Stationary Component in Stock Prices : A Reappraisal of Empirical Findings

This paper re-examines the issue of mean reversion in stock prices by incorporating the structural break effect in the long horizon regression. Before adjusting for structural break, the paper finds that previous studies understate the evidence of mean-reversion. The understatement is mainly due to the clustering heteroskedasticity and autocorrelation in the overlapping returns. After adjusting for structural break(s), no evidence of predictability for value-weighted returns has been documented. However, stronger evidence of mean reversion in stock prices is documented for equally-weighted portfolios. The reverse effect of structural break can be explained by the switch to mean aversion in the last subperiod of value-weighted portfolios while no such switch in equally weighted portfolios