Large and growing gaps in SAT scores, by race and ethnicity, are nothing new. The College Board and educators alike have acknowledged these gaps and offered a variety of explanations, with a focus on the gaps in family income (on average) and the resources at high schools that many minority students attend. And indeed there is also a consistent pattern year after year on SAT scores in that the higher the family income, on average, the higher the scores.
But a new, long-term analysis of SAT scores has found that, among applicants to the University of California's campuses, race and ethnicity have become stronger predictors of SAT scores than family income and parental education levels.
Further, the study has found that all three factors -- race/ethnicity, family income and parental education levels -- now predict one-third of the variance in SAT scores among otherwise similar students, up from a quarter in 1994. In other words, a larger share of SAT variance today than in 1994 may be predicted based on where and to whom a child is born.
The research was done by Saul Geiser and was released by the research center where he works, the Center for Studies in Higher Education at the University of California at Berkeley.
Geiser is quick in the paper to acknowledge that his study is only of the applicant pool for the University of California and that he has not done research on the extent to which these trends play out nationally. However, his study was based on a very large pool: the more than 1.1 million California residents who applied to UC campuses from 1994 through 2011. And his study is based on the current and previous SAT, not the new one about to be unveiled.
But his findings suggest that those who hope for a closing of racial gaps on standardized tests used for college admissions may be in for disappointment.
Much of the study is based on regression analysis of different factors associated with SAT scores. By controlling for some factors, he can find which characteristics have the most influence.
And for those who wish race to play less of a role, Geiser notes that there were some years of hope. The share of score variance attributable to socioeconomic factors fell from 25 percent in 1994 to 21 percent in 1998. But in the years that followed it went back up to 35 percent.
In contrast, socioeconomic factors could not be linked in a major way to variance in high school grade point averages. Socioeconomic factors, including race and ethnicity, accounted for 7 percent of the variance in GPAs in 1994 and 8 percent in 2011.
Geiser considers several possible explanations for the increasing impact of race and other socioeconomic factors in predicting SAT scores. One that he takes seriously is the possibility that links growing rates of "intense segregation" in high schools, with more minority students attending high schools that are overwhelmingly minority and poorly resourced. For instance, the percentage of what researchers call "apartheid" schools -- those where 99-100 percent of students are nonwhite -- has doubled in the last two decades, and now represents one in 14 high schools. So the impact of race and class are, in many cases, combined for the minority students attending those schools.
Unlike some other critics of the SAT, Geiser doesn't push for its elimination as an admissions criteria. He notes that other measures don't necessarily help minority applicants.
However, Geiser does write that the SAT appears to be a poor predictor (especially for black and Latino students) of whether they will graduate from UC. A key caveat here is that the College Board has always stressed that the SAT is a tool for predicting first-year performance, not graduation. Still, Geiser writes that the consideration of the SAT depresses the chances of minority students getting in, while doing little to help admissions officers predict applicant success.
The solution, for Geiser, is to go back to what the University of California did when it adopted the SAT, but which the state's voters have barred it from doing today: considering race in admissions. He writes that if public universities are going to consider SAT scores in a serious way, they should also consider race and ethnicity.
"The continuing dominance of standardized admissions tests in American higher education is one of the most powerful arguments for affirmative action. Much of the original impetus for race-conscious policies grew out of recognition of the severe adverse impact of SAT scores on admission of students of color. Since then, that impact has not only continued but worsened, if the California data are any indication," writes Geiser.
"These findings underscore the continuing relevance of the original, remedial rationale for affirmative action. Rather than a remedy for historical discrimination, however, they show that race-conscious policies are essential to remedy unwarranted disparities in the present day. The adverse racial impact of SAT scores is far out of proportion with their limited capacity to predict how applicants will perform in college," he concludes.
Geiser makes his argument as the U.S. Supreme Court is once again considering the constitutionality of considering race in admissions.
Asked if he thought his findings could influence the justices or colleges, he said via email: "I have no idea whether my results might influence the court, but if other states were to observe the same trends, I think it might make a difference. I’m hoping that other institutional researchers will pick this up. As for the message to colleges, I think the important point is the linkage of affirmative action with standardized testing, not the emphasis on one or the other."
Roger Clegg, president and general counsel of the Center for Equal Opportunity, which opposes affirmative action, had a different take on the study.
"If a test is unreliable for certain races -- and this has long been alleged and long been refuted for the SAT, by the way -- then a school is perfectly justified in not using it, but it should try to find other measures that are reliable," Clegg said via email. "What it should not do is admit students who are less well qualified under any measure in order to reach a particular racial result."