Resolving the question of why the favourite-longshot bias exists in some betting markets but not in others

Abstract

Favourite-longshot bias (FLB) is a phenomenon in betting markets reported over many decades and in many jurisdictions, whereby market prices deviate systematically from their fundamental value; favourites/longshots being under-/over- value. Many studies have sought to explain the enduring presence of the FLB and its absence in some markets, but little consensus has been reached. This paper aims to settle this debate by providing a clear explanation for the FLB’s continued presence in most markets and its absence in others. We develop a model which suggests the bias may result from competition between bookmakers and with betting exchanges, combined with bettors’ greater demand elasticity with respect to favourites. Further, we propose that the FLB will be eliminated when informed traders dominate and transaction costs are low. We use the predictions of our model to develop hypotheses, which we test empirically by analysing how the bias develops over the course of the markets for 6,058 races run between August 2009 and August 2010, requiring the analysis of over 5.5 million market prices in total.

Our results confirm that three factors contribute to the existence of the FLB: the pricing decisions of bookmakers, the availability of information, and the level of transaction costs. This analysis helps to identify the roles played by previously suggested accounts of the bias. In particular, we show that, because of (a) competition between bookmakers themselves and with exchanges, and (b) bettors’ demand for competitive prices on favourites, bookmakers’ optimal pricing decisions necessarily lead to the FLB. Second, we show that there is FLB in exchange prices in the early stages of the market, which is not eliminated because of higher transaction costs (in the form of wider bid-ask spreads). Finally, we draw upon models of prediction markets to suggest that, when informed traders dominate, any FLB in betting exchanges is likely to be short-lived, and we find that this is the case.