We study the behavior of real exchange rates under various official designations of exchange rate arrangements. Examining many currencies, we find important differences across the designations. Most notably, real exchange rate mean reversion is fastest when nominal exchange rates are officially pegged. We also find a large nonlinear effect: adjustment is fastest when the real exchange rate deviates greatly from its mean. This nonlinear effect is also most striking among officially pegged currencies. Finally, we find that nominal exchange rates, rather than prices, do most of the adjusting.
IN MANY PARTS OF THE WORLD, exchange rate arrangements are as hotly debated as they ever have been. Underlying the controversy is the sense that exchange rate arrangements matter, whether for prices, the terms of trade, or economic activity. Yet, how they matter remains an open question. In this study, we explore one important aspect of that question--the empirical link between official exchange rate arrangements and the behavior of real exchange rates themselves.
Our work is in the tradition of Mussa (1986), who stressed the differences in the behavior of exchange rates under alternative arrangements. Mussa argued that the exchange rate float had brought extremely high nominal and real exchange rate variability to many countries. Since then, Grilli and Kaminsky (1991) and Engel and Kim (1999) have reexamined the role of exchange rate arrangements using a century of British and U.S. data. Grilli and Kaminsky found that the float's high volatility seemed to be an artifact of the particular historical period that Mussa studied, rather than a result of floating exchange rates. However, differences across arrangements seemed to reemerge when Engel and Kim separated the permanent and transitory components of the real exchange rate. The transitory component (the most volatile component) was indeed much more volatile when the pound floated than when it was pegged.
The present paper builds on these studies by systematically examining the link between exchange rate arrangements and the real exchange rate using a large panel of countries. Our work also builds on the now numerous panel studies of real exchange rates, including Frankel and Rose (1996) and Taylor (1996), who carefully divide their panels into various subperiods to determine the potential importance that exchange rate arrangements may have on their main findings. However, even within their fairly narrow subperiods, aggregating across countries has meant aggregating across differing kinds of exchange rate arrangements. Here, we classify the arrangements of each individual country in each year. We then study the real exchange rate under the alternative designations of the exchange rate arrangements.
We begin with some summary statistics. We find that the magnitude of the real exchange rate, defined in terms of absolute price levels, is roughly comparable across the designated arrangements. However, we note that its variability differs markedly, though not as greatly as has been commonly supposed. Next, we test for the presence of a unit root in the real exchange rate, and we compare estimates of the speed of its mean reversion under the alternative designations. We find faster reversion in those countries with a dollar peg than in those without one. We also find that the speed of reversion depends positively on the size of the deviation from the mean. The extent of this nonlinearity also differs across the various arrangements, with the adjustment under a peg being the most nonlinear. Finally, we use an error correction framework to examine separately the adjustment of the nominal exchange rate and of relative prices. While we find that both exchange rates and prices adjust, we find that exchange rates carry out the lion's share of the adjustment. This result holds across all of the exchange rate arrangements that we examine, including-most strikingly-those classified as maintaining a peg.
1. REAL EXCHANGE RATES
We examine the log of the real exchange rate, [q.sub.it] = [p.sub.it] - [s.sub.it], where [s.sub.it] is the log of the exchange rate in U.S. dollars per domestic currency unit, [p.sub.it] is the log of relative prices, and subscripts index the country and the year. We obtain these data from the Penn World Tables of Summers and Heston (1991). For our purposes, the primary strength of the Penn World Tables is that they provide price indices using price levels, rather than changes in prices.(1)
To examine the role of exchange rate arrangements, we divide the sample using the exchange rate arrangements reported in the International Monetary Fund's Annual Report on Exchange Arrangements and Exchange Restrictions. Our use of reported arrangements implies that the designations sometimes may diverge from de facto arrangements. Such divergences reflect the somewhat ambiguous nature of exchange rate arrangements, for example, "floating" exchange rates may be managed, and "fixed" exchange rates may be altered. In their study of the effect of exchange rate arrangements on macroeconomic performance, Ghosh et al. (1997) use both reported arrangements and an indicator constructed from the actual behavior of the nominal exchange rate. Unfortunately, as Ghosh et al. point out, their indicator approach is vexed by the difficulty of distinguishing between arrangements and the other determinants of behavior. Yet another conceivable approach to identifying arrangements would be to construct indicators from observable institutional characteristics. Unfortunately, this approach would require identifying the signature characteristics of each type of arrangement. These limitations persuade us to rely on the reported arrangements. We are, however, cautious in interpreting our results too strongly. Our objective here is simply to describe real exchange rates under the official designations.
We divide the sample annually in six distinct ways. First, we separate those countries that reportedly peg their currencies to the U.S. dollar from all other countries. Second, we separate those countries with exchange rates that are "independently floating" from all others. Third, we combine the "floating" group with "managed floaters," those managing exchange rates within narrow bounds; and, we separate the combined group from all others. Fourth, we combine the floaters and managed floaters with those maintaining nondollar pegs; again, we separate the combined group from the remainder. Fifth, we divide the sample using the Bretton Woods breakdown, which we place between 1972 and 1973. Finally, for benchmarking, we separate the G10 and non-G10 countries. We examine a balanced panel of 82 countries from 1961 to 1992. The appendix describes the subsamples in more detail and lists the included countries.
Table 1 provides some basic statistics describing the real exchange rates, beginning with the sample as a whole. The first column gives the mean absolute value. For the sample as a whole, it is quite large, about 0.58. This figure represents a substantial deviation from purchasing power parity. For example, it is about four times that reported by Parsley and Wei (1996) for goods and services sold in various cities within a single country, the United States. Presumably, the much larger real exchange rate measured here reflects, in part, the greater difficulty of goods arbitrage across currencies and international borders.(2)
TABLE 1 SUMMARY STATISTICS: REAL EXCHANGE RATES Standard Percentiles of |[[Bar] Deviation of [q.sub.it] q.sub.it]| [q.sub.it] 5% 95% Full Sample 0.581 0.482 -1.316 0.251 Dollar Peg 0.545 0.354 -1.066 0.031 Other 0.597 0.527 -1.363 0.305 Non-Float 0.573 0.468 -1.200 0.252 Float 0.619 0.537 -1.463 0.204 Non-Flex(a) 0.599 0.449 -1.236 0.192 Flex(a) 0.525 0.559 -1.413 0.335 Non-Flex(b) 0.600 0.400 -1.200 0.119 Flex(b) 0.566 0.539 -1.351 0.320 Pre-1973 0.578 0.372 -1.161 0.048 Post-1972 0.583 0.519 -1.332 0.314 G10 0.188 0.225 -0.355 0.367 Non-G10 0.636 0.466 -1.342 0.178 Standard |[[Bar][Delta] Deviation of Number of q.sub.it]| [[Delta] Observations q.sub.it] Full Sample 0.080 0.126 2296 Dollar Peg 0.064 0.101 677 Other 0.087 0.134 1619 Non-Float 0.079 0.126 1864 Float 0.085 0.125 432 Non-Flex(a) 0.077 0.123 1729 Flex(a) 0.088 0.133 567 Non-Flex(b) 0.070 0.110 1057 Flex(b) 0.088 0.137 1239 Pre-1973 0.050 0.079 656 Post-1972 0.092 0.140 1640 G10 0.069 0.094 280 Non-G10 0.082 0.129 2016 NOTES: 1. [q.sub.it] = [p.sub.it] is the real exchange rate; |[Bar][q.sub.it]| is its mean absolute value; and |[Bar][Delta][q].sub.it]| is the mean absolute change in the deviation. 2. See appendix for an explanation of the exchange rate arrangement classification scheme. 3. The G10 countries include Belgium, Canada, France, …