The sample consists of U.S. firms with available data for six years from 1996 to 2001.
Because our measure of forecast accuracy requires future years’ data (i.e., year t1 and
t2 data) on earnings, however, we use up to year 2003 data. Our final sample consists of
9,261 firm-year observations that have necessary analysts’ forecast data available from
I/B/E/S and financial data available from Compustat. We retain 3,749 firm-year observations
for the forecast dispersion sample due to further data requirements.
Descriptive statistics of the regression variables are provided in Table 1. The mean
(median) forecast accuracy (ACCY) is 0.030 (0.003) in the sample, suggesting that the
mean (median) difference between analysts’ forecasts and corresponding actual earnings is
about 3 (0.3) percent of the lagged stock price. The mean dispersion (DISP) of 0.021