Why Colleges Shouldn't Dump the SAT
By Robert J. Barro
President Richard Atkinson of the University of California recently caused a stir by proposing to eliminate the Scholastic Aptitude Test (SAT) as a required part of the college application. Opponents of the SATs have argued that the test is ineffective for evaluating applicants, but such claims have been largely anecdotal. We now know, for example, that Senator Paul D. Wellstone (D-Minn.) had a stellar college career despite having an SAT score near the 50th percentile. But given the wealth of available data, we do not have to rely on these stories. They show that SAT scores have a lot of predictive power for students' college grades.
Every three years, the Education Dept. conducts the National Postsecondary Student Aid Study (NPSAS). It provides information for a nationally representative sample of colleges on students' grade-point averages (GPAs), admissions-test scores (including the SATS), and other family and school variables. Unfortunately, it doesn't contain high-school grades. I looked at the NPSAS studies for 1990, 1993, and 1996, which provide 33,000 observations, for my own analysis.
In this sample, admissions-test scores strongly predict college grades, though much of the individual variation in grades remains unexplained. Taking into account many other factors (including college attended, race and gender variables, and parental income and education), the t-statistic--a measure of how closely two variables move together--for the admissions test is 60. In comparison, researchers customarily regard a result as significant if this statistic exceeds 2. Therefore, admissions tests have strong predictive power for college grades. They are as good for senior as for freshman grades.
BIASED AGAINST WOMEN? For a given test score and other variables, females have significantly higher GPAs than males of the same ethnic group (by 0.25). Blacks and Hispanics do significantly worse than whites of the same sex (by 0.24 and 0.08, respectively), whereas Native Americans and Asians do not differ much from whites of the same sex. Another result is that the mathematics part of the admissions test is nearly twice as good as the verbal part as a predictor of college grades.
Looking separately by gender, the predictive power of admissions-test scores for college grades is as good for women as it is for men. However, results on admissions tests were, on average, 4 percentiles lower for females than for males of the same ethnic group, even though women had better college grades. Therefore, women may have a legitimate complaint that admissions policies that rely on the SAT have a sex bias.
The same case cannot be made for ethnic-minority groups. After taking into account parental income and education, admissions-test scores for blacks were, on average, 17 percentiles lower than those for whites of the same sex. The comparable gap for Hispanics was 4 percentiles. However, college grades for blacks and Hispanics were lower, on average, than those for whites, even comparing students who had the same scores on the admissions tests.
WAGE FORECAST. The ethnic patterns in admissions-test results have sometimes been attributed to cultural bias. One problem with this argument is that test scores have similar predictive power for college grades within an ethnic group, such as blacks or Hispanics, as they do in the overall population. Moreover, other data indicate that test scores and college grade-point averages help to predict wages earned after college. Therefore, if admissions tests reflect a cultural bias, it must be the same bias that exists in other indicators of success that people care about.
Much of the opposition to the SATs seems to derive from the poor average performance of ethnic minorities, especially blacks. Clearly, a reliance on the tests conflicts with affirmative-action objectives, especially at the University of California, where explicit racial preferences have been ruled out. The results from the NPSAS data do not support preferences for racial minorities as a way to maximize the academic performance of the student body. But maybe the top schools should not be trying to admit applicants who will be the best scholars.
An implicit assumption in academia is that the best faculty--indeed, the best resources overall--should be matched with the most academically talented students. Some evidence suggests that the synergies from this matching are important, and, being a Harvard University professor, I naturally find these arguments persuasive. I am surprised, however, that the notion of partnering the best with the best has generated long-term support from taxpayers in funding such public institutions as the University of California as a meritocracy. After all, most taxpayers and their children are not themselves the best.
Perhaps strong voter support for elite public universities is ending, and the movement away from the SATs is a way of signaling the demise of the University of California's aspirations to meritocracy. But somehow, I do not believe that President Atkinson had this goal in mind.
Robert J. Barro is a professor of economics at Harvard University and a senior fellow of the Hoover Institution (email@example.com).