An analysis of persistence in analyst's relative forecast accuracy

We examine the persistence in analysts' relative earnings forecast accuracy. When analysts are ranked into forecast accuracy quintiles, calculated over all the firms they cover in each year, we find that 52% (45%) of superior (inferior) analysts, i.e. analysts in the lowest (highest) quintile, remain in this quintile in the subsequent period. We show that a variable we develop and denote as forecasting complexity, i.e. the extent to which analysts' earnings forecasts vary when predicting a firm's earnings, is important in explaining variation in the persistence of the relative forecast accuracy of analysts. When we control for forecasting complexity, the probability of analyst relative forecast accuracy to persist is reduced by about half. This reduced persistence, however, measures true forecasting ability. When we form portfolios using recommendations of analysts identified as superior in two consecutive periods, controlling for forecasting complexity, we find significant abnormal returns after adjusting for the Fama--French and momentum factors.