IV New Evidence on the Effect of Exchange Rate Volatility on Trade
- Peter Clark, Shang-Jin Wei, Natalia Tamirisa, Azim Sadikov, and Li Zeng
- Published Date:
- September 2004
As discussed in Section II, theoretical models do not point unambiguously to a negative effect of exchange rate volatility on trade. Moreover, empirical analysis in the existing literature has not uncovered a strong causal impact that is consistently negative. In the empirical analysis reported in this section, there is no obvious negative relationship between aggregate exchange rate volatility and aggregate trade. Turning to bilateral trade, there is some evidence that exchange rate volatility tends to reduce trade. This negative effect, however, is not robust to alternative ways of controlling for factors that could affect trade. The key findings of this empirical analysis are summarized below, and the appendix describes the statistical results in more detail.
The objective of the empirical analysis is to examine the role of exchange rate volatility in trade in a comprehensive manner. Compared to the existing academic literature and the 1984 IMF paper on the topic, the contribution of this analysis lies in exploring the effect of exchange rate volatility on trade along several dimensions:
- by the type of exchange rate volatility: examining a range of different exchange rate volatility measures—short- and long-run, real and nominal, official, IFS based and parallel market based, and conditional and unconditional;
- by country group: testing if the impact of exchange rate volatility differs across country groupings, including industrial and developing countries; and
- by the type of trade: examining the impact not only on aggregate but also on sectoral trade, which allows one to test if the effect of exchange rate volatility varies in direction and magnitude across different types of goods. The role of exchange rate volatility has not yet been explored extensively using disaggregated trade data.
In addition to the disaggregation of the volatility effect, tests were conducted to determine its robustness to alternative ways of controlling for joint causality between trade and exchange rates and for trade-related factors other than exchange rate volatility. Finally, while the focus is on exchange rate volatility, this provides an opportunity to revisit a related topic, the role of a common currency in enhancing trade flows, and to explore the robustness of the finding by Rose (2002) that this positive effect is very large.
Aggregate Volatility and Aggregate Trade—A First Look
It is instructive to look at the time paths of world trade and exchange rate volatility and examine if there is any obvious negative association between the two. Figure 4.1 shows the evolution of world trade since 1970 together with the average real effective volatility for all countries in the sample. There is a clear bulge in exchange rate volatility from 1989 to 1993, which reflects the large fluctuations in the currencies of a number of transition economies during this period in the aftermath of the breakup of the Soviet Union.20 If one excludes transition economies from the measure of world currency volatility, the bulge disappears. What then appears is an upward trend in average volatility from the early 1970s through the end of the 1980s, but a general moderation in the overall level of currency volatility since then.
Figure 4.1.Effective Volatility of the Real Exchange Rate and World Trade
1 Volatility is measured as the unweighted average of the real exchange rate volatility of the countries in the sample (left scale, in percent).
2 World trade is measured as the average of the volume of world exports and imports in billions of 1995 U.S. dollars (right scale).
In comparison, world trade has increased steadily since 1970, and the growth rate is much more smooth than that of exchange rate volatility. Looking at the movement of world trade and aggregate volatility over time, there does not appear to be any clear relationship between them. Therefore, at the aggregate level there is no evidence of a negative effect of real exchange rate volatility on trade.
It may be useful to examine the relationship between the two by breaking down the sample by major country groups (Figure 4.2) and developing countries by geographic region (Figure 4.3) and by type of export (Figure 4.4). In some of the sub-samples and for some of the years, there appears to be a negative association between exchange rate volatility and the level of trade in certain country groupings. This is most evident in the case of transition economies in 1990–94 (lower left graph of Figure 4.2), the Asian crisis in 1997–98 (upper right graph of Figure 4.3), and economies exporting non-fuel primary products in the early 1980s (lower graph of Figure 4.4). This negative association may not reflect a causal relationship, however, but rather a manifestation of the effects of a common set of factors that both raise currency volatility and reduce trade. For example, the Asian crisis led to a large decline in the imports of the affected countries and major movements in their exchange rates, but the fall in domestic demand was the most important factor reducing import volumes—not currency volatility. Similarly, the breakup of the Soviet Union caused widespread dislocations in many transition economies, resulting in substantial falls in output and trade and huge changes in many exchange rates that were part and parcel of the transition process.
Figure 4.2.Effective Volatility of the Real Exchange Rate and Trade by Major Country Groups1
1 Trade is in billions of 1995 U.S. dollars; volatility is average of real exchange rate volatility.
Figure 4.3.Effective Volatility of the Real Exchange Rate and Trade of Developing Countries by Region1
1Trade is in billions of 1995 U.S. dollars; volatility is average of real exchange rate volatility.
Figure 4.4.Effective Volatility of the Real Exchange Rate and Trade of Developing Countries by Type of Export1
1 Trade is in billions of 1995 U.S. dollars; volatility is average of real exchange rate volatility.
In order to estimate the specific impact of exchange rate volatility on trade flows, it is necessary to take account of the separate effects of a myriad of factors that determine the level of exports and imports. The following moves away from aggregate trade and discusses a methodology that exploits the much richer variations in the data on bilateral trade and bilateral exchange rates, which in turn permit the identification of the distinct contribution of volatility on trade.
The Conceptual Framework for Analyzing the Volatility Effect in Trade
To investigate the effect of exchange rate volatility on trade, there are several building blocks to consider. First, there are factors other than exchange rate volatility that affect trade, and it is important to account for them in a way that is consistent with economic theory. Otherwise, one runs the risk of misattributing the effect of these other factors to exchange rate volatility. Second, the measure of exchange rate volatility should be conceptually reasonable. Third, it may be useful to allow exchange rate volatility to have different effects on different types of trade or trade in different country groupings. These building blocks are explained in turn.
As part of the first building block, one must account for the determinants of trade patterns other than exchange rate volatility in a modified gravity model. This model relates trade between a given pair of countries to characteristics of each of the two countries and the characteristics of their relationship to each other. The characteristics that are most important—to which the model owes its name—are the economic mass (i.e., GDP) and the distance between the countries. In addition, the empirical specifications of the gravity model typically control for other factors that augment or reduce trade, such as land areas, cultural similarity, geographical position, historical links, and preferential trading arrangements, all of which tend to affect the transaction costs relevant for bilateral trade and which have been found to be statistically significant determinants of trade in various empirical applications. Typically, the model also controls for the level of economic development, which is expected to have a positive effect on trade because the more developed countries tend to specialize and trade more (e.g., Frankel and Wei, 1993). The gravity model has been empirically successful in terms of its ability to explain a large part of the variations in the observed trade patterns. It also has the merit of being grounded in international trade theories, ranging from those based on country differences in factor endowments or technology to models of increasing returns to scale and monopolistic competition.
A relatively recent development in the theoretical foundation of the gravity model emphasizes “remoteness” or “multilateral resistance” effects. These effects were proposed by Anderson and van Wincoop (2003) and are defined as a function of unobservable equilibrium price indices that depend on bilateral trade barriers and income shares of all the trading partners. In other words, the multilateral resistance effects are catch-all expressions that summarize the effects on a given bilateral trade from differential, possibly unobserved, trade costs between this country pair and all other trading partners. The gravity equation can then be interpreted as indicating that bilateral trade depends on the bilateral trade barrier between the two countries in question, relative to the two countries’ multilateral resistance indices: for a given bilateral trade barrier between the two countries, higher barriers between them and their other trading partners would reduce the relative price of goods traded between them, thus raising bilateral trade. In empirical applications, the multilateral resistance indices can be conveniently proxied by country effects (fixed or time varying). Also included are time effects in the model to control for time-specific factors, such as world business cycles, global shocks, etc.
The second building block is the measure of exchange rate volatility. The focus of the benchmark model is on the long-run measure of IFS-based real exchange rate volatility. Its value in any given year, t, is calculated as the standard deviation of the first-difference of the monthly natural logarithm of the bilateral real exchange rate in the five years preceding year t, which is a conventional measure most commonly used in the current literature on the subject. To check the robustness of results, alternative—yet analogously calculated—measures of exchange rate volatility are examined: long-run IFS-based nominal exchange rate volatility; short-run, contemporaneous IFS-based real and nominal exchange rate volatility; and the short-and long-run volatility of real parallel market rates using data from Reinhart and Rogoff (2002). Also considered as an additional robustness analysis are the conditional volatilities of real exchange rates, which are estimated using a GARCH (1, 1) model. To ensure the stationarity of the GARCH model, countries with hyperinflation episodes, extreme exchange rate fluctuations, and/or incomplete data are excluded, focusing the estimation on 124 industrial, developing, emerging, and transition economies.
The third building block for the model is the consideration of different country groups and different types of trade. An analysis of the exchange rate volatility effect was conducted separately for industrial countries and developing countries. The separate effects of exchange rate volatility are allowed for, depending on the type of product trade—differentiated or homogeneous. The classification of products into differentiated and homogenous varieties follows the strategy in Rauch (1999). Conceptually, Rauch first identifies two types of homogenous products: those traded on an organized exchange (commodities) and those whose prices are reported regularly in a professional trade publication (referenced price products). All other products are then defined as differentiated products.
What the Data Tell Us
The gravity model performs well empirically, yielding precise and generally reasonable estimates. The coefficient on distance is negative and statistically significant, while the coefficient on the economic mass is positive and statistically significant. Most other control variables are also mostly significant and have the expected signs.
Does Exchange Rate Volatility Hamper Trade?
As a benchmark specification using country and time-fixed effects, one finds that the long-run real exchange rate volatility has a statistically significant negative effect on trade (Table 4.1, column 1, row 1). If exchange rate volatility were to rise by one standard deviation (from 0.12 to 0.15, for example, in the sample), trade would fall by about 7 percent (Table 4.1, column 2, row 1).21 This effect is comparable to the estimates found by previous studies, e.g., Rose (2000) and Tenreyro (2003).
|Not Controlling for Joint Causality Between Trade and Exchange Rates||Implied Percentage Change in Trade by One Standard Deviation Increase in Volatility||Controlling for Joint Causality Between Trade and Exchange Rates||Implied Percentage Change in Trade by One Standard Deviation Increase in Volatility|
|With country- and time-fixed effects||−2.37* (0.67)||−6.64||−22.64* (12.50)||−63.39|
|With country-pair and time-fixed effects||−2.40* (0.47)||−6.72||−6.49 (6.24)||−18.17|
|With time-varying country effects||2.89* (1.78)||8.09||−23.82 (28.87)||−66.70|
|With country- and time-fixed effects on the full sample||−1.16 (0.22)||−8.82||…||…|
Is the Negative Effect on Trade Robust to Alternative Ways of Controlling for Factors Other Than Bilateral Exchange Rate Volatility?
The answer is no. On the one hand, a negative effect is still observed when controlling for unobservable cultural, economic, historical, geographical, and other factors specific to a given pair of countries rather than individual countries (Table 4.1, column 1, row 2). On the other hand, no negative effect is found when country-specific effects are allowed to vary over time, as appears justified theoretically, given the dynamic nature of multilateral resistance. Indeed, in some cases this specification could even result in a positive coefficient (Table 4.1, column 1, row 3). While this does not necessarily imply that volatility promotes trade, it suggests that the finding of a negative effect of exchange rate volatility on trade is not robust.
A note of caution is in order here. Recent developments in the theoretical foundation of a gravity specification suggest that it is important to include time-varying country-fixed effects in order to fully absorb the multilateral resistance effects. Otherwise, one might misattribute to exchange rate volatility those effects on bilateral trade that should have been attributed to other factors. At the same time, it should be noted that part of the forces underlining bilateral exchange rate volatility is time varying and country specific. The inclusion of the time-varying, country-fixed effects could also overcorrect. For example, an unexpected increase in one country’s money supply could raise all the bilateral exchange rate volatility involving that country. Even if this increase in volatility depresses all bilateral trade involving that country, a specification that controls for the country’s time-varying fixed effects would not be able to reveal a negative effect of exchange rate volatility on trade. This qualification should be kept in mind in interpreting the result.
Sorting Out Causality
To the extent that countries implement policies aimed at lowering exchange rate volatility in order to increase bilateral trade, the model considered so far would suffer from an endogeneity bias. Two instrumental variable approaches are used to control for this possibility: that proposed by Frankel and Wei (1993), whereby the volatility in the relative quantity of money is used as an instrumental variable for exchange rate volatility; and that implemented by Tenreyro (2003), which relates the exchange rate volatility to the propensity of countries to adopt a common currency anchor. Neither of these approaches is perfect, and each has its advantages: the Frankel-Wei approach appeals to the monetary theory of exchange rate determination and is simple and easy to implement, while the Tenreyro approach appeals to the optimal currency framework, as described in Alesina, Barro, and Tenreyro (2002). There is no significant effect of exchange rate volatility on trade in the models with country-pair and time-varying country effects (Table 4.1, column 3, row 2 and column 3, row 3). The negative volatility effect found in the model with constant country effects, however, survives (Table 4.1, column 3, row 1).
Does the Conclusion Change When Alternative Measures of Volatility Are Employed?
The short answer is no. Table 4.2 reports results from the same regression, which includes our standard long-run measure together with all three alternative measures of exchange rate volatility (as differences from the long-run real IFS-based measure). The short-run volatility in the real exchange rate appears to discourage trade, albeit to a smaller extent than the long-run volatility. The volatility in the parallel market exchange rates has a similar effect on trade as the volatility in the IFS-reported exchange rates, but only in the long run. As shown in Appendix Tables A11 and A12, the volatilities of the nominal and real exchange rates are highly correlated and thus have similar effects on trade. In addition, conditioning the measure of exchange rate volatility on historical information using the GARCH approach, instead of the simple statistical measure of volatility, also preserves the negative relationship with trade. As in Table 4.1, when time-varying country fixed effects are controlled for, there is no longer evidence of a negative and significant association between volatility and trade.
|Long-Run Real Exchange Rate Volatility||Short-Run Real Exchange Rate Volatility2||Long-Run Real Parallel Market Exchange Rate Volatility2||Short-Run Real Parallel Market Exchange Rate Volatility2|
|With country- and time-fixed effects||−3.92*||(1.3)||−2.72*||(1.04)||−1.20*||(0.63)||−0.55||(0.73)|
|Implied percentage change in trade by one standard deviation increase in volatility||−10.98||−6.80||−6.48||−2.15|
|With country-pair and time-fixed effects||−4.72*||(0.76)||−4.15*||(0.55)||−0.42||(0.34)||−1.14*||(0.41)|
|Implied percentage change in trade by one standard deviation increase in volatility||−13.22||−10.38||−2.27||−4.45|
|With time-varying country effects||7.52*||(3.89)||6.70*||(3.24)||−2.20||(2.71)||−1.55||(2.8)|
|Implied percentage change in trade by one standard deviation increase in volatility||21.06||16.75||−11.88||−6.05|
In excess of long-run real official exchange rate volatility.
Does Exchange Rate Volatility Have a Different Effect on Trade in Differentiated or Homogeneous Products?
Recent developments in the economics of trade suggest that a given increase in transaction costs (of which exchange rate volatility is a component) could have a larger, negative effect on trade in differentiated products than on trade in homogenous products. But, as with aggregate trade, the estimation results show that this theoretical prior is not robust. When country and time effects are controlled for separately, exchange rate volatility indeed has a negative effect on trade in differentiated products, but not on trade in homogenous products (Table 4.3, column 1). When, however, time-varying country fixed effects are included (Table 4.3, column 3), the conclusion is overturned, as in the aggregate trade model.
|With Country and Time Effects||Implied Percentage Change in Trade by One Standard Deviation Increase in Volatility||With Time-Varying Country Effects||Implied Percentage Change in Trade by One Standard Deviation Increase in Volatility|
|Trade in homogeneous products||−0.59 (2.12)||−1.65||−2.97 (4.39)||−8.32|
|Trade in differentiated products||−2.89* (1.66)||−8.09||0.98 (3.06)||2.74|
Do Members in Currency Unions Trade More?
The core results confirm the finding of Rose (2000) that common currency arrangements triple trade. Apparently, the trade-enhancing benefits of currency unions exceed by far gains from a reduction in exchange rate volatility and are preserved over time (Appendix Tables A9 and A11). They are also robust to controlling for time-varying country effects, but break down in a model with country-pair fixed effects (Appendix Table A12). This suggests that currency union membership may be correlated with other country-pair characteristics. Once these characteristics are controlled for by the inclusion of country-pair fixed effects, there is no additional trade-promoting effect from currency unions.
Does the Volatility Effect Differ Across Country Groups?
In principle, the effects could be different. Foreign exchange markets are typically less developed and less liquid in developing countries, thereby limiting firms’ opportunities for hedging foreign exchange risk. Indeed, volatile exchange rates are more likely to be associated with the smaller trade of developing countries than with trade among advanced economies in the specification with country-fixed effects. The negative effect disappears for both country groups, however, when country effects are time varying (Appendix Table A15). As hedging instruments become more readily available for the currencies of industrial countries, one might expect that their trade would be less affected by exchange rate volatility. Wei (1999), however, finds little support for the hypothesis that the growing availability of hedging instruments is responsible for the small impact of volatility on trade.
On balance, for both aggregate and disaggregated trade there is empirical evidence pointing to a generally small negative effect of exchange rate volatility on trade. But this evidence is not overwhelming and not robust across different empirical specifications.
Data for transition countries are not reported before 1988 because most of these countries did not exist before 1991. Data are available for Yugoslavia only from 1970 and for Hungary beginning in 1976. The effective volatilities for the major country groupings shown in Figures 4.2–4.4 do include, however, the available bilateral exchange rate data for transition countries, weighted by appropriate trade shares.
This impact is computed as the estimated coefficient in the regression equation multiplied by one standard deviation of the volatility measure, multiplied by 100 to convert to percent.