Financial Integration and FX Trading

This paper addresses international financial integration in a new way. Past work on integration focuses either on specul ative integration or on geographic integration. Our analysis focuses instead on informational integration. We d evelop a multi -currency model of portfolio allocation in the presence of dispersed information (therein making our definition of informational integration precise). We then test the model’s implications using four months of concurrent transaction data on nine major currencies (each of which floats against the US dollar). The model explains 45 to 78 percent of daily returns in all nine currencies. Moreover, the model’s prediction that order flow information should be relevant for determining prices in multiple markets is borne out. Two currencies in particular exhibit substantial informational relevance in other markets: the German Mark and the Swiss Franc. This is in keeping with recent empirical results on information geography (8 of our 9 currencies are European). Finally, we find that order flow information accounts on average for about 80 percent of the (high) unconditional covariance in cu rrency returns. This suggests that currency return covariance is not due to co mmon public news, as traditionally modeled, but is instead due primarily to market aggregation of dispersed information.