Tuesday, November 4, 2014

Combining Interval Forecasts

GABA Anil, TSETLIN Ilia, WINKLER Robert L.
Read the working paper
INSEAD Working 2014/58/DSC

In combining forecasts, a simple average of the forecasts performs well, often better than more sophisticated methods. In a prescriptive spirit, we consider some other parsimonious, easy-to-use heuristics for combining interval forecasts and compare their performance with the benchmark provided by the simple average, using real-life data sets consisting of forecasts made by professionals in their domain of expertise. The relative performance of the heuristics is influenced by the degree of overconfidence in the experts’ intervals. With a moderate to high degree of overconfidence, two of the heuristics outperform the simple average, with the best creating wider combined intervals and locating the intervals better in terms of the accuracy of the midpoints of the intervals. If there is not much overconfidence, the median and simple average perform best. The results provide some easy-to-use alternatives to the simple average, with an indication of when each might be preferable.