Optimal Length of Moving Average to Forecast Futures Basis

Futures prices when combined with a basis forecast provide a reliable way to forecast cash prices. The most popular method of forecasting basis is historical moving averages. Given the recent failure of longer moving averages proposed by previous studies, this research reassesses past recommendations about the best length of moving average to use in forecasting basis. This research compares practical preharvest and storage period basis forecasts for hard wheat, soft wheat, corn and soybeans to identify the optimal amount of historical information to include in moving average forecasts. Only with preharvest hard wheat forecasts are the best moving averages longer than 3 years. The differences in forecast accuracy among the different moving averages are small and in most cases the differences are not statistically significant. The recommendation is to use longer moving averages during time periods (or at locations) when there have been no structural changes and use last year’s basis after it appears that a structural change has occurred