New Evidence on Serial Correlation in Analyst Forecast Errors

Abstract

We reexamine the serial correlation of forecast errors using a method that allows analysts to react differently to good and bad news. Our method also controls for the influence of a normal non-zero, firm-specific component of forecast error. Our results indicate that forecast errors exhibit positive serial correlation when there is bad news in the prior forecast error, negative serial correlation when there is good news in the prior forecast error, and no serial correlation when there is no news in the prior forecast error. These findings are consistent with analysts having optimistic reactions to new information.