Characterizing Exchange Rate Regimes in Post-Crisis East Asia
Author: Mr. Taimur Baig

Contributor Notes

Author’s E-Mail Address: tbaig@imf.org
Publication Date:
01 Oct 2001
eISBN:
9781451857092
Language:
English
Keywords:
WP; U.S. dollar; exchange rate; regime
Download PDF (1 MB)

This paper examines the behavior of the exchange rates of selected emerging market East Asian economies in the aftermath of the Asian crisis. The results suggest that movements in the Asia-5 currencies (Indonesia, Korea, Malaysia, Philippines, and Thailand) were significantly influenced by the U.S. dollar's day-to-day movements before the crisis, and have indeed continued to do so post-crisis. However, comparisons with a range of other currencies suggest that this is a fairly common trait across various regimes. Moreover, results from the post-crisis data do not support the view that the Asia-5 currencies presently have the same characteristics as they did before the crisis.

Abstract

This paper examines the behavior of the exchange rates of selected emerging market East Asian economies in the aftermath of the Asian crisis. The results suggest that movements in the Asia-5 currencies (Indonesia, Korea, Malaysia, Philippines, and Thailand) were significantly influenced by the U.S. dollar's day-to-day movements before the crisis, and have indeed continued to do so post-crisis. However, comparisons with a range of other currencies suggest that this is a fairly common trait across various regimes. Moreover, results from the post-crisis data do not support the view that the Asia-5 currencies presently have the same characteristics as they did before the crisis.

I. Introduction

More than four years after the onset of the Asian crisis, the characteristics of the exchange rate regimes of the Asia-5 (Indonesia, Malaysia, Korea, the Philippines, and Thailand) countries—before and after the crisis—remain a topic of considerable discussion. Recent recommendations have pointed toward a need for free floating rates in emerging market economies in general and in East Asia in particular. Mussa et. al. (2000), for example, argue that given the current international financial conditions, tight management of exchange rates that lead to limited exchange rate volatility in normal times can foster complacency with regards to exchange rate risk. They conclude—

“Thus, for emerging market countries that cannot or choose not to undertake the very strict regimen necessary to sustain pegged exchange rate regimes in an environment of international capital mobility, it is essential that floating exchange rates really do float.”2

After going through steep devaluations and high volatility in 1997-98, the currencies of the region have mostly stabilized over the past couple of years. Some observers, however, have interpreted this stability as evidence that the East Asian currencies are reverting back to de facto pegs against the U.S. dollar. In the context of the Asia-5, McKinnon (2000) argues that the so-called floating exchange regimes of the countries—barring Malaysia, which maintains a peg—are not really floating. Using a regression framework from Frankel and Wei (1994) on exchange rate data from January 1999 to May 2000, McKinnon argues that the evidence points toward a case of high-frequency pegging in Indonesia, Korea, the Philippines, and Thailand. He contends—

“In the year 2000, both the crisis and non-crisis countries of East Asia (with Japan remaining the important exception) have returned to formal or informal dollar pegging, which is statistically indistinguishable from what they were doing before the crisis.”

The prospect of a return of the dollar peg among the Asia-5 countries leads to concerns of fragility buildup through reduced incentive for exchange rate risk management, as well as real exchange rate misalignment. However, a critical and close scrutiny of the data is warranted before the characteristics of the exchange rate regimes in the concerned countries are branded as similar to that of pegged exchanged rates, especially the type that existed before the crisis. Hernández and Montiel (2001) examine the post-crisis exchange rate behavior of the Asia-5 currencies, and their results suggest that the currency regimes of Korea and Thailand have moved in the direction of greater flexibility, but not to the extreme pole of clean floating. They assign lesser weight to their findings with respect to Indonesia and the Philippines, but nevertheless suggest that those two currencies have also become more flexible when compared to the pre-crisis period. With regard to policy implications, they conclude that countries that are not prepared to accept the constraints of a hard peg, a managed float designed to accumulate reserves and resist real appreciation could be preferable in some conditions over the polar extreme of free floating.

This paper is concerned with the questions regarding the methodology of characterizing exchange rate regimes in post-crisis East Asia. What are the appropriate benchmarks to compare the characteristics of exchange rates? What are the empirical pitfalls of dealing with the relevant data, especially given that exchange rate intervention can be unobservable and there can be considerable uncertainty about the authorities’ reaction function? Are there any shortcomings in the standard statistical tests that are traditionally used to examine if an exchange rate is pegged? How can distinctions be made between pegging and “smoothing,” and is such distinction meaningful?3 And finally, is the post-crisis behavior of the exchange rates of the four countries concerned clearly indistinguishable from the pre-crisis behavior? Has there really been a return of the dollar standard in these countries?

The rest of the paper is organized as follows—Sections II and III describe the data and methodology. Section IV compares exchange rates, reserves, and interest rate volatility across the pre-crisis, crisis, and post-crisis periods of the selected East Asian exchange rates against a control group, representing various currency regimes from the industrial and emerging economies. Section V describes the construction of an indicator of exchange rate flexibility, which is examined across the three periods for the selected exchange rates. In Section VI, regressions are carried out across various data frequencies to test for the evidence of a dollar peg. Finally, in Section VII, properties of exchange rate data from the countries concerned are examined for evidence of pegging and/or smoothing. Section VIII contains some concluding remarks.

The results suggest that the Asia-5 currencies assigned statistically significant and large weights to the dollar on their day-to-day movements before the crisis, and have indeed continued to do so post-crisis. However, comparisons with a range of other currencies show that this is a fairly common trait across various regimes. Moreover, results from the post-crisis data do not support the view that the Asia-5 currencies have fully reverted to behavior that is statistically indistinguishable from pre-crisis characteristics.

II. Data

The data requirements for this paper are simple. Four variables—exchange rate, interest rate (overnight money market rates), international reserves, and reserve/base money series—are used for the countries in the sample. The daily exchange rate data was extracted from Bloomberg, whereas the rest of the data (in monthly frequency) were obtained from the International Financial Statistics database. The sample includes the Asia-5 countries, ten selected countries with free floats, and nine other countries from emerging market economies with varying regime history.4

III. Methodology

In order to discern regime-specific behavior, the exchange rate characteristics of the Asia-5 countries are compared against the two control groups described above. Changes in the observations over time within the country in question are also analyzed to track regime switches. A similar exercise is carried out for interest rates and reserves. The exchange rate flexibility indicator, described in Section V, combines the information extracted fromexchange rates, reserves, and base money, and is estimated for each year within the 1995–2000 sample. The results complement the ones obtained in Section IV, and allow for a better comparison across and within regimes.

The question of to what extent the Asia-5 countries have tracked the dollar through the sample period is explored in section VI through a country-by-country regression analysis and two hypothesis tests of coefficient stability. The approach involves using an independent currency as an arbitrary numéraire for measuring exchange rate variation against the U.S. dollar, Japanese yen, and German mark. The robustness of the results is tested by using two numéraires—the Swiss franc, which has been used in several recent studies, and the British pound. The coefficient tests are for the two hypotheses—if the dollar coefficient is statistically indistinguishable from one (implying a peg type behavior), and if the dollar coefficient in 1999 and 2000 for each currency is equal to the respective coefficient estimate in 1996.5

Section VII examines the exchange rate behavior of Indonesia, Korea, the Philippines, and Thailand by testing the currencies’ respective residuals from a random walk regression. The residuals are tested for normality (which would tend to imply an exchange rate without any intervention), and their various properties are examined for the evidence of pegging or smoothing.

IV. Volatility Comparison

A. Exchange Rate Volatility

We begin by looking at the volatility of exchange rates, defined as the standard deviation of the percentage changes of the exchange rates against the U.S dollar. Using daily and monthly data respectively, Tables 1A and 1B illustrate annual estimates of exchange rate volatility between 1995 and 2000. For ease of exposition, the tables group the Asia-5, a selection of floating exchange rates (mostly from developed markets), and a selection other rates from emerging markets separately.

Table 1A.

Exchange Rate Volatility: 1995-2000

Standard deviation of daily movements (percentage changes) against the U.S. dollar

article image

Countries that have been classified by the IMF as having independently floating exchange rate regimes throughout the sample period.

Table 1B.

Exchange Rate Volatility: 1995-2000

Standard deviation of monthly movements (percentage changes) against the U.S. dollar

article image

The extreme swings experienced by the Asia-5 countries during the 1997/98 crisis are clearly reflected in Tables 1A and 1B and Figure 1, with volatility jumping 10-20 times compared to the pre-crisis period of 1995/96. Of course, the very large increases in crisis period volatility estimates appear more dramatic due to the extremely low pre-crisis volatility among the concerned currencies, when they were all managed heavily to track the U.S. dollar—i.e. the exchange rates were pegged de facto. The spike in volatility in 1997/98 among the Asia-5, however, was not an isolated incident. The spillover from the Asian crisis (as well as the subsequent Russian crisis in 1998) is clearly evident among the estimates from the rest of the sample. Currencies with significant Asia-5 exposure, regardless of regime affiliation, also came under pressure, with volatility jumping in Australia, New Zealand, and Singapore.6 Contagion from the crisis also spread, and was reflected in the exchange rate volatility of Mexico, South Africa, and the Czech Republic. Nevertheless, the magnitude of volatility of the Asia-5 was far above than those in the rest of the sample.

Figure 1
Figure 1

Asia 5 Exchange Rates and Volatility; (1/1/1994-2/28/2001)

Right axis: exchange rate; national currency/U.S. dollar

Left Axis: log difference of daillly exchange rate

Citation: IMF Working Papers 2001, 152; 10.5089/9781451857092.001.A001

Figure 2A
Figure 2A

Exchange Rates (against the U.S. dollar)

Citation: IMF Working Papers 2001, 152; 10.5089/9781451857092.001.A001

Figure 2B
Figure 2B

Exchange Rate Volatility (log difference or daily data)

Citation: IMF Working Papers 2001, 152; 10.5089/9781451857092.001.A001

Figure 3A
Figure 3A

Exchange Rates (against the U.S. dollar)

Citation: IMF Working Papers 2001, 152; 10.5089/9781451857092.001.A001

Figure 3B
Figure 3B

Exchange Rate Volatility (log difference of daily data)

Citation: IMF Working Papers 2001, 152; 10.5089/9781451857092.001.A001

The volatility estimates for 1999 and 2000 highlight the key exchange related developments in Asia-5—Malaysia’s nominal fixing of the rate against the U.S. dollar, continued turbulence with the Indonesian rupiah (at a somewhat lower level than in 1997/98, but still over 10 times than seen in 1995/96), and a marked return to stability for the currencies of Korea, the Philippines, and Thailand (although they remained substantially more volatile than during the pre-crisis period). Among the latter three countries, exchange rates of Korea and Thailand showed a further decline in volatility from 1999 to 2000, whereas the exchange rate of the Philippines became more volatile in 2000.

The volatility of the floating exchange rates (in dollar terms), namely Japan and the countries in the sample outside of Asia (Germany, Sweden, Switzerland, and the U.K.), has been roughly unchanged or slightly decreasing since 1995, although a few countries with exposure to the financial crises in the late 1990s had sporadic episodes of increased volatility. While they were relatively stable, the magnitude of the floating currencies’ volatility estimates were in general greater than the post-crisis Asia-5 estimates (with the exception of Indonesia).

B. Interest Rate Volatility

Exchange rate volatility alone may not be sufficient to characterize the exchange rate regime, as this statistic does not account for the extent to which the authorities have targeted the rate through monetary policy and intervention in the foreign exchange market. Thus, two currencies with comparable standard deviations could conceivably represent two contrasting regimes—one could be a stable free-float, the other a dirty float kept in check through interest rates changes or foreign exchange market transactions. Tables 2 and 3 address this issue by looking at the volatility of interest rates (standard deviation of interest rate differences) and reserves (standard deviation of monthly growth rates) during 1995–2000.

Table 2

Interest Rate Volatility: 1995-2000 1/

article image

Standard deviation of differences in interest rates.

Table 3.

Reserves Volatility. 1995-2000 1/

article image

Standard deviation of percentage change in reserves.

The interest rates of the Asia-5 countries, barring Malaysia, are seen to be substantially more volatile than the group of floating countries in the pre-crisis period. This is consistent with the experience of regimes with exchange rate as a nominal anchor, as capital flow related volatility is reflected somewhere else in the economy. Among other countries in the sample, as expected, interest rate volatility in the countries with managed floats or crawling pegs was comparable or more than the Asia-5 countries, which in turn was substantially more than the countries with floating rates.

Coinciding with the severe exchange rate pressure episodes in 1997/98, interest rate volatility is seen to have increased dramatically among the Asia-5, doubling in the case of Korea to over ten times in Indonesia. Among the floaters, New Zealand and South Africa, the countries that were impacted by the crises, show increases in interest volatility as well.7 The same is seen in Brazil, Chile, Czech Republic, and Singapore.

The post-crisis estimates reveal a marked decrease in interest rate volatility in the Asia-5 economies. By 2000, for all of the Asia-5 countries, the standard deviations of interest rates were not only lower than the crises period, they were even lower than in 1995. Among the floating rate countries, virtually no change in interest rate behavior is noticeable, whereas in the other countries a clear pattern to reduced volatility is seen. Overall, across the time period of 1995-2000, with the crises years as exceptions, a broad trend in declining interest rate volatility is seen for the entire sample.8

C. Reserves Volatility

With respect to reserves, the volatility for the Asia-5 jumped as expected during the crisis years as a result of the exchange rate defense and capital outflows (see Table 3). However, the crisis period volatility estimates are comparable or lower than the floating rate sample average. In the post-crisis period of 1999/2000, reserves volatility followed a declining pattern in Korea, Malaysia, and Thailand, whereas both Indonesia and the Philippines saw a decrease in volatility in 1999, only to have it reversed in 2000.9 Among the Asia-5 countries, Korea’s reserves volatility decreased the most.

Among the “other,” nonfloating group in the sample, the volatility estimates illustrate the turbulence in Brazil and Turkey, as well as increasing stability seen in the economies of Chile, Czech Republic, Hungary, Israel, Poland, and Singapore in 1999/2000. These observations echo the interest rate volatility estimates.

The reserves figures presented are subject to two caveats. First, fluctuations in reserves can reflect valuation adjustments, debt repayments, and other factors that do not necessarily represent foreign exchange market intervention. Second, forward market intervention, which is common in some of the countries in the sample, is not fully captured by the gross reserves figures. Spot market interventions show up in the central bank’s balance sheet immediately, whereas forward market interventions remain off-balance sheet, unless fully unwound at a future date.10

Notwithstanding the caveats, the results from this section suggest that while there has been a broad return to stability since the crises, not all Asia-5 countries have followed similar paths. Overall, Korea and Thailand appear to have normalized the most, with sharp reductions in interest rate and reserves volatility, although exchange rate volatility remains higher than in the pre-crisis period. The Philippines have also seen substantial relative change in the volatility of the three variables in question, although there are indications of some increased turbulence by end-2000. Indonesia appears to be lagging in its path to stability, while Malaysia, by virtue of capital controls and a fixing of its exchange rate, appear to have tempered the market volatilities, at least by the benchmarks used in this section.

V. Exchange Rate Flexibility

Exchange rate or reserves movements, in isolation, offer a partial picture of an exchange rate regime. However, they can be combined to produce a more informative indicator of exchange rate flexibility to be used to analyze further exchange rate regime behavior. In this section, we follow the methodology used in Glick and Wihlborg (1997) and Bayoumi and Eichengreen (1998) to create an index of exchange market flexibility.

The flexibility index is constructed by dividing the standard deviation of exchange rate movements by a measure of exchange market pressure, which in turn is a function of reserves volatility, scaled by base money. The precise formula for the index is—

Index=SDEX(SDEX+SDREV)

where,

  • SDEX: standard deviation of exchange rate changes (log difference),

  • SDREV: standard deviation of the ratio of changes in reserves, divided by lagged stock of base money.

By construction, the index ranges from 0.00 to 1.00, with the lower values indicating relative inflexibility of the exchange rate.

The index is calculated for each country for the years 1995–2000. For ease of exposition, Table 4A summarizes the Asia-5 results, and Table 4B breaks down the rest of the sample by the latest IMF classification of exchange rate regimes.

Table 4A.

Exchange Rate Flexibility: 1995-2000 1/ Asia 5

article image

Calculated as SDEX/(SDEX+SDREV), where SDEX is the standard deviation of log differences of exchange rate against the U.S. dollar, and SDREV is the standard deviation of the changes in the central bank’s reserves divided by lagged stock of base money.

Based on International Financial Statistics, International Monetary Fund, June 2001.

Moved from managed float to independent float: August 1997.

Moved from managed float to independent float: December 1997.

Moved from managed float to fixed peg: September 1998.

Moved from Fixed peg to independent float: July 1997.

Table 4B.

Exchange Rate Flexibility: 1995-2000 1/

article image

Calculated as SDEX/(SDEX+SDREV), where SDEX is the standard deviation of log differences of exchange rate against the U.S. dollar, and SDREV is the standard deviation of the changes in the central bank’s reserves divided by lagged stock of base money.

Based on International Financial Statistics, International Monetary Fund, June 2001.

Moved from managed float to independent float: January 1999.

Moved from crawling band to independent float: September 1999.

Reclassified from managed float to crawling band: January 1998.

Reclassified from independent float to managed float: December 2000.

Reclassified from managed float to crawling band: August 1998.

Moved from crawling band to independent float: April 2000.

Moved from managed float to crawling band: June 1998.

The index tracks the pre-crisis lack of exchange rate flexibility of the Asia-5 rather well, and the 1997/98 spike in the index illustrate the crisis-related developments. The post-crisis figures are broadly consistent with the findings in the pervious section—for Korea, the Philippines, and Thailand, the exchange rates have become somewhat less flexible from 1997/98 levels, but they remain consistently more flexible than during the pre-crisis years. Malaysia’s regime has become completely inflexible, and Indonesia’s exchange rate volatility is overwhelmed by its jump in reserves volatility in 2000, thus leading to a lower index value, indicating lower flexibility.

The results summarized in Table 4B reveal salient features of the other regimes. The floating rate regimes come across with relatively high flexibility in all periods, whereas the figures from the other regimes indicate a broad trend toward increased flexibility in recent years. While in general, during the post-crisis period, the Asia-5 currencies appear somewhat less flexible than the sample of other floating currencies, it is noteworthy that one cannot readily discern between various regimes using the above index alone. As evident from the index figures for the floating and other regimes, data and regime-specific idiosyncrasies can lead to difficulties in making cross-country comparisons. The index is susceptible to the same caveats raised about reserves figures in Section IV.C. This would partly explain why Australia and Canada score relatively low in the index, while India scores exceptionally high in 1995/96.

The index can however be additionally useful in within country analysis through examining the changes in the index over years. In this regard, the developments of the Asia-5 currencies can be followed readily, as described earlier. Moreover, the various regime switches that take place among some of the exchange rates in the sample are also picked by the index (Table 4B).

VI. Peg to the Dollar?

The analysis of the previous two sections suggest that there has been a post-crisis decrease in exchange rate volatility, as well as flexibility, among the Asia-5. However, the analysis also highlight that the greater relative stability of the regional currencies does not necessarily imply a reversion to pre-crisis behavior. We continue our examination of post-crisis exchange rate characteristics in this section by addressing the question of whether the currencies have reverted back to their pre-crisis behavior of re-linking their currencies to the U.S. dollar, as claimed in some recent work.11

A test for high frequency pegging was developed by Frankel and Wei (1994), and it has been used subsequently by Ogawa (2001) and McKinnon (2000) in the context of post-crisis exchange rate behavior seen among the Asia-5. In this approach, an independent currency is chosen as an arbitrary numéraire for measuring the exchange rate variation. The goal here is to estimate the weight a currency assigns to another currency for a given frequency. The regression model, where the local currency’s value against the independent currency is regressed against the major world currencies, is—

dlog(LCSF)=β1+β2dlog(USDSF)+β3dlog(JPYSF)+β4dlog(DEMSF)+ε

where,

  • LC: Local currency,

  • SF: Swiss franc,

  • USD: U.S. dollar,

  • JPY: Japanese yen,

  • DEM: German mark.

We begin by estimating annual regressions (using daily data for each year spanning 1995–2000) for the group of countries in the sample.

The regression framework helps in addressing two issues. First, the extent to which the coefficient of the U.S. dollar deviates from unity provides an indication of the flexibility of the currency against the dollar. Second, the pre and post-crisis results can be compared to test the hypothesis of a reversion to pre-crisis behavior.

To facilitate the exploration of the first issue, we apply the Wald coefficient test, for each regression, to test the null hypothesis that the dollar coefficient is equal to one. The second issue is probed by a test for coefficient equivalence, between the dollar estimates of the regressions for 1999 and 2000, against the estimates from the pre-crisis year of 1996.

In order to test the robustness of the results, the entire set of tests are then repeated by using a different numéraire—the British pound.12 Tables 5A, 5B, 6A, and 6B summarize the results (the first two feature results using the franc as numéraire, and latter two feature the pound).

Table 5A

Exchange Rate Regressions: Comovement with the U.S. Dollar 1/ Numéraire: Swiss Franc

article image
article image
article image

Regression Model: dlog(local currency/SF) = b1 + b2 dlog(USD/SF) + b3 dlog(JPY/SF) + b4 dlog(DEM/SF’’); where SF - Swiss Franc, USD - US Dollar, JPY - Japanese Yen, DEM - German Mark.

Null hypothesis for Wald Coefficient test: coefficient estimate on the US dollar is.equal to one.

denote significance at 5 percent and 10 percent levels, respectively.

Table 5B.

Hypothesis Test of Coefficient Equivalence 1/

For the dollar coefficient of regressions; against 1996 estimates; Null: coefficients are equal

article image

Test statistic = [estimate (t1)-estimate(t0)]/standard error of estimate (t1).

90 percent confidence level.

Table 6A

Exchange Rate Regressions: Comovement with the U.S. Dollar 1/ Numéraire: British Pound

article image
article image
article image

Regression Model: dIog(local currency/BP) = b1 + b2 dIog(USD/BP) + b3 dlog(JPY/BP) + M dlogfDEM/BP); where BP - British Pound, USD - US Dollar, JPY - Japanese Yen. DEM - German Mark.

Null hypothesis for Wald Coefficient test: coefficient estimate on the US dollar is equal to one.

denote significance at 5 percent and 10 percent levels, respectively.