View article

Recent studies have shown that composite forecasting produces superior forecasts when compared to individual forecasts. This paper extends the existing literature by employing latent root regression and robust-weighting techniques in composite model building of corporate earnings per share series. Security analysts' forecasts may be improved when combined with time series forecasts for a diversified sample of 648 year-end firms with a 1981-1985 postsample estimation period. The mean square error of analysts' forecasts may be reduced by combining analyst and univariate time series model forecasts in an ordinary least squares regression model. This reduction is very interesting when one finds that the univariate time series model forecasts are produced by ARIMA (0, 1, 1) random walk with drift processes. Latent root regression and robust weighting reduce forecasting errors relative to the ordinary least squares weighting scheme. Multicollinearity exists between analysts' and time series model forecasts and latent root regression techniques are used to estimate composite earnings models. Composite earnings models product statistically significant excess returns.