Because we have no three-year-ahead forecasts, the coefficients on remaining variables that are correlated with the omitted three-year-ahead forecasts must be biased upward. In addition, the coefficients on abnor mal earnings-especially over the longer horizons-will be influenced by the extent to which those regressors proxy for any abnormal earnings that are expected beyond four years but are omitted from the regression.' Precise coefficient values, however, are not a central concern for our pur poses; at issue here is the overall explanatory power of the regressions. The key question is how well one can explain prices with only one book value number and three abnormal earnings forecasts