Optimal Length of Moving Average to Forecast Futures Basis

The question addressed in this study is which length of historical moving average provides the best forecast of futures basis. Differences in observed forecast accuracy among the different moving averages are usually less than a cent per bushel, and most are not statistically significant. Further, the search for an optimal length of moving average may be futile since the optimal length depends on how much structural change has occurred. Our recommendation is to use moving averages when there has been no structural change and to use last year’s basis or an alternative approach if the forecaster perceives that a structural change has occurred.