College football attendance: a panel study of the Football Bowl Subdivision

Abstract

Panel data with an instrumented real ticket price are used to estimate a regular season game-day attendance and per cent of capacity regression equations. Better team performance, whether short term (season wins), intermediate term (bowls games in last 10 years) or long term (lifetime winning percentage), higher undergraduate enrolment, traditional rivalries and video coverage increase per cent of capacity used. Poor weather (more rain or cloud cover), higher travel costs and larger local population decrease it. Fan interest wanes as a season progresses, but this is offset as a team wins more games. Games played near a National Football League stadium, those with conference opponents, non-FBS opponents and non-BCS opponents have lower stadium utilization. The substantive results of the analysis do not change when attendance is used as the dependent variable rather than per cent of capacity.

Keywords:

• college football
• game-day attendance
• ticket price
• stadium utilization

JEL Classification:

• L83
• D12
• D21

Notes

¹ Teams from FBS conferences whose champions do not automatically qualify for a BCS bowl game are identified as a ‘non-BCS team’.

² Stadium capacity is that reported by the university to the NCAA and published each year in the NCAA Yearbook. Capacity is not the physical limits of the stadium. In approximately 25% of the games stadium utilization exceeds 100% of this official NCAA stadium capacity. Assuming that open-admission student sections and standing-room-only sales could account for some cases of ‘overcapacity’, we checked the validity of any observation that exceeded 110% of the NCAA stadium capacity. In all cases, supplementary information indicated that the attendance figures were valid. Take, for example, the 161.7% figure reported in Table 1 where the University of South Florida (USF) hosted West Virginia University on 28 September 2007. The USF plays their home football games at Raymond James Stadium, which has a capacity of 67 857. In 2007, USF used only the lower deck and thus listed their stadium capacity as 41 441. The upper deck was apparently opened for the West Virginia game, and thus attendance was reported as 67 012.

³ For a more complete description of the variables, see ‘Appendix A’. All nominal monetary figures are converted to real dollars using a monthly regional consumer price index.

⁴ Gasoline prices are available for seven regions in the United States: New England, East Coast, Central Atlantic, Gulf Coast, Midwest, Rocky Mountain and West Coast.

⁵ Ticket price information primarily was collected from a school’s athletic department website via: (1) ticket office information; (2) annual ticket brochures; and (3) archived news releases. Some athletic department ticket managers were contacted via e-mail and many kindly responded to specific requests for information.

⁶ Of the 4317 observations, 2278 (53%) had an observed value for single game price. The remainder of the observations (47%) used an imputed price. The several methods used to impute these prices and the per cent of missing observations replaced by each imputation method were: (1) single game price of a different game from the current season at the same school (15%); (2) season ticket price divided by the number of home games from the current season at the same school (25%); (3) single game price from a different season at the same school (47%); (4) season ticket price divided by the number of home games from a different season at the same school (9%); and (5) other schools’ ticket prices in the same season (5%). These percentages do not sum to 100% due to rounding error.

⁷ By comparison, Price and Sen (2003) define their ticket price variable as the ‘ticket price for sideline seating’. This definition declines to elaborate whether this price is the average price for a season ticket holder or the price for a single game reserved seat or a general admission seat includes a license fee or mandatory minimum donation to the university, or where this seat is found relative to the 50-yard line.

⁸ For this set the Big Ten conference is the default conference (one of the co-authors is a Purdue graduate).

⁹ If television executives choose games to broadcast based on fan interest and fan interest also drives attendance then endogeneity may be present in the model. This endogeneity may be muted by several factors. The two decisions (i.e. fan attendance and television broadcasts) are made by two separate groups at two different times (i.e. fans on the day of the game versus television executives months or weeks in advance of game day). The existence of a telecast will be known by fans before they make a decision to attend the game while the television decision is made before actual attendance is known. Although both decisions are driven by fan interest, the television decision is based on expected fan interest which is likely influenced by long run factors such as long run team performance and traditional rivalries. However, the game day fan attendance decision is based on, among other factors, short run phenomena such a recent team performance and current weather.These endogeneity effects are confounded by the measurement of attendance: tickets sold versus people through the turnstiles. Take, for example, a televised game. The option of watching the game on television (a substitute) could reduce official attendance figures if attendance is measured by the number of people entering the stadium – more fans opt to stay at home and watch on television. There would be less impact on official attendance figures if attendance is measured by ticket sales – people purchasing tickets in advance would be counted whether they entered the stadium or watched the game on television. Television broadcasts also could influence the number of fans purchasing tickets on the day of the game.We were unable to determine which method each school chose for reporting attendance (turnstile count or tickets sold) in each year of the sample period with the exception of 2004 when all schools were required to report to the NCAA the turnstile counts as the official figure. In addition, we were unable to determine the number of tickets sold in advance versus those sold on the day of the game.

¹⁰ A substantial majority (87%) of these video broadcasts were either via major network broadcast or cable television. Of all video broadcasts that were recorded in the data: 14% were over-the-air broadcasts on a major network (regional or national coverage); 73% were available via a cable television channel; 4% via a local television station; 4% via pay-per-view; and 4% via internet video streaming. Less than 1% of games were broadcast over four different types of video sources; 1% over three sources; 11% over two sources; and 87% over a single source.

¹¹ The authors also used an alternative set of data to instrument ticket prices. The Equity in Athletics web-site (http://ope.ed.gov.athletics/) provides a comprehensive set of data on athletic department operations including measures of revenue and cost. Statistical testing indicated, however, that stadium capacity did a better job of instrumenting ticket prices than any other set of alternative instruments.