هم انباشتگی و علیت در بازارهای ارز خارجی نوظهور و آسیایی: شواهدی از بحران های مالی دهه 1990

This article examines the cointegration level, changes in the existence and directions of causality of the foreign exchange (FX) rates in the Asian and emerging markets during the 1990s financial crises. Engle and Granger's simple bivariate and Johansen's multivariate cointegrations are applied to the FX rates for the 1994 Mexican, 1997 Asian, 1998 Russian, and 1999 Brazilian crises. In addition, the article conducts the Granger causality test and impulse response analysis to examine the causality pattern in all the FX rates. The analysis shows most of the pre-Mexican causality disappears and significant numbers of new causality emerge in the 1994 Mexican crisis while the 1997 Asian crisis generates significant spillover effects into the later part of the 1998 Russian and 1999 Brazilian crises.

Contagion has been argued as a channel for currency crises spilling across countries in today's turbulence financial markets (see Sander & Kleimeier, 2003 and Reside & Gochoco-Bautista, 1999). Tests on the presence of contagion effects in currency crises include the cross-market correlation coefficient method (see Baig & Goldfajn, 1999 and Park & Song, 2000), cointegration method (Reside & Gochoco-Bautista, 1999), Granger causality method Khalid & Kawai, 2003 and Sander & Kleimeier, 2003, and conditional probability of crisis method (see De Gregario & Valdes, 2001 and Eichengreen et al., 1996). The cointegration test has been used to analyze long-run relationships among nonstationary time series (see Baillie & Bollerslev, 1989, Caporale et al., 2000 and Nieh & Lee, 2001). Prior to the 1997 Asian crisis, cointegration tests were primarily conducted to assess the relative impact of the Japanese yen (JPY) and/or the U.S. dollar (USD) on Asian currencies. For example, Aggarwal and Mougoue (1996) in their analyses found that the JPY is cointegrated with the East Asian and ASEAN (Association of South East Asian Nations) currencies. The authors' findings demonstrate a significant preliminary discussion of yen blocs in Asia. Tse and Ng (1997) reached similar cointegration conclusions among the Japan, Malaysia, the Philippines, Singapore, and Thailand currencies. The cointegration method has been utilized to examine the effect of contagion post-1997 Asian crisis. For example, Reside and Gochoco-Bautista (1999) use an error correction model to examine the cointegration relationships among exchange rates in the Asian market to isolate country specific effects of contagion. The authors' results show stable long-run relationships between exchange rates in the region with contagion effects from other countries to Singapore, China, and Japan. This article extends the cointegration findings of Reside and Gochoco-Bautista to include exchange rates from the Asian and emerging economies accounting for the major currency crises of the 1990s. The Granger causality method has also been widely used to draw inferences on contagion effects of currency crises. Unlike the cross-market correlation coefficient method, cointegration, and conditional probability of crisis methods, which can only be used to identify the existence of contagion, the Granger causality method allows further exploration of changes in the existence and directions of causality among crisis-ridden countries. Previous literature shows that the Granger causality test has been used to analyze the equity, exchange rate, and bond market to establish the interrelationship of these markets among countries (see Khalid & Kawai, 2003). For example, Nagayasu (2001) use Granger causality test to examine the relationship between exchange rate and stock indices during the 1997 Asian crisis. The author found that developments in some sectoral stock indices, including those of banking and financial sectors, have caused upward pressure on exchange rates. The existence of the strong relationships between the exchange rate market and stock indices confirms the importance of financial market linkages as a transmission channel of the 1997 Asian crisis from Thailand to neighboring countries, such as Indonesia, Malaysia, and the Philippines. Sander and Kleimeier (2003) examine causality pattern changes in the 1997 Asian and 1998 Russian crises using the Granger causality method. According to the authors' findings, the 1997 Asian crisis first established new and changed causality patterns that were not present before the crisis on a regional basis. As for the 1998 Russian crisis, it did not have a direct impact on Asia but invoked new short-run causalities mainly between Asian and Latin American countries that were not present before the crisis. The authors' findings highlight the role of international financial markets interdependence and linkage in regional and global financial contagion. This article examines the cointegration level, changes in the existence and directions of causality of the foreign exchange (FX) rates in the Asian and emerging markets during the 1990s financial crises. Engle and Granger's simple bivariate and Johansen's multivariate cointegrations are applied to the FX rates for the 1994 Mexican, 1997 Asian, 1998 Russian, and 1999 Brazilian crises. In addition, the article conducts the Granger causality test and impulse response analysis to examine the causality pattern in all the FX rates. The analysis shows most of the pre-Mexican causality disappears and significant numbers of new causality emerge in the 1994 Mexican crisis while the 1997 Asian crisis generates significant spillover effects into the later part of the 1998 Russian and 1999 Brazilian crises. In addition, this study provides a comprehensive analysis of the impact of each currency crises on the relationships between exchange rates of within and cross-regional currencies. This permits comparison of the cointegration and causality patterns of different currencies across the four different currency crises and provides a general indication of the level of crisis spillover and financial integration over the past decade. The balance of the article is organized as follows. Section 2 discusses the cointegration and Granger causality methodology. Section 3 describes the data used in this article. Section 4 discusses the cointegration and Granger causality test results within the Asian exchange rates, within emerging economies, and between Asian and emerging economies exchange rates, followed by the discussion of the impulse response analysis results. Finally, Section 5 concludes.

Overall, our results show that most of the causality relationships during the pre-Mexican crisis period disappear but significant new causality relationships are generated where formerly unrelated currencies are found to be Granger causing each other in the 1994 Mexican crisis. The causality pattern is even more unpredictable during the 1997 Asian crisis where there exists Granger causality between countries geographically far away and with no significant trade links with Asia. This is especially true when the 1997 Asian crisis fuels the crisis in Russia and Brazil due to risk aversion and herd behavior of investors. The later episodes of the 1998 Russian and especially the 1999 Brazilian crises have generated significant causality relationship to and from Asian currencies, but the number of causality relationships generated is less than those during the 1997 Asian crisis. Many of these causality relationships disappear during the post-crisis period, implying the existence of contagion effect of the currency crises examined. Currency crisis within a region also affects the linkages of exchange rates in a different region. The 1994 Mexican, 1998 Russian, and 1999 Brazilian crises have had some effects on the Asian currencies with new causality relationship comparing to the previous period. The causality relationships are mainly found from South East Asian to East Asian currencies except for the 1999 Brazilian crisis, where causality relationships are mainly within the East Asian currencies and from East Asian to South East Asian currencies. The 1997 Asian crisis generated new short-run causality relationship between the BRR and the MXP within the emerging economies currencies, which did not exist in the previous period. Further examination of the effect of the first crisis hit currency to other currencies during each crisis using impulse response analysis found that the 1997 Asian crisis and the 1998 Russian crisis represent major shocks to a number of exchange rates in East Asia, South East Asia, and the emerging economies in the 1990s. The findings of our study have important implications. First, the evidence of the existence of cointegration and causality patterns in currency crisis undermines the benefits of international risk diversification. Since trade or real linkages cannot explain adequately the causality relationships generated by currency crisis, the act of investing in geographically diverse countries or countries with little trade links does not guarantee risk diversification. Second, policy makers need to introduce measures to improve their fundamental economic structures and pursue policies that strengthen financial market accountability and transparency. This is to soothe the market sentiments to avoid future currency crisis in a financially integrated world. Third, the existence of significant causality relationships during currency crises provides rationale for greater coordination of multilateral and regional action to avoid the spread of shocks or contagion. For example, McKinnon (1998) suggests that coordination of the exchange rate policy is necessary among Thailand, Indonesia, Malaysia, the Philippines, and South Korea so that exchange rate changes in these neighboring countries or large interest rate misalignments will not generate flows of short-term hot money within these countries. As currency crises often spillover to neighboring countries within the region, Rose (1998) suggests a case for the Asian Monetary Fund to provide regional financial support to the crisis-ridden countries to limit the spread of the currency crises. In our study, some of the exchange rate series under investigation are strictly pegged to the USD for certain subsample periods and therefore the Granger causality model used is ineffective in measuring the true causality relationships among these exchange rate series. These exchange rate series includes the CHR and the MYR (in the post-crisis period from 3 January 2000 to 31 December 2001). Despite this limitation, this study does provide new and interesting insight into the interrelationship of the exchange rate across the four major crises in the 1990s as prior research using FX data mostly focuses on the 1997 Asian crisis but fall short of investigating the 1994 Mexican crisis, 1998 Russian crisis, and 1999 Brazilian crisis. Since the effect of currency crisis on Asian and emerging economies currencies have now been extensively researched, future research can be extended to include developed countries currencies, such as European and Canadian currencies, to examine if they are also affected by the currency crises occurring in the Asian and emerging economies. Another interesting research area would be to investigate the sources of the causality relationships found in this study to determine whether they are due to real link, financial links, or market perceptions. Understanding the source and transmission mechanism of crisis are important as it will assist policy makers in developing domestic and international policies to limit the contagion effect of the currency crisis. Finally, future research can test for cointegration for a longer time span, taking structural break into account.