A Long Run Look at FBS Football Attendance

A balanced panel (99 teams over 40 years) is used to estimate three regression models: average attendance via fixed and random effects plus Tobit estimation of percent of capacity. Variables are either stationary or cointegrated. Estimation makes adjustments for serial correlation and endogeneity between several variables. Independent variables measure economic conditions, demographic characteristics, and team performance. Attendance is a normal good. Travel cost is insignificant in two models but positive in one. Weak evidence suggests undergraduate enrollment and county population exert a positive impact. Three team performance measures and power ranking scores exert consistent positive impacts. Stadium renovations that increase capacity also increase attendance while those that reduce capacity are insignificant except in the percent of capacity equation. Attendance responses vary, either in sign or significance, across Power Five and Group of Five teams for eight of 10 independent variables. Specification tests suggest that random effects estimation is preferred.