Study: S.A.T. Questions Are Biased Against Girls
Washington--Questions on the Scholastic Aptitude Test are more likely to favor males than females, a study released here last week asserts.
The report by Phyllis Rosser of the Center for Women Policy Studies marked the latest volley of a long-running battle over whether standardized college-entrance tests are biased against women.
In an analysis of two versions of the sat, Ms. Rosser found many mathematical and verbal questions that statistically favored males--but only a handful of verbal questions on which females had the advantage.
"I'm advocating either changing the test to make it more fair, or coming up with something else," Ms. Rosser said at a press conference.
The sat underpredicts the academic performance of women in college, the researcher argued, noting that studies by the College Board have shown that women's average first-year grades are as good as or better than men's. Yet, she said, females' average sat scores reported in 1988 were 56 points lower, out of a total of 1,600, than the scores of males.
The test is designed to predict first-year college performance.
Hurting the Best Most
In comparing high-school grades and sat scores, the study found that women with the best high-school records were penalized most by the test.
In the College Board's 1988 report on scores, Ms. Rosser noted, the difference between all men and women who took the test was 13 points on the verbal portion and 43 points on the mathematical. But women with A-plus averages scored much lower than men with the same grades--a gap of 23 points on the verbal test and 60 points on the mathematical.
"This may exclude them from the most prestigious colleges that accept their male peers," Ms. Rosser's report maintains, "and may also prevent them from qualifying for merit scholarships and other scholarships that are based on sat scores rather than high-school performance."
Ms. Rosser did an analysis of 1,000 students who took the June 1986 sat and 100,000 students who took the November 1987 test. She found 17 questions on the 1986 test and 23 questions on the 1987 test in which there was a difference of at least 10 percent in the proportion of each sex that answered correctly.
On the 1986 math test, for example, 10 differences of more than 10 percent appeared--all favoring men. One was on a question concerning a basketball team's win/loss record, which was answered correctly by 27 percent more males.
But Cathy Wendler, assistant program director at the Educational Testing Service, which writes the sat, emphasized that the ets examines questions for potential bias.
"Just looking at the percent correct is not a good way to determine bias," she said last week.
The ets and the College Board, which sponsors the sat, contend that the test itself does not reflect or promote bias. Instead, they argue, the racial and gender differences in scores are due to socioeconomic factors and the educational backgrounds of those taking the test.
In another report released last week, the College Board said that men's and women's scores on a wide variety of tests have been converging over the past 20 years.
Based on a review of more than 200 studies, Gita Z. Wilder and Kristin Powell of the ets said there was evidence "to support the hypothesis that men are better at some quantitative tasks, but not the hypothesis that women enjoy a consistent advantage over men on verbal tasks."
Copies of "Sex Differences in Test Performance: A Survey of the Literature " are available for $6 each from College Board Publications, Box 886, New York, N.Y. 10101-0886.
Copies of Ms. Rosser's report, "The sat Gender Gap: Identifying the Causes," are available for $15 each from the Center for Women Policy Studies, 2000 P St., N.W., Washington, D.C. 20036.--mw