Structural Change in Competitive Balance in Big-Time College Football

The idea that balance matters to fans can be traced in economics back to Rottenberg (1956).

Goff (2004) reviews a substantial literature on these other values. The literature that followed included Tucker (2004, 2005), Humphreys (2006), Tucker and Amato (2006), Humphreys and Mondello (2007), Stinson and Howard (2007), Pope and Pope (2008), Smith (2008), and Fort and Winfree (2013).

The BP Approach has been used on all four North American pro sports leagues and the English Premier League. On competitive balance, see Lee and Fort (2005, 2012) and Fort and Lee (2007). On attendance, see Lee and Fort (2008) and Mills and Fort (2014).

Scholarship limits and the determination of the national champion, as well as passing notice about the introduction of the GI Bill, are covered later in the paper.

Presentation of the entire results would result in a paper that is unsuited to the usual journal length. The full test and regression results are in the Data and Statistical Tests Appendix available online at https://​www.​dropbox.​com/​s/​reo9ym5uoohbmym/​SalagaFortAppend​ix.​docx?​dl=​0.

Others might take the history of the Big 12 back to the Missouri Valley Intercollegiate Athletic Association (1907–1927) and the Pac-12 only back to the Athletic Association of Western Universities (1959–1967), rather than including the Pacific Coast Conference (1916–1958) that was broken up due to scandal.

College conference championship games have only been around since Arkansas and South Carolina joined the SEC in 1992–1993. One could envision “post season” access to bowl games, but the purpose of nearly none of them is to crown a champion. Indeed there was no national championship game, per se, until the Bowl Championship Series put one in place for the 2007 bowl season.

The full data and calculation results for historical conference championships are in Tables A1–A5 in the online appendix (see footnote 6).

The “tail likelihood” measure from Fort and Quirk (1995) puts the attention on the outliers (big losers and big winners) rather than on outcomes around the mean of the winning percentage. In the data to follow, RSD takes on values that are much lower than in sports such as baseball or basketball (Fort 2011, Chapter 6), which suggests that there really isn’t as much going on in the tails for college football competitive balance.

Andrews (1993) focused on a single structural change, while Bai (1997, 1999) and Bai and Perron (1998, 2003) moved on to consider multiple structural changes.

Sequential tests allow for the ability to determine the optimal number of breaks within a given series. The UDMax and WDMax tests identify the presence of breaks (with no set maximum number of breaks) against the null of zero breaks. The Sup F _T (k) tests specify statistical significance for a given number of breaks (k) against a null of zero breaks.

Bounded time series such as ours can render unit root tests misleading. However, in actuality, none of our series hit the boundaries so unit root testing remains insightful. Our reference is Cavaliere and Xu (2014).

ADF and PP tests are the standard choice for unit root testing in the BP literature and test results are in Table A6 in the online appendix (see footnote 6).

Break point test results are in Tables A7–A9 in the online appendix for the results for each of \(RSD_{t}\), \(MVR_{t}\), and \(WPC_{t}\), respectively (see footnote 6). As a robustness check, we tried to recreate the significance and direction of the break-point results using generalized linear modeling (GLM). Conference-level measures were the dependent variable, and indicators for the years at which we find break points, plus a constant, and a trend were the explanatory variables. For all of the series where we find a single break point using the BP Approach, our GLM results give the same results for both significance and direction of the break point. For the series where we find two break points, our GLM results give the same the direction of the breaks, and significance occurs in 3 of 6 cases. The BP Approach and GLM results are in general agreement.

The regression results are in Table A10 in the online appendix for the results for each of \(RSD_{t}\), \(MVR_{t}\), and \(WPC_{t}\), respectively (see footnote 6).

The actual and fitted results for all Power 5 conferences, including those not in Figure 1, are in Figures A1–A3 in the online appendix (see footnote 6).

It may be the case that our measures of competitive balance are not robust enough to show the impact of events. However, in the many other works that are referenced as using these variables, the measures were insightful.

To follow part of this discussion, the reader will need to return to Figures A1–A3 in the online appendix (see footnote 6).

Salaga (2015) analyzed the GI Bill (1946), a general national education policy measure thought to have influenced balance in college football. Our work reinforces his episode analysis finding that the GI Bill did not coincide with changes in competitive balance. However, Mills and Salaga (2015) do find breakpoints in NCAA college basketball that align closely with the implementation of the GI Bill.

Quirk (2004) found short-term impacts on balance with conference switching. The work cited above on pro sports also found the pro version of conference switching—expansion and relocation—and integration to coincide with break points. With respect to integration, Mills and Salaga (2015) find a number of breaks proximate to this time period in college basketball.

The grant-in-aid allowed coverage of tuition, room and board, and books. See Falla (1981) and Byers (1995).

The invariance principle holds that the distribution of talent is invariant with respect to who keeps the value created by athletes. Talent goes to its highest valued use whether athletes are paid their marginal revenue product or less due to labor market restrictions. Our finding also reinforces the episode analysis in Salaga (2015) that finds grants-in-aid had no impact on balance in college football.

468 US 85 (1984).

Scholarship limits: Sutter and Winkler (2003). TV deregulation: Bennett and Fizel (1995), Eckard (1998), Siegfried and Burba (2004), and Salaga (2015). The BCS formation is common knowledge and its website was still in operation as of this writing (www.​bcsfootball.​org).

For example, on coaches, no break points were associated with other outlier performances such as: Fielding Yost at the University of Michigan (1901–1926, pre-bowls, 6 National Championships including 4 in a row, 1901–1904); John McKay at the University of Southern California (1960–1975, 9 bowls, 4 National Championships); Bear Bryant at the University of Alabama (1958–1982, 24 bowls, 6 National Championships); or Joe Paterno at Pennsylvania State University (1966–2011, 37 bowls, 2 National Championships).