Divergence of opinion and the information content in point spreads
This study provides a thorough investigation of several aspects of football betting market research and examines two alternative hypothesis of financial market theory. The study is conducted using data from the betting market of the National Football League and the relatively unexplored college football betting market.
In the area of football betting market research, a new specification that corrects for an improper restriction imposed in the previous literature is developed and subsequently examined. The results presented are consistent with efficient markets theory and suggest that the previous specification biased the results towards rejection of market efficiency.
A second specification issue addressed is the use of Ordinary Least Squares. Since the important result in football betting markets is winning a wager, it is argued that logistic regression is a more appropriate methodology to use. All research is performed redundantly using both regression methodologies. The wagering simulation results using forecasts from OLS and logistic models indicate an advantage for the logistic models in producing winning wagers.
Rationality of public wagering is examined in each of the two betting markets using a variable constructed from the information provided by betting line movements. The results presented do not confirm the irrational public wagering results reported by previous authors and suggest that betting line movements are attempts by bookmakers to incorporate new information into the betting lines – a result consistent with the Efficient Markets Hypothesis. The college football betting is further examined using proxy variables constructed from information taken from the Associated Press' AP Top Twenty Poll and the Dunkel Index. Little to no evidence against market efficiency of this information is determined.
Miller's "Divergence of Opinion" is merged with Canes' propositions of individual's wagering behaviors to form joint propositions of potential market inefficiencies related to differences in team fan characteristics. These propositions are examined using variables proxying for the differences between teams fan wealth and/or numbers. The evidence does not support the joint hypotheses formulated; however, the results are consistent with Miller's theory and also the widely-held belief that low income individuals are those individuals more likely to wager.