On the Association between Analysts' Forecast Errors and Past Stock Returns : A Re-Examination

Prior studies (e.g., Lys and Sohn 1990; Ali, Klein and Rosenfeld 1992) have documented a positive association between analysts' forecast errors and past stock returns and suggested cognitive bias on the part of analysts as a possible explanation. In this paper, we separately analyze the association between forecast errors and past negative returns and that between forecast errors and past positive returns. We find that forecast errors are only positively associated with past negative returns and are not associated with past positive returns. These results are robust to a series of sensitivity tests. They are inconsistent with analysts being subject to cognitive bias; instead, they are consistent with several explanations related to accounting conservatism or analysts' incentives: analysts having difficulty in forecasting discretionary charges associated with past negative returns, analysts not exerting effort in forecasting earnings of firms with poor performance, or analysts ignoring bad news in order to please managers