We’ve all heard about the gender gap in higher education: Nationally, women enroll in college and complete degrees at higher rates than men do. But new research reveals that for decades women have been underrepresented at the nation’s most-selective institutions. And the apparent culprits are standardized tests.

Admissions offices’ reliance on SAT scores has created “de facto institutional preferences for men” at the nation’s most-competitive colleges, according to the results of a longitudinal study to be published in a forthcoming issue of Research in Higher Education. In 2004, for instance, women made up more than half of undergraduates attending all types of four-year colleges except for the most-selective ones, as categorized by Barron’s Profiles of American Colleges. Within that tier of 65 institutions, which accepted no more than a third of applicants, women accounted for 47 percent of students that year.

“It’s perplexing,” says Michael Bastedo, an associate professor of education at the University of Michigan at Ann Arbor and a co-author of a report on the study. “You would think that women’s advantages nationally, with their higher high-school grades, would translate into larger advantages at elite colleges.”

Mr. Bastedo and his fellow researchers sought to explain why that hadn’t happened. Using nationally representative student data from the high-school senior classes of 1972, 1982, 1992, and 2004, they tested three hypotheses. One was that women had applied to the most-selective colleges at lower rates than men did. Another was that colleges had built preferences for men into their admissions processes.

But the researchers found evidence for another explanation: gender disparities in standardized measures of academic ability that colleges value highly. On average, men outperform women on the SAT. That, combined with the importance attributed to test scores in admissions offices, they wrote, creates preferences for men “that drive women’s underenrollment in these institutions.”

Mr. Bastedo, who is also director of the Center for the Study of Higher and Postsecondary Education at Michigan’s School of Education, suspects that, to some extent, such preferences may be unconscious. “Part of the problem of doing admissions at elite institutions is finding distinctions that make a difference, and standardized tests are one of the few places where you can see differences,” he says. “As much as people might say standardized tests don’t matter that much, when they’re trying to make distinctions, they actually end up mattering a lot.”

No measure of student achievement is perfect. Some admissions offices put great stock in test scores because many top applicants have superb—and often identical—grade-point-averages. Moreover, grade inflation in secondary schools is a common concern for colleges.

Still, Mr. Bastedo and his co-authors suggest that an overreliance on test scores is problematic. Previous research on the “predictive validity” of the big test has found that SAT scores underpredict women’s first-year grades in college. Given that women tend to outperform men after matriculating, the researchers write, “disproportionately excluding women from institutions on the basis of a criterion that underpredicts their performance is especially difficult to justify.”

Read the report here.