College football attendance in the long run: the FCS

A balanced panel of FCS football teams (61 teams, 38 years) was used to identify variables that drive per-game season attendance. Variables were either stationary or co-integrated. Four regression models were estimated: pooled ordinary least squares, pooled two-stage least squares, fixed effects, random effects. Most coefficients were consistent across models: season win percentage, lifetime win percentage, playoff game appearances, undergraduate enrollment and real gas price per mile driven were positive while real state per capita income exerted a negative impact. Two coefficients displayed mixed results: state unemployment rate was negative but sometimes insignificant; county population was positive but sometimes insignificant.