Technical Analysis as a Source of Exchange-Rate Forecast Irrationality: The Head-and-Shoulders Pattern

Abstract

Much empirical research suggests that exchange-rate forecasts are not rational. This paper identifies a commonly used technical trading signal, the head-and-shoulders pattern, as a potential source of that irrationality. We use an objective, computer-implemented algorithm to identify head-and-shoulders patterns in daily dollar exchange rates during the floating rate period. The resulting profits, replicable in real time, are then evaluated statistically using the bootstrap technique. We impose two conditions for trading-rule rationality: a rule must be profitable and it must not be dominated by other trading rules. We find that head-and-shoulders trading generates statistically significant profits, and thus satisfies the first condition. However, the trading rule is dominated by simple filter rules, so it does not satisfy the second condition.

JEL Classification: F31, G12, G14