Reframing Analyst Forecast Efficiency: An Empirical Bayes Perspective

We develop a rational, non-behavioral framework in which analysts use empirical Bayesian techniques to construct their forecasts. Our approach assumes that analysts construct a single empirical prior by aggregating information across all of the firms that they cover, and then use this single prior to construct a unique forecast for every firm. This allows analysts to enhance efficiency at the portfolio level rather than at the individual firm level. We show that our empirical Bayes framework explains patterns of forecast bias, including downward (upward) adjustments for firms with higher (lower) prior means, and underreaction (overreaction) to information for firms with greater (lesser) prior variance. Our findings advance prior studies’ traditional view that analysts pursue firm-level forecast efficiency, and instead suggest that their pursuit of portfolio-level efficiency explains many of the properties of their forecasting behavior.