Earnings Estimates and the Accuracy of Expectational Data

This paper examines the accuracy of forecasts produced by mechanical forecasting techniques and three groups of analysts. The nine mechanical forecasting techniques are variations of exponentially weighted moving averages, naive models, simple moving averages, and regressions. One-, two- and three-year forecasts are used to evaluate these techniques. The mechanical techniques exhibit statistically significant differences in their ability to forecast earnings per share, with the exponentially weighted moving averages producing the best forecasts. One-year forecasts produced by the best of the mechanical forecasting techniques were compared to the corresponding analysts' projections. No statistically significant difference could be discerned.