Current measures of analyst forecast dispersion have limited association with abnormal earnings announcement trading volume, contrary to the findings in Beaver (1968). One reason for this is that analysts increasingly forecast in the days immediately following earnings. This paper proposes extending the ex ante forecast window to include the significant forecast activity after Q t-1 earnings—roughly 90-91 days prior—and shortening the ex post window to capture analysts’ updated beliefs concurrent with Q t earnings. These adjustments generate robust positive associations between dispersion measures and abnormal volume across various specifications, strongly consistent with Beaver (1968). Amended and IBES Summary dispersion (opinion divergence) measures generate postearnings announcement returns more consistent with the risk-factor explanation of Varian (1985): dispersion is positively associated with future returns, while change in dispersion is negatively associated with concurrent returns. Results using existing measures are inconsistent and inconclusive. Amended forecast windows generate samples twice as large as existing windows and exclude forecasts issued beyond the Q t earnings window. Similar amendments can be applied to the construction of other analyst forecast-based determinants of trading volume and returns around earnings, such as information uncertainty, asymmetry, precision, consensus, and differential interpretation.
Matthew Driskill (Wed,) studied this question.