Can Ex-Post Bias Arise From an Ex-ante Unbiased Forecast?

• In Event: Financial Analysts: Forecast and Revision Behavior

Abstract

Our paper contests the widely held belief that the forecast of an unbiased analyst should have, on average, zero forecast error. In our contesting of this belief, we illustrate that analysts may be literally unable to eliminate all bias from their forecasts. We develop and test a theory that demonstrates that an ex-post biased forecast error can be obtained from an ex-ante unbiased forecast when an analyst forecasts from the weighted average of an asymmetric probability distribution and a uniform distribution. Our theory explains the change in the magnitude of forecast bias over the forecast horizon, the percentage of optimistic and pessimistic forecast errors at long horizons, and seemingly counter-intuitive findings around the implementation of Regulation Fair Disclosure. Our theory may inform regulators as they attempt to reduce or eliminate analyst bias; eliminating incentives to bias may not lead to statistically unbiased analyst forecasts.

Authors

• Kathryn Brightbill, University of Iowa

• Cristi A Gleason, University of Iowa

• Mark Penno, University of Iowa