Technical analysis and central bank intervention

This paper extends the genetic programming techniques developed in Neely, Weller and Dittmar (1997) to show that technical trading rules can make use of information about U.S. foreign exchange intervention to improve their out-of-sample profitability for two of four exchange rates. Rules tend to take positions contrary to official intervention and are unusually profitable on days prior to intervention, indicating that intervention is intended to check or reverse predictable trends. Intervention seems to be more successful in checking predictable trends in the out-of-sample (1981-1996) period than in the in-sample (1975-1980) period. We conjecture that this instability in the intervention process prevents more consistent improvement in the excess returns to rules. We find that the improvement in performance results solely from more efficient use of the information in the past exchange rate series rather than from information about contemporaneous intervention.