Prediction markets vs polls – an examination of accuracy for the 2008 and 2012 elections

Abstract

Which technique is more accurate in predicting the outcome of U.S. presidential elections, polls or prediction markets? Several studies on this have been conducted in the past. We use market data and poll numbers, included adjusted version of the poll numbers, to reexamine this question based on the two last American presidential elections, in 2008 and 2012. We find that the market predictions outperformed the polls for these elections, and that adjusting the polls makes them less accurate relative to prediction markets, if anything.