THE EFFECT OF CONTRACT STRUCTURE ON PREDICTION MARKET PRICE BIASES

Abstract

Prediction markets add value when they produce unbiased forecasts. However, several prior studies find persistent biases when examining prediction market sides contracts. Sides contracts represent bets on whether the score differential between two teams in a contest will be greater or less than a stated value. We propose that inferences generated from examining Tradesports’ sides contracts may be problematic because they are framed exclusively with respect to favorites. If a favorite-longshot (or reverse favorite-longshot) bias causes these deviations from rationality, it may be that non-sports-related (e.g., internal corporate) prediction markets assets do not suffer from the same shortcomings. To evaluate the generalizability of prior findings, we contrast the price efficiency of Tradesports’ sides and totals contracts. In totals wagers, traders take a position on whether the combine score of both teams in a game will be above or below a stated value. We find that the fundamental structural differences between totals contracts and sides contracts partly determine differences in price efficiencies. Relative to those in the sides market, some price biases in the totals market are significantly smaller in magnitude, and others are absent altogether. Results indicate that contract structure plays a significant role in the ability of prediction markets to produce unbiased estimates.