COMBINING FORECASTS FOR FORECASTING U.S. PRESDIENTIAL ELECTIONS
Bucharest Dialogues on Expert Knowledge, Prediction, Forecasting: A Social Sciences Perspective, November 2010.
Abstract. Prior research suggested that accuracy gains from combining forecasts are particularly high if one uses forecasts from different methods that draw upon different data. We tested the magnitude of effects by combining forecasts from four methods (polls, a prediction market, econometric models, and expert judgments) used for predicting the five U.S. Presidential Elections between 1992 and 2008. The gains from combining were substantially larger than the 12% error reduction obtained from a prior meta-analysis. Compared to the typical uncombined forecast, the combining procedures yielded error reductions ranging from 10% to 58%. Combining is probably the most cost efficient way to improve forecast accuracy and to prevent large errors.
Click here for the full paper.
