Are Sports Betting Markets Prediction Markets?

Abstract

Researchers commonly use sports betting lines as predictions of the outcome of sporting events. Betting houses set betting lines conditional on bettors ex ante beliefs about game outcomes, which implies that the predictive power of the sports betting market could be an unintended consequence of betting house profit maximization. Using this insight, the authors propose a new test of the predictive power of the sports betting market, which incorporates a seldom-used piece of complementary betting information: the over/under—the predicted sum of scores for a game. Since the over/under has the same market properties as the betting line, it should be similarly predictive about the actual outcome, while if bettors have different beliefs about this game feature it need not be predictive. Using the universe of betting lines and over/unders on National Football League (NFL), National Basketball Association (NBA), National Collegiate Athletic Association (NCAA) college football, and NCAA college basketball games from 2004 to 2010, the authors test the predictive power of the sports betting market in a seemingly unrelated regression (SUR) structure that allows us to characterize both features of the betting market simultaneously. The joint test reveals that while the betting line is an accurate predictor of the margin of victory, the over/under is a poor predictor of the sum of scores. The authors consistently reject the hypothesis that the sports betting market overall functions well as a prediction market.