Purchasing Power Parity: Basic Concepts
An important determinant of a country’s external payments position is its export competitiveness. The most common approach to the analysis of competitiveness involves a comparison of trends in exchange rates and prices, based on the concept of purchasing power parity (PPP).6 This approach assumes that equilibrium real exchange rates remain constant over time and therefore that nominal exchange rate movements tend to offset relative price movements. Three versions of PPP have traditionally been used in the literature: the law of one price, which relates exchange rates to prices of individual, homogeneous goods in different countries; absolute PPP, which relates exchange rates to overall price levels; and relative PPP, which relates exchange rate changes to inflation rates.
The law of one price states that, assuming there are no transaction costs or trade barriers (such as tariffs or quotas), the prices of identical goods sold in different countries should be the same when expressed in a common currency. Empirically, the law of one price appears to hold quite well for homogeneous primary commodities traded on major organized exchanges, when appropriate adjustments for contract differences and delivery lags are made.7 By contrast, and not surprisingly, differentiated products (such as manufactured goods and services) subject to international competition tend to deviate somewhat from the law of one price, at least over the short and medium run.8
Absolute PPP extends the law of one price to the general price level; under the same assumptions as the law of one price, the same basket of goods and services should cost the same amount in all countries when expressed in a common currency. If the law of one price holds for every good, then absolute PPP between similar baskets of goods should also hold. Absolute PPP, however, requires the parity relationship to hold only on average for all goods— not strictly for each good.
Absolute PPP provides a specific equilibrium concept for the nominal exchange rate, namely, the PPP exchange rate. This is defined as the rate that equalizes the prices of a common basket of goods in two different countries. When the market exchange rate differs from the PPP exchange rate, profits can be made by purchasing goods in one country and selling them in another, a situation that will tend to push the exchange rate back toward its equilibrium value. Such arbitrage in traded goods is a mechanism that in principle anchors the nominal exchange rate and keeps departures of the actual real exchange rate from its PPP level relatively short lived.9
Although the law of one price and its multi-good counterpart—absolute PPP—are intuitively appealing, their usefulness as guides to exchange rate behavior is limited. Transportation and information costs, as well as institutional impediments to trade such as tariffs and quotas, limit the response of consumers and firms to cross-country price differences, even large ones, and thus prevent absolute price levels from being equalized. Relative PPP is an even weaker condition than absolute PPP and assumes only that the rate of change in the nominal exchange rate will be equal to the difference between the domestic and foreign rates of inflation on equivalent baskets of goods.10 If PPP is judged to hold in some particular year, one can obtain from a time series of the inflation differential the implied PPP nominal exchange rate, which can then be compared to the market exchange rate.
Causes and Effects of Deviations from PPP
Most economic theories suggest that PPP should hold in the long run for traded goods, assuming that measurement problems involved in constructing comparable price deflators can be resolved (see the discussion below). Evidence from empirical tests of the validity of PPP is mixed, however. On the one hand, a growing amount of evidence suggests that PPP does represent a reasonable characterization for long-run movements in industrial country nominal exchange rates and traded goods’ prices. For example, Lee (1976), Hakkio (1992), and Lothian and Taylor (1993) have examined exchange rates and prices of traded goods dating back to 1900 and have detected a tendency for the exchange rates of major currencies to revert toward their PPP value over horizons of 3 to 12 years. These findings, however, contrast with other evidence indicating that PPP does not explain the behavior of exchange rates in the short run (see, for instance, Frenkel (1978) and Krugman (1978)). When exchange rates and price movements are compared between industrial and developing countries—be it in the long run or the short run—PPP is clearly rejected.11
In sum, the evidence as a whole indicates that PPP is not a good guide for short- and medium-run exchange rate behavior. There are four main reasons for this: hysteresis effects due to adjustment costs in trade; price rigidities in terms of the currency in which the goods are sold; imperfectly substitutable traded goods; and structural changes in technology and demand, particularly between traded and nontraded goods and services. The first two factors explain why PPP can fail over short horizons, while the last two can be associated with persistent changes in equilibrium real exchange rates.
Temporary deviations from PPP are generated in hysteresis models of trade.12 These models emphasize that trade flows may not fully respond in the short run to a change in the real exchange rate because of the presence of adjustment costs. These costs include marketing new products in a foreign country, expanding or reducing existing production or distribution lines, and altering brand recognition.13 When such adjustment costs are present, only large departures from PPP will trigger a response on the part of exporters that will act to correct the initial relative price change.
A second explanation for temporary deviations from PPP is provided by sticky price models of exchange rates (see Dornbusch (1976)). While originally aimed at explaining short-term volatility of nominal exchange rates, this class of models has gained status as a broad interpretative tool for gauging the effects of monetary policy in a world with sticky prices. This approach recognizes that prices react sluggishly in the short run to unanticipated changes in monetary conditions, thus inducing a compensatory overshooting response in the nominal exchange rate. For example, in response to a permanent increase in the domestic money supply, the nominal exchange rate would first depreciate by jumping to a value lower than its long-run equilibrium level, and then gradually appreciate toward that equilibrium. Overshooting of the exchange rate relative to its equilibrium value results from differences in the speed of adjustment between financial markets and goods markets; the former respond swiftly to exogenous shocks, while the latter adjust over the medium run.14
A third factor generating departures from PPP is imperfect substitutability among traded goods. The existence of differentiated products implies that differences in growth rates and in income elasticities across countries can become important determinants of long-run trends in real exchange rates.15 If, for example, world demand for the goods produced by the home country declines, the prices of these goods will tend to fall and the equilibrium real exchange rate will tend to depreciate. It would be misleading, however, to regard this real depreciation as an improvement in international competitiveness, since both the current real exchange rate and its equilibrium value have declined by the same amount. Thus, it is not always easy to distinguish empirically when a change in the real exchange rate reflects a change in competitiveness or a structural change that leads to a fall in the equilibrium real exchange rate.
A different rate of technological change in the traded vs. nontraded sectors has long been recognized (Balassa (1964)) as a cause of sustained movements in equilibrium real exchange rates and, therefore, as a reason for the failure of PPP to hold. In a world with traded and nontraded goods and services, demand and supply for tradables are brought into equilibrium on a worldwide basis, while demand and supply for nontradables are equated in each country’s domestic market.16 While market competition may keep prices of tradables broadly aligned across countries, prices of nontradables in different countries need not move together. Thus, if the real exchange rate between two countries is computed using deflators that include both tradables and nontradables, such as the consumer price index (CPI) or the gross domestic product (GDP) deflator, then countries with faster growth of productivity in the traded (notably manufacturing) sector than in the nontraded (notably service) sector will exhibit a tendency toward real appreciation.
This phenomenon is often advanced as an important factor underlying the secular (long-term) appreciation of industrial countries’ real exchange rates relative to developing countries.17 Even among industrial countries, different trends in productivity among tradables and nontradables can be significant in explaining movements in equilibrium real exchange rates over horizons that are relevant for policy consideration. Japan is often cited as the classic case of a country where faster growth of productivity in the traded vs. nontraded goods sector, relative to its trading partners, has been associated with a real appreciation of a currency. Chart 1, which displays real effective exchange rates of the yen for both a composite tradable-nontradable index (the CPI) and a tradable-only index (the export unit value index) over 1971-93, highlights this phenomenon. What is noteworthy is the significant real appreciation of the yen in terms of the composite price index and the contrasting approximate constancy for the tradables-only price index.18
Japan: Real Effective Exchange Rates
(1985 = 100)
Sources: IMF, International Financial Statistics, various issues, and staff estimates.1 Does not include data on China from 1975 to 1980.Japan: Real Effective Exchange Rates
(1985 = 100)
Sources: IMF, International Financial Statistics, various issues, and staff estimates.1 Does not include data on China from 1975 to 1980.Japan: Real Effective Exchange Rates
(1985 = 100)
Sources: IMF, International Financial Statistics, various issues, and staff estimates.1 Does not include data on China from 1975 to 1980.Thus, in a world where structural changes lead to shifts of equilibrium real exchange rates, a country’s competitiveness can be fully assessed only by explicit consideration of the factors that drive such changes. See Section III below for an examination of these fundamental economic factors and their effects on equilibrium real exchange rates.
Real Exchange Rate Indicators and Competitiveness
In our context, the main purpose of an analysis of competitiveness is to assess how exchange rate changes affect a country’s external position by examining their relationship to the incentives faced by consumers and investors in their consumption and production decisions. Several measures of real exchange rates have been considered, at the IMF and elsewhere, as useful indicators of competitiveness.19
As hinted at earlier in this section, a real exchange rate index based on the prices of tradable goods could be a useful indicator for assessing changes in a country’s external competitiveness. For exports, such an indicator is defined as the ratio of a weighted average of the home country’s prices of exported goods to a weighted average of its partner countries’ prices of exported goods, all expressed in the same currency. A rise in this indicator would signal a loss of competitiveness for the home country in its export markets.
The main practical shortcoming of an export-based index of competitiveness is that it may be subject to a sampling bias. Specifically, such an indicator does not include all potentially exportable goods; rather it covers only those tradable goods that are priced sufficiently low—at the current exchange rate—to be exported. Although sampling biases may also affect other indicators (for example, a good may be priced too high to be consumed in a particular country and will thus be excluded from the CPI), export deflators are more narrowly defined than are other price indices, and thus are more susceptible to this problem. More important, if traded goods are close substitutes, then the export-based real exchange rate is likely to show little variation; Chart 1 shows that this has been the case in Japan. This is because exporters of homogeneous goods will adjust prices quickly in response to exchange rate changes—even though the relative profitability of producing the goods included in the export basket may vary considerably. Exporters can “price to market” by squeezing profits in the short run, but this strategy is not sustainable in the long run. An export-based index of competitiveness may therefore provide little information on the relative profitability of domestic vs. foreign traded goods, and therefore on the incentives to shift resources between these sectors.
More comprehensive measures of price competitiveness can be obtained by constructing real exchange rates based either on aggregate price deflators or on unit labor costs. For example, because a GDP-based real exchange rate reflects the ratio of the relative prices of nontraded to traded goods at home and abroad, an increase in this index would reflect either a loss of competitiveness in the traded goods market or a greater incentive to allocate resources to the non-traded goods sector at home. Thus, movements in this indicator would usefully reflect movements in important determinants of trade flows and real exchange rate pressure. In addition to GDP deflators, other, broader measures include wholesale prices, value-added deflators, and consumer price indices (CPIs). The main practical disadvantage of wholesale and value-added measures is the lack of cross-country comparability with regard to both concept and commodity composition; also, they are typically available only for the manufacturing sector, and often with a substantial delay. CPIs are available on a more timely basis and with greater frequency but reflect taxes and other institutional distortions, as well as prices of imported goods; this latter feature makes CPI-based measures less indicative of the prices faced by producers.
A real exchange rate index defined in terms of relative unit labor costs (ULC) in the traded goods sector (typically manufacturing) compares the relative profitability of nonlabor factors in producing manufactured goods at home and abroad. Implicit in this measure is the notion that the real exchange rate operates to equilibrate the rate of return of nonlabor factors across countries. In this respect, unit labor cost indicators are useful for assessing competitiveness in terms of incentives to shift non-labor factors domestically and internationally. They also have the advantage of being defined similarly across industrial countries.
Again, however, there are drawbacks. First, because they are defined in terms of unit (or average) labor costs, such measures provide only a rough approximation of the relative incentives for labor allocation at home and abroad, which should be measured by marginal labor costs. Because the relationship between marginal and average labor costs varies with capital/output ratios, changes in unit labor costs may be signaling only changes in capital/output ratios that are unrelated to competitiveness. For example, a rapid shift of resources toward more capital-intensive industries may lead a country’s ULC-based real exchange rate to over-estimate its competitiveness gains. A good case in point is the Irish experience, summarized in Chart 2, in which the rapid inflow of capital-intensive multinational enterprises during the 1980s led to a substantial decline in the ULC-based indicator. The decline, however, reflected neither lower costs at the firm level nor an exchange rate depreciation, but rather to a significant degree a shift in the relative factor content of output.
Ireland: Real Effective Exchange Rates
(1985 = 100)
Sources: IMF, International Financial Statistics, various issues, and staff estimates.Ireland: Real Effective Exchange Rates
(1985 = 100)
Sources: IMF, International Financial Statistics, various issues, and staff estimates.Ireland: Real Effective Exchange Rates
(1985 = 100)
Sources: IMF, International Financial Statistics, various issues, and staff estimates.A second problem associated with the use of ULC-based indicators is that nonlabor costs in production include not only the remuneration to capital but also other costs, such as the rental of land and the cost of intermediate goods and primary commodities. As shown by Lipschitz and McDonald (1991), a suitable correction of the labor cost index for the share of value added domestically can be made in principle but would be difficult to carry out in practice. Furthermore, two practical shortcomings of ULC-based indicators are the relatively long delay in their availability and the likelihood of relatively large errors in the measurement of their components.
Finally, an attempt to forecast future changes in competitiveness, rather than simply assess past developments, underlies the so-called structural forecasting models of exchange rates. These models typically try to predict future movements of exchange rates by using a set of variables derived from such models of exchange rates as the monetary and the portfolio balance approaches.20 In general, the empirical success of these approaches has been limited, and the general perception of this effort has changed little since the negative assessment provided by Meese and Rogoff (1983).
Given the empirical difficulties of constructing models of exchange rates that adequately predict future exchange rate movements, greater effort has been placed, at the IMF and elsewhere, in exploiting market-based indicators of anticipated movements in exchange rates and in their fundamental determinants. These indicators typically include forward exchange rates, interest rate differentials, forward interest rates, yield curves, and option-based estimates of future exchange rate volatility. While these indicators have proven useful in certain circumstances in identifying incipient sources of exchange rate pressure (see the discussion below), their empirical performance is not always adequate. Moreover, because these indicators usually estimate only nominal exchange rate movements, they must be supplemented by forecasts of inflation differentials. In sum, use of market-based exchange rate indicators is a promising but as yet rather undeveloped area of research that may prove helpful in the assessment of exchange rates.
Recent Experience with Real Exchange Rate Indicators
Two frameworks for the construction of measures of competitiveness are currently used at the IMF.21 The first framework involves calculation of CPI-based real effective exchange rates (REERs) for almost 140 countries. Trade weights are constructed using data on trade flows in manufacturing and primary (non-oil) commodities. A second framework uses unit labor costs in manufacturing for 17 industrial countries to construct REERs.
Given the availability of a range of indicators of competitiveness, it is of interest to assess the relative usefulness of different measures as predictors of exchange rate pressure and changes in trade volumes. In this connection, Lipschitz and McDonald (1991) have studied the evolution of Germany’s competitiveness during the 1980s vis-à-vis its EMS partners. They show that the ability of ULC-based real exchange rates to predict changes in European trade shares can be enhanced by adjusting this indicator to account for movements of value-added deflators in Germany and in its EMS partners. The study also suggests that the inability of standard GDP-based and CPI-based indicators of competitiveness to predict the pattern of market shares during this period can be explained by the slower rate of productivity growth in Germany’s traded goods sector vis-à-vis its EMS partners. Differential productivity effects in EMS countries are also analyzed by Micossi and Milesi-Ferretti (1993) and by De Gregorio, Giovannini, and Krueger (1993).
A summary of the behavior of different indicators of competitiveness in developing countries is provided by Wickham (1993). Using data for Colombia and Kenya as examples, that study discusses the difficulties encountered in constructing empirical counterparts to various theoretical concepts of real exchange rates. The study notes that despite the differences among the various approaches, different indicators nonetheless end up being fairly similar in construction and behavior; this is particularly the case for developing countries, where large exchange rate movements tend to outweigh the relative volatility of the price deflators.
Recent studies, including several at the IMF, have tried to establish a link between market-based expectations of exchange rate changes and macro-economic fundamentals, including competitiveness.22 Caramazza (1993), for example, finds that investors’ anticipations of a realignment of the French/German ERM parity from 1987 to 1991 can be explained in large part by several macroeconomic variables, including relative inflation rates and export competitiveness. Similarly, Bartolini (1993) studies the relationship between the competitive position of Ireland and market anticipations of a devaluation of its exchange rate from 1987 to 1993. He finds strong links between anticipated devaluations and both a CPI-based measure of competitiveness and a measure of the expected devaluation of the currencies of Ireland’s main trading partners. Rose and Svensson (1993) also find a strong link between inflation differentials and anticipated exchange rate devaluations, although their results are more mixed for a number of other macroeconomic variables. In general, evidence from these and other studies tends to support the predictions of PPP-based theories on the correlation between expected changes in exchange rates and inflation differentials.23
The links between competitiveness and trade volumes are explored by Marsh and Tokarick (1994). They estimate import and export equations for the major industrial countries from 1973 to 1991 at different levels of aggregation using indicators of real exchange rates based on consumer prices, export unit values, and unit labor costs. They find evidence of a long-run response of trade to real exchange rate behavior but little significant linkage in the short run. Most important, the study finds that none of the three indicators uniformly outperforms the others in explaining trade flows. Thus, the main conclusion of this research is that it is difficult to discriminate among these indicators. One reason appears to be that the large volatility of nominal exchange rates tends to swamp the difference in volatility among the individual underlying indicators of competitiveness.24
To shed further light on these issues, we examined the behavior of three of the real exchange rate indices discussed above and of the trade balance in the two largest industrial countries, the United States and Japan, from 1980 to 1993.25 Chart 3 which graphically summarizes this evidence, reveals two main features. First, because of the higher volatility of nominal exchange rates relative to the volatility of prices, the different indicators of competitiveness tend to be highly correlated in the short run. The main exception is the slow response of Japan’s export unit value index to the appreciation of the yen in the second half of the 1980s, which partly reflects the successful efforts of Japanese firms to maintain market share despite unfavorable fluctuations in the yen/dollar rate. The second notable feature of Chart 3 is that whereas for the United States losses of competitiveness are clearly associated with a declining trade balance, the opposite is true for Japan, where competitiveness losses are contemporaneously correlated with improvements in the trade balance. Clearly, however, this finding should not be interpreted as suggesting that changes in competitiveness produce perverse effects on a country’s trade balance.26 Rather, it shows that examining trends in competitiveness alone—independent of structural and cyclical developments in output and demand, changes in policies, and financial market conditions—generally provides only a partial account of developments in external trade.
United States and Japan: Selected Indicators
Sources: IMF, International Financial Statistics, various issues, and staff estimates.United States and Japan: Selected Indicators
Sources: IMF, International Financial Statistics, various issues, and staff estimates.United States and Japan: Selected Indicators
Sources: IMF, International Financial Statistics, various issues, and staff estimates.In summary, while aggregate indicators of competitiveness may not by themselves be strongly correlated with changes in external imbalances, these indicators nonetheless can signal the emergence of market pressure toward adjustment of nominal exchange rates. Relative costs and prices are an important underlying determinant of external positions, and consequently, large changes in competitiveness generate economic forces that move real exchange rates toward their equilibrium levels. Analyzing international competitiveness, however, constitutes only one ingredient in assessing exchange rates because that approach takes the level of equilibrium real exchange rates as given. Since there is little reason to believe that real exchange rates remain constant over time, it is necessary to incorporate the analysis of international competitiveness in a more comprehensive framework that can explain changes in the equilibrium rates themselves. The next section describes an approach that focuses directly on equilibrium real exchange rates and their determinants.
For a lucid analysis of the theory and empirical evidence on PPP, see Dornbusch (1987).
This result has been documented, for example, by Goodwin, Grennes, and Wohlgenant (1990).
See, for example, Isard (1977) and Kravis and Lipsey (1977).
See Chapters 10 and 11 in Officer (1982) and Chapter 4 in Turner and van’t Dack (1993) for comprehensive surveys of the sizable empirical literature on absolute PPP.
See Ohno (1990) for estimates of equilibrium PPP rates for the yen/dollar and the mark/dollar rates. The Organization for Economic Cooperation and Development provides periodic calculations of relative PPP exchange rates; see OECD (1987) for a description of the underlying methodology.
Evidence on PPP among industrial countries is also mixed when real exchange rates are computed on the basis of price deflators that include prices of nontradable goods (such as the CPI and the GDP deflator). For instance, Officer (1982), Abuaf and Jorion (1990), and a recent study by the Bundesbank (Germany, 1993) report evidence in support of PPP among industrial countries using aggregate deflators inclusive of prices of non-traded goods. Evidence to the contrary is reported in the empirical literature following Balassa (1964). The recent study by Mark (1990) provides negative evidence on a CPI-based PPP, while Dueker (1993) reports inconclusive results. Using a sample of 34 industrial and developing countries, Kravis and Lipsey (1983) show that aggregate price deflators tend to be rather uniform among industrial countries, whereas they decline sharply as the stage of development (measured by real per capita income) falls, with the price of nontradables largely explaining the correlation.
Similar to its use in the physical sciences, the term “hysteresis” has been used in economics to describe a phenomenon (such as a trade imbalance) that fails to disappear once the causes that brought it into existence (such as an appreciated real exchange rate) have been removed.
See, for instance, Baldwin (1988), Baldwin and Krugman (1989), and Dixit (1989). For a brief review of the topic, see Krugman (1990).
A broader class of models (the portfolio balance approach) extends the overshooting literature by allowing for imperfect substitutability among different countries’ financial assets. In this type of model, temporary deviations from PPP may also arise. See Dornbusch (1982) for a discussion of both approaches and Artus (1977) for a related approach.
The mere existence of heterogeneous traded goods, however, does not necessarily invalidate PPP as a long-run proposition. Countries facing less elastic import and export demand (such as Japan) have to some extent offset the resulting tendency to real exchange rate appreciation by growing at a faster rate than their trading partners. See Krugman (1989) for a model that provides a rationale for this tendency that is based on recent developments in the analysis of trade in differentiated products.
Because resource allocation is driven by the relative price of traded to nontraded goods, this price will tend to reflect the relative efficiency of production in these two sectors and the relative demands for traded and nontraded goods. For this reason, the relative price of traded to nontraded goods is often referred to as “the” real exchange rate. For consistency with the previous terminology, we shall continue to define the real exchange rate as the price-adjusted nominal exchange rate. As discussed below, the two are closely related when the GDP deflator is used as the price deflator.
See, for example, Kravis and Lipsey (1983). Recognition of this secular trend also underlies the recent use by the IMF of the principle of PPP to calculate relative national outputs. In this approach, goods and services are converted into a common currency not at the current exchange rate, but rather at the PPP rate that equates domestic prices with world prices for each country’s domestic product. See Guide and Schulze-Ghattas (1992) for a description of this methodology, and Annex 4 in the World Economic Outlook (IMF, 1993c) and Chapter 6 in Staff Studies for the World Economic Outlook (IMF, 1993e) for its application to cross-country real income comparisons.
Marston (1987) provides an extensive discussion of this development.
Six series for real effective exchange rates are published by the IMF in International Financial Statistics. These are based on wholesale, consumer, and export prices, value added, unit labor costs, and normalized unit labor costs. See IMF (1984) and Turner and van’t Dack (1993) for a more complete discussion of these measures of competitiveness. Turner and van’t Dack also assess current trends in international competitiveness based on several of the indicators discussed in this paper.
For a discussion of these models, see Frenkel and Mussa (1985), MacDonald and Taylor (1992), and Taylor (forthcoming).
These frameworks are based in part on the model of imperfect substitutes of Armington (1969), and on its empirical implementation by McGuirk (1987) and Wickham (1987).
In these studies, market-based measures of anticipated devaluations of ERM currencies are obtained by correcting interest differentials with an estimate of anticipated movements of exchange rates within their fluctuation bands. See Svensson (1993) for a discussion of the methodology. Isard (1994) provides a theoretical analysis of unemployment as a determinant of expectations of a change in a currency peg in a policy-optimizing framework.
See, for instance, Thomas (1993) and Chen and Giovannini (1993).
In a study of U.S. trade flows, Marquez (1992) finds that unit labor costs and consumer prices dominate some other indicators of competitiveness, but that the empirical results do not discriminate between these two indicators themselves.
A discussion of the ability of alternative competitiveness indicators to explain external imbalances in selected industrial countries from 1963 to 1983 appears in IMF (1984).
In this connection, it should be noted that there is ample evidence that real exchange rates, after taking proper account of other relevant factors, including J-curve effects due to the lagged response of trade volumes, do have a significant effect on trade balances. See, for example, Hooper and Marquez (1993) and Meredith (1993).
