Analysis of Diffrential Prediction of Law School Performance by Gender Subgroups Based on 2002-2004 Entering Law School Classes (TR-07-01)
by Deborah A. Suto, Lynne L. Norton and Lynda M. Reese
In the law school admission process, it is essential that the criteria used for admission be fair to all subgroups in the applicant population. One method used to evaluate the fairness of the admission process is to compare the predicted and actual first-year grade-point averages (FYAs) at each law school for various subgroups of the applicant population. The purpose of this study is to address questions of differential prediction between male and female first-year law school students based on data from the 2002, 2003, and 2004 entering classes of 183 law schools.
Statistical analyses were used to predict FYAs using three traditional predictors: undergraduate grade-point average (UGPA), Law School Admission Test (LSAT) score, and the best predictive combination of UGPA and LSAT score. A separate analysis was conducted for each law school included in the study (i.e., three prediction equations were derived for each individual law school included in this study).
None of the three prediction equations evaluated were deemed to be problematic. More specifically, none of the prediction equations systematically and significantly overpredicted or underpredicted FYAs for male or female students across the schools studied. The magnitude of overprediction or underprediction across members of each gender group was found to be less than one-tenth of a standard deviation, on average, for each prediction equation. The degree of differential prediction was greatest for the model using UGPA alone, but these differences were still too small to be of practical significance. To describe these results in practical terms, consider a law school that has a 0 to 4.33 grading scale with an observed mean of 3 and standard deviation of .44. Then, for example, an underprediction of one-tenth of a standard deviation for female students at such a law school would mean that their FYAs were underpredicted on average by a factor of .044. Overall, the results of this study do not support the concern that the use of LSAT scores or the traditional combination of LSAT scores and UGPAs results in unfair admission decisions with regard to gender.
While considering the results of this study, the reader should keep in mind that the results refer only to subgroup behavior and not to individuals. That is, individuals within a subgroup that is overpredicted on average may still be themselves underpredicted in terms of their individual law school performance, and vice versa.
Finally, it is worth repeating that the average amount of overprediction or underprediction of FYAs found was always very small, regardless of the prediction equation that was used. In other words, this study provided no evidence that LSAT scores, UGPAs, or combinations of those two measures unfairly predict future law school performance for either gender subgroup.