Abstract
The trade effects of exchange rate variability have been an issue in international economics for the past 30 years. The contribution of this article is to apply meta-regression analysis (MRA) to the empirical literature. On average, exchange rate variability exerts a negative effect on international trade. Yet MRA confirms the view that this result is highly conditional, by identifying factors that help to explain why estimated trade effects vary from significantly negative to significantly positive. MRA evidence on the pronounced heterogeneity of the empirical findings may be instructive for policy: first, by establishing that average trade effects are not sufficiently robust to generalize across countries; and second, by suggesting the importance of hedging opportunities – hence of financial development – for trade promotion. For the practice of MRA, we make a case for checking the robustness of results with respect to estimation technique, model specification and sample.
Acknowledgements
The authors acknowledge helpful suggestions from Tom Stanley and other participants at the International Colloquium on Meta-Analysis in Economics, Sønderborg (September 2007). In addition, this article was greatly improved by guidance from our two anonymous referees. The authors alone are responsible for shortcomings.
Notes
1Although some studies which conclude that the trade effect is negative, nonetheless contain some positive results; for example, Stokman (Citation1995) reports two positive effects and one zero effect among otherwise consistently negative effects.
2The main combinations of keywords used were ‘exchange rate variability’, ‘exchange rate volatility’, ‘exchange rate uncertainty/risk’ and ‘trade effect’.
3Some studies employ more than one measure of exchange rate variability (e.g. by including both current and lagged values). An appendix detailing how the effect size was selected in each such case is available on request. It is excluded here for reasons of length.
4The mean and SD weighted to give each study equal influence on the estimates are −1.21 and 2.55, respectively.
5This procedure relaxes the assumption of independence between observations within the same group, requiring only that observations be independent between groups, and produces ‘correct’ SEs ‘even if the observations are correlated’ (StataCorp, Citation2003; see also Deaton, Citation1997, pp. 73–78 and Baum et al., Citation2003).
6The 573 t-values estimated from 100 or fewer degrees of freedom have a mean of −0.75 (SD = 2.02); the 114 estimated from between 101 and 500 degrees of freedom have a mean of −1.08 (SD = 2.60); and the 148 estimated from more than 500 degrees of freedom have a mean of −3.66 (SD = 4.56).
7In bivariate regressions of ERVES on a constant and the SqRt_df, the R 2measures are 0.27 (unweighted) and 0.16 (weighted), respectively. In comparison, the R 2measures reported in are 0.48 and 0.35, respectively.
8MERV11 is used in one study with 17 observations; MERV8, in three studies with 11 observations; MERV3, in three studies with 24 observations; MERV7, in two studies with 37 observations and MERV13, in one study with six observations.