Biases in analyst forecasts: Cognitive, strategic or second-best?

Financial analysts act in a complex environment, and the incentives they face may make them issue forecasts which appear to be inconsistent with rational expectations. In order to test several explanations for biases, I analyze a large sample of individual analyst earnings forecasts of German companies. The evidence supports the conjecture that biases in point estimates can be useful for communicating information about forecast precision. When analysts believe their clients misconceive the true precision of the forecast, they distort their estimates in a way which gives them a more appropriate weight in the clients' decision process. Game-theoretical models, on the other hand, which rationalize biases by arguing that they are strategic lack support from the data. Violations of forecast rationality are more likely to stem from underreaction and overconfidence, two types of bias discussed in the phychological literature. Judging from the results of simulations, however, these cognitive errors do not seem to be of major economic significance.