Consistent earnings surprises and behavioral biases in analyst earnings forecasts.
详细信息
文摘
Several behavioral finance models suggest that individuals tend to overreact to information signals as similar information signals continue to arrive, and conjecture that such biased behavior can explain important return patterns such as return momentum and return reversal. This paper seeks to evaluate the descriptive validity of this behavioral bias by examining analysts' forecast revisions and subsequent forecast errors as a series of similar earnings surprises develops. I find that analysts react more strongly to consistent earnings surprises (i.e., earnings surprises that are preceded by a sequence of similar earnings news) than to isolated earnings surprises in revising their one-quarter-ahead earnings forecasts. However, such forecasting behavior does not lead to their overreaction to consistent earnings surprises. Specifically, I find analysts underreact to earnings news on average, consistent with prior studies. As a series of positive earnings surprises continues, analysts underreact more to the later news than to the initial good earnings news in the same series. This finding is contrary to the behavior suggested by behavioral finance models. Although I do find analysts underreact less to consistent bad earnings news than to isolated bad earnings news, analysts' asymmetric behavior to consistent good or bad earnings news fails to fully support the behavioral assumption that individuals tend to overreact to information regardless of its nature as similar information signals continue to arrive. Furthermore, it seems that analysts perceive consistent good earnings news as less precise than isolated good earnings news, offering a possible reason why analysts underreact more to consistent good earnings news. Overall, the evidence fails to support the conjecture from behavioral finance theories that investors overreact to a series of consistent earnings signals, suggesting that researchers should be cautious in extrapolating behavioral biases observed in a controlled laboratory setting to real market settings.