On the nature and rationality of analysts' forecasts under earnings conservatism

We examine tile implications of conservatism bias in reported earnings for research on analysts' earnings forecasts. We show that when earnings are conservative, in the sense of [J. Account. Econom. 24 (1997) 3], earnings changes are no longer equal to earnings surprises and the distribution of earnings surprises is left skewed, due to negative transitory items. Moreover earnings changes become partially predictable, due to the fact that transitory negative earnings surprises reverse in the next period. Under these circumstances analysts are faced with a dilemma when producing their earnings forecasts. Should they issue forecasts that focus on the permanent component of earnings or should they issue forecasts that also incorporate an estimate of the transitory component of earnings? We present theoretical arguments that explain why it would be reasonable for analysts to resolve this dilemma pragmatically by issuing initial forecasts that are focused on permanent earnings, but which are revised over the year to come in line with end of year reported earnings. We also report evidence that suggests that a significant proportion of analysts appear to follow this pragmatic approach. We also present evidence that forecast revisions respond asymmetrically to good and bad news consistent with the predictions of our model. Finally we show that optimism bias is associated empirically with conservatism in reported earnings. © 2003 Elsevier Ltd. All rights reserved.