Prediction Market Accuracy: The Impact Of Size, Incentives, Context And Interpretation

Enterprises desiring to utilize prediction markets for decision support must consider numerous design factors for their market deployments. Through logistic regression analyses of more than 350 real and play-money prediction markets, this paper evaluates several design issues in order to identify conditions under which prediction markets can effectively contribute to an enterprise decision support process. Two of these design considerations include the size of the trader pool and the nature of trader incentives. We find that varying the number of market traders has minimal accuracy impact for markets exceeding 10-20 traders and that the impact of financial incentives is contextual and beneficial to market accuracy. When to act upon market results and at what levels of market support must also be considered. Our data shows that acting on market output up to three weeks prior to an event’s occurrence and requiring markets to sustain desired price levels for up to 3 weeks before responding to the market signal did not statistically impact the market’s ability to accurately predict an event’s occurrence. Adjusting the price threshold for market recommendation acceptance to levels between $55 and $80 ($45 and $20) also does not negatively impact market accuracy. A related measure capturing market uncertainty was found to be the leading predictor of a market’s failure to provide accurate predictions.