The price of efficiency: examining the effects of payroll efficiency on Major League Baseball attendance

Abstract

From 1998 to 2008, the Oakland Athletics were the most payroll efficient team in Major League Baseball (MLB), recording the 5th highest win percentage despite having the 7th lowest payroll to work with. Their remarkable success has been widely attributed to the innovative strategies employed by general manager Billy Beane designed to identify undervalued statistics and players. However, these strategies also had the unintended effect of diminishing fan interest. An estimated 1600 fans per game were lost as a result of these payroll efficiency strategies, which produce teams composed of young, relatively unknown players and game tactics perceived as less exciting than more traditional approaches.

Keywords:

• payroll efficiency
• baseball
• attendance
• roster turnover
• Moneyball

JEL Classification:

• L83
• M21

Notes

¹ Unlike many professional sports, MLB does not enforce a salary cap, which creates large disparities in payroll and talent across teams.

² A linear regression model is used for simplicity, but a logarithmic model would provide a better fit since the marginal cost per win increases with each additional win.

³ TENURE is weighted by number of game appearances for each player. TENURE ² was also included to test the possibility that the effect on attendance is quadratic, but the variable was not significant.

⁴ ATTEND was calculated for every MLB team for the years 1991–2008, resulting in a total of 518 observations, excluding the inaugural seasons for the Colorado Rockies, the Florida Marlins, the Arizona Diamondbacks and the Tampa Bay Devil Rays due to lack of prior data. These years were chosen for the availability of team payroll, player salary and average ticket price data.

⁵ Prior to 1994, only the top team from each of the four divisions would make the playoffs, so the GBPLAYOFF was calculated as the difference between wins by the division leader and wins by each team in that division. After the 1994 season, two additional divisions were added, plus two wild card spots for the team in each league with the best record of teams that did not win their respective division. Because the 1994 season ended prematurely due to a player strike and no postseason was held, GBPLAYOFF was calculated the same in this season as in seasons prior to 1994. Since 1995, the variable used the lesser of the games behind the division leader and games behind the wild card leader in its calculation.

⁶ PAYROLL is fixed at the start of each season, so there is no risk of reverse causation.

⁷ While there is the potential for reverse causation with the variable ALLSTARS, it likely has only a weak effect since fan election accounts for less than a quarter (8 of 33) of the players selected to each All-Star roster.

⁸ From 1997 to 2001, club seats were included in computation. Premium seats have been excluded since 2002.

⁹ For seasons in which teams change stadiums midseason, the average of the two stadium capacities is used. In cases where capacity differs for night and day games, the mean capacity is used.

¹⁰ The K&S study used imputed average real ticket price by dividing total gate receipts by attendance, instead of the weighted average ticket price calculated by Team Marketing Research (TMR) used in this study. Because the TMR data are not available for seasons before 1991, the 1990 ticket prices are extrapolated assuming linear growth of ticket prices.

¹¹ The 1990 season is excluded due to lack of ticket price data.

¹² LEAGUE was dropped from Regressions 3, 4 and 5 because it entered insignificantly in each case.

¹³ Specifications using random effects and entity fixed effects were also estimated, but yielded insignificant results. Many of the payroll efficiency variables, especially BBPG and SBPG, reflect entity-specific strategies that remain constant over time, which results in their coefficients being eliminated under random effects and team fixed effects specifications.

¹⁴ The estimated coefficient for ticket price should theoretically be negative, but previous studies including Bruggink and Eaton (1996), Rascher (1999) and Demmert (1973) have reported insignificant or even positive coefficients. It is possible that fan loyalty is so strong that ticket demand is fairly inelastic.

¹⁵ The interval regression method assumes that the censored observations given our explanatory variables follow a homoskedastic normal distribution. The R ² values displayed for Regressions 4 and 5 are McKelvey and Zavoina pseudo-R ²s.

¹⁶ (719.8 fans per game) * ($11.30 average ticket price in 1989 dollars) * (81 home games per season) * (215.3/124 adjusting to 2008 price levels) = $1 143 523.

¹⁷ Oddly, PRICE has a positive impact on attendance in this specification.