Biases in Individual Forecasts: Experimental Evidence
Document Type
Article
Publication Date
4-2008
Abstract
Trueman [1994] provides a model of forecasting behavior in which analysts do not always make forecasts that are consistent with their private information. Using Trueman's model to provide theoretical direction, we conduct six experimental sessions to investigate individual forecasting behavior. In each session, four individuals predict earnings based on possibly divergent information. We manipulate forecast ability so that two individuals are strong analysts and two are weak. In three sessions, forecasts are released simultaneously. We find that forecasts do not always reflect private information. Both weak and strong analysts make forecasts that are inconsistent with private information, although the behavior is much more pronounced for weak analysts. In another three sessions, forecasts are released sequentially. We find that weak second analysts engage in herd behavior: that is, they mimic the reporting behavior of the first analyst. In contrast, strong second analysts are unaffected by the reporting behavior of the first analyst. The overall findings are consistent, in spirit, with the forecasting behavior suggested by Trueman [1994].