11 Exchange Rate Policy and Macroeconomic Management in ASEAN Countries
- John Hicklin, David Robinson, and Anoop Singh
- Published Date:
- July 1997
Determining the lessons to be derived from the successes of the East Asian “miracle” economies has become a minor growth industry in the economics profession. Not only have these economies achieved extremely rapid and relatively equitable growth, but they have also by and large avoided major macroeconomic crises along the way, even when the world economic environment proved inhospitable. Among the many lessons that observers have derived from this experience is that outward orientation is a successful development strategy. While the term “outward orientation” is not well defined, and economists differ in particular with respect to how much encouragement to exports it implies, one feature of an outward-oriented policy package has clearly been the avoidance of prolonged real exchange rate overvaluation. This is an aspect of the East Asian macroeconomic experience that many observers had begun to emphasize by the mid-1980s. Maintaining real exchange rates close to their equilibrium values was credited with helping East Asian countries both to continue to compete successfully in world markets and to avoid the episodes of massive capital flight that aggravated debt problems and destabilized macroeconomic performance in Latin America and else-where during the late 1970s and early 1980s.
Countries in East Asia have received substantial capital inflows during the last few years, and conventional wisdom holds that the arrival of capital inflows should be associated with an appreciation of the equilibrium real exchange rate. Indeed, just such a link was used to make the case prior to December 1994 that the Mexican peso was not overvalued; that is, the observed real appreciation of the Mexican peso was interpreted in some quarters as an equilibrium phenomenon. In East Asia, however, very large capital inflows—comparable in magnitude to those Mexico received—have typically not been associated with a similar appreciation of the actual real exchange rate. What makes this observation important is that real exchange rate performance has been correlated with other aspects of macroeconomic performance among developing countries that have received substantial capital inflows. In cross-country data, larger real appreciation has been associated with a relatively larger increase in the share of consumption in GDP and with relatively slower growth. Many observers have also cited the avoidance of substantial real appreciation in explaining why post-Mexico “tequila effects” did not extend to East Asian countries (see, e.g. Sachs, Tornell, and Velasco, 1996).
There are two ways to interpret the East Asian real exchange rate response to the arrival of capital inflows. The first is to argue that the behavior of the real exchange rate in these countries is due to active management of the nominal exchange rate in the pursuit of a competitiveness objective—in other words, real appreciation has not emerged because the authorities adopted a nominal exchange rate policy geared to the pursuit of a real exchange rate target, rather than a price-level objective. This interpretation is plausible a priori, both because, as mentioned above, maintaining a competitive real exchange rate has been a key component of the outward-oriented development strategy favored by the East Asian countries, and because, unlike many of their Latin American counterparts, these countries had already achieved significant success in inflation stabilization before capital inflows began to arrive. Both factors would lead one to expect that policies influencing the nominal exchange rate would be likely to place relatively more weight on competitiveness than on price stability in the East Asian than in the Latin American context.
However, interpreting the East Asian experience in terms of nominal exchange rate policy leaves some loose ends. In particular, it fails to explain how a nominal policy instrument can be used consistently and over long periods of time to achieve a real economic target. During the recent capital-inflow episode, if conventional wisdom is correct, the arrival of capital inflows would, all other things being equal, have contributed to an appreciation of the equilibrium real exchange rate in the recipient countries. In that case, if East Asian countries have managed to avoid the emergence of an actual real appreciation through the use of nominal exchange rate policy, the currencies of those countries would now be undervalued, and this undervaluation would prove difficult to sustain. If these countries were to persist in maintaining an excessively depreciated real exchange rate, some other macroeconomic variable would have to adjust to perform the equilibrating function that would otherwise have been performed by movements in the real exchange rate. Recent research suggests that the rate of inflation may turn out to be the adjusting factor. In that case, an interpretation of recent experience that relies on persistent undervaluation through nominal exchange policy to explain the behavior of the real exchange rate in these countries would suggest that an acceleration of the rate of inflation may be in the offing unless nominal exchange rate policy begins to accommodate a nominal appreciation.1
The second interpretation would argue instead—as was indeed argued in the Mexican context—that the relative stability of actual real exchange rates in the East Asian countries has been an equilibrium phenomenon. If so, no nominal appreciation or inflation acceleration would be called for. The difficulty with this hypothesis is reconciling it with the view that the arrival of capital inflows has resulted in an appreciation of the equilibrium rate. The obvious way to do this is to note that capital inflows are only one of several fundamental factors that may drive the equilibrium real exchange rate and that other fundamentals—exogenous, policy-determined, or both—may have moved in such a way as to offset the tendency of capital inflows to appreciate the real exchange rate. On the demand side of the economy, such fundamentals may include, for example, changes in fiscal policy, in the traded-nontraded composition of private expenditure, or in the desired level of private absorption relative to income. On the supply side, they would include sectoral differentials in productivity growth. If these factors have, together, evolved in such a way as to cause a depreciation of the real exchange rate, the net result could be a fairly stable equilibrium rate.
Which of these interpretations is correct is clearly of vital importance for the countries concerned, because the sustainability of their exchange rate paths turns on this issue. This paper attempts to shed some light on this question for five countries of the Association of South East Asian Nations (ASEAN)—Indonesia, Malaysia, Philippines, Singapore, and Thailand. It addresses empirically the issue that distinguishes the two views described above—whether the recent behavior of the real exchange rate in these countries is or is not an equilibrium phenomenon.
The evolution of the real effective exchange rate (REER) for the five ASEAN countries included in this study followed a roughly similar pattern during the 1980s (see Figure 1).2 In particular, during the early part of the decade until approximately 1982–84, all of these countries experienced some real effective appreciation of their currencies. A substantial real depreciation followed in all of them from the period roughly spanning the early part of the international debt crisis in 1982–84 until approximately 1988. The onset of the recent capital-inflow episode in East Asia, when several countries in the region began to receive large inflows of capital, is conventionally dated as of approximately this time. The pace of real depreciation slackened, halted, or was actually reversed in all five countries after 1988.
Figure 1.Real Exchange Rates in Five Southeast Asian Countries
Note: An increase represents a depreciation.
Within this group of countries, real exchange rate paths diverged more after 1988 than before, with two broad patterns emerging. In Indonesia, Malaysia, and Thailand, the real exchange rate was roughly stable or slightly depreciating during 1988–94, whereas the Philippines and Singapore experienced some real appreciation. The magnitudes of the real appreciations of the Philippine peso and the Singapore dollar were both relatively moderate, in each case amounting to a cumulative 17–18 percent over the six-year span. By the end of the period, however, the Philippine peso was still depreciated by about 10 percent relative to its value at the beginning of the decade. By 1994, the Singapore dollar had appreciated only by a cumulative 5 percent relative to its value in 1980, despite the substantial capital inflows that economy had received during the current inflow episode.
A conservative conclusion to draw from this experience is that, while there was no Mexican- or Argentine-style real appreciation during the recent capital-inflow episode, the onset of the episode clearly marks a break in the real exchange rate experience of these countries, and the break is in the direction suggested by conventional wisdom: what had been rapid real depreciation before the arrival of capital inflows became much slower real depreciation or slight real appreciation for these five countries.
Role of Exchange Rate Policy
How do we explain these real exchange rate movements? As a first step, we can attempt to identify the role that exchange rate policy may have played in generating these outcomes. The two hypotheses between which we are trying to distinguish have different implications for the role of active exchange rate policy. The hypothesis that nominal exchange rate management explains the ASEAN countries’ real exchange rate experience presumes an active role for nominal exchange rate policy in pursuit of a real exchange rate objective. By contrast, if the path of the real exchange rate in these countries is an equilibrium one, an active nominal exchange rate policy may or may not be involved. Consequently, if we can rule out an active exchange rate policy for these countries, we can also rule out the first hypothesis.
To clarify these issues, it may be helpful to array the alternative combinations of policy activism and exchange rate outcomes in a matrix describing the four potential ways in which these dimensions can interact. Such a matrix is presented in Figure 2. The hypothesis that the currencies of the five ASEAN countries are currently undervalued because they have been managed actively with a competitiveness objective corresponds to the lower left-hand cell. The competing hypothesis—that these currencies have tracked their equilibrium value—is located along the first row, since they may have done so through active (top left-hand corner) or passive (top right-hand corner) exchange rate management. To interpret the ASEAN experience, it is important to discriminate not just between the two rows, but between the columns as well.
Figure 2.Dimensions of Exchange Rate Measurement
Unfortunately, determining whether exchange rate policy has been managed actively is not a trivial matter. There is no one-to-one mapping, for example, between active exchange rate management and the exchange rate regime. Active management is possible when the exchange rate is predetermined (and the announced parity is adjusted) or floating (and monetary policy is conducted with an exchange rate objective). Except when the nominal exchange rate is literally fixed, or when the authorities adhere strictly to a money growth target, it is also difficult to infer the role of active exchange rate management from the authorities’ policy announcements. When some flexibility is accorded to exchange rate management (as it has in all five of the countries examined here), the monetary authorities inevitably face trade-offs between competitiveness and price stability in formulating exchange rate policies. The shifting weights given to these objectives over time as circumstances change may not be fully recorded in the public record.
We are left, therefore, with attempting to infer the role of active exchange rate policy from exchange rate experience itself. As a matter of accounting, movements in the real effective exchange rate can be decomposed into changes in the nominal effective exchange rate (NEER)(the trade-weighted exchange rate of the domestic currency relative to those of its trading partners) and changes in relative price levels, as given by the identity:
where P* denotes the trade-weighted price level in the country’s trading partners, and P is the domestic price level. NEER is a variable that the authorities can in principle control, either directly through exchange rate policy if the exchange rate is officially determined, or indirectly through monetary policy managed with an exchange rate objective. A first-round assessment of the role of nominal exchange rate policy in producing the REER outcomes described above can be generated by separating the behavior of the real effective exchange rate in each of these countries into these two components and determining how much of the variation in the REER for each of these countries is accounted for by changes in the NEER.
The reasoning behind this approach is the following. If indeed the authorities have managed the NEER actively in the pursuit of a competitiveness objective, trying to keep the actual REER more depreciated than its equilibrium level, then one of two outcomes is possible. If the domestic price level does not adjust quickly to restore the equilibrium value of the REER, then movements in the actual REER will be dominated by movements in the NEER. If the domestic price level does respond quickly, frustrating the attempt to cause the actual REER to deviate from its equilibrium value, then neither the NEER nor relative price levels will be closely correlated with the REER, but they will be negatively correlated with each other. Thus, when movements in the REER are dominated by movements in the NEER, or when neither the NEER nor relative price levels are closely associated with movements in the REER, a role for active exchange rate management cannot be ruled out. If, however, the variations in the REER described above for each country are dominated by movements in relative price levels, then it is difficult to argue that the REER outcomes can be attributed to nominal exchange rate management. A more plausible interpretation in that case would be that the behavior of the REER primarily reflects endogenous responses to other macroeconomic variables.
It is worth emphasizing that, although a strong correlation between the REER and relative price levels suggests that the authorities were not actively managing the NEER to achieve a REER objective, the converse is not true. That is, the absence of such a correlation does not necessarily imply active exchange rate management. This is because, while the authorities can in principle set the value of the NEER to attempt to hit an exogenously determined REER target even if they had not chosen to do so during the period under consideration, the NEER and the REER could be highly correlated. This can come about in at least two ways.
First, the NEER itself can be decomposed into two parts, consisting respectively of the price of the intervention currency (the U.S. dollar) in terms of the domestic currency (denoted EXCH below) and the U.S. dollar price of the trade-weighted currencies of the country’s trading partners (DOLLAR):
The former is the policy instrument actually controlled by the domestic monetary authorities, while the latter is a strictly exogenous component of the nominal effective exchange rate. In equation (2), the authorities can choose a target for either NEER (making EXCH endogenous to changes in DOLLAR) or EXCH (making NEER endogenous to changes in DOLLAR). In the latter case, if movements in NEER are dominated by DOLLAR, a close association between the NEER and the REER could emerge despite a relatively passive stance on the part of the authorities regarding the path of the REER. Following the logic of the interpretation of the REER decomposition, an active exchange rate policy would result in situations in which movements in NEER are dominated by movements in EXCH or in which movements in EXCH and DOLLAR are negatively correlated with each other, causing neither to be closely associated with NEER. Thus, as a second filter in detecting active exchange rate management, in cases where NEER and REER are closely related and where active management cannot therefore be ruled out, we can look at the correlation between NEER and DOLLAR. If these two variables are closely related, the presumption would be against active management even if movements in REER appear to arise primarily from changes in NEER.
But suppose that both filters are passed, that is, that the correlation between REER and P*/P as well as that between NEER and DOLLAR is weak. This does not necessarily suggest that the nominal exchange rate was actively managed in pursuit of a competitiveness objective. EXCH and REER could be strongly correlated through NEER even if the authorities did not manage the exchange rate at all (e.g., with flexible rates under a money growth rule with sticky prices and frequent money demand shocks), or if they managed the exchange rate with other objectives, such as achieving an inflation target or attempting to track the equilibrium real exchange rate.
The key point is that while the two decompositions described above may be capable of providing evidence against active exchange rate management in pursuit of a competitiveness objective, they cannot provide evidence for it. Thus, the question is whether the “competitiveness” hypothesis remains plausible in light of the data. Both of these decompositions are presented graphically for each of the five countries.
The decomposition of the REER into the NEER and relative price levels for Indonesia appears in the top panel of Figure 3, and that of the NEER into EXCH and DOLLAR is presented in the bottom panel. The Indonesian experience appears to consist of three subperiods: during 1980–82 the REER appreciated slightly, from the combined influence of faster inflation at home than in Indonesia’s trading partners and appreciation of the NEER. The latter reflected the increased international value of the U.S. dollar, as the rupiah-dollar exchange rate was stable during this time. Thus, by the criteria established above, this was a period of passive exchange rate policy in Indonesia.
Figure 3.Indonesia: REER Decomposition
Note: An increase represents a depreciation.
Indonesia undertook extensive macroeconomic reforms during the mid-1980s to liberalize its economy and increase its export orientation. Nominal exchange rate management was an important part of this reform. Two major discrete devaluations of the rupiah against the U.S. dollar were implemented in March 1983 (by 27 percent) and September 1986 (by 31 percent). These stepwise devaluations are clearly discernible in the lower panel of Figure 3. While DOLLAR tracks the appreciation of the U.S. dollar against major currencies during the first half of the 1980s, as well as its subsequent depreciation during the second half, the exchange rate of the rupiah against the dollar is responsible for most of the variation in the nominal effective exchange rate during mid-decade. Movements in this rate were large enough to compensate for the inflation differential during this time; consequently, the country’s real effective exchange rate under-went a substantial depreciation during 1983–87, although inflation performance at home continued to be less favorable than that abroad. The evidence provided by the decompositions of the real and nominal effective exchange rates during this period is thus quite consistent with the authorities’ avowed policy objective of moving the latter in such a way as to achieve a competitiveness-driven target value of the real exchange rate.
The final period, 1988–94, is one of remarkable stability in Indonesia’s real effective exchange rate. However, this stability is the result of offsetting movements in the nominal effective exchange rate and relative price levels. It suggests either a policy of using the nominal effective exchange rate to offset smooth inflation differentials—that is, of successful real exchange rate targeting on the part of the Indonesian authorities—or an unsuccessful attempt to use depreciation of the nominal effective exchange rate to achieve a real depreciation of the rupiah. In either case, active exchange rate management is implied during 1988–94.
Thus, there is a strong suggestion that the Indonesian authorities sought to achieve a real effective depreciation of the rupiah after the out-break of the international debt crisis, and that they then used the nominal exchange rate to try to maintain the more depreciated value of the real exchange rate by offsetting an inflation differential that was unfavorable to Indonesia. Active exchange rate policy is therefore at least potentially responsible for the behavior of the real effective exchange rate in Indonesia during the capital-inflow period.
As shown in the top panel of Figure 4, the Malaysian experience has some features in common with that of Indonesia. First, both countries are in the group that avoided real appreciation after 1988. Second, as in Indonesia, movements in Malaysia’s REER were dominated by movements in the NEER to a greater extent during the first part of the decade than subsequently. As in Indonesia, price levels in Malaysia moved relative to those of the country’s trading partners along a relatively smooth trend.
Figure 4.Malaysia: REER Decomposition
Note: An increase represents a depreciation.
However, in other ways, the Malaysian experience contrasts with that of Indonesia. First, in Malaysia, the relative price-level trend took the direction of depreciating the real exchange rate; that is, Malaysia experienced lower inflation than its trading partners. Second, the timing of NEER movements, shown in the top panel of Figure 4, was also different. Malaysia’s nominal effective exchange rate appreciated continuously, and sharply, from the beginning of the decade through 1984 despite mild continuous depreciation of the ringgit against the U.S. dollar, reflecting the dollar’s international strength during these years. Thus, although the behavior of the REER mirrored that of the NEER closely (because Malaysia’s inflation rate was quite close to that of its trading partners over this period), it is clear that, until 1985, the real appreciation of the ringgit was the outcome of a passive exchange rate policy and was dominated by the international appreciation of the U.S. dollar.
From 1985 to 1991, Malaysia’s REER depreciated steeply. A third contrast with Indonesia is that this real depreciation was largely the result of the depreciation of the U.S. dollar against the currencies of Malaysia’s trading partners and a strengthening of the price-level differential in favor of Malaysia. However, the bilateral ringgit-dollar exchange rate continued to depreciate during 1985–91, complementing the effects of U.S. dollar depreciation on Malaysia’s NEER. Indeed, by 1991 the value of DOLLAR had almost returned to where it stood at the beginning of the decade, and the entire cumulative depreciation of the NEER up to that point, amounting to nearly 20 percent, was thus attributable to movement in the ringgit-dollar exchange rate.
Malaysia maintained a flexible exchange rate during this period, with an announced policy of accommodating long-term market-determined movements in the nominal exchange rate and intervening only to smooth out short-term fluctuations. In principle, then, the bilateral depreciation of the ringgit could have reflected either a tracking of the equilibrium rate or a monetary policy that gave significant weight to export competitiveness. That market forces played a role during this period is suggested by events in 1992, when the persistence of capital inflows resulted in an appreciation of the ringgit-dollar rate that was fully reflected in both the nominal and real effective exchange rates. In 1993 and 1994, however, the 1985–91 pattern was reestablished, with both bilateral depreciation of the ringgit and continued U.S. dollar depreciation contributing to depreciation in Malaysia’s NEER.
Because of its flexible exchange rate regime, Malaysia is not as clear-cut a case as Indonesia. On the other hand, the close association of the NEER and the REER and the bilateral depreciation of the ringgit over most of 1985–94 are consistent with an interpretation of exchange rate policy in Malaysia after 1984 in which the authorities were concerned with a long-run competitiveness objective and achieved a desired value of the NEER compatible with that objective by supplementing the exogenous contribution of U.S. dollar depreciation with a monetary policy for which competitiveness was an important consideration. On the other band, the appreciation of the ringgit in 1992 in response to capital inflows suggests that market forces also played a role. Overall, although the contribution of active policy is not as clear as in Indonesia, it cannot be discounted in the case of Malaysia.
The Philippines also recorded a substantial real depreciation in the mid-1980s, amounting to about 30 percent from 1982 to 1988 (Figure 5). However, this was achieved in the context of substantial inflation, significantly higher than in its trading partners during all of this period, and with a particularly large inflation differential during politically unsettled times in 1984–85. The achievement of a large real depreciation in the face of high inflation has meant that the NEER has been on a sharply depreciating trend, with step changes in 1984 and 1986 accounting for a substantial part of the cumulative depreciation of the REER through 1988. The changes in the real value of the peso during mid-decade reflect an active exchange rate policy, with the initial real depreciation in 1982–84 emerging despite U.S. dollar appreciation through large nominal devaluations of the bilateral peso-dollar rate, evident in the bottom panel of Figure 5. A slowing of the bilateral peso-dollar rate of depreciation designed to contain inflation in 1985 resulted in a slight real appreciation of the peso, since the inflation differential continued to be unfavorable, but became compatible with the resumption of real depreciation in 1986–88 as a result of better domestic inflation performance and the decline in the international value of the dollar. The upshot is that, during 1982–88, the authorities actively managed the exchange rate in the Philippines, more than succeeding in preventing inflationary erosion of the real exchange rate.
Figure 5.Philippines: REER Decomposition
Note: An increase represents a depreciation.
From 1989 through 1991, relative price-level performance in the Philippines deteriorated once again, but the REER was maintained approximately stable through an acceleration of the rate of bilateral depreciation against the U.S. dollar, representing a continuation of the active exchange rate policy followed previously. In the context of the arrival of capital inflows, however, the bilateral peso-dollar rate stabilized. The NEER continued to depreciate slightly during this period, but primarily as a consequence of continued dollar depreciation. After a period of relative constancy from 1987 to 1991, the REER appreciated slightly over the rest of the sample period.
Except for this most recent episode, an active exchange rate policy cannot be ruled out for the Philippines during most of the period under review. The cumulative real depreciation of the peso over the period, amounting to about 10 percent, was completely accounted for by movements in the NEER, which in turn were overwhelmingly attributable to the behavior of the bilateral exchange rate against the U.S. dollar. However, it is hard to make the case that the failure of the REER to appreciate more substantially in the Philippines through 1994 was the result of nominal exchange rate management. Exchange rate policy appears to have been relatively passive during the last three years of the sample. This may reflect a policy decision to focus more on competitiveness than on the price level in response to a more favorable balance of payments performance or may simply reflect the influence of changing market forces. To the extent that movements in the NEER may have caused the REER in that country to be depreciated relative to its equilibrium value, this outcome has resulted from the structure of the country’s trade and fluctuations in the international value of the U.S. dollar, rather than from an active targeting of the real exchange rate.
Singapore is an outlier in this group of countries, with respect to both the path followed by its REER during this period and how that path was brought about. In Singapore, monetary policy has explicitly been framed in terms of the exchange rate since 1981 and has also explicitly been targeted at a price-level objective. Consequently, exchange rate policy was almost completely passive during the period of U.S. dollar appreciation in the first half of the 1980s, with very little variation in the exchange rate between the Singapore and U.S. dollars until 1986. As a result, the path of the NEER essentially tracked that of the U.S. dollar against the currencies of Singapore’s trading partners, as shown in the bottom panel of Figure 6. This implied a real appreciation during 1980–82, REER stability during 1983–85, and a sharp real depreciation in 1986. The stability of the REER in the middle period reflected Singapore’s favorable inflation performance (top panel of Figure 6), which essentially offset the appreciation of the NEER caused by the behavior of the U.S. dollar during this time. Similarly, the sharp real depreciation in 1986 was the result of the country’s relative inflation performance magnifying the effects of movements in the value of the U.S. dollar.
Figure 6.Singapore: REER Decomposition
Note: An increase represents a depreciation.
After dollar depreciation set in, however, exchange rate policy became quite active. Consistent with its announced price-level objective, but unlike its neighbors, Singapore used nominal exchange rate policy to achieve an appreciation of its NEER, thereby attempting to offset the effect of U.S. dollar depreciation on the domestic currency price of traded goods. In 1987 and 1988, however, the bilateral appreciation of the Singapore dollar was not enough to offset the depreciation of the U.S. dollar internationally, causing the NEER to continue to depreciate, albeit at a more muted pace. After 1988, the NEER tracked the bilateral rate quite closely and both appreciated continuously despite the depreciation of the U.S. dollar (see bottom panel of Figure 6). The REER followed suit during 1988–94, though to an extent that was muted by the favorable relative price-level performance that continued in uninterrupted fashion until the end of the sample period.
Unlike in the other four countries, REER experience in Singapore has been consistent with what many observers have expected in the context of large capital inflows. The cumulative real appreciation of the Singapore dollar during 1988–94 amounted to about 17 percent. However, this change in the real value of the Singapore dollar was the product of active exchange rate policy, entirely accounted for by the change in the bilateral exchange rate against the U.S. dollar. In principle, Singapore’s experience could be interpreted as active management of the nominal exchange rate to track an appreciating equilibrium rate. However, the deliberate pursuit of price-level stability through the exchange rate instrument in that country suggests that tracking the equilibrium exchange rate was not the authorities’ overriding objective. Ironically, then, despite Singapore’s failure to fit the mold, its experience strengthens the a priori case for the hypothesis that exchange rate policy has made the difference in the ASHAN context, because it seems to suggest that nominal exchange rate policy can matter for real exchange rate outcomes over an extended period.
Unlike the other countries in this group, Thailand avoided real appreciation during the early part of the last decade. The REER was approximately stable from 1980 until 1983, with the NEER and relative price levels moving in opposite directions and by roughly comparable absolute magnitudes (see top panel of Figure 7). A bilateral depreciation of the baht against the U.S. dollar more than offset the effects of the international appreciation of the U.S. dollar on Thailand’s NEER during this period, consistent with an active exchange rate policy.
Figure 7.Thailand: REER Decomposition
Note: An increase represents a depreciation.
Like the other countries, however, Thailand registered a large real depreciation in mid-decade. This was achieved through a 15 percent devaluation of the baht in November 1984. After 1986, however, the baht was stabilized relative to the U.S. dollar. The result was a fairly stable REER from 1986 until 1992, with a slightly unfavorable inflation differential offsetting the effects of the depreciation of the U.S. dollar on the REER. Thus, in contrast with the other countries, the stability of the real value of the Thai baht in the late 1980s and early 1990s appears to have been the product of a passive exchange rate policy. During the last two years of the period under review, REER depreciation emerged once again, but as the result of a depreciation of the U.S. dollar, rather than of active Thai policy.
In Thailand, therefore, the apparent passive-active sequence that characterized some of the countries examined above was reversed. Policy activism appears to have been the rule early in the decade, and, by 1986, it had produced a cumulative real depreciation of the baht of more than 30 percent relative to the beginning of the decade. After this point, exchange rate policy became passive, and the stability of the REER during this time reflected not an explicit effort at real exchange rate targeting, but rather the offsetting effects of exogenous changes in the international value of the U.S. dollar and endogenous domestic price-level adjustments. By the end of the period, the cumulative depreciation of the REER stood about where it had in 1986, the result of movements in the NEER composed cumulatively of roughly equal parts bilateral baht devaluation and U.S. dollar depreciation. Nonetheless, as in the Philippines, the stability of the Thai REER during the capital-inflow episode cannot be attributed to real exchange rate targeting through nominal exchange rate policy. To the extent that NEER developments are responsible for REER performance, the depreciation of the U.S. dollar has played a leading role.
In summary, these five countries shared certain common features in their REER experiences. All but Thailand experienced some real appreciation from the beginning of the decade until approximately 1982–84, after which all of them experienced rather large real depreciations until 1987–88, a time roughly coincident with the inception of the current capital-inflow episode. Indonesia, the Philippines, and Thailand all registered cumulative real depreciations of over 30 percent over this period. Malaysia depreciated in nominal effective terms longer than the others, undergoing by 1990 a cumulative depreciation amounting to about 20 percent, comparable in magnitude to that of the other three countries.
In each of these countries, active exchange rate policy helped determine the behavior of the real effective exchange rate when the real depreciation was achieved. In Indonesia, the Philippines, and Thailand, bilateral depreciation overcame the combined effects of U.S. dollar appreciation and unfavorable differentials between domestic and external inflation rates. The differentials were large for Indonesia and the Philippines and much milder in Thailand. Both a favorable inflation differential and depreciation of the NEER contributed to the steady depreciation of Malaysia’s REER. In Indonesia, the Philippines, and Malaysia, exchange rate policy was clearly passive in the early 1980s, but by mid-decade, Indonesia had adopted a more activist stance, and Malaysia and the Philippines may have done so as well. In recent years, however, the Philippines appears to have reverted to a more passive stance. Thailand, by contrast, began the decade with an active policy that allowed it to avoid the real appreciation experienced by the other countries and switched later to a more passive exchange rate policy stance.
Singapore is the outlier among these five countries. Not only did it undergo a smaller cumulative depreciation during this time—of less than 15 percent—but all of that depreciation emerged in the context of an exchange rate policy that was explicitly managed with a price-level objective in mind. Singapore adopted a continuous bilateral appreciation against the U.S. dollar in 1986, an experience that makes it stand apart from the other countries in this group.
In short, although all of these countries may have experienced periods of activism and passivity in exchange rate policy since the early 1980s, it is fair to say that the cumulative movements in the REER in all of them since that time have been consistent with activism in exchange rate policy. However, after capital inflows began to arrive, only Indonesia and Malaysia continued to manage their exchange rate policy in a manner consistent with the hypothesis that real exchange rate targeting through nominal exchange rate policy may account for REER performance in these countries during the inflow period. In terms of Figure 2, we cannot rule out that the real exchange rate outcomes in these two countries belong in the first column. The Philippines and Thailand (especially the latter), on the other hand, maintained a relatively passive exchange rate stance during the inflow episode. The policies of these countries during the inflow period belong in the second column of Figure 2. However, the structure of their trading patterns, combined with the depreciation of the U.S. dollar relative to the currencies of their trading partners, has implied that NEER depreciation may nevertheless have played a role in determining their REER outcomes, despite the passive stance of exchange rate policy.
The experience of the four larger ASEAN countries, therefore, is consistent with a generalized version of the hypothesis that nominal exchange rate policy has made the difference in East Asia; that is, the relatively stable behavior of the real effective exchange rate in these countries has in each case been associated with a depreciating nominal effective exchange rate, driven either by explicit policy choice or by exogenous variations in the international value of the U.S. dollar. The contrasting experience of Singapore only reinforces the view that nominal exchange rate policy indeed makes a difference, because the appreciation of the REER was associated with a policy-driven appreciation of the NEER. The next task is to determine whether these five countries belong in the first or the second row of Figure 2.
Determinants of Equilibrium Real Exchange Rates
Finding that movements in the REER have been associated with movements in the NEER in the same direction does not necessarily imply that changes in the NEER—whether driven by policy or by exogenous events—have been capable of keeping the REER from moving to appreciated equilibrium values dictated by the presence of capital inflows. The question remains: if the path of the REER is not consistent with long-run equilibrium—and if the concept of a long-run equilibrium real exchange rate is a meaningful one—why have relative price levels not adjusted to move the REER back into line with the value determined by fundamentals? Two answers are possible. The first is that price-level adjustment is very slow and occurs with a lag. In that case, as indicated earlier, the observed path of the REER will not be sustainable, because the future evolution of relative price levels will cause the REER to deviate from the authorities’ intended path. To avoid an acceleration of inflation in these countries, therefore, a nominal appreciation would be required. The second answer is that, despite the apparent role of the NEER in determining the path of the REER, it has infact not strayed far from its long-run equilibrium. The issue, then, as posed at the outset is, how well have these countries been tracking, through their nominal exchange rate policies, the underlying path of the equilibrium real exchange rate?
To answer this question, we need to identify the factors that determine the long-run equilibrium real exchange rate (the “fundamentals”) as well as to quantify their effects on the equilibrium REER. Our present state of knowledge is much more advanced with respect to the first of these than to the second. An extensive literature exists identifying the fundamental factors that drive the long-run equilibrium real exchange rate, which is conventionally defined as the value of the real exchange rate that is simultaneously compatible with internal balance (full employment and equilibrium in the market for nontraded goods) and external balance (a sustainable value of the current account). A recent overview of its fundamental determinants is provided in Montiel (1996b). A number of factors have been identified. They are described below.
Domestic Supply-Side Factors
The most venerable theory regarding long-run real exchange rate determination is the Balassa-Samuelson effect, which relies on differential productivity growth between an economy’s traded and nontraded sectors, favoring the traded goods sector. It causes the equilibrium real exchange rate to appreciate over time, both because excess demand is created in the nontraded goods sector and because the trade balance surplus tends to increase as a result of these differential productivity improvements.
Changes in the composition of government spending between traded and nontraded goods affect the long-run equilibrium real exchange rate in different ways. Additional tax-financed spending on nontraded goods, for example, creates incipient excess demand in that market, requiring a real appreciation to restore equilibrium. In contrast, tax-financed increases in spending on traded goods put downward pressure on the trade balance, and a real depreciation is required to sustain external balance.
Changes in the International Economic Environment
Changes in an economy’s external terms of trade, the availability of external transfers, the level of world real interest rates, and the world inflation rate all influence the long-run equilibrium real exchange rate. Improvements in the terms of trade and increases in the flow of transfers received tend to appreciate the equilibrium real exchange rate, the former by improving the trade balance and creating excess demand for non-traded goods, and the latter through positive effects on the current account. Reductions in world real interest rates and increases in world inflation cause the long-run equilibrium real exchange rate to depreciate. The reason for the real interest rate effect is that lower world interest rates induce capital inflows, which reduce the country’s net creditor position over time, and the long-run loss of net interest receipts requires a real depreciation to maintain external balance. Changes in world inflation affect the equilibrium real exchange rate through effects on transaction costs associated with changes in real money balances. In the model described in Montiel (1996a), the direction of this effect is ambiguous and depends on an essentially arbitrary assumption about the form in which monetary transaction costs are incurred.
Finally, trade liberalization is associated with long-run real depreciation. The effect works by channeling resources into the nontraded goods sectors. The emergence of incipient excess supply in the nontraded goods market dictates the nature of the adjustment in the real exchange rate.
Capital inflows do not appear on this list of fundamental determinants of long-run equilibrium real exchange rates. The reason is that capital inflows are an endogenous phenomenon, likely to materialize as a consequence of change in some other fundamental variable. They thus have an association with the real exchange rate that is determined by the source of the shock that generates the inflow. Among the candidates for this role are changes in world economic conditions—most important, interest rates in the major industrial countries—as well as improvements in the domestic policy environment. Most of the empirical evidence points to changes in international rates of return as a key driving force behind the current capital-inflow episode, and, to the extent that this is so, the effects of these inflows on the long-run equilibrium real exchange rate will be captured by external financial variables.3
With the possible exception of the role of external real interest rates, these effects are well known and would command broad agreement. Using these analytical results to estimate changes in long-run equilibrium real exchange rates is a different matter. This task confronts a number of difficulties, not least of which is the dynamic nature of the problem. At any moment, the actual real exchange rate will differ both systematically and randomly from the underlying long-run equilibrium real exchange rate, and the mechanism of adjustment involves change not just in the nominal exchange rate, but, as indicated above, in relative price levels as well. This means that, to extract empirical information about the unobservable long-run equilibrium real exchange rate from a series of observed real effective exchange rates, the full dynamics of price-level adjustment must, in principle, be specified—an issue that is among the most problematic and controversial in modern macroeconomics for industrial countries, not to mention for individual developing countries.
The approach adopted here is to exploit the time-series properties of the variables involved in this exercise. Theory tells us that the long-run equilibrium real exchange rate depends on long-run values of fundamental determinants such as those listed above. Deviations of the actual real exchange rate from this long-run equilibrium value tend to be transitory, because the economy contains automatic mechanisms that tend to eliminate such deviations over time. The time-series properties of the real exchange rate thus depend on the corresponding properties of its fundamental determinants. If the latter are trend stationary, the real exchange rate will tend to be trend stationary as well, and its behavior will be consistent with, at worst, a modified version of the purchasing power parity hypothesis, which allows for a trend in the equilibrium real exchange rate, perhaps reflecting Balassa-Samuelson effects. If the fundamentals driving the long-run equilibrium real exchange rate are themselves integrated time-series processes, however—that is, if such variables experience permanent shocks during the sample period—the real exchange rate will tend to be an integrated time-series process as well, in violation of the purchasing power parity hypothesis. In this case, if the theory linking the real exchange rate to its fundamental determinants is correct, a cointegrating relationship should exist between the real exchange rate and the fundamental determinants identified by the theory. The residual from this cointegrating equation is the gap between the actual real exchange rate and the long-run value predicted by the fundamentals and is itself a stationary process.
The important point for present purposes is that the cointegrating equation is static, and ordinary least squares (OLS) estimates of its parameters are super consistent (i.e., the estimates converge to their true values more rapidly than usual as sample size increases). This means that, if a cointegrating relationship can be found, one can obtain a consistent estimate of the long-run equilibrium real exchange rate without prior knowledge of the full dynamics of adjustment. This fact represents an enormous advantage in situations such as the one at hand, when we can be much more confident about the theoretical specification of the long-run equilibrium relationship than about the empirics of the short-run dynamics.
These observations suggest the following empirical procedure: first, the real effective exchange rate in each of the countries is tested for the presence of a unit root in its time-series representation. If the series is trend stationary, then a reasonable estimate of the equilibrium real exchange rate can be obtained by fitting a constant and a trend to the available series—that is, by using an estimate of the equilibrium real exchange rate that is based on purchasing power parity. If instead the series contains a unit root, then permanent changes in the REER during the sample period are assumed to have been driven by corresponding changes in some subset of the set of its potential fundamental determinants. Because the nonstationary behavior of the REER must be driven by nonstationary behavior in the fundamentals, the next step is to identify the subset of non-stationary fundamentals. This is done by testing the fundamentals individually for the presence of a unit root. Finally, to identify which among this restricted set of non-stationary fundamentals was responsible for driving the behavior of the REER over the sample, cointegrating relationships are estimated between the REER and the relevant subset of the underlying fundamentals. The fitted values of the cointegrating equation for the REER in each country represent the estimate of the long-run equilibrium real exchange rate, and its residuals measure the gap between the actual and the long-run equilibrium real exchange rates.
Although this approach is attractive for present purposes, it also carries some significant pitfalls. Foremost among these is that the consistency properties apply to large samples when the unbounded variance of the regressors associated with their random-walk properties dominates sampling problems. In small samples, parameter estimates of the cointegrating regression may be biased for all the usual reasons. Because the estimates reported below are based on small numbers of annual observations, this problem is potentially serious. As reported below, the approach adopted in response to this problem has been to rely on estimation techniques that do relatively well in small samples, as well as to use algorithms for specifying the fundamentals that minimize the likelihood of specification error.
The first step in the implementation of this methodology is to test for the nonstationarity of the REER. These tests were conducted in log form for each of the five countries and are reported in Table 1. Testing covered the period 1960–94, except in the Philippines, where a complete set of data was available only for 1963–94. As indicated, sample sizes are thus quite small, and, because the power of these tests is low for such small sample sizes, the robustness of the results was checked by running both the Augmented Dickey-Fuller (ADF) and the Phillips-Perron (PP) tests of nonstationarity. The tests were applied to the levels and first differences of log (REER) for each of the countries.
Each of these tests can be conducted either allowing or not for the presence of a constant and/or a deterministic trend in the underlying time series. The distribution of the test statistic for the unit root test depends on the deterministic variables that are included in the regression on which the test is based. To avoid misspecifying the deterministic component of the time-series model, each test was conducted for the most general trend specification (both a constant and a trend in the series) first, and then the trend and constant were eliminated successively if they proved not to be statistically significant. The test statistic reported in the table in each case is for the final specification derived using this procedure.4
The null hypothesis for this test is that the series is nonstationary, that is, that a unit root is present in the time-series characterization of the series. As is evident from Table 1, this hypothesis failed to be rejected in all but one case: that of Singapore when applying the ADF test. My interpretation of this result is that it reflects a small sample problem associated with the change in regime for nominal exchange rate policy in Singapore alluded to in the previous section. Because the Phillips-Perron test fails to reject nonstationarity, I have chosen to proceed on the assumption that log (REEK) is nonstationary in Singapore as well. The last pair of columns in Table 1 confirms that each series contains at most one unit root—when first differenced, each scries is rendered stationary. The presence of a unit root can be rejected with a high degree of confidence in all cases.5
The next step is to generate empirical counterparts for each of the variables in the set of potential fundamentals described above. Unfortunately, it was not possible to do so for all of the variables that theoretically could enter the cointegrating equation. For the group as a whole, the set consisted of the following fundamental variables:
(1) The ratio of government consumption to GDP (CGR).
(2) The ratio of government investment to GDP (IGR)6
(3) An index of the terms of trade (TOT).7
(4) An index of commercial openness (trade liberalization). Because commercial openness is a multidimensional concept, no single empirical measure of the extent to which the trade regime is liberalized exists. A number of proxies have been used in various applications. For present purposes, it was desirable to adopt a proxy that could be constructed for as many of the countries as possible and for as long a time period as possible. The one adopted was the ratio of foreign trade (exports plus imports of goods and services) to GDP, denoted OPEN. For some countries, however, an alternative measure used was the ratio of trade tax receipts to total trade. This variable was denoted OPNTX. Increased openness is associated with an increase in the value of OPEN, but a decrease in the value of OPNTX.8
(5) External real rates of return. Two measures were used for this purpose. The first was the U.S. six-month treasury bill rate calculated in ex post real terms using the consumer price index for the United States (R*). The second was an index of Japanese unit labor costs (JULC). The latter is intended to capture the fact that foreign direct investment from Japan was an important component of the initial wave of capital flows to these countries during the late 1980s, and the factors driving these flows of foreign direct investment may well fail to be captured by R*. An increase in JULC would reduce the marginal product of capital in Japan and would thus play a role equivalent to a reduction in R*.
(6) The world rate of inflation (INFL), calculated using the U.S. consumer price index.
(7) A time trend. The purpose of the trend is to serve as a proxy for sectoral productivity growth differentials that could affect the equilibrium real exchange rate from the supply side (Balassa-Samuelson effect).
In addition to these variables, which are included for reasons out-lined in Montiel (1996b), the population dependency ratio (DEP) was included as a potential indicator of systematic changes in private saving rates, to reflect demographic changes during the sample period that, while not included in the model that motivated the original specification of the set of fundamentals, may nonetheless affect the behavior of the long-run equilibrium real exchange rate. In the model developed in Montiel (1996b), an increase in the dependency ratio can be interpreted as an increase in the rate of time preference. The long-run effect of this shock would be to reduce the economy’s international net creditor position, which would result in an equilibrium real depreciation.
As previously mentioned, while theory implies that permanent changes in the fundamentals affect the long-run equilibrium real exchange rate, nothing requires that shocks to any particular set of fundamentals over any particular period of time have a permanent component. Fundamentals that do not undergo permanent changes during the sample period will not contribute to changes in the long-run equilibrium real exchange rate over this period and hence do not belong in the cointegrating equation. Except for the trend, then, each of these variables was tested for stationarity, and the nonstationary fundamentals for each country were considered candidates for inclusion in the cointegrating equation. The results of these tests are presented in Tables 2 and 3.
Table 2 presents results for the external variables R*, INFL, and JULC. Of these, only INFL and JULC proved to be nonstationary, in both cases containing a single unit root. R*, however, was stationary over the sample period and was thus dropped as a candidate for inclusion in the cointegrating equations. This implies that potential effects of changes in world financial conditions on the behavior of equilibrium real exchange rates in these countries over the sample period will be captured to the extent that these are reflected in JULC.
The results for the country-specific variables are presented in Table 3. With three exceptions, the time-series properties of these variables proved to conform to those of the REER for these countries. For the Philippines, the openness variable and the dependency ratio were stationary, according to both the ADF and PP tests, and were thus dropped from consideration for the cointegrating equation for this country. In Singapore, the dependency ratio was stationary according to the PP test, but nonstationary even in differenced form according to the ADF test. In view of this rather extreme inconsistency, I omitted the dependency ratio from the cointegrating equation for Singapore.
The cointegrating equations can be estimated in two ways. The simplest approach, that proposed by Engle and Granger (1987), involves estimating an OLS regression with REER as the dependent variable and the potential fundamentals as independent variables. Tests for cointegration among these variables would then be based on the time-series properties of the residuals of this regression in the form of an ADF test, with critical values modified to reflect the fact that the regressor is itself an estimated variable. While this procedure yields superconsistent estimates of the parameters of the cointegrating equation, it has some undesirable finite-sample properties. In particular, coefficient estimates may be biased if the explanatory variables are Granger-caused by the dependent variable (i.e., if weak exogeneity fails) and if the residuals of the equation are serially correlated. Moreover, OLS estimates of the standard errors of the parameter estimates will be biased if there is serial correlation in the residuals, invalidating standard hypothesis tests. These are serious problems for present purposes, because the concern is precisely with obtaining reliable estimates of cointegrating parameters using relatively small samples.
An alternative approach is that of Johansen (1988), which relies on maximum likelihood estimates of the vector autoregression:
where y is an n × 1 vector containing the n variables consisting of the REER and the potential fundamentals, DV is an n × 1 vector of deterministic variables (a constant and a trend), A and βj are n × n matrices of estimated coefficients, and u is an n × 1 vector of serially uncorrelated (but possibly mutually correlated) random shocks. If all the components of y are nonstationary with a single unit root, then the right-hand side must be stationary as well. This means that, if the rows of A contain nonzero elements, these rows must represent cointegrating vectors. A test for cointegration can thus be based on whether there exists at least one such row with nonzero elements (i.e., whether the rank of A is at least one), and estimates of the parameters of the cointegrating vectors emerge from the estimated values of the elements of the linearly independent rows of A. Note that biases arising from simultaneity are eliminated by the vector autoregressive specification, which also removes serial correlation in the residuals, thereby rendering the estimates of the standard errors unbiased and permitting standard hypothesis testing. For these reasons, the Johansen approach was used to extract estimates of the parameters of the cointegrating equations.
One complication in implementing this procedure is that both the distributions of the likelihood-ratio statistic for testing hypotheses concerning the rank of A (and thus for establishing the presence of cointegration) and of the estimates of the individual parameters of the cointegrating vectors depend on the specification of DV, the number of lags included in the autoregression term, and the presence of exogenous variables in the vector autoregression. Accordingly, the procedure for implementing these tests goes from the general to the specific, as follows: first, equation (3) was estimated for each country allowing for the presence of both a constant and a trend in DV and setting the maximum value of j at two in a vector autoregression containing all of the nonstationary fundamentals. Statistically insignificant values of the nonfundamental variables were dropped successively (first, the second lag of Δy, then the time trend, and finally the constant) in subsequent estimation. After the specification of these variables was set, fundamental variables were dropped one at a time if (1) the signs of their coefficients were inconsistent with theory, or (2) their estimates were statistically insignificant (in that order of priority). As each potential fundamental was dropped, the nonfundamentals were reintroduced and the previous procedure was repeated. The final specification chosen for the cointegrating equation in each country was one containing a restricted set of nonstationary fundamentals with the properties that (1) the absence of cointegration among those variables and the real effective exchange rate could be rejected by the data, (2) each parameter of the cointegrating regression was of the theoretically appropriate sign, and (3) the estimate of the parameter was statistically different from zero.
Before proceeding to the results, one final finite-sample issue remains to be discussed. As indicated previously, in the analytical model that underlies the interpretation of these results, a permanent reduction in international rates of return on capital, which triggers capital inflows in the short run, produces a depreciation in the long-run equilibrium real exchange rate, an effect that is opposite in sign to that of a permanent increase in the level of transfers received from abroad. The reason is that short-term capital inflows cause the economy’s international net creditor position to deteriorate in the long run, when the net creditor position has reached a new steady-state value. During the transition, however, when net capital inflows exceed debt-service obligations, the real exchange rate temporarily appreciates. Because large inflows of foreign direct investment to the countries under study are of relatively recent vintage, they still dominate service payments generated by the stock of externally owned capital. Thus, over the sample period, capital inflows triggered by changes in JULC may generate data that look like transfers. Consequently, the sign of JULC in the cointegrating equations could be negative (an increase in JULC triggers foreign direct investment inflows and thus causes the real effective exchange rate to appreciate), as would be consistent with an equivalent transfer, or positive, as would be expected with a set of data that spans a more complete adjustment in each economy’s net international creditor position. To handle these possibilities, the sign of JULC was not imposed.
The resulting cointegrating equations are listed in Table 4. For each country, the final specification of the cointegrating equation is reported, together with the likelihood-ratio statistic for the Johansen eigenvalue test for the null hypothesis that the number of cointegrating vectors is at least one. Several comments about these results are in order.
Focusing first on the methodology, cointegrating equations for the real effective exchange rate that are consistent with theory were found for all the countries in the sample. Although consistency with theory was imposed by the way the specification was chosen, this procedure does not guarantee that the residuals of the resulting equations will be stationary. The results of the Johansen rank tests for all of the specifications listed in Table 4, however, are consistent with stationary residuals, that is, with cointegration among the variables included in the regression. However, the algorithm described above for uncovering the parameters of the cointegrating vector did not perform equally well in all countries. For Malaysia, Singapore, and Thailand, convergence was relatively straight-forward, and, as reported in Table 4, cointegrating vectors were found with economically sensible and statistically desirable properties. For Indonesia and the Philippines, however, this was not the case. For both countries, in order to find a cointegrating vector with theoretically appropriate parameter estimates it was necessary to include variables whose coefficients were not statistically different from zero. Surprisingly, for Indonesia the affected variable was LTOT, which can be expected to be an important determinant of the equilibrium real effective exchange rate in that country on a priori grounds. In the Philippines, LJULC played this role. The a priori expectation would have been the opposite in this case: because the Philippines has not been among the region’s primary recipients of foreign direct investment originating in Japan, this variable would not have been expected to be important. Yet without it, a cointegrating relationship could not be found. This raises the possibility that LJULC is serving as a proxy for a variable omitted from the initial list of fundamentals in the Philippines.
The main results are as follows:
(1) Because the countries under study have grown quite rapidly by international standards over a large portion of the sample period, the Balassa-Samuelson effect might have been expected to generate a negative coefficient for the trend term. Yet, the trend term in the cointegrating equations has the wrong (positive) sign in every case in which the trend proved to be significant. There are several ways to interpret these results. First, for the sample period as a whole (which includes the 1960s and the 1970s), the growth experience of these countries was not as favorable in relative terms as it has been during recent years, particularly in view of the important role of Japan in their external trade. During 1960–85, for example, only Singapore (at 5.9 percent) grew faster than Japan (5.5 percent) in real per capita terms, according to Summers-Heston data. Second, of more direct relevance to the Balassa-Samuelson effect, the growth of labor productivity in these countries was lower than that of output per capita and fell significantly short of that in Japan.9 Thus, one interpretation is that these results are indeed consistent with the Balassa-Samuelson effect for these countries, given the length of the sample, the distinction between growth rates of total output and of labor productivity, and the large weight of Japan in the external trade of these five ASEAN countries. An alternative interpretation, however, is that the trend term may have been picking up the effect of omitted variables and/or measurement error arising from the specification of the empirical proxies for the theoretically appropriate variables.
(2) World economic conditions have an important effect on the behavior of the long-run equilibrium real exchange rate in all five countries. The terms of trade are important for Indonesia, the Philippines, and Thailand, while the world inflation rate matters for all of them. Recall that the latter measures the effect of imported inflation on the equilibrium real exchange rate. Imported inflation matters essentially because inflation affects the productivity of domestic production. The sign of the coefficient of this variable was not imposed ex ante, because it depends on whether adverse effects on productivity are more important in the traded or nontraded sectors, an issue on which little can be said a priori. The coefficient of world inflation was positive in Malaysia, Singapore, and Thailand, but negative in Indonesia and the Philippines. In terms of the model in Montiel (1996b), this is consistent with transaction costs being incurred primarily in terms of traded goods—therefore differentially impairing the productivity of the traded goods sector—in the former group of countries and in terms of non-traded goods in the latter group. Finally, as expected, changes in Japanese unit labor costs turn out to be important for this group of countries, entering the cointegrating equations for Malaysia, the Philippines, and Singapore. As for INFL, the sign of this variable was not imposed, for reasons explained above. Yet, in every case it proved to be negative, consistent with the emergence of a persistent net resource inflow into these countries as a consequence of increases in Japanese unit labor costs.
(3) Turning to the effects of domestic macroeconomic variables, fiscal policy mattered for the long-run equilibrium real exchange rate in Malaysia, Singapore, and Thailand. In Thailand, an increase in government consumption causes the REER to appreciate, consistent with the conventional presumption that government consumption is relatively intensive in nontraded goods. In Malaysia and Singapore, in contrast, increases in government investment are associated with equilibrium depreciation.
(4) Finally, changes in commercial policy, proxied by the openness variable, affected the equilibrium real exchange rate in Malaysia and Thailand.
What do these results imply about the relationship between the actual and long-run equilibrium real exchange rates for these five countries during recent years? The answer to this question is given in Figure 8. The panels of this figure depict the time paths of the actual and estimated equilibrium REER for each of these five countries during 1960–94. While the estimated equilibrium REER had a trend component in each country, the estimate of the equilibrium REER is not a simple, smooth trend. It exhibits a substantial amount of variation in every case. The implication is that fundamentals matter—not just in a statistical sense, as indicated by significant coefficients in Table 4, but also in the economic sense that the estimated time path of the equilibrium REER is quite different from what would be predicted from a simple calculation based on purchasing power parity. To focus on the recent history of the REER in these countries, Figure 9 plots the same variables during 1980–94. The solid line in each panel depicts the path of the actual exchange rate, while the dashed lines represent the fitted value of the cointegrating equation, with bands above and below representing one standard deviation of the gap between actual and fitted values. In interpreting this figure, one should keep in mind that both the predicted values and the estimated standard deviations are derived from cointegrating equations estimated over the full sample, which contains 20 more years of observations than the graphs in Figure 9, using a methodology that does not require the sum of the residuals between the actual REER and the fitted values that constitute the estimate of its long-run equilibrium value to necessarily sum to zero, even over the complete sample.
Figure 8.Actual and Fitted Values of Cointegrating Equations, Full Sample
Note: An increase represents a depreciation.
Figure 9.Actual and Fitted Values of Cointegrating Equations, 1980–94
Note: An increase represents a depreciation.
Figure 9 shows that the sharp real effective depreciation that the currencies of these countries underwent near mid-decade was in part a movement back to equilibrium. In all of them, the domestic currency had become overvalued in real effective terms during the first part of the decade. A second observation, however, is that in all of these countries the real depreciation achieved in mid-decade overshot the equilibrium REER. Thus, according to these estimates, the currencies of all five countries were more depreciated than their equilibrium values by, say, 1989 (1988 in Malaysia). This was not, however, necessarily because the arrival of capital inflows caused the equilibrium REER to appreciate. The equilibrium value of the REER continued to depreciate during the capital-inflow period in Indonesia, Malaysia, and Singapore, which implies that the relative stability in the actual REER for these countries in recent years would eventually lead the equilibrium real exchange rate to catch up with and ultimately pass its actual value. This seems to have happened in all three of these countries by 1992, and the estimated equilibrium REER has remained above, but relatively close to, the actual REER in each of these countries since that time. Thus, instead of the countries having currencies that are currently undervalued, as some observers have supposed in light of the sharp real depreciations of the mid-1980s and the subsequent arrival of capital inflows, these results suggest that, because the equilibrium value of the REER has itself depreciated in recent years, the recent movements in the REER for each of these five countries have closely tracked their equilibrium values. Discernible gaps between actual and equilibrium values, where they have arisen in recent years, have been quite small. While the values of the currencies of Malaysia, Singapore, and Thailand have been greater than their predicted equilibrium ones for several years, these gaps are relatively small in all three cases—roughly equivalent to one standard error of the difference between the actual and estimated equilibrium values of the REER. Statistically, then, they cannot be distinguished from zero. In short, these results provide statistical support for the view that the paths followed by the REER in all five of these countries during recent years have closely followed the equilibrium one.
At least two caveats are warranted, however. The first is that, as already mentioned, the cointegrating equations for both Indonesia and the Philippines are problematic. The second is that small sample sizes have made it difficult to estimate long-run equilibrium real exchange rates with precision, and the failure to detect meaningful deviations from the estimated equilibrium REER may in part reflect that lack of precision. However, it is important to recall that the estimated values of the equilibrium REER in Figure 9 are not based on the observations included in the figure only, but rather on the full sample that includes 20 additional years of data. As inspection of Figure 8 confirms, these estimates are quite capable of generating large estimated differences between actual and equilibrium real exchange rates in all of these countries. Using the same criterion as above, for example, the precision of the estimates is sufficient to detect undervaluation in Indonesia in association with oil shocks in 1974–75 and 1979–80, as well as overvaluation in 1985–86; overvaluation in Malaysia in 1982–84; undervaluation in the Philippines in 1973–74 and overvaluation in 1985; undervaluation in Singapore in 1971–72 and 1986–87; and overvaluation in Thailand during 1982–84. The key point is that, during the recent capital-inflow episode, differences between actual and estimated equilibrium values of the real exchange rates have not been of comparable magnitude.
Determinants of Movements in the Actual Real Exchange Rate
What is the operational relevance of the long-run equilibrium real exchange rate estimated in the previous section? Analytically, the claim that the fitted value of the cointegrating equation represents a long-run equilibrium for the real exchange rate implies that deviations from this value cannot be sustained permanently. For this to be so, such deviations must generate mechanisms that tend to move the actual real exchange rate in the direction of equilibrium. This section tests for the presence of such an error-correction mechanism by modeling empirically the behavior of the actual real effective exchange rate in each country. Because the empirical model is that implied by the analytical approach adopted in the previous section, the ability of this model to account for the annual variation in the actual REER in these five countries provides an independent evaluation of the theory underlying the calculation of long-run equilibrium real exchange rates in the previous section.
The methodology involves estimating the error-correction representation of the behavior of the actual REER in these five countries, using the general-to-specific procedure suggested by Hendry (1986). This paper focuses on two aspects of the estimated regressions: first, the sign and statistical significance of the error-correction term, which provides a test for the presence of mechanisms driving the actual real exchange rate in the direction of the long-run equilibrium value estimated in the last section and, second, the overall goodness-of-fit (R2) of these regressions, to indicate how much of the annual change in the actual REER can be explained by the theory that underlies the calculation of the long-run equilibrium REER in the previous section.
The results of the estimation of the error-correction equations are provided in Table 5. Notice first that for all of these countries, the coefficient of the lagged error-correction term EC is negative, as required to move the REER in the direction of the long-run equilibrium rate, and is statistically significant at the 99 percent level of confidence in every case. The influence of the long-run equilibrium rate on the next period’s value of the actual real exchange rate is strongest in the Philippines and weakest in Thailand. In the former, two-thirds of the gap tends to be closed in one year (other things being equal), whereas in the latter about one-third of the gap is eliminated in that time. For the remainder of the countries, approximately half of the gap between the actual and equilibrium rate is eliminated within a year. These results strongly support the view that the long run equilibrium real exchange rate exerts an important influence on the actual rate. The implication is that, given a policy-determined path of the NEER, relative price levels would tend to evolve in such a way as to move the REER to the equilibrium value estimated in the previous section.
How complete is this story as an explanation of year-to-year changes in the actual REER? The answer is that, on average across these countries, this theory accounts for somewhat less than half the annual variation in the REER, and the variation is large across countries. Recalling that the dependent variable is stationary, that the number of observations is small, and that these countries have undergone substantial structural changes during the sample period, this result is perhaps not surprising. The implication is that the theory of the long-run equilibrium real exchange rate on which these estimates are based provides a useful analytical starting point for the understanding of actual real exchange rate movements in these countries, but much still remains to be explained.
This paper was motivated by the question of how to interpret the recent behavior of the real effective exchange rate of five ASEAN countries. From the start of the recent capital-inflow episode, from 1987–88 through 1994, all five countries had fairly stable real effective exchange rates. Where real appreciation materialized, as in the Philippines and Singapore, it was fairly moderate—certainly far smaller than in several Latin American countries during the same time. While, in the absence of shocks, stability of the real exchange rate would be viewed in a positive light, the arrival of capital inflows constituted a shock that has caused many observers to conclude that the underlying equilibrium real exchange rate might well have appreciated for these countries, meaning that their currencies, possibly including those of the Philippines and Singapore, might have been undervalued. The issue, then, is how far the values of the REERs in these countries departed from their long-run equilibrium values during this period.
To address this issue, it is necessary to generate estimates of the long-run equilibrium real exchange rate for each of these countries. While quite a bit of theory is available to suggest the nature of the fundamental factors that influence the long-run equilibrium real exchange rate, empirical implementation of this theory is rendered problematic by, among other things, the difficulty of modeling short-run price-level determination. Fortunately, the time-series properties of the data can be exploited to extract estimates of the long-run equilibrium exchange rate through the estimation of fairly simple static cointegrating equations among the REER and its fundamental determinants. The key findings from this estimation were the following:
(1) For each of these countries, cointegrating relationships can be found relating the REER to some subset of the potential fundamentals suggested by theory. Despite the availability of only short sample periods characterized by structural changes, it was possible to obtain stationary residuals from equations in which fundamentals generally affect the equilibrium REER in the direction suggested by theory and are generally amenable to measurement at standard levels of statistical precision. This indicates that the notion of a long-run equilibrium real exchange rate is empirically meaningful in this application.
(2) The estimated long-run equilibrium real exchange rate responded to different factors in each of these countries and exhibited substantial variability over time, suggesting that more is at work in determining the equilibrium REER than simple purchasing power parity (which would have generated a constant average value of the long-run equilibrium rate) or long-run sectoral productivity differentials, such as those emphasized in the Balassa-Samuelson effect (which would have generated a smooth trending value of the equilibrium REER).
(3) For each of these countries, the long-run equilibrium real exchange rate has exhibited a depreciating trend over time. While this can be reconciled with the presence of a Balassa-Samuelson effect, it may alternatively represent a measure of our remaining ignorance about the determinants of long-run equilibrium real exchange rates, in the form of omitted variables or errors in representing empirically the theoretically appropriate concepts.
(4) The sharp depreciation in the currencies of these five ASEAN countries during the mid-1980s, while in part a movement back to equilibrium from a situation of overvaluation earlier in the decade, appears to have overshot the long-run equilibrium value of the REER by the late 1980s.
(5) The recent capital-inflow episode did not result in an appreciation of the long-run equilibrium real exchange rate in all of these countries. The equilibrium rate appreciated in Singapore after about 1987 and in the Philippines after 1990, but stabilized or continued to depreciate elsewhere.
(6) The key finding is that REER performance in these countries, from the beginning of the capital-inflow period through 1994, is best interpreted as consistent with long-run equilibrium, given that the gaps between actual and estimated equilibrium REER are not statistically significant.
Finally, to assess the empirical relevance of the theory that supports these conclusions, I examined the links between the equilibrium and actual real exchange rates in these countries. The main finding is that the gap that emerges between the actual and equilibrium values of the REER tends to be self-correcting in all of these countries, as the theory would predict, but the factors emphasized by the theory employed for this exercise leave a substantial fraction of the variation in the actual real effective exchange rate still to be explained in all five countries.
AdamsCharles and DanielGros1986“The Consequences of Real Exchange Rate Rules for Inflation: Some Illustrative Examples,”Staff Papers International Monetary FundVol. 33 (September) pp. 439–76.
CalvoGuillermoLeonardoLeiderman and CarmenReinhart1993“Capital Inflows and Real Exchange Appreciation in Latin America,”Staff Papers International Monetary FundVol. 40 (March) pp. 108–51.
CampbellJohn Y. and PierrePerron1991“Pitfalls and Opportunities: What Macroeconomists Should Know About Unit Roots,” in NBER Macroeconomics Annual 1991ed. byO.Blanchard and S.Fischer (Cambridge, Massachusetts: MIT Press).
EngelRobert F. and C.W.J.Granger1987“Co-Integration and Error Correction: Representation, Estimation, and Testing,”EconometricsVol. 55 (March) pp. 251–76.
HendryDavid F.1986“Econometric Modelling with Cointegrated Variables: An Overview,”Oxford Bulletin of Economics and StatisticsVol. 48 (August) pp. 201–12.
JohansenSoren1988“Statistical Analysis of Cointegrating Vectors,”Journal of Economic Dynamics and ControlVol. 12 (September) pp. 231–54.
LizondoSaul1990“Real Exchange Rate Targets, Nominal Exchange Rate Policies, and Inflation,”Documentos de Trabajo No. 164 (Buenos Aires, Argentina: Instituto Torcuato di Tella, Centro de Investigaciones Económicas).
MarkNelson1990“Real and Nominal Exchange Rates in the Long Run: An Empirical Investigation,”Journal of Monetary EconomicsVol. 28 (February) pp. 115–36.
MontielPeter J.1996a“Managing Economic Policy in the Face of Large Capital Inflows: What Have We Learned?” in Private Capital Flows to Emerging Markets After the Mexican Crisised. byG.CalvoM.Goldstein and E.Hochreiter (Washington: Institute for International Economics).
MontielPeter J.1996b“The Theory of the Long-Run Equilibrium Real Exchange Rate” (unpublished; Williamstown, Massachusetts: Williams College).
MontielPeter J. and Jonathan D.Ostry1991“Macroeconomic Implications of Real Exchange Rate Targeting in Developing Countries,”Staff Papers International Monetary FundVol. 37 (December) pp. 872–900.
SachsJefferyAaronTornell and AndrésVelasco1996“Financial Crises in Emerging Markets: The Lessons from 1995,”NBER Working Paper No. 5576 (Cambridge, Massachusetts: National Bureau of Economic Research).