Arbitrage in Political Prediction Markets

Abstract

Online prediction markets are a powerful tool for aggregating information and show promise as predictive tools for uncertain outcomes, from sporting events to election results. However, these markets only serve as effective prediction tools so long as the market pricing remains efficient. We analyze the potential arbitrage profits derived from such mispricings in two leading American political prediction markets, PredictIt (for the 2016 and 2020 elections) and the Iowa Electronic Markets (for the 2016 election), to quantify the degree of mispricing and to show how market design can contribute to price distortion. We show that contracts hosted by PredictIt, compared to the IEM, are chronically mispriced, with large arbitrage profits in the 2016 election markets and non-negligible profits for the 2020 markets. We discuss the role of profit fees and contract limits, the primary differences between the PredictIt and IEM, in distorting pricing on PredictIt by limiting the ability of traders to capture arbitrage profits. Additionally, we examine the association between arbitrage and margin-linking, increased liquidity, and the number of unique contracts PredictIt's markets. This research provides cautionary evidence of potential inefficiencies in prediction markets with the intention of improving market implementation and enhancing market predictiveness.