Skip to main content

You’re Fired!: The Impact of Race on the Firing of Black Head Coaches in Major College Football


It is well known that the hiring rate of black head coaches in major college football is not representative of the number of student-athletes that are black. However, less obvious is the fact that black head coaches may be treated unfairly when decision-makers decide whether to retain or fire their institution’s current head coach. In this paper, I use a rich dataset of National Collegiate Athletic Associate football coaches from 1990 to 2012, containing measures of coaching performance and school expectations, as well as information on each coach’s race and whether he was fired or retained in each season. Using this data, I estimate a discrete-time hazard model of the probability that a head coach is fired, allowing the hazard rate of black head coaches and white head coaches to differ, and find that black head coaches are 5.28 percentage points more likely to be fired than their white counterparts. Additionally, I find that black head coaches are more likely to be fired in the initial 3 years of tenure, and again in their seventh and eighth years, but that the difference in the probability of release between black and white head coaches in the fourth through sixth years of tenure is small and statistically insignificant.

This is a preview of subscription content, access via your institution.

Fig. 1


  1. 1.

    There have also been three Latino and two Asian/Pacific Islander head coaches hired in FBS history (Hightower 2012).

  2. 2.

    A 2006 study by Cunningham, Bruening, and Straub found that there was no significant difference between black and white assistant football coaches in their aspirations to become head coaches or in their intentions to apply for head coaching positions in the future. Thus, I have subsequently ruled out that a lack of interest among blacks for becoming head coaches is an explanation for their underrepresentation in head coaching.

  3. 3.

    This excludes those Offensive or Defensive coordinators that also serve as the head coach.

  4. 4.

    Scully (1973) makes the point that this positional segregation is largely the result of blacks being excluded from positions involving a large amount of leadership and decision-making and mentions that blacks may be forced into positions requiring less coaching if discriminatory white coaches refuse to work closely with blacks. This process may take hold on black players early on in their youth.

  5. 5.

    In the study by Sagas and Cunningham (2005), human and social capital are measured by age, education, college coaching experience, and professional playing experience.

  6. 6.

    In the time period studied, 12.7 % of all white coaches that are involuntarily released are rehired compared to only 4.3 % of black head coaches involuntarily released.

  7. 7.

    Five white coaches with exceptionally long tenures were involuntarily released since the end of the 2009 season at least in part due to severe NCAA infractions. If these coaches are included in the total, the 56 nonblack coaches released since the end of the 2009 season had an average tenure with their teams of over 7 seasons.

  8. 8.

    I define a coach to be fired if they are either fired outright or forced to resign involuntarily. This is discussed further in “Data” section.

  9. 9.

    These are both endogenous determinants of whether a coach is fired. Athletic directors may be fired when the football team has exceptionally poor performance, perhaps because they are blamed for bringing in the wrong head coach. Controlling for a new athletic director, who is more likely to bring in his own head coach, would ignore the fact that the hiring of the new athletic director was the direct result of poor performance by the head coach. To be fair, Holmes (2011) notes that one should be cautious when observing specifications that control for a new athletic director for this very reason. Likewise, while it is one of the head coach’s primary responsibilities to recruit good players and prepare them for the NFL, and the outcome of this is on-the-field results, Holmes (2011) surmises that coaches with better players should be held to a higher performance standard. However, this argument discredits the head coach for the part that he played in the recruitment and development of these players.

  10. 10.

    The inclusion of a large number of school fixed effects may bias the results due to the nature of maximum likelihood estimation using a binary logistic regression function (Green 2004). Computationally, including school fixed effects forces me to disregard observations from ten different schools because these institutions did not fire a coach in the time period of the data. Nonetheless, I undertake this estimation as an additional robustness check; the results are shown in Table 5, column (6).

  11. 11.

    There are three Latino and two Asian/Pacific Islander head coaches in the data. I include these coaches with white coaches since their characteristics and outcomes more closely resemble that of white coaches than that of black coaches. Including them in separate race categories has no substantive impact on the coefficient of interest, and the coefficient on the indicator variable for Asian/Pacific Island or Hispanic is very close to zero.

  12. 12.

    I define a team’s inverse AP ranking as “the inverse of a team’s associated press ranking for ranked teams and zero otherwise.”

  13. 13.

    For schools that played at a level below Division I FBS in the previous ten seasons (but not in the past three seasons) prior to hiring their coaches, only their record in FBS games was included. If a school played at a level below Division I FBS in the previous three seasons prior to hiring its coach, the institution’s overall record for all three previous seasons is included.

  14. 14.

    In order to accurately measure the quality of outside coaching candidates it might make sense to use year to year variation in the firing rate by including year fixed effects. However, the number of firings is roughly increasing over time, so using year fixed effects instead of a linear function in year has no substantive impact on the estimate of interest.

  15. 15.

    Interaction terms between Previously Not FBS and a quadratic function in tenure are not statistically significant, and their inclusion has no substantive impact on the estimate of interest.

  16. 16.

    A few coaches of retirement age immediately took jobs as assistant coaches following their release. In these cases, I consider the release to be involuntary as it is difficult to argue that the coach left in order to retire.

  17. 17.

    Interestingly, college coaching experience has no impact on whether college coaches are fired after controlling for the set of tenure indicator variables.

  18. 18.

    In addition to the robustness checks shown, I estimate an additional robustness check that controls for alumni giving per student. Inclusion of this variable has no impact on the estimate of the black-white firing differential.

  19. 19.

    Testing the joint significance of Black, Black*Tenure, and Black*Tenure2 using a chi-squared test with three degrees of freedom gives a chi-squared statistic of 4.02.


  1. Bozeman B, Fay D. Minority football coaches’ diminished careers: Why is the “pipeline” clogged? Soc Sci Q. 2013;1:29–58.

    Article  Google Scholar 

  2. Cunningham GB, Bruening JE, Straub T. The underrepresentation of African Americans in NCAA Division I-A head coaching positions. J Sport Manag. 2006;20:387–413.

    Google Scholar 

  3. Dillard, Z.: Y.O.C.O. (You Only Coach Once): college football’s black coach problem. The Shadow League. 2013. Retrieved from

  4. Fredrickson JW, Hambrick DC, Baumrin S. A model of CEO dismissal. Acad Manage Rev. 1988;13(2):255–270.

  5. Green W. The behaviour of the maximum likelihood estimator of limited dependent variable models in the presence of fixed effects. Econ J. 2004;7(1):98–119.

    Article  Google Scholar 

  6. Henderson, J.: Jon Embree fired as Colorado Buffaloes Head Coach, says he needed more time. The Denver Post. 2012. Retrieved from

  7. Hightower, K.: Study: FBS leadership remains mostly white, male. USA Today. 2012. Retrieved from

  8. Holmes P. Win or go home: why college football coaches get fired. J Sports Econ. 2011;12(2):157–78.

    Article  Google Scholar 

  9. Kahn L. Race, performance, pay, and retention among National Basketball Association head coaches. J Sports Econ. 2006;7:119–49.

    Article  Google Scholar 

  10. Lapchick R, Anjorin R, Nickerson B.: Striving for sustained positive change: the black coaches and administrators hiring report card for NCAA FBS and FCS football head coaching positions 2011–12. The Institute for Diversity and Ethics in Sport. 2012. Retrieved from

  11. Lapchick R, Agusta R, Kinkopf N, McPhee F.: The 2012 racial and gender report card: college sport. The Institute for Diversity and Ethics in Sport. 2013. Retrieved from

  12. Madden JF. Difference in the success of NFL coaches by race, 1990–2002: evidence of last hire, first fire. J Sports Econ. 2004;5:6–19.

    Article  Google Scholar 

  13. Madden J, Ruther M. Has the NFL’s Rooney rule efforts “leveled the field” for African American head coach candidates? J Sports Econ. 2011;12:127–42.

    Article  Google Scholar 

  14. McDonald S, Lin N, Ao D. Networks of opportunity: gender, race, and job leads. Soc Probl. 2009;56(3):385–402.

    Article  Google Scholar 

  15. Miller, T.: Coaches see progress in hiring. 2012. Retrieved from

  16. Mixon FG, Trevinõ LJ. How race affects dismissals of college football coaches. J Lab Res. 2004;25:645–56.

    Article  Google Scholar 

  17. Sagas M, Cunningham GB. Racial differences in the career success of assistant football coaches: the role of discrimination, human capital, and social capital. J Appl Sport Psychol. 2005;35:773–97.

    Google Scholar 

  18. Scully GW. Economic discrimination in professional sports. Law Contemp Probl. 1973;38(1):67–84.

    Google Scholar 

  19. Turner, R.: College football: minorities trying to break into top jobs. The Salt Lake Tribune. 2009. Retrieved from

Download references

Author information



Corresponding author

Correspondence to Nolan A. Kopkin.

Appendix A. Misspecifications in models from past studies

Appendix A. Misspecifications in models from past studies

The results presented in this paper differ from those found in Mixon and Treviño (2004) and Holmes (2011). Misspecifications in their models led each to come to an incorrect conclusion: that black coaches face a lower dismissal probability than white coaches in the case of Mixon and Treviño (2004), and that the dismissal probability for black coaches is only higher than that of white coaches in years seven and eight in the case of Holmes (2011). Mixon and Treviño (2004) use an overly simplistic model that includes no indication of the tenure of each head coach other than an indicator for whether the coach is in his first season at the school; their model only captures performance through each coach’s cumulative winning percentage, the change in winning percentage from year t-1 to year t, and the number of games played in each year. In column (1) of Table 6, I estimate this model using data from 1990 to 2000, the same years that Mixon and Treviño (2004) use. Estimation of this model leads to the conclusion that blacks are less likely to be fired, but that this result is statistically insignificant; this is the conclusion presented in Mixon and Treviño (2004). In column (2) of Table 6, I estimate the same model using my full dataset. The coefficient on black is now positive although still statistically insignificant.

Holmes (2011) comes closer to the true result, noting that black head coaches are more likely to be fired in years seven and eight of their tenure than white coaches, and are no more likely to be fired in years four through six. However, he mistakenly concludes that black coaches in years one, two, and three of their tenure are no more likely to be fired than white head coaches. He comes to this incorrect conclusion because he chooses a specific functional form of tenure-dependence for the hazard rate, a quartic polynomial, instead of allowing the hazard rate to vary freely with tenure, as I do in this paper. Furthermore, he only allows tenure-dependence of the hazard rate for black coaches to differ from that of white coaches through a quadratic polynomial in tenure, another misspecification of the true model. I further explore the impact of the tenure-dependence restrictions made in Holmes (2011) in columns (3)–(6) of Table 6. In column (3), I estimate my full model from Table 2, column (4), making two adjustments present in Holmes (2011): I restrict tenure-dependence of the hazard rate to take on a quartic polynomial in tenure instead of allowing the hazard rate to vary freely with tenure, and I only allow the hazard rate for blacks to differ from that of whites through a quadratic function in tenure. The coefficients on Black, Black*Tenure, and Black*Tenure2 are all statistically insignificant in the model, and are also jointly insignificant.Footnote 19 If I simply check whether the hazard rate for black coaches varies from that of white coaches at a constant rate independent of tenure but continue to restrict tenure-dependence to take on a quartic polynomial in tenure, as I do in column (4), I find that black head coaches are 4.68 percentage points more likely to be fired than white head coaches; this estimate is statistically significant at the 5 % level. Next, as shown in column (5), I test how the hazard rate for blacks varies with tenure using the same tenure spline for blacks that I use in Table 3: years one–three, years four–six, and years seven and eight. Here, I find the familiar result from Holmes (2011) —that the hazard rate for blacks is larger in years seven and eight, statistically significant at the 10 % level, and that the difference is statistically insignificant in earlier years. However, this conclusion is the product of restricting tenure-dependence of the hazard rate to take the form of a quartic polynomial. Relaxing this assumption so that the hazard rate varies freely with tenure gives the result found in column (4) of Table 3.

Table 6 Misspecifications of the Hazard Rate for Black Coaches

About this article

Verify currency and authenticity via CrossMark

Cite this article

Kopkin, N.A. You’re Fired!: The Impact of Race on the Firing of Black Head Coaches in Major College Football. Rev Black Polit Econ 41, 373–392 (2014).

Download citation


  • Discrimination
  • Unfair firing practices
  • Black head coaches
  • College football
  • NCAA
  • Racial differences