Exchange Rate Regime

3.1 Methodology

In the last decade, the literature has revealed that in many economies, the de jure exchange rate regime announced by the central bank differs substantially from the de facto regime in operation.This has motivated a small literature on data-driven methods for classifying exchange rate regimes (Reinhart and Rogoff 2004; Levy-Yeyati and Sturzenegger 2003; Calvo and Reinhart 2002). This literature has attempted to create datasets identifying de facto exchange rate regimes across all countries in recent decades, using a variety of alternative algorithms. While these databases are useful for many applications, they have limited usefulness in measuring the finer characteristics and structures of intermediate regimes. For instance, Reinhart and Rogoff classify the Indian rupee as a single exchange rate regime from 1993 onwards, but as this paper will subsequently show, India has had an intermediate regime since 1993, which yields fresh insights into the drivers and consequences of the exchange rate regime and monetary policy framework.

A valuable tool for understanding the de facto exchange rate regime is a linear regression model based on cross-currency exchange rates (with respect to a suitable numeraire). Used at least since Haldane and Hall (1991), this model was popularized by Frankel and Wei (and is hence also called the Frankel-Wei model). Recent applications of this estimation strategy include Bénassy-Quéré, Coeuré, and Mignon (2006), Shah, Zeileis, and Patnaik (2005) and Frankel and Wei (2007). In this approach, an independent currency such as the Swiss Franc (CHF) is chosen as an arbitrary ‘numeraire'. If estimation using the Indian rupee (INR) is desired, the model estimated is:

This regression picks up the extent to which the INR/CHF rate fluctuates in response to fluctuations in the US$/CHF rate. If there is pegging to the US Dollar (US$), then fluctuations in the Japanese Yen (JPY) and the German Deutsche Mark (DEM) will be zero. If there is no pegging, then all the three betas will be different from 0. The RM² of this regression is also of interest; values near 1 would suggest reduced exchange rate flexibility.

To understand the de facto exchange rate regime in a given country for a given time period, researchers and practitioners can easily fit this regression model to a given data window, or use rolling data windows. However, such a strategy lacks a formal inferential framework for determining changes in the regimes. This has motivated an extension of the econometrics of structural change, for the purpose of analyzing structural change in the Frankel-Wei model (Zeileis, Shah and Patnaik 2008). This involves extending the familiar Perron-Bai methodology (Bai and Perron 2003) for identifying the dates of structural change in an OLS regression. Through this, dates of structural change in the exchange rate regime are identified. We focused on the period after 1976, and utilized weekly changes in exchange rates for these estimations. Values shown in brackets are t-statistics.

For each country, a set of sub-periods were identified. In each sub-period, the R² of the regression served as a summary statistic of exchange rate flexibility. Values near 1 convey tight pegs. Floating rates have values of between 0.4–0.5.

Using this classification scheme, we were able to do the following:

• Measure and quantify the fine structure of intermediate regimes using a real-value measure of exchange rate inflexibility (the regression R²), which naturally suggests a real-value measure of exchange flexibility.
• Specify dates at which the exchange rate regime changed. We implemented these methods using weekly percentage changes in exchange rates, which yielded break dates to the resolution of the week. Through this, a time-series of exchange rate flexibility was obtained for each country, of the value of the R², which prevailed at a point in time.
• Determine the number of breaks and the placement of breaks based on sound inference procedures.

3.2 Evidence on Exchange Rate Flexibility of Asia-11

We applied this methodology to examine the de facto exchange rate regimes of the Asia-11 economies. For each country, a time-series of currency flexibility was obtained, providing summary statistics on exchange rate flexibility.

In India, the rupee began its life as a ‘market-determined exchange rate' in March 1993. However, this date is not identified as a structural break in the data analysis. Instead, a sub-period for the exchange rate regime is found, from 1976– 1998. During this period, the rupee was de facto pegged to the dollar with a certain degree of exchange rate flexibility, with an R² of 0.84.

After the Asian Financial Crisis, India embarked on a tight rupee-dollar peg. From 28 September 1998 to 19 March 2004, the US$ coefficient reverted to 1.01. The other coefficients were not statistically significant. The R² rose to 0.97. During this period, the exchange rate regime in India was similar to PRC's after July 2005.

Table 5: India's de facto Exchange Rate Regime [ PDF 95.3KB | 1 page ]

Table 6 [ PDF 25.6KB | 1 page ] shows the results of this estimation strategy for the Chinese Renminbi. It finds that the first period runs from 9 Jan 1981 until 1 November 1985. This was a period with bigger currency flexibility by Chinese standards, with the R² at 0.89. Subsequently, PRC moved to a tight US dollar peg. While there have been some minor changes in the exchange rate regime, it remains primarily a simple peg, with a US$ coefficient of 1 and an R-squared ˜ 1.

In some respects, these results are consistent with official announcements and a simple examination of the exchange rate. The break date of 22 July 2005 that is derived from the regression is consistent with that announced by the authorities. The results for PRC therefore suggest that the econometric analysis is broadly on the right track.

At the same time, it is noteworthy that after 22 July 2005, no further structural changes were evident from the econometric analysis. This contradicts a variety of official claims regarding the evolution of the exchange rate away from the US dollar peg towards a basket peg, and towards greater exchange rate flexibility.

The regression results suggest that remarkably little has changed in the actual prevailing exchange rate regime. The US$ coefficient has dropped to 0.949. A statistically significant Euro coefficient has emerged, with a small value of 0.06 where the null hypothesis of zero can be rejected. The residual standard deviation has more than doubled to 0.243, but the R² has dropped only slightly to 0.974. While there was more exchange rate flexibility in this period, the change in the exchange rate regime was extremely small.

Finally, Table 7 [ PDF 25.6KB | 1 page ] shows the evolution of the exchange rate regime in Korea. From 1981 until early 1995, the country ran a de facto peg to the US dollar. In 1995, a big increase in currency flexibility came about and the R² dropped to 0.65. This is a regime with greater flexibility than India's.

Figure 7 [ PDF 18.6KB | 1 page ] shows the average and the median value of the R² for the Asia-11 economies. At each point in time and for each country, the de facto exchange rate regime was identified, and the R² value from that sub-period was utilized.

The average R2started out with a high value of 0.9. There was a small increase in flexibility in 1980 and 1981. Subsequently, however, there was a sustained period of exchange rate rigidity. From 1982 until 1997, the average R² was above 0.9. This exchange rate inflexibility, coupled with increasing de facto capital account openness, helped trigger the Asian Financial Crisis, which involved firms and banks borrowing in foreign currency based on expectations of exchange rate rigidity.

During the crisis, exchange rate flexibility increased. In 1998, the average R2dropped to 0.61. However, immediately after that, exchange rate rigidity went up. This empirical fact was brought to prominence by Calvo and Reinhart (2002), who emphasised that after the crisis, little had changed with exchange rate regimes in Asia. This perspective was further amplified by the ‘Bretton Woods II' hypothesis, which tried to rationalize this exchange rate rigidity (Dooley, Folkerts-Landau, and Garber 2003).

Our evidence offers a somewhat different perspective in two respects. First, while exchange rate inflexibility in Asia-11 rose after the crisis subsided, it reverted to lower values when compared to what prevailed before the crisis. The mean R² was 0.93 in 1997; post-crisis, this changed to 0.88 over the 2002–2004 period.

The second interesting observation is that since 2002, exchange rate flexibility in Asia-11 has been slowly rising. The mean R² dropped slightly from 0.886 in 2002–2004 to 0.85 in 2009. This suggests that while Asia-11 economies continue to have considerable exchange rate inflexibility, there has been a gradual movement towards greater flexibility. With a mean R² of 0.85 in 2009, the environment has improved when compared with the mean of 0.93 in 1997.

Download this Paper [ PDF 359.1KB| 29 pages ].

Comment(s)

There are [0] comment(s) for this entry. Post a comment.