SAT-Math Gender Bias: Causes & Consequences

University Testing

The recent settlement of FairTest's complaint charging the Educational Testing Service (ETS) and the College Board with gender bias for developing and administering the National Merit Scholarship qualifying test, the Preliminary SAT (see Examiner, Fall, 1996), has reinvigorated interest in determining why tests like the SAT consistently underpredict the academic performance of females.


By agreeing to alter the content of the PSAT to give young women a fairer chance at more than $25 million a year in scholarships, the test-makers implicitly conceded that their exam was discriminatory. But ETS and the College Board have made no attempt to address gender bias in the SAT or in tests of a similar design such as the GRE.


Since nearly the entire test score gender gap comes from the math portion of the SAT (the PSAT contains previously administered SAT items), that is where most attention has focussed. Conservative commentators continue to maintain that a largely multiple-choice, "objective" exam cannot possibly be biased -- hence, the problem must be that high school and college math teachers consistently give female students top grades they do not deserve. But they have no facts to back up that theory.


Those who contend that the SAT-Math test is inherently flawed now have substantial new evidence in the form of two "meta-analysis" reports by Cathy Kessel and Marcia Linn of the University of California - Berkeley School of Education. In a paper delivered at an American Education Research Association (AERA) symposium and a companion journal article, Kessel and Linn document the impacts of the SAT's gender bias and analyze its most likely causes.


According to Kessel and Linn, there is no question that the test is at fault. Summarizing more than a dozen studies of large student groups and specific institutions such as MIT, Rutgers and Princeton, they conclude that young women typically earn the same or higher grades as their male counterparts in math and other college courses despite having SAT-Math scores 30-50 points lower.


Among the causes documented in their literature review:


- The SAT's emphasis on speed: "Since these tests require students to solve 25 to 35 problems in 30 minutes, there is a premium on efficiency. In contrast, reformulated mathematics courses that include an emphasis on solving complex, ill-posed, and personally-relevant problems place a premium on sustained reasoning."


- multiple-choice questions: "mathematics examinations in the Netherlands, England, Australia, and other countries that require solutions to several long problems seem unbiased with regard to gender."


- Simplistic content: "ability to refine complex problems and design appropriate solutions goes beyond the skills measured in college entrance examinations."


- Self-confidence: "Studies suggest that males, compared to females, expect to answer correctly, independent of actual performance. This perspective results in less answer checking and reflection, useful skills for speeded tests."


- Stereotype vulnerability: "there was a gap in average scores for males and females when the experimenters announced that they expected gender differences." Similar correlations between expectations and performance have been reported for African American students by researcher Claude Steele (see Examiner, Fall/Winter 1995-96)


As a result of these factors, young women receive lower SAT-Math scores and lose out on college admissions, scholarships, placement into advanced courses, prestige, and career opportunities. Yet, given a chance, they are likely to outperform their male counterparts. Thus, Kessel and Linn favorably cite FairTest's list of schools where test scores are optional in the admissions process as one way to address the problem. They also endorse new forms of assessment that make use of portfolios and projects. Of course, eliminating reliance on the SAT for any high-stakes decisions would benefit an even larger number of academically able students.


* "Grades or Scores: Predicting Future College Mathematics Performance," Educational Measurement: Issues and Practice. Vol. 15, No. 4, Winter 1996 (special issue on gender and mathematics performance).


* "Participation in Mathematics Courses and Careers: Climate, Grades, and Entrance Examination Scores," presented at the 1995 AERA Annual Meeting, Division D Symposium on gender and mathematics performance.