Do analysts manage earnings forecasts to 'confirm' their own recommendations?

We propose and test a form of bias in earnings forecasts that arises from analysts' desire to be perceived as accurate. Our hypothesis suggests that an analyst with a buy (sell) recommendation has an incentive to report a downward (upward) biased earnings estimate so that the company is more (less) likely to beat the consensus forecast and experience an earnings surprise that appears “in line” with the analyst's outstanding stock recommendation. Consistent with this prediction, we find that stock recommendations prior to earnings announcements significantly and positively predict subsequent earnings forecast errors, and that this predictability is concentrated in situations where the motivation for such strategic behavior is particularly strong. Together, the results provide evidence that the desire to appear accurate does not always lead to more accurate forecasts.