excessive payroll tax burden, fiscal war, regressive taxation, vertical imbalances, cascading taxation, unclearly defined competencies between levels of government in many areas of public policy. The studies also agree on the consequences: adverse effects on the productive sector, particularly on the export sector, missing coordination between different levels of government in tax collection and fiscal expenditure, maintained or even worsened income inequality. The origins of these problems can be traced back to the extensive fiscal reforms undertaken by the Military Government in 1967 and the by 1988 Constitution. Moreover a large number of small changes have significantly altered the fiscal structure. The main features are shown in Figure 1, depicting the development of the Brazilian tax burden and the composition of fiscal expenditures for the three levels of government. First of all one will 2

See for example: Afonso and De Mello (2000), Afonso et al. (2000), Aguirre (1999), Varsano (1995;1999;2002) and Werneck (2000;2002).

Fiscal Reform and Federal Relations

391

note that almost during the whole period fiscal unbalances can be observed. However, with just a few exceptions, the federal government registered a fiscal surplus compared to primary expenditure whereas state and municipality levels showed a deficit. These differences can be explained by transfers and debts. Until 1998, states and municipalities where allowed to borrow from state owned banks. On the other hand, transfers have been a common instrument of the federal government since the advent of the republic in 1889 to cope with regional imbalances. In 1967, they became compulsory by constitution and the federal government started to transfer automatically a significant amount of its revenues to states and municipalities. Figure 1: Development of Fiscal Composition in Brazil 90% 80%

1982

1929 1945

1988

1964

70% 60% 50% 40% 30% 20% 10% 0% 1900 1907 1914 1921 1928 1935 1942 1949 1956 1963 1970 1977 1984 1991 1998 Federal Tax Burden State Tax Burden State Expenditures

Federal Expenditures Muncipality Tax Burden Municipal Expenditures

Source: 1900-1945: IBRE/FGV-IPEA. 1946-1990: IBGE

However, the most striking fact appearing in Figure 1 is the coincidence of major political events in Brazil with changes in fiscal structures. Before 1929, when states where totally autonomous to raise taxes and to borrow, we see an increasing trend of state revenues and expenditures while those of the federal government decreased. Between 1930 and 1945, a period in turmoil caused by internal and external events, we observe a reversal of this tendency, mainly due to the measures adopted by the Vargas dictatorship in order to increase the power of the central government power relative to states. During the re-democratization period

392

Basilia Aguirre and Guilherme Dias

between 1945 and 1964, we watch relative stability. After the fiscal reform implemented by the Military Regime in 1966, state governments suffered a huge decline in their tax share while municipalities increased their share in expenditures due to larger transfers from federal tax collection. The re-democratization process, starting in 1982 with the first election for state government in 18 years, changed again the pattern. Federal government began to loose in participation while state and municipalities gained. After the 1988 Constitution, the trend is again altered with federal government trying to recuperate losses made between 1982 and 1988. In summary, Figure 1 shows a century long distributional conflict between the different levels of government. Since 1970, the Brazilian public indebtedness has been at the centre of fiscal problems. Until 1998, all three levels of government were allowed to borrow without restrictions even from their own financial institutions. The traditional literature on Brazilian fiscal structure problems concentrates on proposals to solve them without paying attention to the fact that economic policy is a matter of political outcomes. Usually significant allocative inefficiencies are diagnosed and technically designed reforms are proposed to solve them. Moreover, if this does not happen it is because of a lack in political will. In our view, the problem is much more complex. It involves distributional conflicts between levels of government. Missing cooperation mechanisms in inter-government relations have been dampening the possibilities of reform. A better understanding of the conflict is necessary in order to tackle the existing problems more successfully. A good evidence of the point made above is the sequence of proposals for fiscal reform during the last ten years. In this period, Brazilian society witnessed at least five reform attempts by the federal government, none of them were adopted. They were initiated in 1997 during the Cardoso government. Werneck (2002; 2003) analyses the proposed reforms: “They seem to share the same basic diagnosis on what is wrong with indirect taxation in the country. They are strikingly similar in what concerns taxes to be eliminated. And are not so dissimilar in what concerns taxes to be created.” (Werneck, 2003, p. 9) The proposals mainly suggested a

Fiscal Reform and Federal Relations

393

Table 1: Fiscal reform proposals in Brazil Proposal

Taxes to be eliminated

Taxes to be created

Executive’s Late 1997 Proposal October 1997

All turnover and cascading taxes (Cofins, PIS-Pasep), except tax on financial transactions Federal tax on manufactured products (IPI) State VAT (ICMS) Service tax charged by local governments (ISS) All turnover and cascading taxes (Cofins, PIS-Pasep), except tax on financial transactions Federal tax on manufactured products (IPI) State VAT (ICMS)

Nationally-managed VAT

Executive’s Late 1999 Proposal October 1999

Executive’s Proposal August 2000

Service tax charged by local governments (ISS) Federal tax on manufactured products (IPI) State VAT (ICMS)

Service tax charged by local governments (ISS) All turnover and cascading Special taxes (Cofins, PIS-Pasep, Committee’s CPMF) Proposal March Federal tax on manufactured 2000 products (IPI) State VAT (ICMS) Service tax charged by local governments (ISS) Non-voted All turnover and cascading Rapporteurs’s taxes (Cofins, PIS-Pasep, Proposal March CPMF) 2000 Federal tax on manufactured products (IPI) State VAT (ICMS) Service tax charged by local governments (ISS) Source: Werneck (2003).

Federal excise tax on goods and services Retail sales tax (IVV)

Federal VAT

State excise tax on goods and services Municipal retail sales tax (IVV)

Federal tax on goods and services (IBS) Nationally uniformed state VAT Municipal retail sales tax (IVV) Dual VAT (coexisting federal and state VATs) Municipal retail sales tax (IVV)

Dual VAT (coexisting federal and state VATs) Non-cumulative excise tax

394

Basilia Aguirre and Guilherme Dias

reform of the VAT tax, which is today the main revenue source of state governments, the creation of a federal VAT tax and the elimination of a bundle of taxes considered to have negative effects on the economy. Table 1 presents the main features of these proposals. However, we do not agree with Werneck stating that: “…Considering that two of those proposals stemmed from the Executive and two from Congress that seems to point out to a surprising degree of agreement on the main line of the required reform.” (Werneck, 2003, p. 9)3 In our view there are significant differences between the proposals that emerge from the federal executive and those that came from the Congress. If we rearrange the information in Table 1 highlighting the proponents the picture changes considerably. Table 2 presents the same information, but it focuses on the origin of the proposals. Starting with the taxes to be eliminated, we see that executive and legislative powers agreed that three taxes should be eliminated: Federal tax on manufactured products (IPI), Service tax charged by local governments (ISS) and State VAT. But this is the only agreement that appears. The two parties disagreed on the elimination of the tax on financial transactions (CPMF) which is one of the cascading federal taxes. Legislative power proposed to abolish it while the executive would never have accepted its elimination. The federal government was not prepared to renounce its major source of revenues although it damages the productive structure of the country. With respect to the creation of additional taxes, we do not find a single agreement between executive and legislative power. The two legislative proposals maintained the idea of a Dual VAT that should be collected by both federal and state levels. In contrast, the executive proposal would have transformed state VAT into federal VAT. Only the last proposal for reform, that was not supposed to be a serious offer because it came too late in the legislature, federal government proposed a state VAT which should be nationally uniform. The idea of a nationally uniform VAT is something that in a way violates the very concept of federalism which should allow states some degree of power to collect and to legislate on taxes.

3

The author does not make reference to the fifth proposal because he considers that it was only to end the society’s outcry for fiscal reform at the end of the Fernando Henrique government.

Fiscal Reform and Federal Relations

395

Table 2: Another picture of the reform proposals Executive Legislative Taxes to be eliminated 2 All turnover and cascading taxes (Cofins, PISPasep), except tax on financial transactions All turnover and cascading taxes (Cofins, PIS2 Pasep, CPMF) 3 2 Federal tax on manufactured products (IPI) 3 2 State VAT (ICMS) Service tax charged by local governments 3 2 (ISS) Taxes to be created 2 Dual VAT (coexisting federal and state VATs) 1 Federal excise tax on goods and services Federal VAT 1 Nationally-managed VAT 1 Federal tax on goods and services (IBS) 1 1 Nationally uniformed state VAT 2 1 Municipal retail sales tax (IVV) 1 Non-cumulative excise tax 1 Retail sales tax (IVV) 1 State excise tax on goods and services Source: own elaboration based on Werneck (2003).

III. The Fiscal Federal Structure as Allocation of Property Rights between Government Levels The discussion above highlights the fact that fiscal reform in Brazil is a matter of intergovernmental relations. In this section we will argue that federations are institutional structures that need coordination mechanisms because of the important distributional conflict between federal members. The inclusion of conflict into the analysis unveils a completely different world, improving the understanding of fiscal federal relations. For that purpose, let us first recall the concept of fiscal federalism as used traditionally in economics. The literature on fiscal federalism is concerned with the efficiency of federal arrangements. Adopting a normative attitude it seeks to define what should be the optimal level of fiscal decentralization. The main problem faced in the fiscal federalism literature is the existence of externalities. Oates (1972) was aware of the problem when he stated that: “...the optimal degree of fiscal decentralization

396

Basilia Aguirre and Guilherme Dias

will vary substantially among different societies”. (Oates, 1972, p. 31). The solution proposed by Oates (1972) was the inclusion of transfers from central government in the analysis since that could handle the externalities problem through matching revenues and jurisdictions. This suggestion is based in the traditional Pigouvian solution to the externalities problem. Transfers from central to lower levels of government would reallocate income from jurisdictions that impose negative externalities to those that suffer it. However this process is not costless. As Coase (1960) demonstrated, if the use of the price system has costs then the Pigouvian solution may not be the best. Transfers will demand negotiations between the involved jurisdictions that may compensate the welfare gains from decentralization. Although many authors tried to deal with the problem of externalities related to federalism,4 none of was able to present a reasonable solution within the limits of conventional economics. Oates´ approach to this problem continues to be the main contribution to the question. According to Coase (1937) the costly nature of the price system is mainly due to uncertainty. If the actors knew which would be the best agreement they would not need to negotiate to get to the best alternative. It will be automatic as in the case of the so called Coase Theorem. It is also important to remember that when talking about fiscal federalism we deal with a problem that involves politics. This brings us to the contribution of Dixit (1996, p. 20) to the problem of transaction costs in political transaction. Laws and constitutions last long, economic circumstances change a lot during constitution life, conflicts about contingencies are much more frequent in the political arena, and there is a larger number of interests involved. Besides constitutions should be built in sufficiently general terms as to be applicable to many circumstances. On the other side, there is a wide array of interpretation of the meaning of orderings and thus constitutions allow much more manipulation than other kinds of contracts. These characteristics impose two important consequences. On the one hand there is the “temporal consistency” in politics, on the other hand the multi agency problem. Problems of temporal consistency arise because contracts in politics are very loose. Property rights are very blurred. Principals have no means as to make politicians or bureaucrats account4

For example Besley and Coate (1999), Inman and Rubinfeld (1997) and Inman and Fitts (1990).

Fiscal Reform and Federal Relations

397

able. Actually it is not desired that politicians become fully accountable since that would mean that vote would not be secret any more in democracies. Moreover, there are many difficulties to establish a clear relationship between agents and principals. Principals are usually the voters. This group, however, does not have a single interest. Even if some groups exhibit a common interest one can find always other kind of interests where they would have conflicting views. Agents are a heterogeneous group as well. Conflicts are also vivid among them. Conflicts are not only important between politicians and bureaucrats but also within the groups of politicians and bureaucrats. When the federalism problem is incorporated into the analysis the complexity increases enormously. Applying the above argument to federalism, one can say that a federal structure is a contract between society and government, but moreover, it is a contract within the state between autonomous units of policy making. (In countries organized on an unitary basis, this characteristic does not arise.) This means that decisions depend not only on central government decisions but also on the decisions of lower level governments, which makes it necessary for government levels to coordinate their action in order to diminish inefficiencies. However, considering the problem of incompleteness of contracts in the political arena, this is an issue that should not be handled as a maximization problem as the conventional theory would do. Probably a more adequate approach to deal with this issue in our case is to use North’s idea of uncertainty reduction. Following his arguments, we should look at the problem of establishing rights and competencies for various government levels, in a federal structure, as a means to reduce conflicts and to improve coordination. Institutional design in this case is not an issue of finding an optimal design. Instead it is a question of finding what set of rules could attain the objectives pointed out above in a specific political and economic environment. Two kinds of distributional conflicts can be observed in this case. The first one is about distribution of tax collection and competencies between levels of government; the second is about income distribution between private and public sectors. Acemoglu et al. (2004), discussing the importance of institutions for economic development, presents a framework of analysis that incorporates distributional conflicts and highlights the importance of political institutions. His arguments emphasize political power. Political power has two components: de jure political power and de facto

398

Basilia Aguirre and Guilherme Dias

political power. De facto political power is determined by the distribution of resources whereas de jure political power depends on political institutions. The combinations of the two political powers determine both economic institutions in the present period and political institutions in the next period. In doing so the authors endogenize all the variables into the model and highlight the importance of political institutions and political power to the determination of economic policy and economic performance. Very important for their argument is the idea that the choice of economic institutions is a problem of collective decision. Different groups have different preferences for institutions, because different sets of institutions will be responsible for different distribution of resources. In a situation like this no optimal solution is possible. This is why the authors emphasize that political power is crucial to the outcome. The question that arises at this point is: which will be the set of equilibrium institutions? And the answer to this question is: the set of institutions that is preferred by the group with the largest political power. This solution is not necessarily optimal in the sense that no one loses, it is, however, an equilibrium solution since no group will wish to devote additional resource to change the situation unless some of the initial conditions change. What changes in initial conditions could alter the result? To answer this question we have to understand better what constitutes political power. Political power according to the definition of the authors has two elements. The first one, de facto power, involves two aspects: economic resources and capacity to solve problems of collective action. Economic resources are usually taken into consideration in economic theory when dealing with questions like this. The capacity to deal with problems of collective action is not considered. This has to do with what Olson (1971) called our attention in the Logic of Collective Action, meaning group’s action efficiency in the presence of differences in common and individual interests. The second element in the analysis of Acemoglu et al. (2004) is de jure political power. This element is not common in economic theory. It deals with political rules and allocation of political rights. They establish the reach of the agents’ action in the political arena and in doing so they are also responsible for determining the possible outcomes. Using the insights of Dixit (1996) and Acemoglu et al. (2004), it is possible to build a framework of analysis for federal relations and

Fiscal Reform and Federal Relations

399

distribution of rights among levels of government. For the sake of simplicity it is possible to start considering only two government levels that are independent units of decision making. However, their actions are inter-dependent which means that there are externalities problems. Federal relations are not a natural cooperative relation. There is a huge potential for conflict in these situation. Federal members compete mainly for tax basis, but they also compete for the provision of public goods. So we have a situation of distributional conflict in an area where contracts are very loose as pointed out by Dixit (1996). Each unity would prefer to have the largest share of tax receipts and, at the same time to provide the minimum amount of public goods that would not risk losing votes. The rational behavior would be to try to make the other government sphere to provide the goods without losing taxes or votes. When this is the behavior of all members, the outcome would be a minimum provision of public goods. To solve this problem some kind of coordination mechanism should be introduced. In this case it is possible to define two main ideal kinds of relationship between levels of government. We should call them: cooperative 5 and competitive. The difference between these two forms of federalism lies in the nature of the coordination mechanism. In cooperative federalism, a policy executed in the territory of a member of the federation should be agreed between central and lower governmental levels. On the other hand, competitive federalism arises when levels of government compete for tax bases and supply of services for each federal member. The important point in this classification of federalism types is that they have very different consequences for the transaction costs of inter-governmental relationships. On the one hand, with cooperative federalism high ex-ante transaction costs arise during the period prior to the agreement that determines decision rights among 5

Inman and Rubinfeld (1997) offer also a definition of cooperative federalism which involves the unanimity rule. They define cooperative federalism as the principle to “prefer the most decentralized structure capable of internalizing all economic externalities, subject to the constitutional constraint that all central government policies are agreed to unanimously by the elected representatives from each of the lower-tier governments.” (Inman and Rubinfeld (1997) p. 48) This definition of cooperative federalism is based on the American federalism structure and is not useful for our purpose.

400

Basilia Aguirre and Guilherme Dias

federal members. However, there should be low ex-post transaction costs in the period of policy implementation and execution, if we can suppose that in this stage members do not compete. This of course will depend on the specific rules that defines federal relations. If competencies are clearly established in the constitution, one can expect that ex-post transaction costs will stay at a minimum.6 On the other hand, competitive federalism will exhibit minimum ex-ante transaction costs since there is no initial negotiation between federal government levels, however, there will be high ex-post transaction costs due to the lack of coordination for the provision of public goods between federal members. This should produce inefficiencies in the provision of public goods mainly due to externalities. As a consequence, a priori it is not possible to conclude on the superiority of cooperative federalism over competitive federalism or vice-versa. The trade-off between ex-ante and ex-post transaction costs may make both forms of federalism similar in terms of total transaction costs. However, competitive and cooperative federalism are pure Weberian ideal types. They are just abstract representations of federal arrangements that help us think about current federal structures in terms of solving conflicts in the process of intergovernmental relations. We can also think of them as extremes of a continuum which allows many other federal organization forms. Each of the extremes is a situation where ex-ante or ex-post transaction costs are at such a low level as to produce the lowest total transaction costs possible since the cooperation mechanisms at work are the best possible. Intermediate positions along the continuum represent total transaction costs higher then on the extremes. This means that various combinations of elements of competitive federalism with cooperative federalism will be less efficient alternatives to various degrees. Why should the intermediate positions be less efficient? Federal organizations are contract networks. As mentioned above, they provide members with the basic rules for their relationship. When these rules offer conflicting signals, the functioning of the system may suffer. Cooperative federalism is an arrangement that is close to the idea of an organization, like a firm, in Coase’s definition where things are settled through agreements. Competitive federal6

However, in this case there may be high ex-post transaction costs for institutional changes due to difficulties to change rigid constitutions.

Fiscal Reform and Federal Relations

401

ism is close to the idea of a market where solutions are reached through agent’s individual efforts to improve his or her performance, independently of the others. If the arrangement allows agents to seek their individual interests within an agreement, the result would be at the expense of cooperation. There are a few basic characteristics that distinguish cooperative and competitive federalism. They are reflected in the fiscal structure. To be more precise, in economic terms cooperative federalism operates in a system of tax-sharing. Taxes are collected by the central and/or state governments but are usually distributed according to the responsibilities of each member. It does not matter much which level is collecting. The allocation of funds is governed by the principle of equal capacity to provide public or collective goods in the territory of each member state and this principle is usually stated in the constitution or by law. In competitive federalism, members of the federation make individual efforts to collect taxes and there is no agreement on sharing the results of these individual collection efforts. In cooperative federalism, the tax base is also a matter of negotiation. Each federal level is attributed a specific base which cannot be taxed by other levels. In competitive federalism, federal and state levels compete for a common tax base like, for example, income. The definition of competencies in cooperative federalism is stated in the constitution or by law while in competitive federalism it is left to each member to decide what to do. Finally, with respect to transfers, in cooperative federalism there are compulsory transfers that can be expended freely by lower government levels. Usually, they constitute compensatory transfers to attain a similar capacity of expending in all states. Transfers in competitive federalism are voluntary, earmarked and they are paid mainly to individuals and organizations rather than to other government levels. Table 3 summarizes the main features of each federalism ideal type. Table 3: Main Characteristics of Ideal Federalism Types cooperative

competitive

tax-sharing

exclusive taxation

definition of tax base per level of government

competition for tax base

definition of competencies compulsory transfers free transfers compensatory transfers to governments

common competences voluntary transfers earmarked transfers transfers to individuals

402

Basilia Aguirre and Guilherme Dias

Each of these ideal federalism types has its own cooperation mechanism and the characteristics described above. In the real world, we neither find a perfectly cooperative nor a perfectly competitive federalism. We can observe a variety of combinations of competitive and cooperative federalism. Each of them will show transaction cost that are higher than that of the ideal types. Some countries exhibit a higher degree of competitive federalism, like for example the US, while others are mainly characterized by cooperative federalism, like Canada or Germany. The question arises which are the minimum transaction costs possible for a given country? The minimum transaction costs possible for a given situation will emerge from the combination of features that reduces conflicts between jurisdictions, permits to internalize the maximum of externalities and provides the highest level of public goods. Let us discuss this question in the context of the situation in Brazil. We argue that in the Brazilian case there are certain features of both competitive and cooperative federalism that are combined and that are responsible for increasing transaction costs due to the weakening of the cooperation mechanisms. It will be shown that this result is a consequence of a path dependent process in fiscal structure that has evolved since the Brazilian Military Government and whose deadlocks still wait to be solved. IV. Evolution of Fiscal Structure and Path Dependence Since the 1988 Constitution a series of fiscal changes were introduced although no major fiscal reform was implemented. The 1988 Constitution transferred a significant amount of resources from federal to state and municipal levels, however, without transferring competencies. The new situation was one of state and municipal enrichment contrasted by federal level impoverishment. This situation was aggravated by the nonexistence of controls over state and municipal indebtedness. Figure 2 shows the trends in fiscal revenues and disposable revenues by government level.

Fiscal Reform and Federal Relations

403

Figure 2: Revenue and Disposable Revenue in per cent GDP by Government Level 1960-2001 25

20

15

10

5

0 1960

1970

1980

1989

1991

1993

1995

1997

1999

2001

2003

Central Revenues % of GDP Central Disposable Revenues % of GDP State Revenues % of GDP State Disposable Revenues % of GDP Local Revenues % of GDP Local Disposable Revenues % of GDP

Source: IBGE

From 1980 to 1991 federal government revenues shares show a variable pattern while state and local levels increase their shares. This was the result of changes in government level’s property rights. Constitutional federal government transfers to both state and local government increased as well as transfers from state to local government. Local government was the level that received the largest gain (see Figure 3). This period was characterized by hyperinflation. This new fiscal structure was a result of a successful mobilization for decentralization made mainly by local government level politicians that managed to influence legislators. It imposed a heavy burden on the Union since the majority of responsibilities were still poorly defined and the federal government maintained all its responsibilities. Figure 3 offers a good picture of the intergovernmental distributional conflict. The principal trend shows a continuous federal loss, fairly stable tendency for state governments and a continuous gain of local levels. However, there are some moments in this period, when we see a reaction from federal government and others when state or local governments succeeded to regain terrain. Consequently, in Brazilian federalism periods of cen-

404

Basilia Aguirre and Guilherme Dias

tralization and decentralization follow each other. The important issue to investigate is the mechanisms used by the three government levels in this conflict. In other words, to consider how intergovernmental rights evolved. Figure 3: Revenues Gains and Losses by Government Level 1960-2001 25

20

15

10

5

0 1960

1970

1980

1989

1991

1993

1995

1997

1999

2001

2003

Central Revenues % of GDP Central Disposable Revenues % of GDP State Revenues % of GDP State Disposable Revenues % of GDP Local Revenues % of GDP Local Disposable Revenues % of GDP

Source: IBGE.

Looking at Figure 3, we can identify 5 periods after 1980 when changes in fiscal structures occurred. Each of these changes represents an attack from one or more federation levels on the revenues share of another. From 1980 to 1988 both federal and state governments lost shares to the local level. Between 1988 and 1990 the federal government managed to recuperate part of its loss, mainly at the expenses of the states, while local governments suffered a slight loss. The period 1990 to 1994 exhibits a certain stability that was interrupted in 1995 when federal government suffered a huge loss. Only after 1997, federal government recuperated in part its share of revenues. These developments make clear that a fierce battle between federal members, which could not coordinate their action, took place. The only way to defend itself was seen in attacking others. We will show that this process is the result of the absence of negotiating power among federation members.

Fiscal Reform and Federal Relations

405

To change a situation like this is necessary to change the constitution in Brazil. There are two ways to do this: a constitutional revision or a constitutional amendment. A constitutional amendment requires the approval of 2/3 of the Congress. A constitutional revision requires only a 50 per cent approval but cannot be done anytime. It takes place in a wider process that is not limited to the issue at stake. In 1993, the constitution was revised and federal government suffered another unfavorable impact on its tax shares. Urgent macroeconomic problems seem to have determined the apathy of federal government concerning fiscal problems. An alternative explanation is that federal government was, at that time, envisaging other formulas to solve its financial crisis. It therefore decided not to enter into difficult negotiations with congress which might have even worsened its situation. From 1988 to 1990, federal government succeeded to reverse the situation mainly by reducing non constitutional transfers and increasing tax rates of non-shared taxes. This was a reaction to the trend after the early 1980s with the re-democratization process. Unfortunately there is no disaggregated data on federal revenues for this period. Nevertheless, we observe in Figure 2 that federal government revenues increased much faster than the revenues of other levels of government. This result can only be attributed to an increase in taxes. It became clear that the success of federal government was only obtained because it imposed revenue losses to other government levels without negotiations. In the period 1990 to 1994 certain stability appears. The possibility of changing the constitution in 1993 might explain this outcome. But instead of ameliorating federal finances the constitutional revision imposed another loss on federal government.7 This loss was mainly due to the prohibition to collect no other taxes than states VAT from public utilities like electricity, inter-state and municipal transportation, on communication and on fuel and minerals.8 This was another victory of lower government levels since this tax is collected by state governments and shared with local governments.

7

8

Constitutional Revision Amendment No 3/1996 was introduced in Article 155 that defines states taxes. Also for this period there is no disaggregated data on tax collection available.

406

Basilia Aguirre and Guilherme Dias

It took three years for a federal government to reaction. In the meantime two other important events occurred. The first was the renegotiation of state debt and the other was the beginning of an establishment of clear responsibilities between federal members. The debt of states and municipalities accounted for 11,5 per cent of GDP in 1991, in 1996 it reached 5,8 per cent of GDP. After the macroeconomic stabilization process it became inevitable to control the debts of states as part of stabilization efforts. The federal government attempted to force states to control debts by offering them a negotiation on debts. According to Rezende and Afonso (2000) state bonds were exchanged by federal bonds, real interests rates were set to a maximum of 7,5 per cent and there was also a provision that interest payments could not exceed 11 per cent of states tax collection. As a result, after 1997 states interest payments declined significantly while federal interest payments increased sharply. The debts of states increased at a much lower rate than federal debts. (See Figures 4 and 5) Once again federal government suffered another loss. In 1996, a new set of norms was passed which intended to provide a clear definition of responsibilities of government levels concerning health services. The NOBSUS-96, as it was called, established that the provision of public health services should be done by local level governments and the costs should be shared between the three levels. Later in 1998 another set of norms established responsibilities for education policy following the same principles. Both initiatives represented an additional burden on federal government since it became responsible for an important part of the financing.

Fiscal Reform and Federal Relations

407

Figure 4: Interest Payments of Federal Government, States and Municipalities, 1995-2003, per cent of GDP 10 9

Federal

8

States and Municipalities

7 6 5 4 3 2 1 0 J-98

J-99

J-00

J-01

J-02

J-03

J-04

J-05

J-06

J-07

Source: Brazilian Central Bank.

Figure 5: Debts of Federal Government, States and Municipalities, 1991-2004, per cent of GDP. 40 35

Federal and Central Bank States

30

Municipalities

25 20 15 10 5 0 J-91

J-93

J-95

J-97

Source: Brazilian Central Bank.

J-99

J-01

J-03

J-05

J-07

408

Basilia Aguirre and Guilherme Dias

Federal government started to react in 1999, when federal taxes increased by 7.5 per cent. Between 1996 and 2000 federal taxes increased at an average rate of 4.5 per cent. In the same period, taxes of states and municipalities increased by 3.4 and 2 per cent respectively. The increase in federal taxes resulted from the elevation of tax rates for taxes not shared with states and municipalities as well as the creation of new taxes. The outcome of these efforts was completely appropriated by the federal government without having to be involved in exhaustive negotiations with states and municipalities in Congress. Furthermore, having experienced successive defeats in Congress in fiscal matters, it was easier for federal government to raise taxes that did not require involvement of states and municipalities. Figure 6: Shared and non-shared taxes of Federal Government and States 10% 9% 8% 7% 6% 5% 4% 3%

Federal non-shared Federal shared

2%

State non-shared State shared

1% 0% 1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

Source: Brazilian National Treasury

The process described so far has two main undesirable consequences. First, a high burden was imposed on the productive sector of the economy as the increasing non-shared taxes were all cascading taxes which impose heavy costs on the Brazilian economy. Besides, the total tax burden increased without evidence for the public that provision of public services had improved. Second, intergovernmental relations showed no coordination any more, but rather became a tug of war. Each time one actor reacts he causes damage to a third party. The general public cannot recognize any relationship between an improvement in public service provision

Fiscal Reform and Federal Relations

409

and the continuous increase in tax burden. The reaction chain continues indefinitely. About the first undesirable consequence much has been written by Brazilian economists. About the second there is a widespread unawareness. In our view, the second consequence is the principal determinant of the costs of a larger tax burden. V. Conclusions The intergovernmental relations in Brazil evolved in a way not favorable for cooperation. As states and municipalities accumulated victories over the federal government in Congress they developed the belief that it was not necessary to negotiate and that they could always solve their financial problems by imposing further losses to federal government. The federal government started to react in 1991 by increasing and creating taxes. Both fighting parties used their instruments as power mechanisms to reach their goals. The problem is that this is a zero or negative sum game for the public sector and the economy. Either one or more government levels lose or the private sector pays the bill through an increasing tax burden. It is clear that this process involves two distributional conflicts. The first conflict is between different levels of government and the other between public and private sector. In this situation a deadlock in intergovernmental relations appears. The discussion only partly referred to responsibilities. There is no mandatory provision that changes in revenues should be accompanied by changes in responsibilities. Lower level governments as well as the federal government, each with their own mechanisms, changed tax structures without having to account for the use of the additional resources. Brazilian federalism presents the main characteristics of cooperative federalism, as defined earlier. It has tax-sharing, a definition of tax base per level of government, definitions of competencies, compulsory transfers and compensatory transfers. However it also bears some features of competitive federalism such as exclusive taxation for certain taxes, competition for tax base between federal and state governments and also among state governments, and voluntary transfers. This structure gives opportunities for predatory competition between levels of government with the consequences pointed in the literature. The solution of this problem has to emerge from negotiation in Congress on how to diminish potential conflicts in intergovernmental relations. A lower conflict potential can only be reached through allocation of property rights between federation members in order

410

Basilia Aguirre and Guilherme Dias

to prevent anyone to impose losses on other(s). One possible way could be to prohibit the federal government to increase tax rates without involvement of the Congress and not permitting the reallocation of taxes without the consent of the federal government. Such changes would induce federal units to enter into negotiation whenever changes are desired by one of them. Any intensive period of negotiations between federal members is a process full of uncertainty. To cope with this problem two points are essential: first it is unreasonable to expect that the process can be finished in a short span of time, it is wise to define a schedule long enough to guarantee full awareness of the consequences of each step; second there should always be the possibility of revise decisions and a compensatory framework to avoid heavy losses to any party involved in order to ensure participation.

References

Acemoglu, Daron, Johnson, Simon, Robinson, James (2004), Institutions as the Fundamental Cause of Long-Run Growth, NBER Working Paper no. 10481, Cambridge, MA,  http://www.nber.org/papers/w10481 (last visit: 07.12.2007). Afonso, Roberto Rodrigues, Araújo, Erika, Rezende, Fernando, Varsano, Ricardo (2000), A Tributação Brasileira e o Novo Ambiente Econômico, Revista do BNDES, June 13, Rio de Janeiro. Afonso, Roberto Rodrigues, De Mello, Luiz (2000), Brazil: An Evolving Federation, IMF/FAD Seminar on Decentralization, Washington, DC, November 20-21. Aguirre, Basilia M. B. (1999), Federalism and Institutions: A Comparative Perspective, Revista de Economia Aplicada 13, São Paulo. Besley, Tim, Coate, Stephen (1999), Centralized versus Decentralized Provision of Local Public Goods: A Political Economy Analysis, NBER Working Paper no. 7084, Cambridge, MA. Coase, Ronald (1937), The Nature of the Firm, Economica 4 (16), 386-405.

Fiscal Reform and Federal Relations

411

Coase, Ronald (1960), The Problem of Social Cost, Journal of Law and Economics 4(1), 1-44. Dixit, Avinash K. (1996), The Making of Economic Policy: A Transaction-Cost Politics Perspective, Cambridge, MIT Press. Inman, Robert P., Rubinfeld, Daniel (1997), Rethinking Federalism, Journal of Economic Perspectives 11(4), 43-64. Inman, Robert P., Fitts, Michael A. (1990), Political Institutions and Fiscal Policy: Evidence from the U.S. Historical Record, Journal of Law, Economics and Organization 6 (special issue), 79-132. North, Douglass C. (2005), Understanding the process of economic change, Princeton, Princeton University Press. North, Douglass C. (1990), Institutions, Institutional Change and Economic Performance, Cambridge, Press Syndicate of the University of Cambridge. Oates, Wallace (1972), Fiscal Federalism, New York, Harcourt Brace Jovanovich. Oates, Wallace (1994), Federalism and government finance, in: Oates W. (ed.), The Economics of Fiscal Federalism and Local Finance, Northampton, Massachusetts, Edward Elgar. Olson, Mancur (1971), The Logic of Collective Action, Cambridge, Harvard University Press, chap. 1 and 2, pp. 1-65. Rezende, Fernando, Afonso, José R. (2000), A federação brasileira: fatos, desafios e perspectivas, Rio de Janeiro, IPEA, : http://info.worldbank.org/etools/docs/library/229990/ Rezende%20e%20Afonso.pdf (last visit: 07.12.2007). Tiebout, Charles (1956), A Pure Theory of Local Expenditures, Journal of Political Economy 64(5), 416-424. Varsano, Ricardo (1995), A Tributação do Comércio Interestadual: ICMS Atual Versus ICMS Partilhado, IPEA Discussion paper No. 382, Brasília. Varsano, Ricardo (1999), Subnational Taxation and Treatment of Interstate Trade in Brazil: Problems and a Proposed Solution, Paper presented to the ABCD-LAC Conference, Valdivia, Chile, July 1999.

412

Basilia Aguirre and Guilherme Dias

Varsano, Ricardo (2002), Sistema tributário para o desenvolvimento, in: Ana Célia Castro (ed.), Desenvolvimento em debate: Painéis do desenvovimento, Rio de Janeiro, Mauad: BNDES. Werneck, Rogério L. F. (2000), A Nova Proposta de Reforma Tributária do Governo: Limites do Possível e Incertezas Envolvidas, Revista de Economia Política 20 (1), 77. Werneck, Rogério L. F. (2002), Reforma tributária: urgência, desafios e descaminhos, in: Ana Célia Castro (ed.), Desenvolvimento em debate: Painéis do desenvolvimento, Rio de Janeiro, Mauad: BNDES. Werneck, Rogério L. F. (2003), An evaluation of the 2003 tax reform effort in Brazil, Working paper No. 488, Rio de Janeiro, Pontifícia Universidade Católica.

List of Authors

Prof. Dr. Basilia Aguirre, Department of Economics, University of São Paulo, Brazil. Prof. Dr. José Luis Arrufat, Department of Economics and Institute of Economics and Finance, National University of Córdoba, Argentina. Prof. Sergio V. Barone (Lic.), Department of Economics, Faculty of Economic Sciences, National University of Córdoba, Argentina. Valeria J. Blanco (Lic.), Department of Economics, Faculty of Economic Sciences, National University of Córdoba, Argentina. Michael Brei (Dipl.VW), Bonn Graduate School of Economics, Department of Economics, Germany. Prof. Dr. Alberto M. Díaz Cafferata, Department of Economics and Institute of Economics and Finance, Faculty of Economic Sciences, National University of Córdoba, Argentina. Prof. Dr. Rinaldo Antonio Colomé, Department of Economics and Institute of Economics and Finance, Faculty of Economic Sciences, National University of Córdoba, Argentina. Dr. Felicitas Nowak Lehmann Danzinger, Ibero-America Institute for Economic Research, Georg-August University Göttingen, Germany. Prof. Guilherme Dias, Department of Economics, University of São Paulo, Brazil. Prof. Dr. Alberto J. Figueras, Department of Economics and Institute of Economics and Finance, National University of Córdoba, Argentina. Prof. Marcelo Florensa (Msc), Department of Economics and Institute of Economics and Finance, Faculty of Economic Sciences, National University of Córdoba, Argentina.

414

List of Authors

Jorge Alberto Fornero (Lic.), Research Institute for European Affairs, Vienna University of Economics and Business Administration. Fernando M. Giuliano (Lic.), Department of Economics, Faculty of Economic Sciences, National University of Córdoba, Argentina. Matteo Grazzi, Department of Economics, Bocconi University, Milano. Dr. Dierk Herzer, Department of Economics, Johann-Wolfgang Goethe University, Frankfurt. Dr. Wolfgang Hetzer, Adviser to the Director General, European Anti-Fraud Office (OLAF), Brussels. M. Dolores De La Mata (Lic.), Department of Economics, Faculty of Economic Sciences, National University of Córdoba, Argentina. Prof. Antonella Mori, Department of Economics, Bocconi University, Milano. Prof. Dr. Ángel Enrique Neder, Department of Economics and Institute of Economics and Finance, Faculty of Economic Sciences, National University of Córdoba, Argentina. Laura Márquez-Ramos, Department of Economics, Universitat Jaume I, Spain. Prof. Dr. María Luisa Recalde, Department of Economics and Institute of Economics and Finance, Faculty of Economic Sciences, National University of Córdoba, Argentina. Prof. em. Dr. Ulrich Peter Ritter, Johann-Wolfgang Goethe University, Frankfurt. Jonatan Saúl (Lic.), Department of Economics, Faculty of Economic Sciences, National University of Córdoba, Argentina. Julieta Schiro (Lic.), Department of Economics, Faculty of Economic Sciences, National University of Córdoba, Argentina. Prof. Dr. Gabriele Tondl, Research Institute for European Affairs, Vienna University of Economics and Business Administration, Austria.

List of Authors

415

Prof. Fernando Zarzosa Valdivia, University of Antwerp, Faculty of Applied Economics and Department of Economics, National University of Córdoba, Argentina. Sebastian Vollmer (Dipl. Math.), Ibero-America Institute for Economic Research, Georg-August University, Germany. Prof. Inmaculada Martínez-Zarzoso, Department of Economics, Applied Economics, Universidad Jaime I, Spain and GeorgAugust University Göttingen, Germany.