SYSTEM-BASED WEIGHTS VERSUS SERIES-SPECIFIC WEIGHTS IN THE COMBINATION OF FORECASTS

System-based combination weights for series r/step-length h incorporat e relative accuracy information from other forecast step-lengths for r and from other series for step-length h. Such weights are examined ut ilizing the West and Fullerton (1996) data set-4275 ex ante employment forecasts from structural simultaneous equation econometric models fo r 19 metropolitan areas at 10 quarterly step-lengths and a parallel se t of 4275 ARIMA forecasts. The system-based weights yielded combined f orecasts of higher average accuracy and lower risk of large inaccuracy than seven alternative strategies: (1) averaging; (2) relative MSE we ights; (3) outperformance (per cent best) weights; (4) Bates and Grang er (1969) optimal weights with a convexity constraint imposed; (5) unc onstrained optimal weights; (6) select a 'best' method (ex ante) by se ries and; (7) experiment in the Bischoff (1989) sense and select eithe r method (2) or (6) based on the outcome of e experiment. Accuracy gai ns of the system-based combination were concentrated at step-lengths t wo to five. Although alternative (5) was generally outperformed, none of the six other alternatives was systematically most accurate when ev aluated relative to each other This contrasts with Bischoff's (1989) r esults that held promise for an empirically applicable guideline to de termine whether or not to combine.