Order Flow and Exchange Rate Dynamics in Electronic Brokerage System Data

We study the association between order flow and exchange rate returns in five years of high-frequency intraday data from the leading interdealer electronic broking system, EBS. While the association between order flow and exchange rate returns has been studied in several previous papers, these have mostly used relatively short spans of daily data from older bilateral dealing systems and, usually, transaction counts instead of actual trading volume. Using a substantially longer span of recent high-frequency data and measuring order flow as actual signed trading volume, we find a strong positive association between order flow and exchange rate returns at frequencies ranging from one minute to one day, and a more modest but still sizeable association at the monthly frequency. We find, however, no evidence that order flow has predictive power for future exchange rate movements beyond, possibly, the next minute. Focusing on the behavior of order flow and exchange rates at the time of scheduled U.S. economic data releases, we find that the surprise components of these announcements are associated with order flow at high frequency immediately after the data releases. This finding seems inconsistent with a simple efficient markets view of how a public news announcement is incorporated into prices