Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Currency War of 2009–11 wikipedia , lookup
Bretton Woods system wikipedia , lookup
Currency war wikipedia , lookup
Foreign exchange market wikipedia , lookup
Capital control wikipedia , lookup
Purchasing power parity wikipedia , lookup
Fixed exchange-rate system wikipedia , lookup
Exchange rate wikipedia , lookup
International monetary systems wikipedia , lookup
Preliminary Draft Comments appreciated February 2004 The Asian Crises Reexamined∗ Thomas D. Willett, (The Claremont Colleges)** Ekniti Nitithanprapas, (Ministry of Finance, Thailand) Isriya Nitithanprapas, (National Economic & Social Development Board, Thailand) Sunil Rongala (The Claremont Colleges) For the Asian Economic Panel Meetings Seoul, Korea October 9-10, 2003. ∗ This paper was written primarily by the senior author but draws heavily on the research in the Claremont Ph.D. dissertations by the co-authors and in subsequent research papers as well as on Claremont dissertations by Aida Budiman, Gab-Je Jo, John Thomas, and Hui Zhang. The policy discussions are the sole responsibility of the senior author and not necessarily reflect the view of the Thai government. Financial assistance for the underlying research from the Freeman Program in Asian Political Economy at the Claremont Colleges, the Haynes Dissertation Fellowship Program at Claremont Graduate University, and the National Science Foundation is gratefully acknowledged as are comments in earlier drafts by participants in the Claremont workshop on International Money and Finance and the Claremont Political Economy Seminar. ** Corresponding author: [email protected] , Horton Professor of Economics and Director, Claremont Institute for Economic Policy Studies. I. Introduction and Overview The series of crises that rocked Asia in 1997 has had a profound impact on how much of the economic profession views the behavior of international financial markets and the determinants of financial vulnerability and has heavily influenced the debate over reform of the international financial architecture. The causes of the crises have been subject to wide differences in interpretation, for example, from the view that most of the crises were due to panic contagion hitting innocent victims to the almost diametrical view that international financial markets were highly efficient and that the crises were due primarily to national government and IMF generated moral hazard. These interpretations in turn have heavily colored the lessons drawn. Given the ongoing influence of these different interpretations on current policy debates, a reassessment of what we have learned about the Asian crises should be well worthwhile. For such a reassessment there are now literally hundreds of studies on which we can draw. In reviewing this literature we find that many of the sharpest differences in interpretation were based on quick and dirty highly oversimplified analyses that were heavily colored by the commentators’ initial preconceptions. The more serious detailed research that has been undertaken does not remove all scope for differences of opinion among reasonable people, but it does narrow the range considerably. For example, there is still considerable disagreement about the extent to which Indonesia was an innocent victim of contagion from Thailand. Our analysis, which rests heavily on the role of wake-up calls and the importance of prospects of political instability on demands to hedge uncovered international financial positions, suggests that the degree of panicked contagion hitting Indonesia was relatively slight, but this is a judgment about which the evidence isn’t clear cut and reasonable people can differ. On the other hand, it is not consistent with the data to argue 2 that the Korea crisis that hit full stride in October was the result of panicked contagion from Thailand. Indeed to the extent that Korea was hit by contagion, this came much more from Taiwan and Hong Kong than directly from Thailand. Thus paying attention to the time dimensions of the unfolding of the Asian crises allows us to reject some of the most popular stories that treat the entire Asian disturbance as one causal chain. This tendency has in turn led to excessive fears of the normal magnitude of pure contagion in international financial markets and resulted in ill-conceived reforms such as the IMF’s ill-fated Contingent Credit Line. We focus on the events in Asia in 1997, but use the plural crises rather than the more common crisis for two reasons. One is to stress that there were both currency and financial crises and these acted to reinforce one another. Secondly, we wish to challenge the widespread view that these was just one crisis strung out over time as countries toppled like dominoes one after the other. We will argue that imperfections in financial markets, both domestic and international, did play an important role in all of the Asian crises, but that this was not due primarily to the panic and contagion that many commentators have alleged. Instead, we argue that the major problems in the operation of financial markets surrounding the Asian crises were not ill-founded speculative attacks, but excessive euphoria prior to the crises by both domestic and international investors. Furthermore, national government policies did as much or more to foster excessive international flows and over heated domestic financial markets as to counter them. Examples include fiscal subsidies such as those that were provided by the Bangkok International Banking Facility, poor sequencing of liberalization such as Korea’s liberalizing short-term before long term capital inflows, disincentives to cover foreign borrowing because of beliefs that the absence of sharp currency depreciations were guaranteed, as well as excessive domestic credit creation 3 and the generation of numerous perverse incentives for lending and borrowing. Nor were there tendencies sufficiently controlled or offset by prudent national and international risk management strategies. A major contributor to the view that the “spread” of the Asian crises was predominantly due to panic and pure contagion was the generally strong macroeconomic fundamentals of the countries in question. This more than moral hazard explains the poor record of international financial markets in predicting the crises. Perhaps the most important lesson to draw from the Asian crises and the Mexican crisis before them is that the important fundamentals include more than the traditional major domestic macroeconomic indicators. To these must be added domestic financial considerations, exchange rates and balance of payments flows, and international liquidity considerations. Contrary to Furman and Stiglitz’s (1998) influential early study, we find that exchange rate and current account positions have important explanatory power for the Asian crises as do the ratios of short term foreign debt to reserves emphasized by both Furman and Stiglitz and Radelet and Sachs (1998). Furthermore once we focus on the combination of exchange rates and current account balances rather than considering them independently, we find that our statistical results are quite robust. Of course, the limited set of variables that we analyze leave enough unexplained that there is still considerable possible role for panic and pure contagion. A look at the timing of the subsequent crises casts doubts on the power of these explanations, however. For example, the heavy speculative attacks on Korea came months after the Thai depreciation. We find considerable evidence that international financial flows don’t manage themselves, but we argue that there has been great abuse of the word “panic” in describing the international capital flows 4 surrounding the Asian crises. We also find that contrary to the emphasis placed by many commentators on the role of unstable portfolio investment, the biggest outflows during the crises were from the banking sector. This casts doubt on the importance of the Calvo-Mendoza model of contagion based on rational ignorance. We discuss how the evidence seems to fit with a number of types of crisis models and hypothesis, including a variety of moral hazard explanations, and attempt to clarify some of the confusion that has been generated in some discussions of second generation crises models. Our focus is on explaining the causes of the crises. While we make a few comments in passing, we do not attempt to systematically analyze either cause of the depth of the crises across countries or the debate about appropriate policy responses. We do discuss, however, the implications of our analysis of causes of the crises for crisis prevention strategies and the roles of capital controls, exchange rate regimes, reserve management, and IMF lending policies. We also find that contrary to numerous pronouncements, careful study of the Asian crises really doesn’t allow us to draw many conclusions about exchange rate policy. It has been common place to put heavy emphasis on the role of pegged exchange rates as a cause of the currency crises. This is surely correct with respect to Thailand with its fixed narrow band peg to a basket of currencies dominated by the dollar. We thus do have further confirmation that Bretton Woods type narrow band sticky adjustable pegs are not compatible with high capital mobility. Countries such as Indonesia and Korea had managed floats, however, and operated these as de facto crawling bands. While this managed flexibility wasn’t sufficient to allow these countries to avoid crises, it’s far from clear that they were a major cause. While the Thai baht had become obviously overvalued prior to its crises, this was not clearly true for either Indonesia 5 or Korea. Estimates of equilibrium exchange rates for these countries have varied substantially, with some arguing that the won had been seriously overvalued for some time, while others argued that it was close to equilibrium of even a little undervalued. Thus we are doubtful that excessive exchange rate management was a cause of serious currency overvaluation for all the crisis countries. Likely more important were the effects of these high managed regimes in encouraging uncovered foreign borrowing. These in turn contributed greatly to the depth of the crises once they were triggered. Our statistical analysis shows that depending on whether these intermediate regimes are classified as pegged or floating, one can find any conclusion one wants about the role of pegged rates in the crises. The conclusion we draw is that simple two-way classifications of exchange rate regimes into pegged and flexible are too simple and no reliance should be placed on any study using a simple schema. Greater differentiation among types of regimes is needed. We discuss a number of the issues involved in developing such richer classifications. This has become the subject of a number of studies in recent years. Again, we find that the Asian experience does not offer enough observations to give us much insight into the breath of the unstable middle of intermediate exchange rate regimes. Our analysis also raises questions about how much the Asian experience tells us about the role of capital controls in stimulating or avoiding crises1. It is often argued that capital controls allowed China and India to escape the Asian crises. It is true that these countries did have substantial controls and that they were not hit hard by the crisis, but the causal connection between these two facts is not so clear. According to our statistical analysis, both China and 1 We do not discuss the issue of Malaysian capital controls in this paper. For more detailed studies on this issue, see Kaplan and Rodrik (2001) and Cook and Devereux (2002). 6 India had a sufficient combination of high reserves and strong fundamentals so that they wouldn’t have been hit even if they hadn’t had controls. Like several other recent studies, our statistical analysis finds a positive rather than negative correlation between capital controls and currency crises. These results strongly conflict with the popular view that premature liberalization of capital controls was a major cause of the Asian crises. In reviewing the quantitative measures of capital controls used in recent studies, we were surprised to find that most of the studies classified Asian crises countries such as Thailand and Korea as high control countries. This hardly fits with most judgments. Much of the problem comes from using simple zero-one classifications of capital controls. As with exchange rate regimes, we believe that absolutely no weight should be given to the results of studies based on such simple classifications. Fortunately, for both capital controls and exchange rate regimes, we now have available several datasets that give finer classifications. We review these capital controls measures and find that unfortunately they also seem to have an upward bias in their classification of the degree of controls of the Asian crises countries. Comparing these datasets with our judgmental assessments of the degree of pre-crisis capital controls, we conclude that by far the most accurate dataset is that developed by a team at the IMF led by Barry Johnston. Unfortunately, however, it is available for only one year. The new thirteen point classification now published by the IMF as a result of Johnston’s study is a substantial improvement over the IMF’s old zero-one classification, but comparison of it with Johnston’s fuller index finds that even the thirteen unit classification has a substantial upward bias. In short, we now have better, but still not terribly good quantitative measures of the tightness of countries’ capital controls. Until we get better measures that distinguish between controls on inflows and outflows, it is going to remain difficult, if not impossible, to learn much 7 from large N statistical studies in this area. Given the importance of capital controls issues, there is a great need for improved data. The IMF seems to be the obvious institution to lobby. We begin our review of the debate over the causes of the Asian crises with a discussion of panic and contagion then turn to moral hazard and perverse financial liberalization before to the more traditional macro economic fundamentals. We then consider the roles of capital controls and exchange rate regimes. We close with a discussion of policy implications, including the need to reform IMF lending practices. II. Panic and Contagion An important lesson of the Asian crises is the need to think more precisely about the meaning of panic, stabilizing versus destabilizing speculation and the distinctions between liquidity and solvency problems. There can be little question that there was a strong element of liquidity crises in the Asian crises and undoubtedly there was some panic, but careful consideration of the evidence rules out the hypothesis of the spread of the Asian crises from Thailand being a pure panic or liquidity crisis. Yet this description is widely used. Perhaps because of a desire to sound dramatic and/or suspicion of markets, political scientists writing on the Asian crises have tended to emphasis the role of panic and of foreign investors in generating contagion. Consider a few quotes from one recent collection of papers on the crisis by Horowitz and Heo (2001)2. The discussion of Korea by Heo (2001) refers to the “foreign exchange panic” (p. 151) and puts its focus on the “exodus of foreign capital” (p. 157). Likewise, Kimberly Niles’ (2001) analysis of Indonesia also stresses panic, “…investors confidence in Indonesia was based on assumptions that high capital inflows would continue and the exchange rate would 2 Other examples may be found in Pempell (1999). 8 hold…Once Thailand devalued, investors in the region and pulled out of other Asian countries in a panic…” (pp. 119-120). Niles does, however, give a more balanced treatment of the role of foreign investors versus domestic residents…“domestic firms were quick to panic and short the rupiah at the first sign of political uncertainty. There is ample evidence that domestic firms were at least as culpable as foreign currency speculators in the rupiah’s collapse. These domestic firms probably had better day-to-day knowledge of rising political vulnerabilities than foreign currency traders had” (p. 120). Niles also usefully stresses the importance of political factors and how failure to consider them sufficiently had contributed to the over optimism of investors. “The initial misreading of the situation was due to a failure to recognize the role of political actors and institutions in exacerbating the economic crisis…” (p. 111). The importance of political factors has not escaped perceptive economists as well. For example, Takatoshi Ito (2000, p. 294) has argued, “Indonesian problems cannot be understood without investigating political and social shocks.”3 Many economists have also adopted the panic description of the recent crises. Philippe Delhaise (1998) describes “The panic that swept through the region like a brush fire or a contagion epidemic devastated the currencies and equities well beyond what was reasonable” (p. 2) although he goes on to argue that “the panic did have its origins in a crisis of fundamentals” (p.2). Other examples are given by two book length treatments recently published by major university presses, Desai (2003) and Rakshit (2002), as well as the early analysis by Furman and Stiglitz (1998) and Radelet and Sachs (1998). Desai writes “The fall of the baht set off…a full blown panic attack by speculators on the Malaysian, Indonesian, and Philippine currencies” (p. 3 This is also noted by Kawai, Newfarmer, and Schmukler (2001). 9 121). Unlike many writers, however, she notes that the South Korean won was an exception to this “panic.” Later he refers to “irrational casino fever” (p. 245). Rakshit also adopts a non-fundamentals view, arguing, “…neither the emergence of the crisis nor its spread and depth can be reasonably explained in terms of the ‘fundamentals’ emphasized in the literature” (p. 85). Earlier, however he had noted that “Economists have identified two prominent sources of the Thai crisis…stagnation of export earnings … (and) the growing manifestation of financial troubles, as non-performing assets of banks and other financial institutions recorded as sharp rise” (p. 56). Indeed, we believe that most economists who have studied the Asian crises agree that the cause of the Thai crisis are over determined, with both external imbalances from an overvalued currency and internal financial imbalances being sufficient causes. Indeed the main issue of depute is primarily whether the resulting “twin crises” spread more from currency crisis to domestic financial crisis or vice versa. (For examples of initial differing emphasis contrast Corbett and Vines (1999) with Willett (2000)). Kawai, Newfarmer, and Schmukler (2001, p. 23) argue that “In Korea, investors simultaneously came to note the risk of large short-term external debt ... And began to sell the won”, but then go on to assert that “any individual investor could ignore the stampede only at his peril” giving the impression that it was herd behavior rather than rational calculations that were stimulating the outflows. This impression is strengthened by references to “investor panic”. Another frequent characteristic of the panic view is to focus primarily on the role foreign investors. An example is given by Park and Wang (2001) where they argue that “the East Asian crisis should in a large part be attributed to the panic reaction and the herd behavior of foreign investors rather than to a deterioration in fundamentals. For this reason, the Thai crisis was found to be much more contagious than otherwise thought.” Much less attention has been focused on 10 the behavior or Asian residents, yet there can be little question that their (in our view quite rational) scramble to close out open foreign currency positions led to substantial capital outflows. Makin (1999) argues that “ the panic decisions of resident East Asians investors to suddenly withdraw fund from their own financial institutions – the classic ‘bank run’ scenario – turned what could have been a sharp, yet orderly, correction of asset prices into a full blown financial crisis” (p. 411). He goes on to note that these withdrawals may well have been rational. Macleod and Garmout (1998) describe the developments in Indonesia as “In Indonesia’s case…the sudden float and devaluation of the baht in neighboring Thailand, and then the ringgit in Malaysia, triggered a re-evaluation of risk exposure of all kinds which, in turn, brought on precisely the changes in asset and liability values that investors feared.” They go on to further say that “The first such re-evaluation is Indonesia related to exchange risk. Private sector foreign debt was of the order of USD 70 billion, 83% of it unhedged…borrowers began to buy dollars in early July 1997, as did domestic and foreign entities holding rupiah denominated deposits and other financial assets.” K, N, & S (2001) argue that “with the possible exception of Thailand, foreign investors apparently did not play a large role in triggering the crises” (p. 13) and while they make many references to panic and contagion, they, as do we, express a sympathy with the wake up call interpretations of the spread of the crisis. They suggest, “The Thai crisis may have led global investors and bank creditors to revise the model of East Asian development and to reduce their exposure in the region.” (p. 21)… “Uninformed investors reevaluated the risk associated with Indonesia, Malaysia, and Korea and when they saw problems arising in Thailand and decided to pull their investments out of neighboring countries” (p. 13). This view, to which we subscribe, is far from consistent with standard high information efficient market theories, but it is also quite 11 different from stories of blind panic. We have no doubts that there was some element of psychological panic at work, but we are doubtful that this was a primary explanation of the major speculative attacks and runs for safety in Asia during 1997. Barry Eichengreen (1999) is skeptical of the wake up call argument and suggests that it simply begs the question “… Rarely is an effort made to explain why this wake up call was so loud and startling,” (p 166) which is a highly appropriate comment with respect to arguments based on efficient markets views. From our behavioral finance perspective, however, there is a simple answer. The effects of the wake up call were so large because quite a large proportion of market participants had been asleep or living in a dream world where miracle economies were assumed not to be subject to the problems of mortal economies. And this is indeed what Morris Goldstein (1998) had in mind when he coined the term. Indeed he specifically points out that “I refer to it as a wake up call because from most market indicators of risk, private creditors and rating agencies were asleep prior to the outbreak of the Thai crisis.” (p. 19). Most of the countries eventually hit by serious crisis had large unhedged short-term debts. The lack of hedging largely reflected government generated expectations that large exchange rate changes would be avoided. The substantial depreciation of the baht was a “wake up call” that led to substantial reevaluation of this assumption by economic agents in other Asian countries. It did not require panic, only prudent risk management, to scramble to cover one’s foreign liabilities in light of this “news.” Such covering, in turn, generates capital outflows which in turn generate pressures in the foreign exchange market and can look like speculative attacks, even though they are caused by increased hedging rather than overt speculation. Likewise, even though most of the lending by foreign banks was denominated in foreign currency and hence immune to pure exchange risk, 12 many borrowers were sufficiently exposed that a sharp depreciation could seriously erode their balance sheets. Thus the currency risk of the borrower was transformed into the credit risk of the lender. This international nexus interacted with substantial shifts in expectations about the domestic basis of the credit worthiness of many Asian corporations and financial institutions. One does not need to resort to panic to explain how shifts in expectations would lead to a substantial change in net capital flows. Since most of the crisis countries had large current account deficits matched with large capital inflows, a drying up of capital inflows, much less an actual shift to outflows, would make currencies that had been initially valued at roughly equilibrium rates, such as arguably was the case with the Indonesian rupiah and the Korean won, quickly become substantially overvalued4. This process of self-fulfilling shift in expectations hit Indonesia in July and August and Korea in October-November. At least with Korea there was no question of national insolvency. Thus from the aggregate level, the problem could be seen as being primarily a liquidity crisis. However, from the standpoint of many individual agents there were quite reasonable concerns about solvency as well. Tirole (2002) argues, “There is never illiquidity without at least some suspicion regarding insolvency” (p. 111). We wouldn’t want to go so far as to argue that there can never be a pure liquidity crises, but real world crises involve varying combination of both, and in our judgment the Asian crises fits this description. On this interpretation, the virulence of the Asian contagion was due to the large number of countries with financial sectors in the vulnerable zone. Here second generation crisis models can give considerable insight. 4 See Corden (2002). Kwon (1998) on the other hand argues that “Government interference in the foreign exchange market and the substantial overvaluation of the Korean won were causes of the financial crisis” (p. 337). 13 There has been some controversy about the extent to which second generation crisis models fit the Asian episodes, however. 5 We argue that some of the disagreements can be eliminated by emphasizing that second generation models make two important innovations. One is in switching from a passive to active modeling of government behavior. Thus expectations about government reaction functions become an important aspect of the speculative calculus. This aspect is emphasized in Furman and Stiglitz’s argument that second generation models don’t fit the Asian crisis. The second innovation is the possibility of multiple equilibrium, which opens the door to the possibility of rational self-fulfilling speculation. Shifts in expectations about government reactions are one, but not the only possible cause of such shifts in expectations. Furman and Stiglitz (1998) were quite right that the causes of the Asian crises was quite different from those of the European monetary crisis in the early 1990s that the second generation models were originally developed to explain, but their dismissal of the relevance of second generation models to the Asian crises is misleading. The key to understanding this is recognizing the two important innovations that the second-generation models made. One, on which Furman and Stiglitz focus, is the introduction of a government reaction function. Furman and Stiglitz correctly point out that the Asian crises countries were hit so hard that they didn’t really have a choice not to devalue (Hong Kong and Taiwan are important exceptions). Thus the market’s expectations about government reactions to speculative pressures didn’t play as important a role in Asia as they did in Europe. However, this needn’t imply that the multiple equilibrium facet of second generation models wasn’t still at work. We believe that it was, with 5 For surveys of first and second generation crisis models, see Eichengreen (1999, Appendix ___), Krugman (1998) and Rajan (2001). 14 wake up calls about financial positions that were much more illiquid than previously recognized and in many cases insolvent as well. The second important innovation of the second-generation models was highly relevant to Asia, however. This is the role of multiple equilibrium. In the first generation models fundamentals came in two flavors – good and bad – and the only role for financial markets in the case of bad fundamentals was to determine the timing of the speculative attack. Second generation models make the much more realistic assumption that fundamentals can also come in a third, in between flavor where a country’s situation is such that it is vulnerable to a crisis, but not so bad that a crisis is inevitable. With good luck they’ll make it through. With bad luck they won’t. Good and bad luck can come in the form of shocks or changes in perceptions or arbitrary changes in the “animal spirits” and moods that may at times govern financial markets. The crisis models, to date, do not explicitly address the causes of shifts in expectations. Thus as Morris and Shinn (1999) stress, they are incomplete theories. It is perhaps unfortunate that many economists have adopted the convention of referring to positions in this vulnerable zone as sunspot equilibria, suggesting that shifts from a “good” to a “bad” equilibrium are arbitrary. While the formal models allow any cause such as animal spirits or sunspots to shift market expectations from a “good” to a “bad” equilibrium, we argue that in practice such shifts in expectations are usually based on reasonable interpretations of new information. What is important here is not that actual fundamentals have necessarily changed, but that perceptions of, or information about them, have changed. In the Asian crises there was a long series of such bad news showing that various financial positions were much worse that had been thought. In our interpretation the speculative inefficiencies were not so much in overreacting to news and rumors 15 after the crises began, but in not paying enough attention to signals and lack of good information prior to the crises. For example, K, N, & S (2001) incorrectly state that “The second generation models of currency crises…..argue that currency crises can occur even in countries with sound fundamentals owing to self-fulfilling expectations.” (p. 1)6. Likewise, in a major survey paper, Krugman (1998) appears to equate self-fulfilling speculation with unjustified speculation. As was emphasized by Maurice Obstfeld (1998), one of the originators of second-generation crisis models, in his comments on Krugman’s paper, this is not the case. In these models the fundamentals limit the range over which arbitrary shifts in expectations can shift equilibria through self-fulfilling speculation. To us, the puzzle is not that there were major shifts in expectations, but why these didn’t begin to occur sooner. Such behavior is consistent with the concept of confirmation bias emphasized in the literature on behavioral finance. Note also that second generation models don’t support the often-expressed view that destabilizing speculative capital flows overwhelmed even countries with strong fundamentals7. In second-generation rational models, speculative attacks only occur in countries with poor or intermediate fundamentals. This is important to emphasize because even some top economists have slipped into implying that because speculation can be self-fulfilling, even countries with strong fundamentals could be subject to attack in second-generation models. In our view many of the crisis countries fell into this intermediate category. They weren’t entirely innocent victims but they were also victims of bad luck. The unusually large number of Asian countries in this 6 An example of a similar statement from a political scientist is Leslie Armijo’s (1999) comment that “The Asian financial crisis showed that even ‘well behaved’ countries such as Thailand, Indonesia, and South Korea could suddenly be attacked by huge credit outflows.” (p. 15). 7 Stress that second generation models are also not consistent with the influential model developed by Sachs, Tornell, and Velasco (1996) to explain the Mexican crisis. In the STV model, high reserve levels can fully protect a country from weak fundamentals. In second generation models high reserves can protect a country with intermediate fundamentals, but not one with weak fundamentals. We find that the Mexican and Asian crises do not offer enough degrees of freedom for us to clearly discriminate between these two views. This will require a larger dataset. 16 vulnerable zone helps explain why the number of crises that were generated after the Thai crisis was followed by more crises as compared to what happened in Central and Eastern Europe after the Czech crisis of the same year. The greatest substantive area of disagreement is about the causes of the crises in the other Asian countries after Thailand was hit. Here is where the pure contagion and innocent victims views are most frequently expressed. Radelet and Sachs (1998) expressed the view that if Thailand had not fallen, the rest of Asia would have escaped crises. There is evidence that at least some investors view Asia as a single entity and had little knowledge of the differences across countries. For example, Delhaise (1998) argues “The transmission belt of the crisis was perception…international lenders and investors wrongly perceived as a single market” (p. 12). Even as astute an observer as Barry Eichengreen (1999) argues that it is “hard to see what the countries hit by contagion had in common other than physical proximity” (p.151). A deeper look, however, suggests that while the whole region felt some effects from the Thai crisis, the strength of these effects were strongly differentiated.. Furthermore, both the statistical analysis presented later and our quantitative discussions will suggest a rough correlation between the strength with which countries were hit and the magnitude of their problems. This would be largely missed, however if analyses were limited to traditional macroeconomic variables. From a pure macroeconomic perspective Radelet and Sachs (1998) are quite right that “Indonesia appears to be the clearest case of contagion in the region.” For Indonesia, however, political instability was the key problem and as Haggard and Macintyre (2001) emphasize “Uncertainty about the future policy environment can be a trigger for financial crises…” (p. 57). 17 Likewise, while some Korean economists have argued that their country was an innocent victim, and its macroeconomic variables were certainly enviable, Krueger and Yoo’s (2002) careful analysis of the Korean financial sector concludes that “the vulnerability of the system was extreme” and “the Korean crisis was a disaster waiting to happen” (p. 638). Finding rough correlation between broader fundamentals and the patterns of crisis, of course, doesn’t rule out that some degree of indiscriminate contagion may also have been at work. It does, however, support the view of Takatoshi Ito (2000) that we need to look carefully at each country since the causes of crises were not uniform. It should also be noted that not all of the Asian crisis countries had been free of at least mild macro economic problems. Inflation was definitely not a problem, but as will be discussed later many countries were allowing excessive credit creation and in some cases such fiscal policy was judged to be too loose. For example, in their careful examination of Indonesia, Korea, Malaysia, and Thailand, Alba et al (1999) argue that “all four countries saw a pickup in domestic demand pressure during 1994-96. The demand pressures were manifested primarily in a widening of the current account.” (p. 35). They go on to argue that “while monetary policy was being tightened, fiscal policy actually turned pro cyclical in all four countries.” (p. 41). More popular, than the hard work of careful analysis are much more simplistic analyses that, at their worst, treat the whole rash of crisis in the late 1990s as a single process of spreading contagion. As will be discussed further below, this tendency to fail to distinguish between relatively mild ripple effects in currency and financial markets and major speculative attacks contributes to the tendency to exaggerate the role of pure contagion. For example, Noble and Ravehill (2000) argue “The contagion effects of the crisis spread to the financial systems beyond East Asia – first Russia and then Brazil suffered massive capital outflow.” (pp 1-2). Likewise, 18 Rakshit (2002) states that “…by October 1997, the crisis had acquired a global dimension, with practically all economics in the world suffering a meltdown in their share prices and/or exchange rates” (p. 75). This is greatly exaggerated. If “suffering a meltdown” were replaced with “had felt some effects” the statement would have been far less dramatic, but far more accurate. Close observers generally accept that the Asian “crisis” had at least two distinct phases, one following the Thai depreciation in July 1997 and then hitting primarily Indonesia, Malaysia, and the Philippines, and a second following Taiwan’s depreciation in October that helped spark attacks on Hong Kong and South Korea. Many commentators, however, lump everything into one Asian crisis caused primarily by panic contagion. And some see the whole set of crises in the late 1990s in terms of a fall of dominos. Habits of expression can contribute inadvertently to this image just by talking in terms of the spread of crisis – which implicitly suggests causal connections. Thus, for example, Blejer and Škreb (2002) in discussing crises in emerging markets say “It started with the Mexican crisis in 1994-95. After Mexico, the crisis traveled a long way, slowly, as it moved to the other side of the globe to Southeast Asia in 1997. Starting with the fall of the Thai baht in July 1997, the crisis rapidly spread from Thailand to Indonesia, Malaysia, South Korea, and (to a lesser degree) the Philippines….From Southeast Asia the crisis spread, with a time lag of one year, to Russia and later on to Latin America…(and) Turkey.” (p. 2). Blejer and Škreb do not offer a pure contagion view. Indeed they argue “neither the character nor the causes of all these events is the same” (p. 3), but use of the “spread” terminology can easily yield the image of strong causal connections. We should hasten to add that we would not argue that there were no causal connections. For example, the real effects of the Asian crises contributed to a slump in the price of raw 19 materials which in turn hurt Russia’s export proceeds. This is a very different channel from panic contagion, however. Furthermore, the fall in demand for Russian exports, while not trivial, was certainly not the major cause of the Russian crisis. Nor can pure herding panic could hardly explain the time lags in the attacks on Hong Kong and Korea. In fact it can’t even fully explain the pressures on Malaysia and Indonesia. As Rakshit (2002) argued “The Thai crisis spread fairly quickly to the Philippines. By contrast, during (July)…the extent of depreciation and fall in share prices were relatively mild in Malaysia and Indonesia. In fact, until the second week of August the Indonesian stock market remained immune to the currency crisis and the depreciation of the rupiah was also quite minor” (p. 100). Rakshit elaborates on his more nuanced view, developing a three-stage depiction of the crisis. In the first stage, which he puts as lasting about a month “the rest of Asian economies remained relatively unaffected” (p. 102). What Rakshit refers to as “The lag in contagion” (p. 114) cannot plausibly be explained by pure panic. It is also difficult to explain with the information based models of rational herding that have been developed by Calvo and Mendoza (2000), and others. Part of the tendency to emphasize panic may be due to the common tendency to refer to financial crises as financial panics. Radelet and Sachs (1998) use this terminology arguing that “Each of these episodes (of the Asian crises) displays elements of a self-fulfilling crisis in which capital withdrawals by creditors cascade into a financial panic” (p. 1). They go on to note that “the panic may be rational on the part of individual creditors…” (p. 1). Their discussion of rational panic, while not without an analytic basis, invites confusion with the common use of panic in psychological terms such as the dictionary definition of panic as “a sudden overwhelming fear that produces hysterical or irrational behavior, and that often spreads quickly through a group of persons or animals,” Random House College Dictionary (1998, p. 961). 20 Radelet and Sachs and George Soros (1998) both talk of international financial markets as being prone to wild swings but while Soros posits to irrational emotional swings from over optimism to excessive pessimism, Radelet and Sachs’ argument that “international financial markets are intrinsically unstable” (p. 2) is based on theories of liquidity crises based on rational bank runs in the absence of a preventive institutional structure. This argument that due to information asymmetries and collective action externalities a competitive banking system cannot be relied upon to manage itself is accepted by a huge majority albeit not by every monetary and financial expert. Radelet and Sachs’ argument that the current international institutional environment is much less adequate to overcome these problems than the typical domestic system, points to much more focused policy prescriptions than the much broader Soros view that all (or at least most) types of international financial markets are inherently unstable. Rudiger Dornbusch (2000) in his discussion of one version of the Calvo-Mendoza analysis, argues “Here is a theory of speculative attacks caused by masses of investors who find it far more profitable to run away that to ascertain whether rumors are true. It is an uncomfortable conclusion but not altogether an implausible one, since the world does appear to warmly welcome emerging market assets one day and then, on a sheer rumor, desert those assets at the drop of a hat” (p. 41). Once the crisis was in progress there is no question that rumors played a role and that some of them were false. Furthermore, markets may have displayed exaggerated responses to such “news.” For example, Rakshit argues that once the crisis was underway, “participants in the market tended to be highly risk averse and their expectations revised disproportionately in response to new information or policy announcements” (p. 107). False rumors did not generate the Asian crises, however, nor is it easy to think of any other recent currency crisis that was generated by false rumors. 21 This led to markets in the “boom” stages to discount negative news – as a result of the tendency toward “confirmation bias” emphasized in the newly developing literature on behavioral finance that draws on the psychology literature. Thus markets waited too long to recognize the deteriorating financial situation in a number of Asian countries. We believe that there are strong reasons to reject the standard economic assumption of farsighted ideally efficient financial markets, especially when dealing with developing countries. By and large, however, critics of these assumptions have been rather cavalier about the alternative assumptions they suggest. There is a substantial need for more systematic research on international financial markets that adopt the behavioral finance perspective that such markets are neither wildly irrational nor fully efficient. (For an initial effort along these lines, see Willett (2000)). There is indeed important evidence of market overreaction, but not as initiators of crisis. We have argued that the lags in the spread of the crisis is hard to square with either simple panic theories or the rational ignorance theories of portfolio investment instability developed by Calvo and Mendoza. Contrary to the excessive contagion view, we see as more important the harshness of the punishments meted out by the market to vulnerable countries. In our interpretation the Asian crises were unusual because of the combination of the sharp increase in risk aversion generated by shattered mental models (see Denzau and North (1994)) and the large number of countries in the vulnerable zone. As Willett (2000) has argued, the largely unexpected nature of the crisis generated a sharp increase in uncertainty as investors realized that this understanding of the situation had been seriously faulty. Given the vulnerability of a large number of countries domestic and international financial positions, these shifts in expectations were accompanied by substantial exchange rate overshooting in most of the crisis countries. 22 We should openly admit that in discussing the shift in expectations following the Thai crisis as resulting from broken mental models or wake up calls, we have used the terminology least at variance with traditional efficient market views. In truth it may be misleading to dignify some of the views shaping discussions to invest in Asia as being based on mental models. In many cases, these “models” seem not to have been more sophisticated than one shouldn’t miss out on the Asian miracle and that since everyone else was doing it, it must be a good idea. The actual information level of many portfolio managers appears to have been quite low. It is very hard to get a clear overall picture of the decision making processes that led to the huge capital inflows into Asia prior to the crisis and undoubtedly there were a wide range of rationales, including the quite sensible one of international diversification, as well as degree of information, but discussions with a few market participants is sufficient to make it clear that we cannot rule out the hypothesis that quite limited information and highly simplified mental models helped to contribute to the size of the large pre-crisis capital inflows. Indeed to the extent that there was herding behavior, we believe that it’s quite likely that this was stronger during the inflows period than after the crisis hit. Lamfalussy (2000) says “harmonizing…views through osmosis … traders and dealers are a remarkably talkative lot, and so, in a more dignified way, are analysts, bank economists, investment bankers, and even CEOs of major banks” (p.78). He goes on to note “the capacity of emerging markets to absorb capital inflows in an orderly way just does not keep pace with the potentially explosive increase of such inflows” (p.97) and that “in the end these periods of euphoria were at the heart of process which lead to the trouble.” (p.165). This view can also help explain why the crises were to such a large extent unanticipated in market prices. (Another possible explanation, moral hazard, will be discussed in the following section). We believe that its no coincidence that most of the good news coming out of Asia in the 23 pre-crisis periods were in easy to obtain macroeconomic indicators and that most of the bad news involved much harder to obtain and analyze information such as the rise of the bad loans and other financial sector weaknesses. Thus it is not easy to dismiss the description of the Asian crises as being in large part of the pricking of a speculative bubble. As Krugman (1998) has argued, there is “an overwhelming array of direct evidence that suggests that foreign exchange markets do not make use of all available information.” (p. 366). He goes on to argue that it is “hard to avoid the suspicion that financial markets were simply myopic in the run up to both the ERM and the Mexican crises…this conclusion wreaks havoc with all of the currently popular models.” (p. 375). A large number, but far from all market participants anticipated the loosening Thai crisis by late 1996 or early 1997, but for the other Asian crisis countries, we believe that Krugman’s judgments for the earlier crises needs no amendment.8 Our tentative judgment is that both the bubbles and second generation models have explanatory power and that more attention needs to be given to their interrelations. Crisis modelers have begun to pay a good deal of attention to bank run models, but we aren’t aware of similar attention being paid to bubble models in this context. Another apparent lacuna involves the role of shifting expectations once a crisis has begun. In the formal crisis models, the focus is on the initiation of the crisis. The determinants of restoration of confidence and what it takes to climb out of bad equilibria are at the heart of practical policy making and are topics that deserve greater attention. 8 There is evidence of much more anticipation of the latter crises in Argentina, Brazil, and Turkey. In the Russia case, there were widespread expectations of current depreciation but not of default. Thus the Russia default in another example of broken mental models. 24 III. The Roles of Moral Hazard and Perverse Financial Liberalization One possible explanation for the failure of the market to display more anticipation of the risk of crisis is moral hazard. If your investments are fully guaranteed, then even if you anticipate a crisis, as long as you’re earning a good rate of return, then there’s no incentive to move your money out. In this view, especially popular on the right, financial markets were efficient, but their proper functioning was undermined by the prospect of government intervention. Few topics have generated more heat in recent financial discussions than that of moral hazard. It can be a confusing topic because there are in fact a number of different types and disputants have often talked (or shouted) past each other. Let us try to put a perspective on this debate. There is little question that moral hazard generated by the action of Asian governments played a major role in the development of financial vulnerability in these countries. Furman and Stiglitz (1998) are quite right that, unlike the traditional Latin American style crises due to huge budget deficits and monetary accommodation, the Asian crises was about private sector decisions gone wrong. However, this observation could be consistent with fully efficiently functioning financial markets where decisions that are bad from a social perspective result from rational, well-informed decisions by private actors who are responding to distorted incentive structures. Prospects of government protection against losses can generate excessive risk taking and reduce the incentives for careful evaluation of lending and borrowing decisions. Such moral hazard considerations have not traditionally been included among the fundamentals analyzed in the balance of payments and open economy macro models taught in our graduate schools, but that is changing rapidly as a result of the recent rash of international financial crises. Indeed in a recent discussion of fundamentals versus panic views of these recent crises, Tirole (2002) argues 25 that “the dominant ‘fundamental view’ emphasizes government bailouts” (p. 43)9. In our view Tirole’s statement is too strong, but it certainly indicates the increased attention being given to this topic by economists. There was certainly a lot of poor risk management by both Asian crises country firms and by industrial country lenders, but perverse incentives generated by governments also played a major role. Government-fostered beliefs that major exchange rate changes would be avoided generated large unhedged foreign borrowing; not only did these countries tend to have poor oversight systems for regulating prudential risk, but in some cases such as the Bangkok International Banking Facility, foreign borrowing was actually subsidized. In Korea, contrary to standard economic theories of sequencing financial liberalization, short-term foreign borrowing was liberalized before long-term borrowing. Indeed in separate work, the senior author has argued that perverse financial liberalization lay at the heart of the Asian crises 10 . By perverse liberalization we mean liberalization that either generates or gives private sector actors greater scope to respond to incentive structures that give distorted signals from the standpoint of overall efficiency. Three of the most common examples of such perverse liberalization are the bad incentives for borrowing, lending generates by moral hazard and cronyism, and inappropriate sequencing of financial liberalization. The problems of effective allocation of capital were magnified by the size and speed of the financial flows into Asia in the 1990s. As Delhaise (1998) describes it “The sheer size of money awaiting opportunities to partake in the phenomenal growth of Asia was vastly superior to what Asia could comfortably swallow…Asia was literally swamped with money.” (p. 9 Haggard (2000) refers to holders of this views as the new “fundamentalists”. See Auerbach and Willett (2003). 10 26 16). “The capital directed at Asia was never priced efficiently.. capital was diverted from its most efficient uses.” (p. 17) While there is considerable dispute in the literature about optimal sequencing, there is a good deal of consensus about many things to avoid and the Asian experience offers many textbook examples of these. As mentioned above, one is the liberalizing of short term before long-term capital accounts in Korea. A second is the fiscal incentives for capital inflows offered in Thailand by the Bangkok International Banking Facility; to varying degrees, in all of the liberalizing countries there was inadequate attention paid to the infrastructure of law, prudential regulations, and experienced professionals needed to make financial markets work well, combined with widespread domestic moral hazard due to implicit or explicit government guarantees. Furthermore, implicit guarantees against large exchange rate movements in a number of countries generated incentives for excessive unhedged foreign borrowing. To some extent such perverse liberalization was the result of policy mistakes that could have been avoided by better analysis. But much more important in our judgment were the roles of rent seeking by important private sector and government actors and the influence of prevailing doctrines of development. Thus we largely agree with Eichengreen (1999) that the causes of these perverse liberalizations were not primarily “incompetence but the logical outgrowth of the government cultivation of a bank-centered financial system”. (pj. 159). To save the fully efficient version of international financial market behavior, moral hazard incentives would have had not just to be widespread, but all-encompassing. Otherwise, lending that is not likely to be blessed with government bailouts would begin to dry up while bailout subject funds would continue to flow in. As both Radelet and Sachs (1998) and Willett (1999) have demonstrated, such sharp differentiations in the patterns of different types of capital 27 flows are not observed prior to the crisis. Certainly, stock market investors had no reason to believe that they would be bailed out in the event of crisis and indeed they were not. Indeed the Institute of International Finance has estimated that in the Asian and Russian crisis, international investors lost over $300 billion. (Cited in Council on Foreign Relations (1999b)). One popular variant of moral hazard arguments puts blame on the IMF and US Treasury for having increased moral hazard through the Mexican rescue in 1995. The prospect of such large international bailouts could on the one hand make governments more lax in their efforts to avoid vulnerable positions and make international investors more willing to provide finance. Given the substantial economic and political costs typically incurred by crisis countries, the effects of prospective international bailouts on government behavior seem likely to be small. In effect, the insurance offered is one with a large co-payment. With respect to private sector expectations, it isn’t at all clear ex ante that the Mexican experience should have given rise to expectations of large bailouts in Asia. The strength of the Congressional opposition generated in the US and the European operation in the IMF could plausibly have led market actors to expect that future bailouts would be smaller, not larger, and indeed one can make a strong argument that the initial IMF program for Thailand was too small rather than too large. Nor if moral hazard were all-pervasive would we have expected such large reversals in capital flows when the crisis hit. Much of the continuing disagreement among international financial experts about the importance of moral hazard stems from a failure to clearly distinguish between the proposition that moral hazard exists and is of substantial quantitative importance versus the proposition that moral hazard was such a dominant factor that without it the recent international financial crises would likely not have happened. Thus the facts clearly demonstrate 28 that some aspects of moral hazard were quite important, but that the variants of moral hazard influence cannot provide a full explanation. As Hutson and Kearney (1999) suggest, “the suddenness of the withdrawal of the funds suggests that creditors did not perceive any guarantees.” (p. 406). More sophisticated versions of the moral hazard argument have been put forward by Michael Dooley (2000a, 2000b) and others which focus on calculations by investors of the backing available by governments’ explicit and implicit guarantees. In this view, sometimes called the third generation of crisis models, the prospect of IMF bailouts increases the amount of capital inflows, but since neither national nor IMF resources to back guarantees is unlimited, if contingent liabilities become too great, then capital flight will be generated. Such models do make important contributions to our understanding of recent international financial crisis, but there has been a tendency to exaggerate the extent of their explanatory power.11 Formal testing for the importance of various types of moral hazard is extremely difficult because it is often not possible to derive predictions that are not equally consistent with other hypotheses. For example, if interest rates fall after an IMF program is announced this would be consistent with the generation of moral hazard, but it would be equally consistent with the 11 Dooley (2000) presents a number of facts that he argues are consistent with his model, but he does not clearly distinguish between arguments that his model has some explanatory power and that it has dominant explanatory power. One falsification of the complete explanation version is reported by Dooley himself. He notes that Indonesia had reserve coverage of foreign debt of 116% and that this is an outlier (the other Asian crises countries on his calculations having ratios averaging well below 100 percent). What he does not explicitly state is that with reserve coverage above 100 percent, according to his model a country should not suffer from a crisis. Several other empirical implications discussed by Dooley also fail to fit the data well. He describes the provision of external financial assistance to support liberalization programs as “the second important exogenous source of insurance” (p. 264), but in his sample of countries, Russia was the only one to receive financial assistance from the IMF before a crisis hit. Dooley argues that another implication of his model is that “crises will be spread over time and move from poorly to well-regulated financial systems” (p. 265). This seems a quite plausible conjecture, but again it does not seem to fully fit the facts. It is hard to argue that Indonesia and Russia had better regulated financial systems than Mexico and Thailand. 29 hypothesis that the program helped restore confidence12 . Still, it is difficult to disagree with Eichengreen (2002) that “the pattern of financial bailouts has distorted the operation of financial markets” (p. 9) and that in at least some cases actual or prospective international aid “has permitted governments to cling longer to unsustainable policies, allowing economic and financial vulnerabilities to build up and creating the incentive for very severe political and social dislocations when support is ultimately withdrawn” (p. 9). In our judgment, some of the best evidence of the importance of moral hazard comes not from formal statistical tests, but from the observations of investors themselves. Of course economists are trained to be skeptical of such types of evidence, but sometimes it’s the best that we have. As Eichengreen (2002) argues, “Investment banks newsletters contain compelling evidence of moral hazard” (p. 51). Indeed, with respect to Russia and Ukraine, the term “moral hazard play” was widely bounced about in the investment community. This description certainly seems to fit well the effects of IMF loans to Argentina, Brazil, Russia, and Turkey to help maintain pegged exchange rate regimes. Such analysis seems to have much less applicability to Asia, however. Most of the countries hit by the Asian crises had not been on IMF programs prior to the crisis. Thus while its true that countries such as Thailand clung to poorly conceived policies that contributed to greatly increasing the burden of the crisis when it finally hit. There was a great deal of moral hazard generated by the Thai government’s policies but in our judgment relatively little of the overall moral hazard that was generated resulted from expectations about IMF policies. Such a statement would clearly not be true for the Russian crisis, however. Thus we should be wary of applying broad generalizations about moral hazard to specific cases. 12 On this issue see Jeanne and Zettelmeyer (2001). 30 IV. The Role of Fundamentals Beliefs that the Asian crises were due largely to panic in international financial markets have been heavily influenced by the view that the macroeconomic fundamentals of these countries were strong and hence the speculative attacks were unjustified. Such general perceptions were reinforced by the influential article by Furman and Stiglitz (1998) that applied several of the then state of the art statistical crisis models to the pre-crisis Asian situation and found that they had little explanatory power. A number of different approaches have been taken to the empirical study of the statistical correlates and predictors of currency crises.13 While some studies have sought to predict the timing of crisis, others have investigated to what extent economic and financial variables can help explain the pattern through which crisis spread from one country to another. Our focus is on this latter question. The initial contribution to this literature was a highly influential paper by Sachs, Tornell, and Velasco (1996). STV analyze which countries were vulnerable to currency attacks as a result of the 1994 Mexican Peso crisis by testing a cross-section of 20 emerging countries.14 They regress a crisis index, the variance-weighted average of the percentage change in the nominal exchange rate and international reserves15 , on three fundamental factors: real exchange rates, banking fragility proxied by the percentage change in bank loans to the private sector, and the ratio of M2 to international reserves. (Their specific formulation and our modification and extensions are discussed in Appendix 1). 13 For recent contributions and surveys of the literature see Berg and Patillo (1999a and b), Bussière and Mulder (1999), Goldstein and Reinhart (1999), and Kaminsky et al. (1998). 14 The list of countries is Argentina, Brazil, Chile, Colombia, India, Indonesia, Korea, Jordan, Malaysia, Mexico, Pakistan, Peru, Philippines, South Africa, Sri Lanka, Taiwan, Thailand, Turkey, Venezuela, and Zimbabwe. 15 In their original development of a crisis index Eichengreen, et al (1996) also included the interest rate, but the lack of high quality data led to the inclusion of interest rates in the indices for developing countries in the initial studies. 31 STV argued that a country with real exchange rate appreciation and banking problems would suffer a severe crisis only if the country had both weak fundamentals and low international reserves, where weak fundamentals were defined as real exchange rate appreciation and the growth in the bank loans are in the highest three quartiles of the sample.16 Likewise low international reserves were defined as ratios of M2 to foreign reserves in the highest three quartiles of the sample. Thus it would seem better to label them as the absence of high reserves and the absence of strong fundamentals, respectively. They found that real exchange rate appreciation and lending booms had positive effects on the severity of crises in countries with low international reserves and weak fundamentals, and did not have significant effects in countries with low reserves and strong fundamentals. The STV paper presented an innovative analysis that heavily influenced a good deal of subsequent research. Two aspects of the analysis are particularly open to question, however. One is the benchmark for distinguishing strong and weak fundamentals and high and low reserves. The spirit of the second generation crisis models suggests that a three way classification of fundamentals into strong, weak, and intermediate or vulnerable would be more useful. (The STV model was developed before the use of second generation models had become widespread). The second is the argument by STV, for which they find empirical support, that high reserves will fully protect a country with weak fundamentals. In second generation models high reserve levels could protect vulnerable countries with intermediate fundamentals but not those whose fundamentals are weak. Since three-fourths of the sample is defined as having weak fundamentals, the STV analysis essentially doesn’t distinguish between weak and intermediate fundamentals. (This will be discussed further below). 16 Several authors have questioned the use of rankings within a sample as opposed to absolute measures. See the comments by Furman and Stiglitz (1998) and comment by Cooper on Sachs et al (1996). 32 Our statistical analysis of the role of fundamentals in the Asian crises is based on the testing in Nitithanprapas and Willett (2000) that includes both the Mexican and Asian crises (combined and separately) and like STV excludes interest rates from the crisis indices. When we turn to the role of exchange rate regimes and capital controls, we draw on more recent analysis by Nitithanprapas, Rongala, and Willett (2002) that considers only the Asian crises and which use crisis indices both with and without interest rates. A particular reason for including interest rates is that without them the data shows no evidence of the strong speculative attack on the Hong Kong dollar following the depreciation of the Taiwanese dollar in October 1997. In the absence of good estimates to determine weights in the crisis index (see Appendix 1) we acknowledge the arbitrariness by using equal weights. In table 4 we present monthly data in all three crises index variables for a number of crisis countries from March 1997 through March 1998 to help readers reach their own judgments about the spread of the crises. Furman and Stiglitz (1998) applied the STV approach along with several others to explain the severity of the East Asian crises. Furman and Stiglitz’s sample included 34 emerging countries, 14 more than STV used.17 Furman and Stiglitz found the STV model to have little explanatory power. The R-squared was only 5 percent and the adjusted R-squared was –16 percent. Furman and Stiglitz conclude, “Our theoretical and empirical analysis strengthens the presumption that the East Asian crisis was a novel event and that, given the knowledge at the time, it probably could not have been predicted” (p. 19). While they find some evidence that rapid expansion of domestic credit played a part, they conclude, “There is little evidence that real exchange rate appreciation played more than a very small role” (p. 19). If correct, this would 17 The addition 14 countries are Bangladesh, Botswana, China, Cote d’Ivoire, Ecuador, Egypt, Ghana, Israel, Kenya, Mauritius, Morocco, Singapore, Trinidad and Tobago, and Tunisia. 33 largely absolve pegged exchange rates from having played the major role in the crisis that many critics have argued. The role of exchange rate regimes will be discussed in detail below. Radelet and Sachs (1998) were more inclined to the view that exchange rates played a major role and argued that “real effective exchange rates appreciated by more than 25 percent in Indonesia, Malaysia, the Philippines and Thailand between 1990 and early 1997; in Korea the appreciation was about 12 percent” (p. 9). Their statistical analysis, however, also failed to find a significant role of exchange rate appreciation. In an independent analysis conducted at about the same time, Berg and Patillo (1998) found results that were slightly more encouraging, but not by much. To the question, could the Asian crises have been predicted, their answer was yes and that some of the models beat chance, but not by a lot. They likewise found appreciation not to have played a major role. Somewhat later analysis using the STV approach by Corsetti et al (1999) Tornell (1999), and Nitithanprapas and Willett (2000) found much greater evidence of the fundamentals having played an important role. These results met with some skepticism about their robustness, however. Besides finding that data revisions had substantial effects on some of the STV estimates, Berg and Patillo conclude that “A variety of apparently small modifications characterizes the difference between the specification STV and Tornell (1998), and yet these respecifications apparently make the difference in predicting the incidence of the 1994 crisis ‘out of sample’. This suggests that specification uncertainty can be as important as parameter uncertainty across crisis episodes…” (p. 53). This is consistent with Furman and Stiglitz’s emphasis that “…the degree of real misalignment is very sensitive to the measure used” (p. 15). In a similar vein, in discussing the Corsetti et al paper, Richard Portes (1999) comments “The original version of the chapter reported 14 regressions on a cross-section of 24 countries 34 each regression having typically 18-19 degrees of freedom; there were clearly many unreported regressions that had been run. All this comes too close to data mining for my taste, and I would suspect that the results would not be robust: either to alternative specifications, or as indicators or interpretations of the next round of crisis” (p. 164). While we cannot speak directly to the robustness of the specific models of Corsetti et al (1999) and Tornell (1999), we have found a great deal of robustness to the variant of the STV approach developed by Nitithanprapas and Willett (2000) and refined by Nitithanprapas ( 2002) and by I. Nitithanprapas, Rongala, and Willett (2002). There has been considerable controversy about the role of fundamentals in general and exchange rate appreciation and current account deficits in particular in contributing to the Asian crises. Both Furman and Stiglitz (1998) and Berg and Patillo (1999a & b) find insignificant effects of currency appreciation on the Asian crises, while Tornell (1998) finds a positive effect. As Begg (1999) and Bussière and Mulder (1999) suggest, one likely reason for these divergent findings are the substantial differences in the calculations used in different indices. For example, the range for China was -7.0 to -35.9, for Hungary from 0 to +31.4, and for Indonesia from -10.0 to +9.6. While not changing signs, calculations for China ranged from –7.0 to –35.9 and for Hungary from 0.0 to +31.4. The calculations for Thailand were much more tightly bunched, ranging from +5.5 to +7.2. Malaysia ranged from +4.5 to +11.3 while the Philippines ranged from +11.9 to +28.4 and Hong Kong from +7.7 to +17.6. In our sensitivity analysis presented in Appendix 2 we test a number of appreciation measures and find that while our composite variables in which they are included always have the expected positive signs, the level of significance varies widely across the measures. 35 In our estimation model (again see Appendix 1) we adopted the basic format of STV but replaced their exchange rate variable with a composite variable that includes both exchange rate appreciation and the current account deficit. There are two rationales for our use of such composite variables. One is that with quite limited degrees of freedom it would be questionable to include separate exchange rate and current account variables. Of course, if both variables provide the same information then either one could be used. For example, Dominick Salvatore (1999) excluded the real exchange rate variable from his preferred indicators because he argues it does not add information to that contained in the current account variable. On the other hand, Kaminsky et al. (1998) exclude the current account from their preferred set of warning indicators of currency crisis on the basis of its poor performance in earlier studies. They suggest that it does not add information to that embodied in the real exchange rate variable. These opposite views clearly highlight the need to pay more attention to potential interactions between the current account and the real exchange rate. There is no reason to suppose that appreciation and current account deficits will always provide the same information, nor is their relevance always independent of each other. External equilibrium indicates that a set of variables are mutually adjusted to one another so that there are no major pressures for change. There are many different constellations of variables that can produce equilibrium or disequilibrium. Currency appreciation can be good or bad depending on the circumstances. The same holds for current account deficits. A current account deficit should be more worrisome if it is accompanied by a real exchange rate appreciation. The basic idea is that a real exchange rate appreciation due to capital inflows under a flexible exchange rate or domestic inflation above foreign levels under a pegged rate will cause a loss of competitiveness, thus worsening the 36 current account deficit. Hence, the more appreciated the real exchange rate is, the higher the rate of depreciation or devaluation that is needed to generate a more sustainable current account position. On the other hand, if real appreciation is due primarily to higher rates of productivity growth, or rapid growth in the demand for exports, then this should not be a source of declining competitiveness and future balance of payments problems. Thus it is when real appreciation and large current account deficits are found together that the situation should be most worrisome. The current account deficit can thus act as a crude screen to help us separate "bad" real exchange rate appreciations from "good" ones. The "good" real exchange rate appreciation, resulting from an increase in productivity in the tradable sectors relative to the non-tradable sectors, generates a current account surplus and need not be cause of worry. On the other hand, when the real exchange rate appreciation is accompanied by a current account deficit, it may signal a "bad" appreciation that would reduce the competitiveness and worsen the current account deficit.18 Recent crises have highlighted the volatility of many types of capital flows. The higher a country's current account deficit, the more dependent it is on continued capital inflows to avoid rapid adjustment. Hence we can think of large current account deficits as an indicator of vulnerability, with their vulnerability being greater, the larger is the accompanying exchange rate appreciation. On this logic we would be less concerned by current account deficits that are financed by sources of inflows that are likely to be more stable. Direct investment flows tend to 18 The importance of this has not escaped the IMF. The May 1999 World Economic Outlook notes that “a variable that may no be significant in isolation may be important because of its interactions with others; conversely, a variable that ,ay appear relevant on its own may no longer be so when other variables are considered” (p.79) and gives the example of real exchange rate appreciation due to productivity growth. They group variables into composite indication of external imbalances, internal imbalances, trade spillovers, and financial vulnerability. For external balances they us a weighted average of measures of real appreciation, productivity growth in the export sector and current account deficits, but do not consider the threshold approach. Comparing the indicators for crisis and non-crisis countries in a large number of 20 industrial and 41 emerging market countries they conclude that “External imbalances were particularly important for differentiating between crisis and non-crisis countries for emerging markets and during the Asian and Russian crisis” p. 80. 37 be stable than flows of financial capital and this certainly proved to be the case in the Asian crises. Thus in place of the full current account deficit, one possibility is to use only the portion of the current account deficit not financed by direct foreign investment. The statistical analysis by Nitithanprapas and Willett (2000) finds little difference in explanatory power between the two measures for the Mexican and Asian crises. This approach can be implemented through a threshold analysis either by including the current account deficit when real appreciation is above a threshold or by including the amount or real appreciation when the current account deficit is above a threshold. Alternatively, one could multiply the values of appreciation and the current account deficit when their values are above a threshold, with values below the threshold as zero. One way does not seem clearly preferable to another on theoretical grounds so in various studies we have made use of all three.19 One view, sometimes called the Lawson Dogma after the former U.K Chancellor of the Exchequer who expressed it, holds that current account deficits are only a problem if they are accompanied by government budget deficits. The relatively strong budget positions of Mexico in 1994 and most of the Asian crises countries in 1997 clearly illustrate the falseness of this view and this is confirmed by formal testing in Nitithanprapas and Willett (2000).20 The basic resu1ts from N&W (2000) show that with low levels of reserves, both the external composite variable and the lending boom variables are significant in both the Mexican 19 The original work by Nitithanprapas and Willett (2000) used only the first two measures. A commentator suggested the third. Independently Corsetti et al (1999) developed a similar measure which they found to have substantial explanatory power for the Asian crises. 20 We should note, however, that several economists have argued that prospective budget deficits may have played a role in the crises. For example, Calomiris (2003) argues, “Expectations of future government expenditures, not current expenditures, often drive crises. Financial sector imbalances … produce fiscal imbalances through the off balance sheet contingent liabilities of the government. In a world in which banking sector collapses often produce fiscal costs in excess of 20 percent of GDP … a focus on macroeconomic flows as measurements of fundamentals may leave the prince out of the play.” (p 266) See also Burnside, Eichenbaum, and Rebelo (2001) for an explicit application of this point to the Asian crisis. 38 and Asian crises and with sufficiently high reserves countries seems to be protected from both external and internal vulnerability. (The latter conclusion will be discussed further below). They found that the model’s explanatory power for the Asian crises was a good bit lower than for the Mexican crises, adjusted R2 of 0.28 compared with 0.64, but it is much higher than the -0.16 found by Furman and Stiglitz (1998). Furthermore, the Chow structural stability test failed to find a significant difference between the periods. Nitithanprapas and Willett (2000) found the estimated coefficients to be remarkably stable in response to a wide range of stability tests including the use of fixed and random effects models, the reversal of the composite variable to measure the real appreciation if the current account deficit is above a given threshold, varying the thresholds for the composite variable, varying the weights of exchange rate versus reserve changes on the crisis indices, and varying the time periods over which the variables are measured. (See Tables 1-10 in Nitithanprapas and Willett (2000)). While the Mexican and Asian crises countries were doing well on the inflation front, rapid increases in monetary and credit expansion were likely contributors to the crises in both cases. Indeed, while recent crisis models have tended to use credit expansion to the private sector as an indicator of banking problems, traditional balance of payments models have focused on monetary expansion as an indicator future inflation. Recent analysis has also focused on the role of money and credit expansion in generating asset bubbles. The Asian experience clearly illustrates that it is not safe to wait for acceleration of inflation as an indicator of excessive money and credit growth. While there have been considerable disagreements about the extent to which the Asian countries could or should sterilize capital inflows, it is a commonly expressed view that is a number of countries the large 39 capital inflows from greater than could be effectively utilized in the short run and that the consequence was not only excessive credit creation overall, but also that a good bit of this credit was allocated to less than efficient users. For example, Makin (1999) judges that “from 25-40 percent of total bank lending in Thailand, Indonesia, and Malaysia was for speculative property development. Moreover, substantial ‘connected lending’ and ‘directed lending’ was undertaken by local banks in the absence of sufficient capital banking.” (p. 419). We cannot use the STV formulation to tell whether to avoid crises, countries need to hold particularly high reserves or just avoid holding particularly low reserves. To investigate this question N&W (2000) tested variations of the definition of high reserves from the first quartile of ratios of M2 to reserves used by the STV to the second and third quartile as well. Table 1 Regression Results Explaining the Severity of the Mexican and Asian Crises (From N&W (2000) 1994 Mexican 1997 East Asian 1994 and 1997 Crisis Crisis Crises -4.03 -0.004 -2.06 (2.49) (2.91) (1.81) COM3b with an absence of -4.91* -4.99* -4.96* high reserves (1.38) (1.76) (1.02) Lending booms with an 0.23* 0.22* 0.22* absence of high reserves (0.06) (0.08) (0.04) b2+b4 COM3 with high reserves 1.79 -1.57 -1.26 b3+b5 Lending booms with high 0.01 0.01 0.01 R-Squared 0.71 0.41 0.56 Adjusted R-Squared 0.64 0.28 0.52 p-value 0.77 p-value 0.58 p-value 0.57 Estimated Independent Variables Coefficient b1 b2 b3 constant reserves Summary Statistic Addendum:Wald Test Ho: b2 + b4 = 0 40 Ho: b3 + b5 = 0 p-value 0.67 p-value 0.93 p-value 0.74 Note: Standard errors are shown in parenthesis. One asterisk (*) indicates statistical significant at a 5 percent level. Two asterisks (**) indicate statistical significant a 10 percent level. a. b. The dependent variable is the crisis index from November 1994 to March 1995 for the Mexican crisis period, and from June 1997 to October 1997 for the Asian crisis period. COM3 is the share of average current account deficit to GDP between 1990 and 1994 for the Mexican crisis and between 1992 and 1996 for the Asian crisis if the real effective exchange rate appreciates by more than 10 percent; otherwise, COM2 is set to zero. The results showed little change when going from the first to second quartile. The coefficients of the independent variables also changed little when we went to the third quartile, but the equation performed less well. Unlike the STV findings for the Mexican crisis, in these runs we did not find strong support for the proposition that high reserves fully offset weak fundamentals, even for the most stringent STV definition of high reserves. We also replaced the dummies based on the ratios of M2 to reserves with ones based on whether reserve levels exceed or fall short of short term foreign debt. Again the equation performed well, but the STV hypothesis of strong reserve positions fully offsetting weak fundamentals did not receive consistent support. Thus we believe there is good reason to doubt the initial STV conclusion that high reserves can fully offset the effects of weak fundamentals. There is need for further study of the interrelations between the effects of fundamentals and reserve levels in reducing countries’ vulnerability to crises, but this should be undertaken using a much larger data sample21. V. Capital Controls and Exchange Rate Regimes in the Asian Crises 21 The estimation results of Bussière and Mulder (1999) suggests that higher liquidity (as presented by the level of reserves over short-term debt) can offset weak fundamentals (as represented by the current account deficit and the appreciation of the exchange rate) and limit the vulnerability of countries in periods of contagion. The existence of such an ability to offset weak fundamentals by high liquidity contrasts sharply with a fundamentals only view of the world, in which solvent countries are always able to exploit their solvency by borrowing their way out of a potential crisis, and thus have no need for a cushion of liquidity. 41 One common interpretation of the Asian crises is that they primarily reflect the problems of one-way speculative gamble in a world of adjustably pegged exchange rates and high capital mobility. In this view the underlying cause of the Asian crises was the same as of the breakdown of the Bretton Woods adjustable peg system in the early 1970s and the crisis in the European Monetary System in the early 1990s. In other words, these are examples of the unstable middle hypothesis. Economists are generally agreed that with high capital mobility and no capital controls, narrow band adjustable peg systems such as were practiced during the Bretton Woods years and by Thailand are accidents waiting to happen. However, here is where the consensus ends. Many prominent economists have embraced the two-corners hypothesis under which the unstable middle is so wide that one must go all the way to one extreme or the other – genuinely fixed or highly flexible rates--in order to avoid currency crises. Others argue that intermediate systems such as crawling bands can be stable and point to examples such as Chile, Israel, and Poland to support their point. Clearly this is an issue which requires serious empirical attention. How much pegged rates contributed to the Asian crises depends in no small part on what one means by pegged rates. Many have talked of all of the hardest hit Asian currencies as having pegged exchange rates. Thailand clearly fit the bill of the Bretton Woods type narrow band adjustable peg. The baht was pegged to a basket of currencies rather than to the dollar alone, but while the weights of the basket were never officially announced, there is general agreement that the weight given to the dollar was at least eighty or ninety percent. A wide range of countries listed themselves as independent or managed floats, but displayed what Calvo and Reinhart (2002) have labeled “fear of floating” and intervened heavily in the foreign exchange markets. While commentators frequently refer to the pegged rates of Indonesia and Korea, both officially listed themselves as having managed floats. Behaviorally, 42 Indonesia and Korea both had crawling bands. Thus little weight should be given to official classifications. The IMF has begun to produce an alternative classification based on staff judgments of actual exchange rate policies. Backdating of these measurements to the Asian crises years has recently been made public (See Bubula and Robe (2002)). The problem with official classifications has led several economists to develop behavioral classifications of exchange rate regimes based on statistical measures. Two of the authors have been entrants into this derby and we draw upon our own as well as several other measures of exchange rate regimes. We find substantial differences in the classifications for a number of countries. To compare only two recent studies, Levy-Yeyati and Sturzenegger (2000a) classify both India and Malaysia prior to the Asian crises as floating, while Grier and Grier (2001) classify them as having pegged rates. Clearly such differences can have a major influence on the lessons drawn from the experience of the Asian crises. Below is a table on the different measures of exchange rates that we used in this study in our sample set of countries in 1996. Table 2 Exchange Rate Regimes ARGENTINA BRAZIL CHILE CHINA CHINA, HONGKONG COLOMBIA EGYPT HUNGARY INDIA INDONESIA JORDAN KOREA MALAYSIA MEXICO MOROCCO PAKISTAN Two-Way Classification Peg Float/Peg Float/Peg Peg Peg Float Peg Float Float/Peg Float/Peg Peg Float/Peg Peg Float Float Peg 43 Three-Way Classification Peg Crawl Float/Crawl Peg Peg Float Peg Float/Crawl Crawl Crawl Peg Crawl Crawl Float Float Peg PERU PHILIPPINES SINGAPORE SOUTH AFRICA SRI LANKA THAILAND TURKEY URUGUAY VENEZUELA ZIMBABWE Float Peg Float/Peg Float Float Peg Float Float Float/Peg Float Two-Way Classification – Countries are classified as having either a floating or pegged regime with some countries being classified as biased to both regimes. Countries are classified as either having a floating, crawling, or pegged regime with some countries being classified as being biased to two regimes. Three-Way Classification - Float Peg Crawl/Peg Float Crawl Peg Float/Crawl Float Crawl Crawl Another key area of dispute is the role of capital controls. There is wide-spread agreement that poorly designed and ineptly implemented financial liberalization played a major role in the Asian crises, but the implication of the crisis for the desirability and effectiveness of capital controls are less clear. There is no question that the Asian crises halted the momentum that had been building in support of freedom of capital flows. Indeed, this crisis has quite appropriately focused attention on the dangers of poorly managed financial liberalization, but what lessons we should draw about capital controls are much less clear. There is a widespread perception that it was capital controls which played an important role in protecting China and India from the Asian contagion. For example, the Wall Street Journal (1998) has noted that “some of the holdout nations that refused to adopt the free-flow-ofmoney orthodoxy of the 1990s -- China, India and, to a lesser degree, Chile -- now stand out as the countries least affected by the current crisis” while Paul Krugman (1999) has argued that “nearly everyone is glad that not all developing countries managed to liberalize their capital accounts before the 1997 crisis hit; in particular, China, thank heavens, still has a nonconvertible 44 capital account.” (p. 74). The data is, of course, consistent with this conjecture, but its validity is far from obvious. For example, China had huge reserves and this may have been what really protected them, and while China did have a tightly pegged exchange rate; India in the years between the Mexican and Asian crises had adapted more flexible exchange rates. Might this not have been what gave India its real protection? We may never be able to answer such questions to the satisfaction of all sensible people, but it is clear that our only hope of providing answers is to undertake multi-variate analysis for a substantial number of countries. Otherwise there is almost no possibility of having enough degrees of freedom to distinguish among alternative plausible hypotheses. Indeed several recent large N studies have investigated the relationship of capital controls to currency crises and have tended to find that the prevalence of capital controls tends to make currency crises more rather than less likely. The standard explanation for such a positive correlation is that capital controls may signal bad policy and hence make crises more likely, while the absence of controls signals good policy and is associated with fewer crises. Leblang (2001a) makes the important distinction between successful and unsuccessful speculative attacks and finds that, while capital controls make speculative attacks more likely, they also help countries maintain their exchange rate regimes in the face of speculative attacks. Such results would be consistent with capital controls being more effective in the sense of having some restraining power on capital flows, but also being subject to bad signaling effects. One of the typical problems with large N studies is that a good deal of institutional detail and nuance is lost. This can be provided by careful case studies. Thus we see case studies and large N empirical studies as complements rather than substitutes. A major theme of our paper is that the value of large N studies can sometimes be greatly enhanced by paying careful attention 45 to the quantitative variables being used. We argue that there are important conceptual issues raised in the measurement of both exchange rate regimes and capital controls. Until recently most of the empirical studies of capital controls used zero-one dummies based on information provided by the IMF. It has become increasingly recognized, however, that most of the interesting policy issues cannot be addressed with such a blunt proxy. Unfortunately, most of the studies of the relationships between currency crises and capital controls have used such limited proxies. One contribution of our work is to compare the results from using such zero-one proxies with two more differential approaches – the procedure for proxying the intensity of capital controls developed by Dennis Quinn (1997) and measures of the breadth of controls developed by Barry Johnston et al (1999) at the IMF22. While it seems clear that poorly executed domestic financial liberalization played a major role in the crisis, the role played by international liberalization per se is not as obvious. There is a general impression of substantial international liberalization by the Asian countries and there were undoubtedly large capital inflows. Thus it is not uncommon to encounter arguments such as Joseph Stiglitz’s given at a speech in Manila where he said that “even with the buildup of vulnerability, it is unlikely that the crisis could have occurred without the liberalization of capital accounts. It is worth observing that some of the countries with the weakest financial sectors and the greatest lack of transparency were hardly touched by the contagion from East Asia. These were countries with closed, or at least more closed, capital accounts.” (1998) 22 The debate over whether Malaysia’s adoption of capital controls helped speed its recovery is outside the scope of this study. For discussion on that, see Edison & Reinhart (2000) and Kaplan & Rodrik (2001). Our focus is on the effects of controls (and exchange rate regimes) on the strength of speculative pressures, i.e. capital flight, felt by countries during the Asian crises. We likewise do not investigate the costs of these crises in terms of the lost output (and human suffering) generated. The latter were heavily influenced by the degree of financial sector problems, not just the extent of the currency crises. 46 The data from the quantitative indicators of capital controls, reported in table 3, suggests another story, however. For example, using a zero-one dummy variable, Glick and Hutchison (2000) classify only Hong Kong and Singapore among the Asian crises countries as having liberalized capital accounts while the IMF’s more detailed measure of the breadth of capital controls lists Indonesia, Korea, and the Philippines at 0.92, Malaysia at 0.85 and Thailand at 0.77 out of a maximum at 1.0. (For example, India was 1.0 and China 0.85, while Hong Kong was 0.08, and Singapore was 0.23). Nitithanprapas, Rongala, and Willett (2002) extended the statistical analysis to consider the roles of capital controls and exchange rate regimes in the Asian crises. In this section we present updated results. Our econometric model controls for what has now become a widely used set of variables in crisis models. The inclusion of measures for credit booms and the ratio of short-term foreign debt to international reserves are quite straightforward. Our non-standard variable, labeled ‘composite’ was developed because many earlier studies tended to find inconsistent results when using measures of the appreciation of the real exchange rate and size of current account deficits. For this study we use the composite variable defined as the amount of real exchange rate appreciation when the current account deficit is above five percent. Below is a table on the different measures of capital controls that we used in this study and which indicate the degree of controls present in 1996 in our sample set of countries. Table 3 Capital Control Variables 47 Countries ARGENTINA BRAZIL CHILE CHINA COLOMBIA EGYPT HONGKONG HUNGARY INDIA INDONESIA JORDAN KOREA MALAYSIA MEXICO MOROCCO PAKISTAN PERU PHILIPPINES SINGAPORE SOUTH AFRICA SRI LANKA THAILAND TURKEY URUGUAY VENEZUELA ZIMBABWE IMF 0.38 0.77 1 0.85 0.85 0.38 0.08 1 1 0.92 0.92 0.92 0.85 0.85 0.77 0.92 0.31 0.92 0.23 0.92 1 0.77 0.69 0.08 0.31 1 G&H 0 1 1 1 1 1 0 1 1 1 1 1 1 1 1 1 0 1 0 1 1 1 1 0 0 1 Quinn 12.25 7.65 8.6 6 10 11.75 12.5 5.1 6.1 12 6.8 7.75 12 11.3 4 3.3 12.55 10.2 12.5 6 7 8.85 9.1 9.5 7.2 5.5 Modified Quinn 1.75 6.35 5.4 8 4 2.25 1.5 8.9 7.9 2 7.2 6.25 2 2.7 10 10.7 1.45 3.8 1.5 8 7 5.15 4.9 4.5 6.8 8.5 JIMF 0.19 0.6 0.89 0.73 n/a 0.3 n/a 0.57 0.87 0.5 n/a 0.7 n/a 0.36 0.72 0.66 n/a 0.47 n/a 0.56 n/a 0.63 0.36 0.13 n/a n/a AbiadMody Abiad-Mody Fin 0 2 1 n/a 6 8 2 n/a 2 1 0 n/a 7 7 0 n/a 2 1 n/a 10 6 n/a 2 1 1 1 1 0 1 0 1 1 1 1 n/a 6 5 4 6 7 5 6 2 0 8 4 5 n/a 1 1 9 9 IMF – Coded based on methodology by Jacques Miniane. G&H – Coded based on methodology by Reuven Glick and Michael Hutchison. Quinn – Coded based on methodology by Dennis Quinn. Modified Quinn – Coded based on methodology by Cheryl Vanden Handel. JIMF – Coded based on methodology by Barry Johnston et al. Abiad-Mody Coded based on methodology by Abdul Abiad and Ashoka Mody. Abiad-Mody Fin – Coded based on methodology by Abdul Abiad and Ashoka Mody. (For a brief explanation of coding methodologies, please refer to the table 6. For a detailed explanation, refer to Rongala (2003)). Most studies of emerging markets have used crisis indices based on weighted averages of changes in exchange rates and international reserves, where the weights are based on the previous variances of each variable. We have shown elsewhere (Appendix 2) that this weighting system tends to give a downward bias to estimates of the strength of unsuccessful speculative 48 attacks on fixed rate systems23. In this section we stress another problem, the need to take the behavior of interest rates into account. While interest rates were included in the indices developed for the industrial countries, they have typically been left out of studies on developing countries because of problems of data availability and quality. Without looking at interest rates, however, one would entirely miss the important unsuccessful speculative attacks on Hong Kong in October 1997. The peg was saved and measured reserves actually rose – in large part because of the sky-high increase in interest rates forced on the monetary authorities by strong speculative attacks triggered by the devaluation of the Taiwanese currency. Determining the appropriate weight for interest rates in a crisis index is, if anything, even more difficult than determining the relative weights of exchange rate versus reserve changes. First we would need to distinguish what proportion of the changes in nominal interest rates was due to changes in policy, i.e., is exogenous, as is assumed in the Mundell-Fleming model and the Keynesian analysis versus the portion that was endogenous due to private sector expectations as is assumed in monetary models. Then, for the exogenous change, we would need to determine the degree of capital mobility under a pegged rate system or the resulting change in the exchange rate under a freely flexible system. The former is extremely difficult to estimate (see Willett, Keil, and Ahn (2002)) while under flexible rates, the nature of the interest rate-exchange rate nexus has been subject to considerable debate. The implication of the Dornbusch overshooting model is that an initial change in the exchange rate is a multiple of the interest rate change conflicts sharply with Stiglitz’s view that the effects may quite often be negative rather than positive. Empirical estimates across studies have proven to be highly unstable Careful measurement of the strength of crises is also relevant for testing fine-grained hypotheses about the relationships between exchange rate regimes and currency crises. One of 23 Nitithanprapas, Nitithanprapas, and Willett (2002). 49 the most interesting developments of recent international financial behavior is the finding that hard pegs of the currency board variety are not in themselves sufficient to protect countries from currency crises. This was evidenced by the unsuccessful attack on Hong Kong in 1997 and by several unsuccessful attacks on Argentina before the final abandonment of its currency board in 2001. These important developments could easily be missed in large N statistical studies that paid insufficient attention to the proxies being used. To examine the conditional links between currency crisis and capital controls and exchange rate regimes, we regressed the crisis index on capital control measures, exchange rate regime measures, composite variable, lending boom, and the ratio of short-term debt to reserves. To check for robustness the analysis was based on three methodologies for calculating the crisis index and we supplement the three-way classifications of pegged, intermediate crawling peg/band, and floating exchange rate regimes with a two-way classification of floating regimes and peg regimes for comparison with other studies. (The estimating equations and results are presented in Appendix 1.) We find the composite variable to be highly significant throughout the spectrum. This result echoes Summers (1995) and Goldstein (1997) who indicated that one of the indicators of an impending currency crisis is a current account deficit of more than five as a percentage of GDP. Though this number is arbitrary, we tested the variable with a threshold of three percent and found reduced significance. The composite variable is the real effective exchange rate appreciation number if the country has a current account to GDP deficit of more than five percent.24 24 Rongala (2003) used two other composite variables. The first one is the current account deficit number when RER appreciation is greater than 10 percent. The other is derived by taking into account countries that have both RER appreciation higher than 10% and current account deficit greater than 5 percent and multiplying both numbers. The results were not consistently significant for the two methodologies of composite variables. Eyeballing the data 50 For sensitivity analysis, we tested seven different real exchange rate measures as part of the composite variable. These different measures were calculated based on work done in earlier papers. The composite variable was calculated by taking the real exchange rate appreciation for countries which had a current account deficit of more than 5 percent. (The different measures and results are presented in Appendices 1 and 2). The composite variables always have the expected positive sign, but do not retain significance across all of the measures. This is consistent with the difficulty of measuring equilibrium exchange rates and the controversy over whether the Korean won was overvalued before the crises. The lending boom variable remains highly significant in all of the regressions. The short term debt to reserves variable showed more significance in the three way classification when the bias is to a crawling or pegged regime. (See table 2 and 3). This could possibly be explained by the fact that most of the Asian crises countries were on either on a crawling or a pegged regime. However, when we ran regressions omitting the capital controls and exchange rate variables the short-term debt to reserves showed a much higher significance. Table 4 Using a 3-way Classification (Countries are Classified Biased Toward Floating Regime) Dependent Variable: CRIEND shows that countries in the sample that had a current account deficit higher than five percent of GDP also had real exchange rate appreciation, while not all countries that had real exchange rate appreciation had large current account deficits. While the current account deficit is generally accepted as not being good for a country, the case on real exchange rate appreciation is not that clear. Real exchange rate appreciation isn’t necessarily bad because this appreciation could be due to other reasons other than just a pegged regime. Rongala (2003) also tested the current account deficit and real exchange rate appreciation variables separately and found the current account deficit variable to have higher significance. 51 Independent Variables and Coefficient Composite (COM1) b1 Short term debt to reserves b2 Lending Boom b3 Control measures b4 Dummy float b5 Dummy Crawl b6 R2 Adjusted R2 N 1. 2. 3. 4. Miniane measure GlickHutchison measure Quinn measure AdaptedQuinn measure Johnston measure Abiad– Mody measure 1.39 ** (0.06) 4.89 (0.19) 1.42 ** (0.06) 5.03 (0.19) 1.51 (0.06) 3.39 (0.31) 1.63 * (0.04) 4.62 (0.22) 3.54 * (0.00) 3.78 (0.27) 1.82 * (0.02) 6.22 ** (0.10) AbiadMody Fin measure 1.35 ** (0.07) 10.26 ** (0.07) 0.12 * (0.04) 0.11 ** (0.06) 0.13 (0.01) 0.11 * (0.03) 0.21 * (0.00) 0.14 * (0.02) 0.18 * (0.01) 17.20 * (0.04) 12.25 ** (0.06) 0.60 (0.46) 3.30 (0.25) -0.14 (0.99) 11.87 * (0.01) 3.06 ** (0.09) -0.66 (0.88) 0.28 (0.95) -0.70 (0.90) -1.43 (0.81) 0.09 (0.98) -3.54 (0.57) 3.62 (0.62) 12.17 (0.17) 0.60 0.47 26 13.70 (0.14) 0.60 0.47 26 14.86 (0.17) 0.53 0.38 26 14.76 (0.15) 0.55 0.40 26 23.47 * (0.02) 0.71 0.53 17 6.59 (0.44) 0.62 0.46 22 10.63 (0.30) 0.58 0.41 22 One asterisk (*) indicates statistical significance at a 5 percent level. Two asterisks (**) indicates statistical significance at a 10 percent level. Numbers in parentheses are probability values. Criend = An equal weighted average of the change in value of the percentage change in exchange rate with respect to the U.S. dollar, the change in value of the percentage change in foreign reserves, and the value of the percentage points change in interest rates from June of 1997 to December of 1997. 52 Table 5 Using a 3-way Classification (Countries are Classified Biased Toward Pegged Regime) Dependent Variable: CRIEND Independent Variables and Coefficient Composite (COM1) b1 Short term debt to reserves b2 Lending Boom b3 Control measures b4 Dummy float b5 Dummy Crawl b6 R2 Adjusted R2 N 1. 2. 3. 4. Miniane measure GlickHutchison measure Quinn measure AdaptedQuinn measure Johnston measure Abiad– Mody measure 1.69 * (0.00) 7.13 ** (0.07) 1.73 * (0.00) 7.46 * (0.05) 1.89 * (0.00) 6.03 ** (0.10) 1.97 * (0.00) 7.17 ** (0.09) 3.11 * (0.00) 7.23 (0.18) 2.02 * (0.00) 7.97 ** (0.06) AbiadMody Fin measure 1.57 * (0.00) 11.39 * (0.02) 0.17 * (0.00) 0.17 * (0.01) 0.18 * (0.00) 0.17 * (0.00) 0.20 * (0.03) 0.17 * (0.01) 0.21 * (0.00) 15.20 * (0.05) 11.33 * (0.05) 0.07 (0.93) 2.53 (0.37) 3.29 (0.85) 11.73 * (0.01) 2.75 * (0.04) -5.04 (0.40) -4.74 (0.45) -4.70 (0.51) -5.33 (0.46) -5.92 (0.58) -6.56 (0.38) -0.46 (0.95) 11.85 (0.12) 0.64 0.52 26 13.06 ** (0.10) 0.65 0.53 26 14.74 ** (0.10) 0.58 0.45 26 14.33 ** (0.09) 0.59 0.47 26 14.74 (0.11) 0.62 0.39 17 7.36 (0.42) 0.65 0.52 22 12.30 (0.14) 0.62 0.48 22 One asterisk (*) indicates statistical significance at a 5 percent level. Two asterisks (**) indicates statistical significance at a 10 percent level. Numbers in parentheses are probability values. Criend = An equal weighted average of the change in value of the percentage change in exchange rate with respect to the U.S. dollar, the change in value of the percentage change in foreign reserves, and the value of the percentage points change in interest rates from June of 1997 to December of 1997. The three-way classification of exchange rate regimes shows a much higher level of consistency than the two-way classification and is far more robust. The results indicate that for the two-way classification the dummy for floating regime can be either positive or negative depending upon whether ambiguous cases are counted as pegged or floating. Grier and Grier (2001) found that countries are more susceptible to a currency crisis when exchange rates are 53 pegged., but this appears to be the result of their treating most ambiguous cases as pegged. That one get can get any sign one wants depending on the methodology used shows the inherent weakness of this type of classification. In the three-way classifications, we consistently find the sign of the dummy for a floating regime to be negative, while the sign for the dummy for intermediate crawling regimes is positive, indicating that they made crisis more likely than pegged rates which in turn were more crisis prone than highly flexible rates. These results essentially capture the pattern of crises during the Asian shock. However because of the ambiguous nature of so many of the regimes and lack of unambiguous evidence of substantial overvaluation for Indonesia and Korea before the crisis, we are hesitant to draw strong inferences form the Asian experience about he stability of different types of exchange rate regimes in general.25 The signs of the capital controls measures are always positive, suggesting that the presence of capital controls makes countries more vulnerable to currency crises. However, only three measures of controls consistently show significance. These are the IMF, Glick, and AbiadMody measures. The positive relationship between controls and crises has been attributed by some to the signaling effect that countries impose controls because there is something wrong with the economy.26 This interpretation does not fit with the data, however, since the market’s perception that these countries had strong economies led to large inflows prior to the crises. The behavior of capital flows appears more consistent with the qualitative discussions suggesting low 25 The senior author is engaged in a major NSF funded project to investigate such issues over a broader set of experiences with David Leblang at the University of Colorado. 26 See Bertolini and Drazen (1997), Drazen (1997), Eichengreen (2000), and Wihlborg and Willett (1997). Edwards (1999) suggests that policy makers tend to be complacent after the imposition of capital controls. 54 levels of capital controls in the major crisis countries than with the story of high levels of controls implied by lost of the quantitative proxies for controls.27 Finally, The Abiad-Mody Fin measure, which is an index of domestic financial liberalization, is also positive and significant. These results suggest that countries with lower degree of domestic financial liberalization are more susceptible to a currency crises. This is consistent with our view that how liberalization takes place is as important as how much liberalization takes place. For a country to be less susceptible to a crisis it should liberalize its domestic economy, but in an orderly manner that minimizes perverse liberalization. VI. Concluding Comments We find considerable evidence that in the Asian crises international financial markets behaved with less than ideal efficiency, but we are skeptical of interpretations that stress panic or other forms of pure contagion as a major cause of the spread of the Asian crises. Undoubtedly there were immediate and somewhat indiscriminate ripple effects generated throughout much of Asia by the abandonment of the Thai pegged rate regimes. It is important, however, to distinguish between ripple effects in financial markets and serious speculative attacks. There is, of course, no unambiguous dividing line between one and the other, there will be gray areas, but there can be little question that the impacts on Hong Kong and Korea in July 1997 following the floating of the baht, uncomfortable as they may have been, were far milder than the strong speculative pressures which they felt in the fall of that year. It’s hard to envision a theory of panic contagion that would have such a delayed impact. 27 In N,R,W (2002), interactions with the capital controls and measures and exchange rate regimes (two-way) were run and there wasn’t much significance. We plan to run interactions with the three-way classification as well as with the short-term debt to reserves and lending boom variable. 55 A major cause of the view that panic spread serious contagion to innocent victims was the strong macroeconomic fundamentals of the crisis countries. More in-depth analysis has clearly indicated the importance of including international payments and exchange rate positions, domestic financial factors, and political considerations among the important fundamentals. Unlike the influential initial empirical study by Furman and Stiglitz (1998) that found that conventional crisis models could explain little if any of the spread of the Asian crises; subsequent analysis has found that fundamentals did play an important role in explaining the pattern of subsequent speculative pressures. Our empirical analysis strongly suggests that the combination of substantial real exchange rate appreciation with large current account deficits was an important source of crisis vulnerability. Our research to date does not give a clear indication of the best way of combining these variables to distinguish between safe and unsafe appreciations and deficits, but we believe that we have established the importance of looking at these two variables in relation to each other as a screening device. Our analysis also helps shed light on previous contradictory findings about the roles of appreciation and current account deficits in the Asian and other crises. Another major finding is that while inflation did not play a major role in the Asian crises, unwise credit expansion did. Thus we cannot take low inflation rates as a safe indicator of financial soundness and the absence of overheating in the economy. This suggests that strict inflation targeting needs to be complemented with a broader view. A related lesson is that policies that prudent for a good while can stray off course rather quickly. The Mexican crisis provided a strong example of this where tight monetary and fiscal policies that had brought down inflation from over 100 percent to less than 10 percent were loosened prior to the 1994 election, Likewise, while Thailand had been following laudable monetary and fiscal policies for some 56 time, a number of observers have argued that if the government had tightened fiscal policy in 1996 in light of the large capital inflows, the subsequent overheating of the Thai economy could have been avoided.28 Our statistical analysis, as that of Furman and Stiglitz (1998) and Radelet and Sachs (1998) before us, provides strong evidence that there was a substantial element of liquidity crisis to the Asian contagion. The ratios of M2 and short-term foreign debt to reserves have substantial explanatory power for the pattern of exchange market pressures. The need for governments to pay careful attention to their country’s international liquidity positions is thus firmly established. Unfortunately, however, our statistical analysis is not able to provide much more precise guidelines. Because of the limited degrees of freedom at our disposal we cannot say much about the tradeoffs between specific levels of reserve holdings and the risk of crises or about the interrelationships among fundamentals and reserve levels in preventing crisis. Analysis of a much larger data set may allow us to develop better guidelines for international liquidity management. Our analysis of the role of capital controls in the Asian crises does not lead to sharp policy conclusions. While it has been popular to argue that capital controls saved China and India from the Asian contagion, our statistical analysis suggests that fundamentals may have been of greater importance. Indeed, like several other recent studies, we find a positive rather than negative correlation between capital controls and currency crises. Using Abiad and Mody’s indices of domestic financial liberalization we find further results that conflict with conventional wisdom. In our regressions more financial liberalization is negatively rather than positively correlated with crisis. This is consistent with our view that how liberalization takes place is as important as how much liberalization takes place. 28 See, for example, Kawai, Newfarmer, and Schmukler (2001). 57 Perhaps our most important conclusion concerning capital controls is that the indices used in all of the empirical studies of which we are aware need to be viewed with considerable suspicion and how we should not base strong policy conclusions on the quantitative studies that have been undertaken to date (including our own). Contrary to the standard qualitative interpretation that premature capital account liberalization was a major contributor to the Asian crises, the available quantitative summary indicators of capital controls typically report high levels of controls for the Asian crises countries. The one quantitative measure that seems to fit qualitative judgment fairly well for Asia, the detailed classification developed by Barry Johnston et al (1999) for the IMF is unfortunately available for 1996 only. The 13-point scale now regularly published by the IMF based on Johnston’s study and backdated by Jacques Miniane, yields much less reliable signals. With respect to exchange rate regimes, our analysis illustrates the need to be careful in classification and the dangers of the simple division of regimes into fixed and floating. With such two-way classifications it is quite unclear how the pre crisis regimes of countries like Indonesia and Korea should be classified. While frequently referred to as having been pegged, they were officially managed floats and behaviorally crawling bands. The Asian crises does provide further evidence of the dangers of attempting to run Bretton Woods type narrow band adjustable peg regimes in a world of substantial capital mobility, but not on the workability of more flexibly oriented intermediate regimes. Beyond the problems with narrow band pegged regimes, we also cannot draw strong lessons from the Asian crises about the potential stability of more flexible intermediate regimes. Because of low inflation rates intermediate regimes have been more durable than in many other areas. The de facto crawling bands of Indonesia and Korea could not protect these countries 58 from crisis, but neither was over valuation clearly a major cause of the crisis in these countries as it was for Thailand and the Philippines. The major speculative attacks on Hong Kong and later Argentina illustrate that “hard pegs” of the currency board variety are not sufficient by themselves to generate confidence that pegged rates won’t be changed. On this count, the breadth of the unstable middle must be seen as extending further in the fixed rate direction than previously anticipated. On the other hand, the scope for fairly flexible crawling bands to avoid generating crisis must be judged still open on the evidence now at hand. Our analysis, however, offers little grounds for optimism that a common basket pegged regime would offer a viable way of securing Asian monetary integration on the cheap. Nor are the Asian economies close to making an optimal currency area that would justify a common currency. 29 For most Asian countries at the current time, flexible inflation targeting with managed flexibility of exchange rate regimes is likely the best option.30 The narrow band pegs of Malaysia and China should be sources of particular concern to their Asian neighbors as is the apparent heavy management of the flexible rate regimes of Korea and Thailand. With the current strong external positions of these countries the immediate concern is not so much the danger of another 1997 style set of currency crises as the strains that these countries under valued currencies are placing on the functioning of the international monetary system. As noted above, desirability of holding substantial levels of international reserves is an important lesson of the Asian crises and certainly a substantial rebuilding of Asian reserves was quite appropriate. However, the magnitude and duration of these accumulations by several 29 30 See, for example, the analysis and work referenced in Kwack et al (2003) and Willett and Maskay (2003). See Willett (2003). 59 Asian countries are beginning to raise major policy concerns in Europe and the US. Thus there is a strong need for increases in the flexibility with which exchange rate regimes are operated by some Asian countries. The role of the IMF in the Asian crises has been the subject of considerable debate and cannot be dealt with in depth here. Two particular aspects of the debate are highly relevant to the analysis in our paper. These involve IMF lending programs and the question of whether IMF conditionality should have been extended from its traditional areas of macro and exchange rate policies to financial sector issues. The IMF’s major new innovation to reform the international financial architecture in response to the Asian and other crises was the creation of the Contingent Credit Line whose purpose was to protect countries with strong fundamentals from unwarranted contagion. On our interpretation, countries with strong fundamentals (including financial, political, and international, not just domestic macroeconomic fundamentals) are seldom the victims of more than mild contagion. It is with the vulnerable, but not hopeless, countries that the IMF is likely to be able to be of the most help. The track record of the IMF in lending to help countries preserve pegged rates is poor, with Argentina, Brazil, Russia, and Turkey being important cases. In each of these cases the market turned out to be right in questioning the viability of these currencies. Thus we think it likely that the IMF (or other international lending bodies) can play a more useful role in lending to reduce the costs of crises under flexible rates, than in lending to prevent crises under pegged rates. The effects of the latter have tended to be more to allow domestic and foreign capital to flee a country at favorable rates, than in heading off the avoidable crisis that might be generated by the existence of multiple equilibria. 60 Thus, in our judgment, efforts to continue to defend the exchange rates of countries such as Indonesia and Korea would have been doomed to failure. Once flexible rates had been adopted, however, there was a good deal that could have been done to avoid the massive overshooting of exchange rates that followed. While far from pure liquidity crises, there were large liquidity crisis elements to many of the crises. Typically, IMF programs have been associated with too little adjustment being undertaken by the recipient countries. This clearly was not the case for most of the Asian crises countries. The turnarounds of their current account positions were both rapid and huge, on average exceeding ten percent of GDP. While the headline totals for the IMF loan packages were large, most of the money was not available in the early phase of the crises. To us there was a substantial liquidity crisis element to the crises with the markets becoming very risk averse. While even large loans prior to exchange ate adjustments could likely have been swamped by market outflows, once markets were freed the marginal elasticity of speculations dropped sharply and interventions could likely have redeemed a good deal of the exchange rate overshooting. This would have been an excellent case for classic lender of last resort type temporary lending. Alternatively, private sector involvement via standstills could have been tried. Park and Wang (2002, p. 110) conclude that “The Korean experience does suggests that internationally legalized payment stand stills could make debt restructuring proceed more smoothly and in a more orderly fashion in the absence of an international bankruptcy court or procedures for sovereign debtors.” We agree with this argument for the Korean case, but are not sure that it would have worked for the less concentrated patterns of lending and borrowing of some of the other Asian countries. The optimal mix of PSI and LOLR actions is likely to vary substantially from one situation to another. This does not detract however, from the need to have 61 an improved international institutional framework for dealing with PSI issues. Sadly to date little progress has been made in this score. As Krugman (1998) put it “the vociferous objections of banks have largely shut down discussions of private sector involvement.” While major financial reforms should have been required as a part of IMF conditionality, it has been a major point of contention. We agree strongly with Martin Feldstein’s (1998) argument that IMF conditionality had grown too broad an intrusive. Indeed the IMF has accepted that its conditionality needed to be “streamlined”. As Feldstein (1998) put it, “The IMF would be more effective in its actions and more legitimate in the eyes of emerging market countries if it pursued the less ambitious goal of maintaining countries’ access to global capital markets and international bank lending.” (p. 32). In a rejoinder, however, Stanley Fischer (1998) argued that financial sector reform was at the heart of the problem and therefore essential to restoring confidence. As he put it “weak financial institutions, inadequate bank regulations and supervision, and the complicated and non-transparent relations among governments, banks, and corporations, lie at the heart of the economic crisis in each country. It would not serve any lasting purpose for the IMF to lend to these countries unless these problems were addressed. There is considerable merit to this argument, but it is important to understand that few issues would be more politically sensitive than financial reforms and that this is something that could not be implemented nearly as quickly as tax or interest rate increases. Furthermore, as both Feldstein and Sachs have argued, emphasis on structural problems could worsen market confidence. In Feldstein’s view “When the foreign exchange crisis hit Korea, the primary need was to persuade foreign creditors to continue to lend by rolling over existing loans as they came due..(the need was) to persuade lenders that Korea’s lack of adequate foreign exchange reserves was a temporary shortage, not permanent insolvency. By emphasizing the structural and 62 institutional problems of he Korean economy, the fund’s program and rhetoric gave the opposite impression…Given the magnitude of the prescribed changes, lenders might well be skeptical about whether Korea would actively deliver the required changes…”. (p. 31). Given its standard lending procedures, there was no good way for the IMF to deal with this problem. That is why the senior author has recommended a major restructuring of IMF lending programs to create separate facilities for short run LOLR loans and medium term conditionality lending. The LOLR facility (see Willett (2003b)) could provide bridge loans while longer run programs with credible national ownerships could be developed. Feldstein (1998) likely had a similar thought in mind when he argued “What Korea needed was coordinated action by creditor banks to restructure its short-term debts.” but “The IMF could have helped by providing a temporary bridge loan then organizing the banks into a negotiating group.” (p. 25). The IMF has moved towards a good deal more front loading of its loans in the face of capital account crises, but this is not enough. Under current procedures the IMF is not allowed to make the types of bridge loans described by Feldstein. There are considerable political and bureaucratic pressures arranged against such a fundamental reform in the way the IMF does business, but to us the need to consider such options seriously should be one of the major lessons of the Asian crises. Stanley Fischer, the Deputy Managing Director of the Fund at that time did initiate discussion of such issues, but since his departure official interest appears to have waned. In our view that is a great pity. 63 References Abiad, Abdul and Ashoka Mody. 2002. “Status Quo Bias in Financial Reform.” IMF Working Paper: October, International Monetary Fund, Washington D.C. Ahluwalia, Pavan. 2000. “Discriminating Contagion: An Alternative Explanation of Contagious Currency Crises in Emerging Markets.” IMF Working Paper: WP/00/14, International Monetary Fund, Washington D.C. Alba, Pedro, Amar Bhattacharya, Stijin Claeseens, Swato Gosh, and Leonardo Hernandez. 1999. in Pierre-Richard Agénor et al. eds. The Asian Financial Crisis: Causes, Contagion and Consequences. Cambridge; New York and Melbourne: Cambridge University Press. Armigo, Leslie Elliott ed. 1999. Financial Globalization and Democracy in Emerging Markets. Palgrave. Auerbach, Nancy, and Thomas D. Willett. 2003. “The Political Economy of Perverse Financial Liberalization” Claremont Working Paper. Baig, Taimur and Ilan Goldfajn. 1999. “Financial Market Contagion in The Asian Crisis”. IMF Staff Papers. Volume 46, Number 2 Bartolini, Leonardo and Allan Drazen. 1997. “Capital Account Liberalization as a Signal.” American Economic Review 87: 138-54. Bayoumi, Tamim, and Barry Eichengreen. 1998. “Exchange Rate Volatility and Intervention: Implications of the Theory of Optimum Currency Area.” Journal of International Economics 45(2):191-209. Begg, David. 1999. “The Asian crises.” IMF Working Paper 199/138, October. Berg, Andrew and Catherine Pattillo. 1998. “Are Currency Crises Predictable? A Test” IMF Working Paper No. 98/154. Berg, Andrew, and Catherine Pattillo. 1999a. “Are Currency Crises Predictable? A Test.” IMF Staff Papers (46): 107-138. ______. 1999b. “Predicting Currency Crisis.” Journal of International Money and Finance, 18, pp 561-586. Blejer, Mario, and Marko Skreb. 2002. Financial Policies in Emerging Markets. Cambridge and London: MIT Press. Brune, Nancy, Geoffrey Garrett, Alexandra Guisinger, and Jason Sorens. 2001. “The Political Economy of Capital Account Liberalization.” Manuscript Collection, Yale University, New Haven, Conn. 64 Bubula, Andrea and Inci Otker-Robe. 2002. “The Evolution of Exchange Rate Regimes Since 1990: Evidence From De Facto Policies” IMF Working Paper No. 02/155. Budiman, Aida. 2002. The role of external factors, the volatility analysis and the impact of the Asian crisis on various types of flows in selected Asian countries. Unpublished Ph.D. Dissertation, Claremont Graduate University. Bussière, Matthieu and Christian Mulder. 1999. “External Vulnerability in Emerging Market Economies: How High Liquidity Can Offset Weak Fundamentals and the Effects of Contagion.” IMF Working Paper No. WP/99/98. July, International Monetary Fund, Washington D.C. Calvo, Guillermo, and Carmen Reinhart. 2002. “ Fear of Floating.” Quarterly Journal of Economics CXVII(2):379-409. Calvo, Guillermo, and Enrique G. Mendoza. 2000. “Rational Contagion and the Globalization of Securities Markets”. Journal of International Economics, No 51, p79-113. Caramazza, Francesco, Luca Ricci, and Ranil Salgado. 2000. “Trade and Financial Contagion in Currency Crises.” IMF Working Paper No. 00/55. Washington D.C.: International Monetary Fund, March. Cho, Dongchul, and Kenneth D. West. 2003. “Interest Rates and Exchange Rates in the Korean, Philippine, and Thai Exchange Rate Crises.” in Michael P. Dooley and Jeffery A. Frankel eds. Managing Currency Crises. National Bureau of Economic Research. Chicago and London, The University of Chicago Press: 11-30 Cook, David and Michael B. Devereuz. 2002. “Capital Controls in Malyasia: Effectiveness and Side Effects,” Asian Economic Paper, Vol.1, No.1: 49 – 82 Cooper, Richard N. (1996). Comments and Discussion on Sachs, Jeffrey, Aaron Tornell, and Andrés Velasco, “Financial Crises in Emerging Markets: The Lessons from 1995.” Brookings Papers on Economic Activity, No.1. Corbett, Jenny, and David Vines. 1999. “The Asian crises: Lessons from the Collapse of Financial Systems, Exchange Rates and Macroeconomic Policy.” in Pierre-Richard Agénor et al. eds. The Asian Financial Crisis: Causes, Contagion and Consequences. Cambridge; New York and Melbourne: Cambridge University Press. Corden, Max. 2002. Too Sensational On the Choice of Exchange Rate Regimes. MIT Press. Corsetti, Giancarlo, Paulo Pesenti, and Nouriel Roubini. 1999. “The Asian crises: An Overview of the Empirical Evidence and Policy Debate.” in Pierre-Richard Agénor et al. (eds.) The Asian Financial Crisis: Causes, Contagion and Consequences. Cambridge; New York and Melbourne: Cambridge University Press. 65 Council on Foreign Relations. 1999. “Safeguarding Prosperity in a Global Financial System: The Future International Financial Architecture.” (Carla Hills and Peter Peterson, cochairs; Morris Goldstein, Project Director). Task Force Report. Washington, D.C.: Institute for International Economics. Dekle, Robert, Cheng Hsiao, and Siyan Wang. 2001. “Interest Rate Stabilization of Exchange Rates and Contagion n the Asian crises Countries.” in Reuven Glick, Ramon Moreno, and Mark M. Spiegel eds. Financial Crises in Emerging Markets. Cambridge, Cambridge University Press: 347-83 Delhaise, Philippe F. 1998. Asia in Crises. Singapore, John Wiley and Son (Asia). Denzau, Arthur, and Douglas North. 1994. “Shared Mental Models: Ideologies and Institutions,” Kyklos, 47(1); 3-31. Desai, Padma. 2003. Financial Crisis, Contagion, and Containment: From Asia to Argentina. Princeton University Press. Domaç, Ilker and Soledad Martinez-Pería. 2000. “Banking Crisis and Exchange Rate Regimes: Is There a Link?” World Bank Policy Research Department Working Paper No. 2489. Washington D.C.: World Bank, November. Dooley, Michael P. 2000. “A Model of Crises in Emerging Markets”. The Economic Journal, 110, p.256-272. ---------. 1999. “Are Capital Inflows to Developing Countries a Vote for or against Economic Policy Reforms?” in Pierre-Richard Agénor et al. eds. The Asian Financial Crisis: Causes, Contagion and Consequences. Cambridge; New York and Melbourne: Cambridge University Press. ---------, and Inseok Shin. 2000. “Private Inflows When Crises are Anticipated: A Case Study of Korea”. Paper presented at Claremont Conference on The Political Economy of International Financial Crisis, December 1-2, 2000. (http://www.spe.cgu.edu/econ/peifcconference/papers/index.htm) Dornbusch, Rudiger. 1998. “Capital Controls: An Idea Whose Time is Past.” in Peter B. Kenen (ed.) Should the IMF Pursue Capital-Account Convertibility? Essays in International Finance. No 207, May, International Financial Section, Department of Economics, Princeton University. Eichengreen, Barry; Rose, Andrew K.; Wyplosz, Charles. 1996. “Contagious Currency Crisis” National Bureau of Economic Research Working Paper: 5681 July 1996. Eichengreen, Barry. 1999. Toward a new international financial architecture: A practical postAsia agenda. Washington, D.C.: Institute for International Economics. 66 Eichengreen, Barry. 2002. Financial crises and what to do about them. Oxford and New York: Oxford University Press. Edison, Hali and Carmen Reinhart. 2000. Capital Controls During the Financial Crises: The Case of Malaysia and Thailand.” International Financial Discussion Papers No. 662. Board of Governors of the Federal Reserve System. Edison, Hali and Francis E. Warnock. 2003. “A Simple Measure of the Intensity of Capital Controls.” Journal of Empirical Finance. 10: 81-103. Edwards, Sebastian. 1999. “The Mirage of Capital Controls.” Foreign Affairs, May/June. Feldstein, Martin. 1998. “Refocusing the IMF” Foreign Affairs, March/August. Fischer, Stanley. 1998. "The IMF and the Asian Crisis." Presented at The Forum Funds Lecture at UCLA. Friedman, Thomas. 1999. The Lexus and the Olive Tree. (Farrar Straus, 1999). Furman, Jason and Joseph E. Stiglitz. 1998. “Economic Crises: Evidence and Insights from East Asia.” Brookings Papers on Economic Activity v0, n2: 1-114. Girton, Lance and Don Roper. 1997. “A Monetary Model of Exchange Market Pressure Applied to the Postwar Canadian Experience” American Economic Review v67, n4 (Sept. 1977): 537-48. Glick, Reuven, Peter Kretzmer, and Clas Wihlborg. 1995. “Real Exchange Rate Effects of Monetary Disturbances Under Different Degrees of Exchange Rate Flexibility: An Empirical Analysis” Journal of International Economics 38 (3-4), 249-73. Glick, Reuven and Michael Hutchison. 2000. “Stopping "Hot Money" or Signaling Bad Policy? Capital Controls and the Onset of Currency Crisis." Paper presented at the Claremont Conference on the Political Economy of International Financial Crises, Claremont, December 2000. available at <http://spe.cgu.edu/econ/peifcconference/papers/index.htm>. Goldstein, Morris. 1997. “The Causes and Propagation of Financial Instability: Lessons for Policymakers: Commentary.” Maintaining financial stability in a global economy: A symposium sponsored by the Federal Reserve Bank of Kansas City, Jackson Hole, Wyo. Aug. 28-30, 1997: 97-116. Goldstein, Morris. 1998. The Asian financial crisis: Causes, cures, and systemic implications. Policy Analyses in International Economics, vol. 55. Washington, D.C.: Institute for International Economics. 67 Goldstein, Morris and Carmen Reinhart. 1999. Forecasting Financial Crises: Early Warning Signals for Emerging Markets. Washington D.C.: Institute for International Economics. Grier, Kevin and Robin Grier. 2001. “Macroeconomic Indicators, Exchange Rate Regimes, Currency Depreciation and Stock Market Performance in the 1997 Crisis.” Economic Inquiry, Jan:139-48. Haggard, Stephan. 2000. The Political Economy of the Asian Financial Crisis Institute for International Economics, Washington D.C. _____ and MacIntyre. 2001. “The Politics of Moral Hazard: The Origins of Financial Crisis in Indonesia, Korea and Thailand.” in Arvid John Lukauskas and Francisco L. Rivera-Batiz, Editors. The political economy of the East Asian crisis and its aftermath: Tigers in distress 85-109, Elgar. Hausmann, Ricardo, Ugo Panizza, and Ernesto Stein. 2001. “Why Do Countries Float the Way They Float?” Mimeo. Inter-American Development Bank. Hernández-Trillo, Fausto. 1995. “A Model-Based Estimation of the Probability of Default in Sovereign Credit Markets.” Journal of Development Economics, 46: 163-179. Hernandez, Leonardo and Peter Montiel. 2001. “Post-Crisis Exchange Rate Policy in Five Asian Countries.” Paper presented at the IMF Institute’s High Level Seminar on Exchange Rate Regimes, March 2001. Holden, Paul, Merie Holden, and Esther Suss. 1979. “The Determinants of Exchange Rate Flexibility: An Empirical Investigation,” Review of Economics and Statistics 61(3): 32733. Heo, Uk.. 2001. “South Korea: Democratization, Financial Crisis, and the Decline of the Developmental State.” in Shale Horowitz and Uk Heo, Editors. The Political Economy of International Financial Crisis:Interest Groups, Ideologies, and Institutions. Rowman and Littefield. Hutson, Elaine and Colm Kearney. 1999. “The Asian Financial Crisis and The Role of The IMF: A Survey.” Journal of the Asia Pacific Economy, 4(3): 393-412. International Monetary Fund. “Annual Report on Exchange Arrangements and Exchange Restrictions.” IMF, Washington D.C. Ito, Takatoshi. 2000. “Capital Flows in Asia” in Capital flows and the emerging economies: Theory, evidence, and controversies NBER Conference Report series. Chicago and London: University of Chicago Press, 255-96. 68 Jeanne, Olivier and Jeromin Zettelmeyer. 2001. “International Bailouts, Moral Hazard and Conditionality” Economic Policy: A European Forum v0, n33 (October 2001): 407-24. Jo, Gab-Je. 2000. The Roles of International Capital Flows and Financial Contagion in the Asian Financial Crises. Unpublished PhD Dissertation, Claremont Graduate University. Jo, Gab-Je and Thomas Willett. 2000. “Currency Market Reactions to Good and Bad News During the Asian Crisis,” Seoul Journal of Economics, Winter 2000, pp. 407-420. Johnston, Barry et al. 1999. “Exchange Rate Arrangements and Currency Convertibility: Development of Issues.” International Monetary Fund. Washington D.C. Kahler, Miler. 2000. “The New International Financial Architecture and Its Limits.” In Gregory W. Noble and John Ravenhill eds. The Asian Financial Crises and the Architecture of Global Finance. Cambridge, Cambridge University Press: 235 – 260. Kaminsky, Graciela, Saul Lizondo and Carmen Reinhart. 1998. “Leading Indicators of Currency Crisis.” IMF Staff Papers. 45, No. 1, pp 1-48. ----------, and Carmen M. Reinhart. 2000. “On Crises, Contagion, and Confusion”. Journal of International Economics, No 51, p145-168. -----------, and Sergio Schmukler. 1999. “What Trigger Market Jitter.” Journal of International Money and Finance18, 99: 537 – 600. Kaplan, Ethan and Dani Rodrik. 2001. “Did the Malaysian Controls Work?” Working Paper. Harvard University. Kawai, Masahiro, Richard Newfarmer and Sergio Schmukler. 2001. “Crisis and Contagion in East Asia: Nine Lessons.” World Bank Working paper. Krueger, Anne O., and Jungho Yoo. 2002. “Chaebol Capitalism and the Currency-Financial Crisis in Korea.” in Sebastian Edwards and Jeffery A. Frankel eds. Preventing Currency Crises in Emerging Markets. National Bureau of Economic Research. Chicago and London: The University of Chicago Press: 601-661 Krugman, Paul. 1998. “What Happened to Asia?” MIT mimeo. http://web.mit.edu/krugman/www/DISINTER.html Krugman, Paul. 1999 “The Return Of Depression Economics.” Foreign Affairs v.78 no. 1 (Jan/Feb 1999). 69 Kwack, Sung Yeung, Choong Yong Ahn and Young-Sun Lee. 2003. Monetary Cooperation in East Asia: Exchange Rate, Monetary Policy, and Financial Market Issues. Korea Institute for International Economy Policy. Seoul, Korea. Kwon, O. Yul. 1998. “The Korean Financial Crisis: Diagnosis, Remedies, and Prospect.” Journal of the Asia Pacific Economy, 3(3): 331-357. Lamfalussy, Alexandre. 2000. Financial Crises in Emerging Markets. Yale University Press. Leblang, David. 2001a. “The Political Economy of Speculative Attacks in the Developing World,” Manuscript, Department of Political Science, University of Colorado. Leblang, David. 2001b. “To Devalue or to Defend: The Political Economy of Exchange Rate Policy .” Manuscript Collection, Department of Political Science, University of Colorado, Boulder. Colo. Levy Yeyati, Eduardo, and Federico Sturzenegger. 2000a. “Classifying Exchange Rate Regimes: Deeds vs. Words.” Mimeo. Universidad Torcuato Di Tella, available at http://www.utdt.edu/~ely. ______. 2000b. “Exchange Rate Policy and Economic Performance.” CIF Working Paper #7. Argentina: Universidad Torcuato Di Tella. Makin, Tony. 1999. “The Great East Asian Capital Flow Reversal: Reasons, Responses and Ramifications.” The World Economy: 407-419. Miniane, Jacques. 2000. “A New Set of Measures on Capital Account Transactions.” Manuscript Collection, John Hopkins University, Baltimore: Md. Niles, Kimberly. 2001. “Indonesia: Cronyism, Indecisiveness, and Economic and Political Meltdown.” in Shale Horowitz and Uk Heo, Editors. The Political Economy of International Financial Crisis:Interest Groups, Ideologies, and Institutions. Rowman and Littefield. Nitithanprapas, Ekniti and Thomas Willett. 2000. “A Currency Crises Model That Works: A Payments Disequilibrium Approach,” Working Paper, Claremont Graduate University. ______, Isriya Nitithanprapas, and Thomas D. Willett. 2002. “Classifying Exchange Rate Regimes.” Working Paper, Claremont Graduate University. Nitithanprapas, Isriya. 2002. Empirical Linkages Between Exchange Rate Regimes and Currency Crises. Unpublished Ph.D. Dissertation, Claremont Graduate University. ______, Sunil Rongala, and Thomas D. Willett. 2002. “The Role of Capital Controls and Currency Regimes in the Asian crises.” Presented at the 2002 annual meetings of 70 the American Political Science Association in Boston, MA, Working Paper, Claremont Graduate University. Nitithanprapas, Ekniti. 2000. Composite fundamental indicators of vulnerability to financial contagion. Unpublished Ph.D. Dissertation, Claremont Graduate University. Noble, Gregory and John Ravenhill. 2000. The Asian financial crisis and the architecture of global finance. Cambridge Asia-Pacific Studies. Cambridge; New York and Melbourne: Cambridge University Press. Pempel, T.J. 1999. The Politics of the Asian Economic Crisis, Ithaca and London: Cornell University. Porison, IIelene. 2000. “How Do Countries Choose Their Exchange Rate Regime.” IMF Working Paper No. 01/46. Washington D.C.: International Monetary Fund, April. Portes, Richard. 1999. “Global Financial Markets and Financial Stability: Europe's Role” CEPR Discussion Paper: 2298 November. Quinn, Dennis. 1997. “The Correlates of Change in International Financial Regulation.” American Political Science Review, 91, Issue 3: 531-551. Radelet, Steven; Sachs, Jeffrey D. 1998. The East Asian Financial Crisis: Diagnosis, Remedies, Prospects. Brookings Papers on Economic Activity v0, n1 (1998): 1-74. Rajan, Ramkishen. 2001. “(Ir)relevance of Currency-Crisis Theory to the Devaluation and Collapse of the Thai Baht”, Princeton Studies in International Economics No.88, International Economics Section, Princeton University, February. Rajan, Ramkishen, Rahul Sen, and Reza Siregar. 2002. “Hong Kong, Singapore and the East Asian Crisis: How Important were Trade Spillovers?” The World Economy, Vol.25, No.4, April: 503-538. Rakshit, Mihir. 2002. The East Asian Currency Crisis. Oxford University Press. Reinhart, Carmen, and Kenneth Rogoff. 2002. “The Modern History of Exchange Rate Arrangements: A Reinterpretation.” Mimeo, International Monetary Fund. Washington D.C. Rodrik, Dani. 1998. “Who Needs Capital-Account Convertibility?” in Stanley Fischer and others, Should the IMF Pursue Capital-Account Convertibility? Essays in International Finance No. 207, International Finance Section, Department of Economics, Princeton University, May 1998. Rongala, Sunil. 2003. Capital Controls and the Asian Currency Crisis. Unpublished Ph.D. Dissertation, Claremont Graduate University. 71 Rossi, Marco. 1999. “Financial Fragility and Economic Performance in Developing Economies: Do Capital Controls, Prudential Regulation and Supervision Matter?” IMF Working Paper No. 99/96, International Monetary Fund, Washington D.C. Sachs, Jeffrey, Aaron Tornell, and Andrés Velasco. 1996. “Financial Crises in Emerging Markets: The Lessons from 1995.” Brookings Papers on Economic Activity, No.1:147215. Salvatore, Dominick. 1999. “Could the Financial Crisis in East Asia Have Been Predicted?” Journal of Policy Modeling 21: 341-347. Soros, George. 1998. “The Crisis of Global Capitalism.” Public Affairs, NY. Stiglitz, Joseph. 1998. “Sound Finance and Sustainable Development in Asia.” Keynote Address to the Asia Development Forum, Manila, the Philippines, March 12, 1998. Summers, Lawrence. 1995. “Summers on Mexico: Ten Lessons to Learn.” The Economist December 23 1995-January 5 1996. pp 46-48. Tirole, Jean. 2002. Financial crises, liquidity, and the international monetary system. Princeton University Press, Princeton and Oxford. Tornell, Aaron. 1999. “Common Fundamentals in the Tequila and Asian Crises.” Working Paper 7139. NBER Van den Handel, Cheryl. 2002. “The Quinn Openness Scale and The Magnitude of Restrictions On Current and Capital Accounts.” Claremont mimeo. The Wall Street Journal. 1998. “Currency Controls Are Getting A Hearing Amid Asian crises.” By Wessel, David and Robert Davis, September 4, 1998. Weber, Axel. 1995. “Exchange Market Mayhem: Discussion.” Economic Policy 21:300-306. Wihlborg, Clas and Thomas Willett. 1997. “Capital Account Liberalization and Policy Incentives: An Endogenous Policy View.” In Christine Ries and Richard Sweeney eds. Capital Controls in Emerging Economies Westview Press, Colo. Willett, D. Thomas. 1999. “Did the Mexican Bailout Really Cause the Asian Crisis?” Claremont, Claremont Policy Briefs (March). ----------. 2000. “International Financial Markets as Sources of Crisis or Discipline: The Too Much, Too Late Hypothesis”. Princeton Essays in International Finance (May). 72 ----------, Manfred W. Keil, and Young Seok Ahn. 2002. “Capital Mobility for Developing Countries May Not Be So High.” Journal of Development Economics v68, n2 (August 2002): 421-34. ----------.2003. “Fear of Floating Need Not Imply Fixed Exchange Rates,” presented at the Fordham CEPR Conference on Dollarization and Euroization, Open Economies Review ----------, and Aida Budiman, Arthur Denzau, Gab-Je Jo, Cesar Ramos, and John Thomas “The Falsification of Four Popular Hypotheses About International Financial Behavior During the Asian Crisis,” The World Economy (forthcoming). ----------, and Nephil Maskay. 2003. “Some Conceptual Distinctions Relevant for Applied OCA Analysis,” presented at the annual meetings of the Western Economic Association, Denver, July 2003. Zhang, Hui. 2001. The Causes and Impacts of International Financial Liberalization. Unpublished Ph.D. Dissertation, Claremont Graduate University. 73 Table 6 Coding Methodology of Capital Controls Measures Name of Measure Author/Coder Scoring Methodology Number of Countries and Time Period of Data Availability Capital Ninety developing countries Capital 1975-1997. Glick Hutchison and Reuven Glick and 0 = No Michael Hutchison. Controls. 1 = Controls.31 Johnston IMF Barry Johnston et al. Continuous measure from 0 to 1 where a higher number indicates more capita controls. Miniane IMF Jacques Miniane Continuous measure from 0 to 1 where a higher number indicates more capita controls. Quinn-Type Dennis Quinn Continuous measure from 0 to 14 where a higher number indicates a higher degree of financial liberalization. Includes both the current and capital account. Abiad-Mody Abdul Abiad and Measures capital Ashoka Mody account liberalization on a scale of 0-3 with increments of 1. A higher number indicates a higher degree of liberalization. Abiad-Mody Fin Abdul Abiad and Measures financial Ashoka Mody liberalization on a scale of 0-15 with increments of 1. A 31 Fifty two countries Data available only for 1996. Thirty two countries. 1983-1994 Fifty countries 1995-2000. Forty four developing countries. 1970-1999. Thirty six countries. 1973-1996. Thirty six countries. 1973-1996. Leblang (2001b) uses the same scoring methodology for ninety emerging market countries from January 1985 to December 1998. 74 Adapted Quinn Cheryl Handel Van higher number indicates a higher degree of liberalization. Den Continuous measure from 0 to 4 where a higher number indicates a higher degree of liberalization on the capital account. Also has separate capital inflows and outflows measures 75 Forty eight countries. Data complete from 1996-1999. Data for some countries available before 1996. Table 7 Dependent and Independent Variables Countries ARGENTINA RER -4.62 CA/GDP -2.86 Lboom StdRes CriMax CriEnd CriEndNoI 15.62 1.44 1.63 -3.41 -6.53 BRAZIL 22.14 -0.84 42.36 0.82 11.77 11.20 6.33 CHILE 12.83 -3.7 41.92 0.51 4.70 1.13 1.95 CHINA 7.67 0.25 39.52 0.25 -1.36 -5.93 -8.18 COLOMBIA 22.8 -3.17 50.65 0.67 8.52 8.23 11.24 EGYPT 18.89 3.99 43.06 0.17 0.57 0.33 0.29 HONGKONG 9.99 2.5 17.11 2.72 -6.18 -12.83 -18.65 HUNGARY 0.34 -5.86 -21.04 0.48 2.97 1.95 3.52 INDIA -4.25 -1.17 24.46 0.31 5.58 5.55 6.80 INDONESIA 3.22 -2.29 49.37 1.78 53.19 45.08 54.12 JORDAN -3.31 -8.29 17.51 0.42 -2.57 -11.72 -17.74 KOREA 3.81 -1.62 44.09 2.12 47.15 47.15 65.54 MALAYSIA 7.93 -5.77 104.62 0.60 26.74 25.69 37.98 -19.62 -4.12 -43.18 1.23 -0.22 -7.20 -9.78 MOROCCO 5.32 -1.87 29.81 0.40 3.32 1.16 2.39 PAKISTAN -1.16 -4.35 12.19 2.52 4.68 4.63 6.60 PERU 1.87 -5.83 127.10 0.64 1.70 0.19 -0.17 PHILIPPINES 17.29 -3.9 127.06 0.93 28.30 28.13 38.61 SINGAPORE SOUTH AFRICA 6.83 13.78 52.96 2.43 11.39 11.39 14.39 -7.65 -0.01 26.27 3.03 2.50 -2.00 -2.87 SRI LANKA 5.28 -5.12 18.06 0.23 5.09 -2.81 -6.21 THAILAND 6.71 -6.5 55.41 1.51 37.47 31.04 49.86 TURKEY 30.06 -0.97 76.46 0.86 14.69 10.06 11.33 URUGUAY 5.52 -1.46 26.32 2.71 -1.07 -5.24 -6.91 VENEZUELA 17.82 2.03 -50.94 0.27 2.38 -1.73 -2.50 ZIMBABWE 15 -5.62 -2.33 1.66 48.97 44.85 63.77 MEXICO RER – Real exchange rate appreciation. CA/GDP – Current account to GDP deficit. Lboom – Lending Boom StdRes – Short term debt to reserves. CriMax – Crisis Index calculated from the maximum value of the percentage change from June 1997 to December 1997 (monthly) in exchange rate, reserves, and interest rates. CriEnd Crisis Index calculated from the end value of the percentage change between June 1997 to December 1997 in exchange rate, reserves, and interest rates. CriEndNoICrisis Index calculated from the end value of the percentage change between June 1997 to December 1997 in exchange rate and reserves. 76 Appendix 1 The STV Crisis Model and Extensions The model that we used in this paper is a variant of the model developed in the highly influential paper by Sachs, Tornell, and Velasco (1995). The model was also used in Furman and Stiglitz (1998) and Bussiere and Mulder (1999). STV analyzed countries which would be vulnerable to currency attacks as a result of the 1994 Mexican Peso crisis by testing a crosssection of 20 emerging countries. They regressed the crisis index which is the variance-weighted average of the percentage change in the nominal exchange rate and international reserves32, on three fundamental factors: real exchange rates, banking fragility proxied by the percentage change in the bank loans to the private sector, and the ratio of M2 to international reserves. The setup of the STV model is as follows: IND = b1 + b2 (RER) + b3 (LB) + b4 (DLR * RER) + b5 (DLR * LB) + b6 (DLR * DWF * RER) + b7 (DLR * DWF * LB) + e. Where • IND is a weighted average of the devaluation rate with respect to the US dollar and the percentage change in foreign exchange reserves. (between end November 1994 and June 1996). A higher IND means a higher devaluation or a greater fall in reserves. • RER (real exchange rate appreciation) and LB (lending boom) capture the effects of fundamentals on the crisis index in countries with high reserves (DLR = 0) and strong fundamentals (DWF = 0). A country has high foreign reserves if its ration of M2 to reserves is in the lowest quartile of the sample. (In this sample, high foreign reserves is defined as when M2 to reserves is less than 1.8). A country has strong fundamentals if its real depreciation is 32 In their original development of a crisis index Eichengreen, et al (1994) (1995) also include the interest rate, but sufficient high quality data is not available to include this for developing countries. 77 in the highest quartile in the sample and its bank lending boom is in the lowest quartile. (In this sample, high fundamentals are defined as when LB is below 0% and the RER appreciation is less than 5%).33 STV indicate that a country with real exchange rate appreciation and banking problems would suffer a severe crisis only if the country has both weak fundamentals and low international reserves. STV found that real exchange rate appreciation and lending booms have positive effects on the crisis severity in countries with low international reserves and weak fundamentals, and do not have significant effects in countries with low reserves and strong fundamentals. The model that we use in this paper is a variant model that was used in Nitithanprapas and Willett (2000) which in turn was a variant of the STV model. The innovation of the N&W model was the use of the composite variable which was intended to capture the effects of certain variables based on a threshold level of another variable. This model used 26 countries in its sample.34 The setup of the model of the N & W (2000) is as follows: IND = b1 + b2 (COMi) + b3 (LB) + b4 (COMi * Dhr) + b5 (LB * Dhr) + ε • IND = Indicator-weighted average of the exchange rate change with respect to the U.S. dollar and the percentage change in foreign reserves. • COM1 = Composite variables of balance of payments disequilibrium (i = the number of the composite variable), where COM 1 = (Current Account/GDP) if (Budget Deficit/GDP) > x Percent35, where x is initially taken as 3 percent,36 and 0 otherwise. 33 Several authors have questioned the use of rankings within a sample as opposed to absolute measures. See comments by Furman and Stiglitz (1998) and comment by Cooper on Sachs et al (1996). 34 The complete list of countries is Argentina, Brazil, Chile, China, Colombia, Egypt, Hong Kong, Hungary, India, Indonesia, Jordan, Malaysia, Mexico, Morocco, Peru, Pakistan, Philippines, Singapore, Sri Lanka, South Africa, South South Korea, Thailand, Turkey, Uruguay, Venezuela and Zimbabwe. 35 In all cases, the size of the threshold effects assumed are somewhat, but not entirely, arbitrary. 78 • COM2 = Real Effective Exchange Rate if (Current Account/GDP) < y percent, where y is initially taken as –5 percent, 37 and 0 otherwise • COM3 = (Current Account/GDP) if Real Exchange Rate Appreciation > percent, where z is initially taken as 10 percent, 38 and 0 otherwise. • COM4 = (Current Account/GDP)– (FDI/GDP) if Real Exchange Rate Appreciation > w percent, and 0 otherwise. • LB = Percentage change in the ratio of the claims on private sectors by banks to GDP over the preceding four-year period.39 • Dhr = Dummy for high international reserves, where Dhr = 1 if the ratio of M2 to the stock of foreign reserves is less than the lowest quartile of the sample40 and 0 if Otherwise. The multiplication of the COMi and Dhr, and the LB and Dhr are the interaction terms. This means that when a country lacks high international reserves, the effect of a composite external imbalance and a lending boom would be captured by b2 and b3. When a country has high international reserves, on the other hand, the effects of a composite external imbalance and a lending boom are captured by b2 + b4 and b3 + b5. The hypotheses on the signs and the significant effects of coefficients are as follows. Firstly, a composite variable (COMi) matters when a country has a lack of high international reserves, i.e., b2 > or < 0, but does not matter 36 Salvatore (1999) argues, “although there is no specific level of the budget deficit that spells inevitable trouble, a budget deficit of 3 percent of GDP or larger can usually be taken as a serious warning signal of possible future problems for the nation”(p. 343). 37 Summers (1995) suggests “that close attention should be paid to any current account deficit in excess of 5% of GDP” (p. 47). 38 The optimal threshold of real exchange rate appreciation for both currency and banking crises has been estimated to be 10 percent (Goldstein, 1997). 39 This period of lending booms is used so as to be able to compare with the previous studies. However, different time periods were used to test for the sensitivity analysis. 40 This definition is used so as to be able to compare with the previous studies. Different definitions were used in the sensitivity analysis. 79 when a country has high international reserves, i.e., b2 + b4 = 0. Secondly, a lending boom (LB) also only matters when a country lacks high international reserves, i.e., b3 > 0, but it does not matter whether a country has high international reserves, i.e., b3 + b5 = 0. Our model does not have the interactive term between fundamental variables and the dummy for weak fundamentals, which is created by ranking the same fundamental factors in the quartile scale. By excluding the dummy for weak fundamentals, we are able to aggregate other important fundamental factors into the composite variable. The ideal estimation would be to consider all of the important factors and the dummy for weak fundamental factors simultaneously. Nonetheless, given the small sample size, this would cause problem from lack of degrees of freedom. Thus, to focus on interrelations among fundamental variables that are of particular interest, the estimation method considers only the interaction between the composite of fundamental indicators and the dummy for high international reserves. In our crisis index, we included the change in the interest rate. The reason for including interest rates is that without them the data shows no evidence of the strong speculative attack on the Hong Kong dollar following the depreciation of the Taiwanese dollar in October 1997. For comparison with other studies, however, we also use a crisis index without interest rates. 41 The composite variable is the same as COM2 in N&W and we used short-term debt to reserves instead of M2 to reserves. The lending boom variable that is used in our paper is for a thirty six month period as opposed to the forty eight month period used in N&W. Since our empirical work is also focused on capital controls and exchange rate regimes, apart from the fundamental variables, those two variables were included. Our exchange rate variable has two classifications, namely the two-way and three-way. Our initial results indicate that current account deficits and 41 The construction and evolution of the crisis indices in explained in Appendix II. 80 high lending booms tend to increase the vulnerability of a country to a currency crisis. It also indicates that the presence of capital controls increases vulnerability. The sample of countries used in our study is the same as N&W. Our basic estimating equation based on two-way grouping of exchange rates is: CRISIS = b1 + b2 (COMPOSITE) + b3 (LB) + b4 (STD/RES) + b5 (CONTROL) + b6 (FLOAT) +ε Our estimating equation based on three-way grouping of exchange rates is: CRISIS = b1 + b2 (COMPOSITE) + b3 (LB) + b4 (STD/RES) + b5 (CONTROL) + b6 (FLOAT1) + b7 (CRAWL1) + ε • CRISIS = There are three different crisis indices used. They are • Crisis Index (Maximum) = An equal weighted average of the maximum value of the percentage change in exchange rate with respect to the U.S. dollar, the minimum value of the percentage change in foreign reserves, and the maximum value of the percentage points change in interest rates42 between June of 1997 to December of 1997. • Crisis Index (End) = An equal weighted average of the change in value of the percentage change in exchange rate with respect to the U.S. dollar, the change in value of the percentage change in foreign reserves, and the value of the percentage points change in interest rates from June of 1997 to December of 1997. • Crisis Index (No Interest) = An equal weighted average of the change in value of the percentage change in exchange rate with respect to the U.S. dollar, and the 42 Money market rates are generally used for most countries. Lending rates are used where money market rates are not available. In the case of Hong Kong, the T-bill rate is used. 81 change in value of the percentage change in foreign reserves from June of 1997 to December of 1997. • COMPOSITE = Real effective exchange rate appreciation if (Current Account/GDP) < y percent, where y is initially taken as –5 percent43, and 0 if otherwise. The real effective exchange rate was calculated using several methodologies. The different real exchange rate measures used were: • Caramazza et al (2000) – It is the three year log change of the 12 month pre-crisis average of the real exchange rate. • Improved RER44 – It is the percentage change of the real exchange rate from May of 1993 to May of 1997. • Bussière and Mulder (1999) - It is the percentage change of the real exchange rate from June of 1993 to June of 1997. • Sachs, Tornell, and Velasco (1996) – It is the percentage change of the real exchange rate from 1992 to 1996 and RER index is the weighted average of bilateral RER in relative to the U.S dollar, Deutsche Mark, and the Japanese Yen. • Ahluwalia (2000) – It is the percentage change in the real exchange rate between December of 1993 and December of 1996. • INS45 – It is the percentage change in the real exchange rate between the monthly average of 1992 and the monthly average of 1996. 43 The real exchange rate appreciation is calculated as the 3-year log change of 12-month pre-crisis average of real effective exchange rate(from IMF database). The current account deficit is calculated as the 4-year pre-crisis average of current account balance to GDP (from world development indicator database). 44 This methodology was used in Nitithanprapas and Willett (2000). 45 This methodology was used in Nitithanprapas and Willett (2000). 82 • Radelet and Sachs (1998) - It is the percentage change in the real exchange rate three years prior to the crisis. It is calculated as ratio of trade-weighted average of foreign CPI to domestic CPI. • LENDING BOOM = Percentage change in the ratio of the increase in banking sector credit to the non-government sector (in real term) over the preceding thirty-six month period. • STD/RES = The ratio of short-term debt to reserves. (using the data from the joint database of the of OECD, BIS, and IMF). (For data on the above four variables, see table 1.) • CONTROL = The controls on capital transactions. There are seven different controls used. (For a detailed explanation for the measures, see Appendix 4. The data for the controls is in table 2.) • FLOAT = The dummy for floating regimes in the two-way classification. • FLOAT1 = The dummy for floating regimes in the three-way classification. • CRAWL1 = The dummy for crawling regimes in the three-way classification. • PEG = The dummy for pegged regimes. (For a more detailed discussion of the classification of exchange rate regimes see Appendix 3. See tables 3 for classifications of countries in sample.) The results of our empirical work are: Two-way classification of Exchange Rates (pegged vs flexible). Crisis Index (Maximum) The composite variable has a positive relation with the crisis variable and it shows a very high level of significance irrespective of the bias of the exchange rate. There is a positive relation between short-term debt to reserves and the crisis variable but there is little significance and the 83 lending boom variable is positive and is significant irrespective of the bias of the exchange rate. All the capital controls variables are positive but only the IMF, Glick, and Abiad-Mody measures are significant irrespective of the bias of the exchange rate while the Abiad-Mody Fin measure is significant only when the exchange rate is biased to a pegged regime. The floating exchange rate regime is positive and not significant and the pegged regime variable is negative and not significant. The adjusted R2s vary from 0.21 to 0.42 and the coefficients of variables are stable. See tables 7-8. Crisis Index (End) The results for the regressions that use the crisis index (end) as the dependent variable and very similar to the results when crisis index (maximum) is the dependent variable with the exception that the Abiad-Mody Fin measure is significant when the exchange rate is biased both to a floating regime as well as a pegged regime. The adjusted R2s vary from 0.21 to 0.48 and the coefficients of variables are stable. See tables 5-6. Crisis Index (No Interest Rate) The results of the regressions using the crisis index (no interest rate) as the dependent variable are similar to the results when crisis index (end) is the dependent variable. The adjusted R2s vary from 0.26 to 0.51 and the coefficients of variables are stable. See tables 9-10. Three-way classification of Exchange Rates (Pegged, Intermediate, and Flexible). Crisis Index (Maximum) The composite variable has a positive relation with the dependent variable and has a high level of significance irrespective of the bias of the exchange rate. There is a positive relation between short-term debt to reserves and the crisis variable but there the significance depends on the bias 84 of the exchange rate. There is increased significance when the bias is to a crawling or pegged regime. The lending boom variable is positive and significant as well. The capital controls measures are all positive but only the IMF, Glick, and Abiad-Mody measures are all significant irrespective of the bias of the exchange rate while the Abiad-Mody Fin measure is significant only when the exchange rates are biased to the crawling and pegged regimes. The sign of the floating exchange rate variable is negative while the sign of the crawling exchange rate regime is positive. Both variables show no significance. The adjusted R2s vary from 0.35 to 0.52 and the coefficients of the variables seem to be stable. See tables 14-16. Crisis Index (End) The results of the regressions where the crisis index (end) is the dependent variable are very similar to the results when crisis index (maximum) is the dependent variable with the exception that the Abiad-Mody Fin measure is significant irrespective of the bias of the exchange rate. The adjusted R2s vary from 0.39 to 0.53 and the coefficients of variables are stable. See tables 11-13. Crisis Index (No Interest Rate) The results of the regressions where the crisis index (no interest) is the dependent variable are very similar to the results when crisis index (end) is the dependent variable. The adjusted R2s vary from 0.37 to 0.52 and the coefficients of variables are stable. See tables 17-19. The results when the different composite variables were run using the three different crisis measures (dependent variables) and three-way classification of exchange rates with lending boom and the short term debt to reserves ratio as the other independent variables. The results indicate that while all the composite variables had a positive sign, only the composite variables 85 that used the Caramazza et al, Bussière-Mulder, and INS methods were significant irrespective of the crisis measure used or the bias of the exchange rate. The Caramazza et al methodology was used in Nitithanprapas, Rongala, and Willett (2002) with good results and is used in this study. A further sensitivity analysis was done using the Caramazza et al method where the current account deficit threshold was reduced to 3 percent. The results showed the composite variable had vastly reduced significance and low adjusted R2s. This is consistent with Summers (1995) who stated that a current account deficit of 5 percent was a good indicator of a country’s increased vulnerability to a currency crisis. 86 Appendix 2 Crisis Indices46 In order to determine the degree of vulnerability to financial contagion or the severity of crisis, an operational definition of crisis is needed. A currency crisis is defined as speculative pressures in the foreign exchange markets. Following Eichengreen, Rose, and Wyplosz (1996), the speculative pressures can be reflected in a change in nominal exchange rate (devaluation/depreciation), a change in international reserves, or a change in interest rates.47 The basic idea is that when there are speculative runs on currency, the government has three policy choices. First, it can let the exchange rate depreciate. Second, it can intervene in the foreign exchange markets by selling international reserves. Lastly, it can increase interest rates to entice capital inflows in order to offset the speculative pressures on domestic currency. Some countries may use a combination of these three policy options to absorb speculative pressures. Because of the limited availability of comparable interest rate data for emerging countries, early studies of the Asian crises excluded the change in interest rate from the crisis index. Thus, the crisis indices are a weighted average of the depreciation rate of the nominal exchange rates and the percentage change in international reserves. An increase in the crisis index, caused by an increase in the depreciation/devaluation rate or a drop in foreign exchange reserves corresponds to a more severe crisis. 48 The difficulty with this concept is how to weight the importance of changes in reserves versus changes in exchange rates (and interest rates where included). As Eichengreen et al (1996) argue, ideally the weights should be derived from the excess demand for foreign 46 This appendix is based on Nitithanprapas and Willett (2000). This concept is essentially the exchange market pressure variable advanced by Girton and Roper (1977). 48 Some papers use discrete measures of crisis, typically when the index value deviates from its mean by more than two or three standard deviations. 47 87 exchange from an empirical model of the exchange rate, i.e. by the slope coefficient that reflects how much official intervention (change in reserves) would be required to avoid a one percentage point change in the exchange rate. The problem is that, as they note, there is little agreement within the profession about the most appropriate theoretical model of the foreign exchange market and none of our models fit well empirically. Thus they note that it is important to check the sensitivity of the results to the weights used. Unfortunately, however, most of the subsequent literature has paid little attention to this important warning.49 Instead most studies have followed Eichengreen et al’s suggestion to use calculations of precision, the inverse of the variance, of the variables as weights without either reporting these weights or checking the sensitivity of the results to different weights.50 Despite their widespread use, precision weights are inappropriate for use in the calculation of indices of currency crisis. They would be entirely appropriate for the calculation of indices of volatility or crisis based on averages of free market variables such as stock market indices for a number of different countries. The idea is to capture the degree of volatility in a particular market at a particular time in relation to its normal volatility. If, for example, one variable is normally much more volatile than another, then an unweighted average would be excessively heavily driven by the behavior of the more volatile series. Many foreign exchange markets, however, are managed by governments. The relative variance of exchange rate changes and reserve changes will be heavily influenced by the exchange rate regime being followed and 49 An exception is Glick and Hutchison (2000). The only instances of which we are aware that the weights are reported are in Weber’s (1995) discussion of Eichengreen et al’s (1995) original paper. There the weights were 7.5 for changes in the exchange rate and 51.9 for changes in interest rates where changes in reserves are given a unit weight. Weber questioned these weights and suggested an equal weighting instead. 50 88 precision measures will reflect the government’s reaction function, not the slope of the excess demand schedule in the foreign exchange market. As a consequence, precision weights will substantially understate the severity of unsuccessful speculative attacks under fixed exchange rates such as hit Argentina following both the Mexican and Thai crisis and hit Hong Kong during the Asian crises. This is easy to see. Under a regime of narrow band fixed exchange rates, almost all of the effects of incipient payments imbalances will fall on changes in reserves rather than changes in exchange rates. Since reserve changes will have a much higher variance, precision weights will give reserve changes little weight. Thus as long as the speculative attacks were unsuccessful, their magnitude would be understated. By the same token, if a country were successfully knocked off a narrow band peg, the heavy weight given to the subsequent currency depreciation would overstate the magnitude of speculative pressure. One approach to guesstimating the weights would be to assume a foreign exchange market with rational expectations and calculate how much of a change in the exchange rate would be required to bring about an adjustment in the current amount equal to the size of the intervention. This would correspond to a flow equilibrium on the assumption that the intervention would be repeated each period. In this case we could use the Marshall-Lerner conditions to derive the slope coefficient as a function of the sum of the country’s demand elasticity for exports and imports. Since the crisis indices focus on the percentage change in reserves, the initial level of reserves also needs to be specified. Taking the ratio of reserves to imports is also a convenient formulation in relation to the Marshall-Lerner equation.51 51 See Nitithanprapas (2000). 89 A high weight on reserves corresponds to such a long run flow equilibrium for a country with fairly high elasticities and/or ratio of reserves to imports. During crises, however, speculative considerations and risk aversion could lead to short-run slope coefficients that differed substantially from these long-run trade-based calculations. These considerations could cut in either direction. With strongly held expectations, speculative capital flows could add to the elasticity of the excess demand schedule, thus increasing the weight that should be given to changes in reserves. On the other hand, a substantial increase in uncertainty associated with a crisis could make the speculative and excess demand schedules much less elastic, increasing the weight that should be given to exchange rate changes in the crisis index. (For arguments that this was frequently the case during the Asian crises see Willett (2000)). Thus there appears to be a substantial range of plausible weights for the crisis index. Fortunately, however, sensitivity analysis in Nitithanprapas and Willett (2000) found that major results were robust with respect to substantial changes in the weights in the crisis index.52 We believe that it is important, however, to recognize that the precision of precision weights is largely spurious. We also believe that despite the difficulties of getting good data it is important to include interest rates in crisis indices. Indeed, looking first in changes in reserves and exchange rates the major speculative attacks on Hong Kong in October 1997 would be entirely missed in the standard emerging market crisis index, yet many point to this as the event that sparked the currency crisis in Korea. It also had much larger effects on financial markets outside of Asia than did the Thai crisis. The speculative pressures ended up showing up in published statistics only in interest rate increases and stock market declines. 52 While we report only the results for the one to one, four to one, and one to four ratios, as we think these are a plausible range, we also checked ratios running from ten to one to one to ten and found that even variations this large did not have major effects on the results. 90 Several recent studies have included interest rates in emerging market crisis indices53. Again, it isn’t clear what weights should be assigned to interest rates in the composite indices. Even if one goes with the equal weights strategy it will make a difference whether changes in interest rates are specified in terms of percentage change or basis point changes from the base level, i.e., should an increase in interest rates from say 4 to 6 percent be equated with a change in exchange rates of 2 or of 50 percent. We believe analytically that use of the basis point change is more relevant and use this formulation in our index since to hold the incentives for capital flows equal the increase in interest rates would require a making 2 rather than 50 per cent expected depreciation of the exchange rate. Another important issue that has been given little explicit attention in the literature is the issue of the duration of speculative pressures. Where there is some reversal of changes than merely looking at changes pre crisis to an arbitrary end date could substantially understate peak pressures. There is also the problem that variables shortly before the crisis hits may already reflect some crisis expectations. Our overall conclusion is that there is no really good operational way to construct composite indices and that their use in large n mechanical ways needs to be supplemented by careful study of the behavior of each of the relevant variables. 53 See Degregorio and Valdes (1999) and Caramazza et al (2000). 91