ON THE SELECTION OF ERROR MEASURES FOR COMPARISONS AMONG FORECASTING METHODS

Clements and Hendry (1993) proposed the Generalized Forecast Error Sec ond Moment (GFESM) as an improvement to the Mean Square Error in compa ring forecasting performance across data series. They based their conc lusion on the fact that rankings based on GFESM remain unaltered if th e series are linearly transformed. In this paper, we argue that this e valuation ignores other important criteria. Also, their conclusions we re illustrated by a simulation study whose relationship to real data w as not obvious. Thirdly, prior empirical studies show that the mean sq uare error is an inappropriate measure to serve as a basis for compari son. This undermines the claims made for the GFESM.