Run length and the predictability of stock price reversals

Abstract

Abstract Survival analysis is used to estimate time‐varying probabilities of price reversals using daily data for the Australian All Ordinaries Price Index. Lagged price changes lead to persistence (shortening) in a price run if they are of the same (opposite) sign as the run. An increase in the number of runs observed in the previous 30 days also increases the probability of price reversal. The predictive accuracy of the models is assessed using a probability scoring rule. Consistent with market efficiency, the estimated models are less accurate than the random walk model in predicting the length of individual price runs out‐of‐sample.