Analysts forecast error: A robust prediction model and its short term trading

We examine the profitability of implementing a short term trading strategy based on predicting the error in analysts' earnings per share forecasts using publicly available information. Since large earnings surprises may lead to extreme values in the forecast error series that disrupt their smooth autoregressive dynamics, we propose to use robust fixed effect panel estimation methods as an alternative to panel least squares or the pooled least absolute deviations estimator. For the I/B/E/S data from 1998 to 2010, we show that the strategy of taking a long (short) position in stocks with the most pessimistic (optimistic) consensus forecast and closing the position on the first post announcement day has an annual gross abnormal return of 16.56%, after correcting for market risk, size, book-to-market and price momentum effects. A key insight is that the profitability of the trading strategy stems from using robust forecasting methods.