Predicting Earnings Surprises in European Countries

Abstract

Despite the incredible amount of time and effort spent on forecasting company earnings, it is baffling that the average (consensus) earnings forecast is consistently biased. Even more surprising is that we can identify a range of characteristics that are systematically related to these forecast biases. The number of analysts following a stock, the dispersion among their forecasts, the sector affiliation, and the nationality of a stock can be used to predict an upcoming earnings surprise. Predicted earnings surprises are significantly related to subsequent stock return, although the value of the prediction varies from sector to sector. A long-short portfolio strategy based on the prediction rule would therefore work well for some sectors, if not for others. Although on average there is value in the signal, a judicious implementation is required to fully benefit from its potential.