Monetary Policy and Macroeconomic Stabilization in Latin America 3540255834, 9783540255833

Citation preview

Monetary Policy and Macroeconomic Stabilization in Latin America

Rolf J. Langhammer Lucio Vinhas de Souza (Eds.)

Monetary Policy and Macroeconomic Stabilization in Latin America

Springer

Rolf J. Langhammer Institut fiir Weltwirtschaft Kiel D-24100Kiel langhammer @ ifw-kiel.de

Lucio Vinhas de Souza European Commission Avenue de Beaulieu, 1 B-1160, Brussels [email protected]

I S B N 3-540-25583-4 Springer-Verlag Berlin Heidelberg N e w York Cataloging-in-Publication Data: Library of Congress Control Number: 2005923823 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer-Verlag. Violations are liable for prosecution under the German Copyright Law. Springer-Verlag is a part of Springer Science+Business Media springeronline.com © Springer-Verlag Berlin • Heidelberg 2005 Printed in Germany The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Hardcover-Design: Erich Kirchner, Heidelberg SPIN 11416425

42/3153-5 4 3 2 1 0 - Printed on acid-free paper

Contents

Preface

ROLFLANGHAMMER

VII

LUCIO VINHAS DE SOUZA

Reducing Inflation through Inflation Targeting: The Mexican Experience Comment

How Has NAFTA Affected the Mexican Economy? Review and Evidence

MANUEL RAMOS-FR ANCIA

1

and ALBERTO TORRES PAULO SPRINGER DE FREITAS

30

M. AYHAN KOSE, GUY M.MEREDITH,

35

and CHRISTOPHER M . TOWE

Comment

LUCIO ViNHAS DE SOUZA

82

Argentina: Monetary Policy by Default

GEORGE T. MCCANDLESS

87

Comment

ALICIA GARCIA HERRERO

113

Do Exchange Rates Matter in Inflation Targeting Regimes? Evidence from a VAR Analysis for Poland and Chile

FELIX H AMMERM ANN

115

Comment

MARCEL FRATZSCHER

149

JORGE BRAGA DE MACEDO

153

Argentina and Brazil Risk: A "Eurocentric" Tale Comment

and MARTIN GRANDES GEORGE T. MCCANDLESS

173

VI

CONTENTS

Macroeconomic Shocks, Inflation, and Latin America's Labor Market Comment

Monetary Policy Rules in Emerging Market Economies: Issues and Evidence Comment

List of Contributors

ANA MARIA LOBOGUERRERO and UGO PANIZZA

177

ANSGARBELKE

198

M.S. MOHANTY

205

AND MARC KLAU RAINER SCHWEICKERT

246

253

Preface Latin America is a very important region of the globe, which has been buffeted by successive waves of economic instability within the last decades. These waves have caused several episodes of hyperinflation or near hyperinflation, and several currency and financial crises, which, in certain moments, have even spilled over and affected other emerging markets. This has resulted in huge costs in terms of lost potential growth, and, as is inevitable, the markets most affected by this have been the least capable of defending themselves. In a region plagued by still considerable rates of social exclusion, with some of the highest rates of income concentration in the whole globe, the human costs of these crises have been very substantial. Starting in the early 1990s, the slow implementation of reforms, plus the resumption of more sustained growth—to a substantial degree linked to the increase in commodity prices, especially since the early 2000s—seems to have resulted in a more stable situation. Initially, in early reformers like Chile, later in the larger economies of the region, like Brazil and Mexico, a consensus— embraced by both sides of the political spectrum—towards integration in global markets, both in their trade and financial components, floating exchange rates, independent monetary authorities, and sustainable fiscal policies has emerged. This process has suffered some temporary setbacks, for instance, the Argentinean crisis of 2001-2002, but the overall trend has been strongly positive. These developments have also been paralleled by the spread of democracy in the region, as the interrelationship between human freedom and economic welfare has been increasingly recognized. Nevertheless, probably the most important questions that still need to be addressed refer to the future of the reform process: What institutions and what policies can make the gains achieved sustainable in the long run, and can increase and spread more evenly those gains, within a democratic framework? This concern is clearly expressed in the papers collected in this conference volume. The first chapter, by Manuel Ramos-Francia and Alberto Torres, from the Central Bank of Mexico, reviews the role of monetary policy in the disinflation process that has taken place in the Mexican economy over the last years. The main purpose of the chapter is to show that, given a sound fiscal position, an inflation targeting framework could be seen as an efficient mechanism to impose discipline on monetary policy and thus effectively reduce inflation, even in a commodity-exporting economy like Mexico. This conclusion is supported empirically both by the results of a vector autoregressive analysis and by the estimation of monetary policy rules.

VIII

PREFACE

The second chapter, by Ayhan Kose, Guy Meredith and Christopher Towe, from the International Monetary Fund, provides a comprehensive assessment of the impact of NAFTA on growth and business cycles in Mexico. They conclude that the agreement spurred an unprecedented increase in trade andfinancialflows between Mexico and its NAFTA partners, and also affected Mexican economic growth and business cycle dynamics. In addition, this chapter provides new evidence—using estimates from a dynamic latent factor model—^that the free trade arrangement has been associated with a sharp increase in the importance of North American regional factors in driving business cycles in Mexico during the past decade. Simulations of a dynamic stochastic general equilibrium model are then used to illustrate that this result is consistent with the reduction in trade frictions that would be expected from NAFTA. The chapter concludes by drawing lessons from Mexico's NAFTA experience for policymakers in developing countries in the Western Hemisphere and elsewhere. The foremost of these is that in an increasingly global trading system, bilateral and regional free trade arrangements should be used to accelerate, rather than postpone, needed structural reform. In the third chapter, George T. McCandless, from the Central Bank of Argentina, provides a brief review of the developments that led to the momentous Argentinian crisis of 2001-2002 and of the limited channels of monetary policy —essentially the exchange rate, the expectations and the money channels—that where left to the Argentinian monetary authority in its aftermath. In the forth chapter, Felix Hammermann, from the Kiel Institute for World Economics, studies alternative monetary policy strategies in flexible exchange rate regimes for Chile and Poland, which are used as a starting point for an empirical investigation of the role of the exchange rate in the inflation targeting regimes in those two countries. Vector autoregressive models for Poland and Chile reveal that the Polish monetary policy of the 1990s shows a clear break when the exchange rate as the nominal anchor is replaced by inflation targeting. Still, his results show that exchange rate targeting is not abandoned completely in Poland, while in Chile inflation targeting is in place for the entire sample period and exchange rate policy is used only in times of internationalfinancialturmoil. The fifth chapter, by Jorge Braga de Macedo, of the Faculty of Economics, Nova University and Tropical Research Institute, Lisbon, and Martin Grandes, of the American University of Paris, examines to which extent the European experience with earning rather than importing policy credibility, using a framework like the Exchange Rate Mechanism (ERM), is relevant outside of the EU or OECD membership, suggesting that it would apply especially to Argentina and Brazil, if these two countries wished to promote regional integration in the context of a regional organization like the Mercosur. In the sixth chapter. Ana Maria Loboguerrero and Ugo Panizza, from the Inter-American Development Bank, look at how macroeconomic volatility is

PREFACE

IX

transmitted to the labor market. They estimate employment, unemployment, and wage Okun coefficients and use them to show that, compared with industrial countries, Latin American countries adjust more through wages than through employment. The chapter also shows that inflation plays some role in explaining the difference between employment elasticities in Latin America and in industrial countries, but that there is a difference between these two regions that cannot be explained away by differences in inflation. When focusing on Latin America, the chapter finds that within Latin America, inflation increases labor market flexibility in countries that have highly regulated labor markets and that enforce regulations. In the final chapter of our book, M. S. Mohanty and Marc Klau, from the Bank for International Settlements, review the recent conduct of monetary policy and the central banks' interest rate setting behavior in selected emerging market economies. Using a standard open economy monetary policy reaction function, they test whether central banks in emerging economies react to changes in inflation, output gap and the exchange rate in a consistent and predictable manner. They find that in most emerging economies the interest rate responds strongly to the exchange rate, and that for some economies, the response is higher than that to changes in the inflation rate or the output gap. This result is robust to alternative specification and estimation methods. They conclude that this outcome highlights the importance of the exchange rate as a source of shock and supports the "fear of floating" hypothesis. They also find some evidence that in some countries the central bank's response to inflation shocks is asymmetric, i.e., the reaction to a negative inflation shock might be weaker than to a positive one. The papers were presented at the conference on "Monetary Policy and Macroeconomic Stabilization in Latin America," held at the Kiel Institute for World Economics (IfW), in Kiel, Germany, September 11-12, 2003. This conference benefited not only from the uniformly high quality of its speakers, but also from their representative cross-sectional origin, as it included speakers from the monetary authorities of all the major Latin American economies and from almost all the major multilateral policy institutions that habitually deal with Latin America. Finally, we would like to express our thanks and gratitude to the generous cosponsorship of the Deutsche Bundesbank, the European Central Bank, and the European Commission. It goes without saying that editing this conference volume in a professional manner would have been impossible without both gentle pressure and strong support from our colleagues from the editorial group, namely Paul Kramer, Kerstin Stark, and Korinna Werner-Schwarz. Kiel, April 2005

Rolf Langhammer Lucio Vinhas de Souza

Manuel Ramos-Francia and Alberto Torres

Reducing Inflation through Inflation Targeting: The Mexican Experience Abstract This paper reviews the role of monetary policy in the disinflation process that has taken place in the Mexican economy in recent years. The purpose is to show that, once an economy establishes a sustainable fiscal position, an inflation targeting framework can be seen as an efficient mechanism to impose discipline on monetary policy and, thus, to reduce inflation. This paper describes the measures that were taken after the 1995 crisis to stabilize the economy and that prevented the possibility of a fiscal dominance situation from arising. Consequently, the role of monetary policy in reducing inflation is analyzed, in particular its response to different inflationary shocks. Results show that in conducting the successful disinflationary process. Banco de Mexico's responses to inflationary shocks have been consistent with inflation targeting principles.

1

Introduction

This paper reviews the role of monetary policy in the disinflation process that has taken place in the Mexican economy in recent years. The purpose is to show that once an economy establishes a sustainable fiscal position, an inflation targeting framework can be seen as an efficient mechanism to impose discipline on monetary policy and, thus, to reduce inflation in a sustainable way. This paper describes the measures that were taken after the 1995 crisis to stabilize the economy and that prevented a fiscal dominance situation from arising. Then, the analysis evaluates whether the response of monetary policy to different inflationary shocks was consistent with inflation targeting principles. A brief Remark: The authors thank Daniel Chiquiar, Alejandro Diaz de Leon, Rodrigo Garcia, Daniel Samano, Julio Santaella, Paulo Springer de Freitas, Juan Pedro Trevino, participants at the Kiel Institute for World Economics' Workshop on Monetary Policy and Macroeconomic Stabilization in Latin America, and an anonymous referee for their comments. Julieta Alemdn, Lorenza de Icaza, Santiago G6mez and Jessica Rold^ provided excellent research assistance. Opinions expressed in the article are solely those of the authors and do not necessarily represent those of Banco de M6xico.

2

MANUEL RAMOS-FRANCIA AND ALBERTO TORRES

description of the episodes of monetary tightening precedes an analysis that uses a structural VAR approach to characterize the associated inflationary shocks as either supply or demand driven. Afterwards, monetary policy rules are used to formally test the consistency between monetary policy and the inflation targeting principles. In recent years, several economies have moved toward flexible exchange rate regimes. This trend has been accompanied by a set of guidelines or recommendations to monetary authorities known as inflation targeting principles. For monetary policy to effectively perform the role of nominal anchor of an economy, inflation targeting stresses the importance of two elements: the appropriate response of monetary policy to inflationary shocks, and transparency in its implementation. This paper shows that in conducting the successful disinflationary process that took place once a sounder fiscal position had been ensured after the 1995 crisis. Banco de Mexico's responses to different inflationary shocks were consistent with inflation targeting principles. The Mexican experience is interesting because it represents an emerging market economy that was able to reduce inflation from close to 52 percent in 1995, to just under 4 percent by 2003 under a flexible exchange rate regime. Furthermore, since the evidence presented shows that monetary policy has been conducted in accordance with inflation targeting principles, Mexico's experience suggests that, given a situation of no fiscal dominance, inflation targeting frameworks in emerging market economies can be useful to impose discipline on monetary policy and, thus, to ensure that it performs the role of nominal anchor of the economy. The rest of the paper is organized as follows. Section 2 discusses the measures adopted after the 1995 crisis in order to stabilize the economy and to attain a fiscal position that would contribute to restoring credibility in the financial system and in monetary policy. Section 3 describes the evolution of monetary policy toward an inflation targeting framework and identifies three episodes of monetary policy tightening during the disinflation process. Section 4 describes and characterizes the inflationary shocks that took place during the same period. Then, in Section 5, results from Sections 3 and 4 are used to discuss the consistency of monetary policy decisions with inflation targeting principles. Finally, Section 6 concludes.

2

Stabilization of the Economy

By the end of 1994, Mexico was forced to float the currency and abandon a target zone for the exchange rate. Since the Mexican economy experienced an

REDUCING INFLATION THROUGH INFLATION TARGETING

3

important financial crisis throughout 1995, the situation was precarious. Under the new flexible exchange rate regime, Banco de Mexico faced the challenge of providing the economy with a nominal anchor. As expected during a crisis, the credibility of Mexican financial and monetary institutions was not at its best. Not only did the Central Bank's policy have to become the nominal anchor of the economy, but it had to do so at a time when there was widespread uncertainty about the Bank's commitment and ability to achieve financial and price stability. There are various papers that analyze the origins and aftermath of the crisis (Carstens and Werner 1999; Gil-Diaz 1998; Calvo and Mendoza 1996). Briefly, the build-up to the crisis was related to the following issues: (1) a rigid exchange rate regime leading to large short-term speculative capital flows and considerable relative price distortions; (2) weak banking regulation and supervision, which led to an inefficient intermediation of the large capital inflows in the years preceding the crisis; (3) an overspending economy with a large current account deficit; (4) as a result of worsening macroeconomic conditions, the government was forced to gradually concentrate its debt on short-term dollar-denominated instruments (i.e., borrowing to defend). All of the above issues interacted with each other, leading to gradually inreasing disequilibria in the economy and, eventually, to a run against the currency. As the crisis unfolded, the authorities faced three central challenges for economic policy: to meet all maturing government debt obligations; to conduct an orderly macroeconomic adjustment to the reversal of capital inflows; and, to maintain the solvency and viability of the banking system. As a result, a comprehensive macroeconomic stabilization program was implemented. First, given the magnitude of the problem that meant meeting the government's short-term dollar-denominated obligations, its solution clearly laid outside the scope of any realistic macroeconomic policy adjustment. Thus, the government negotiated a financial support package of over 50 billion dollars with international financial institutions. Second, to induce an orderly macroeconomic adjustment, consistent with the drying up of capital inflows, the challenge was to induce an orderly contraction of domestic aggregate demand, for which a large real exchange rate depreciation and expenditure reducing policies were required. Once the exchange rate depreciated, fiscal and monetary policies focused on containing the inflationary pressures induced by the initial devaluation. Fiscal policy in particular was oriented towards increasing public savings through a rise in the value added tax rate, increases in public prices, and a contraction of public expenditures. Simultaneously, monetary policy was geared towards restoring orderly financial markets and to reduce exchange rate volatility. For that purpose, the Central Bank established a limit to the expansion of net domestic credit, thus signaling that it would not "validate" further runs on the currency, and the resulting re-

4

MANUEL RAMOS-FRANCIA AND ALBERTO TORRES

Figure 1: Annual CPI Inflation, 1994:12-2003:12

strictive monetary policy stance induced considerable increases in nominal and real interest rates. Despite these policies, given the magnitude of the exchange rate depreciation (approximately 120 percent from December 1994 to December 1995), annual CPI inflation rose from 7 percent in December 1994 to 51.9 percent by December 1995 (Figure 1). Finally, the accumulated fragility of banks and the previous indebtedness of firms and households, combined with the damaging effects of the crisis, seriously threatened the viability of the financial system. Thus, in order to avoid a systemic run on the banks, three measures were put in place. First, to stop the run on the external liabilities of commercial banks, Banco de Mexico established a dollar liquidity facility. Second, the exchange rate depreciation drastically augmented the peso value of the dollar-denominated loan portfolio, causing the capital-asset ratios of many Mexican banks to fall. To address this situation, undercapitalized banks were required to issue subordinated debt, convertible into equity, which was purchased by the government. Third, the loan portfolios of banks were deteriorating significantly as a result of an overexpansion of credit in the previous years and because of the large increase in interest rates in the aftermath of the crisis. To cope with this, the government offered debtors discounts on interest rate payments on performing loans and also offered to acquire a fraction of banks' loan portfolios at book value if banks brought in additional resources to increase their capital.

REDUCING INFLATION THROUGH INFLATION TARGETING

5

It is important to mention that right from the beginning of the crisis, a continuous effort was made by the authorities to asses and quantify the costs associated with the financial and debtor support programs, and to explain to markets that the public finances would be able to absorb these costs. The objective was to show that a fiscal dominance situation would not arise. This, as it turns out, would give monetary policy in the following years the opportunity to concentrate its efforts on reducing inflation. As explained above, at first monetary policy actions were directed towards restoring orderly financial markets. Later on, the Central Bank started to adopt more direct measures intended to curtail inflationary pressures in the economy. In 1996, in addition to the limit to the expansion of net domestic credit. Banco de Mexico established a nonnegative target for net international reserves accumulation. These two measures not only reassured its commitment not to "validate" a run against the currency, but would also allow the remonetization of the economy, albeit through its external accounts. Similarly, in 1996 the Central Bank started to publish, at the beginning of every year, a projection of monetary base mainly as a reference to guide inflation expectations, although it did not adopt a formal intermediate target on narrow monetary aggregates. Another element of the strategy to restore credibility in monetary policy consisted of a set of pre-announced rules through which Banco de Mexico intervened in the foreign exchange market. After the crisis, one of the challenges was to restore the stock of international reserves to build up credibility on the currency so as to strengthen the viability of the flexible exchange rate regime. The main source of international reserve accumulation came from the exports of the state oil company Petroleos Mexicanos (PEMEX). PEMEX's dollars are sold directly to Banco de Mexico at the market exchange rate. This mechanism has two benefits. First, it allows the Central Bank to accumulate international reserves without discretionary interventions. Second, it isolates the foreign exchange market from variations in the international price of oil. Later on, in 1996 the Foreign Exchange Rate Commission introduced an instrument explicitly designed to foster international reserves accumulation. ^ It consisted of selling put options once a month to sell dollars to the Central Bank on any working day at the "reference interbank exchange rate" (called "FIX" rate), determined on the previous working day, provided such rate did not exceed the average of the FIX rate in the 20 days prior to the exercise date.^ The instrument was designed so that market participants would have the incentive to exercise the options when

The Commission consists of three members of the Ministry of Finance and three of Banco de Mexico, with the Ministry of Finance casting the deciding vote in case of a tie. See Banco de Mexico (1996) for a detailed discussion of the instrument.

6

MANUEL RAMOS-FRANCIA AND ALBERTO TORRES

the exchange rate appreciated; thus. Banco de Mexico accumulated reserves without interfering with the floating exchange rate regime. Afterwards, in 1997 a third instrument was introduced to reduce exchange rate volatility in times when the foreign exchange market experienced low levels of liquidity. This consisted of an automatic mechanism through which the Central Bank would auction 200 million dollars in the event that the exchange rate depreciated on any given day more than 2 percent with respect to the FIX exchange rate of the previous day.^ This instrument added symmetry with respect to the direction in which the Central Bank could intervene in the foreign exchange market. The strategy to accumulate international reserves, which relied on non-discretionary interventions since all three mechanisms followed previously announced rules, proved to be successful in restoring a considerable stock of international reserves in light of the expected remonetization process that the economy underwent as a result of its successful stabilization.^ Furthermore, to guarantee that the accumulation of international reserves would not cause an overexpansion of the monetary base. Banco de Mexico sterilized any impact on the monetary base beyond money demand growth. One of the goals of the economic program was to stabilize the economy as orderly and as quickly as possible, so as to guarantee that a fiscal dominance situation would not arise. In doing so, the three challenges posed by the crisis were met. First, the government fulfilled all its obligations. Second, the economy adjusted swiftly to a new macroeconomic environment, whereby the current account deficit fell from 7.1 percent of GDP in 1994 to 0.61 percent in 1996 and 1.8 percent in 1997. Third, a breakdown of the financial system was avoided. Henceforth, as the policies and programs began to yield results, economic activity resumed its growth path. After a contraction of 6.1 percent in 1995, GDP resumed growth at 5.1 percent and 6.7 percent in 1996 and 1997, respectively. A key element to this development was the declining pattern of inflation, from 51.9 percent in 1995 to 27.7 percent in 1996 and to 15.7 percent in 1997 (Figure 1), which resulted in a downward trend in nominal and real interest rates; clearly, this contributed to gradually easing the private debt overhang problem. Furthermore, as the cost of the financial and debtor support programs was incorporated into the long-term fiscal policy agenda, the possibility of fiscal dominance The behavior of the exchange rate in late 1995 revealed that depreciations of around 2 percent and more occurred when conditions in the foreign exchange market tumed highly illiquid. Thus, the purpose of the instmment was to prevent this type of situation. See Banco de Mexico (1997) for a detailed discussion of the instmment. In 2001 the Foreign Exchange Rate Commission decided to abandon the use of the dollar put option and automatic dollar auction mechanisms. Then, in 2003, the Commission defined a set of mles to reduce the rate of accumulation of intemational reservesftomPEMEX.

REDUCING INFLATION THROUGH INFLATION TARGETING

7

eventually stopped representing a threat for macroeconomic stability. All these elements were fundamental in allowing monetary policy to later on focus on actively pursuing long-run price stability.

3

Monetary Policy

As the economy stabilized, Banco de Mexico concentrated its efforts on reducing inflation. It can be said that the strategy consisted mainly of three elements: to improve transparency in the implementation of monetary policy, to maintain a clear restrictive bias in order to induce a sustainable reduction in inflation and to respond appropriately to inflationary shocks.

3.1

Transition towards a Fully Fledged Inflation Targeting Framework

Since 1995 Banco de Mexico has defined as its main instrument to affect interest rates a target for the cumulative balance of commercial bank's current accounts at the Central Bank ("corto")-^ With this instrument the Central Bank is able to affect interest rates by inducing (through open market operations) an overdraft in these accounts (of one or several banks), since banks have to pay a penalty on the amount of the overdraft; this operational procedure is similar to the non-borrowed reserves target strategy followed by the U.S. Federal Reserve in the early 1980s.^ Although this is an instrument referred to quantities, given the relatively small magnitude of the target for the said cumulative balances, it should be interpreted more as a signaling device in terms of the direction in which the Central Bank wants interest rates to move. As it turned out, three characteristics of this instrument proved to be useful through the disinflation process. First, starting in 1998, the Bank was able to signal its continuous bias toward a restrictive monetary policy stance to reduce inflation by setting a negative target on commercial banks' current accounts (i.e., overdraft). Second, to cope with inflationary pressures related to higher frequency shocks (among these, those related to the cyclical phase of the economy), the Bank was able to follow a "lean against

^ ^

See Banco de M6xico (1996, 2000) for a description. When the Central Bank induces an overdraft in the system, to avoid paying the penalty (i.e., higher interest rates) each bank tries to avoid being the one that ends up with the overdraft, and therefore has the incentive to raise interest rates on its deposits and/or loans. Since positive balances receive no interests and the penalty for negative balances is defined as two times the market short-term interest rate, the instrument is symmetric and gives the incentive to banks to maintain zero balances.

8

MANUEL RAMOS-FRANCIA AND ALBERTO TORRES

the wind" policy by signaling changes in interest rates through changes in the objective for the balance of commercial banks' current accounts, that meant a larger or smaller target for the overdrafts. Third, since the precise level of interest rates is determined by market participants, this instrument allows interest rates to decrease as fast as risk premiums decrease. By 1998, announcements of changes in the instrument ("corto") began to be accompanied with a discussion of the reasons behind the decision to modify it. This strategy enhanced transparency in the implementation of monetary policy and improved the effectiveness of the instrument to signal the stance of monetary policy. Henceforth, changes in the "corto" began to be associated with changes in the stance of monetary policy, that is, with the direction the Central Bank wanted interest rates to move. Later on, transparency was further enhanced with several announcements or measures adopted by Banco de Mexico. In 1999, the medium-term inflation objective for CPI inflation was first defined as "... the convergence of domestic inflation towards external inflation by 2003" (Banco de Mexico 1999: 137). Then, in 2000 the Central Bank started publishing quarterly inflation reports, containing, among others, detailed discussions of the sources of inflationary pressures. In that same year. Banco de Mexico introduced the concept of core inflation as an important element in the analysis of inflation, particularly in the evaluation of inflationary shocks. The process towards transparency was reinforced in 2002, when Banco de Mexico announced that it was formally adopting an inflation targeting framework. Since 1995, the limits to the expansion of net domestic credit had resulted to be nonbinding, given that the yearly targets for the nonnegative accumulation of net international reserves had been exceeded. In effect, every year the accumulation of net international reserves surpassed the expansion of the monetary base, and thus, year after year net domestic credit decreased. Therefore, in 2002 these two elements were excluded from the monetary policy program announced for that year. Also, in 2002 a long-term inflation target was defined at 3 percent for CPI inflation, with an interval of variability of +/-1 percentage point established mainly to reflect the transitory impact on inflation of relative price adjustments. Finally, starting in 2003, monetary policy announcements have been made at preestablished dates. As a result of the transition towards a fully fledged inflation targeting framework, the objective and implementation of monetary policy has become more transparent and open to public scrutiny. Similarly, this process has fostered the accountability of Banco de Mexico and has helped to anchor inflation expectations.

REDUCING INFLATION THROUGH INFLATION TARGETING

3.2

9

Episodes of Monetary Policy Tightening

While transparency in the implementation of monetary policy is helpful to restore credibility and to anchor inflation expectations, it is certainly not enough to reduce and stabilize inflation at low levels. Clearly, an appropriate response of monetary policy to inflationary shocks is needed to consolidate the role of monetary policy as nominal anchor of the economy. Since 1995 Banco de Mexico's main instrument to affect interest rates has been the "corto." As explained in the previous section, this type of instrument implies that short-term interest rates are determined by the market and not directly by the Central Bank. This could suggest that in order to identify changes in the monetary policy stance, it would be enough to look for the dates when the Central Bank has announced a change in the "corto." However, since the instrument is intended only as a signal of the direction the Central Bank wants interest rates to move, these do not always react immediately after the announcement. Thus, to account for changes in the monetary policy stance, short-term interest rates clearly also have to be considered. Figure 2 shows the short-term nominal interest rate (daily interbank interest rate) and the target for the cumulative balance of commercial banks' current accounts at the Central Bank ("corto") for the 1995-2003 period. It is important to note that during 1995 and 1996, Banco de Mexico modified the instrument several times within very short intervals on different occasions (for example, eight times during November 1995). Since some of these changes were in one direction (restriction), and others in the opposite (relaxation), it is difficult to argue that the Bank was trying to signal changes in the stance of monetary policy. As discussed in the previous section, after the crisis it can be argued that the main concern of monetary policy was to restore order in financial markets. In this context, changes in the "corto" during 1995 and 1996 should be viewed mainly as actions intended to influence interest rates for a very short-time horizon (days or weeks), so as to reestablish orderly conditions in financial markets. By 1998, the main challenges posed by the crisis had been mostly dealt with; i.e., throughout 1995, 1996, and 1997 all government debt obligations had been met, an orderly macroeconomic adjustment took place, and financial and debtor support programs contributed to maintaining the solvency and viability of the banking system. Therefore, the possibility of a fiscal dominance situation arising had basically disappeared. This allowed Banco de Mexico to start using monetary policy in a more pro-active and forward-looking way to combat inflation, which at the time had already declined from 51.9 percent in 1995, to 15.7 percent in 1997. Figure 2 shows that in 1998, the Bank started to signal its permanent bias toward a restrictive monetary policy stance through a permanent negative

10

MANUEL RAMOS-FRANCIA AND ALBERTO TORRES

Figure 2: Nominal Interest Rate and Changes in the Instrument of Monetary Policy, 1995:4-2003:12

•800 inioioa)0)0>0)0)0>0>0>0>0)0>0>OOOOOOOQ;?OOOQ^ < < Q < < Q < < Q < < Q < < Q < < Q < < Q < < Q < < Q

(overdraft) target on the cumulative balance of conmiercial banks' current accounts. Likewise, as explained in the previous section, in 1998 changes in the instrument began to be accompanied with an announcement of the reasons behind the decision to modify the stance of monetary policy. In what follows, the discussion concentrates on the 1998-2003 period. Following the type of approach in Friedman and Schwartz (1963) and Romer and Romer (1989), we combine statistical information (interest rates, inflation, exchange rate, "corto," etc.) among other things, with anecdotal evidence included in official statements that accompanied the announcements of monetary policy in order to identify episodes in which monetary authorities restricted the stance of monetary policy. Three episodes of monetary policy tightening are thus identified. Figures 3.1 and 3.2 show nominal (weekly average of daily interbank interest rates) and ex ante real interest rates (computed using the weekly average of the 28-day treasury bill interest rate and the twelve months ahead inflation expectations reported in the weekly Infosel survey), with vertical lines representing changes in the "corto."^ The three shaded areas correspond to the episodes in which, it is Infosel is a private firm that reports financial information online and also conducts a weekly survey on inflation expectations among market participants.

REDUCING INFLATION THROUGH INFLATION TARGETING

11

Figure 3.1: Nominal Interest Rate and Changes in the Instrument of Monetary Policy, 1998:1-2003:12

Note: Solid lines represent the dates when the instrument changed toward restriction and dotted lines represent changes toward relaxation. Shaded areas correspond to episodes of monetary policy restriction.

Figure 3.2: Ex-Ante Real Interest Rate and Changes in the Instrument of Monetary Policy, 1998:1-2003:12

Note: Solid lines represent the dates when the instrument changed toward restriction and dotted lines represent changes toward relaxation. Shaded areas correspond to episodes of monetary policy restriction.

12

MANUEL RAMOS-FRANCIA AND ALBERTO TORRES

argued, monetary policy was tightened. All three episodes have some characteristics in common. First, within each episode, the instrument of monetary policy ("corto") was increased on several occasions, signaling a more restrictive stance.^ Second, both nominal and real interest rates present an upward trend during these three episodes. Third, monetary policy announcements on the dates the instrument was modified describe specific shocks that Banco de Mexico considered posed risks to the disinflation process.^ During the first episode (1998:03-1999:01), inflationary pressures arose from an increase in the country's risk perception andfroman exchange rate depreciation due to contagion from financial crises in other emerging market economies (East Asian economies in 1997, Russia in 1998, and Brazil in early 1999). Throughout the second episode (2000:01-2001:01), inflationary pressures arose mainlyfroma rapid expansion of aggregate demand, coming bothfromU.S. demand for Mexican exports and, possibly, from the cyclical increase in expenditures that typically takes place in the months preceding a federal election. Finally, during the third episode (2002:09-2003:03), the driving force behind inflationary pressures was a step hike in public prices, in particular in residential electricity tariffs. These inflationary shocks are discussed in the following section.

4

Inflationary Shocks

One of the main purposes of identifying the sources of inflation is to determine whether inflationary pressures come from the supply side in the form of costpush shocks or whether they are derived from demand shocks. As will be explained in the following section, this distinction is important, since, as is well known, according to the inflation targeting principles the appropriate monetary policy response could be different in each case. A casual inspection of Figures 4.1 and 4.2, in which the inflation expectations gap and the output gap are shown, is helpful to start analyzing the nature of the inflationary shocks during the aforementioned episodes.^^ The first and third °

The increase in the instrument on February 2002 is not considered as part of the third episode of monetary policy restriction because on April 2002 Banco de M6xico announced a change in the "corto" in the opposite direction. ^ See Banco de Mexico (2003). ^^ The inflation expectations gap is defined as the difference between the monthly average of twelve months 5iead expected annual inflation and the twelve months ahead annual inflation target. Inflation expectations are obtained from the Infosel survey, although results are similar if inflation expectations are obtainedfromthe Banco de Mexico survey. The target is computed through a linear interpolation of the end-of-year annual inflation targets announced by Banco de Mexico. The output gap

REDUCING INFLATION THROUGH INFLATION TARGETING

13

Figure 4.1: Inflation Expectations Gap, 1998:1-2003:12

Note: Shaded areas correspond to episodes of monetary policy restriction. Figure 4.2: Output Gap, 1998:1-2003:12

Note: Shaded areas correspond to episodes of monetary policy restriction. is computed using a monthly index of economic activity (IGAE) and a trend defined as the average between linear and Hodrick-Prescottfiltertrends.

14

MANUEL RAMOS-FRANCIA AND ALBERTO TORRES

Table 1: Unit Root Tests, 1998:01-2003:12 Dickey-Fuller t-statistic

Phillips-Perron t-statistic

Critical Value at 95 percent

-1.67

-1.76

-3.47

Pt

-2.48

-1.90

-3.47

^yt

-3.92

-10.24

-2.90

-4.63

-4.63

-3.47

Variable yt

^Pt

Note: For y^, p^, and Apf test includes constant and trend. For Ayf test includes constant and no trend.

episodes are characterized by an upward trend in the inflation expectations gap and a downward trend in the output gap. This suggests that cost-push (supply) shocks are likely to have been the main source of inflationary pressures. On the other hand, during the second episode, both the inflation expectations gap and the output gap presented (for the most part) an upward trend, suggesting that inflationary pressures could have been demand driven. To formally discuss the origin of these inflationary shocks, a structural VAR methodology on monthly data is used. The two variables included in the VAR are output, y^, based on IGAE, and prices, Pf, based on the CPI. Table 1 presents the results of Dickey-Fuller and Phillips-Perron unit root tests, which suggest that both variables are integrated of order one.^^ The long-run properties of output and prices suggest that the VAR representation can be expressed as follows:^^ (1)

B(L)x,=v,,

where B(L) is the matrix on the lag operator, Xf =[yf, Pf] and v^ is a vector 1

2

with the reduced-form innovations of the VAR: [v, ,v^ ] . E(L) is then defined as B(L)-^, and the moving average representation of the VAR is given by (2) x,=E(L)v,, where the variance-covariance matrix of reduced-form innovations is given by ^ ^ Variables have been seasonally adjusted and expressed in logs. ^^ Results of the Johansen cointegration test (not shown) suggest that at 5 percent significance level the null hypothesis of zero cointegrating vectors is rejected and that it is not possible to reject the null hypothesis of one cointegrating vector. Therefore, the VAR is specified in levels.

REDUCING INFLATION THROUGH INFLATION TARGETING

(3)

15

2' = v,v/.

The reduced-form innovations, v^, are defined as linear combinations of the structural shocks, e/ and e^^(v/= [e/,^/^]): (4)

V, = 5 e , ,

where 5 is a two-by-two invertible matrix. Then, C(L) is defined as E(L)S, and the structural moving average representation of the V AR is given by (5)

x,=C(L)8,.

To identify the structural shocks e/ and e^^, the matrix S has to be identified. Since the matrix S consists of four elements, four restrictions are needed. The standard approach in the literature is to assume that the structural shocks are orthogonal: (6)

£,£/=/.

Since the variance-covariance matrix £ is symmetric, three of the four needed restrictions are obtained from the combination of (6), (4), and (3): (7)

Z = SS\

The fourth restriction can be obtained either by directly making an assumption on one of the elements of matrix 5, or by imposing another restriction on the elements of matrix S. A standard assumption in the literature is to assume that element Sn is equal to zero (recursive identification).^^ However, in the present application it is important to let the structural shocks, e^ and e^ , have contemporaneous effects on output and prices (no zeros on matrix 5), since the purpose of the exercise is precisely to identify the immediate response of these two variables to the structural innovations. Therefore, the fourth restriction is obtained from the assumption that, in the long run, the structural shock e^ has no effect on output, that is, the fourth restriction implies that element Cnil) of matrix C(l) is equal to zero (Blanchard and Quah 1989 and Clarida and Gali 1994). It is important to mention that, in general, this identification strategy has been used over long samples to analyze the contribution of supply shocks, in the form of technology innovations, and cyclical demand shocks, to output and prices. In that case, the structural shock with the restriction of having no long-run effects on output usually has the characteristics of a demand shock. The key element

^^ This is the well known Cholesky decomposition.

16

MANUEL RAMOS-FRANCIA AND ALBERTO TORRES

is that long samples allow for the identification of low-frequency technology shocks. Therefore, demand shocks, which usually have a higher frequency than technology shocks, are identified through the structural shock on which the longrun restriction is imposed. For the purpose of the exercise presented in this paper, technology shocks are of no particular interest, since in general they do not cause inflationary pressures. In our analysis, the relevant type of supply shock is of the cost-push form, since they usually have a once-and-for-all effect on prices and a temporary effect on output. This presents an important implication concerning the identification of the structural shocks. In this case, it is possible that demand shocks are not identified with the structural shock on which the long-run restriction is imposed for two reasons. The first reason is that, since the sample used in the exercise is relatively short (1998:01-2003:12), the effects of demand innovations on output could be seen as lasting for relatively long periods, with respect to the size of the sample. Alternatively, it is possible for cost-push shocks to be of a higher frequency than demand shocks. Thus, when only these two types of shocks are considered, the one with longer effects on output could be the demand shock. The VAR is estimated with three lags.^"^ To characterize the inflationary shocks experienced by the Mexican economy over the sample period, the estimated VAR is used to identify the contribution of each type of structural shock to output and prices over the sample period. Figures 5.1 and 5.2 show the combined contribution of both structural shocks, e/ + e^ , to output and prices, respectively. Then, these combined effects are decomposed into the effect that each type of shock had on output and prices. Figures 5.3 and 5.4 show the contribution of the structural shock e/ to output and prices, respectively. Similarly, Figures 5.5 and 5.6 show the contribution of the structural shock e^ to output and prices, respectively. Results of this exercise suggest that the contribution of the structural shock e^ to output and prices goes in the same direction, as if this type of innovation has its origin in the demand. In the case of the structural shock E^, results show that the contribution to output and prices goes in opposite directions, which suggests that this type of innovation comes from the supply side. With respect to the inflationary shock during the first episode of monetary policy tightening. Figures 5.1 to 5.6 suggest that in 1998 inflationary pressures came primarily from the supply side (structural shock E^ ). Financial problems in East Asian economies, Russia, and Brazil caused volatility in the international financial markets. As a result, country risk perception for Mexico increased considerably, while the exchange rate experienced an important depreciation (Figure 6.1). ^^ The Akaike criteria suggest three lags. However, results are similar for specifications with different numbers of lags.

REDUCING INFLATION THROUGH INFLATION TARGETING

Figure 5.1: Combined Contribution of Structural Shocks e] and e} to Output, 1998:12003:12

Note: Shaded areas correspond to episodes of monetary policy tightening. Figure 5.2: Combined Contribution of Structural Shocks e] and e} to Prices, 1998:12003:12

Note: Shaded areas correspond to episodes of monetary policy tightening.

17

18

MANUEL RAMOS-FRANCIA AND ALBERTO TORRES

Figure 5.3: Contribution of Structural Shock ej to Output, 1998:1-2003:12

Note: Shaded areas correspond to episodes of monetary policy tightening. Figure 5.4: Contribution of Structural Shock e} to Prices, 1998:1-2003:12

Note: Shaded areas correspond to episodes of monetary policy tightening.

REDUCING INFLATION THROUGH INFLATION TARGETING

Figure 5.5: Contribution of Structural Shock e? to Output, 1998:1-2003:12

Note: Shaded areas correspond to episodes of monetary policy tightening. Figure 5.6: Contribution of Structural Shock e? to Prices, 1998:1-2003:12

Note: Shaded areas correspond to episodes of monetary policy tightening.

19

20

MANUEL RAMOS-FRANCIA AND ALBERTO TORRES

Figure 6.1: Exchange Rate and Country Risk Perception, 1998:1-2003:12

Note: Shaded areas correspond to episodes of monetary policy tightening. Figure 6.2: CPI, Public Prices, and Fuel Inflation, 1998:1-2003:12

Note: Shaded areas correspond to episodes of monetary policy tightening.

REDUCING INFLATION THROUGH INFLATION TARGETING

21

Of course, this tightening of the economy's external financing constraint increased the cost of capital. Furthermore, during 1998 domestic fuel prices were increased considerably (Figure 6.2).^^ Thus, the evidence points to a strong adverse cost-push shock. In this case, it is interesting to note that despite the restriction on monetary policy, inflation rose from 15.7 percent to 18.6 percent during 1998 (Figure 1). However, since the restriction induced sharp increases in nominal and real interest rates (Figures 3.1 and 3.2) the adverse cost-push shock had no second-round effects of consequences on prices, and inflation resumed its downward trend in 1999. In the case of the second episode of monetary policy tightening, Figures 5.1 to 5.6 show two shocks taking place in late 1999 and 2000. The first one is a favorable cost-push shock, with an expansion on output and a contraction on prices (structural shock Ef ). This result represents the reversal of the adverse costpush shock experienced in 1998. In this case, both country risk perception (Figure 6.1) and fuel inflation (Figure 6.2) returned to the levels they had before the 1998 events. On the other hand. Figures 5.3 and 5.4 (structural shock Sf ) suggest that during 2000 the expansion of aggregate demand induced inflationary pressures. This phenomenon is explained by a rapid expansion of industrial production in the United States and, consequently, in the demand for Mexican exports (Figure 7.1); and, by an important increase in domestic expenditure (private and public consumption and investment) that took place during 1999 and 2000 (Figure 7.2). The fact that inflation presented a downward trend from 1999 to 2001 (Figure 1 in Section 2) suggests that monetary policy actions prevented inflationary pressures, originated from a possible demand shock, from having an important effect on prices. Finally, Figures 5.1 to 5.6 suggest that inflationary pressures during 2002 and early in 2003 (third episode of monetary policy tightening) once again arose from the supply side. The contribution of the structural shock St shows an increasing effect on prices and a decreasing effect on output. In those years the economy experienced an important shock to public prices, including electricity and gas (Figure 6.2). Clearly, this evidence points to an adverse cost-push shock which resulted in inflation rising from 4.4 percent in 2001 to 5.7 percent in 2002 (Figure 1). Again, the restriction on monetary policy induced increases in interest rates (Figures 3.1 and 3.2) and, since no second round effects of consequence occurred, inflation decreased to 3.9 percent by December 2003 (Figure 1). Summing up, the results suggest that inflationary pressures during the first episode in our sample in which Banco de Mexico restricted the stance of monetary policy, came from the supply side in the form of an adverse cost-push shock. 15

This increase in the internal price of fuel deemed from an attempt to increase public revenues.

22

MANUEL RAMOS-FRANCIA AND ALBERTO TORRES

Figure 7.1: Industrial Production in the USA and Mexican Exports, 1998:1-2003:12

Note: Shaded areas correspond to episodes of monetary pohcy restriction. Figure 7.2: Domestic Expenditure, 1998:1-2003:12

Note: Shaded areas correspond to episodes of monetary poHcy restriction.

REDUCING INFLATION THROUGH INFLATION TARGETING

23

Then, during the second episode, the expansion of aggregate demand induced inflationary pressures. Finally, in the third episode in which the Central Bank restricted monetary policy, inflationary pressures came again from the supply side as adverse cost-push shocks. In what follows, these results are used to evaluate if the tightening of monetary policy that took place during the said episodes was consistent with inflation targeting principles.

5

Inflation Targeting, Monetary Policy, and Inflationary Shocks

The consensus in the inflation targeting literature (e.g., Bernanke et al. 1999, Clarida et al. 1999, Svensson 1997, 2000) is that, in the long run, the cost of maintaining inflation low and stable is minimized when a central bank follows two general guidelines. First, when inflationary pressures have their origin in a demand shock, the recommendation is not to accommodate the effect of the shock. Second, when inflationary pressures have their origin in a cost-push (supply) shock, the recommendation is to accommodate the shock and let it have a once-and-for-all effect on the price level and, thus, a temporary effect on inflation (first-round effects). However, if the temporary effect on inflation contaminates inflation expectations, so that further variations in prices can take place (second-round effects), then the recommendation is to modify the stance of monetary policy until inflation expectations get in line again with the inflation target.^^ The conclusion is that a central bank that systematically follows these two recommendations will significantly increase the probability of monetary policy being an effective nominal anchor. In what follows, the results from the previous two sections are used to evaluate monetary policy in light of the inflation targeting principles described above. The first episode of monetary policy restriction (1998:03-1999:01) coincides with the adverse cost-push shock experienced in 1998. Therefore, to evaluate the consistency of monetary policy with inflation targeting principles, it is necessary to consider inflation expectations. Figure 4.1 shows that the inflation expectations gap increased. This suggests that the restriction of monetary policy during this episode was in line with inflation targeting principles. Although inflationary pressures had their origin in the supply side, inflation expectations became contaminated by the shock and represented a threat to the disinflationary process. This prescription should be interpreted carefully in the case of an economy which is undergoing a disinflationary process, i.e., which has not converged to its long-run inflation target. In that case, monetary policy should have a clear restrictive bias at all times, independently of the nature of the shocks the economy faces.

24

MANUEL RAMOS-FRANCIA AND ALBERTO TORRES

The second episode of monetary policy restriction (2000:01-2001:01) coincides with inflationary pressures coming from a possible demand shock. As in the previous episode, the inflation expectations gap increased (Figure 4.1). Since inflationary pressures had their origin in the demand side, the restriction to the stance of monetary policy by Banco de Mexico was consistent with inflation targeting principles. The third episode of monetary policy tightening (2002:092003:03) took place while the economy was experiencing a series of adverse cost-push shocks in 2002 and early 2003. In this case, the inflation expectations gap (Figure 4.1) reversed its downward trend and increased through all of 2002 and early 2003. Again, since inflation expectations became contaminated, the tightening of monetary policy was in line with inflation targeting principles. To formalize the previous analysis, it is useful to refer to the literature on monetary policy rules (Taylor 1993; Clarida etal. 1999). The consistency of monetary policy with inflation targeting principles can be analyzed through a monetary policy rule of the following form:^^ (8)

t; = )3o + A < + Piinf^n -1 and P^>0, is consistent with inflation targeting principles and, in the long run, induces inflation to converge to its target and the output gap to zero. This is so, since, in response to a positive demand shock when both the inflation gap and the output gap increase, the rule reconmiends an increase in the real interest rate. However, when inflationary pressures come from a cost-push shock the recommendation is, in general, not to adjust the interest rate, since the increase in the inflation expectations gap is offset by a drop in the output gap. Nevertheless, if inflation expectations are contaminated by the shock, the increase in the inflation expectations gap will not be fully offset by the reduction in the output gap, and the recommendation is to raise interest rates so as to avoid second-round effects on prices. To formally test whether monetary policy in Mexico has been consistent with the inflation targeting principles, a rule like (8) is estimated and the parameters ^2 and J83 are tested statistically to be larger than one and zero, respectively. Given that the adjustment of interest rates by a central bank is usually only ^^ In general, the literature on monetary policy rules has found that simple specifications like (8), where the interest rate depends on the inflation expectations gap and on the output gap, are useful to characterize monetary policy.

REDUCING INFLATION THROUGH INFLATION TARGETING

25

gradual, the approach proposed by Clarida etal. (1999) to estimate (8) is followed. It is assumed that the actual interest rate, if, is determined as a weighted average of the desired interest rate, /*, and the actual interest rate observed in the previous period if_i, plus an exogenous interest rate shock (v^) with zero mean: (9)

/, = ( l - p ) i ; + p/,.i+v„

where parameter p takes values between 0 and 1 and measures the degree of interest rate smoothing. Then, combining the desired interest rate (8) with the smoothing equation (9), the relationship to be estimated is the following: (10)

/, = (1 -pXPo + ATT; + )32(7rf^i2 - K

ON

rx

--. 00 ^0

Csi

ON

od «o -^ >

C/3

I

»^

in in m

too

OS m ON r^ od t-^ o '-H ^ -^ CM CM

m

^c

cd c/a

o

^

1 n

p ON in t-^ iri CO NO .-;

g

X

S o C/3 D

13

«o^ ^ jd ^% ^

o

tScS

'^ & Cr
CM en cs

s

e:: o

M. AYHAN KOSE, GUY M . MEREDITH, AND CHRISTOPHER M . TOWE

Figure 4a: Trade with NAFTA Partners, 1980-2002 1980 = 100 1.600

1.600

. Exports 1.200 4

- Imports

1.200

800

800

400

400

0 I • , 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002

Figure 4b: Trade with NAFTA Partners, 1980-2002

r60

49

50

How HAS NAFTA AFFECTED THE MEXICAN ECONOMY?

Figure 4c: Trade with NAFTA Partners, 1980-2002

that NAFTA was not a trade diverting agreement, i.e., the expansion of trade was not at the expense of other countries. Krueger (1999) uses a gravity model and biannual data for the period 19871997 to analyze the impact of NAFTA on Mexico's trade performance. She finds that most of the increase in Mexican trade after NAFTA was driven by factors other than the agreement, including Mexico's unilateral reduction of tariffs following its entry into GATT in 1986 and the collapse of the Mexican peso in 1994. Krueger (1999, 2000) also studies the changes in the volume and patterns of trade flows between NAFTA partners and between NAFTA partners and the rest of the world using disaggregated data for the period 1990-1996. She finds that the categories in which Mexican exports to the United States registered the largest increase overlapped with those in which they rose most rapidly with the rest of the world. She interprets this result as an indication that NAFTA was not a trade-diverting agreement. Lederman et al. (2003) also use a gravity model and conclude that NAFTA did not have a significant impact on trade flows between the member countries and did not induce trade diversion. Studies employing export and import demand equations find that NAFTA contributed to the growth of Mexico's exports to the United States. For example, the United States International Trade Commission (USITC) (1997) estimates export and import demand functions using aggregate monthly series for the period 1989-1996. The results suggest NAFTA boosted Mexican (U.S.) exports to the United States (Mexico) by 1.0 (1.3) percent in 1994, 5.7 (3.8) percent in 1995,

M. AYHAN KOSE, GUY M . MEREDITH, AND CHRISTOPHER M . TOWE

51

and 6.4 (3.3) percent in 1996. The Congressional Budget Office (CBO) (2003) employs a similar methodology with quarterly aggregate data for the period 1969-2001 and finds that NAFTA boosted U.S. imports from Mexico by 8 percent and raised U.S. exports to Mexico by just over 11 percent in 2001. Some recent studies using sectoral data series find that NAFTA has had a more significant impact on trade flows. For example, Romalis (2002) examines the impact of tariff preference afforded by the United States to the NAFTA partners on different industries employing disaggregated data series covering the period 1980-2000. Using the tariff preference afforded to Canada and Mexico in 1999, he finds that between 25 to 50 percent of the increase in U.S. imports from Mexico after 1993 was driven by Mexico's preferential treatment associated with NAFTA. He also reports that Mexico's exports of conmiodities with the greatest NAFTA advantage grew much faster than others. Moreover, these commodities constituted a significant fraction of the U.S. demand, implying that NAFTA produced substantial trade diversion. Romalis argues that his results are different than those in Krueger because he uses longer and more disaggregated time series enabling him to analyze the impact of tariff preference on trade flows more precisely. In a related study, Agama and McDaniel (2002) conclude that NAFTA had a positive impact on the growth of trade flows in the region. They focus on the time-varying dimension of the U.S. tariff preference afforded to Mexico using the data over the period 1983-2001. They estimate that a 1 percentage point increase in tariff preference induced around a 4 percent increase in the volume of exports from Mexico to the United States and a roughly 6 percent increase in U.S. exports to Mexico during the period 1993:4-2001:4. Some other studies focus on the impact of NAFTA on particular sectors (USITC 1997; Burfisher et al. 2001; Fuako et al. 2002). These studies conclude that NAFTA resulted in trade diversion in textiles and apparel industries. Some other studies rely on computable general equilibrium (CGE) models to analyze the impact of NAFTA on trade flows. Unlike ex post studies, which employ econometric methods with historical time series data, ex ante studies with CGE models utilize various simulation methods to analyze a calibrated model economy for a particular base year. These studies estimate that NAFTA's longrun impact on Mexico's exports to the United States was between 3 and 16 percent (CBO 2003). Kouparitsas (1998) argues that these studies are not able to account for the dynamic effects of the agreement on trade flows because they use static models. He constructs a dynamic model and finds that the increase in trade flows between Mexico and its NAFTA partners was around 20 percent.^ A recent paper by Kehoe (2003) argues that CGE models severely underestimated NAFTA*s impact on the volume of regional trade. For example, a static CGE model

52 4.3

How HAS NAFTA AFFECTCD THE MEXICAN ECONOMY? How Has the Nature of Trade Changed after NAFTA?

Mexico's export base shifted toward manufactured goods following NAFTA's introduction. Although the share of manufactures in total exports had been increasing since at least 1980, the pace of diversification accelerated after the inception of NAFTA, as the average manufacturing share increased to more than 80 percent during the post-NAFTA period (1994-2002), from around 37 percent in the pre-NAFTA period (1980-1993) (Figure 5a). As a result, Mexico's export and import base has become one of the most diversified among emerging market economies. Vertical specialization (i.e., the value of a country's imports that are embodied in its exports) has increased among the NAFTA partners. Hunmiels et al. (2001) focus on the maquiladora trade and conclude that vertical specialization has played an important role in the growth of Mexico's exports since 1979. Maquiladora firms, which are mostly located along Mexico's northern border, import inputs from the United States, process them, and re-export them back to the United States. These firms specialize in the manufacture of electronics, auto parts, and apparel. Maquiladora firms grew substantially after the early 1980s, with the share of maquiladora exports in Mexico's total exports rising from 15 percent in 1980 to roughly 50 percent in 2001 (Figure 5b). The growth of the maquiladora sector accelerated during the 1990s. The average growth rate of real value added produced by the maquiladora sector was around 10 percent in the period 1990-2002, over three times the average growth of real GDP during the same period (Hanson 2002). Although there was a significant increase in employment during the first five years of NAFTA, recent research suggests that this increase reflected cyclical factors rather than the tariff preference afforded by NAFTA (Gruben 2001). Indeed, the maquiladora sector has been going through a recessionary period since early 2001 (Gruben 2004), which is explored further in Chapter 8. Intraindustry trade between Mexico and its NAFTA partners rose significantly. Intra-industry trade, which is defined as trade in similar but differentiated products, is also closely associated with maquiladora trade. OECD (2002: 161) reports that the share of intraindustry trade in Mexico's manufacturing sector rose from 62.5 percent in 1988-1991 to 73.4 percent in the period 1996-2000. OECD (2002: 162) also argues that most of the increase in intra-industry trade flows was related to NAFTA. Clark et al. (2001) find that a number of manufacby Brown et al. (1992) estimates that Mexico's exports (imports) relative to its GDP would increase roughly 51 (34) percent during the period 1988-1999, while the data suggest that the relative increase in exports (imports) was larger than 140 (50) percent in the same period. Kehoe alsofindsthat the models were unable to account for much of the increase in sectoral trade flows.

M. AYHAN KOSE, GUY M. MEREDITH, AND CHRISTOPHER M. TOWE

53

Figure 5a: Diversification of Exports, 1980, 1993, and 2000

mm w/A'y////:4'Z-%

"^^m

Figure 5b: Maquiladora Exports and Imports, 1980-2002 Share of G D P (percent) 60 -,

60

50 H

50

40

40

30

30

20 \

V 20

10

\ 10

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

turing industries registered substantial increases in intra-industry trade flows between Mexico and the United States after the inception of NAFTA. NAFTA also boosted intrafirm trade in the region. Intrafirm trade is defined as cross-border trade between multinational companies and their affiliates. In the

54

How HAS NAFTA AFFECTED THE MEXICAN

ECONOMY?

case of Mexico, most of these affiliates are maquiladora firms. OECD (2002: 159-170) reports that intrafirm exports from Mexico to the United States rose by more than 3 percent in the period 1992-1999 and accounted for more than twothirds of total exports in 1999. NAFTA has also resulted in a substantial increase in the variety of products traded between Mexico and its partners. Hillbery and McDaniel (2002) analyze the impact of the increase in the variety of products on the volume of trade between Mexico and the United States. They find that almost 25 percentage points of the 190 percent increase in Mexico's exports to the United States were attributable to the increase in the number of traded good varieties, while more than 8 percentage points of the 93 percent increase in the exports of the United States to Mexico were accounted for by the growth of product variety. In a related paper, Kehoe (2003) documents that the sectors with very small trade in 1988 experienced the largest increases in exports in the period 1988-1999. For example, the share of the "motor cars for transport of passengers and engines" sector in total Mexican exports to the United States increased from under 1 percent in 1988 to 15 percent in 1999.

4.4

Growth of Financial Flows

Foreign direct investment (FDI) flows between Mexico and its partners strengthened after NAFTA. The agreement contained various provisions that improved the relative standing of investors from the partner countries in Mexico and expanded the sectors in which they could operate. These changes helped boost FDI flows to Mexico from 12 billion U.S. dollars over 1991-1993 to roughly 54 billion U.S. dollars in the 2000-2002 period and increased the share of FDI flows in domestic gross fixed capital formation (investment) from 6 percent in 1993 to 11 percent in 2002, mainly on account of inflows from Mexico's NAFTA partners (Figure 6). Net portfolio flows also increased rapidly after the inception of NAFTA. Recent research suggests that the NAFTA membership significantly affected FDI inflows to Mexico. Cuevas et al. (2002b) employ panel regressions using the data of 45 countries for the period 1980-1999 to analyze the impact of NAFTA membership on FDI flows. They find that Mexico's participation in NAFTA led to roughly a 70 percent increase in FDI flows. In a related study, Waldkirch (2003) examines data from 11 countries for the period 1980-1998 and finds that NAFTA induced a 40 percent increase in the volume of FDI flows. He argues that NAFTA's impact on FDI inflows to Mexico was the result of increased vertical specialization as well as the effect of the agreement on Mexico's commitment to liberalization and reform programs.

M. AYHAN KOSE, GUY M . MEREDITH, AND CHRISTOPHER M . TOWE

55

Figured: Gross FDI Flows, 1980-2002

1950 1982 1984 19S6 1988 1990 1992 '

5

Dynamics of Macroeconomic Volatility

One of the major policy challenges for an emerging market economy like Mexico is to establish a stable macroeconomic environment. A burgeoning literature has documented a negative relationship between macroeconomic instability (volatility) and growth (Kose et al. 2005). There are several channels through which NAFTA membership could help Mexico to establish and sustain a stable economic environment, some of which are discussed below.

5.1

Economic Integration and Volatility: Theory and Recent Empirical Studies

The theoretical impact of increased trade and financial flows on output volatility depends on various factors, including the nature of financial flows, patterns of specialization, and the sources of shocks. For example, if trade openness is associated with increased interindustry specialization across countries and industryspecific shocks are important in driving business cycles, this could lead to an increase in output volatility. However, if increased trade is associated with increased intra-industry specialization across countries, which leads to a larger volume of intermediate inputs trade, then the volatility of output could decline

56

How HAS NAFTA

AFFECTED THE MEXICAN ECONOMY?

(Kose et al. 2003a). In addition, economic theory suggests that increased access to international financial markets should dampen the volatility of consumption, while inducing an increase in investment volatility. Recent studies are unable to establish a clear empirical link between stronger economic linkages and macroeconomic volatility. While some of these studies find no significant relationship between the increased degree of economic interdependence and domestic macroeconomic volatility, some others find that an increase in the degree of trade openness leads to higher output volatility, especially in developing countries. Kose et al. (2003a) find that while trade openness increases the volatility of output, income, and consumption in emerging market economies, it reduces the relative volatility of consumption, implying that it improves the consumption-risk-sharing possibilities. They also document that increased financial integration is associated with rising relative volatility of consumption, but only up to a certain threshold.

5.2

Changes in Macroeconomic Volatility in Mexico after NAFTA

Macroeconomic volatility in Mexico declined markedly after the inception of NAFTA. This can be seen in the uniform and sizeable decline in the variance of several macroeconomic aggregates between the 1980-1993 and 1996-2002 periods (Figure 7a).^^ In particular, output volatility decreased by almost 30 percent and the volatility of investment by more than 40 percent in the latter period. Consistent with the prediction of economic theory, increased trade and financial linkages also led to a reduction in the volatility of consumption in Mexico. ^^ In addition, consumption became slightly less volatile than output in the 1996-2002 period. This result, along with the increased cross-country consumption correlations documented in the next chapter, suggests that Mexico has been able to utilize the risk-sharing benefits of increased trade and financial linkages with its NAFTA partners. The decreased volatility of the Mexican economy during the past eight years could be the result of several factors, including some associated with NAFTA. In particular, it could be the result of NAFTA's effect on intraindustry and vertical trade, and the increased importance of regional rather than country-specific shocks in driving the Mexican business cycles, which are further explored in Section 6.3. Torres and Vela (2003) argue that the advent of NAFTA led to an ^^ Volatility is measured as the standard deviation of the Hodrick-Prescott filtered quarterly series. 11 Prasad et al. (2003) show that welfare gains associated with intemational consumption risk sharing are quite large in emerging market economies, including Mexico.

M. AYHAN KOSE, GUY M. MEREDITH, AND CHRISTOPHER M . TOWE

57

Figure 7a: Volatility of Macroeconomic Aggregates, 1980-2002

Figure 7b: Volatility of Policy Variables and Prices, 1980-2002

g 198(^1993 G1994^2002

increase in the correlation of the Mexican exports and imports over time,which in turn dampened the volatility of net exports during the period 1991-2001. The decrease in the volatility of business cycles could also have been the result of increased stability of domestic macroeconomic policies. For example, there was a uniform decline in the volatility of several exogenous policy variables and prices (Figure 7b). Cuevas et al. (2002b) also find that the volatility of Mexico's output decreased by almost 50 percent during the 1990s relative to the previous decade. They argue that the substantial decline in volatility could be the result of the implementation of sound monetary and fiscal policies over the period 19962001.

58

How HAS NAFTA

AFFECTED THE MEXICAN ECONOMY?

6

Dynamics of Comovement

6.1

Economic Integration and Comovement: Theory and Recent Empirical Studies

In theory, increased trade linkages have ambiguous effects on the comovement of business cycles. On one hand, stronger trade linkages can result in more highly correlated business cycles, since they generate both demand- and supplyside spillovers across countries. For example, on the demand side, an investment or consumption boom in one country can generate increased demand for imports, boosting economies abroad. Moreover, if stronger trade linkages are associated with increased intraindustry specialization across countries, and industry-specific shocks are important in driving business cycles, then business cycle comovement would be expected to increase. However, the degree of comovement might diminish if increased trade is the result of a rise in interindustry trade and industry-specific shocks are important in driving business cycles (Kose and Yi 2004). Increased financial flows also have an ambiguous theoretical effect on business cycle correlations. For example, stronger financial linkages could result in a higher degree of synchronization of output fluctuations by generating large demand-side effects. Contagion effects that are transmitted through financial linkages could also lead to heightened cross-country spillovers of fluctuations. However, financial linkages could stimulate specialization of production through the reallocation of capital in a manner consistent with countries' comparative advantage. This type of specialization, which could result in more exposure to industry- or country-specific shocks, could lead to a decrease in the degree of output correlations while inducing stronger comovement of consumption across countries (Kalemli-Ozcan et al. 2003). Several recent studies suggest that trade linkages result in greater business cycle synchronicity. For example, using the results from cross-country or crossregion panel regressions, Frankel and Rose (1998), Kose and Yi (2004), and others show that, among industrialized countries, pairs of countries that trade more with each other exhibit a higher degree of business cycle comovement. Calderon et al. (2002) find that the impact of trade intensity on business cycle comovement is positive but smaller in a sample including both industrialized and developing countries. Calderon (2003) documents that the impact of trade intensity on cross-country business cycle correlation is larger if the two countries have a free trade agreement. Recent empirical studies also show that stronger financial linkages could lead to higher cross-country output and consumption correlations. Kose et al. (2003b) study the impact of increased financial linkages on the correlations between

M. AYHAN KOSE, GUY M . MEREDITH, AND CHRISTOPHER M . TOWE

59

fluctuations in individual country aggregates (output, consumption, and investment) and those in corresponding world (G-7) aggregates. They report that countries that are more open to financial flows have higher business cycle correlations with the G-7 aggregate.

6.2

Comovement of Business Cycles after NAFTA

The agreement appears to be associated with an increased degree of comovement of business cycles of Mexico and its NAFTA partners.^^ The increase in comovement can be seen from the marked increase in cross-country correlations of the major macroeconomic aggregates, including output, consumption, and investment (Figures 8a and 8b). In particular, the output correlation between Mexico and its NAFTA partners rose from almost zero in the pre-NAFTA period to around 0.75 during the post-crisis period. There was a significant increase in consumption correlations, which could be a reflection of Mexico's being able to diversify its consumption risk more effectively after NAFTA. Cross-country correlations of exports and imports also increased significantly after the inception of NAFTA, which could be associated with the jump in intraindustry trade in the region. In addition, there was a substantial increase in the cross-country correlations of manufacturing and industrial production, which could be the result of the increase in the trade of manufactured goods. The increase in business cycle comovement in the region is documented by several studies. Torres and Vela (2003) find that business cycle correlations between Mexico and the United States rose as trade linkages between the manufacturing sectors of the two countries became stronger after the inception of NAFTA. Using quarterly data for the period 1991-2001, they document that the cyclical dynamics of Mexico's exports, imports, and output became more responsive to the changes in U.S. exports and imports. Cuevas et al. (2002a) use the quarterly output series of the NAFTA members for the period 1981-2001 and find that business cycles in Mexico became more synchronized with the cycles in Canada and the United States after NAFTA. They also document that there was an increase in the correlations of sectoral business cycles in Mexico and the United States during the period 1997-2001, with the correlation of manufacturing sector output rising from 0.28 to 0.97.

^^ Comovement is measured as the cross-country correlation of the Hodrick-Prescott filtered quarterly series of main macroeconomic aggregates (output, consumption, investment, exports, and imports ) for Mexico, Canada, and the United States.

60

How HAS NAFTA AFFECTED THE MEXICAN ECONOMY?

Figure 8a: Comovement of Economic Variables in Mexico and the United States, 1980-^2002

m

ii

m m Pi

I

mi

•

mi

IP mm

m

I

m

iybf4:i-2Ui'i-:4

i

-m

II ii:

i

g .yy

Figure 8b: Comovement of Economic Variables in Mexico and Canada, 1980-2002

M. AYHAN KOSE, GUY M . MEREDITH, AND CHRISTOPHER M . TOWE

6.3

61

How Have the Regional Cycles Changed?

Simple correlations do not allow conclusive statements about the changes in the degree of business cycle comovement. First, cross-country correlations capture only the contemporaneous comovement in macroeconomic variables, and do not account for common fluctuations associated with "leads" and "lags." Second, correlations can account for the degree of comovement in only a single macroeconomic variable. Moreover, correlations are not helpful in analyzing the relative importance of different types of factors and/or shocks in explaining business cycle comovement. To overcome these problems, a dynamic latent factor model that captures the dynamic comovement in output, consumption, and investment series of the NAFTA partners is estimated here. The model helps account for contemporaneous as well as temporal covariation among the variables and enables us to study how common (regional) and country-specific factors affect the fluctuations in different macroeconomic variables. A very brief explanation of the model is presented here.^^ There are K dynamic, unobserved factors thought to characterize the temporal comovements in our cross-country panel. Let A^ denote the number of countries, M the number of time series per country, and 7 the length of the time series. Observable variables are denoted by y,^^, for / = 1,... ,M x A^, i t = l,...,r. There are two types of factors: N country-specific factors (/„^^""''^, one per country), and the single regional factor ( / ^ ) . Thus, for observable /

yu-a,^bf^f,''+br""^nr"^+e,,,,

Ee„ej,_,=0

for

i*j,

where n denotes the country number. The coefficients b/ are called "factor loadings," and reflect the degree to which variation in y^,^ can be explained by each factor. We use output, consumption, and investment data for each NAFTA member, so there are MxA^ (3-3 = 9) time series to be "explained" by the / / + 1(3 + 1 = 4) factors. The "unexplained" idiosyncratic errors e,,^ are assumed to be normally distributed, but may be serially correlated. The estimated regional factor model explains some of the major economic events since 1980. For example, the behavior of the regional factor is consistent with the recessions of the early 1980s and 1990s, and the expansionary period of the late 1980s. The Mexican country factor is quite successful in replicating some important episodes of cyclical fluctuations as it is consistent with the recessions of the early 1980s, 1982-1983, 1986, and the mid-1990s.

^^ Kose et al. (2003c) provide a detailed discussion of these models. Kose et al. (2003c, 2004d) document that there is a significant common component explaining business cycles in industrialized countries.

62

How HAS NAFTA AFFECTED THE MEXICAN ECONOMY?

The importance of the regional factor in driving business cycles is analyzed using variance decompositions. To measure the relative contributions of the regional, country, and idiosyncratic factors to variations in aggregate variables in each country, the variance of each macroeconomic aggregate is decomposed into the fraction that is due to each of the two factors and the idiosyncratic component in three different time periods. Specifically, the fraction of variance of each macroeconomic aggregate explained by the following factors is computed: (i) a regional factor that is conmion across all variables/countries, (ii) countryspecific factors which are common across the main aggregates within a country, and (iii) factors specific to each variable. Regional factors became more important in driving business cycles in Mexico with the advent of NAFTA. The proportion of output volatility explained by the regional factor in Mexico rose from less than 1 percent in the period 1980-1993 to more than 19 percent in period 1994-2002, while the variance of investment accounted for by the regional factor increased almost tenfold during the same period (Figure 9a). A significantly large fraction of consumption fluctuations in Mexico was driven by the regional factor after the inception of NAFTA. While the importance of the country-specific factor remained relatively stable during the later period, the idiosyncratic factor became less important. Examination of business cycles in manufacturing and industrial production using a set of single factor models suggests that the regional factor also played a more important role in explaining these variables after NAFTA (Figure 9b). These findings are consistent with recent empirical studies. Kose et al. (2003c) examine the roles of world-, regional-, and country-specific factors in a 60-country sample that includes both developed and developing countries. Their results suggest that the North American regional factor played an important role in driving macroeconomic fluctuations in Mexico, Canada, and the United States. Cuevas etal. (2002b) employ simple regression models to analyze the responsiveness of Mexican business cycles to changes in economic activity in the United States. They show that changes in U.S. output growth accounted for a larger fraction of the variation in the growth rates of output and industrial production in Mexico during the period 1997-2001 than before this period.

6.4

Changes in the Channels of Business Cycle Transmission

A multi-country dynamic stochastic general equilibrium (DSGE) model is constructed to illustrate the channels through which NAFTA could effect business cycle spillovers among its participants. The multi-country DSGE model is a natural setting for this purpose because it accounts for the demand- and supplyside spillover channels that are critical in transmitting business cycles. The model, developed by Kose and Yi (2004), extends the two-country free trade.

M. AYHAN KOSE, GUY M. MEREDITH, AND CHRISTOPHER M . TOWE

63

Figure 9a: Variance in Output, Consumption, and Investment Explained by the Regional Factor, 1980-2002

mm

1993:4

J21994:1-2002:4

Figure 9b: Variance in Manufacturing and Industrial Production Explained by the Regional Factor, 1980-2002

^^^H

64

How HAS NAFTA AFFECTED THE MEXICAN ECONOMY?

complete market model of Backus etal. (1994) by including three countries, trading frictions (tariffs and transportation costs), and allowing for international financial autarky. The model economy includes a traded intermediate goods-producing sector and a nontraded final goods-producing sector. Perfectly competitive firms in the intermediate goods sector produce traded goods according to a Cobb-Douglas production function. When the intermediate goods are exported to other countries, they are subject to transportation costs, which are considered as a proxy for tariffs and other nontariff barriers, as well as transport costs. It is assumed that each country is completely specialized in the production of an intermediate good. Each country's output of intermediates is used as an input into final goods production. Final goods firms produce their goods by combining domestic and foreign intermediates via an Armington aggregator. These assumptions imply that, in the model economy, imports from Mexico are used as intermediate inputs to produce final consumption and investment goods in the United States and Canada. In each country, there are representative agents who derive utility from consumption and leisure. The model is calibrated to reflect some basic structural features of the NAFTA members. Since the objective is to analyze the interdependence of business cycles in Mexico and its NAFTA partners, it is assumed that the three countries in the model are Mexico, its NAFTA partners, which are an aggregate of Canada and the United States, and the rest of the world, represented by an aggregate of the members of the European Union and Japan. It is assumed that Mexico accounts for 4 percent of the world economy and each of the other two countries for 48 percent of the world economy. The elasticity of substitution between domestic and foreign goods is set at 1.05. The impact of NAFTA is simulated by changing the level of transportation costs (trading frictions) between the member countries. The model is solved following the standard linearization approach in the international business cycle literature. The results suggest that agreements such as NAFTA that lower trade frictions can magnify the impact of external shocks on the Mexican economy. To analyze the responses of macroeconomic aggregates in Mexico to shocks originating in Canada and the United States, the impulse responses of Mexico's variables to a temporary productivity (supply) shock in Canada and the United States were computed. The results indicate that the responses of Mexican output, consumption, and investment to the external shock become larger after the inception of NAFTA (Figures 10a, 10b, and 10c). In addition, pre- and post-NAFTA simulations illustrate the substantial increase in Mexican exports that results from the lowering of tariffs after the advent of the agreement (Figure lOd). In other words, the reduction in trade frictions in the model results in greater trade intensity in the region, which in turn leads to a higher degree of business cycle interdependence.

M. AYHAN KOSE, GUY M. MEREDITH, AND CHRISTOPHER M . TOWE

a

(O VH O

c o^ .-< a t i T3

c

^•^

€ Si s

.00

psi

tj

% 0) :S 0

w

0

d

13 t^ 0 0 ex x- C

W c«

^-^

>^

•0

(U C/J

0

rt 1)

0

§ u 0 -§ c«

0

0, ACR > 0), magnifying the necessary increase in domestic interest rates (A/) to avoid capital flight: A/ = AFP + ACR. Is there any causality between FP and CP? Could it be that FP and CR share a common generating factor? In the hope of finding a causality relation between FP and CRy Powell and Sturzenegger (2000) analyzed their relation in light of dollarization, i.e., when FP = 0. When the local currency is abandoned in favor of a hard currency, what is the effect on CR, if any? Choosing a date when an event had an effect on FP, they estimate the evolution of abnormal country risk return, that is to say the effect on CR, given a model predicting expected return based on portfolio theory. They observe the direction of ACP relative to AFP in specific institutional shocks, described in Table 1. For Austria, Belgium, and Ireland, seen as EMU insiders, a decline in the forward premium is associated with a decline in country risk. On the contrary, for Portugal, Denmark, and Sweden, a decline in the forward premium is associated with an increase in country risk. This increase in CR actually lowered the probability of EMU because the three were seen as outsiders. Spain and Finland fall in-between. In Latin America, the correlation is generally positive. The zero correlation in Chile and Colombia may be due to a lesser degree of dollarization. A high degree of liabilities dollarization may contribute to increase exposure to external shocks, lower policy degrees of freedom dampen of other

ARGENTINA AND BRAZIL RISK: A "EUROCENTRIC" TALE

165

transmission mechanisms. Vulnerability to higher sovereign spreads may work through the balance sheet channel. Intra-eurozone FP collapses to zero, but CR can go up or down. Figure 1 plots CR (based on bond yield spreads involving all maturities) for two alleged EMU outsiders. Even before FP goes to zero, there is a sharp decline in the CR of Portugal. When the list of EMU members is announced in May 1998, markets expect higher CR in the shortest term, rising from minus 50 basis points to 50 basis points in September 2003, passing through zero in January 2001. After EMU starts in 1999, CR converges but the volatility of Denmark's is greater. If the correlation between FP and CR is indeed positive, a rise in the probability of sovereign default will be magnified in the currency premium rather than offset by it. Causality can go in the other direction, i.e., an increase in the forward premium can bring about wider sovereign default spreads, due to maturity and/or currency mismatches in the balance sheets. Marcio Garcia calls "cousin risk" the case of positive correlation. Garcia and Lowenkron (2003) investigates how widespread it is and finds that it is not present in all emerging markets. Comparing descriptive statistics in a sample of 25 countries, reproduced in Table 2, they find no effect on cousin risk for external balance (rows 1-5) or government solvency (rows 6-10) variables. But the degree of financial deepening (measured by domestic credit to private sector, row 12) and the government's currency mismatch (measured by net external liabilities, row 11) is a significant determinant of "cousin risks". Political Noise and Spreads Decoupling in Mercosur For countries that are not financially integrated and have credit risk, there may be an additional risk, due to the differences between domestic ("onshore") financial regulations and international ("offshore") legal standards. In this onshore-offshore premium we have the same issuer, the same currency but a different jurisdiction. International bonds are usually issued in financial centres and governed by local law. This jurisdiction risk {JR) is often apportioned between FP and CR, otherwise: / - /* = FP^CR^ JR. In an application using daily data for Brazil between January 2000 and July 2003 (excluding outliers), Grandes (2003) calculates the forward premium using the 1 year forward rate, assuming strict covered interest rate parity and a flat fwd yield curve beyond 1 year (some term structure effects unaccounted for). There are no secondary market prices or redemption yields for on-shore issues in foreign currency. Notas do Tesouro Nacional are dollar-indexed instruments, but these are primary issues. Country risk is measured by the yield differential between Brazilian and US Treasury bonds at similar maturities (between 10 and 20 years). In anticipation of the Argentine collapse, the forward premium more

166

JORGE BRAGA DE MACEDO AND MARTIN GRANDES r eo/60/62 eO/90/62

eO/10/62

8

[- 20/60/62

cs

I

20/90/62 20/10/62

ex

10/60/62 ON

OS

10/90/62

I

h 10/1.0/62

0; Si /(I - 55) > 1, 53 < 0; and 54 $0. We call (4) the baseline model, and to check robustness properties of the result we estimate an alternate version where actual inflation and the exchange rate are replaced by their deviations from trend values (the gap model).^^

^^ In symbols, this is given by /, =Oo +0i(7tt -^)+02>'r ^O^ixr^ -jcr)+04(xr,_i -xr)

MONETARY POLICY RULES IN EMERGING MARKET ECONOMIES

223

According to Taylor (2001), the exchange rate is likely to have only marginal significance in (4), with 54 approximately equal to -53 P-^ As in Table 3, the implicit assumption is that shocks to the exchange rate represent temporary deviations from its long-run value. To smooth exchange rate movements, the central bank might raise interest rates in response to currency depreciations in the current period. But because such depreciations do not have a significant impact on the central bank's inflation forecast, it would lower rates in the next period to partially or fully offset the rate increase. However, if shocks to the exchange rate are large and persistent and the central bank places a higher weight on exchange rate stability, we would expect significant negative coefficients on both current and lagged values of the exchange rate in (4).

3.2

Results of the Baseline Model

Table 6 presents results of the baseline model for the 13 countries. The output gap was estimated by using a Hodrick-Prescott measure of trend output, and in some cases we controlled for known episodes of crises by using suitable dunmiy variables. We used the short-term interest rate—daily interbank rate in most cases—as the relevant left-hand variable of the reaction function. This choice was guided by the fact that many central banks until recently did not have an official policy rate. Furthermore, to the extent that monetary operating regimes varied considerably during the sample period, a short-term market rate was thought to be more appropriate in capturing the variety of operating procedure than the actual policy rate. In any case, the correlation between the policy rate and the short-term rate is uniformly high in all countries, except India, which follows a multiple instrument approach to influence the call money rate. The results suggest that simple rules fit the interest-rate-setting behaviour of emerging economies reasonably well. For most countries (4) explains between 70 and 90 per cent of the actual movement of short-term interest rates. The estimates are free from problems of autocorrelation, and all coefficients have the expected signs, although some are only weakly significant. As Figure Al in the annex shows, the model closely tracks the historical policy path and captures most of the turning points since the mid-1990s. The errors seem to be somewhat higher for Brazil and South Africa: in the former case the estimated rule mostly overpredicts the short-term rate since the devaluation of the real in 1999. In the latter case, the problem seems to be underprediction in much of the period between the 1998 Asian financial crisis and the sharp fall of the rand towards the ^^ Our model differs from Taylor's to the extent that we assume that central banks respond to changes rather than levels of the real exchange rate.

224

M.S. MOHANTY AND MARC KLAU

Table 6: Taylor Reaction Function (Baseline Model)^ Const

TCf

yt

Axr,

Axr,_;

/,_/

-0.09 (-3.44)

-0.09 (-2.00) -0.08 (-1.70)

0.70 (7.25) 0.57 (3.80)

R^

BG-LM

088

0.02

0.76

1.13

0.28 (2.74)

0.71

0.35

0.83 (9.10)

0.96

0.30

0.86

0.30

India

1.11 (1.66)

0.13 (3.10)

Korea

1.15 (1.31)

0.66 (1.54)

0.13b (2.70) 0.29 (1.78)

Philippines

4.35 (4.17)

0.51 (3.77)

0.35 (0.78)

-0.15 (-2.25)

Taiwan

0.49 (1.31) 1.19 (1.85)

0.23 (1.85)

0.13 (2.04)

-0.04 (-0.78)

0.06 (1.01) 0.01 (0.40)

0.56 (1.86)

0.37 (2.61)

-0.11 (-3.28)

-0.20 (-2.41)

0.58 (3.48)

Brazil

4.12 (1.08)

0.08 (3.35)

-0.33 (-2.20)

0.97 (4.87)

0.35 (2.78)

0.72 (5.11) 0.32C (4.03)

2.20

0.32 (0.25)

0.23 (1.28) -0.35c (-2.40)

0.81

Chile

0.75

0.98

Mexico

1.79 (1.28)

0.55 (4.71)

0.98 (0.93) 0.32 (1.25) 0.74 (2.09)

-0.63 (-2.93)

-0.16 (-2.03)

0.50 (3.93)

0.86

1.15

Peru

2.57 (1.44)

0.19 (2.37)

0.15 (1.09)

-0.15 (-2.05)

-0.23 (-4.11)

0.86 (12.98)

0.92

0.46

Czech Republic

0.56 (1.36)

0.33 (2.69)

-0.06 (-1.11)

0.04 (0.49)

0.66 (5.98)

0.96

0.28

Hungary

0.96 (1.03)

0.20 (1.76)

-0.19 (-2.02)

0.04 (0.53)

0.75 (5.69)

0.97

2.11

Poland

2.13 (1.94)

0.17 (3.31)

0.20 (2.37) 0.35d (1.11) 0.66 (4.89)

-0.07 (-1.16)

0.02 (0.31)

0.75 (8.38)

0.96

0.29

-0.59 (-0.42)

0.08 (1.09)

0.04d (0.21)

-0.06 (-2.86)

-0.06 (-2.46)

0.98 (11.50)

0.74

0.37

Thailand

South Africa

-0.21 (-5.04)

^t-statistics in parentheses, n = annual percentage change in consumer prices (for India, wholesale prices); }; = output gap; Axr = change in the real effective exchange rate (an increase means appreciation); / = interest rate (dependant variable); BG-LM = Breusch-Godfrey Serial Correlation LM test (the F-statistics are given). — ^Industrial production. — ^t-4. — ^t-l.

end of 2001. Overprediction bias appears to have been a common phenomenon since the beginning of the current cycle, especially in Asia, where monetary policy has been eased sharply. Table 7 presents the relevant short- and long-run responses. One general finding is that emerging economies central banks seem to adjust interest rates by small steps. The simple average response to an inflation or output shock is little above one-third in the quarter in which the shock occurs, and is about one-sixth

MONETARY POLICY RULES IN EMERGING MARKET ECONOMIES

225

Table 7: Interest Rate Response from a Simple Reaction Function^ Inflation

Output

Exchange rate

Short term

Long term

Short term

Long term

Short term

Long term

India Korea Philippines Taiwan Thailand

0.13 0.66 0.51 0.23 0.56

0.43 1.53 0.71 1.35 1.33

0.13 0.29 0.35 0.13 0.37

0.43 0.67 0.49 0.76 0.88

-0.18 -0.29 -0.09 -0.03 -0.31

-0.60 -0.67 -0.13 -0.18 -0.74

Brazil Chile Mexico Peru

0.08 0.97 0.55 0.19

0.29 1.43 1.10 1.36

0.98 0.32 0.74 0.15

3.50 0.47 1.48 1.07

-0.10 0.00 -0.79 -0.38

-0.36 0.00 -1.58 -2.71

Czech Republic Hungary Poland

0.12 0.20 0.17

0.75 0.80 0.68

0.32 0.35 0.66

2.00 1.40 2.64

0.03 -0.15 -0.05

0.19 -0.60 -0.20

South Africa

0.08

4.00

0.04

2.00

-0.12

-6.00

^Based on the Taylor rule specification of Table 6 using the inflation rate.

to an exchange rate shock. But, as suggested by the coefficients on the lagged interest rate, the response rises through subsequent quarters and is eventually much higher. The degree of interest rate smoothing is particularly high in Peru, Poland, South Africa and Taiwan (China), but small in Chile and the Philippines.^^ Another important finding is that the monetary policy response to inflation seems to be higher in Asia and Latin America than in Central Europe, and is particularly strong in South Africa. In a number of countries the coefficient on inflation exceeds one, indicating that central banks do not accommodate inflationary pressures. Within Asia, the estimated long-run inflation coefficients are low in India and the Philippines, perhaps explaining their relatively high inflation rates. In Latin America, the reaction coefficients point to a non-accommodating monetary policy stance in Chile, Mexico and Peru, whereas in Brazil

^^ The high value of the coefficient on the lagged interest rate in some cases could reflect a number of facts. On the one hand, it might suggest nonstationarity of the interest rate series and the related problem of spurious statistical relation. While more conclusive evidence requires testing the order of integration of the interest rate series, the usefulness of such a test in our case is limited by the small sample size. On the other hand, as shown by Rotemberg and Woodford (1999), the coefficient on a lagged interest rate close to or even exceeding one could be an optimal policy when the private sector is sufficiently forward-looking. By committing to raise rates sufficiently in future in response to an initial inflation pressure, the central bank can have a large effect on aggregate demand without having to raise the rates substantially in the current period.

226

M.S. MOHANTY AND MARC KLAU

the response appears to be weaker. The relatively weak monetary policy response in Central Europe, however, needs to be interpreted with some caution. During the 1990s, inflation rates were significantly influenced by large relative price movements brought about by the gradual removal of price controls. The results confirm the findings of other studies that central banks in transition economies may have accommodated some of the non-monetary price pressures in order to reduce the output costs.^^ In so far as output stabilization is concerned, the evidence is mixed. The output gap is a statistically significant determinant of short-term interest rates in the Czech Republic, India, Korea, Mexico, Poland, Taiwan, and Thailand. In other countries, it is either weakly significant or not significant. However, the coefficients may be downward biased, since our estimates of the output gap may not adequately measure demand gaps. It is well recognized that estimating an accurate measure of potential output is more difficult for emerging economies than for industrial economies, given the relatively greater importance of supply shocks in output developments in the former. The implied long-term responses suggest that monetary policy response to output is stronger in Latin America and Central Europe than in Asia (Table 7). With the exception of Chile, the long-term response of monetary policy to the output gap exceeds one for all countries in the former two regions. This result might be related to the role of other polices in output stabilization. For instance, fiscal policy has played an important role in Asia since the 1997-1998 financial crisis, perhaps reducing the need for a more aggressive response of monetary policy. In contrast, many recent studies suggest that the fiscal policy response to output in Latin America has been extremely weak or procyclical.^^ As a result, central banks may have played an active role in output stabilization. To cite an example, Sidaoui (2003) discusses the dilemma posed by a procyclical fiscal policy response to the central bank in Mexico. He argues that because fiscal policy tended to accentuate rather than attenuate demand shocks, the central bank had to be more aggressive in responding to demand fluctuations. The results also provide evidence on central bank reactions to exchange rate movements. First, they strongly reject the hypothesis that central banks do not react to exchange rate volatility. In all countries, except for Chile, current period real exchange rate changes have uniformly negative signs in the reaction function, suggesting that central banks "lean against the wind" by raising rates when the exchange rate depreciates and vice versa. This relationship is statistically significant in all countries except for the Czech Republic, Poland, and Taiwan.

2^ See, for example, Coorey et al. (1998) and Pujol and Griffiths (1998). 2^ See, for example, Gavin and Perotti (1997) and IMF (2001).

MONETARY POLICY RULES IN EMERGING MARKET ECONOMIES

227

Second, as noted earlier, the coefficient on the lagged exchange rate term provides information about whether exchange rate shocks are persistent or temporary and about the nature of central bank reaction to them. A significant and positive coefficient on the lag exchange rate is an indication of mean-reverting exchange rate movements, and implies that the central bank reverses its interest rate action in this period. As may be seen from Table 6, although the coefficient on the lagged exchange rate is positive in a number of countries, it is not statistically significant. On the other hand, whenever the lagged exchange rate was negative it was also statistically significant, indicating a high degree of persistence of exchange rate shocks in Korea, India, Mexico, Peru, Thailand, and South Africa. In Chile, the coefficients on the current and lagged exchange rates have wrong signs, although they offset each other, suggesting that monetary policy is neutral with respect to exchange rate movements. Third, taken together, the results suggest a high degree of interest rate response to the exchange rate. For instance, in India, Mexico and Peru, the long-term response of the interest rate to the exchange rate exceeds that for inflation and output gap and is also high in many other countries. Another measure of relative importance of each variable for monetary policy is provided by its contribution to the mean and standard deviation of the estimated short-term interest rate, represented by coefficients r] and s in Table 8. The 7] coefficient is a measure of contribution to interest rate volatility, which is computed by weighting the ratio of the standard deviation of each determinant to the standard deviation of the interest rate with its respective regression parameter. In a similar way, the e coefficient evaluates the contribution to the mean. Thus, r]j=Sj(Sj/si), ej=Sj(fij/^i),

where Sj, Sj, and //^are the regression coefficient, standard deviation and mean of the determinant j , respectively, while Si and jLLf are the standard deviation and mean of the interest rate, respectively. As the table shows, in all countries, exchange rate volatility is a significant contributor to interest rate volatility. The large rj values not only reflect the relatively large reaction coefficient on the exchange rate but also a considerable degree of exchange rate volatility. The contributions of inflation and output gap to interest rate volatility are comparatively smaller. This is also true for the contribution to the mean interest rate during the sample period.

228

M.S. MOHANTY AND MARC KLAU

Table 8: Contribution to Interest Rate Responses^

n

Exchange rate

Output

Inflation

e

India Korea Philippines Taiwan Thailand

0.16 0.22 0.40 0.23 0.22

0.08 0.27 0.29 0.07 0.22

n 0.22 0.15 0.13 0.10 0.22

OlO 0.24 -0.08 -0.10 0.68

^ 11.68 19.47 7.57 1.63 14.36

-0.40 0.31 0.34 0.10 0.96

Brazil Chile Mexico Peru

0.28 0.54 0.76 0.15

0.05 0.49 0.46 0.04

0.10 0.10 0.24 0.05

-0.00 -0.00 -0.09 -0.00

3.61 0.00 42.44 13.28

0.40 -0.00 -0.25 0.02

Czech Republic Hungary Poland

0.17 0.19 0.27

-0.26 0.16 0.12

0.10 0.05 0.15

0.01 -0.01 -0.08

0.84 3.05 2.23

-1.04 -0.30 -0.14

South Africa

0.23

0.09

0.28

0.80

15.71

3.89

£

£

^rj = evaluates the contribution to the standard deviation; e = evaluates the contribution to the mean.

4

Robustness Checks

In this section, we conduct several robustness checks on our baseline results to see whether they stand up to alternate specifications and whether the estimated relationship has undergone significant changes. There are several potential sources of instability to the estimates. First, our baseline estimates were based on absolute inflation rate and exchange rate changes rather than on the deviations of these variables from the target or the trend, as might be the case in other models. Second, there are known sources of instability, given that a number of countries moved to inflation targeting or significantly changed their emphasis on inflation and the exchange rate during the past three to four years. However, a specific problem in our case is the short sample size and the very recent experience of inflation targeting of these countries, reducing the power of the usual stability tests. Third, the baseline estimates were obtained in the context of a reactive interest rule whereby central banks respond to observed rather than expected variables. Do the results remain valid in the context of a forward-looking policy setting?

MONETARY POLICY RULES IN EMERGING MARKET ECONOMIES

4.1

229

Sensitiveness to Measurement of Variables

As a first test, we estimated the baseline model by replacing the absolute values of inflation and exchange rate changes by their respective deviation from the trend. Given the short experience with inflation targeting in emerging economies, the announced inflation targets could not be used in the estimates of the gap model. We instead used an Hodrick-Prescott measure of trend inflation, which fits target inflation rather well because such filtering allows a trend drift in long-run inflation, which the actual inflation targets seem to follow. The results are given in Table 9. As may be seen from the table, there are few major changes to the signs and magnitude of the parameters, although the explanatory power of the model and the statistical significance of individual parameters declined in the case of some countries. With the notable exception of Mexico and South Africa, interest rate response to inflation and output declined in the gap model compared to the baseline model. It is important, however, to note that the results validated the finding of the baseline model with respect to the exchange rate. The coefficient on the current period exchange rate is negative in all countries excepting Chile, and significant in the majority of them. As in the baseline model, the results confirm the importance of exchange rate shocks and central banks' response to them in a number of countries.

4.2

Stability of Estimates

Our short sample size imposes major constraints on conducting satisfactory tests of parameter and model stability. Notwithstanding this caveat, we conducted Chow breakpoint tests for each country assuming that the date of a break to the relationship, if any, is the start of a new monetary era in 1999, after the outbreak of the Asian crises. Although such a breakpoint might look arbitrary for all countries, especially Latin America and Central Europe, there appears to be a general consensus that emphasis on inflation control improved significantly around this time. With the exception of Chile, the Czech Republic, Hungary, India, Peru, South Africa, and Taiwan there is evidence of shifts in the estimated equations in many countries (Table 10). These results are not surprising, as most of the countries falling under this category are in Asia or Central Europe, which have undergone big monetary regime changes. The Chow breakpoint test, however, assumes that the estimated relationship is stable up to a specific point, at which it might make a discrete change to a new but thereafter stable structure. As this point is unknown, in most cases the advantages of the Chow test diminishes. Therefore, a more appropriate technique, namely, a recursive coefficient estimates test was introduced where the coefficients

230

M . S . MOHANTY AND M A R C K L A U

Table P.Taylor Reaction Function (Gap Model)^ India Korea Philippines Taiwan Thailand Brazil Chile Mexico Peru Czech Republic Hungary Poland South Africa

Const

Ttgapf

yf

/^rgapf

Axrgap^^i

if_i

R^

BG-LM

139 (2.07) 2.40 (2.26) 6.60 (3.75) Oil (0.42) 1.12 (1.27)

016 (2.46) 015 (0.45) 0.53 (2.05) 015 (1.08) 0.10 (031)

0.16^ (3.46) 0.19 (1.36) 0.94 (2.05) Oil (1.89) 0.34 (2.11)

-0.09 (-3.13) -0.26 (-5.70) -0.30 (-2.94) -0.04 (-1.21) -0.09 (-1.98)

-0.09 (-1.93) -0.06 (-1.44) -0.09 (-1.06) 0.06 (1.60) -0.10 (-1.09)

072 (7.13) 0.73 (6.27) 0.43 (2.98) 0.94 (17.37) 0.86 (7.40)

088

OH

0.80

1.14

0.43

0.87

0.88

2.42

0.82

2.00

4.77 (1.31) 3.72 (2.72) 7.97 (3.78) 1.83 (1.03)

0.08 (3.27) 0.93 (1.20) 1.09 (5.43) 0.33 (1.84)

1.04 (1.00) 0.14 (0.49) 0.40 (1.11) 017 (1.26)

-0.34 (-2.02) 0.42 (3.36) -0.60 (-4.67) -0.15 (-1.79)

0.24 (1.20) -0.16^ (-0.67) -0.23 (-2.04) -0.22 (-3.94)

0.73 (5.68) 0.43^ (3.61) 0.56 (5.92) 0.92 (16.38)

0.79

2.08

0.62

0.69

0.82

0.99

0.95

0.53

0.86 (1.79) 0.40 (0.35) 1.75 (1.70)

0.12 (1.27) 0.06 (0.41) 0.32 (2.60)

0.32 (3.42) 0.19 (0.47) 0.46 (2.67)

-0.04 (-0.73) -0.21 (-1.95) -0.06 (-1.35)

0.07 (0.65) 0.06 (0.76) -0.02 (-0.52)

0.84 (13.76) 0.93 (11.86) 0.87 (14.92)

0.95

0.57

0.97

2.59

0.95

1.03

1.04 (0.89)

0.31 (3.42)

0.28 (1.40)

-0.08 (^.61)

-0.08 (-3.96)

0.93 (11.20)

081

0.37

^t-statistics in parentheses, ngap = consumer price gap (for India, wholesale prices); y = output gap; Axrgap = change in the real effective exchange rate gap (an increase means appreciation); / = interest rate (dependant variable); BG-LM = Breusch-Godfrey Serial Correlation LM test (the F-statistics are given). — ^Industrial production. — ^t-4.

of each parameter specified in (4) were re-estimated recursively and the plots of the coefficients were explored. The results do not reveal obvious in-sample parameter instability for many countries. However, as shown in Figure 2, the coefficients for Brazil, Mexico, and Korea display some instability towards the end of the 1990s. These results should be interpreted with caution, as the time frame is quite short. Using a recursive residuals test as a third stability test, with the exception of Brazil and Mexico, the residuals did not go outside the standard error bands for any of the other countries, suggesting that the parameters are stable (Figure A2 in the appendix).

MONETARY POLICY RULES IN EMERGING MARKET ECONOMIES

231

Table 10: In-Sample Chow Breakpoint Stability Test^ Country

F-Statistic

Country

F-Statistic

India Korea Philippines Taiwan Thailand South Africa Brazil

0.53 2.69* 2.99* 1.91 2.19* 1.62 18.48*

Chile Mexico Peru Czech Republic Hungary Poland

0.56 5.05* 1.07 1.12 1.39 3.95*

^The Chow breakpoint tests have been done for 1999:1 (except for Mexico 1998:1, Poland 1998:1, and Thailand 1997:3). — *lndicates that the null hypothesis of no-structural break has been rejected.

Figure 2: Recursive Estimates for Selected Countries and Coefficients^ Brazil Inflation

Exchange rate

Inflation

Exchange rate

Interest rate

Mexico Interest rate

HO.25

Korea Inflation

10

Exchange rate

Interest rate

97 98 99 00 01 02

97 98 99 00 01 02

R.A 10

t''J

-6

1

1

I

L

97 98 99 00 01 02

-10

' Recursive estimates ' 2 standard deviations

^Based on the specification of Table 6.

232

M.S. MOHANTY AND M A R C K L A U

4.3

Sensitive to Different Estimation Methods

As a last step, we estimated the reaction function with expected rather than actual variables, assuming that monetary policy is preemptive rather than reactive. We used monthly data starting from 1998 for the 13 countries. Shortening the sample to the most recent period provides a further test on the structural changes in the relationship and thus supplements the stability test above. Wherever applicable this also allowed us to use the actual inflation target announced by the central bank. The model was estimated through the generalized method of moments (GMM), with the lagged short-term interest rates and current and lagged values of money supply growth, export growth, the exchange rate and the output gap as the relevant instruments. The lag length varies across countries, depending on the underlying dynamics. We chose an appropriate weighting matrix that would satisfy the orthogonality condition of zero correlation between parameter estimates and instruments^^ and is also robust to heteroskedasticity and autocorrelation of unknown form. To check overidentification conditions. Table 11 presents the relevant p-values of J-statistics. As can be seen from the table, for most countries the null of overidentification is rejected at over the 90 per cent confidence level. The results of the GMM estimates are reported in the annex (Table Al). Again, it is important to note that the forward-looking reaction function confirms most findings of the simple function, indicating the robustness of the model for Table 11: Overidentifying Restriction Test^ Country

p-value

Country

p-value

India Korea Philippines Taiwan Thailand South Africa Brazil

0.990* 0.892* 0.968* 0.936* 0.992* 0.987* 0.981*

Chile Mexico Peru Czech Republic Hungary Poland

0.865* 0.966* 0.993* 0.943* 0.900* 0.970*

^Based on the GMM inflation rate specification (for the Czech Republic and South Africa, the inflation gap specification). — *Indicates that the null hypothesis of overidentifying restrictions using the J-statistics has been rejected.

^^ The model is first estimated using two-stage least squares to generate the initial covariance matrix for subsequent iterations to find the optimal weighting matrix.

MONETARY POLICY RULES IN EMERGING MARKET ECONOMIES

233

emerging economies. There are few sign reversals when we used expected rather than actual variables. The size of reaction coefficients, however, underwent some change: the estimated interest rate response to inflation and output declined significantly in Asia and Central Europe (excepting Poland with respect to inflation) when using expected variables. By contrast, the reactions in Latin America to inflation and output generally increased. This, however, is not true for the exchange rate response, which is substantially higher, particularly in Korea, the Philippines, Taiwan, Brazil, Peru and all Central European countries.

5

Testing for Non-Linear and Asymmetric Reactions

Given the constraints imposed by the linear structure of the model and the quadratic loss function (as implied by (1) to (3)), the estimated reaction function assumed a linear and symmetric response by central banks to inflation and output shocks. Specifically, the estimated reaction coefficients imply that central banks give similar weights to positive and negative price pressures as well as economic upswings and downswings and that their reaction coefficients do not vary with the size of the shocks. However, if these assumptions do not hold, the reaction function is misspecified. In this section, we therefore test whether central banks' response depends on the size and sign of inflation and output deviations. Do central banks respond differently to a negative as against a positive deviation of inflation from the target? Do larger shocks imply a stronger response than smaller shocks? The literature has identified two potential sources of asymmetric and nonlinear monetary policy responses. First, asymmetry may be induced by a nonlinear relationship between inflation and output in the presence of significant price and wage rigidities. For example, nominal wages may be sticky downwards but flexible upwards, producing a convex Phillips curve.^^ While a positive inflation deviation leads to a faster rate of increase in prices in the subsequent periods because of upward wage flexibility, the effects of a negative inflation deviation will be dampened by the downward rigidities of wages. In such circumstances, Dolado et al. (2002) show that an optimal policy rule involves some degree of asymmetric reaction to offset the non-linear Phillips curve. As a result, the central bank may penalize positive deviations of inflation more severely than negative deviations, and the response may depend on the size of the shocks.^^ ^* Akerlof et al. (1996) point out that this is likely when inflation is low and workers resist nominal wage cuts in response to higher unemployment. ^^ Gerlach (2000) provides evidence of asymmetric monetary policy response in the context of industrial countries. Dolado et al. (2002) confirm this result for Europe

234

M.S. MOHANTY AND MARC KLAU

A second source of asymmetric monetary responses may arise when central banks' loss function is non-linear with respect to the sign and size of inflation and output deviations. Some have argued that, to protect their credibility, independent central banks are likely to be biased towards undershooting rather than overshooting their inflation targets, giving rise to a deflationary bias in monetary policy. Goodhart (1999) cites two, possibly offsetting, asymmetries in central bank policy. While central banks have an asymmetric bias towards tightening, faced with multiplicative uncertainty, they tend to delay tightening decisions longer than easing decisions. Others argue that political accountability and uncertainty about future economic developments might encourage central banks to adopt a policy of greater aversion to recession than expansion (Blinder 1998; Cukierman 1999).^^ To test whether interest rate responses change with the sign of inflation deviations, we included a slope dummy in (4), with the dummy taking a value of 1 in the quarters in which inflation was below the average during the sample period. For these quarters, central banks' reaction to inflation is given by the combined effect of the usual inflation coefficient and the slope dummy coefficient. A negative and significant coefficient implies a weaker response to a negative inflation deviation, while a positive coefficient indicates the opposite. The results are reported in Table 12. The dummy is strongly or weakly significant in Chile, the Czech Republic, Korea, Poland and Thailand. Of these, Poland, with a positive coefficient, appears to be an outlier, indicating a stronger central bank response to a negative inflation shock than to a positive shock. In other countries, the dummy coefficient is negative: it is as high as the coefficient on inflation in Korea and Thailand, suggesting that monetary policy has remained neutral during periods when inflation fell below its average value. In Chile the coefficients suggest that interest rate responses to negative inflation deviations were about two-thirds lower than the responses to positive deviations. To test for size asymmetry, (4) was augmented by the squared deviations of inflation and the output gap:^"^ (5)

it =5o +5i7rf +S2yt -^S^Axn +54A^r^_i +55/^-1 •^S^int)'^+5'j(yt)^.

Testing for symmetry implies testing the parameter constraints 56=57=0, using the Wald test of joint significance of the parameters. Acceptance of the and find both a sign and size asymmetry in central banks* reaction to inflation and output shocks. ^^ The model of Cukierman (1999) generates an inflation bias in monetary policy similar to that of the Barro-Gordon type without the condition that the central bank aims to achieve an output level in excess of the economy's potential. ^^ See Surico (2003) for the formal derivation of the test.

MONETARY POLICY RULES IN EMERGING MARKET ECONOMIES

235

Table 12: Tests for Asymmetric Response to Positive and Negative Inflation Deviations^ Const India Korea Philippines Taiwan Thailand Brazil Chile Mexico Peru Czech Republic Hungary Poland South Africa

^t

yt

Axr^

AJ^^/-I

^'r-l

dumXTtt

1.00 (1.38) 2.50 (1.67) 3.89 (2.30) 0.50 (1.31) 1.84 (2.44)

0.14 (2.37) 0.61 (1.82) 0.44 (2.95) 0.24 (1.81) 0.62 (2.32)

0.12^ (2.72) 0.36 (2.30) 0.44 (0.84) 0.16 (1.74) 0.35 (2.55)

-0.09 (-2.57) -0.20 (-4.26) -0.14 (-1.67) -0.04 (-0.81) -0.11 (-3.39)

-0.09 (-1.77) -0.12 (-2.50) 0.02 (0.34) 0.02 (0.42) -0.21 (-2.63)

0.70 (7.29) 0.45 (2.83) 0.22 (1.55) 0.83 (9.03) 0.52 (3.21)

0.02 (0.28) -0.63 (-1.69) 0.12 (0.57) -0.11 (-0.51) -0.62 (-1.70)

5.51 (1.26) 2.34 (1.36) 0.62 (0.33) 2.36 (1.14)

0.60 (3.16) 0.76 (3.08) 0.55 (4.53) 0.17 (1.56)

0.61 (0.63) 0.51 (1.83) 0.64 (1.58) 0.15 (1.06)

-0.38 (-2.28) 0.36 (2.54) -0.61 (-3.01) -0.15 (-2.00)

0.21 (1.46) -0.38C (-2.18) -0.15 (-1.93) -0.23 (-3.98)

0.49 (3.49) 0.35c (3.32) 0.51 (3.84) 0.87 (10.22)

-0.14 (-0.44) -0.50 (-1.79) 0.15 (1.03) -0.07 (-0.33)

2.20 (2.01) 2.20 (2.05) 2.17 (2.27)

0.48 (4.20) 0.18 (1.56) 0.27 (5.87)

0.09 (0.90) 0.29^ (0.86) 0.38 (2.92)

-0.08 (-1.72) -0.20 (-2.01) -0.08 (-1.57)

-0.01 (-0.07) -0.01 (-0.14) 0.01 (0.28)

0.46 (3.14) 0.72 (5.28) 0.63 (8.54)

-0.31 (-1.99) -0.07 (-1.44) 0.17 (4.34)

-0.13 (-0.08)

0.05 (0.68)

0.06^ (0.27)

-0.07 (-2.76)

-0.06 (-2.47)

0.97 (11.06)

-0.04 (-0.58)

^t-statistics in parentheses. ;r= annual percentage change in consumer prices (for India, wholesale prices); y = output gap; Axr = change in the real effective exchange rate (an increase means appreciation); / = interest rate (dependant variable); dum = inflation dummy ( = 1 when inflation rate minus average inflation is negative). — ^'Industrial production. — ^t-4. — ^t-l. constraint implies accepting the hypothesis that central banks do not view a larger shock differently from a smaller shock when changing their interest rates. The results are given in Table 13 with the probability value of the Wald test statistics. Excepting for the Czech Republic, Brazil, Peru, and Thailand, the pvalues of the Wald test failed to reject the constraints of zero coefficients on squared inflation and output gaps, suggesting no significant difference in re-

236

M.S. MOHANTY AND MARC KLAU

Table 13: Tests of Dependence of Response to the Size of Inflation and Output Shocks^ Const India Korea Philippines Taiwan Thailand Brazil Chile Mexico Peru Czech Republic Hungary Poland South Africa

^/

1.67 -0.20 (2.32) (-0.90) 0.66 0.75 (1.62) (0.55) 3.14 1.05 (1.22) (1.60) -0.13 0.53 (1.23) (-0.94) 1.04 0.52 (1.06) (1.17)

yt

0.10^ (1.90) 0.43 (2.25) 0.10 (0.38) 0.22 (2.98) 0.39 (2.49)

Axr,

Axr,_i

-0.09 -0.11 (-3.27) (-2.20) -0.12 -0.20 (-3.93) (-2.00) 0.06 -0.13 (1.25) (-1.79) 0.03 -0.03 (-0.88) (1.17) -0.16 -0.10 (-2.76) (-1.89)

h-i

Tt^

y'.

-0.02 0.02 0.76 (7.04) (1.48) (-0.44) -0.01 0.03 0.49 (0.43) (3.08) (-0.09) -0.04 0.21 0.38 (1.87) (-0.68) (2.17) -0.05 0.07 0.87 (2.45) (-1.60) (2.78) -0.03 -0.01 0.68 (2.93) (-0.14) (-0.92)

3.83 (1.21) -0.58 (-0.30) 2.84 (1.25) 3.68 (1.94)

0.64 (3.03) 1.42 (2.08) 0.68 (2.63) 0.85 (3.34)

-0.27 0.54 -0.00 0.22 -0.39 0.70 (1.42) (4.17) (-2.66) (-0.61) (0.74) (-2.19) -0.34c 0.33c -0.03 0.34 -0.15 0.38 (3.73) (-0.63) (-2.29) (1.44) (2.88) (-2.43) -0.12 0.00 -0.57 -0.19 0.38 0.94 (0.02) (-1.36) (2.81) (2.35) (-2.36) (-2.86) -0.21 0.78 -0.05 -0.03 -0.17 0.09 (0.75) (-2.24) (-3.90) (10.30) (-2.64) (-0.45)

1.02 (1.48) 1.54 (1.34) 1.75 (1.76)

0.15 (0.65) 0.21 (1.72) 0.39 (2.87)

0.04 -0.09 0.25 (0.65) (2.38) (-1.22) -0.01 0.18^ -0.19 (0.59) (-1.82) (-0.18) -0.01 0.54 -0.08 (3.65) (-1.35) (-0.11)

1.51 (1.04)

0.42 -0.04^ -0.07 (2.69) (-0.14) (^•74)

-0.06 (-2.87)

-0.01 0.67 0.01 (0.82) (-0.53) (7.39) 0.00 0.04 0.70 (0.14) (5.27) (0.05) 0.04 -0.00 0.69 (0.51) (7.51) (-1.75) 0.91 (10.38)

-0.01 (-1.11)

0.06 (0.27)

Wald test (p-value) 0.107* 0.911* 0.089* 0.636* 0.004 0.029 0.062* 0.293* 0.025 O.OIO 0.988* 0.109*

0.518*

*t-statistics in parentheses. *Indicates that the Wald test failed to reject the null hypothesis of parameter of squared inflation = parameter of squared output gap = 0. ;r= annual percentage change in consumer prices (for India, wholesale prices); y = output gap; Axr = change in the real effective exchange rate (an increase means appreciation); / = interest rate (dependant variable). — I'Industrial production. — ^t-4. — ^t-1.

sponse to the size of shock. Even for these four countries the coefficients on the squared inflation and output term are mostly negative, which is not economically meaningful. Including the squared inflation and output terms also led to some odd sign changes of other parameters in India, Taiwan, and South Africa.

MONETARY POLICY RULES IN EMERGING MARKET ECONOMIES

6

237

Summary and Conclusions

As set out in the introduction, one of the objectives of this paper was to review the central banks' interest rate setting behaviour in emerging market economies and collect evidence on the consistency of their monetary policy decisions. For this purpose the paper focussed on the experience of 13 major emerging market economies following a movement to independent monetary policy. Against the backdrop of a series of emerging market financial crises during the last decade, the debate has increasingly shifted to building monetary policy credibility. In this context, suggestions such as subjecting central banks to "constrained discretion" and increasing the predictability of central bank policy actions have assumed importance. Monetary reforms in emerging economies in recent years appear to have moved in this direction. As countries have adopted inflation targeting, there have been attempts to improve policy transparency. In some countries, central banks have also announced guidelines for setting interest rates. At the same time, while monetary policy has increasingly been focussed on price stability, other objectives have remained significant as well. One such objective highlighted in the paper is the stabilization of the exchange rate. Episodes of recent exchange market pressures reviewed in the paper reveal that central banks often intervene to stabilize the exchange rate; in some countries, this may have at times dominated interest rate developments. The empirical evidence presented in the paper confirms that central banks' rate setting behaviour is consistent. In many countries, the interest rate reaction to inflation exceeds one, suggesting a non-accommodating stance of monetary policy towards price pressures. However, in most countries the interest rate response to exchange rate movements is also strong, with exchange rate volatility significantly contributing to interest rate volatility. This reflects joint effects of two factors: first, exchange rate shocks appear to be more persistent; and second, central banks' preference for stabilizing the exchange rate by using monetary policy instrument remains strong. Evidence presented in the paper also suggests that in some countries central banks may respond more aggressively to positive than to negative inflation deviations. There is, however, little evidence that interest rate response depends on the size of the inflation and output shocks.

238

M.S. MOHANTY AND MARC KLAU

Appendix Table Al: Reaction Functions from GMM Estimates^ Const

n,

/,_,

R^

-O.OS'^(-4.02)-0.06'^(-4.95) 0.05 (2.69)-0.10 (-5.15) -0.15 (-3.63)-0.12 (-^.76) -0.06 (-1.86)-0.05 (-2.33) -0.19 (-6.70) 0.09 (4.74)

0.85(22.39) 0.87(78.25) 0.53(10.93) 0.89(23.39) 0.70(52.41)

0.86 0.91 0.38 0.92 0.92

(1.84) -0.10 (-1.34)-0.12 (-1.86) (6.89) 0.21 (2.28)-0.22 (-3.10) (4.31) -0.23 (-2.62) 0.13 (2.42) (1.53) -0.32^(-7.00) -0.11^(-5.48)

0.85(44.15) 0.75(10.53) 0.78(11.15) 0.95 (9.68)

0.91 0.63 0.92 0.97

0.21 (1.75) 0.08 (0.92) 0.69 (1.96)

0.11 (5.21) 0.03 (3.07) -0.15 (-7.95) 0.10 (5.65) 0.92(50.27) 0.02 (2.09) 0.01^ (1.66) -0.12 (-3.43) 0.02 (1.45) 0.97(75.11) 0.34 (7.99) 0.07 (3.26) -O.04 (-1.12) -0.04 (-2.89) 0.81 (19.35)

0.98 0.98 0.96

0.86 (4.52)

0.04 (7.09) 0.07

0.88

India Korea Philippines Taiwan Thailand

0.85 (3.73) 0.48 (6.14) 2.23 (6.48) 0.33 (1.81) 0.60(10.76)

0.04 0.04 0.37 0.10 0.09

(2.06) (1.23) (9.24) (2.60) (3.67)

Brazil Chile Mexico Peru

1.36(2.92) 2.20 (4.10) 1.15 (2.90) 0.31 (1.49)

0.16(19.80) 1.15 (4.21) 0.20 (2.48) 0.22 (6.98)

Czech Republic Hungary Poland South Africa

y, O.OS*''^ (8.01) 0.03 (3.76) 0.03 (1.70) 0.03 (3.40) 0.01 (0.63) 0.17 0.48 0.42 0.01*^

djcr,

(7.52) -0.04 (-5.84)

Axr,,

0.03

(7.18) 0.88(76.41)

^-statistics in parentheses, n = annual percentage change in consumer prices (for India, wholesale prices, for the Czech Republic and South Africa, the inflation gap); y = industrial production gap; Axr = change in the real effective exchange rate (an increase means appreciation); / = interest rate (dependant variable). List of instruments: annual percentage change in consumer prices; industrial production gap; change in real effective exchange rate; annual percentage change in broad money; annual percentage change in exports. — ^Industrial production. — ^ / - 1 .

MONETARY POLICY RULES IN EMERGING MARKET ECONOMIES

Figure Al: Actual versus Calculated Interest Rates (from the Simple Reaction Function Specification),^ 1995-2002 Chile

Brazil

Hungary

r'v

_

[

-

V

*^vA ^».

\-

^ NT'/^*^

J

1

1

1

1

1

1

South Africa

97 ^ — —

Actual rate Fitted rate

99

01

-

^