In contrast to positive earnings forecasts, the negative earnings forecasts of security analysts are grossly optimistic. We adjusted negative earnings forecasts downward by varying amounts and evaluated forecast performance according to (1) forecast accuracy relative to the consensus, (2) the frequency of being closer to actual earnings than the consensus, and (3) the frequency with which adjusted forecasts underestimate actual earnings, thereby jeopardizing the analyst's relations with corporate managers. Relative forecast accuracy and the probability of beating the consensus are improved, without an inordinate increase in the probability of underestimating earnings, by adjusting negative forecasts downward by a small amount.