Rationality and Analysts' Forecast Bias

Authors

  • Terence Lim

    Search for more papers by this author
    • Goldman Sachs and Dartmouth College Tuck School. I am especially grateful to Andrew Lo, my dissertation advisor at MIT, for generous guidance and many useful discussions, and to S. P. Kothari, Jeremy Stein, René Stulz, Jiang Wang, Kent Womack, an anonymous referee, and seminar participants at Dartmouth College, MIT, the University of Arizona, and the Ninth Annual Conference on Financial Economics and Accounting (NYU) for helpful comments. All errors are my own. The views expressed do not necessarily reflect those of Goldman Sachs. Data on analysts' forecasts were provided by I/B/E/S Inc., under a program to encourage academic research.

ABSTRACT

This paper proposes and tests a quadratic-loos utility function for modeling corporate earnings forecasting, where financial analysts trade off bias to improve management access and forecast accuracy. Optimal forecasts with minimum expected error are optimistically biased and exhibit predictable cross-sectional variation related to analyst and company characteristics. Empirical evidence from individual analyst forecasts is consistent with the model's predictions. These results suggest that positive and predictable bias may be a rational property of optimal earnings forecasts. Prior studies using classical notions of unbiasedness may have prematurely dismissed analysts' forecasts as being irrational or inaccurate.

Ancillary