II A Brief Review of the Theoretical and Empirical Literature
- Peter Clark, Shang-Jin Wei, Natalia Tamirisa, Azim Sadikov, and Li Zeng
- Published Date:
- September 2004
Since the appearance of the 1984 IMF study on the effects of exchange rate volatility on trade, two survey papers covering the literature on this topic have appeared, Côté (1994) and McKenzie (1999). In addition, the U.K. Treasury recently commissioned a number of studies (U.K. Treasury, 2003) and invited submissions from numerous academics to provide data for an assessment of the desirability of joining EMU. Therefore, it is not necessary to present a comprehensive discussion of the many contributions to the field. Rather, the focus here will be on certain key issues that explain why it has been difficult to reach clear-cut conclusions on the impact of exchange rate variability on trade flows. The study will also focus on some of the more recent work in the area. The two aforementioned surveys conclude that, from a theoretical perspective, there is no unambiguous response in the level of trade to an increase in exchange rate volatility because differing results can arise from plausible alternative assumptions and modeling strategies. The same ambiguity pervades much of the empirical literature, which may reflect the lack of clear-cut theoretical results as well as the difficulty in arriving at an appropriate proxy for exchange rate risk. Nonetheless, some recent studies, as well as some of the evidence presented here, appear to suggest that the data support a negative relationship.
Theoretical Aspects of the Relationship Between Exchange Rate Volatility and Trade
It is useful to begin with the example of a rudimentary exporting firm to illustrate how (real) exchange rate volatility can affect the level of its exports. The simplest case, described by Clark (1973), considers a competitive firm with no market power, producing only one commodity sold entirely to one foreign market, and that does not import any intermediate inputs. The firm is paid in foreign currency and converts the proceeds of its exports at the current exchange rate, which varies in an unpredictable fashion because there are no hedging possibilities by assumption, such as through forward contracts. Moreover, because of costs involved in adjusting the scale of production, the firm must make timely production decisions in advance of any subsequent exchange rate movements. Therefore, it cannot alter its output in response to favorable or unfavorable shifts in the profitability of its exports arising from these movements. In this situation, the variability in the firm’s profits arises solely from the exchange rate. Also, where the managers of the firm are adverse to risk, greater volatility in the exchange rate—with no change in its average level—leads to a reduction in output, and hence in exports, in order to reduce the exposure to risk. This basic model has been elaborated by a number of authors, such as Hooper and Kohlhagen (1978), who reach the same conclusion of a clearly negative relationship between exchange rate volatility and the level of trade.
This strong conclusion, however, rests on a number of simplifying assumptions. First, it is assumed that there are no hedging possibilities either through the forward exchange market or through offsetting transactions. For advanced economies, where there are well-developed forward markets, specific transactions can be easily hedged, thus reducing exposure to unforeseen movements in exchange rates.5
Moreover, there are numerous possibilities for reducing exposure to the risk of adverse exchange rate fluctuations other than forward currency markets. The key point is that for a multinational firm engaged in a wide variety of trade and financial transactions across a large number of countries, there are manifold opportunities to exploit offsetting movements in currencies and other variables. For example, there is a clear tendency for exchange rates to adjust to differences in inflation rates, and recent evidence suggests that such adjustments may be quicker than indicated by earlier studies. Thus, if exports are priced in a foreign currency that is depreciating, the loss to the exporter from the declining exchange rate is at least partly offset by the higher foreign-currency export price (Cushman, 1983 and 1986). In a similar vein, as noted by Clark (1973), to the extent that an exporter imports intermediate inputs from a country whose currency is depreciating, there will be some offset to declining export revenue in the form of lower input costs. In addition, when a firm trades with a large number of countries, the tendency for some exchange rates to move in offsetting directions will provide a degree of protection to its overall exposure to currency risk. Finally, as analyzed by Makin (1978), a finance perspective suggests that there are many possibilities for a multinational corporation to hedge foreign currency risks arising from exports and imports by holding a portfolio of assets and liabilities in different currencies.
One reason why trade may be adversely affected by exchange rate volatility stems from the assumption that the firm cannot alter factor inputs in order to adjust optimally to movements in exchange rates. When this assumption is relaxed and firms can adjust one or more factors of production in response to movements in exchange rates, increased variability can in fact create profit opportunities. This situation has been analyzed by Canzoneri and others (1984), De Grauwe (1992), and Gros (1987). The effect of such volatility depends on the interaction of two forces at work. On the one hand, if the firm can adjust inputs to both high and low prices, its expected or average profits will be larger with greater exchange rate variability because it will sell more when the price is high and vice versa. On the other hand, to the extent that there is risk aversion, the higher variance of profits has an adverse effect on the firm and constitutes a disincentive to produce and to export. If risk aversion is relatively low, the positive effect of greater price variability on expected profits outweighs the negative impact of the higher variability of profits, and the firm will raise the average capital stock and the level of output and exports. In a more general setting that analyzes the behavior of a firm under uncertainty, Pindyck (1982) has also shown that, under certain conditions, increased price variability can result in increased average investment and output as the firm adjusts to take advantage of high prices and to minimize the impact of low prices.
One aspect of the relationship between trade and exchange rate volatility that must be mentioned is the role of “sunk costs.” Much of international trade consists of differentiated manufactured goods that typically require significant investment by firms to adapt their products to foreign markets, to set up marketing and distribution networks, and to set up production facilities specifically designed for export markets. These sunk costs would make firms less responsive to short-run movements in the exchange rate because they would tend to adopt a wait-and-see approach, stay in the export market as long as they can recover their variable costs, and wait for a turnaround in the exchange rate to recoup their sunk costs. Following the finance literature on real options, McDonald and Siegel (1986), Dixit (1989), and Krugman (1989) have explored the implications of sunk costs in the context of an “options” approach, which has been applied by Franke (1991) and Sercu and Vanhulle (1992). The key idea is that an exporting firm can be viewed as owning an option to leave the export market, and a firm not currently exporting can be regarded as owning an option to enter the foreign market in the future. The decision to enter or exit the export market involves considering explicit fixed and variable costs—but also the cost of exercising the option to enter or leave the market. The greater the volatility in exchange rates, the greater the value of keeping the option; hence the larger the range of exchange rates within which the firm delays action by staying in the export market or staying out if it has not yet entered. This suggests that increased exchange rate volatility would increase the inertia in entry and exit decisions.
In most theoretical models, that which is being studied is the volatility of the real exchange rate, as opposed to that of the nominal exchange rate. The two are distinct conceptually but do not differ much in reality: prices of goods tend to be sticky in local currency in the short-to-medium run. In this case, real and nominal exchange rate volatilities are virtually the same for practical purposes. For this reason, after reviewing the literature on the effect of real exchange rate volatility, this study does not present a separate discussion on the effect of nominal exchange rate volatility. The exceptions are episodes of high inflation, when nominal exchange rate volatility tends to be bigger than real exchange rate volatility. For this reason, the empirical analysis that will be presented later examines explicitly whether or not real versus nominal exchange rate volatilities have different effects on trade.
Up to this point, the discussion of the impact of volatility on trade has been within a partial equilibrium framework, i.e., the only variable that changes is some measure of the variability of the exchange rate; all other factors that may have an influence on the level of trade are assumed to remain unchanged. Those developments that are generating the exchange rate movements, however, are likely to affect other aspects of the economic environment, which will in turn have an effect on trade flows. Thus, in a general equilibrium framework it is important to take account of the interaction of all the major macroeconomic variables to get a more complete picture of the relationship between exchange rate variability and trade.
Such an analysis has been provided recently by Bacchetta and Van Wincoop (2000). They develop a simple, two-country general equilibrium model, where uncertainty arises from monetary, fiscal, and technology shocks, and they compare the level of trade and welfare for fixed and floating exchange rate arrangements. They reach two main conclusions. First, there is no clear relationship between the level of trade and the type of exchange rate arrangement. Depending on the preferences of consumers regarding the tradeoff between consumption and leisure, as well as the monetary policy rules followed in each system, trade can be higher or lower under either exchange rate arrangement. As an example of the ambiguity of the relationship between volatility and trade in a general equilibrium environment, a monetary expansion in a country would depreciate its exchange rate causing it to reduce its imports, but the increased demand generated by the monetary expansion could offset part or all of the exchange rate effect. Thus, the nature of the shock that causes the exchange rate change can lead to changes in other macroeconomic variables that offset the impact of the movement in the exchange rate. Second, the level of trade does not provide a good index of the level of welfare in a country, and thus there is no one-to-one relationship between levels of trade and welfare in comparing exchange rate systems. In their model, the authors determine trade by the certainty equivalent of a firm’s revenue and costs in the home market relative to the foreign market, whereas the welfare of the country is determined by the volatility of consumption and leisure.
Obstfeld and Rogoff (1998) also provide an analysis of the welfare costs of exchange rate volatility. They extend the “new open-economy macroeconomic model” to an explicitly stochastic environment, where risk has an impact on the price-setting decisions of firms and, hence, on output and international trade flows. They provide an illustrative example whereby reducing the variance of the exchange rate to zero by pegging the exchange rate could result in a welfare gain of up to one percent of GDP. Bergin and Tchakarov (2003) provide an extension of this type of model to more realistic situations involving incomplete asset markets and investment by firms. They are able to calculate the effects of exchange rate uncertainty for a wide range of cases and find that the welfare costs are generally quite small, on the order of one-tenth of one percent of consumption. They explore the implications of two cases, however, where risk does matter quantitatively, on the order of the effect in the example cited above by Obstfeld and Rogoff: first, where consumers exhibit considerable persistence in their patterns of consumption, such that welfare is adversely affected by sudden changes in consumption; and second, where asset markets are asymmetric in that there is only one international bond, such that the country without its own bond is adversely affected.
Finally, Koren and Szeidl (2003) develop a model that brings out clearly the interactions among macroeconomic variables. They show that what matters is not the unconditional volatility of the exchange rate as a proxy for risk, as used in many empirical papers in the literature, but rather that exchange rate uncertainty should influence trade volumes and prices through the covariances of the exchange rate with the other key variables in the model. In this general equilibrium context, they stress that it is not uncertainty in the exchange rate per se that matters, but rather whether this uncertainty magnifies or reduces a firm’s other risks on the cost and demand side, and ultimately whether it exacerbates or moderates the risk faced by consumers.
Empirical Results on the Relationship Between Exchange Rate Volatility and Trade
The early empirical work on the effect of exchange rate variability on trade surveyed in the 1984 IMF study did not yield consistent results, with much of the work yielding little or no support for a negative effect. For example, the early work by Hooper and Kohlhagen (1978) used the model of Ethier (1973) for traded goods and derived equations for export prices and quantities in terms of the costs of production, reflecting both domestic and imported inputs, other domestic prices, domestic income, and capacity utilization. Exchange rate risk was measured by the average absolute difference between the current period spot exchange rate and the forward rate of the last period, as well as by the variance of the nominal spot rate and the current forward rate. The authors examined the impact of exchange rate volatility on aggregate and bilateral trade flow data for all G-7 countries except Italy. In terms of the effect of volatility on trade flows, they found essentially no evidence of any negative effect. Cushman (1983) uses a model similar to that of Hooper and Kohlhagen but extends the sample size and uses real as opposed to nominal exchange rates. Of 14 sets of bilateral trade flows between industrial countries, he found a negative and significant effect of volatility for six cases. Finally, IMF (1984) uses a simplified version of Cushman’s model to estimate bilateral exports between the G-7 countries from the first quarter of 1969 to the fourth quarter of 1982, with real GNP, the real bilateral exchange rate, relative capacity utilization, and variability measured as the standard deviation of the percentage changes in the exchange rate over the preceding five quarters. In only two cases did variability have a significantly negative coefficient, while positive coefficients were significant in several cases.
A number of factors may have contributed to the lack of robust findings in this early work. First, as noted above, theoretical considerations do not provide clear support for the conventional assumption that exchange rate volatility has a negative impact on the level of trade. Second, the sample period over which exchange rates showed significant variation was relatively short. Finally, the specification of the estimating equations was typically rather crude, consisting of a few macro variables from standard trade equations in use at the time.
McKenzie (1999) surveys a large number of empirical papers on the topic, most of which appeared after the IMF study. He stresses the point made above that, at a theoretical level, models have been constructed that lead to negative or positive effects of variability on trade, and that a priori there is no clear case that one model is superior to another. His survey of the empirical work leads to the same mixed picture of results, with many studies finding no significant effect or, where significant, no systematic effect in one direction or the other. He finds, however, that the most recent contributions to the literature have been more successful in obtaining a statistically significant relationship between volatility and trade, which he attributes to more careful attention to the specification of the estimation technique and the measure of volatility used. Similarly, U.K. Treasury (2003) cites a number of recent studies—De Grauwe (1987), Rose (2000), Dell’Ariccia (1999), Anderton and Skudelny (2001), Arize (1998), and Fountas and Aristotelous (1999)—that find a negative link, but these effects are not very large: complete elimination of volatility would raise trade by a maximum of 15 percent compared to the consensus estimate of the effect as typically less than 10 percent.
Recent work on this topic employing the gravity model has found some significant evidence of a negative relationship between exchange rate variability and trade.6 The gravity equation has been used widely in empirical work in international economics and has been highly successful in explaining trade flows.7 In its basic form, the gravity model shows bilateral trade flows between countries as depending positively on the product of their GDPs and negatively on their geographical distance from each other. Countries with larger economies tend to trade more in absolute terms, while distance can be viewed as a proxy for transportation costs, which act as an impediment to trade. In addition, population is often included as an explanatory variable as an additional measure of country size. In many applications, a host of dummy variables are added to account for shared characteristics that would increase the likelihood of trade between two countries, such as common borders, common language, and membership in a free trade association. To this basic equation researchers add some measure of exchange rate variability to see if this proxy for exchange rate risk has a separate, identifiable effect on trade flows after all other major factors have been taken into account.
The work by Dell’Ariccia (1999) provides a systematic analysis of exchange rate volatility on the bilateral trade of the 15 EU members at that time and Switzerland over the 20-year period from 1975 to 1994, using four different measures of exchange rate uncertainty: the standard deviation of the first difference of the logarithm of the monthly bilateral nominal exchange rate, that of the real exchange rate, the sum of the squares of forward errors, and the percentage difference between the maximum and minimum of the nominal spot rate. In the basic regressions, exchange rate volatility has a small but significantly negative impact on trade: eliminating volatility to zero in 1994 would have increased trade by an amount ranging from 10 to 13 percent, depending on the particular measure of variability.8 The results for both nominal and real variability are very close, which is not surprising given that in the sample the two exchange rate measures are highly correlated.
Dell’Ariccia then goes on to take account of the simultaneity bias that can result from central banks’ attempts to stabilize their exchange rates with their main trading partners. If they were successful, there would be a negative association between exchange rate variability and the level of trade, but it would not reflect causation from the former to the latter. He first uses an instrument (the sum of squares of the three-month logarithmic forward error) for the measures of exchange rate volatility to account for possible endogeneity in this variable. The results confirm the negative relationship between volatility and trade, with the magnitude of the effect about the same as before. In addition, he uses both fixed effects and random effects estimation methods to account for the simultaneity bias. In this case, the effect is still significant, but the magnitude is much smaller: total elimination of exchange rate volatility in 1994 would have increased trade by only 3–4 percent.
Rose (2000) also employs the gravity approach and uses a very large data set involving 186 countries for the five years 1970, 1975, 1980, 1985, and 1990. His main objective in the paper is to measure the effect of currency unions on members’ trade—an issue that is dealt with at length below—but he also uses his model to test for the effects of exchange rate volatility on trade. His primary measure of volatility is the standard deviation of the first difference of the monthly logarithm of the bilateral nominal exchange rate, which is computed over the five years preceding the year of estimation. In his benchmark results using the pooled data, he finds a small but significant negative effect: reducing volatility by one standard deviation (7 percent) around the mean (5 percent) would increase bilateral trade by about 13 percent, which is similar to the finding of Dell’Ariccia described above.9 This result is robust when using three alternative measures of volatility, but not when the standard deviation over the previous five years of the level of the exchange rate is used. When random effects are incorporated in the estimation, however, the magnitude of the effect of volatility on trade is reduced to about a third of the benchmark estimate, or roughly 4 percent. Thus the estimation results of Rose and Dell’Ariccia appear to be quite consistent.
A recent paper by Tenreyro (2003) casts some doubt, however, on the robustness of these results. She uses a gravity equation similar to that of Rose for a broad sample of countries, using annual data from 1970 to 1997. The measure of volatility is the same as that employed by Rose, except that the standard deviation of the log change in monthly exchange rates is measured only over the current year. Her main objective is to address several estimation problems in previous studies on the effect of volatility on trade. When these problems are not addressed and the ordinary least squares method is used, she finds a small effect: reducing volatility from its sample mean of about 5 percent to zero results in an increase in trade of only 2 percent. When the more appropriate method is used, but without taking account of endogeneity, eliminating exchange rate uncertainty leads to an estimated 4 percent increase in trade. When endogeneity is taken into account through the use of instruments, however, volatility has an insignificant effect on trade: a result that is robust on the choice of instruments.
Finally, it should be noted that there has been some recent work looking at the effects of exchange rate volatility on disaggregated trade flows. Broda and Romalis (2003) find that volatility decreases trade in differentiated products relative to trade in commodities, although the effect is rather small: eliminating all real exchange rate volatility would increase trade in manufactures by less than 5 percent and total trade by less than 3 percent. They note, however, that some countries with particularly volatile exchange rates, especially developing countries, would experience a more pronounced increase in trade. Koren and Szeidl (2003) also use disaggregated data and find small effects: eliminating exchange rate variability would result in a change in export prices of only a few percentage points.
For an analysis of the effects of forward cover on the level of trade, see Ethier (1973), Kawai and Zilcha (1986), and Viaene and de Vries (1992). Wei (1999), however, does not find empirical support for the hypothesis that the availability of hedging instruments reduces the impact of exchange rate volatility on trade.
In 1994, the average standard deviation of the monthly nominal exchange rate change was roughly 5.5 percent, and over the sample period the annual average bilateral trade growth was 3.5 percent.
Parsley and Wei (2001) look at the effect of reducing nominal exchange rate variability on relative price variability across countries and find that reducing the former diminishes the latter. They also find, however, a much stronger effect arising from participation in a hard peg, such as a currency union, which is consistent with Rose’s finding of a large impact of a currency union on trade.