Since the breakdown of the Bretton Woods agreement, the trade effect of exchange rate variability (ERV) has been contentious. However, neither the theoretical nor the empirical literature provides unambiguous guidance on the trade effect of ERV. This article applies metaregression analysis to the empirical literature and finds evidence of: modest publication bias; a range of authentic empirical effects that are highly conditional, even with respect to sign and, correspondingly, pronounced heterogeneity in the reported results. Investigation of this heterogeneity reveals that the results are significantly influenced both by authors' modelling strategies and by the contexts of their investigations. In particular, researchers are most likely to find an adverse trade effect by investigating low-frequency real exchange variability and trade between less developed economies, hence, beyond the reach of hedging opportunities (as suggested by previous studies). In general, our most important advice for policy makers is that economic research does not reveal a single representative effect size.