A reexamination of how athletic success impacts graduation rates: comparing student-athletes to all other undergraduates - The University

American Journal of Economics and Sociology, The, April, 2003 by Patrick James Rishe

I

Introduction

McCORMICK AND TINSLEY (1987) found that athletic success served as an important marketing tool for universities, in that it attracted students with higher incoming SAT scores. Shughart, Tollison, and Goff (1986) found that athletic success hurt the academic productivity of economics professors in terms of fewer journal pages per faculty member. Tucker (1992) found a similar result, in that undergraduate graduation rates were inversely correlated to the success of football, though Tucker and Amato (1993) found that changes in incoming freshman SAT scores over time was positively correlated to football success.

This paper also focuses on how athletic success impacts college graduation rates. The departure from the papers mentioned above is that this paper separates the student-athlete graduation rate from the graduation rate of all other undergraduates, and it looks at a more complete sample of Division I schools. Separating the graduation rate of student-athletes from all other undergraduates allows us to address three questions: Is the academic success of the consumers of college athletics (undergraduates) impacted by the success of their school's athletic program? Do pressures to produce athletically impact the academic success of the producers of college athletics (student-athletes)? Is the relative academic success of student-athletes compared to all other undergraduates sensitive to the success of the school's athletic program and/or the school's perceived status as a major Division I athletic power?

II

Literature Review

McCORMICK AND TINSLEY (1987) measured athletic success with a dummy variable distinguishing the 63 universities in 1984 that belonged to either major athletic conferences or were a major independent (e.g., Notre Dame) in Division I. In their dynamic model of explaining changes in SAT scores over time, they reduced their sample to just the 63 "big-time" universities and measured athletic success as a 14-year trend in football winning percentages.

Tucker (1992) and Tucker and Amato (1993) introduced a new method of measuring athletic success that involved aggregate football and basketball rankings by the Associated Press. They were uncomfortable with classifying athletic success with a dummy variable because the disparity in athletic success of teams from the same conference could not be distinguished (such as the quality of Florida's athletic program versus Vanderbilt's program). The "rankings" measurement of athletic success allowed a more accurate, quantitative measure of athletic success than did past studies.

But Tucker and Amato only considered the 63 "big-time" schools examined in earlier papers, thereby not allowing for a more comprehensive cross-sectional comparison across Division I schools. Perhaps more important is that Tucker (1992) did not distinguish the athlete graduation rate from the graduation rate of all other students. If the student-athlete graduation rate differs significantly from that of all other students, one obtains a muddied picture concerning whose academic performance is impacted by athletic success when the graduation rate variable is defined to encompass all students.

The graduation rate for student-athletes for the entire sample is 58.15 percent, whereas the graduation rate for all other undergraduates is 54.62 percent. A paired t-test confirms that this difference is statistically significant. This graduation gap in favor of athletes would be greater if not for the increasing phenomenon of college athletes in football and basketball leaving school early to play professionally. Hence, it seems important to separate the two graduation rates from each other in order to obtain a clearer picture of which students are influenced the most by athletic success.

III

Empirical Analysis

THE GENERAL MODEL SPECIFICATION is:

Y = [beta]X [epsilon],

where X is a vector of non-academic and academic variables, [beta] is a vector of coefficients on these variables, and [epsilon] is a normally distributed error term with zero mean and constant variance. Three dependent variables are used to address the three separate questions posed in the introduction: the graduation rate of all non-athlete undergraduates, the graduation rate of student-athletes only, and the difference between the graduation rate of student-athletes and the graduation rate of all other undergraduates.

The graduation rate used is the average graduation rate of the four freshman cohorts from 1988 to 1991, and means that a student earned his or her diploma within six years of his or her freshman year. Using a four-year average graduation rate makes this variable more resistant to an unusual cohort's graduation rates. All transfer students are omitted from consideration. The data is from the 1998 NCAA Division I Graduation-Rates Report.

Several variables are used to proxy athletic success. ALLSPT measures the number of Sears' Director's Cup points a school accumulated between 1993 and 1997, the years in which the freshman cohorts in question were competing in NCAA competition. The top 64 teams in a given sport are awarded points. The top team receives 64 points, the runner-up receives 63 points, and so on through the top 16 teams. Teams 17 through 25 receive 40 points, teams 26 through 44 receive 20 points, and teams 45 through 64 receive 10 points. All sports are weighted equally so that, for example, a national champion in men's basketball is awarded the same number of points as the women's field hockey champion.