trevino by ajizai



     Racial Discrimination and Employment in
                  College Football

                        LEN J. TREVIÑO AND TAISA C. MINTO
                        The University of Southern Mississippi–U.S.A.

    The work of Blinder [J. of Human Res., 1973] and Flanagan [Rev. Econ. & Stat., 1973]
builds a solid foundation of labor economics studies measuring racial discrimination. To date,
there has been little examination, outside of the popular media, of racial discrimination in
the hiring and Þring process in intercollegiate sports (for example, college football coaching).
The disproportionate percentage of black head football coaches (4 percent) has prompted the
NCAA Board of Directors to establish a Football Oversight Committee to examine diversity
    The authors examine the decision by universities to terminate football coaches using a
pooled, annual time series of institution-level cross-sections for 1990-2000. The data cover
81 institutions, totaling 759 observations. Following Caudill et. al. [Appl. Econ., 1995]
and Mixon [Appl. Econ., 2001], it is assumed that the institutional tendency to terminate its
football coach at the end of year t is given by an unobservable variable, Y ∗ . What is observed
is the outcome (at the end of year t) of the employment decision, Y . If Y ∗ ≥ 0, the coach
is terminated and Y = 1. If Y ∗ < 0, the coach is retained (or resigns), and Y = 0. Assume
that a reduced-form model of the institutional tendency toward termination can be written:
Y ∗ = Xβ +ε. Here, X is a vector of exogenous variables affecting this tendency (at the end of
year t), including the coach’s cumulative winning percentage at school i (CU W IN), the year
of each observation (Y R), number of games coached (in year t) minus 11 (GAM ES), coach’s
winning percentage in year t minus his winning percentage in year t−1, when both years were
coached at school i (W IN LAG), and a dummy variable equal to 1 for black coaches, and 0
otherwise (RACE). The employment decision is analyzed as a discrete-time hazard model,
with a likelihood function composed of probabilities of two types [Caudill et. al., 1995]. A
linear probability model (LPM) is employed to estimate the regression equation.
    As expected, the coach’s cumulative winning percentage (CU W IN) had a negative and
signiÞcant (0.01) effect on the probability of termination. Since Þnancial gains from intercol-
legiate athletics have become increasingly important in the modern era, the authors expected
and found a positive (though insigniÞcant) relationship between Y R and Y . Given an ex-
pectation of rewards for year-to-year improvement, they expected and found a negative and
signiÞcant (0.05) relationship between W IN LAG and Y . GAM ES is a likely indicator of
post-season play. The authors expected and found a negative and signiÞcant (0.01) rela-
tionship between GAM ES and Y . The signiÞcance of the RACE parameter empirically
addresses the question of racial discrimination. There was a positive, although insigniÞcant,
relationship between RACE and Y . The sign of this coefficient indicates that the likelihood
of a black coach being Þred, ceteris paribus, is greater than that for a non-black coach. This
jointly signiÞcant regression produced an R of 0.0586. For completeness, future work should
concentrate on other estimation methods and statistical decomposition techniques [Jackson
and Lindley, Appl. Econ., 1989]. (JEL J7); Atlantic Econ. J., 31(1): p. 118, Mar. 03. ° All
Rights Reserved.


To top