An analytical perspective on sporting events attendance: The 2007–2008 US NCAA college bowl games

Abstract

Modern statistical and spatial statistical methodology—based upon generalized linear, mixed and semivariogram/autoregressive modeling theory—offers modeling techniques that are particularly suitable for analyzing sporting event attendances. Counts of attendees must be non-negative, and should follow a Poisson frequency distribution. The rapid expansion in recent years of US NCAA college football bowl games raises the practical research question of whether or not bowl game attendance (i.e., counts) can be predicted from readily available simple measures, which include: a team's win–loss record, distance separating a team's school from the city hosting its bowl game, and the payout of a bowl game. These three sets of variables are employed as covariates for predicting the geographic variation from city-to-city of college football bowl game attendance for the 2007–2008 season. A principal finding of this study is that bowl game attendance appears to be predictable with a contemporary spatial statistical model that is a special case of a Poisson probability model, whose mean is a linear combination of payoff levels and distance of the closer team to a stadium, two factors over which individual bowl organizing officials have some control. This analysis supplements and extends research findings pertaining to the mapping of intercollegiate sports and the geography of visitor attendance at college football games, and offers insight into factors over which cities hosting bowl games have some control.