Markets vs. polls as election predictors: An historical assessment

Prediction markets have drawn considerable attention in recent years as a tool for forecasting elections. But how accurate are they? Do they outperform the polls, as some scholars argue? Do prices in election markets carry information beyond the horserace in the latest polls? This paper assesses the accuracy of US presidential election betting markets in years before and after opinion polling was introduced. Our results are provocative. First, we find that market prices are far better predictors in the period without polls than when polls were available. Second, we find that market prices of the pre-poll era predicted elections almost on par with polls following the introduction of scientific polling. Finally, when we have both market prices and polls, prices add nothing to election prediction beyond polls. To be sure, early election markets were (surprisingly) good at extracting campaign information without scientific polling to guide them. For more recent markets, candidate prices largely follow the polls.

► This article compares the performance of prediction markets and vote intention polls as election predictors.
►  The analysis goes back in time to assess the historical performance of presidential election markets in the US since 1880.
► Using the data, it is possible to evaluate the performance of prediction markets both before and after the introduction of scientific polling.
► Surprisingly, markets predicted better without polls, partly because pre-poll markets were good at extracting essential information about electoral forces at work.
► Once polls were introduced, market prices mostly followed the polls, even when they erred dramatically, as in 1948.