real and financial linkages and asia-pacific exchange rates - CiteSeerX

kobepapr 5.9.1997

REAL AND FINANCIAL LINKAGES AND ASIA-PACIFIC EXCHANGE RATES by Menzie David Chinn University of California, Santa Cruz and NBER

July 13, 1997

Acknowledgements: Paper prepared for the conference "International Economic Links and Policy Formation: Where Do We Go from Here?" organized by Shigeyuki Abe, Mordechai E. Kreinin and Michael G. Plummer, held at Kobe University, June 12-13, 1997. I thank YinWong Cheung, Chia Siow Yue and the discussant Eric Ramstetter for valuable comments. This paper was completed while the author was a resident fellow at the Rockefeller Foundation’s Bellagio Study and Conference Center, Bellagio, Italy. Portions of the paper were also written while the author was a visiting scholar at the Center for Pacific Basin Monetary and Economic Studies in the Economic Research Department of the Federal Reserve Bank of San Francisco. Financial support of faculty research funds of the University of California, Santa Cruz, is gratefully acknowledged.

Address for Correspondence: Menzie D. Chinn, Department of Economics, Social Sciences I, University of California, Santa Cruz, CA 95064. Tel.: (408) 459-2079; Fax: (408) 459-5900. E-mail: [email protected].

REAL AND FINANCIAL LINKAGES AND ASIA-PACIFIC EXCHANGE RATES ABSTRACT This paper surveys the state of goods and financial market integration in Asia-Pacific. It then examines the evidence for a productivity-based explanation for real exchange rate behavior. Using sectoral output and employment data, relative prices and relative productivity levels are calculated for China, Indonesia, Japan, Korea, Malaysia, Philippines, Singapore, Taiwan and Thailand. Time series regressions of the real exchange rate on relative prices indicate a role for prices for Indonesia, Japan and Korea. When examining real exchange rates and relative productivity ratios, one identifies the posited relationship for Japan, Malaysia, the Philippines and, when augmenting the regressions with real oil prices, Indonesia and Korea. The roles of monetary and real factors are compared using quarterly data. While there is some evidence that monetary factors are important, sensible estimates are obtained only when a indicator variable for relative prices of nontradables versus tradables is included. The importance of sectoral productivity differentials in real exchange rate determination, in combination with widely diverging rates of productivity growth among the East Asian economies, suggest that attempts to manage medium to long term exchange rate movements will have to be very carefully thought out.

Keywords: exchange rates, productivity, tradables, nontradables, purchasing power parity, interest rate parity

JEL: F31, F41

1 Introduction Exchange rates are a key relative price in the open economy. The level of exchange rates is related to exports, imports, the trade balance, and through a variety of direct and indirect channels, domestic output and prices. Expectations regarding future movements in exchange rates also influence capital flows, and asset stocks. Consequently, it is no surprise that policy makers in Asia Pacific economies have actively sought to manage exchange rates either through explicit pegs, or through intervention in foreign exchange markets. Yet, as the recent decision to float the Thai Baht demonstrates, the era of exchange rates managed independently of domestic monetary policy is drawing to a close. This paper examines the determinants of exchange rates, both real and nominal, against the backdrop of continuing economic integration in the region. In doing so, it sheds light on future prospects for exchange rate behavior and policy in the region. The study begins in Section 2 with a survey of the prerequisites for the application of modern asset-based models of exchange rates, namely real and financial market integration.1 The purchasing power parity and interest rate parity conditions are used in this discussion. In Section 3 the long run real exchange rate is modeled. The modeling strategy follows the approach of Harrod (1933), Balassa (1964) and Samuelson (1964). Their models imply that real exchange rate appreciation should be correlated with differentials in tradable/nontradablesector productivity growth. Previous papers have also examined these issues for the industrialized countries; however, with the exception of Japan, one would not expect the Harrod-Balassa-Samuelson (hereafter HBS) effect to be very apparent.2 I examine not only the HBS effect in East Asian countries, but also the roles of demand side factors such as government spending ratios, and real oil prices. The East Asian economies provide an excellent opportunity to search for the HBS effect, since they are characterized by rapid growth, presumably due to rapid manufacturing (and hence traded) sector productivity growth. Unfortunately, past researchers have been constrained by the unavailability of sectoral productivity data, and have either restricted themselves to examining only portions of the of the HBS hypothesis or have assumed that overall productivity growth is highly correlated with the tradable/nontradable productivity differential. A contribution of this paper is to use data more consistent with the underlying economic hypotheses. Section 4 details the application of monetary models of the exchange rate to AsiaPacific currencies. To the extent that the HBS model is relevant, the conventional models typically applied to industrialized countries are unlikely to be very useful. Drawing on a sticky price monetary model due to Frankel (1979), modified to allow long run deviations from purchasing power parity (PPP), I find that such models do have some empirical content. Section 5 concludes with an overview of results and discussion of policy implications. To anticipate the results, I find that the relative price of tradables to nontradables explains some, but not all, of the variation in real exchange rates. Somewhat surprisingly, productivity ratios do appear to be related to the real exchange rate, but only in certain cases. However, at the shorter horizon, nominal exchange rates are influenced by monetary factors, although the size 1

of the effects is imprecisely estimated. Certain policy implications arise from these empirical findings. First, attempts to peg or stabilize currencies around certain levels expressed solely in nominal terms may induce short to medium term deviations from long run equilibrium exchange rates. However, the duration of the deviation will hinge upon the extent of price stickiness, a parameter of uncertain magnitude. If countries do manage or peg their exchange rates, the reference levels need to be adjusted for sectoral productivity differentials, as well as for the inflation differentials typically considered. Second, the model forwarded in this paper cannot be applied in too dogmatic a fashion. Although it is able to explain some proportion of both real and nominal exchange rate behavior, a large degree of uncertainty still surrounds the strength and stability of the relationships examined. Third, because the process of financial liberalization and goods market integration is continuing, the ability of these economies to pursue independent targeting of exchange rates and monetary aggregates/interest rates is destined to be ever more circumscribed.

2 Goods and Financial Market Integration Modern asset based models of exchange rate determination rely upon a number of assumptions. Two key assumptions are that both the domestic goods and financial markets are integrated with their world counterparts. It is a commonplace to assert that the East Asian newly industrializing countries (NICs) are relatively open to both goods and capital. Yet a more nuanced view of the region’s economic history will reveal a wide diversity of experience over time and across countries. And to a certain degree, “openness” is a relative concept; certainly by contrast to policies pursued in Sub-Saharan Africa, and in Latin America up until the 1990s, the Asia-Pacific economies have been fairly open. (Relative to the OECD economies, they are not.) In the following two sub-sections, explicit definitions will be used to examine the degree to which these two markets are integrated in the region.

2.1 Goods Market Integration Goods market integration is usually identified with some version of the Law of One Price: abstracting from transport costs, identical goods in different countries have the same price, when expressed in common currency terms (1)

where st is the exchange rate in US$/foreign currency unit, pit is the price of a US widget, and pi*t is the price of a foreign widget in foreign currency units. An arbitrage argument is often proffered to explain why this condition should hold. At this juncture, data limitations intrude. Typically, one does not have prices for 2

identical goods; rather one observes price indices, p , for bundles of goods. These indices do not usually ascribe the same weights to each good, nor are the quality attributes of these goods (say a car made in the US and one made in Korea) the same, so that direct testing of integration is not possible. What one can test is how well purchasing power parity (PPP) holds, up to a constant which depends upon the base year for the price indices. (2)

That means that one cannot directly measure the extent of the absolute deviations from purchasing power parity, only the version of PPP in equation (2). Nonetheless, a cottage industry has grown around testing this hypothesis (See Breuer, 1994 for a survey); the consensus is that PPP clearly does not hold continuously, and perhaps does not hold over long periods, when one interprets the price index as one pertaining to a broad set of goods and services. Since some of these items (like haircuts) are not tradable and subject to price pressures from international trade, this result is not completely unexpected (see Rogoff, 1996).On the other hand, it is reasonable to believe that the parity condition should hold for price indices measuring tradable goods, pT ; hence I calculate a deviation using as a proxy the producer price index (PPI), which covers goods considered to be tradable.3 One approach is to consider the mean absolute value of (3)

where µ is estimated. This approach suffers from the fact that the measure will differ depending upon the magnitude of shocks perturbing prices, even if the degree of integration is held fixed. Consequently, I consider an alternative measure: the rate at which gaps between the price levels in the two relevant countries converge. The faster the rate of convergence, the more quickly “gaps” are eliminated. If there is no convergence, then one could say that the two markets are not integrated. To implement this procedure, I estimate the following Augmented Dickey-Fuller (ADF) regression, (4)

which yields an estimate of the rate of reversion toward PPP of -Φ. Table 1 presents the results of estimating equation (4) for two subsamples: an early (1974.1-85.4) and late (1986.196.4). The countries covered include Indonesia, Korea, Malaysia, the Philippines, Singapore, Taiwan, and Thailand. Malaysian PPI data is only reported from 1985 onwards, so there is no estimate for the early subsample in this case. In the early subsample, only Korea and the Philippines evidence statistically significant convergence to PPP for the PPI.4 Consider the Korean estimate for Φ of -0.154. This value 3

means that each quarter, 15.4% of the gap in US and Korean tradable goods prices, as measured in a common currency, is eliminated each quarter. The implied half-life of a deviation from price equality is about a year. Phylaktis and Kassimatis (1994) obtained a similar estimate of a year long half life for deviations from PPP in their examination of eight Pacific Basin PPI deflated real exchange rates. Interestingly, in the later subsample the economic and statistical magnitudes of the estimated rates of reversion generally increase. Furthermore, in the Singapore case, not only is the unit root null hypothesis rejected, the half-life of a deviation is only about 1.6 quarters. The Philippines constitutes the sole exception to this pattern, although it should be noted that the reversion coefficient is so imprecisely estimated during the early subsample that the two standard error band easily encompasses the point estimate obtained for the later subsample. An overall reading of these statistics suggest increasing integration of the goods market. The degree of integration is fairly high in the later subsample, with price deviations rapidly eliminated in most cases.

2.2 Financial Market Integration Financial market integration refers to the ease with which assets are traded across borders and currency denominations. A decomposition of the interest differential is helpful at this point. (5)

where it is the interest rate, st is the (log) spot exchange rate at time t, f t is the (log) forward exchange rate at time t (in US$/foreign currency units), for a trade at time t+1, and set+1 is the (log) spot exchange rate expected for period t+1, as of time t. The term fd t is called the forward discount, and (set+1 - st ), the rate of expected depreciation. This identity states that the interest differential can be broken up into a “covered interest differential” in the first set of brackets, and the “exchange risk premium” in the second set, and expected depreciation. The existence of a covered interest differential is often taken as a manifestation of “political risk”, caused either by capital controls, or the threat of their imposition. In the absence of these barriers, such differentials should not exist because they imply unlimited arbitrage profit opportunities. Frankel (1983) terms the condition of a zero covered interest differential “perfect capital mobility.” The absence of an exchange risk premium constitutes “perfect capital substitutability”. This condition arises when government bonds, denominated in differing currencies, are treated as perfect substitutes. Investors will act this either when they are risk neutral, or when government bonds are actually identical in all important aspects. Clearly, these conditions are 4

implausible, especially in the Asia-Pacific context. While a plethora of studies too numerous to mention have examined both issues,5 an accurate breakdown is difficult. First, expected depreciation ( set+1 - st ) is never directly observed. Second, in the Asia-Pacific context, analysis is complicated by the fact that forward market do not exist, or are very thin, for most of the currencies. Consequently, one cannot easily measure the extent of capital mobility. In Table 2, Panel 1, the mean and mean absolute covered interest differential is reported for three currencies: Malaysia, Singapore, and for comparison purposes, Japan. Malaysian absolute covered differentials shrink somewhat, while Singapore’s increases slightly. What one can observe for a wider set of countries is the interest differential. In this case, the interpretation of the measure is less clear-cut. If one assumes that spot exchange rates follow a random walk6, then expected depreciation is zero, and the interest differential represents a composite variable of : (6)

Summary statistics, drawn from Chinn and Dooley (1997), are presented in Panel 2 of Table 2. Malaysia’s mean differential (mean absolute differential) moves from -0.51% to +0.20% (2.13% to 2.99%), and Taiwan’s from -2.64% to +1.04% (2.69% to 2.74%). Singapore’s gap shrinks from -2.62% to -2.02% (2.62% to 2.02%). On the other hand, the differentials rise for Indonesia, Korea, the Philippines and Thailand. On the basis of this information, one might conclude that integration is not advancing. On the other hand, with the exception of the Indonesian and Philippine differentials, the figures are not very wide. Compare for instance the mean absolute differentials in Western Europe, a relatively well developed and integrated region, for the 1987:01 to 1990:04 period: 6.28% for Italy-Germany, 6.34% for UK-Germany, and 10.88% for Portugal-Germany (see Frankel, Phillips and Chinn, 1993). A reasonable view, then, is that the gaps are not shrinking because they are already quite small by international standards. Work by Chinn and Frankel (1994) does suggest that the degree of covariability between US and local market interest rates is increasing over time; this is an aspect not captured by the summary statistics. Furthermore, using measures of expected depreciation derived from surveys of market participants, they also find increasing covariability between local interest rates, and US interest rates in common currency terms, over the 1988-92 period. Finally, Chinn and Frankel (1995) find that Asia-Pacific real interest rates are definitely linked with their US and Japanese counterparts in the long run. Hence, the weight of the evidence argues that capital market in the region are increasingly linked.7 Now that the foundations have been set, one can now proceed to examining the implications for exchange rate determination.

5

3 The Real Exchange Rate

8

3.1 The Model The starting point for most investigations of the linkage between the relative price of nontradables and the real exchange rate relies upon the following construction. Let the log aggregate price index be given as a weighted average of log price indices of traded (T) and nontraded (N) goods: (7) where α is the share of nontraded goods in the price index. Suppose further that the foreign country's aggregate price index is similarly constructed: (8) Then the real exchange rate is given by: (9) where st is the log of the domestic currency price of foreign currency. For α = α*, the following holds: (10) Although there are many alternative decompositions that can be undertaken, equation (10) is the relevant one since most economic models make reference to the second term as the key determinant of the real exchange rate. Isard and Symansky (1996) undertake a slightly different decomposition, allowing the nontraded shares (α's) to change over time. This procedure yields a breakdown of the real exchange rate changes into three components: (i) changes in the relative price of traded goods; (ii) changes in the relative price of traded to nontraded goods; and (iii) changes in share weights (α's). They obtain the result that the first term accounts for almost all of the movements in the real rate for China, Indonesia, Japan, Philippines, and Thailand. Chinn (1996) examines Asia-Pacific exchange rates in this accounting sense, and using cointegration techniques finds that there is some evidence that in the long run, relative prices of tradables and nontradables do explain real exchange rates, at least for certain currencies. In this study, one long run connection examined is the cointegrating relationship (11) obtained when (s - pT + pT*) is assumed to be integrated of order 0, or I(0). This approach, 6

assuming PPP holds for tradable goods prices in the long run, is consistent with that adopted in Section 2.1. In order to truly test the HBS assertion, one can assume that the relative prices of nontradables and tradables are determined solely by productivity differentials, which is the case under the stringent conditions that capital is perfectly mobile internationally, and factors of production are free to move between sectors. (12)

where ai is total factor productivity in sector i, and θ's are coefficients in each sector’s CobbDouglas production functions.

(13)

Invoking again the assumption that the common currency price of tradables is I(0),9 where the production functions in the tradable and nontradable sectors are assumed to be the same, so that the θs cancel out in equation (12). Equations (11) and (13) will motivate the empirical work. 10

3.2 Data and Variable Construction An important contribution of this portion of the paper is the construction and use of appropriate data. Thus, the details of the data and the construction of the empirical counterparts to the theoretical variables merit some discussion (additional details are in the Data Appendix). The annual data encompasses the 1970-92 period, for the following countries: China, Indonesia, Japan, Korea, Malaysia, the Philippines, Singapore, Thailand, Taiwan and the US. The data are drawn from a variety of sources. The nominal exchange rate is the bilateral period average, expressed against the US$ (in $/foreign currency unit). Price levels are GDP deflators drawn from the World Bank's W orld Tables. The real exchange rates are then expressed in terms of GDP deflators, rather than the typical CPIs. However, the difference in the behavior of the variables is minor.

The prices of tradables and nontradables are proxied by the price deflator for manufacturing, and for "other", which includes mostly services, construction, and transportation, respectively. These data are calculated as the ratio of nominal to real (in 1987 domestic currency units) sectoral output, as reported in W orld Tables, except in certain cases, 7

such as Malaysia and Taiwan, where the Asian Development Bank's Key Indicators of Developing A sian and Pacific Countries are used. Note that this ratio omits the agricultural sector. Hence, unlike the approach adopted in Engel (1995a) and Chinn (1996), the overall price deflator is not a weighted average of the two sectoral deflators. I have also examined an alternative measure of the relative price of tradables and nontradables, where a weighted average of agricultural and manufacturing prices is used in place of manufacturing prices. Although the time series pattern for Thailand, and the trends for China, Indonesia and Korea, do change, none of the statistical results for cointegration hinge upon this alternative definition. For specific results, see Appendix 1. The exchange rate series and the respective relative price series are displayed in Figures 1 through 9; the average annual rate of change for these series are also reported Panel 1 of Table 3. To make ocular regressions easier to conduct, the exchange rates are expressed in inverse terms (foreign currency unit/ US$) and normalized to 1974 = 0 in these figures; hence the country pairs should covary positively according to the theory described above. With the exception of China and Taiwan, there does appear to be some comovement of the posited nature. As remarked above, attempts to test the Harrod-Balassa-Samuelson model's predictions regarding Asia-Pacific productivity levels and currencies have been hampered by the lack of sectoral productivity data. Consequently, previous researchers have resorted to proxies, such as GDP per capita,11 or aggregate productivity. However, in such cases, either the α coefficients are not identified or an omitted variable situation arises, and cointegration is unlikely to be found, even if the model in (13) holds true.12 It is therefore critical to find adequate proxies for the sectoral productivity levels. I calculate average labor productivity as the proxy for sectoral total factor productivity; average labor productivity is obtained by dividing real output in sector i by labor employment in sector i: (14) where i takes on the same categories as before. The output series are drawn from the same sources used for the relative price deflators. The employment figures are drawn from the W orld Tables, the ADB Key Indicators, and for the developed countries, the International Labour Office's Y earbook of Labour Statistics. Two limitations of the data should be stressed. First, since these labor employment statistics are not adjusted for part time workers, one may very well have qualms about the reliability of these proxies for labor productivity. To cross- check the results, I have compared the calculations for manufacturing productivity against those reported by the W orld Tables for several countries. These figures match quite well. Furthermore, for the US and Japan, we have data on labor productivity in the traded (industrial output, transportation) and nontradable 8

(services, construction) sectors from the OECD's International Sectoral Data Base. These series also match quite well for manufacturing vs. tradables, and "other" versus nontradables. These results serve to improve one's confidence that the proxies we use are not wholly inadequate. Second, the proxy variable is labor productivity, rather than total factor productivity as suggested by the model. Canzoneri, Cumby and Diba (1996) have argued that use of labor productivity is favored because it is less likely to be tainted by mis-estimates of the capital stock. In any event, there is little possibility of circumventing this problem. To my knowledge, almost all calculations of Asia-Pacific total factor productivity over long spans of time have been conducted on an economy-wide basis, with the exception of Alwyn Young's (1995) recent analysis. The (inverse) exchange rate and relative productivity variables are displayed in Figures 10-18 (rescaled to a common base year equal zero); the average annual rates of change are reported in Panel 2 of Table 3. Once again, China and Taiwan fail to exhibit the expected correlation. In the other cases, the covariation is not very pronounced, although in a eclectic model of relative price determination, demand side factors such as government spending comes into play. Hence, it appears worthwhile to proceed with the statistical analysis. The other variables are government consumption spending, in nominal terms. The variable used in the regressions is the log of the ratio of nominal government consumption to nominal GDP. It would be preferable to have real government consumption divided by real GDP as the government spending measure. In the absence of these data (and the likelihood that such government spending deflator data would be even more flawed than developed country equivalents), I use the nominal measure. Real oil prices are measured as the $ price per barrel of Dubai Fateh, deflated by the US CPI (in log terms). Almost all the series appear to be integrated of order 1, according to ADF tests (with constant and trend) with lags of order 1 and 2. The exceptions are the US-China productivity differential, the US-Malaysia relative price ratio (although only 12 observations are available for this series), the Singapore real exchange rate (all at the 5% MSL) and the US-Taiwan relative productivity differential (at the 1% MSL). I will proceed under the presumption that all series are I(1).

3.3 The Real Exchange Rate and Relative Prices 3.3.1 Econometric Specification If the relative price of traded goods is stationary, then the series q, pN -pT , pN* -pT* , are cointegrated with the cointegrating vector implied by theory of (1 -α +α*). Best current practice is to use the multivariate maximum likelihood technique of Johansen (1988). Unfortunately, this procedure yields very dispersed estimates (for reasons suggested by Stock and Watson (1993)). Moreover, since there are very few observations, few instances of 9

cointegration are found after adjusting the critical values for the loss of degrees of freedom.13 Given the implausible estimates derived from the Johansen procedure, I implement a single equation error correction model, as suggested by Phillips and Loretan (1991). One always has to worry about endogeneity in any single equation regression. Fortunately, in this model, it is plausible to assume a recursive structure such that (pN -pT ) and (pN* -pT* ) are weakly exogenous for q. Hence, in this paper I adopt a single-equation error correction modeling approach, viz:

(15)

This equation is estimated using nonlinear least squares (NLS) The asymptotic distribution of the t-statistic associated with the φ parameter is approximately Normal under the above assumptions. 3.3.2 Empirical Results Regressions of the form of (15) were estimated for all real exchange rates. In general k=m=1, except where noted. The results in Table 4 indicate that the real exchange does exhibit some mean reversion. In all cases except China, Taiwan and Thailand, the coefficient on the lagged error correction term appears to be statistically significant. In these cases the estimated rate of reversion toward equilibrium is above 25%. For the exception of China, the statistical results are unsurprising, given the behavior of the real exchange rate over the sample period, and the clear violations of the assumptions of the Harrod-Balassa-Samuelson model -- full capital mobility, and free intersectoral factor mobility. For Taiwan and Thailand, the failure to find reversion is more puzzling. The coefficient on relative nontradable/tradable prices is less precisely estimated. The three significant estimates appear in the plausible range of 0.28 to 0.76 (recall this coefficient has the interpretation of the share of total expenditures on nontradables). Even the nonsignificant estimates are plausible, with the exception of Singapore and Taiwan. In light of Young's (1995) findings regarding the close rates of productivity growth in manufacturing and service sectors in both these countries, it would be unsurprising to find a lack of a role for relative prices.14 , 15 Interpreting deviations from equilibrium are due to movements in the common currency price of tradables, Korea evidences the most rapid rate of convergence. The implied half life of a deviation is less than a year. On the other end of the spectrum, the implied half life for a Indonesian deviation is 2.3 years.

10

One must be forthright about the limited results obtained here. For China, Taiwan and Thailand, there is no statistically substantive evidence for reversion of any sort. For Malaysia, the Philippines and Singapore, there is some evidence that there is reversion, but the effect of relative prices is too imprecisely estimated to make any definite conclusions. Only for Indonesia, Japan and Korea are the posited effects identified; and for Japan, one is already aware of the evidence in favor of the relative price effect (Strauss, 1996; Chinn and Johnston, 1996; Chinn, 1996).

3.4 The Real Exchange Rate, Productivity and Demand Shocks 3.4.1 Econometric Specification Since the productivity data span even shorter periods than the relative price data, conservation of degrees of freedom is at a premium. In principle, the NLS procedure adopted in the previous section is feasible. However, estimation of the implied error correction models yielded highly implausible estimates. Kremers, Ericsson and Dolado (1992, hereafter KED) propose a test where the cointegrating vector is imposed a priori. The test for cointegration can then be applied quite simply by evaluating the t-statistic on the error correction term. As Zivot (1996) has pointed out, the distribution for this t-statistic depends upon a number of assumptions, most importantly on the validity of the imposed a priori cointegrating vector. Since in the absence for a role for demand side shocks, (16) I substitute relative productivity terms for relative prices, and impose the constraint α = 0.5. This implies that the share of nontradables in the aggregate price index is one-half. Then I estimate the equation, allowing the level of relative productivity to enter in separately. Zivot shows that if this variable enters in significantly, then the imposed cointegrating vector is invalid. For the instances in which this is true, I estimate the equation unconstraining the cointegrating vector using NLS. To summarize, the unconstrained specification is:and the

(17)

constrained specification is: (18)

3.4.2 Empirical Results Preliminary regressions indicated that the α = 0.5 restriction was violated by the data 11

in only two cases -- Singapore and Taiwan. Hence, regressions of the form of (17) were estimated for all real exchange rates except those two currencies, in which case equation (18) was implemented. In general k=m=1 (except where noted). The results in Table 5 indicate that the real exchange does appear to be cointegrated with relative traded/nontraded productivity levels for Japan, Malaysia and the Philippines.16 Exchange rates for Singapore and Taiwan appear to be unexplained by relative productivity differentials, as indicated by the zero reversion and wildly implausible coefficient estimate on productivity in the former, and the positive estimate on productivity in the latter. These results for Taiwan contrast with those obtained by Wu (1996) who finds that the real exchange rate is cointegrated with relative productivity levels and relative unit labor costs, expressed in a common currency. However, the point estimates on the relative productivity variables imply that over 100% of the output in the US and Taiwan are nontradables. Notice that the formulation embodied in these regressions assumes that supply shocks, in the form of productivity growth, are the only determinants of the relative price of tradables to nontradables. Dropping that assumption, I allow for government spending to affect the contemporaneous real exchange rate. If government spending falls mostly on nontradables, then a change in government spending should induce a current appreciation in the real exchange rate. I also check to see if oil prices have an effect. In previous productivity-based studies of the real exchange rate, real oil prices have entered significantly (Chinn and Johnston, 1996; DeGregorio and Wolf, 1994). In the Japanese and Korean cases, where all oil is imported, the channel could either work its way through the wealth effect, or via shifts in the production function The results of these re-specifications are reported in Table 6. Inclusion of the government spending variable reduces the statistical significance of the coefficient on the error correction term for several currencies, but increases it for Indonesia and Korea (as long as oil prices are included). At the same time, the government spending variable often enters in the regression with unexpected sign. For Malaysia and Taiwan, changes in the local government spending to GDP ratio cause depreciation of the local currency. This is consistent with most government spending falling on tradable, rather than nontradable, goods. On the other hand, US government spending, when enters significantly, does appreciate the US$ (the Indonesian and Korean cases). Real oil prices prove to be a significant factor in three cases -- Indonesia (a substantial oil exporter) and Japan and Korea (two wholly dependent oil importers). Log real oil prices enter into the cointegrating vectors for Indonesia and Japan, and into the short run dynamics for Korea. The direction of effects is consistent with priors: Increases in oil prices appreciates the Indonesian Rupiah against the US$, and depreciates the Yen (in both short and long run) and Won (in the short run). It appears that the evidence, after accounting for the role of oil and government spending, is fairly strong for a productivity-based model of real exchange rates in the AsiaPacific region. Indonesia, Japan, Korea, Malaysia and the Philippines appear to fit the general 12

pattern. However, China, Singapore, Taiwan and Thailand do not appear to fit into this framework.

4 NOMINAL EXCHANGE RATES AND MONETARY VERSUS REAL FACTORS17 Thus far the discussion has been couched in terms of real variables, and long run relationships. However, for many questions of policy, one is interested in the behavior of nominal magnitudes over both the short and long run. Hence, it is useful to consider the role of monetary factors in the context of the real relationship identified in the previous sections.

4.1 The Conventional Monetary Model The asset-based approach to monetary models of the exchange rate, as examined for instance by Mark (1995), relies upon a money demand equation of the form: (19) where mt is the (log) nominal money stock, pt is the (log) price level, yt is (log) income and it is the interest rate. Hence, the demand for real balances is an increasing function of income, and a decreasing function of the interest rate. The parameters have structural interpretations. φ is the income elasticity, and λ the interest semi-elasticity, of money demand Building upon this money demand function, the condition that expected depreciation equals the nominal interest differential (i.e., perfect capital mobility and substitutability), and continuous purchasing power parity one obtains the flexible price monetary model: (20) where st is the logarithm of the exchange rate (US$ per unit of foreign currency) and * denotes the local country variable. The intuition for this type of model is as follows: the larger the money supply, m, relative to the determinants of the money demand, such as income and interest rates, the weaker (the higher) the exchange rate. When income rises, more money is demanded for transactions purposes, and the exchange rate falls (or appreciates). Interest rates decrease money demand, so that the higher the domestic interest rate, the lower the demand for money and hence the weaker the domestic currency. The model above imposes an assumption that prices are perfectly flexible; for many purposes this is an undesirable restriction. One way to relax this assumption is to allow prices to adjust slowly, and monotonically, so as to close the gap between the current price and the 13

long run price. Allowing for secular inflation, the exchange rate behaves as: (21)

where Πt is the expected CPI inflation rate from t to t+1, and Θ is the rate of reversion of the price level to its long run value. Equation (21) states that the exchange rate depreciates whenever the current exchange rate is stronger than its long run value (denoted by the overbar). Imposing rational expectations, one then obtains the following model due to Frankel (1979) following Dornbusch (1976): (22)

Now interest rate differentials and inflation rate differentials enter in separately whereas in the previous formulation, the inflation differential and the interest differential were one and the same. When the money supply is increased in this model, domestic real money balances increase too, since prices are sticky. To equilibrate the money market, domestic interest rates must fall. However, since uncovered interest parity must hold, the exchange rate must be appreciating over time, even though in the long run the exchange rate must be depreciated relative to its initial value. Hence, the response of the exchange rate to a interest rate decrease is an instantaneous depreciation of the currency. The more rigid prices are, the larger this “overshooting” effect. This model is more appropriate than the flexible price version if only because it accords better with casual empiricism -- higher interest rates, ceteris paribus, are associated with stronger currencies.

4.2 Problems in Application to Asia-Pacific Because the monetary approach is built on perfect capital mobility and substitutability, it is unreasonable to expect that these models would hold very well for Asia-Pacific NICs. Indeed, one has to look long and hard for such models to work for even developed country currencies (Chinn and Meese, 1995; Mark and Choi, 1994). As discussed in Section 2.1, some of these NICs are only now removing restrictions on the capital account, e.g., Korea (see Chinn and Maloney, 1996, for details). Hence, perfect capital substitutability and mobility as defined by Frankel (1983) are unlikely to hold. If, however, the risk premium associated with holding domestic bonds is constant, or nearly so, then this violation of the assumptions will not pose a large problem for estimation. A related issue pertains to existence of a "market determined" interest rate. While interest rate series are reported for all of these NICs, many of them are regulated, or at least 14

subject to informal intervention by the monetary authorities. When the money market rate is not freely determined, then it is unclear whether the market clearing conditions assumed in the monetary approach hold. Perhaps a more fundamental issue pertains to the stability of the money demand functions. Recall that equations (20) and (22) are built upon the money demand equation in (19). Even in developed countries, there is substantial debate over the stability of such equations in light of financial innovation. In LDCs, where economies are subject to monetization, increasing financial intermediation, or financial repression, a priori assertion of money demand stability is even less sustainable. Tseng and Corker (1991) assert that a stable cointegrating relationship does hold for several of these East and Southeast Asian countries. A key assumption of the monetary approach is that purchasing power parity holds in the long run. For the reasons elucidated in the Section 3, this is an undesirable assumption. Using equations (7) and (8) from Section 3, one can write the relative price level as: (23)

Now let the nominal exchange rate behave as follows: (24)

In other words, tradable goods inflation, ΠT, is what is relevant. Further, the long run nominal exchange rate is given be purchasing power parity only in tradable goods, (25)

so that one then obtains the following expression:

(26)

where the price of nontradables (tradables) is proxied by the CPI (PPI). In the empirical section, Π = ΠT for simplicity. In principle, one would like to substitute out for the determinants of the relative price variable in the square brackets, especially since the price of tradables is likely to be endogenous to the exchange rate. Chinn (1997a) undertakes regressions where productivity in 15

manufacturing is used as a proxy for relative sectoral productivity, for the US, Canada, Germany, Japan and the UK. However, such productivity data is not readily available for many of the countries being investigated. Hence, I proxy the HBS effect with this relative price variable, which is calculated as: (27)

4.3 Empirical Preliminaries Before embarking upon a statistical analysis of the exchange rate equations, it might be of interest to see how well the relationship holds in a cross-section, over long periods of time. First I calculate the fundamentals F assuming perfect price flexibility, which is appropriate for the long run comparison being made. The fundamentals are defined as:18 (28) The relationship between the exchange rate and the fundamentals is one-for-one if the fundamentals follow a random walk. If the fundamentals follow some stationary autoregressive process, say an AR(1), then the relationship should be less than one-for-one, with the deviation from unity dependent upon the autoregressive coefficient. In Figure 19, the average quarterly change in nominal exchange rates (in US$/currency unit) is plotted against the change in the fundamentals for the 1985.1-94.4 period. There is a definite positive relationship, with the slope coefficient of 0.60 (the relationship is significant at the 5% level). However, even a cursory examination indicates that some currencies fall far above or below the regression line, indicating a wide diversity of experience. Indonesia in particular falls one percent below the line; one percent does not seem like a lot, but averaged over ten years, this represents a cumulative deviation of 40 percent.

4.4 Econometric Evidence In order to investigate the relationship more carefully, time series regressions using the Phillips-Loretan nonlinear least squares technique (see Phillips and Loretan, 1991) are implemented. This procedure involves running a regression including at least one lead of all the right hand side variables, to account for endogeneity of the regressors. (29)

Where X is a vector of explanatory variables.

16

The samples are chosen so that they contain only periods when the exchange rates are floating. In certain instances, this truncates the sample considerably; Indonesia allows its exchange rate to float only after 1985. The results are reported in Table 7; they indicate that there is indeed a wide variety of experiences. Korea, Malaysia, Taiwan, Thailand and possibly Singapore all exhibit statistically significant mean reversion (Japan is included for purposes of comparison).19 However, the cointegrating vectors for Taiwan and possibly Singapore do not conform to priors. In particular, the coefficients on money are perversely signed. The Korean Won is the only currency for which money and income have correctly signed coefficients, and at the same time exhibit a correctly signed, statistically significant, relative price variable. For all exchange rates, the interest and inflation rate coefficients are not generally significant. What can one make of these results? One positive aspect of the results is that all the coefficients on the relative price variable are positive, with the exception of Singapore's and Thailand's (neither significantly though). This is hardly a resounding confirmation of the theory; however, if the relative price variable is omitted from the regressions, then the results are even less conducive to adducing a role for monetary variables. Money coefficients are more often incorrectly signed, with statistical significance, while the standard errors on the interest and inflation coefficients are much larger. In one sense, the relatively poor results are unsurprising. Portions of the samples encompass periods marked by capital controls. These capital controls could have effectively segmented the domestic money markets and the foreign exchange markets, so that exchange rates could be manipulated independently of monetary policy. One implication of increasing liberalization is that such manipulation will become less and less viable over time.

5 Conclusions In the first portion of this paper, the strength of goods and financial linkages was assessed. While integration is far from complete, there is a substantial degree of goods market integration, and financial capital mobility evident. These empirical conclusions allow me to proceed to the next sections of the paper, namely the application of neoclassical models of the exchange rate. I first examine the evidence for a productivity-based explanation for long run movements in Asia-Pacific real exchange rates. I find, somewhat in contrast to Isard and Symansky (1996), that there is some evidence in favor of such explanations. However, the evidence is by no means conclusive, especially as it relates to some of the countries now growing the most rapidly -- China, and Thailand. Furthermore, it must be admitted that the real exchange rate exhibits large swings 17

away from the equilibrium rate as predicted by either the relative price of tradables to nontradables, or relative productivity levels. The most plausible explanation for this finding -that relative prices of traded goods can exhibit substantial persistence -- suggests that closer examination of the changing composition of Asia-Pacific exports and imports over time is warranted.20 The finding that some real exchange rates appear to move in response to relative productivity differentials suggests that nominal pegging of exchange rates is an enterprise fraught with pitfalls. To understand this contention, consider what would happen if Korea were to peg the Won against the US$, and the initial peg were set at "equilibrium". In the absence of price level adjustment, the real exchange rate would be constant. However, the US-Korea trend relative productivity differential over the 1970-90 period is approximately 1.3% per annum (see Table 1); given a share of nontradables in overall output of 0.5, the trend real appreciation of the Won should be about 0.6% per year. A fixed peg that survives five years would induce a 3% undervaluation. Since nominal prices are usually considered to less flexible than exchange rates, such a peg inhibits economic adjustment. Further, to the extent that other Asia-Pacific countries also peg to the US$, this problem could be even more substantial. If both Korea and Thailand were to peg to the US$, then with a trend relative productivity differential between the two Asia-Pacific countries of approximately 4% per annum, the implied trend real appreciation of the Won relative to the Baht is about 2%. Clearly, none of these countries is likely to adopt a rigid peg (either explicit or implicit) against the US$; however, these concerns still arise, albeit with less force, if monetary authorities attempt to stabilize their currencies around mis-estimated trending reference values, such as was the case with the Mexican crawling peg. The contentious issue is not whether such stabilization is in principle desirable; rather it is how in practice to identify the appropriate trends for target exchange rates in a region characterized by rapid productivity growth. The empirical results from this paper, and recent experience with nominal pegs, suggest that policy makers will have to proceed on this front with great caution. What is also true is that as these countries proceed with financial integration (including the liberalization of domestic banking systems) and the differential in sectoral productivity growth slackens, monetary factors will become more important in determining exchange rates. As the textbook Mundell-Fleming model makes clear, then the simultaneous manipulation of exchange rates and money supplies to hit independent targets will become more difficult over time.

18

References Bahmani-Oskooee, Mohsen, and Hyun-Jae Rhee, 1996, "Time-Series Support for Balassa's Productivity-Bias Hypothesis: Evidence from Korea," Review of International Economics 4: 364-370. Balassa, Bela, 1964, "The Purchasing Power Parity Doctrine: A Reappraisal," Journal of Political Economy 72: 584-596. Breuer, Janice Boucher, 1994, "An Assessment of the Evidence on Purchasing Power Parity," in John Williamson (ed.), Estimating Equilibrium Exchange Rates (Washington, DC: Institute for International Economics): 245-277. Canzoneri, Matthew, Robert Cumby and Behzad Diba, 1996, "Relative Labor Productivity and the Real Exchange Rate in the Long Run: Evidence for a Panel of OECD Countries," NBER W orking Paper #5676. Cheung, Yin-Wong and Kon. S. Lai, 1993, "Finite-Sample Sizes of Johansen's Likelihood Ratio Tests for Cointegration," Oxford Bulletin of Economics and Statistics 55(3): 313-328 Chinn, Menzie, 1997a, “On the Won and Other East Asian Currencies,” Mimeo (University of California, Santa Cruz, September). Chinn, Menzie, 1997b, "Paper Pushers or Paper Money? Empirical Assessment of Fiscal and Monetary Models of Exchange Rate Determination." Journal of Policy Modeling 19(1): 5178. Chinn, Menzie, 1997c, “The Usual Suspects: Productivity and Demand Shocks and AsiaPacific Real Exchange Rates,” NBER W orking Paper #6108. Chinn, Menzie, 1996, "Asian-Pacific Real Exchange Rates and Relative Prices," Working Paper #358 (Dept. of Economics, Univ. of Calif.: Santa Cruz, July). Chinn, Menzie and Michael Dooley, 1997, "Asia-Pacific Capital Markets: Measurement of Integration and the Implications for Economic Activity," forthcoming in T. Ito and A.O. Krueger (editors) Regionalism versus Multilateral Trading Arrangement (Chicago U. Press for NBER: Chicago), pp. 169-196. Also NBER W orking Paper #5280 (September); Chinn, Menzie and Jeffrey Frankel, 1995, "Who Drives Real Interest Rates around the Pacific Rim: The US or Japan?" Journal of International Money and Finance 14(6): 801-821. Chinn, Menzie and Jeffrey Frankel, 1994, "Financial Links around the Pacific Rim: 19821992," in R. Glick and M. Hutchison (editors) Exchange Rate Policy and Interdependence: Perspectives from the Pacific Basin. (Cambridge: Cambridge Univ. Press), pp. 17-26. 19

Chinn, Menzie and Louis Johnston, 1996, "Real Exchange Rate Levels, Productivity and Demand Shocks: Evidence from a Panel of 14 Countries," NBER W orking Paper #5907 (August). Chinn, Menzie and William Maloney, 1996, "Financial and Capital Account Liberalization in the Pacific Basin: Korea and Taiwan." NBER W orking Paper #5814 (November); forthcoming, International Economic Journal. Chinn, Menzie and Richard Meese, 1995, "Banking on Currency Forecasts: Is Change in Money Predictable?" Journal of International Economics 38 (1-2): 161-78. DeGregorio, Jose and Holger Wolf, 1994, "Terms of Trade, Productivity, and the Real Exchange Rate," NBER W orking Paper #4807 (July). Dornbusch, Rudiger, 1976, "Expectations and Exchange Rate Dynamics," Journal of Political Economy 84: 1161-76. Engel, Charles, 1995a, "Accounting for US Real Exchange Rate Changes," NBER W orking Paper #5394. Engel, Charles. 1995b, "The Forward Discount Anomaly and the Risk Premium: A Survey of the Recent Experience," NBER Working Paper #5312 (October); forthcoming, Journal of Empirical Finance. Frankel, Jeffrey A., 1991, “Quantifying International Capital Mobility in the 1980s,” in D. Bernheim and J. Shoven (editors) National Saving and Economic Performance (Chicago: Chicago University Press): 227-60. Frankel, Jeffrey A., 1983, "Monetary and Portfolio Balance Models of Exchange Rate Determination," in J.S. Bhandari and B. Putnam (eds.) Economic Interdependence and Flexible Exchange Rates, MIT Press: 84-115. Frankel, Jeffrey A., 1979, "On the Mark: A Theory of Floating Exchange Rates Based on Real Interest Differentials," American Economic Review 69: 610-622. Frankel, Jeffrey A., Steven Phillips and Menzie Chinn, 1993, "Financial and Currency Integration in the European Monetary System: The Statistical Record," in F. Torres and F. Giavazzi (editors) A djustment and Growth in the European Monetary Union. (Cambridge: Cambridge University Press/CEPR): 270-305. Harrod, Roy, 1933, International Economics (London: Nisbet and Cambridge University Press). Hsieh, David, 1982, "The Determination of the Real Exchange Rate: The Productivity 20

Approach," Journal of International Economics 12 (2) (May), 355-362. Isard, Peter and Steven Symansky, 1996, "Long Run Movements in Real Exchange Rates," Chapter 2 in Takatoshi Ito, Peter Isard, Steven Symansky and Tamim Bayoumi, Exchange Rate Movements and Their Impact on Trade and Investment in the A PEC Region, Occasional Paper 145 (Washington, DC: International Monetary Fund, December). Ito, Takatoshi, Peter Isard and Steven Symansky, 1996, "Economic Growth and Real Exchange Rate: An Overview of the Balassa-Samuelson Hypothesis in Asia," paper presented at the Seventh Annual NBER-EASE Conference "Changes in Exchange Rates in Rapidly Developing Countries: Theory, Practice, and Policy Issues," Hong Kong, June 19-22, 1996. Johansen, Søren, 1988, "Statistical Analysis of Cointegrating Vectors," Journal of Economic Dynamics and Control 12: 231-54. Kakkar, Vikas and Masao Ogaki, 1994, "Real Exchange Rates and Nontradables: A Relative Price Approach," Working Paper 94-06 (Department of Economics, Ohio State University: Columbus). Mark, Nelson, 1995, "Exchange Rates and Fundamentals: Evidence on Long-Horizon Predictability," American Economic Review 85: 201-18. Mark, Nelson and Doo-Yull Choi, 1994, "Real Exchange Rate Prediction over Long Horizons," mimeo (Department of Economics, Ohio State University: Columbus, OH, December). Marston, Richard, 1990, "Systematic Movements in Real Exchange Rates in the G-5: Evidence on the Integration of Internal and External Markets," Journal of Banking and Finance 14(5) (November): 1023-1044. Meese, Richard and Kenneth Singleton, 1982, “On Unit Roots and the Empirical Modelling of Exchange Rates,” Journal of Finance 37: 1029-1035. Obstfeld, Maurice and Kenneth Rogoff, 1996, Foundations of International Macroeconomics (Cambridge: MIT Press). Phylaktis, Kate and Yiannis Kassimatis, 1994, "Does the Real Exchange Rate Follow a Random Walk? The Pacific Basin Perspective," Journal of International Money & Finance 13(4) (August):476-495. Rogoff, Kenneth, 1996, “The Purchasing Power Parity Puzzle,” Journal of Economic Literature 34: 647-88. Samuelson, Paul, 1964, "Theoretical Notes on Trade Problems," Review of Economics and 21

Statistics 46: 145-154. Strauss, Jack, 1996, "The Cointegrating Relationship between Productivity, Real Exchange Rates and Purchasing Power Parity," Journal of Macroeconomics 18(2): 299-313. Strauss, Jack, 1995, "Real Exchange Rates, PPP and the Relative Price of Nontraded Goods," Southern Economic Journal 61(4): 991-1005. Tseng, Wanda and Robert Corker, Financial Liberalization, Money Demand, and Monetary Policy in A sian Countries, Occasional Paper 84, Washington, DC:IMF, 1991. Wu, Hsiu-Ling, 1996, "Testing for the Fundamental Determinants of the Long-Run Real Exchange Rate: The Case of Taiwan," NBER W orking paper #5787 (October).

22

Data Appendix Data for Section 2.2: Interest rates: The US, UK and Japanese 3 month Eurocurrency deposit rates were the arithmetic average of the bid and offer rates in London at close of market, as reported by Bank of America up to October 6, 1986, and Reuters' Information Service thereafter, and recorded by DRI in the DRIFACS database. Where the source is both WFM and DRI, then WFM is source until 1989:10, at which time DRIFACS is source. Country

Source

DRI Code

Description

Variable Name

US

DRI

USD03

3 month Eurodollar rate

IEUS

Indonesia

WFMr

1 month interbank deposit rate

IIN

Japan

WFM,DRI

JAGBDS03 JACD03B,A

3 month Gensaki bond rate 3 month CD rate IJP up 86.09, IJP2 after

IJP IJP2 IJP3

Japan

DRI

JAD03

3 month EuroYen rate

IEJP

Korea

WFMr

MSB until 1991:12; call money thereafter

IKO5

Malaysia

WFMr

3 month interbank deposit rate

IMA

Philippines

WFMr fm IFS

3 mo. T-bills at tender

IPH

Singapore

WFM,WFMr

3 mo. banker's acceptances to Aug.87; 3 month commercial bills thereafter

ISI

Taiwan

WFMr

90 day bankers acceptances

ITI

call money rate until 91:12, BIBOR thereafter.

ITH2

Thailand

Exchange rates: All exchange rates (except those indicated below) are London 3 PM, arithmetic average of bid and offer rates as reported by Barclay's until end of March 1990, at which time the series is no longer recorded by DRIFACS. Thereafter, the London close rate is used, as reported by Reuter's Information Services. A consistent series is not used (i.e., the London close all the way) because the London close series only begin in 1986. The exchange rates for Indonesia, Korea and Thailand were obtained from the International Monetary Fund's International Financial Statistics database, and are London midday rates.

Data for Section 3: Exchange rates: period average: line rf, IMF, International Financial Statistics, November 1996 CD-ROM; and for Taiwan, Bank of China, Financial Statistics, various issues. Consumer price indices: line 64, IMF, International Financial Statistics, November 1996 CDROM; and for Taiwan, Bank of China, Financial Statistics, various issues. Nominal GDP and real GDP: World Bank, W orld Tables, 1995 CD-ROM; and Asian 23

Development Bank, Key Indicators of Developing A sian and Pacific Countries, various issues; and Bank of China, Financial Statistics, various issues. Sectoral output, at factor cost, in nominal and real terms: World Bank, W orld Tables, 1995 CD-ROM; and for Malaysia, and Taiwan, Asian Development Bank, Key Indicators of Developing A sian and Pacific Countries, various issues. Sectoral employment: Asian Development Bank, Key Indicators of Developing A sian and Pacific Countries, various issues; and International Labour Office, Y earbook of Labour Statistics, various issues. Labor productivity: For all Asian LDCs, sectoral output/sectoral employment: "Tradable" = manufacturing; "Nontradable" = other. For US and Japan, indices from OECD International Sectoral Data Base (ISDB). "Tradable" = industry, mining, transportation, agriculture. "Nontradable" = services, construction, government. Alternate manufacturing labor productivity measure: World Bank, W orld Tables, 1995 CDROM. Government spending. Government consumption, from World Bank, W orld Tables 1995 CDROM. (G## is government consumption spending, divided by GDP). Oil prices: IMF, International Financial Statistics, CD-ROM. (LRPOIL is log price of oil, IFS line 76aad, deflated by US CPI). Real exchange rate: (Q##) Nominal exchange rate adjusted by aggregate GDP deflators. (Q1##) Nominal exchange rate adjusted by CPIs. "Relative prices": (other sector deflator/manufacturing sector deflator) in log terms.

"Relative productivity": (manufacturing sector productivity/other sector productivity) in log terms.

Data for Sections 2.1 and 4: Data is from IMF, International Financial Statistics, March 1997 CD-ROM, except for data for Taiwan: Bank of China, Financial Statistics, various issues, as recorded in Federal Reserve Bank of San Francisco electronic database.

24

Money is narrow money, IFS line 34, in national currency units. Income is real GDP, IFS line 99b.r, in 1990 national currency units. Malaysia income is proxied by industrial production. Taiwanese GDP, in 1991 New Taiwan $. For Indonesia, and Singapore, GDP data is a centered 4 quarter moving average of annual data. Interpolation is via the formula:

Interest rates are short term, interbank interest rates, IFS line 60b, in decimal form. Indonesian interest rates are long term, from Morgan-Guaranty database. Missing values interpolated using an estimated AR(1). Consumer price index, IFS line 64, 1990 = 100. Producer price index, IFS line 63, 1990 = 100. Indonesian data excludes petroleum prices. Inflation is 4-quarter difference of log(CPI). Relative price variable:

[which is appropriate if α = 0.5, and CPI contains one half nontradables].

25

ENDNOTES 1. A correlation between productivity growth and the real exchange rate can arise even if capital is immobile. See Obstfeld and Rogoff, 1996: 215-16. 2. In fact, it is difficult to detect evidence in support of purely supply-side effects on the real exchange rate in time series data. Rather, one finds that demand side shocks, such as government spending, must be included if one is to find a role for traded/nontraded relative productivity levels (see DeGregorio and Wolf (1994) and Chinn and Johnston (1996) in a panel context). 3. See Engel (1995a) for a contrasting view of the behavior of “tradables” prices. 4. Since these series are plausibly integrated of order one [I(1)], it might be argued that the ADF test should be used to determine statistical significance of the reversion coefficient. On the other hand, there is some evidence of mean reversion in the literature, and further, unit root tests have notoriously low power; hence I use conventional critical values in evaluating the estimates. 5. For the latter, see Engel (1995b) for a survey. 6. This assumption, which is at first glance rather odd, is consistent with the empirical finding that the exchange rate appears to follow a random walk. See Meese and Singleton (1982) for an early study. 7. Frankel (1991) argues that real interest rate parity holds when both the financial and goods markets are integrated; the fact that real interest rate parity holds in the long run implies that the relevant conditions (uncovered interest parity and purchasing power parity) hold in the long run. 8. This section draws upon Chinn (1997c). 9. In order to obtain equation (13) then, the tradable goods market must be integrated, as well as the capital market. Econometrically, one is allowing for short run deviations from both PPP for tradable goods, as well as short run deviations from real interest rate parity. 10. Both equations have been exploited extensively. Equation (11) has been examined most recently by Kakkar and Ogaki (1994) for several exchange rates. Equation (13) has been estimated by Hsieh (1982), Marston (1990), and the references in footnote 2. 11. An alternative interpretation of the correlation between the real exchange rate and income levels, consistent with this specification, is what Samuelson has termed the "Penn effect": higher income levels are associated with greater demand for nontradables such as services, and hence higher relative prices for nontraded goods. 12. In fact, if the aggregate price index is a Cobb-Douglas function of tradable and nontradable prices, then the equation using aggregate productivity can only represent a cointegrating relation if the cointegrating coefficients are allowed to time-vary nonlinearly with aT /aN .

26

13. There are at most 22 observations. If one adjusts the critical values for the loss of degrees of freedom due to estimating the parameters, then one cannot in general reject the null hypothesis of no cointegrating vectors. See Cheung and Lai (1993). 14. An alternative interpretation is that the Taiwanese deflators are very mis-measured. Young (1995: 662-63) notes the unusual manner in which Taiwan accounts for public sector employee productivity. On statistical grounds, the failure to find cointegration is also not surprising since US-Taiwan relative productivity appears to be I(1) according to ADF tests. 15. Standard econometric practice is to include deterministic time trends in such regressions; inclusion of such time trends in these regressions usually drastically reduces the statistical significance of the point estimates on the relative price variable. See Zivot (1996) for conditions where this restriction is appropriate 16. In the case of the Philippines, I report the estimates using the World Bank measure of manufacturing productivity because of uncertainty about the accuracy of the calculated productivity measure (see the Appendix to Chinn, 1997c). Using the measure used in the other regressions raises the estimated rate of reversion and the goodness of fit statistic. Hence I have been conservative in my approach. 17. This section draws upon Chinn (1997a) 18. The calculations of the fundamentals require that some assumption be made regarding the income elasticity of narrow money demand. I assume that this is unity for the US. For Indonesia, Korea, Malaysia, the Philippines, Singapore and Thailand, this parameter takes on the values of 1.2, 0.8, 1.1, 0.7, 0.9 and 1.0, respectively (see Tseng and Corker, 1991, Table 1). For Taiwan, I used Dynamic OLS estimates (Stock and Watson, 1993) over the 1981-96 period, which indicate an income elasticity of 1.5. 19. An error correction model, incorporating single lags of first difference terms, and estimated using NLS is used for Indonesia and Japan. Indonesia has too short a sample period to apply the PL estimator, while I assume that for Japan the right hand side variables are weakly exogenous with respect to the exchange rate. 20. Models of imperfect competition and pricing to market may be of relevance. For instance, Ito, Isard and Symansky (1996) have linked machine export ratios to real exchange rate appreciation.

27

Appendix 1 Impact of Alternative Relative Price Measures This appendix discusses the results of using an alternative measure of tradables, which includes both manufacturing and agricultural prices. In the text, the relative price of nontradables, (pN - pT), is measured by the log ratio of the price of “other” to the price of “manufacturing”,

Since a large portion output for several of the Asia-Pacific countries is in agriculture, especially in the early period, a natural concern is that the empirical results are being driven by the neglect of agricultural prices. I calculate an alternative measure of relative prices,

where pAg.+Mfg. is a price index for both agriculture and manufacturing. The weights used are the average proportion of output in each sector over the sample period,

where VA is value added. The results of using this measure of relative prices is reported in Appendix Table A1. They do not differ very much from those reported in Table 3.

28

Appendix Table 1 NLS Error Correction Regressions Real Exchange Rates and Alternative Relative Nontradable/Tradable Prices

Coeff CH

IN

JP

KO

MA

PH

SI

TA

TH

ECT

-0.069 (0.054) (0.154)

-0.218*** (0.059)

-0.396** (0.167)

-0.244† (0.146)

-0.773*** (0.159)

-0.358* (0.181)

-0.431*** (0.096)

-0.186† (0.113)

-0.176

pN-pT

-0.367*** (0.128)

-0.771* (0.429)

-0.191 (0.403)

-0.297 (0.302)

0.494 (0.880)

0.004 (0.180)

1.986† (1.335)

0.987

qt-1

0.054 (0.154)

0.468* (0.240)

0.585** (0.239)

0.282 (0.149)

0.217 (0.272)

0.796*** (0.141)

0.569** (0.212)

0.078

pt-1

0.213 (0.354) (0.269)

1.001*** (0.255)

-0.731 (0.728)

-0.078 (0.303)

-0.229 (0.242)

-0.445 (0.397)

-0.481** (0.213)

0.342† (0.251)

0.210

_ R2 N LM(2) p-val Iter. Smpl 90

.93 21 0.253 [.780] 7 1972-92

.69 21 0.237 [.792] 4 1972-92

.85 21 2.072 [.163] 5 1972-92

.45 12 0.553 [.607] 5 1973-84

.51 21 2.303 [.137] 4 1972-92

.82 21 0.914 [.424] 5 1972-92

.88 17 0.465 [.641] 3 1976-92

.72 19 1.402 [.284] 4 1972-

-6.092† (4.131) (1.433) 0.437* (0.226) (0.331)

.97 20 0.599 [.564] 3 1973-92

Notes: Regression coefficients from Nonlinear Least Squares. †{*}(**)[***] indicates significance at the 20%{10%}(5%)[1%] MSL. "LM(2)" is the Breusch-Godfrey TR2 test for serial correlation of order 2. "N" is the effective number of observations included in the regression. "Iter." is the number of iterations necessary for convergence. "Smpl" is the sample period.

Table 1 Measures of Real Integration

1974.1-85.4 s.e.

1986.1-96.4 s.e.

Indonesia

-0.129†

(0.088)

-0.276**

(0.126)

Japan

-0.184**

(0.081)

-0.240***

(0.083)

Korea

-0.154*

(0.078)

-0.118**

(0.049)

Malaysia

na

na

-0.157

(0.088)

Philippines

-0.345***

(0.126)

-0.144*

(0.080)

Singapore

-0.095†

(0.065)

-0.347***

(0.114)

Taiwan

-0.025

(0.055)

-0.166***1/

(0.075)

Thailand

-0.030

(0.058)

-0.060†

(0.039)

Notes: Coefficients from ADF regression incorporating one difference term. The value - is the rate of reversion; s.e. is standard error. †{*}(**)[***] indicates significance at the 20%{10%}(5%)[1%] MSL using conventional critical values. Bold Italics indicates significant at 10% MSL under null hypothesis of unit root, using MacKinnon critical values. 1/ Estimate is the coefficient on error correction term, augmented with one lag each of differences in exchange rate and prices.

Table 2 Measures of Financial Integration Panel 2.1: Covered Interest Differentials (in percentage points)

Mean Japan1/ Malaysia Singapore

1982:09-88:04 Mean Abs.

-0.08 1.41 0.47

0.14 1.51 0.49

Mean

1988:05-94:11 Mean Abs.

-0.01 0.01 0.91

0.05 1.28 0.91

Notes: Figures in percentage points, estimated by regressing the end-of-month covered interest differential on a constant. Maturity is 3 months. Source: Chinn and Dooley (1997). 1/ CD-Euroyen differential. Data up to 1986:09 is Gensaki rate. Estimate is from a regression with a dummy for Gensaki series, so that the differential is interpretable as that pertaining to a CD rate.

Panel 2.2: Interest Differentials (in percentage points)

Mean Indonesia Japan1/ Korea2/ Malaysia Philippines Singapore Taiwan Thailand3/

1982:09-88:04 Mean Abs.

-9.24 2.90 -6.27 0.51 -10.47 2.62 2.64 -2.05

9.24 2.90 6.27 2.13 10.47 2.62 2.69 2.51

Mean

1988:05-94:11 Mean Abs.

-10.94 1.17 -7.97 -0.20 -11.08 2.02 -1.04 -2.85

10.94 1.80 7.97 2.99 11.08 2.02 2.74 3.01

Notes: Figures in percentage points, obtained by regressing the US Eurodollar minus local interest differentials on a constant. Maturity is 3 months. Source: Chinn and Dooley (1997). 1/ Euroyen-CD differential. Data up to 1986:09 is Gensaki rate. Estimate is from a regression with a dummy for Gensaki series, so that the differential is interpretable as that pertaining to a CD rate. 2/ MSB until 1991:12; call money thereafter. Estimate is from a regression with a dummy for call money series, so that the differential is interpretable as that pertaining to a MSB rate. 3/ Call money rate until 1991:12; BIBOR thereafter. Estimate is from a regression with a dummy for BIBOR series, so that the differential is interpretable as that pertaining to a call money rate.

Table 3 Changes in Real Exchange Rates, Relative Prices, and Relative Productivity Panel 3.1: Changes in Real Exchange Rates and Relative Prices

Country

Inverse Real Exchange Rate (- q)

Relative Prices p

Obs. Sample

China Indonesia Japan Korea Malaysia Philippines Singapore Taiwan Thailand

0.0569 0.0018 -.03484 -.02231 -.02144 -.00039 -.01505 -.01135 0.0035

-0.0216 0.11517 -0.0265 -0.0378 0.00680 0.00584 0.01756 -0.0093 0.00583

20 20 21 21 12 21 21 17 20

1971-91 1970-90 1970-91 1970-91 1971-83 1970-91 1970-91 1974-91 1970-90

Panel 3.2: Changes in Real Exchange Rates and Relative Productivities

Country

Inverse Real Exchange Rate (- q)

Relative Prod. a

Obs. Sample

China Indonesia Japan Korea Malaysia Philippines Singapore Taiwan Thailand

0.0763 0.0479 -.03484 -.02419 0.0014 0.0262 -.00108 -.01135 0.0142

0.01118 0.08165 -0.0049 -0.0125 0.00877 0.02058 0.03513 0.02898 0.02962

11 14 21 20 18 12 18 17 13

1978-89 1977-90 1970-91 1970-90 1971-90 1977-78, 80-91 1973-91 1974-91 1977-90

Notes: Annual change, expressed in log differences. Real exchange rate expressed as ln(1/Q) or as -q.

Table 4 NLS Error Correction Regressions Real Exchange Rates and Relative Nontradable/Tradable Prices

Coeff CH

IN

JP

KO

MA

PH

SI

TA

TH

ECT

-0.024 (0.058)

-0.259*** (0.070)

-0.404** (0.166)

-0.523*** (0.171)

-0.699*** (0.153)

-0.488* (0.273)

-0.433*** (0.100)

-0.184† (0.110)

-0.286 (0.191)

pN-pT

-0.575 (9.183)

-0.464*** (0.120)

-0.761* (0.391)

-0.280** (0.104)

-0.703 (0.387)

-0.407 (0.663)

-0.010 (0.171)

2.203 (1.423)

-0.281 (0.773)

qt-1

0.467* (0.236)

0.469* (0.237)

0.556** (0.226)

0.282 (0.149)

0.286 (0.284)

0.776 (0.144)

0.545** (0.206)

0.452 (0.299)

-0.372 (0.433)

-0.405 (0.212)

0.292 (0.215)

-0.130 (0.301)

.51 21 5.296 [.071] 4 1972-92

.80 21 2.348 [.309] 4 1972-92

.88 17 2.556 [.279] 4 1976-92

.70 19 0.735 [.692] 3 1972-90

qt-2

0.476* (0.251)

pt-1

0.632** (0.233)

pt-2

0.486 (0.194)

_ R2 N LM(2) p-val Iter. Smpl

.97 20 3.167 [.205] 1 1973-92

.94 20 0.824 [.662] 1 1973-92

-0.798 (0.690)

-0.207 (0.288)

.70 21 0.986 [.611] 4 1972-92

.85 20 5.150 [.080] 1 1973-92

.55 12 2.571 [.277] 1 1973-84

Notes: Regression coefficients from Nonlinear Least Squares. †{*}(**)[***] indicates significance at the 20%{10%}(5%)[1%] MSL. "LM(2)" is the Breusch-Godfrey TR2 test for serial correlation of order 2. "N" is the effective number of observations included in the regression. "Iter." is the number of iterations necessary for convergence. "Smpl" is the sample period.

Table 5 Kremer-Ericsson-Dolado Error Correction Regressions Real Exchange Rates and Relative Tradable/Nontradable Productivity Levels

Coeff CH

IN

JP

KO

MA

PH1/

SI2/

TA2/

TH

ECT

-0.059 (0.080)

-0.071 (0.135)

-0.260** (0.120)

-0.050 (0.146)

-0.302* (0.157)

-0.258* (0.123)

0.000 (0.167)

-0.244† (0.172)

-0.183 (0.171)

aT-aN

-0.5

-0.5

-0.5

-0.5

-0.5

-0.5

-142.109 (57848.94)

1.309† (0.796)

-0.5

0.443† (0.290)

0.150 (0.279)

0.310† (0.209)

0.407† (0.289)

0.412* (0.235)

0.243 (0.307)

0.528*** (0.230)

0.643** (0.248)

0.239 (0.300)

qt-1 qt-2

-0.351 (0.295)

at-1

0.348 (0.249)

0.605† (0.391)

-0.261† (0.161)

at-2 _ R2 .94 N 12 LM(1) 0.177 [.674] Iter. -Smpl 1979-90

-0.218 (0.177)

-0.289** (0.124) .88 15 0.195 [.659] -1978-92

.69 21 0.046 [.830] -1972-92

.85 20 1.746 [.186] -1972-91

.57 17 0.428 [.513] -1973-89

.16 14 0.177 [.674] -1978-92

.82 17 2.045 [.153] 3 1976-92

.87 16 0.095 [.758] 5 1977-92

.76 13 0.150 [.700] -1978-90

Notes: Regression coefficients from Nonlinear Least Squares. †{*}(**)[***] indicates significance at the 20%{10%}(5%)[1%] MSL. LM(1) is the Breusch-Godfrey TR2 test for serial correlation of order 1 [p-values in brackets]. N is the effective number of observations included in the regression. Iter. is the number of iterations necessary for convergence. Sample is the sample period. 1/ Philippines regression uses World Bank measure of manufacturing productivity. See text. 2/ Estimated using NLS.

Table 6 Kremer-Ericsson-Dolado Error Correction Regressions Real Exchange Rates, Relative Tradable/Nontradable Productivity Levels, Government Spending Changes and Oil Prices

Coeff

IN

JP

KO

MA

PH1/

SI

TA

TH

Method

OLS

OLS

OLS

OLS

OLS

NLS

OLS

OLS

ECT

-0.601** (0.260)

-0.216** (0.100)

-0.130† (0.098)

-0.077 (0.122)

-0.173 (0.133)

-0.128 (0.179)

0.032 (0.071)

0.087 (0.372)

aT-aN

-0.347** (0.152)

-0.5

-0.5

-0.5

-0.5

0.310 (1.004)

-0.5

-0.5

poil

0.386*** (0.096)

0.381† (0.275)

--

-0.369 (0.298)

qt-1

0.393† (0.255)

0.401* (0.229)

0.665*** (0.229)

0.458*** (0.157)

0.535* (0.257)

0.899*** (0.227)

0.095 (0.457)

at-1

0.208* (0.105)

0.818** (0.298)

-0.168† (0.123)

poilt

--

--

-0.092**2/ (0.040)

-0.205*** (0.071)

-0.106*** (0.037)

-0.092**1/ (0.040)

gt

-1.071† (0.660)

0.409 (0.728)

-1.287† (0.919)

0.168 (0.422)

0.084 (0.429)

-0.217 (0.966)

g*t

0.216 (0.221)

-0.767*** (0.195)

0.116 (0.293)

-0.191† (0.131)

-1.074*** (0.322)

-0.512 (0.561)

.89 20 2.576 [.108] -1972-91

.81 17 0.001 [.972] -1973-89

.20 14 0.879 [.348] -1978-92

.79 18 1.143 [.285] 3 1975-92

.91 17 0.349 [.554] -1976-92

.73 13 0.012 [.913] -1978-90

_ R2 N LM(1) Iter. Smpl

-0.210*** (0.060)

.914 14 0.242 [.623] 6 1979-92

.79 21 1.248 [.264] 6 1972-92

Notes: Regression coefficients from Nonlinear Least Squares. †{*}(**)[***] indicates significance at the 20%{10%}(5%)[1%] MSL. LM(1) is the Breusch-Godfrey TR2 test for serial correlation of order 1 [p-values in brackets]. N is the effective number of observations included in the regression. Iter. is the number of iterations necessary for convergence. Sample is the sample period. 1/ Philippines regression uses World Bank measure of manufacturing productivity. See text. 2/ Oil price change lagged once.

Table 7 Phillips-Loretan Error Correction Regressions Nominal Exchange Rates, Monetary and Real Factors

Coeff

pred

IN1/

JP1/

KO

MA

PH

SI

TI

TH

ECT

(-)

-0.071 (0.255)

-0.221*** (0.064)

-0.132** (0.057)

-0.419** (0.151)

-0.112 (0.167)

-0.148† (0.091)

0.320*** (0.100)

-0.777*** (0.250)

m-m*

(+)

0.293 (0.987)

1.554*** (0.516)

0.636* (0.364)

0.025 (0.215)

0.457 (0.687)

-1.222† (0.782)

-0.437*** (0.140)

0.069 (0.094)

y-y*

(-)

-3.712 (14.232)

-0.001 (0.102)

-0.391† (0.304)

-0.092 (0.467)

3.993 (8.544)

1.830 (1.658)

-0.013 (0.281)

-0.081* (0.044)

i-i*

(-)

1.980 (6.934)

-2.190† (1.544)

1.220 (1.396)

0.206 (1.045)

3.359 (5.480)

-4.939† (3.582)

-0.032 (1.359)

0.007 (0.366)

- *

(+)

0.916 (4.187)

0.605 (2.133)

0.331† (1.308)

-1.324 (1.495)

-3.330 (7.507)

-2.200 (2.318)

-0.930 (0.861)

0.160 (0.457)

(+)

1.115 (6.127)

3.546*** (0.636)

2.803* (1.685)

0.623 (1.037)

3.375 (4.360)

-0.365 (0.634)

0.552 (0.777)

-0.047 (0.200)

(?)

-0.0027† (0.0017)

.981 83 74.3-95.1 1.842 [.029] 54

.992 58 81.3-95.4 2.018 [.119] 124

.804 41 85.1-95.1 0.861 [.517] 82

_ R2 N Smpl LM(4) Iter.

.998 27 86.4-93.2 0.952 [.495] 123

0.0031*** (0.0012) .975 91 73.4-96.2 0.820 [.517] 49

.976 62 80.3-95.4 2.127 [.100] 52

.885 41 84.3-94.3 2.116 [.147] 16

.990 45 81.2-95.4 0.387 [.815] 69

Notes: Regression coefficients from Nonlinear Least Squares. “Pred” is predicted sign. †{*}(**)[***] indicates significance at the 20%{10%}(5%)[1%] MSL. LM(4) is the Breusch-Godfrey TR2 test for serial correlation of order 1 [p-values in brackets]. N is the effective number of observations included in the regression. Iter. is the number of iterations necessary for convergence. Sample is the sample period. 1/ Estimated using ECM(1).

1.5

2.0 Real Exchange Rate (Yuan/$) 1.5

Relative Price

1.0 1.0 0.5

0.5 0.0

0.0

Real Exchange Rate (Rupiah/$)

Relative Price -0.5

-0.5 70 72 74 76 78 80 82 84 86 88 90 92 ZIQCH

-1.0 70 72 74 76 78 80 82 84 86 88 90 92

ZRELPTNCH

Figure 1: China

ZIQIN

ZRELPTNIN

Figure 2: Indonesia

0.4

0.2 Real Exchange Rate (Yen/$)

0.2

Real Exchange Rate (Won/$)

0.0 -0.2

0.0 -0.4 -0.2 -0.6 -0.4 Relative Price -0.6 70 72 74 76 78 80 82 84 86 88 90 92 ZIQJP

-1.0 70 72 74 76 78 80 82 84 86 88 90 92 ZIQKO

ZRELPTNKO

Figure 4: Korea

0.4

0.2

-0.8

ZRELPTNJP

Figure 3: Japan

Relative Price

0.4 Real Exchange Rate (Ringgit/$)

0.3

Real Exchange Rate (Peso/$)

0.2 0.0

0.1 Relative Price

0.0

-0.2

Relative Price

-0.1 -0.4 70 72 74 76 78 80 82 84 86 88 90 92 ZIQMA

ZRELPTNMA

Figure 5: Malaysia

-0.2 70 72 74 76 78 80 82 84 86 88 90 92 ZIQPH

ZRELPTNPH

Figure 6: Philippines

0.4

0.2

0.3 0.2

0.1 Real Exchange Rate (Singapore $/$)

0.0

0.1

Real Exchange Rate (NT$/$)

-0.1 Relative Price

0.0

-0.2

-0.1

-0.3

-0.2 70 72 74 76 78 80 82 84 86 88 90 92

-0.4 70 72 74 76 78 80 82 84 86 88 90 92

ZIQSI

ZRELPTNSI

Figure 7: Singapore

ZIQTA

ZRELPTNTA

Figure 8: Taiwan

0.4

0.2

Relative Price

1.2 1.0

Real Exchange Rate (Baht/$)

Real Exchange Rate (Yuan/$)

0.8 0.6

0.0 Relative Price

0.4 Relative Productivity (1978 = 0)

0.2

-0.2

0.0 -0.4 70 72 74 76 78 80 82 84 86 88 90 92 ZIQTH

-0.2 70

72

74

76

ZRELPTNTH

Figure 9: Thailand

78

80

82

ZIQCH

84

86

88

90

92

ZRELA3CH

Figure 10: China

1.2

0.4

1.0 0.8

0.2

Relative Productivity (1977 = 0)

Relative Productivity (1974 = 0) 0.0

0.6 0.4

-0.2

Real Exchange Rate (Rupiah/$)

0.2

Real Exchange Rate (Yen/$)

-0.4

0.0 -0.2 70 72 74 76 78 80 82 84 86 88 90 92 ZIQIN

ZRELA3IN

Figure 11: Indonesia

-0.6 70

72

74

76

78

80

82

ZIQJP

Figure 12: Japan

84

86

88

ZRELA3JP

90

92

0.2

0.0

0.4

Real Exchange Rate (Won/$)

0.3

Relative Productivity (1974 = 0)

0.2 -0.2 0.1 -0.4

0.0

Relative Productivity (1974 = 0) -0.6 70 72 74 76 78 80 82 84 86 88 90 92 ZIQKO

-0.1 70 72 74 76 78 80 82 84 86 88 90 92

ZRELA3KO

Figure 13: Korea

Real Exchange Rate (Ringgit/$)

ZIQMA

ZRELA3MA

Figure 14: Malaysia

0.4

0.4 Relative Productivity (1977 = 0)

0.3

Real Exchange Rate (Singapore $/$) 0.2

0.2 0.1

0.0 Relative Productivity (1974 = 0)

0.0 Real Exchange Rate (Peso/$)

-0.1

-0.2 70 72 74 76 78 80 82 84 86 88 90 92 ZIQPH

-0.4 70 72 74 76 78 80 82 84 86 88 90 92

ZRELA3PH

Figure 15: Philippines

0.6

0.4

-0.2

ZIQSI

Figure 16: Singapore

0.4 Relative Productivity (1974 = 0)

0.3

0.2

0.2

0.0

0.1

-0.2

ZRELA3SI

Real Exchange Rate (NT $/$)

-0.4 70 72 74 76 78 80 82 84 86 88 90 92 ZIQTA

Figure 17: Taiwan

ZRELA3TA

Relative Productivity (1977 = 0)

Real Exchange Rate (Baht/$)

0.0

-0.1 70 72 74 76 78 80 82 84 86 88 90 92 ZIQTH

ZRELA3TH

Figure 18: Thailand

0.02

TI

DLX94

0.01

SI

KO TH

0.00 MA

PH -0.01 IN -0.02 -0.03

-0.02

-0.01

0.00

0.01

DFUN2194 Figure 19: Average change in nominal exchange rates against average change in fundamentals, 1985.1-1994.4