Abstract This paper conducts an empirical analysis of the approaches to obtaining linear combinations of forecasts. Simulated quarterly earnings were modeled using three ARIMA models. One-quarter ahead forecasts were then developed. These forecasts were combined using alternative approaches. The most accurate forecasts were obtained by adding a constant term and not constraining the weights to add up to one. The differences in the accuracy rankings were found to be statistically significant. The results are similar to those obtained by Granger and Ramanathan (1984).