Earnings Skewness and Analyst Forecast Bias

Abstract

Statistically optimal forecasts need not be unbiased. If analysts' objective is to provide the most accurate forecast through minimizing the mean absolute forecast error, the optimal forecast is the median instead of the mean earnings. When earnings distribution is skewed, the median is different from the mean and forecast bias is observed. Thus, analyst forecast bias could be a natural result of analysts' effort to improve forecast accuracy combined with skewed distribution of earnings. We find that earnings skewness explains a significant amount of variation in analyst forecast bias across firms, across fiscal quarters and across time. Moreover, the market appears to understand at least part of the skewness-induced bias and adjusts accordingly. One salient feature of our explanation is that we predict not only forecast optimism for firms with negatively skewed earnings, but also pessimism for firms with positively skewed earnings, thus providing a more coherent explanation of analyst forecast bias.

JEL Classification: G10, G29, M41