POLICY RESEARCH WORKING PAPER 2259 Predicting Currency Markets have had limited success predicting crises and Fluctuations and Crises might do better by drawing on private information . . ~~~~~~~~~~~~~~~~available tO resident Do Resident Firms Have an available tonaesident enterprise managers, who Informational Advantage? seem to know better than markets about future Daniel Kaufmann movements in exchange Gil Mehrez rates. Sergio Scbmukler The World Bank World Bank Institute Governance, Regulation, and Finance U December 1999 | POLICY RESEARCH WORKING PAPER 2259 Summary findings Kaufmann, Mehrez, and Schmukler investigate whether conventional forecasts of future volatility and changes in resident enterprise managers have an informational the exchange rate, which are based on economic advantage about the countries in which they work. They fundamentals or interest rate differentials. propose a method for extracting information available to They find that the local business community perceived resident managers but unknown to investors and in advance the recent crises in the Republic of Korea, forecasters. Russia, and Thailand, but not those in Indonesia and They test their hypothesis of informational advantage Malaysia. using a unique data set, the Global Competitiveness Markets have had limited success predicting crises and Survey. The survey asks local managers about their might do better by drawing on private information outlook for the country in which they reside. available to resident enterprise managers, who seem to They find that local managers do have useful private know better than markets about future movements in information. Local managers' responses improve on exchange rates. This paper - a product of Governance, Regulation, and Finance, World Bank Institute - is part of a larger effort in the institute to understand the roles of transparency and governance. Copies of the paper are available free from the World Bank, 1818 H Street, NW,Washington, DC20433. Please contactEmily Khine, roomMC3-341, telephone 202-473-7471, fax 202-522-3518, email address kkhine@worldbank.org. Policy Research Working Papers are also posted on the Web at www.worldbank.org/research/workingpapers. The authors may be contacted at dkaufmann@aworldbank.org, mehrezg@gunet.georgetown.edu, or sschmukler@worldbank.org. December 1999. (34 pages) The Poliy Research Working Paper Sedbes disseminates the findgs of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentationis are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the autbors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Produced by the Policy Research Dissemination Center Comments Welcome Predicting Currency Fluctuations and Crises: Do Resident Firms Have an Informational Advantage? Daniel Kaufmann, Gil Mehrez, and Sergio Schmukler+ The World Bank, Georgetown University, and The World Bank JEL Classifilcation Codes: F3, F4, Gl Keywords: expectations; informational advantage; asymmetric information; local investors; managers; financial crises; exchange rate; prediction; survey; currency forecast + We thank Ian Wilson for the data on Mutual Funds and Ilan Goldfajn for the data on currency forecasts. We gratefully thank Akiko Terada and Pablo Zoido-Lobaton for excellent research assistance. Contact address: The World Bank, 1818 H Street NW, Washington, DC 20433. E-mail address: sschmukler@worldbank.org L Introduction The recent crises in Mexico, East Asia, Russia, and Brazil have caught many economists, policymakers, and financial markets by surprise. Mexico was forced to devalue its currency shortly after joining the OECD. East Asian countries, with highly regarded policies and long history of high growth and stability, suddenly suffered balance of payments crises. Russia was abruptly cut off from international credit, having to restructure its sovereign debt. While the crisis in Brazil was at the end widely expected, the economic recovery following the devaluation of the real has been, however, surprisingly fast. In the aftermath of each crisis, the efficient functioning of international financial markets and the nature of crises are revisited. Are crises self-fulfilling? Do investors act as a herd? Are international investors less informed? At the heart of these questions is the flow of information and transparency. Information is a key element in the theoretical analysis of recent crises. For example, Calvo (1999), Calvo and Mendoza (1999), Kodres and Pritsker (1998), and Rigobon (1998) show that costly information about international investments can produce herding and contagion effects. One explanation goes as follows. As investment opportunities grow, investors diversify their portfolio, holding assets in countries in which they have little information. Therefore, a crisis in one country might prompt uninformed investors to revise their expectations with respect to other countries. Moreover, the fact that informed investors sell assets from a crisis country, for reasons unrelated to fundamentals, might lead uninformed investors to pull out from other emerging markets. Uninformed investors try to extract information from the action of informed market participants. This type of reaction generates contagion or spillover effects across countries. Fundamentally sound economies are affected by panics, herd behavior, and self-fulfilling expectations. Empirically, these inferior information theories are consistent with many findings. Frankel and Schmukler (1996, 1998) report instances of local investors exiting their markets before international investors at times of crisis - consistent with locals being better informed. Brennan and Cao (1997) look at U.S. equity flows to a sample of developed and emerging markets. They find evidence consistent with asymmetric information. Choe, Kho, and Stulz 1 (1999) and Kim and Wei (1999) analyze data from Korea to study trading patterns by resident and intemational investors. They conclude that international investors engage in positive feedback trading. They sell past losers and buy past winners. The evidence from Korea suggests that traders outside the country have less information about the Korean economy than resident investors do. In the present paper, we use an exclusive data set, the Global Competitiveness Survey (GCS), to study the existence of informational advantages. The GCS is a questionnaire answered by managers located in countries around the world. The survey is collected for the Global Competitiveness Report, produced by the World Economic Forum of Davos and the Harvard Institute for International Development. The survey gathers each manager's perspective on the economic, political, and institutional situation of the country in which each manager resides. The data comprises responses from surveys conducted at the end of 1995, 1996, and 1997. The GCS data set gives us a unique opportunity to directly test the informational advantage by resident managers. The previous studies on asymmetric information are forced to infer asymmetric information (by different groups of investors) from aggregate flows or price series. Those papers need to assume which groups of investors determine certain flows or prices. On the other hand, the GCS directly reveals what managers think about each country. The timing of the questioners is such that they precede the crises in Asia, Russia, and Brazil. Hence, the survey can be used to test not only general informational advantages, but also whether managers predicted recent crises. The objective of this paper is twofold. First, we take advantage of the unique data set to investigate whether local managers have private infonnation, not captured by available macro data or by other market indicators. Second, we propose a way of extracting the information available to local managers, but unknown to investors. The paper is organized as follows. Section II describes some evidence of asymmetric information by looking at recent predictions of crises at different market levels. Section HI describes the econometric technique used to extract managers' information. Section IV describes the data and the econometric results. Section V concludes. 2 II. Predicting Financial Turmoil-The Case of the Asian Crisis This section analyzes how different groups of investors and market participants have predicted problems in the economy. We focus on three groups: market indicators, market analysts, and local managers. We compare these indicators across the four Asian countries hit hard by the 1997-98 crisis, namely Indonesia, Korea, Malaysia, and Thailand. We observe these indicators before, during, and after the crisis. IL a Market Indicators To study the market indicators we look at holdings of dedicated emerging market funds, international bank exposure, and country funds. Mutual Funds: Figure 1 plots holdings and flows (net buying) of dedicated emerging market mutual funds. Just before the crisis started, dedicated emerging market mutual funds held 6.66 billion dollars in Indonesia, 9.43 in Korea, 9.01 in Malaysia, and 4.11 in Thailand. Mutual funds holdings changed drastically by December 1997. Mutual fund holdings decreased to 1.71 billion dollars in Indonesia, 2.31 in Korea, 2.21 in Malaysia, and 2.02 in Thailand. Mutual fund holdings had decreased in Thailand and Korea even before the crises. This decrease, however, was mainly due to the decline in stock market prices in these countries and not due to net selling. When controlling for the decline in the stock market, we do not see large selling prior to the crisis. The only exception is Malaysia, were net selling of 1.5 billion per quarter started two quarters before the crisis and another 1.5 billion was sold in the third quarter of 1997, after the crisis erupted. In sum, the evidence from mutual funds does not suggest that investors and fund managers were anticipating a crisis in Asia. International Banks: Commercial banks have played an important role in emerging economies since the 1980s. Their large exposure during the 1982 Latin American debt crisis prompted U.S. banks 3 to reduce their exposure as a share of their capital. Also, the relative participation of commnercial banks in emerging markets has declined as other institutional investors increased their involvement. Commercial banks, still however, have a large exposure in emerging markets. According to the Bank for Intemational Settlements (BIS), intemational banks had 622 billion dollars in claims in all developing countries by the end of 1995. Around 70 percent of this exposure was held by European and Japanese banks. In order to look at the exposure of international banks in Asia we use the data published by the BIS. Figure 2 reports outstanding claims and changes in net claims in each Asian country. The figure displays consolidated cross-border claims in all currencies and local claims in non-local currencies of international banks in Indonesia, Korea, Malaysia, and Thailand between the end-1995 and mid-1997. The exposure is divided into "short term" (up to one year) and "long term" (over one year). The total exposure in Indonesia, Korea, Malaysia, and Thailand was around 260 billion dollars by mid-1997. In Korea, Indonesia, and Malaysia, banks' exposure has been increasing since the end-of 1995, while in Thailand the exposure has been mostly unchanged, after increasing in the beginning of 1996. Among these four countries, only Malaysia has had an increasing proportion of short-term debt. The BIS figures show that, in Thailand, Indonesia, and Malaysia, most of the lending went to the non-bank private sector. While in Korea, most of the lending went to banks. In General, most of the lending in Asia came from European and Japanese banks. The Japanese presence has been high in Asia. However, its relative presence has been declining over time, while the European banks increased their relative importance in the area. One interesting fact is the evolution of net claims (claims minus liabilities). Figure 2 shows that net claims decrease substantially in Korea and Thailand. Given that claims do not decrease, net claims reflect the increase in liabilities of international banks. People from Korea and Thailand increased their claims against the international banking sector. This increase in liabilities seems to be reflecting capital outflows by Korean and Thai investors. In sum, international banks did not decrease their exposure in these countries before the crisis, which suggests that there was no change in banks' expectations between 1995 and the outset of the crisis. At the same time, there seems to be an outflow from domestic residents to 4 the international banks. CountyFunds. Country funds can also describe the reaction of international investors versus domestic investors. Country funds are traded in New York in the secondary markets. Their price is denominated in U.S. dollars. Another value of the fund, the net asset value (NAV), is calculated on a weekly basis. The NAV reflects the value of the underlying assets, mostly traded in the local markets. Small international investors are the usual holders of country funds, while domestic investors and large international investors are the usual holders of the underlying assets. The lack of perfect arbitrage enables us to compare the reaction of prices and NAVs during crisis times. Discounts-equal to the percentage difference between prices and NAVs-reflect the small international investors' sentiment relative to the local and large international investors' sentiment. The analysis of country fund discounts before the Asian crisis in Frankel and Schmukler (1998) show that in the two Thai country fund discounts turned into premia by the end of 1996. From early-1997, the premia of these funds increased steadily. In the case of Indonesia, discounts turned to premia right before the devaluation. When looking at the Korean and Malaysian funds, one can observe that discounts were shrinking right before the devaluation, but turned into large premia only after each country's currencies were forced to devalue. In sum, the evidence indicates that the holders of the underlying assets turned more pessimistic than the Thai country fund holders well before the Thai crisis erupted.' In the case of the other Asian countries, the country fund data do not suggest that expectations between the two groups of investors were divergent in the month previous to the Asian crisis. iH b Market Analysts We now turn to the financial analysts. We investigate how their expectations and evaluations evolved over time, particularly around the Asian crisis. We look at two main groups: currency forecasters and rating agencies. 'Similar evidence was found in Frankel and Schmukler (1996) for the case of the Mexican crisis of 1994-95. 5 Currency Forecasters: The data set from the Currency Forecasters Digest contains the average exchange rate forecast made by multinational companies and by currency traders. The forecasts are for 1, 3, 6, and 12-month horizons. Figure 3 displays the average forecasts from the beginning of 1996 up to April 1998 for Indonesia, Korea, Malaysia, and Thailand. The graphs suggest that in Indonesia forecasts are very much in line with the current spot exchange rate until the end of July 1997. At the end of July 1997, the 12-month forecast was 20 percent above the current spot rate. The 12 month-forecast was 2,900 ruphias per dollar, while the exchange rate was 2,434. By the end of October 1997, the exchange rate had already jumped to 3,405. At that time, expectation had adjusted and the 12-month forecast moved to 4,545 ruphias per dollar. Even though expectations adjusted after the crisis erupted, currency forecasters did not expect the extent of the continuing depreciation of the ruphia. By December 1997, the exchange rate was 5,402 and by January 1998 the exchange rate already reached 12,950 ruphias per dollar. The data from Korea show a similar picture. The exchange rate forecast moved with the spot exchange rate. Forecaster expected a small depreciation of the Korean won. Up to the end of November 1997, the 12-month expected depreciation was around 5 percent, never exceeding 10 percent. The expected depreciation at 1-month, 3-month, and 6-month was even lower. However, between November and December 1997 the Korean won devalued close to 40 percent. In the case of the Malaysian ringgit, a comparable story can be observed. Up to the end of June 1997, the forecasted devaluation was usually below 2 percent, even for the 12-month horizon. During August and September 1997, the ringgit devalued close to 11 percent. In October 1997, expectations adjusted. The 12-month forecasted depreciation jumped to 13 percent. A related story can be told about the Thai baht, the long-term expected depreciation jumped to 12 percent in October 1996 and June 1997, while in February 1997 it was close to zero. In July 1997, the Thai baht devalued 28 percent. To sum, the data from the Currency Forecasters Digest suggests that in countries like Thailand there was a perception that the exchange rate might move in the near future. 6 Currency forecasters perceived some exchange rate movements in Indonesia and Korea, with expected devaluation below 10 percent. In Malaysia, currency forecasters hardly perceived any future devaluation even for the following 12-months. Currency forecasters predicted a small change in the exchange rate. None of the forecast, however, predicted the magnitude of the devaluations triggered after July 1997.2 The currency forecasts appear to revise their predicted exchange rate whenever the spot rate moves. RatingAgencies: Now we turn to investigate the analysts' evaluations of the economy. We use the following sources: Standard & Poor's McGraw-Hill Global Risk Service, Moody's, and Standard & Poor's (S&P) ratings. Standard & Poor's McGraw-Hill Global Risk Service produce ratings by specialists and analysts with a regional focus. The ratings are on a scale of 1 to 100. The ratings represent a probability measure, i.e. the risk that something "bad" might happen. The higher the score, the larger the probability of a crisis. The data includes 33 immediate risk events. In the present, we work with the short-term risk of a 5 percent domestic demand fall and with the short-term risk of a 25 percent equity price fall. Figure 4 plots the evolution of these two variables. The dark gray bars represent values for Indonesia, Korea, Malaysia, and Thailand from the third quarter of 1996 to the fourth quarter of 1997. For comparison, the light gray bars represent values for other South East Asian countries, Latin American countries, and OECD countries. Figure 4 shows that, before the crisis, the probability of a 5 percent demand fall remains fairly steady in all Asian countries as well as in the other regions. The probability increases slightly in Korea, Malaysia, and Thailand in the quarters before the crisis. However, the probabilities remain below 10 percent. On the other hand, after the crisis, the probabilities are updated. For instance, in Indonesia and Thailand, the probability jumps to 70 percent. Figure 4 also displays the probability of a 25 percent equity price fall. These probabilities do not increase before the crisis, except a very small increase in Indonesia, where the probability remains below 35 percent. In Malaysia, the probability is totally flat. In 2 Tlis evidence is consistent with the findings by Goldfajn and Valdes (1998). 7 Indonesia and Thailand, the probability decreases, even after the crisis began. On the other hand, in Korea, the probability remains flat at 10 percent. It only increases in the fourth quarter of 1997 to 35 percent. After the crisis started the probabilities in all other regions increase. Additionally, we analyze the Standard and Poor's (S&P' s) sovereign debt ratings.3 Figure 5 shows how better ratings were obtained in mid 1990s. These ratings dropped sharply in all countries only after the crisis began. Although not reported, the figures from Moody's do not tell a different story. In fact, Ferri, Liu, and Stiglitz (1999) econometrically show that credit rating agencies aggravated the East Asian crisis, after having failed to predict it. I. c Survey of Local Managers We use the GCS data to analyze the perspective of local businessmen.4 Figure 6 shows the responses to two questions asked to local managers about the likelihood of a future recession and the future volatility of the exchange rate. Respondents provide a rating of the country on a scale of 1 (high volatility) to 7 (stable). The data show that businessmen expectations were worse in 1997 relative to 1996 in Korea and Thailand, and to some extent in Malaysia. The most stfiking change occurs in Thailand, where expectations deteriorated sharply between 1996 and 1997. Businessmen thought that a recession and exchange volatility were more likely during 1997 than during 1996, as if they were anticipating future problems. Meanwhile, in the other Asian countries (including Indonesia), in Latin America, and in OECD countries expectations improved. After the crisis, in December 1997, expectations deteriorated dramatically in Indonesia and Malaysia. Indonesia was the country hit the hardest by the crisis. Expectations on the likelihood of a recession increased in the other regions of the world as well. The evidence suggests that managers' expectations were correct. However, one should be careful from drawing conclusions based solely on managers' response for several reasons. First, it is possible that managers simply use the available macro data. That is, if we look at the 3 Both S&P's and Moody's provide different ratings. There are long-term and short-term ratings. There exist ratings on debt, financial sector, and currencies. 4The data is fully described in Section 4. 8 macro data we will reach the same conclusions as the managers. In this case, managers do not provide any information that is not captured by the available information. Second, even if managers have valuable private information, it is possible that their interpretation of the macro data is incorrect. Thus, their response, which is based on their private information and their interpretation of the macro data, might be incorrect. In this case, even though managers' expectations are wrong, their private information is still very valuable. Third, it is possible that managers' characteristics affect their response. For example, managers of export-oriented firms might expect higher exchange rate devaluations than managers of import-oriented firms. These issues are taken into consideration in the next section, where we do the econometric analysis. II d Comparison across Groups The evidence presented so far suggests that there are signs of asymmetric information. On the one hand, neither the market indicators nor the market analysts appeared to have anticipated the Asian crisis. International mutual funds, international banks, currency forecasters, and rating agencies did not react as if they were expecting the crisis. The country fund data and the assets of domestic residents in the international banking sector suggest that there was a divergence in expectations in Korea and Thailand. The data suggests that local residents were aware of deteriorating local conditions in those countries. This is confirmed by the survey of local managers. We have already presented descriptive evidence of informational advantage by local residents. Even though the evidence seems suggestive and is consistent across measures, we would like to formally test differences in information. As mentioned in the introduction, other papers have inferred asymmetric information using data on prices or flows. Here, we use the GCS survey to test econometrically whether managers have informational advantages. 9 II Methodology - Extracting Managers Private Information from the Survey Survey of local business people expectations and beliefs may reveal valuable information, unobservable by policy makers or other investors. Local business managers may possess valuable unobserved information because of poor transparency at the micro level-e.g. the financial stability of institutions or firms-or at the aggregate level-e.g. monetary policy, fiscal policy, or the aggregate economic conditions. One, therefore, should incorporate information known to managers, in addition to all other available information, when evaluating the economic conditions. In this section we describe how to extract the managers' private information from the survey. The managers' response is a combination of their private information and their forecast based on available information. If managers form their expectations rationally and have all available information, then there is no gain in extracting their private information. We could simply use their response as the sole explanatory variable. If managers, however, do not form expectations rationally or do not have all available information, then predicting the economic conditions based solely on their response may be misleading. For example, suppose that managers form expectations on exchange rate volatility based only on GDP per capita, in addition to their private information. That is, managers in countries with high income expect a stable currency, while managers in countries with low income expect a volatile currency. Managers' response by itself, in this case, does not reveal any information about future volatility. One must, therefore, decompose the managers' response into the managers' expectations based on available data and the managers' private information. In other words, managers' response may be incorrect and misleading, but their private information may still be very valuable. Let y, be manager i expectations at time t. The expectation is a function of available information, the manager's characteristics (such as the size of the firm), and the manager's private information e. Yt = f(available information, managers' characteristics) + e . Managers' response, however, is categorical. That is, managers' response is as following: 10 1 if yt < u 2 if < cy < p2 response'= 3 if 82