ABSTRACT
The role of outcome uncertainty in increasing spectator appeal is a central focus of the academic literature on the conduct and regulation of professional sports leagues. If the result of a match is hard to predict, in that the two teams have similar ex ante probabilities of winning, then a close contest is likely and, because they value suspense, extra spectators are liable to be attracted to the stadium, raising club and league revenue. This simple hypothesis has been at the heart of the defense case in many court
proceedings, in America and Europe, where competition authorities have challenged
restrictive business practices such as collusion in the selling of television rights or limitations on the freedom of movement of players. Very often the defense has been successful, courts accepting that special measures are necessary to equalize financial, and therefore, playing resources across clubs. This in turn, leagues have argued, will promote outcome uncertainty, ensure closely fought games and therefore attract more spectators. But do evenly balanced matches in fact attract greater audiences? In the large liter-
ature reporting attempts to build regression models of attendances at matches in professional sports leagues, a measure of outcome uncertainty is typically included as one of the covariates. Borland and Macdonald (2003) tabulate the results of 18 such match-level studies (in various sports and on different continents) and reveal very mixed results, with only three cases where the hypothesis, that the size of crowd is affected by outcome uncertainty, is supported. Szymanski (2003) likewise describes the evidence as “far from unambiguous.” Out of 22 cases cited by Szymanski, only ten offer clear support for the outcome uncertainty hypothesis. For such a popular hypothesis, there is, then, curiously little empirical support. It
might be concluded that, within the range of outcome uncertainty actually observed in sports leagues, variations in it from game to game are not in fact important to spectators considering whether or not to attend. The defense case in competition proceedings against sports leagues would then look very thin. There is, however, an alternative possibility (unexplored in previous studies) to
account for the lack of explanatory power of outcome uncertainty measures in crowd size models. This is that the proxies for outcome uncertainty used by the various authors typically fail to capture accurately the likelihood that a particular game will be close. Our starting point, in the spirit of the rational expectations paradigm (whereby the average forecast among a large number of people tends to be as correct as it could be given available information), is that sports fans in the aggregate are likely to have good intuition as to how closely contested a particular game is likely to be (if indeed this is capable of being forecast at all). A valid measure of outcome uncertainty (that will capture the behavior of potential spectators) will then be one that is effective as a predictor of a close game. The attendance literature features several different measures for ex ante outcome uncertainty but none of the studies report testing whether the proxy used is in fact highly correlated with match outcomes. In this chapter, we will review the several measures of outcome uncertainty found
in statistical analyses of attendance. For each, we will test whether it is actually an effective forecasting tool, where the object is to forecast closeness of contest. Further, we will assess whether any one measure outperforms the others to an extent such that findings based on it should be accorded greater weight when reviewing the existing attendance literature. Our work may also provide guidance to future researchers as to the most suitable outcome uncertainty measure for adoption in statistical modelling of crowd size.
