SAT Scores Hurt Berkeley Women

University Testing

A recent study at the University of California (UC) at Berkeley concluded that use of the SAT in the college's undergraduate admissions formula biases the process against females and reduces the number of women entering each class by 5%, or about 200-300 students.


In a paper delivered at the 1995 American Educational Research Association conference in San Francisco, authors David K. Leonard and Jiming Jiang report, "not only are the tests flawed in their predictions but . . . this problem has been known to insiders for over half a century." Leonard is a past Chairman of Berkeley's Academic Senate Committee on Admissions and Enrollment.


For admissions purposes, the UC system converts a student's SAT-I and SAT-II scores together with high school grade point average (GPA) into a numerical Academic Index Score (AIS), designed to predict college GPA. At Berkeley, the first 50% of admission decisions have been based solely on AIS -- those with high enough scores are admitted automatically. Students with AISs below that first round cut-off, but still within a set score range, were put into the "Special Promise Read Pool," where their essays and personal achievements are considered in the admission decision. A third pool of candidates were admitted based on other factors including racial diversity, unusual talents and family associations with the school.


Leonard and Jiang compared 10,000 Berkeley students' AISs to their actual college GPAs after graduation (or other termination from the school). Their findings include the following:

  • Many women who missed the automatic-admission cut-off nevertheless obtained the GPAs the cut-off is meant to predict;
  • Women with lower AISs than men obtained equal GPAs in all majors except Chemistry;
  • Women with identical AISs to men obtained higher GPAs (by a tenth of a point) across majors.

SAT proponents often assert that the test underpredicts female GPAs because females select less difficult academic majors and courses. The Berkeley authors addressed this claim by comparing admissions data with academic performance within particular programs of study. Results showed that course difficulty differences account for, at most, half of the GPA underprediction. The rest of the underprediction is attributable to the SAT's relatively low validity for females and its role in the AIS.


Women admitted to Berkeley's Engineering and Chemistry colleges in spite of low-end AISs, based on Special Read Pool qualitative criteria such as essays and recommendations, earned higher grades than male peers from the same AIS group. This means non-test criteria are better predictors for women than modest test scores in these academically rigorous (and male-dominated) fields.


Leonard and Jing conclude that women should have about 140 points added to their AIS to compensate for the SAT's underprediction at Berkeley. They also suggest several other ways to level the playing field for women: Qualitative indicators should figure more prominently in women's admissions; the school could alter the relative weight of AIS factors, making high school GPA far more important; it could vary the admissions formula by gender to compensate for the bias, as the Massachusetts Institute of Technology (MIT) does; or it could drop the SAT requirement altogether.


The report concludes, "Given the history of the College Board's obfuscation and avoidance of this issue for the last 50 years, users would be unwise to assume that a good faith effort will be made to correct the problem without external prodding."


. Leonard, David K. & Jiang, Jiming. Gender Bias in the College Predictions of the SAT. Paper presented to the annual conference of the American Educational Research Association, San Francisco, CA, April 21, 1995. (David Leonard, University of California, 210 Barrows, Berkeley, CA 94720).