A re-examination of analysts’ superiority over time-series forecasts

In this paper, we re-examine the widely-held belief that analysts’ earnings per share (EPS) forecasts are superior to forecasts from a time-series model. Using a naive random walk time-series model for annual earnings, we investigate whether and when analysts’ annual forecasts are superior. We also examine whether analysts’ forecasts approximate market expectations better than expectations from a simple random walk model. Our results indicate that simple random walk EPS forecasts are more accurate than analysts’ forecasts over longer forecast horizons and for firms that are smaller, younger, or have limited analyst following. Analysts’ superiority is less prevalent when analysts forecast large changes in EPS. These findings suggest an incomplete and misleading generalization about the superiority of analysts’ forecasts over even simple time-series-based earnings forecasts. Our findings imply that in certain settings, researchers can reliably use time-series-based forecasts in studies requiring earnings expectations. We thank workshop participants at the University of Akron, University of Arkansas, Florida State University, and Ohio State University for helpful suggestions and comments. James Myers gratefully acknowledges financial support from the Ralph L. McQueen Chair and Linda Myers gratefully acknowledges financial support from the Garrison/Wilson Chair, both at the University of Arkansas.