Powered by

How The $200+ Million Settlement For COA Payments Was Calculated

By Daniel A. Rascher, Ph.D. (OSKR, LLC and University of San Francisco), Andy Schwarz (OSKR, LLC)

As economists who have studied questions surrounding the need for agreements among schools as to the maximum allowed grant-in-aid (GIA), we’re often asked to figure out exactly which schools can “afford” to pay athletes more than they do now.  Specifically, we wonder whether college sports function if more market-oriented means of recruitment and compensation were employed for athletes, just as they are for coaches, staff, and administrators.  The NCAA’s annual finances often show a deficit between so-called “generated revenues” and listed expenses. From this, ADs and other insiders seem to conclude that hardly any school has the resources to pay a dime more than they do now.  Until 2015-16, there’s never really been a direct way to put that question to the test — can schools that seem to be losing money, at least according to the MFRS data they send into the NCAA, afford to increase their scholarship expenses?


However, such an opportunity finally arrived with the decision by the Division I schools to change the way scholarship caps are determined.  In August of 2014, the NCAA adopted a new form of governance, where the Power 5 (P5) schools were allowed, for the first time, some degree of freedom in the terms of how they set the GIA maximum.  Rather than requiring a majority vote of all of Division I, now the P5 could vote among itself, and in January 2015, they did so, agreeing to allow Full Cost of Attendance (COA) GIAs, rather than the previous “full ride” that was not full.  No school was required to do so (although a few conferences then did choose to mandate COA for their members) but as of August 2015, schools in each of the P5 conferences began offering Full COA scholarship, and non-P5 conference then had the freedom to choose whether to make similar offers or to stay at the old GIA level instead.


Additionally, there were economic data to test whether having an accounting deficit was really an impediment to higher compensation or if instead, schools would act as economics would predict, making an investment decision to shift money to athletes if the return on that investment, financially or otherwise, so dictated.  This article lays out how Dan (with help from Andy) built a sophisticated but elegant econometric model (i.e., a statistical model based on economic theory) designed to test whether schools’ conduct could be predicted on the basis of the same sort of economic variables that would drive any other business decision regarding the price of labor, or whether the existing budgetary constraints made schools immune to those business forces.


In the recent GIA Cap antitrust case,[1] Dan was asked to estimate which schools and teams would have paid Full COA back as early as the 2009-10 academic year if it had been allowed.  The analysis looked at the economic context in 2015-16 for the schools that did and did not immediately offer Full COA, created an econometric model based on identifying key economic factors that correlated with the yes/no decision, and applied that model to the 2009-10 economic context of those schools.  The idea was first to identify the best economic factors that predicted which schools would choose to compete more intensely for athletes if allowed (by adopting COA), and which would continue to compete, but less intensely (by remaining at COA).  This was done for both men’s basketball and football (with a simpler, but related analysis done for women’s basketball).


The analysis was based on an econometric technique known as “probit regression.”  Probit is a form of what economists call a “discrete choice model” that is well suited to regressions focused on a yes/no decision, e.g., to adopt GIAs above the maximum athletic aid cap from the pre-2015 cap or not.  The probit regression was used to model past and present COA adoption rates.  As a first test, the model was used to “predict” which schools would adopt COA in 2015-16, without telling the model which schools had actually done so.  The model was very accurate, only incorrectly predicting a school would immediately adopt COA when it did not 6% of the time (with many of those schools adopting (or stating an intention to adopt) COA in subsequent years).  This showed that the model’s estimation of the economic factors underlying the decision were good predictors of schools’ conduct.  The model was then used to provide a reliable prediction of which schools within FBS football and Division I basketball would have paid athletic grants-in-aid at levels in excess of the pre-2015 cap for the period from 2009-10 forward, based on data contemporaneous with each year’s prediction.


Generally speaking, the results of this model identified schools as likely to adopt COA starting in 2009-10 if they adopted COA in 2015-16, but not simply because we knew the adopted it in 2015 (once calibrated, the model had no idea which schools had or had not adopted COA in the calibration year).  Instead, the model used factors pertaining to revenues, expenses, recruiting success, etc., to develop a predictive means of assessing a school’s competitive situation and generating a predicted yes/no decision.


The independent variables (those used to predict whether a school would offer COA) included:


  • The sum of the full scholarships equivalents given to overall counters from the Squad Lists for 2014-15, as a direct measure of a school’s payments to its athletes. The lagged year is used in order to be able to compare to the earlier years in the damages period, where both years are unaffected by COA payments.
  • The average across all members of the school’s conference (other than the school itself) of each schools’ total recruiting stars divided by the FBS (or D1) average number of stars. This provides information on how competitive the schools’ conference is compared to the FBS or D1 average.  Also, the “school’s recruiting success” as measured by the total stars (as measured by rivals.com) of the new recruits who committed to the school.
  • The number of conferences during the damages period that the school (team) was in (i.e., a school that stays in the same conference the entire period would have a 1 for this variable),[2] as well as whether the school changed conferences during the given year.
  • The ratio of athletic department revenue to expenses during 2014-15. Also, the difference in the athletic department’s revenues and expenses as well as the team’s revenues and expenses during 2014-15. 2015-16 data was not yet available.
  • The team’s budget during 2014-15. 2015-16 data was not yet available.
  • The ratio of the athletic department’s expenses compared to the median during 2014-15, as well as the team’s ratio compared to the median. 2015-16 data was not yet available.
  • The change (in dollars) per year in the team’s budget (i.e., 2014-15 minus 2013-14). Also, the percentage change in the same variables (to account for differences in size of programs and across sports).
  • The compounded annual growth rate in the athletic department’s budget for 2012-13, 2013-14, and 2014-15, as well as that for the team’s budget.
  • The COA Gap (i.e., the difference b/w the new maximum GIA and the old maximum) multiplied by either 85 for football or 13 for basketball divided by athletic department expenses during 2014-15. This is a measure of the cost of providing COA payments compared to what is already being spent in athletics.  Also, the COA Gap divided by the recent growth in athletics department expenses, to account for the growth in investment in athletics each year.
  • A measure of whether or not a school was on probation in the year in question and was thus limited in its ability to give scholarships to its athletes.

The result is a model which provides a prediction of schools likely to have adopted COA payments in 2009-10 had the alleged restraints in suit never been enacted by Defendants.


As part of the work on that model, we developed also an algorithm to identify all class members based on Plaintiffs’ proposed class definition.  By combining the results of the algorithm for identifying class members, and the econometric method of assessing impact, the model was able to demonstrate a class-wide method for assessing impact for every class member, and then also to demonstrate how their damages would be estimated by means of a reasonable and non-speculative formula.


This was done using data produced by the Defendants and their member schools, and with government data on COA gaps in those limited cases where discovery was still incomplete.  (Those data were provided confidentially, and so we cannot share the full details of the model, but we hope to recreate as much as possible of our work using public data.) The COA gap was defined as being equal to the difference between (a) the relevant average Cost of Attendance for each class member for each academic year and (b) the sum of an estimate of all athletic and non-athletic financial aid provided to the athlete for that same academic year.  Excluded from these calculations were the receipt of Pell Grants and/or any payments identified as coming from the Student-Athlete Opportunity Fund (SAOF) or the Student Assistance Fund (SAF), so that the measurement of this gap was not affected by the receipt of these funds.  (That is, if two students each had a $3,000 gap and one of them got SAF money of $2,500 and also had a $5,700 Pell Grant, the gap remained at $3,000 for each).


Using that definition and those assumptions, the estimated total class-wide damages for the three classes of athletes (prior to any trebling) attending schools identified by the probit model’s predictions was approximately $210 million to $220 million (or about five to six thousand dollars per four-year athlete ).  This total was based on the academic years 2009-10 through 2015-16.  The parties ultimately settled this piece of the litigation for a little more than $208 million, with this payment designed to cover the athletes’ losses and attorneys’ fees.


Leaving aside the litigation outcome, from an economic perspective, the most interesting conclusion is how many schools were able to come up with the funding to cover some or all of the COA “wage” bill, and how consistent their conduct has been with how a normal for-profit business would act.  Schools with low revenue generation potential were far less likely to make the jump, even if their listed “deficit” was comparable to other schools with higher revenue potential.  That is, rather than asking whether the school had cash lying around to give to athletes, the decision process looks, from the outside, to be a much more clear-eyed question of whether the payment made sense economically — would the school’s net position be better by making the investment and reaping the resulting benefit of access to higher quality talent at a higher price, or would the payoff not be worth the price.  How the cost was funded varied widely — some schools solicited new donations or relied on the increased revenue compared to the year before, some cut spending elsewhere, some simply requested and received more funding from the central fund. But, as the model revealed, generally speaking the Athletic Directors who chose COA did so because talent seems to drive revenue, and those that did not mostly chose the lower cost route because their likely payoffs were lower than the new costs.  That is, they acted like any business would, weighing costs and benefits and adjusting budgets accordingly, not treating budgets as hard-and-fast barriers to making wise investments.


[1] In Re: National Collegiate Athletic Association Athletic Grant-In-Aid Cap Antitrust Litigation

[2] Those schools not in a conference are independent (which is rare), but are also given a minimum of a 1 for this variable.