ABSTRACT
Empirical models of exchange rate determination, especially at intermediate estimation horizons, have frustrated economists at least since the Meese and Rogoff (1983a, 1983b) result that macrobased exchange rate models under-perform a random walk model in predictive ability. In the empirical finance literature there is, however, a long tradition of studying the higher frequency relationship between features of prices of financial assets and measures derived from trading activity.1 Simple analysis of trading volume, however, does not help resolve the Meese-Rogoff problem, not least because volume is directionless, i.e. a change in volume cannot predict the direction of FX changes.
