eThesis - School of Business electronic theses

PURPOSE OF THE STUDY:

Modern exchange betting markets represent an interesting phenomenon for economists since they closely resemble traditional financial markets. As such, they provide a practical real-life setup for testing the efficient market hypothesis. Unlike many financial assets, bets have finite life periods after which their true payoffs are revealed. Nevertheless, research on exchange betting markets is scarce to date. This thesis aspires to add to the previous literature by providing a throughout investigation of the efficiency of the exchange betting market for men's professional tennis. In particular, I examine whether the market is characterized by a favorite-longshot bias.

DATA AND METHODOLOGY:

My analyses are based on a novel sample of pre-match exchange betting market odds for 7,555 men's professional singles tennis matches from seasons 2010-2013. The market odds in my dataset are market lines, or odds quotes that prevailed at the close of betting in Betfair's betting exchange. I obtain the data from Fracsoft, an official Betfair data vendor.

I perform three distinct statistical tests of market efficiency. First, I classify the observations into odds groups and test whether the subjective probabilities derived from market odds equal the empirical probabilities estimated from ex-post match results. I use multiple odds group classifications to improve the robustness of my results. Second, I compute the rates of return of betting at different odds levels. Finally, I model the relationship between the objective and the subjective probability. To begin with, I estimate a linear model based on the classified data by the OLS technique. Then I proceed to formulate a logit model where I explain the match results by the market odds.

RESULTS:

By and large, my results cannot reject market efficiency. None of the statistical tests that I perform can reject market efficiency. The exchange betting market for men's professional tennis appears to function well. Overall, the subjective probabilities predict the objective probabilities very well indeed. Notably, I do not find evidence of a favorite-longshots bias. Thus my results overshadow many previously suggested explanations for the FLB. As such, my findings support cost and information-model explanations for the favorite-longshot bias.