Goods Arbitrage and Real Exchange Rate Stationarity

Recent evidence suggests that while real exchange rates exhibit mean reversion, the reversion only sets in once a minimum "threshold" distance from the mean has been exceeded. The non-linearity has generally been attributed to costly arbitrage, which requires a minimum divergence before the costs of arbitrage can be recouped. In this paper, we examine this reasoning. If arbitrage was indeed the cause of mean reversion, one would expect to see a quantity respose of trade flows at the thresholds. We conduct an array of formal and informal tests of the presence of such quantity responses. We fail to unearth any significant evidence of trade flows changing at the points of mean reversion. Alternative explanations of mean reversion, notably exchange market intervention, would thus seem to warrant further attention.