WORLD BANK LATIN AMERICAN AND CARIBBEAN STUDIES Ti a~poialts work in progress 19607 for public discussion June 1999 Adjustments after Speculative Attacks in Latin America and Asia: A Tale of Two Regions? G(zille/e;o E Perrrv aid Dainiel LelerinM WORLD BANK LATIN AMERICAN AND CARIBBEAN STUDIES Viewpoints Adjustments after Speculative Attacks in Latin America and Asia: A Tale of Two Regions? Guillermo E. Perry and Daniel Lederman Office of the Chief Economistfor Latin America and the Caribbean The World Bank Washington, D.C. (C 1999 The International Bank for Reconstruction and Development / THE WORLD BANK 1818 H Street, N.W Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First printing June 1999 The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. The World Bank does not guarantee the accuracy of the data included in this publication and accepts no responsibility whatsoever for any consequence of their use. The boundaries, colors, denominations, and other information shown on any map in this volume do no imply on the part of the World Bank Group any judgment on the legal status of any territory or the endorsement or acceptance of such boundaries. The material in this publication is copyrighted. Requests for permission to reproduce portions of it should be sent to the Office of the Publisher at the address shown in the copyright notice above. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to copy portions for classroom use is granted through the Copyright Clearance Center, Inc., Suite 910, 222 Rosewood Drive, Danvers, Massachusetts 01923, U.S.A. The complete backlist of publications from the World Bank is shown in the annual Index of Publications, which contains an alphabetical title list (with full ordering information) and indexes of subjects, authors, and countries and regions. The latest edition is available free of charge from the Distribution Unit, Office of the Publisher, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A., or from Publications, The World Bank, 66, avenue d'Iena, 75116 Paris, France. Guillermo Perry is the Chief Economist for Latin America and the Caribbean Region, the World Bank. Daniel Lederman is Economist in the Office of the Chief Economist for Latin America and the Caribbean Region, the World Bank. Perry, Guillermo. Adjustments after speculative attacks in Latin America and Asia: a talc of two rcgions? / Guillermo E. Perry, Daniel Lederman. p. cm. - (World Bank Latin American and Caribbean studies. Viewpoints) Includes bibliographical references (p. ). ISBN 0-8213-4541-9 1. Structural adjustment (Economic policy) -Latin America. 2. Structural adjustment (Economic policy) - Economic conditions - 1982- 4. Asia - Economic conditions - 1945- 5. Recessions - Latin America. 6. Recessions - Asia. I. Lederman, Daniel, 1968- . II. Series. HC125.P453 1999 338.9-dc2l 99-34548 CIP 8 ~~~~~~~~S~DNThEL1Th .XSISA'IVNV NOIIVhIHHOD-}NVH JAIISUDDfIS V ' XICNIddV .. SNOIIVIDZdXa ThVd -ThDNVH:DX1 HLIIAI JCaOW NVISINA[Nk XOOLXJI V ONISfl NOIJVHISI7YHI NV :SVWWHflJU ADI1Od -OHDVW CNV SNDVILV 1AJIV'If1DJdS V XiaNddV . SNOIIVDIIdWI ADI-1Od CNV SNOISflYIDNOD A 9z ....... AuiS 11odxo pu? uoi t31ui IrUOIX3I jo Qlo.' 3L ................. ...... SUOIIFUtLA IEJ-3SUeL?JJX3 JO SID; Jj;) pUE SDLUUUkp U?IUt .................................. Uddutj 1i pip AiM t;ol Auld osuodsQi iiodxo M1O PIG 61 ...... ,,arol E AtJld purtLupjo uoiiisodnuo: WIviiu! 3tp PCI Li ........................................................... . lo Auld aim mj s 1 iurug lE lui Ztl pI Abilod puE si3mnssaid oAlp1?fJds jo ladeUi ;ip p3Iijdaw suoilipuoOD jlilUI ;1snED;aq t s11 bl. sDoilod luosi ................... . ......... -i 3uousoA ssuodsi Ajod IIDsg pum AJriouoLU QsfleD;)q ii suJ 8 .. . . . . . . . . . sDnss_ld oAmi1e1lds jo sJ_ns7e3aJ .g....... 3AuoXas S 1sfultupsnfpu Iuaia?3 31 lO sainssoid 3Ai1efnDods 311 3snr3oq it s?A 8 .. YVISV Nl IIA O'I S STIIIHAODOTh TIJHI CINV H dEII S SE ISIJ HI aL NIA\ AH A Al s . VISV NI YIAl&O 'IS SJIRHJAODTd JHI aNv dJdja a EI[1tAS SaSuJD IIFJI :SHSIHSD NVISV CINV NVDOITIWV NIJV'I HH I JO HIVWl JV 1 HHI H1II .....IN. ONODI IIVJd JIHI OINO NDVLLV ZIAIIV'IflDZIdS V JO NOISSIWSNVdI OLI S'IiNNVHD 'II !!A .......................................... SI N W9 a T4a ON > 9V S IN3INOD !!U iv NOTES ........................................................ 40 LIST OF TABLES Table 1. Sacrifice Ratios .10 Table 2. Fiscal Policy Before and After the Crises .15 Table 3. Fiscal Policy After the Crises .15 Table 4. Initial Financial Structure .17 Table 5. Decomposition of GDP Growth Rates after Speculative Attacks .20 Table 6. Intraregional Trade in the Americas .28 Table 7. Intraregional Trade in Asia .28 Table 8. Average Export Similarity Indexes .29 Table 9. Summary of Policy Recommendations .32 Table B1. Country Rankings .37 LIST OF FIGURES Figure la. Average Quarterly GDP Growth Rate .6 Figure lb. Year-on-Year Quarterly GDP Growth Rate .6 Figure 2. Decline in Growth of Industrial Production .7 Figure 3a. Index of Speculative Pressure 1 .9 Figure 3b. Index of Speculative Pressure 2 .9 Figure 4a. ISP1 and Economic Decline Rankings .11 Figure 4b. Change in CAB/Y and Economic Decline Rankings .11 Figure 5. Real Ml .12 Figure 6. Real Domestic Credit .12 Figure 7. Change in Real Deposit Rate Relative to the Level Before the Crises .13 Figure 8. Change in Real Money Market Rate Relative to the Level Before the Crises . 13 Figure 9. Tight Money and Economic Decline Rankings .14 Figure 10. Fiscal Impulse and Economic Decline Rankings .16 Figure 1 la. Financial Structure and Economic Decline Rankings .18 Figure 1 lb. Ex-ante Credit Growth and Economic Decline Rankings .19 Figure 12. Ex-ante GDI Rate and Economic Decline Rankings .20 Figure 13. Index of Exports Volume .21 Figure 14. (Low) Export Volume Growth and Economic Decline Rankings .22 Figure 15a-f. Recursive Inflation Drift and Inertia (Argentina, Brazil, Mexico, Indonesia, Korea, Thailand) ................................................................ 24-25 Figure 16. Index of the "Success" of Nominal Exchange-Rate Variations .............................. 26 Figure 17. Index of Imports Volume ................................................................ 27 Figure 18. Japan: Nominal Exchange Rate, and Growth Rates of Import Volume and Industrial Production ................................................................ 28 Figure 19. Intraregional Trade in Latin America, 1990-1996 ................................................... 32 Figure Al. An Illustration of Policy Dilemmas in a Textbook Keynesian Model with Exchange-rate Expectations ................................................................ 34 A TALE OF TWO REGIONS? * V ABSTRACT The paper analyzes the adjustment process in the paper concludes that the large size of short-term aftermath of speculative attacks against the cur- external debt relative to GDP, the higher inci- rencies of six countries-Argentina, Brazil, Mex- dence of leverage and currency mismatches, the ico, Indonesia, Korea, and Thailand. Even though higher rates of investment, plus the regional char- the Mexican and Argentine crises of 1995, and the acter of the Asian crisis and the high export simi- Brazilian adjustment after October 1997, were larity across the Asian economies contributed to certainly costly, the Asian crises of 1997 were the deeper economic downturn. However, with deeper and the recovery of the real economy was rising regional trade and financial development, slower. The paper relies on rank-correlation the aftermath of future speculative attacks in Latin analysis, and economic indexing and growth America may look more like the recent East Asian decomposition techniques to evaluate the tales of adjustment. The paper draws policy impli- explanatory power of several hypotheses of why cations for reducing the costs of the macroeco- the recessions in Asia were more severe. The nomic adjustment after currency crises. A TALE OF Two REGIONS? * vii ACKNOWLEDGMENTS We are indebted to Pablo Fajnzylber, Depart- research assistance. We received invaluable com- ment of Economics at the Federal University ments and suggestions from Joe Stiglitz, Luis of Minas Gerais in Belo Horizonte, Brazil. Serven, Eliana Cardoso, Bill Maloney, Jose He made substantial contributions to a related Gonzalez, Sergio Schmukler, Neile Quintero, presentation delivered by G. Perry at the 1998 and numerous participants in the World Bank Meeting of the Latin American Econometric Latin American and Carribean Regional Studies Society in Lima, Peru. We are also grateful Seminar. The opinions expressed in this docu- to Alejandro Guerson and Ana Maria ment do not necessarily represent the opinions Menendez for substantial comments and of the World Bank. 1 NUMEROUS ARTICLES HAVE BEEN WRITTEN about the causes of financial crises in emerging markets, especially the Mexican and Argentine crises of 1995 and the East Asian crises of 1997. Much less attention has been devoted to the effectiveness of alternative policy responses and the ensuing process of recovery of the real economy, although Radelet and Sachs (1998), Furman and Stiglitz (1998), Krugman (1998b), and Corden (1998) reflect an emerging uneasiness with contractionary macroeconomic policy responses to the Asian crises. This paper analyzes the adjustment process in the example of a "successful" defense of the currency aftermath of speculative attacks in six countries- (Argentina 1995), while another (Brazil after Argentina, Brazil, Mexico, Indonesia, Korea, and October 1997) is an example of a temporarily suc- Thailand. As implied by the title, the main ques- cessful defense followed by an incomplete adjust- tion to be addressed is whether the tales of adjust- ment program. The others experienced dramatic ment in these Latin American and Asian currency devaluations. This small sample of economies were similar. This comparison is inter- episodes of adjustment also offers variety in the estlilg for several reasons. The six countries came magnitude of the ensuing economic decline. under the aegis of adjustment programs supported Even though the Mexican and Argentine crises by international financial institutions, and the of 1995, and even the Brazilian adjustment after associated policy prescriptions have been at the the October 1997 attack against its currency, center of attention .2 Of the six cases, one is an were certainly costly, the Asian crises have been 2 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: deeper and the recovery of the real economy has contained devaluation), were less effective in Asia been slower. than in Latin America in terms of rebuilding There are several plausible explanations of investor confidence and bringing a speedy recov- the different depths and speed of recovery in Latin ery of the real economy. With rising regional trade America and East Asia. We conclude that some integration and rising financial integration and initial conditions, such as the larger size of short- development in Latin America, it is possible that term external debt relative to Gross Domestic the aftermath of future speculative attacks in the Product (GDP), the higher incidence of leverage region will look more like the recent East Asian and currency mismatches, the higher rates of experiences. The policy implications are that investment (which tend to fall to a greater extent sound institutions for maintaining a healthy finan- than consumption during crises), plus the regional cial system (i.e., that prevent excessive credit character of the Asian crises (including the reces- growth, asset-price bubbles, and currency mis- sion and financial crisis in Japan), and the higher matches) and to improve corporate governance, similarity of their export structures contributed to prudent debt (and exchange rate) management, the deeper economic downturn and the relatively and establishing liquidity cushions are necessary slower recovery in Asia. These factors may also both to diminish the likelihood of future specula- explain why traditional recipes for recovering con- tive attacks and to reduce the costs of such crises in fidence, such as tight monetary policies (and/or a the future. 3 NNNE |of TRANSMISSION SPE TIVE ATTACK TO AL ECONOMY IN THEORY, CURRENCY DEVALUATIONS NEED NOT be contractionary, although the historical experience in Latin America seems to indicate that they are contractionary in the short run.3 In practice, speculative attacks and devaluations are often associated with increases in nomi- nal and real interest rates in the short run. The increase in the costs of borrowing dampen invest- ment and interest-sensitive consumption, and may have additional adverse effects through its effect on the domestic financial sector. Calvo (1998b) identifies two channels through take productive investments. A speculative attack which the real economy can be affected. The first against a currency could have limited effects on channel is a Keynesian effect, where contractions of domestic nominal interest rates, especially when aggregate demand in the presence of sticky prices the authorities do not defend the exchange rate. are translated into declines of output and employ- However, when the prices of nontradables fall, the ment. The other channel is the "Fisherian" chan- ex-post real interest rate rises, rendering investment nel, alluding to the early contributions of Irving unprofitable. A further aggravating consideration is Fisher to the theory of interest. The central tenet of that the Fisherian channel affects the financial sec- Fisher's theory was that the "rate of return over tor, which may lead to declines in credit availability cost" of capital (what Keynes called the "marginal or even credit crunches.4 efficiency of capital") has to be higher than the A currency devaluation can ameliorate the interest rate in order for economic agents to under- Keynesian contractionary effects by improving the 4 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: competitiveness of domestic tradable goods and extent interest rate increases result in financial dis- services. It may reduce Fisherian effects by raising tress and credit crunches. the marginal product of investmnents and reducing The effects of devaluations and interest rate real interest rates through its effect on the domes- hikes do not end with these Keynesian and Fisher- tic price level. However, if the domestic financial ian effects; there are also neo-Keynesian effects. and/or corporate sectors have significant liabilities First, a devaluation reduces the stock of wealth, denominated in dollars, the currency devaluation either due to the resulting increase in the price may actually exacerbate the Fisherian effects by level and/or by raising the domestic currency value raising the needed marginal productivity of capital of foreign liabilities. This wealth reduction effect required to service a given stock of liabilities. can lead to a decline in aggregate consumption Currency devaluations may also have negative and/or investment. Second, interest rate hikes raise repercussions in the context of dynamic inflation- the value of the liabilities of indebted banks and ary expectations and wage indexation. Since the corporations, thus reducing their net worth. Con- domestic price level is determined by a weighted sequently, private sector creditworthiness and average of nontradable and tradable prices, cur- investment decline. rency devaluations can lead to higher expected Regarding policy responses, contractionary rates of inflation. Moreover, if nominal wages are monetary and fiscal policies can raise the confi- indexed to the general price level, then nominal dence of international investors, reducing capital wages will rise proportionally to the currency outflows. Tight monetary policies tend to raise the devaluation, the exact magnitude depending on returns of domestic assets, but they also tend to the share of domestic absorption dedicated to increase domestic financial distress and dampen tradable goods.5 This means that a devaluation can domestic investment, both of which reduce the be followed by immediate wage and price hikes expected returns from investing in domestic firms. affecting the nontradable sector. In the extreme Tight fiscal policy can also have contradictory case, the nominal devaluation will lead to imrmre- effects on the confidenice of internatioiial investors. diate increases in nontradable prices, thus main- On the one hand, it reduces the public sector's taining a constant real exchange rate (RER) and borrowing requirements, which may lead to a negating the competitiveness effect of the devalu- lower perceived risk of default, and sends a signal ation. to the market that the fiscal authorities will con- The positive effects of a nominal devaluation tribute to the anti-inflation effort. On the other can be partially or wholly overtaken by the nega- hand, tight fiscal policy contributes to the fall in tive effects of financial distress whenever firms in domestic aggregate demand. The net effect of the nontradables sector and financial institutions macro policy responses, therefore, is a subject for have high unhedged foreign currency liabilities. In empirical research, and it may depend on country- turn, the severity of the resulting real interest rate specific conditions. increase will depend both on the sensitivity of Before proceeding to the analysis of the fac- investment and durable-goods consumption to tors that may have determined the magnitude of interest rates and on the degree of leverage the economic downturn in Latin America and (debt/equity ratios of corporates) and the health of Asia, we assess the process of adjustment in our the financial sector, which will determine to what six case studies. 5 l 2 i | i~~~~MAT] theX . 0 tTIN 2 CAI D AS IAN g|i~EP | the VERES FIGURE IA SHOWS THE GDP GROWTH RATES BASED on four-quarter moving averages, while Figure lb shows the year-on-year quarterly growth rates.6 From Figure la, it is clear that the ex-ante growth rates in Argentina were comparable to those of Indonesia, and a bit higher than Korea's. However, both Mexico and Brazil were growing at significantly lower rates than the other cases. Argentina and Mexico initiated their recoveries by quarters four and five, while the Asian economies continued their economic decline. In the Further evidence of the deeper economic case of Brazil, as of the fifth quarter (1998Q4), the downturn in Asia relative to the Latin American cases economic decline had been continuous but moder- is provided by Figure 2, which shows an index of the ate relative to the other countries. Figure lb shows difference between the average monthly growth rate that Mexico experienced the fastest decline, in the of industrial production in the year preceding the sense that its year-on-year quarterly growth rate was cn'ses minus the actual year-on-year monthly growth the lowest in quarters one and two (1995Q1-Q2), rates of the same variable in the aftermath of the but its recovery began by quarter three. In contrast, crises 7 Korea's and Thailand's growth of industrial Indonesia's GDP decline began with a significant production began to recover after 8-9 months, at lag, and the precipitous decline began in earnest by about the same time as in Mexico and Argentina, but quarter three (1998Q1). Argentina and Korea fol- after experiencing a deeper decline. For Indonesia, lowed similar paths up to quarter four (1995Q4 and the available information does not show a recovery 1998Q3, respectively), but Argentina's recovery was yet. Brazil experienced the least dramatic change in much more dramatic thereafter. the growth of industrial production. 6 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: Overall, then, quarterly GDP and monthly rates unseen in the Latin American cases. More- industrial production growth rates show that the over, Argentina and Mexico experienced speedy Asian economies experienced deeper economic recoveries by the fifth quarter after their episodes declines, since their growth rates fell to negative of speculative attacks. Figure la. Average Quarterly GDP Growth Rate -4 3- - -1 9 3 4 5 1 Sources: Central Banks and M nistries of Fnance. Figure lb. Year-on-Year Quarterly GDP Growth Rate V _ _S _ _ _ _S 1- 0 -0- -5 -r-Argentina + Braz XX -20 + ~~Mexico - Indonesia Ls Korea -25 3 -2 -1 1 2 3 4 5 6 Sources; Central Barks and Ministr es of Finance. A TALE OF Two REGIONS? * 7 Figure 2. Decline in Growth of Industrial Production (growth averaged in the 12 months leading to crisis minus actual year-on-year monthly growth) 15 . -. -~~Aroentina +Brazi! + Mexco A* -20 -25 -30 vJ -3 5 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Months After the Crisis Sources. Argentina: INDEC; Brazi IBGE; Mexico: INEGI: Indoresia: Bank of indonesia and BPS (Statistical Agencyi; Korea: Bank of Korea; Thailand: Bank of Thailand. 8 LEPER aE ECOVERIE S THE FOLLOWING SUBSECTIONS ATTEMPT TO ANSWER this question. In the first place, we evaluate whether the magnitude of the speculative pressures or the extent of the reversal in current account balance can explain why some countries suffered greater economic declines than others. Subsequently, we look at the potential role played by macroeconomic policies, the initial financial structure, the initial composition of aggregate demand, and exports. Each subsec- tion includes a suggestive (although certainly not definitive) analysis of the correlations between the country rankings of economic decline and the country rankings of the relevant explanatory vanables. These rankings and the procedures for deriving them are explained in Appendix B. WAS IT BECAUSE THE SPECULATIVE average of changes in the stock of international PRESSURES OR THE EXTERNAL reserves, nominal exchange rates, and domestic ADJUSTMENTS WERE STRONGER? interest rates.8 An important consideration in the construction of such an index of speculative pres- Measures of speculative pressures sure (ISP) is the criteria used for weighing each component.9 A measure of the magnitude of speculative pres- The main purpose of using the ISP here sures is necessary to answer this particular ques- is to measure the magnitude of the speculative tion. Eichengreen, et al. (1996) use an indicator of pressures on Mexico and Argentina (around speculative pressures that consists of a weighted December 22, 1994), Brazil (around October 22, A TALE OF Two REGIONS? * 9 1997), Thailand and Indonesia (around July 2, reflect the impact of the changes in the three 1998), and Korea (around Oct. 22, 1997). Given variables. the various choices available for constructing Figure 3a shows the evolution of the ISPI, such indicators, we chose to rely on two sets of and Figure 3b shows the evolution of the ISP2. indicators. One (ISP1) is constructed with the This evidence shows that the speculative pressure standard deviation of the variables (divided by on Mexico's currency was of comparable-to-greater their sum) as the corresponding weight. The magnitude than that faced by the Asian economies. other index (ISP2) is constructed by giving equal But the Asian economies were definitely harder hit weight to each variable, which would then purely than Argentina in 1995.10 Figure 3a. Index of Speculative Pressure 1 33 7D 60 rgenrina -Erazil -+-Mexico .c 0 ndoresia Korea -*The an- 4C 20 / 30 -4 -3 2 -1 0 1 2 3 4 Monrhs since the Crises Source: Authors' calculations basec on data from IMfF International Firance Statistics. Figure 3b. Index of Speculative Pressure 2 7CC 60 - 0 -- Argerina -Brazil ---- Mexico 30 - ndonesia -5-KaKorea --n-Tha and 20 - ° 20 _3D -4 -3 -2 1 c 2 3 W-cnths srce - te Cr,ss Source: Based on catE from IMF, riternationa Finance Stat stics. 10 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: The degree of speculation against a currency Brazil experienced the smallest reversal in the pri- may not necessarily lead to a greater external mary trade balance. It seems, therefore, that Thai- and/or domestic adjustment in the medium term. land and Korea experienced the largest swings in Hence, we also look at the resulting reversal of the net capital inflows and in net resource transfers. It current account and primary trade balance after is possible, therefore, that, at least in these two one year as an indicator of the magnitude of the cases, the severity of the domestic economic adjust- external adjustment required by the sudden stop of ment was due to the high magnitude of the reversal capital inflows. in capital inflows. One possible explanation of the larger reversals in Thailand and Korea is their very External adjustments and measures of high share of short-term inflows relative to total "sacrif ce" flows, which is reflected in the large shares of for- eign claims maturing in less than a year as a share Table I contains the relevant information. The of total foreign claims as reported by the Bank for first column shows the periods corresponding International Settlements.12 to the growth rates used in the second column." Large external adjustments can result from The third and fourth columns show the change exogenous reversals in capital flows, but they can in the current account (CAB) and primary trade also be driven by domestic adjustments in the level (TB) balances of the balance of payments as a and composition of aggregate demand (i.e., the share of the ex-ante GDP in U.S. dollars. The demand for tradable versus nontradables). The change in the CAB variable is useful because the fourth and fifth columns of Table 1 show the ratio turnaround in the CAB is associated with rever- of the ex-post GDP growth rate divided by the sals in net capital inflows, plus the difference change in the CAB or TB (as a percentage of the in the change in reserves. The TB is a measure initial GDP in U.S. dollars). Indonesia is clearly of the net resource transfer to the rest of the the case where the domestic sacrifice was the world; a surplus represents a net transfer to the largest relative to its external adjustment. Argentina outside. is a distant second, while Thailand and Mexico Thailand and Korea had the largest reversals in share the third and fourth- spots. The high domes- the CAB and TB, followed by Mexico. Argentina tic sacrifice relative to the external adjustment experienced the smallest reversal of the CAB, while of Argentina could be explained by its lack of Table 1. Sacrifice Ratios Period after GDP Growth Change in Change in % GDP Growth/ %/c GDP Growthl the attack (%) CAB/GDP (%) TB/GDP (%) Change CAB' Change TB2 Argent na 1995 -4.02 2.19 2.15 -1 84 -1.87 Mexico 1995 -6.22 6.67 6.70 -0 93 -0.93 Braz I 1997 Q4-' 998 Q3 1.22 -0.25 0.41 -4.87 3.00 ndonesia 1997 Q4-1998 Q3 -8.13 2.89 6.24 -2.81 -1.30 Karea 1998 -6.50 11.20 8.52 -0,58 -0.76 Thailand 1998 -8.00 9.44 2.16 -085 -3.71 'The Sacrifice Rat os are computed as tol ows: g / [(CAB1 - CABO) / GDPO] where g is the GDP growth, CAB1 ard CABO are the Cirrent Account Ea ance for the year ending at the qjarter indicated, and GDPO corresponds to tne GDP in LS dollars for the four quarters r ght before the crisis. 2Same as ir 1. except that tne TB stands for the Trade Balance, Sources: Internationae Monetary Fund, Irterrational Financia Sta.istics Bank of Korea, Bank of Indonesia, Bank of Thai and, Bank of Brazi , M nistry of Finance of Argentina A TALE OF Two REGIONS? * 11 adjustment of its exchange rate. In contrast, Brazil associated with higher external adjustments, which seems to have experienced the smallest domestic were perhaps associated with harsh punishment sacrifice relative to its (also low) external adjust- inflicted by international capital flows. In turn, it is ments, but this small sacrifice is probably the result possible that the imposed external adjustments of a rise in consumption that, paradoxically, may be were related to the incidence of short-term foreign a symptom of the lack of credibility of the adjust- liabilities. In the case of Indonesia, it is more likely ment program (see below). that the severe economic downturn was due to the In synthesis, the Asian economies experi- effects of either policy responses or to the impact enced either the largest external adjustments of the high interest rates and exchange-rate depre- (Thailand and Korea) or the largest relative ciation on the profitability of domestic investment domestic sacrifice (Indonesia). Hence, in the for- and national wealth. mer two cases the slow economic recovery in the Figures 4a and 4b show the extent to which aftermath of the speculative attacks is likely to be the rankings of the ISP1 and change in the CAB/Y Figure 4a. ISP1 and Economic Decline Rankings (D1 1.7) Figure 4b. Change in CAB/Y and Economic Decline Rankings (Di= 1.3) t 0 a1 12 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: help explain the rankings of economic decline. WAS IT BECAUSE MONETARY The deviation index (DI) of the rank correlation AND FISCAL POLICY RESPONSES based on the ISPI and the CAB reversal are 1.7 WERE STRONGER? and 1.3, respectively. Whether these DIs are low or high has to be assessed in comparison to the In this section we describe the monetary and fiscal explanatory power of other plausible hypotheses. policies implemented in our case studies and assess Figure 5. Real MI (deflated with CPI) 3 -_A g-ti sBraz l 3 ndoresia _ 80 ,. , e-= o -. = -<- Moro . . - Ir a-re z U . v . 12- -3 I' 8- 7 4.2 2 1 0 2 6 7 8 O 21 12 W"/Ihs 9 nce -e Drie Soirce: Based an data from tre IMF (Intemational Monetary Fund) Irternational Financial Statistics. Figure 6. Real Domestic Credit (deflated with CPI) 250 ,30 -_ _ _ _ _ __ _ _ _ _ _ --*Argentna -Brazo +- n2ones6 a -- Korea, Rep. - -r--- Mexico -Thaiiand 3 50 . . . - --_ 90 _ 73 -12 11 -13 9 -8 7 -e 5 -4 3 -2 -1 o 2 3 4 5 r 7 8 9 13 1 12 Mcnthsrn^e the Cise- Source: Based on data from IMF (International Monetary Fund), nternational Financ al Stat s. cs. A TALE OF Two REGIONS? * 13 Figure 7. Change in Real Deposit Rate Relative to the Level Before the Crises (deflated using annualized monthly change in CPI) 25 O -- -- ---_ __ __ _ - 200 -2: | 1 -0enin -0 a -7dDes -0 -_ _4 -l- 0 e5C-- -, F -Tr . 20 0 -' 1.1 9 -3 7 6 5 - -4 -3 2 .1 D 2 3 4 5 6 7 6 g GO Zl IL Source: Based on data fromr IMF (International Monetary Fund), Internationai Financial Statistics. Figure 8. change in Real Money Market Rate Relative to the Level Before the Crises (deflated using annualized monthly change in CPI) 500 - + Arg-bnt - B-a I A 220 - Indons - Korea Rep t - 1 a 102 - ___ -000 -- -___ -~ ____ -200 30C 12 -11 o 9 8- 7 5 4 . 2 .1 0 1 2 3 4 e 9 7 01 11 12 Source: Based on data from IMF ( nternational Monetary Fund), International Flnancral Statistics the extent to which the "tightness" of these poli- where real Ml levels fell 20 percent below the cies can explain why some countries suffered level it had at the beginning of its crisis, as illus- more than others. trated in Figure 5. 1-lowever, Indonesia experi- enced a significant increase in real Ml. A similar Monetary policies pattern can be depicted from the evolution of real domestic credit. Figure 6 shows that the two cases Figures 5 through 8 show the evolution of mone- that experienced the smallest domestic credit tary variables before and after our six episodes of expansion were Argentina, followed by Brazil. The speculative attacks. In some cases, monetary con- most salient case is Indonesia, where a year after traction was especially important, as in Korea, the crisis real domestic credit had more than 14 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: doubled. We should be cautious about our inter- real money market rate than the Latin American pretation of this evidence, because it is possible cases, although the behavior of Brazilian rates was that the rise in the demand for liquidity was greater very similar to Indonesia's in the sense of their in Indonesia and in the other Asian episodes than gradual decline. in Latin America.13 Figure 9 shows the relationship between the However, a similar conclusion emerges from economic decline and "tight" money rankings- looking at the evolution of real interest rates. the DI is 3.2. The main conclusion from this sec- Two different real interest rates are considered in tion, therefore, is that monetary policy was not Figures 7 and 8 to analyze developments in both clearly tighter in the Asian crises than in the Latin deposit and liquidity markets. Figure 7 clearly American cases. The extent to which the reces- shows Indonesia, Brazil, and Mexico on top of the sions following the crises were deeper, and their ranking in terms of the rise in real deposit interest consequences longer lasting, is probably related to rates during months 1-3. Real deposit rates rose other factors, some of which are discussed below. dramatically in Indonesia by month 4, when the change in its rate reached the levels seen in Brazil Fiscal policies during the earlier months. Mexico and, to a lesser extent, Korea experienced real deposit rates that Table 2 shows the primary fiscal deficits as a were below their precrisis levels at some point share of GDP during the years before and after during the 12 months following their crises. The each episode. Table 3 presents a measure of "fis- changes in the real money market interest rates, cal effort," defined as the change in the primary taken as a measure of the liquidity of the financial balance as a share of the initial GDP The nega- systems, tell a similarly ambiguous story. Figure 8 tive numbers reflect declines in primary fiscal shows that, of the Asian cases, only Indonesia had balances. Indonesia experienced by far the great- a more pronounced and persistent increase in the est deterioration of its primary fiscal accounts, Figure 9. Tight Money and Economic Decline Rankings (DI= 3.2) 0 1 2 3 4 5 6 Tg ght k'4o4e Rankhigs A TALE OF Two REGIONS? * 15 Table 2. Fiscal Policy Before and After the Crises (primary fiscal balance as a % of GDP) Before the Crisis After the Crisis Argentina 0.10 -0.91 Brazil 0.57 -1.23 Mexico -2.10 -4 70 ndonesia0b 12.05 -214 Koreab 5.52 1.62 Thailandoc -2.10 -2.80 'Since the fiscal year for Indonesia begirs in April, the oata before the crisis corresponds to the period Apri 1996-March 1997, and the data afterthe cr"sis to the perod Apri 1997-March 1998. bCent,al government on y, net includ ng lica governments. 0The data before the crisis correspord to overall 1997, and afterthe crisis correspond to an estimation for 1998 made by IMF staff. Table 3. Fiscal Policy After the Crises Fiscal Effort Fiscal Impulse Argentina -0,98 0.10 Brazil -1.88 2.15 Mexico -2.31 0.89 Indonesia -13.99 12.64 Korea -3.95 2.40 Tha land -0.39 -1.00 Fiscal Effor = [Primary F scal Balance () - Pr mary Fiscal Balance (t-1)] / GDP (t-1) Fiscal Impulse = Growth of GDP ' (Gov. Revenues/GDP) - FPscas Effor Source: Based or nat ona and f sca accounts data at constant local prices. followed by Korea. Thailand, in contrast, was the deterioration of the primary fiscal account that was country with the least expansionary change in the not due to the decline in GDP The numerator in primary fiscal account. equation (1) is equal to the change in public rev- It is significant that in all cases the primary enues (t) minus the change in expenditures (G): fiscal accounts deteriorated. This phenomenon may be due to the impact of the so-called automatic sta- (2) (PFB, - PFBd)/Y = (A Tt - AGd)/Yt, bilizers of the fiscal accounts, that is, the declines in economic activity reduce public revenue (see Price Here we can introduce two simplifying assump- and Chouraqui 1983). As mentioned, we use the tions: Let G be determined by exogenous spend- following measure of fiscal effort: ing policies, while T depends on the revenue rate t-also determined by exogenous policy choices- (1) Flscal Effort = (PFB, - PFBt )/Y', and on the level of real income, Y. The primary fiscal balance (PFB) at time t is the (3) (PFB1 - PFBT,)/YIt = ((ttYt - tt_,r, )-AG,)IY,_, inflation-adjusted PFB, which makes it compara- ble in terms of purchasing power to the PFB at In turn, the change in revenues can be decom- time t-1. posed as follows: With a simple accounting exercise we can derive the "fiscal impulse," or the portion of the (4) tII.- t ,Y, = t,AY+ Y>At 16 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: Equation (4) can then be substituted into (3): The second column in Table 3 shows our estimates of fiscal impulse. Indonesia was by far (5) (PFBI - PFBt_7)/Y, I = ((tAY + Yt ,At) - AG,)/Yr, the country that implemented the most expan- sionary fiscal policy. Only Thailand had a contrac- The concept of fiscal impulse is meant to capture tionary fiscal stance, while Argentina's was virtu- the portion of the change in the primary balance ally neutral. As in the realm of monetary policy, that was not due to the change in income Y In therefore, the Asian countries, especially Indone- other words, a portion of the right-hand side of sia, do not seem to have "suffered" from tighter equation (5) contains the (negative of the) fiscal fiscal policies more than the Latin American impulse (FI), which is due to changes in the rev- countries. enue rate t minus the change in expenditures G, Figure 10 shows the relationships between relative to the ex-ante GDP: the economic decline rankings and the country rankings of fiscal impulse. While the DI of 2.0 is (6) - Fl = (Y,_,At - AG)/t,_ lower than the one for "tight" money, it is still comparatively high. Therefore, equation (5) can be rewritten as follows: WAS IT BECAUSE INITIAL CONDI- (7) (PFB - PFB,_)/Y, -F/ + (t,AY/KYt TIONS AMPLIFIED THE IMPACT OF SPECULATIVE PRESSURES AND Our measure of Fl can be derived by adjust- POLICY RESPONSES? ing our measure of fiscal effort by subtracting it from the product of the ex-post revenue rate (ti) It is clearly difficult to explain why the Asian multiplied by the real growth rate as follows: cases suffered a faster and deeper economic decline than the Latin American cases based on indicators (8) Fl, = [(t,A /Y,_) - [(PFB, - PFBt_)/1Y of their policy responses. However, the effect of Figure 10. Fiscal Impulse and Economic Decline Rankings (DI=2.0) o3t~~~~~~~~~~~~~~~i-~ E- / '.... Sascal baon Randis Source: Based on off cial data from central banks and m inistries of finance. A TALE OF Two REGIONS? * 17 monetary and fiscal policies on the real economy exposures acquired by Indonesia, for example, depends on initial conditions, including the finan- were held primarily by the nonbanking private sec- cial structure and the composition of demand .14 tor.6 In the case of Thailand, the net foreign liabil- The following subsections address these issues. ities accumulated by its banking system exceeded 21 percent of its GDP This means that a 50 per- Did the initialfinancial structure play cent devaluation would have led to an increase in a role? the net foreign liabilities of its banking system in excess of 10 percent of GDP (assuming that the The size (and speed of recent growth) of the GDP stays constant). Table 4 also shows the out- domestic financial sectors and the structure of standing foreign loans to the private sector (as domestic financial assets and liabilities may be cru- reported by the Bank for International Settlements) cial elements in determining the magnitude of the as a share of GDP for each country prior to the economic decline in the aftermath of speculative crises. Again, Thailand clearly faced the prospects attacks.'5 Table 4 shows the initial level of factors of significantly larger negative effects from nominal that may determine the magnitude of negative devaluations. Indonesia and Korea also had signifi- Fisherian and net worth effects of currency devalu- cantly higher foreign loans (FL) relative to the size ations and interest-rate increases. The first variable of their economies than the three Latin American is the net foreign liability (NFL) positions of the cases. In addition, as mentioned earlier, the Asian banking systems as a share of GDP This evidence economies had higher shares of short-term debt indicates that only Thailand in July 1997 had a sig- (with a maturity of one year or less) than the Latin nificantly higher net foreign liability position than American economies. In general, the potential for Argentina and Mexico before December 1994 and harmful effects emanating from currency devalua- higher than Brazil inJuly 1997. However, it is tions was significantly higher in the Asian markets well known that much of the foreign currency than in Latin America. 7 Table 4. Initial Financial Structure NFLIGDP PrvtL/GDP STL/GDP PSC/GDP Private (%/0) (%0) (%0/)(% Credit Growth (%) Argentina 2.1 7.8 6.2 17.1 29.4 Brazil 4.4 6.5 5.2 30.7 0.0 Mexico 4.0 9.8 7.5 7.6 46.1 Indones:a. 2.3 20.3 13.5 49.8 53,9 Korea 1.9 17.4 12.4 60.5 29.0 Thailand 21.7 36.4 24.6 95.6 34.2 NFUGDP = Net Foreign Liabil .ies of the bank ng system civided by Gross Domestic Product PrvtUGDP - Private oars (c a ms as reported by fhe B S) divded by Gross Domestic Product STUGDP = Shot-term fore gn iaDilit es ddvied by GDP PSCiGDP= Prvate Sector Credit dOv ded by Gross Domestic Product Private Credit Srovrtn= cumrj ative growth of tota: credit to private sector at conrstant loca pr ces during 24 months pror to -oe crises Observations for NFL: Argertira and M.exico: average for Noverrber 199l: Braz I and Korea: average for Septerrber 1997: indones a and Thailano: Average for June 1997 STL is as of end-June of vear indicated, except for the STU/GDP rat os for Braz.l Indonesia. Korea, and Thailand, where data as of end-June 1 937 were used. Sources: IMF ( nternaticnal Monetary Furid) literietiu cil Firiaricial Stat st vva PrvsL frorTl BIS. 18 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: The effects of interest-rate increases depend to which these factors could have played a role in on the extent of leverage in an economy. When determining the magnitude of the ex-post eco- corporations are highly indebted and their debt-to- nomic decline. equity ratios are high, an increase in interest rates The evidence concerning the initial incidence can be catastrophic. The firms' creditors, including of private sector credit presented in Table 4 portrays domestic banks, can either attempt to absorb the dramatic differences between the Latin American shock themselves, which would be reflected in and Asian cases. The negative effects of a given deteriorating capital and liquidity to liabilities interest-rate hike in Asia would have naturally ratios, or they can pass on the increased costs of resulted in greater negative consequences for the real borrowing to their clients. If their corporate clients economy than in the Latin American countries."9 already have high debt-to-equity ratios, their Another factor that may explain the nega- capacities to pay back more exTensive credit will tive effects of high interest rates is the quality of deteriorate further, and the quality of the banks' credit to the private sector. In principle, this fac- loan portfolios will also deteriorate. World Bank tor can be assessed by looking at the initial share (1998) showed that while the incidence of leverage of nonperforming loans of the banking system. varied across industries and across countries in In practice, however, this type of information is East Asia, Indonesia, Korea, and Thailand had not strictly comparable across countries. For some of the highest average corporate debt-equity these reasons, we look at the evolution of real ratios in the world during 1988-96.18 Due to data credit to the private sector in the 24 months pre- limitations concerning comparable debt-to-equity ceding the crisis episodes. The underlying ratios of firms in our sample of countries, we rely assumption is that the quality of credit to the on a macroeconomic proxy, namely, the incidence private sector tends to deteriorate when it grows of credit to the private sector, to assess the extent too fast. The last column of Table 4 shows the Figure 11a. Financial Structure and Economic Decline Rankings (DI=0.3) ~t,_ I3 Suc It rs c a F I.I F a t .: sf'R - Sojce. Authors' calcu ations. A TALE OF Two REGIONS? * 19 Figure I1b. Ex-ante Credit Growth and Economic Decline Rankings (DI= 1.3) 3 1 2 3 4 5 sod I G3rowth Rassk sos Source: Authors'ca culations. accumulated growth rate of inflation-adjusted situations, are also important explanations of the credit to the private sector in each case. It is deep Asian recessions. Figure 12 shows the rank- striking that Brazil had an accumulated growth correlation of the relationship between ex-ante rate of credit to the private sector of zero during investment rates and the subsequent economic the two years leading to its episode of speculative decline-the DI is only 0.7. pressure. The other five economies experienced A simple accounting exercise reveals the rela- a substantial growth of real credit, led by tionship between ex-ante investment and con- Indonesia and Mexico. sumption shares and income growth: Figures 1 la and 1 lb show the correlation between the rankings derived from our financial (9) AY/Y',, = (C1 !/Y, * A/CL, + (i1_1Y ,) * A/,//t, + structure and ex-ante credit growth indicators X,_Y, I* /jXI1t, - (M1->/,Y ,) * AMM, and the rankings of economic decline. The DIs are 0.3 and 1.3, respectively. Hence the initial If the fall in consumption during crisis episodes financial structure seems to be an important were of similar magnitude (in terms of the per- explanation of why the Asian economies suffered cent change) to the fall in investment, then it greater economic declines than the Latin Ameri- would be difficult to assert that higher ex-ante can countries. investment rates (I/Y) explain the deeper down- turn in Asian countries.20 But Table 5 shows Did the initial composition of demand play that in all of our cases, gross domestic investment a role? fell more than consumption. Since the Asian economies initially had higher investment rates, a High investment rates, combined with the given percent fall of investment accounted for high responsiveness of investment to crisis larger shares of the decline in GDP In fact, the 20 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: Figure 12. Ex-ante GDI Rate and Economic Decline Rankings (DI =0.7) fJ 1 2 3 5 6 LA-te GOI Rate R-kJg, Source: Authors' calculat ors. contribution of consumption was only larger in exchange-rate regime are not consistent, eco- the case of Argentina, where consumption nomic agents will tend to increase their con- declined by 5.8 percent, but its share of GDP ex- sumption, especially imports, in anticipation of ante was 0.84, thus contributing -4.9 percent to its the fall of the currency. Combined with the fall in GDP evidence discussed below concerning the rise of In Brazil, investment declined by 3 percent, inflationary inertia in Brazil (Figure 15b) and the but consumption rose by a similar 3.2 percent, fact that fiscal policy in Brazil was expansionary and imports (not shown in Table 4) rose by 8.3 in the aftermath of the attack of October 1997 percent. This combination of declining invest- (Table 3), it is clear that the Brazilian predica- ment with rising consumption and imports is ment of 1999 is partly due to the failure to symptomatic of an adjustment program that lacks implement a credible adjustment program during credibility. When adjustment policies and a rigid 1998. Table 5. Decomposition of GDP Growth Rates after Speculative Attacks Argentina Brazil Mexico Indonesa Korea Thailand Ex-post GDP Growth (%o) -4.02 1.22 -6,22 -8.13 -'.91 -8.00 Ex-ante GDI Rate 0.23 0.21 0.22 0.34 0.32 0.35 Ex-post GDI Grovvth (°%) -15.97 -3.00 -34.95 -29.17 -55.04 -39.00 Contribution of Investment (%) -3.64 -0.64 -7.77 -9.83 -17.78 -13.61 Ex-ante GDC Ra:e 0.84 0.81 0.83 0.69 0.63 0.62 Ex-post GDC Growth (%) -5.78 2.39 -8.44 0.37 -11.04 -10.49 Contribution of Consumption (%) -4.86 1.94 -6.99 0.25 -6.91 -6.56 Contribution of Net Exports 4.48 -0.09 8.54 1.45 18.77 12.17 Source: Based on Data from Miiistries of Finance and Centra Banks. A TALE OF Two REGIONS? * 21 The main lesson from this analysis is that crises. However, given the evidence presented the initial composition of demand seems to have concerning real interest rates and trends in mone- played a crucial role in determining the extent of tary aggregates, it is more likely that other factors the economic decline after the speculative attacks. played a role. A Keynesian analysis leads us to In particular, given the high sensitivity of invest- focus on inflation dynamics, as does the considera- ment to uncertainty, high initial rates of invest- tion of inflationary expectations and wage indexa- ment tend to be associated with harsher economic tion, because these are factors likely to affect the declines. success of currency devaluations in terms of pro- ducing real exchange rate variations. In the follow- DID LOW EXPORT RESPONSE ing paragraphs we look at relevant evidence about PLAY A ROLE? WHY DID IT the extent to which our cases suffered from infla- HAPPEN? tion inertia and assess the "success" of the cur- rency devaluations in Asia and Mexico. In turn, we The slower recovery in Asia can be partially look at the evidence concerning the role of explained by the relatively lackluster response of regional integration and export similarity. merchandise exports (Figure 13). Figure 14 shows the relation between the (low) export growth and Inflation dynamics and the effects of the economic decline rankings; the DI is 0.7, exchange-rate variations which is quite low But, why did export volumes grow so slowly in the Asian economies? The lack- Obstfeld (1995) uses simple inflation auto- luster merchandise export performance could be a correlations to examine whether inflation persis- result of the severe foreign liquidity constraints tence (or inertia), which can be caused by either faced by Asian producers in the aftermath of the credibility problems and/or contract indexation Figure 13. Index of Exports Volume (12-month moving average) 140 Argent na - Brazi + Mexico 13 3 - _ _ 1Irdonesia -Korea -Thailand 29-- <2 , -i i EI - -7 -6 -5 4 -3 2 -1 3 1 2 3 4 5 6 7 3 9 J 11 12 klcnths S nce te Cr see Source. Based on International Finarcia Sta; stocs data, IMF 22 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: Figure 14. (Low) Export Volume Growth and Economic Decline Rankings (DI =0.8) 5 - /_ _ _ 02t - /_ ___ _ _ _ _ _ _ _. __ _ __ _ __ _ 0~~~~~~~~~~~~~~~~~~~~ Source. Authors' calculations. schemes, tended to be higher in industrial coun- time trend of the level of inflation. For example, tries with flexible exchange-rate regimes than in consider disturbances -t (perhaps a devaluation) those with fixed exchange rates. For our purposes, that raises the consumer price level at t= I and countries with high inflationary inertia can be another that raises it at t=2. In this case, the infla- expected to gain less from nominal exchange-rate tion level at t=2 would bc: adjustments, since domestic prices will probably catch up quickly afterwards, thus nullifying the (17) i2 = + 2xa +c, + F, potentially expansionary "switching" effect of the devaluation. Of course, the extent to which a disturbance , t The estimation of the extent of inflationary affects the level of inflation in subsequent periods inertia can be done using Ordinary Least Squares, will vary across countries. The actual regression assuming that the true inflationary process in these equation to be estimated is: economies is a random walk with a drift: (12) at=a+ Jt.1' (10) ir = a+i t 1 +c1, where & is the estimated drift coefficient and 1 where the true value of a is not zero, and £, is is the estimated autocorrelation or inflation iner- normally distributed with mean zero. In other tia coefficient. The main point, however, is that a words, we model inflation dynamics assuming currency depreciation at time t can affect subse- that contemporaneous disturbances affect the quent levels of inflation; the greater this effect, A TALE OF Two REGIONS? * 23 the less likely that such a depreciation will pro- throughout the period. This evidence indicates duce a real depreciation. that even when the levels of inflation were lower Figures 15a through 15f show the recursively in Asia on the eve of the crises, by late 1996 estimated coefficients of the constant (drift) on Indonesia and Thailand actually had high degrees the left-hand side graphs and the autocorrelation of inflation inertia. This is certainly the case for (inertia) coefficient for the one-year lag estimates Brazil and, to a lesser extent, Mexico. Only Korea on the right-hand side graphs.2" The recursive seems to have had significantly lower inflationary estimation technique allows us to examine inertia than the Latin economies and, based on whether the drift and autocorrelation coefficients this evidence, could have benefited from nominal were stable over time. Argentina (in Figure 15a) exchange-rate devaluation if other factors had not shows stable coefficients throughout the period; been present. the drift is zero and the autocorrelation is steady An index proposed by Goldfajn and Gupta at about 0.25. Brazil (in Figure 15b) shows a ris- (1998) can be used to assess the extent to which ing drift coefficient until mid-1995, followed by a the real exchange rate devaluations were brought rapid decline reflecting the success of the Real about by the nominal devaluations: Let the nomi- Plan in reducing the level of inflation. Brazil also nal depreciation success index, S = %ANEER/ seems to have experienced a significant rise in %AREER, where NEER is the nominal effective inflationary inertia, which is reflected in the rise exchange rate and REER is the real effective of the autocorrelation coefficient from a signifi- exchange rate. If S > 1, then the real depreciation cantly negative point estimate in mid-1995 to was brought about by a combination of nominal about 0.40 in 1998, although it seems to be statis- devaluation plus some domestic inflation greater tically significant only after early 1997. Mexico (in than foreign inflation. If S < 1, then real depreci- Figure 15c) had a stable and statistically significant ation was brought about by nominal devaluation drift coefficient of about 0.005 (or 0.5 percent plus some domestic deflation, and S =0 when the iimontlhly inflaJon) until early 1998, when it expe- noiiinlal devaluation is zero. S < 0 when the rienced a sudden upward trend that was probably nominal and real exchange rates move in opposite due to the inflationary consequences of the nomi- directions. For example, when the nominal nal depreciations experienced after the Asian exchange depreciates, but domestic inflation is crises. Mexico's inflationary inertia was statisti- greater than foreign inflation by a margin larger cally significant and positive beginning in 1994, than the nominal depreciation, the real effective and steadily climbed to about 0.35 in late 1997. In exchange rate appreciates and S acquires a nega- 1998, Mexico's inflationary inertia became statisti- tive value. cally insignificant. Indonesia (in Figure 15d) Figure 16 shows the evolution of S, each shows a lower inflation drift but a steadily rising observation being measured with respect to the degree of inflation inertia that was near 0.35 by level of NEER and REER in month zero. It is late 1996, continuing to climb thereafter, includ- clear that our three Asian economies experienced ing the period after the currency crises in Asia. nominal devaluations that effectively brought Korea (in Figure 15e) shows a sudden jump in the about real exchange rate depreciations, as reflected drift in early 1995, but an autocorrelation coeffi- in their S values near 1. In contrast, Mexico expe- cient near zero beginning at that point. Thailand rienced nominal depreciations that were diluted (in Figure 15f) shows a relatively stable drift (at by high domestic inflation relative to foreign 0.002) and autocorrelation coefficients (at 0.4) inflation. 24 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: Figures 15a-f. Recursive Inflation Drift and Inertia Figure 15a. Argentina 0.06- 0.5. 0.04_ 0.4P, 0.02 -- OOC ___-______ .3_ 0 OC 0. C'S _ _, ~ ~ ~ 0.2 _ _ _ ------------------_ -0.02 - . ----- 004- w 01 ' r0 -0.06- 0.0 92 93 94 95 96 97 98 92 93 94 95 96 97 9 - RecursveC(1) Estir ateC-- 2.E. - Recuive C(2) Estriates ---±2S.E. Figure 15b. Brazil 0.4 06 0.2~~~~~~~~~~~~~~~~~~~. 031 0.0~~~~~~~~~~~~~~~~~~~~~. 92 93 94 95 96 97 98 902 93 4 95 98 -Reirse C (1) Estirmtes --- -2S E Rec,rs e C(2) Estimates --- +2S E| Figure 15c. Mexico 1.5 6000 1 0 6. O ,--- 0--- 000-_ 0.0~ ~ ~ ~~~~~~~~~~10 _ _ _ _ _ _ ~~~~~~~~~~~~~COOQ5- ~0.0 6 1 92 93 94 95 96- 97........ 93 QC1Q 2 93 94 90 96 91 98 |-Recurs.eC(2) Ertimates 2SE Recars e (): s ates ±- 02SF. A TALE or Two REGIONS? * 25 Figure 15d. Indonesia C.C12 -.8 D.C08 ' C.CO6 '.,, ~~~~~ ~ ~ ~ ~~~~~~0 4. _ C.2 C CC4 74 C92 -C.C022 92 93 94 95 96 97 E8 92 93 94 95 96 97 98 R-R ecjrs ve C(1 ) stimates ---- ±2 S.E. R | ecursive C(2) rstimates --- 2 S.E. Figure 15e. Korea 3C3 2 - 2952 : I , 1 -.002 -1 Q 92 92 94 95 98 97 98 92 93 94 9S 96 97 98 - Rer N % NR ii 4i oqz°* .Qo 02J1 ~ ~ a-2-morth moving average of growth. Source: Based on data from il\AF (International Monetary Fund) International Financial Statistics Table 6. Intraregional Trade in the Americas (percent of total exports, average 1990-1996) United States 7 LAC Countries Total Argentina 11 31 42 Brazil 20 15 35 Chile 17 14 31 Colombia 39 15 54 Mexico 81 3 84 Peru 21 13 34 Venezuela 53 11 64 United States - 13 13 Source: Based on data from IMF (International Monetary Fund), D rection of Trade Statistics Table 7. Intraregional Trade in Asia (percent of total exports, average 1990-1996) Japan 5 Crisis Countriesa Rest of Asia's LDCs Total Indonesia 33 10 19 62 Korea 15 11 19 44 Malaysia 14 6 38 58 Philippines 18 7 14 39 Thailand 17 6 23 46 Japan - 16 22 38 alnconesia, Korea, Malaysia. 'hi ippines, and Thai and. Source: Based or data from IMIF (International Monetary Fund), Direction of Trade Statistics. A TALE OF Two REGIONS? * 29 Table 8. Average Export Similarity Indexesa Asia 5b LAC 7c Asia 5b 0.36 0.21 LAC 72 _ 0.22 Japan 0.31 0.19 China 0.39 0.24 2Based on Finger and Kre n n (1979). Ca cu:ations performed using data on the compost on of total exports (at the 4-digit evel) from the U,N. Comtrade database for the period 1990-96. blrdones a, Korea Malays a, Phi ipoines and Thai and. ~Argentna Brazil Ch le, Colomb a, Mexico, Peru and Venezue a. 30 i)NXONN IOS and tI tIMPLICATIONS THE FACT THAT RECENT LATIN AMERICAN CRISES have been followed by relatively quick recoveries should not be a Justification for inaction; even the recent Latin American crises have been terribly costly in terms of declines in GDP and industrial production. In the future it is likely that Latin American adjustments after speculative attacks will more resemble those of Asia in 1997-98 than previous Latin American crises, due to several reasons. As is well known, capital flows, including short-term loans and equity investments, have been rising sharply since the early 1990s, and they will probably continue to do so. Second, the privatization of domestic banking systems and the end of high inflation have also In Perry and Lederman (1998) we concluded been associated with increases in the size of bank- that sound institutions and incentives limiting the ing and financial markets relative to GDP This extent of currency mismatches, excessive leverage, part of the story is well known, but the interested credit growth, and stocks of short-term debt were reader can examine the evidence presented in key policy areas for preventing speculative attacks. chapter 2 of Burki and Perry (1997). Thus, unless The evidence reviewed in this paper, especially the adequate safeguards are put in place, one can realization that the size and structure of financial expect both higher debt-to-equity ratios, currency systems and high ratios of short-term debt to GDP exposures, and large stocks of short-term debt held seem to be key determinants of the magnitude of by the private sector in the future. the economic decline after crises, reinforces those A TALE OF Two REGIONS? * 31 conclusions, as such institutions and policies are 1. Inflation inertia and recent histories of also necessary to reduce the costs associated with high inflation may indicate that policymakers the postcrisis adjustment process. Sound financial should be cautious about excessive exchange-rate and corporate governance institutions, flexible devaluations, which can lead to an outburst of exchange rates, and prudential regulations on inflation. In such cases, other things being equal, short-term capital inflows, in addition to strong there is a stronger case for interest rate defenses of macro fundamentals and prudent public debt the currency. management, would buy significant insurance 2. On the other hand, if the initial financial against the likelihood and potential cost of finan- structure is characterized by a high incidence of cial crises. leverage, and there are moderate foreign currency Purchasing the right to tap liquidity cush- exposures and a previous history of price stability, ions in times of distress (like the Argentines have it is preferable not to undertake overly restrictive done since 1995) can also be an important instru- monetary and fiscal policies, thus letting the ment to reduce both the likelihood of a specula- exchange rate depreciate accordingly. tive attack and enhance the credibility of tight 3. If, on the contrary, an economy has high monetary policies aimed at defending the value of foreign currency exposures but low levels of the national currency. In effect, the availability of leverage, then policymakers should be concerned liquid lines of credit from abroad can be viewed about dramatic currency devaluations. Therefore, as additional international reserves that can be under these conditions, tight macro policies are tapped in the context of a speculative attack. In appropriate. turn, these additional reserves provide additional 4. The worst scenario is one where domestic time for monetary authorities to maintain a vigor- leverage is high combined with high, unhedged ous defense of the exchange rate, assuming, of foreign currency exposures, because in this type of course, that the domestic financial system can situation both high interest rates and currency withstand a period of high real interest rates. depreciations have severe Fisherian and net worth High (remunerated) liquidity reserve require- effects on the real economy. Thus, any necessary ments on bank deposits would also help reduce institutional reforms should be made to prevent the likelihood and eventual cost of capital flow the emergence of this combination of factors in reversals. the first place. The main conclusion about monetary and 5. Finally, higher initial rates of investment fiscal policy responses to speculative attacks is that may also require laxer macro policies to maintain there may be difficult tradeoffs during a crisis. We real interest rates at more-moderate levels, other found that monetary and fiscal policies in Asia things being equal. were not necessarily tighter than in the Latin It is also more likely that Latin American American cases in the aftermath of the speculative economies in the future will experience financial attacks. However, macro policy responses should crises with a regional character, due both to consider that certain initial conditions seem to be increased financial and trade integration. Figure 19 key determinants of the magnitude of the eco- shows, for example, that the share of Argentine nomic recessions after speculative attacks. There- and Brazilian merchandise exports going to Latin fore, it is quite likely that macro policies in Asia American markets, including themselves, has been should have been less contractionary than they rising steadily since the early 1990s. Mexico is actually were. Table 9 summarizes our main policy slightly different, due primarily to its preexisting recommendations in this area: high dependence on the U.S. market. Also, 32 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: Table 9. Summary of Policy Recommendations Initial Conditions Tight Pl4onetary Policy Recommended Tight Fiscal Policy Recomnmended Recent high inflation Yes Yes High inflation ineria Yes Yes High domestic aeverage No No High fore gn currency exposures Yes Yes High domestic and foreign liabi ities Diff cu t tradeoffs Difficu t tiadeoffs Low capitalization and/or liquidity reserves of the banking system No No High investment rates No No Calder6n, et al. (1998), show that fluctuations in cies. Moreover, they are likely to benefit from Latin American industrial production seem to have regionally coordinated policy responses because an increasing correlation with variations in eco- if all countries respond to their crises with nomic activity in developed economies. This is a expansionary monetary and fiscal policies in iso- likely result of the process of trade liberalization lation from the others, they may find that the and financial integration that has progressed extent of currency depreciation will be larger quickly during the 1990s. than intended. In other words, the extent of When countries face regional crises, the nominal exchange-rate depreciation for each policy responses should be coordinated. If country necessary to achieve a given adjustment regional partners share similar economic and of the real effective exchange rate will be less if financial characteristics, all countries are likely the corresponding fiscal and monetary expan- to benefit from undertaking similar macro poli- sions are coordinated. Figure 19. Intraregional Trade in Latin America, 1990-1996 (exports to LAC7a as a % of total exports) Argeriav 2LAC7 =Arceor ne. Brazil, C[rie, C~oirlob a, Mexico, Peru ano Venezue a Sour-ce: Ca der6n et a. (998), based on' dats from [MF. Direction of Trade Statistics. A TALE OF Two REGIONS? * 33 APPENDIX A. Speculative Attacks and Macro-Policy Dilemmas: An Illustration Using a Textbook Keynesian Model with Exchange-Rate Expectations There has been a recent resurgence of inter- Assuming that nontradable inflation is deter- est in textbook Keynesian macroeconomic policy mined by the rate of change of nominal wages (w) analysis (Corden 1998, Krugman 1998a, Frankel and by the extent of excess aggregate demand rela- 1997). We find it useful to analyze the effects of tive to the full-employment income (Y*), speculative attacks using a simple model presented in Krugman and Obstfeld (1994), which incorpo- (4) mre = w + b(D - Y*), rates exchange-rate expectations into the tradi- tional Mundell-Fleming analysis. where b is the slope of the aggregate supply curve We begin by specifying the equilibrium con- in the price-quantity space. Assuming that dition in the goods and services market: expected tradable inflation is equal to the expected rate of nominal depreciations (e), (1) D (RER, Y-t(Y), r - 7r) =C (Y-t(Y). r - 7E) + (r - n) + CA (RER, Y-t(Y), r- -) (5) 7t' = e, where D is aggregate demand, expressed as a func- and by substituting equations (4) and (5) into tion of the real exchange rate, disposable income equation (3), we see that domestic inflation is (Y-t(Y)), and the real interest rate (r-7c). Note that determined by: the tax bill, t, is also a function of income, Y In an open economy with perfect capital (6) ite = aw + ab(D -Y*) + (1-a) e mobility, the nominal exchange rate is determined by the covered interest rate parity condition: In the goods market equilibrium, income equals demand. Also, since RER=EP*/P, we can (2) E = Ee/(I + r- r*), substitute equation (2) into equation (1), and also by substituting (6) into (3) and then into where Et is the expected future nominal exchange (1), income is determined by the following rate, and r* is the world interest rate. function: The domestic inflation can be determined by a variety of functions. To introduce the roles of (7) Y = D [El P*/{P(l +r-r*)}, Y-t(Y), r-aw-ab(D-Y*)-(,-a)e] exchange-rate expectations and wages, we can model the expected rate of inflation itc as follows: In the money market, the demand for liquidity, L, is a function of the income, Y, (3) Te = a TENI (/-a) iTi,e and the nominal interest rate, r. While the nominal money supply, Ms, is assumed to be where he,N stands for the expected rate of nontradable exogenous (determined by policy), the real inflation and T for tradables, and a is the share of money supply is simply the ratio of Ms to domestic consumption dedicated to nontradables. domestic prices, P: 34 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: Figure Al. An Illustration of Policy Dilemmas in a Textbook Keynesian Model with Exchange-rate Expectations r LM E 4Y E1 1FE Y (8) Ms/P = L (Y, r) to defend the currency. The idea is that the rise Diagrammatically, we have an extended in Ee is thought to be temporary, and thus the Mundell-Fleming model, where the domestic defense of the currency and showing a willing- interest rate (or the nominal spot exchange rate) ness to defend it eventually eliminate the expec- is determined in the foreign exchange market. tations of devaluation. When this happens, the Figure Al shows the foreign exchange market on LM curve can return to its original position after the left side, and the traditional IS-LM schedules the attack ends. on the right side. A speculative attack is then Alternatively, the authorities could let the represented by an exogenous increase in Fe, exchange rate go. If the RER changes, the demand which is illustrated here as an upward shift of the for home-produced goods and services rises, as equilibrium foreign exchange schedule. After r equation 7. Ignoring the effects of this policy this happens the national economic authorities on w and on We (as per equation 6), the IS curve have several options. First, they can intervene in would shift to the right, while the equilibrium r the foreign exchange market, which, assuming falls as E approaches Ee. Once we consider the no sterilization, results in a decline in Ms. This fact that such a policy would raise P - a once- is illustrated by the leftward shift of the LM and-for-all increase, since we are ignoring equa- schedule until the domestic interest rate equals tion 6 - we would reach a point such as C, the world interest rate plus the expected rate of which would result in lower domestic interest depreciation of the currency at point B. This is rates (both nominal and real) and a slight expan- the equivalent of a tight monetary policy aiming sion of income. A TALE OF Two REGIONS? * 35 When we consider the inflation dynamics decline of the real money supply, leading to a specified in equation 6, we see that the expected point B, despite the nominal devaluation. That rate of inflation would rise, depending on how is, the IS would not shift, because the RER did this policy affects the rate of growth of nominal not change, but the LM shifts to the left as a wages (w) and the extent of excess demand for consequence of the rise in domestic prices. home-produced goods. If Ire rises then domestic Finally, an increase in the demand for liquidity inflation would rise in tandem with expectations, for a given range of Y could arise as a conse- diluting the real exchange-rate depreciation. quence of financial distress, which would also When this happens, it is possible that Ee may rise lead to an upward shift of the LM curve. This again, especially in cases where the original illustration helps to shed some light on why it is increase in Et had been caused by a perceived likely that increases in domestic interest rates overvaluation of the real exchange rate. In the may be associated with both successful and extreme case there would actually be no change unsuccessful speculative attacks against emerging in the RER, with rising inflation resulting in a market currencies. 36 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: APPENDIX B. A Suggestive Rank-Correlation Analysis Throughout this paper we have attempted to calculated by ranking the resulting averages. The evaluate whether the performance of simple country that was ranked the highest in terms of macroeconomic indicators can explain why one the average of the two previous rankings was the set of countries experienced a more severe eco- country with the fastest and deepest fall in eco- nomic decline than another set of countries. nomic activity. There is one explanation that we can discard at the outset: Since the nominal devaluations in Asia CO UN TRY RANKING S: were more successful than in Latin America in EXPLANATORY VARIABLES terms of producing real exchange-rate variations, we cannot conclude that inflation inertia in Asia Indexes of Speculative Pressure. Each country (which was high in Indonesia and Thailand) was was ranked according to the average levels of the an important explanation of the deeper economic ISPI during months 0-4, based on the evidence downturn experienced by these economies. To presented in Figure 3a. summarize our findings, we look at the extent to Change in CAB/Y Each country was ranked which the country rankings produced by our according to the magnitude of the change in the "explanatory" variables conform to the country CAB relative to GDP, based on the data presented rankings of economic decline, which are shown in in Table 1. The country with the highest change in Table B1. The following paragraphs explain these CAB relative GDP received the first ranking. country rankings. Fiscal Impulse. Countries were ranked accord- ing to the magnitude of the fiscal impulse as pre- ECONOMIC DECLINE RANKINGS sented in Table 3, with the country with the small- est fiscal impulse receiving a ranking of "1". The first task in this exercise is to rank our Tight Money. Each country was ranked cases according to the magnitude of the economic according to the level of the real Ml and credit decline. For this purpose, we would like to capture indexes at month 12, and according to real deposit both the speed of the decline and the overall and money market interest rate indexes at month decline after a given period of time. These rank- 3. These four rankings were then averaged, and ings, therefore, were constructed from two rank- the final ranking was derived from these averages. ings. First, based on the evidence presented in Fig- The country with the lowest Ml-credit indexes ure 2, each country was ranked by how low the received a one, and the country with the highest average index of the decline in the growth rate of real interest rate indexes received a "1'. IP was during the 12 months after the crises. Sec- Financial Structure. Based on the information ond, based on the evidence presented in Figure la, provided in Table 2 (concerning the incidence of each country was ranked by the magnitude of the foreign liabilities relative to GDP, net foreign percentage decline of GDP during the first year asset positions of the banking systems over GDP, after the crisis.24 These two rankings were then the incidence of short-term foreign liabilities, averaged, and the economic decline ranking was and the ratio of credit to private sector over A TALE OF Two REGIONS? L 37 Table Bs. Country Rankings Economic Fiscal Tight Financial Ex-ante Change in Ex-ante Exports Decline ISP I Impulse Money Structure Credit Growth CAB/DP Investment Rate Volume Argentina 5 6 2 2 6 4 6 4 6 Brazil 6 2 4 2 5 6 4 6 4 Mexico 4 1 3 1 4 2 3 5 5 Indonesia 2 3 6 5 2 1 5 3 2 Korea 3 5 5 4 3 5 2 2 3 Thaland 1 4 1 5 1 3 1 1 1 Sotirce; Author's calculatioris GDP), each country was ranked according the Initial Composition of Demand. Countries magnitude of these indicators. These rankings were ranked according to their ratios of Gross were averaged, and a new set of rankings was Domestic Investment over GDP in the year (or derived from the latter. four quarters) immediately preceding the Ex-ante Credit Growth. Countries were crises. ranked according to the ex-ante cumulative The rank-correlation analysis can be done growth rate of real credit to the private sector as with boxes with 45 degree lines that show how reported in Table 2. The country with the highest closely the rankings of each one of the possible index was ranked number "1". explanatory variables match the rankings of eco- Low Growth of Export Volume Rankings. Coun- nomic decline. Each figure presented in the text tries were simply ranked according to the level of is accompanied by a DI that is the average of the the index of export volume at month 7, as pre- deviations from the 45 degree line; the higher the sented in Figure 4. The country with the lowest DI the lesser the explanatory power of a particu- index was ranked number 1. lar set of variables. 38 * ADJUSTMENTS AFTER SPECULAI IVE AT IACKS IN LATIN AMERICA AND ASIA: REFERENCES Burki, Javed S., and Guillermo Perry. 1997. from the Thailand Industrial Survey." Paper pre- The Long March: A Reform Agendafor Latin America sented at the Competitiveness Conference, and the Caribbean in the Next Decade. World Bank Bangkok, Thailand, 20-22 May 1998. Latin American and Caribbean Studies View- Edwards, Sebastian. 1989. Real Exchange Rates, points. Washington, DC: The World Bank. Devaluation, and Adjustment: Exchange Rate Policy in Calder6n, Cesar, Norman Loayza, and Hum- Developing Countries. Cambridge, MA: The MIT berto Lopez. 1998. "Common Cycles and Trans- Press. mission of Shocks: Recent Stylized Facts in Latin . 1996. "Exchange Rate Anchors, America." Processed. World Bank Latin American Credibility, and Inertia: A Tale of Two Crises, and Caribbean Regional Studies Program, The Chile and Mexico." American Economic Review, World Bank, Washington, DC. Papers and Proceedings 86(2): 176-80. Calvo, Guillermo. 1998a. "Capital Flows and . 1998. "Two Crises: Inflationary Iner- Capital-Market Crises: The Simple Economics of tia and Credibility." 7he EconomicJournal 108 Sudden Stops." Processed. University of Mary- (May): 680-702. land, College Park. Eichengreen, Barry, Andrew K. Rose, and . 1998b. "Understanding the Russian Charles Wyplosz. 1996. "Contagious Currency Virus: With Special Reference to Latin America." Crises." National Bureau of Economic Research Processed. University of Maryland, College Park. Working Paper 5681. Claessens, Stijn, Simeon Djankov, and Larry Finger, J. M., and M. E. Kreinin. 1979. 'A Lang. 1998. "Corporate Growth, Financing, and Measure of'Export Similarity' and Its Possible Risks in the Decade before East Asia's Financial Uses." The EconomicJournal 89 (December): 905-12. Crisis." Policy Research Working Paper 2017, The Frankel, Jeffrey A. 1997. "Sterilization of World Bank, Washington, DC. Money Inflows: Difficult (Calvo) or Easy Corden, W Max. 1998. "The Asian Crisis: Is (Reisen)." Estudios de Economia 24(2): 263-85. There a Way Out? Are the IMF Prescriptions Furman, Jason, and Joseph E. Stiglitz. 1998. Right?" Processed. Johns Hopkins University, "Economic Crises: Evidence and Insights from School of Advanced International Studies, Wash- East Asia." Brookings Papers on Economic Activity ington, DC. 2: 1-136. Corsetti, Giancarlo, Paolo Pesenti, and Goldfajn, Ilan, and Poonam Gupta. 1998. Nouriel Roubini. 1998. "Paper Tigers? A Prelimi- "Overshooting and Reversals: The Role of Mone- nary Assessment of the Asian Crisis." Processed. tary Policy." Processed. International Monetary Yale University, Princeton University, and New Fund, Washington, DC. York University. Kraay, Aart. 1998. "Do High Interest Rates Ding, Wei, Ilker Domac, and Giovanni Ferri. Defend Currencies During Speculative Attacks?" 1998. "Is There a Credit Crunch in East Asia?" Processed. The World Bank, Washington, DC. Processed. The World Bank, Washington, DC. Krugman, Paul. 1998a. "Latin America's Dollar, David, and Mary Hallward-Driemeier. Swan Song." Processed. Massachusetts Institute of 1998. "Crisis, Adjustment, and Reform: Results Technology, Cambridge, MA. A TALE OF Two REGIONS? * 39 1998b. "Saving Asia: It's Time to Get World Bank Latin American and Caribbean Studies Radical." Fortune, Sept. 7, 1998. Viewpoints. Washington, DC: The World Bank. Krugman, Paul, and Maurice Obstfeld. 1994. Price, Robert W R., and Jean Claude International Economics: Theory and Policy, 3rd ed. Chouraqui. 1983. "Public Sector Deficits: Prob- New York: Harper Collins. lems and Policy Implications." OECD Occasional Obstfeld, Maurice. 1995. "International Cur- Studies (une): 13-44. rency Experience: New Lessons and Lessons Radelet, Steven, and Jeffrey D. Sachs. 1998. Relearned." Brookings Papers on Economic Activity 1: "The East Asian Financial Crises: Diagnosis, 119-96. Remedies, Prospects." Brookings Papers on Economnic Perry, Guillermo, and Daniel Lederman. Activity 1: 1-74. 1998. Financial Vulnerability, Spillover Effects and Con- World Bank. 1998. East Asia: Tle Road to tagion: Lessonsfrom the Asian Crisesfor Latin America. Recovery. Washington, DC: The World Bank. 40 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: NOTES 'Among the many analyses of the causes of estimates provided by World Bank staff in Decem- the Asian crises, see Corsetti, et al. (1998), Radelet ber 1998. and Sachs (1998), and Perry and Lederman (1998). 7Indonesia's data are quarterly. Corsetti, et al. argue that the Asian crises were 8 The extent of speculative pressure against a caused by weaknesses in Asian "fundamentals," currency must be reflected in nominal deprecia- while Radelet and Sachs place greater emphasis on tions, and/or declines in reserves and/or increases the instability of international capital flows. Perry in domestic interest rates. and Lederman emphasize the interaction between 9 Corsetti, et al. (1998), use a weighted aver- a perverse incentive structure affecting domestic age of only the change in reserves and nominal private agents and the partial (and ill-sequenced) exchange rates due to (1) fears about the compara- liberalization of (volatile) international capital bility of interest-rate data across countries, and (2) flows. The three studies emphasize the important nominal interest rates are often affected by steril- role played by rollover or liquidity risks, usually ization operations, thus concealing some of the reflected in high ratios of short-term external debt effects of speculative pressures. Frankel (1997) over reserves. argues that in a simple textbook imodel of an open 2 Brazil accepted an IMF adjustment program economy with less-than-perfect capital mobility; in November 1998, but not immediately after its sterilization operations do not affect the domestic currency was attacked in October 1997. interest rate when the source of the original dis- 3 See, for example, chapter 8 in Edwards turbance was a change in international interest (1989). rates. Our estimates of speculative pressure with- 4 According to Ding, et al. (1998), a "credit out interest rates were very similar to the others. slowdown" can be defined as a decline in credit 10 Argentina's low ISP may be due to the fact growth, and a "credit crunch" implies a reduction that they do not include changes in the reserves or of credit availability for a given range of interest dollar deposits of the banking system. rates. " We use 1997Q4-1998Q3 for Indonesia due 5 See Appendix A for a simple formalization to the delayed response of its GDP growth rate. of these arguments. For Korea we use 1998 for similar reasons. For 6 In both graphs, quarter number 1 corre- Thailand, we use 1998 due to the lack of quarterly sponds to the quarter immediately following the data. currency crises; that is, 1995Q1 for Argentina and 12 Table 2 shows the initial share of short- Mexico, 1997Q3 for Thailand and Indonesia, and term debt relative to GDP for our six cases. As an 1997Q4 for Brazil and IKorea. The estimates are example, consider two countries. One has a CAB based on quarterly GDP data at constant domestic deficit of 100 dollars, which is fully financed by prices. For Thailand, we use annual GDP growth short-term inflows; the other has a deficit of the rates in Figure la only, due to the lack of quarterly same size, but all of it is financed by loans matur- data. For Korea, the series is based on actual quar- ing after one year. Both countries experience a terly observations through 1998Q3 (i.e., through sudden stop of capital inflows, whereby the 100 quarter number 4 in Figures la and lb), but the dollars of inflows suddenly go to zero. In the case observation for 1998Q4 is based on preliminary of the first country with short-term debt flows, the A TALE OF Two REGIONS? * 41 CAB would have to move to a surplus of 100 dol- consideration of the accompanying policy lars if it does not want to declare a moratorium on announcements has been a major weakness of its debt payments. The second country would only econometric studies such as Kraay (1998). need to have a CAB of zero, because no debt pay- 15 There is some econometric evidence to ments would be due the year following the crisis. support this proposition. Goldfajn and Gupta Hence, while it is true that for a given sudden stop (1998) analyze the effectiveness of tight monetary of capital inflows, the composition of the inflows policy in the aftermath of currency crises for 80 does not affect the magnitude of the internal countries between January 1980 and January adjustment, the composition can influence the 1998. They find that tight monetary policy (i.e., magnitude of the current account reversal. raising interest rates) increases significantly the 13 This is a common problem of studies that probability of reversing a real exchange-rate have attempted to evaluate the usefulness of undervaluation through nominal appreciation "tight" monetary policies in crisis situations. Kraay rather than through higher inflation. However, (1998) is an example of an econometric study of they also find that in economies with weak bank- many episodes of speculative attacks that attempts ing sectors (i.e., countries that simultaneously to show that "tight" money is neither a necessary faced a banking crisis) the opposite result holds: nor a sufficient condition to defend an exchange- tight monetary policies and high interest rates rate parity. But the author relies on crude defini- reduce the probability of a reversal through the tions of tight money: increases in the discount rate nominal exchange rate, and inflation becomes the and reductions in the growth rate of domestic endogenous mechanism that eventually returns credit. The former is imperfect because it is not the real exchange rate to equilibrium. Thus, high the only monetary policy instrument, and the lat- interest rates help defend a currency only if the ter is a weak proxy because it is possible to observe banking sector is strong enough, say, in terms of increases in credit growth with simultaneous capital and/or liquidity provisions, to absorb dete- increases in market-determined interest rates. riorating loan portfolios. 14Another important initial condition may 16 See Perry and Lederman (1998). be the recent history of inflation. Based on 13 17 A more detailed microeconomic analysis episodes of temporarily high rates (i.e., when rates would be required to ascertain the types of firms rose more than 10 percentage points for at least 5 (i.e., tradable versus nontradable sectors) that were days), Furman and Stiglitz (1998) conclude that responsible for the foreign liabilities. An industrial interest-rate defenses tend to be more successful survey of Thailand conducted by Dollar and Hall- in high inflation countries than in low inflation ward-Driemeier (1998), however, shows that a situations Furman and Stiglitz suggest that high large portion of short-term foreign liabilities were interest rates signal the authorities' willingness to held by financial companies with loans issued to contain inflation, but this positive effect is only nonexport firms. relevant for economies with recent inflation prob- 18 See Claessens, Djankov, and Lang (1998). lems. It should be noted that in general, but espe- 19 World Bank (1998) and Perry and Leder- cially in cases with high inflation rates, the effec- man (1998) argue that the buildup of leverage by tiveness of monetary policy cannot be judged in the private sector in Asia was associated with weak isolation, because it is likely that the fiscal policies corporate governance and financial institutions. implemented or announced during speculative 20 Again, Calvo (1998a) points out that a attacks can either reinforce or weaken the signal- given percent reduction in consumption can be ing effects of the monetary stance. The lack of more contractionary than an equivalent fall in 42 * ADJUSTMENTS AFTER SPECULATIVE ATTACKS IN LATIN AMERICA AND ASIA: investment when consumption is more intensive 23 The fact that the Asian economies are in nontradables (or more labor-intensive). more integrated and their exports more similar 21 As far as we know, Edwards (1996, 1998) than is the case in LAC is consistent with the was the first to use recursive estimates of inflation- findings of Calder6n, et al. (1998), who show ary inertia, but he controls for the influence of that the correlation of growth rates of industrial 'excess' money supply production within Asia are higher than within 22 It is possible, however, that merchandise LAC. import volumes do not decline in the face of major 24 Due to the slow response of GDP growth declines in domestic denmand when the adjustment in Indonesia and Korea, we use tlheir growtlh rates is translated into declines in the prices of imports. after five quarters. 4i, THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433 USA Telephone: 202-477-1234 Facsimile: 202-477-6391 Telex: MCI 64145 WORLDBANK MCI 248423 WORLDBANK World XVide Web: http://wwNvworldbank.org/ E-mail: books@worldbank.org ___ I.'I ISBN 0-8213-4541-9