BY NORRIS M. HAYNES
The US Supreme Court has recently ruled that colleges and universities can no longer use race as a factor in college admissions. This ruling was the result of a specific challenge to admissions policies at Harvard University and the University of North Carolina. In their complaint, the plaintiffs, an organization called Students for Fair Admissions, claim that affirmative action amounted to discrimination against non-Black students, particularly white students and students of Asian backgrounds with higher standardized test scores and high school grade point averages (GPAs). The plaintiffs and their supporters maintain that using race as a factor essentially lowers admissions standards to colleges and universities to benefit Black students.
Affirmative action has come to connote, in the minds of many, unfair preferential treatment in favor of some minority groups, mainly Black students, and against others, particularly white and Asian students, in college admissions. This misconstrued view of affirmative action assumes that the best and ultimate measures of students’ academic competencies are scores on tests such as the SAT or the American College Test (ACT), which purportedly assess cognitive capacity and college readiness. Scores on these tests are numbers based on constructed measures that contain sources of human error and that cannot and do not reflect the human variations of ability and creative capacities among individuals. The main arguments among those who have opposed affirmative action have been focused on test score disparities and GPA differences, while neglecting the fact that these measures are flawed and imprecise and contain significant bias.
Studies of the predictive validity of the SAT and ACT indicate that they are not strong predictors of college performance, yet they are still very widely used. Analyzing data from over 23,000 students, Jeff Allen and Steven B. Robbins reported a positive and significant correlation of 0.53 between ACT composite scores and college GPA. This means that only 25 percent (.53 squared) of variance observed on GPA for the 23,000 students could be explained by ACT scores. In another study, Jennifer L. Korbin and colleagues analyzed data for 50,000 students and reported correlations of 0.41 for the SAT, suggesting that only 16 percent (.41 squared) of the variance observed on college GPA could be explained by SAT scores. The results for the ACT in this study were remarkably similar. Such results show that the eminence given to test scores in college admissions is over-amplified and unjustified.
The results on the predictive validity of the SAT and ACT for Black and Hispanic students show that these tests fare even worse in terms of their predictive strength. Jeff Allen and Steven B. Robbins reported a similar correlation between SAT scores and first-year college GPA for Black and Hispanic students (r = 0.34), suggesting that only about 12 percent (.34 squared) of the variance in college performance could be explained for these two groups of students. A report by Jonathan Beard and Jessica Marina indicated that the SAT is a stronger predictor for white students compared to Black and Hispanic students. California state universities and colleges no longer considers applicants’ standardized test scores. April Grommo, CSU assistant vice chancellor for enrollment management services, said, “The decision by the Board of Trustees aligns with the California State University’s mission of access and our efforts to provide high-quality college degrees for students of all backgrounds. . . . We are eliminating a high-stakes test that can cause great stress on students and their families and does not add any additional predictive value over high school GPA. The CSU being test-free will better meet the needs of our future students.”
Researchers have noted that high school GPA is more strongly correlated with college completion than standardized test scores. Wesnick and colleagues reported correlations as high as .58 between high school GPA and college performance, when corrected for sample size, explaining about 34 percent of the variance in observed college performance. A major issue with the use of high school GPA in college admissions is that many admissions officials attach value to GPAs based on high school characteristics such as perceived rigor, location, cost to attend, and private or public, all of which lead to reputational weighting of GPAs, taking us back to the problems of disadvantage for students who attend public schools in urban school districts.
Because test scores and high school GPA explain only approximately 50 percent of the observed variance in college performance, one must be concerned about the fact that these two metrics are so very heavily favored and used in college admissions.
Unequal access to test preparation opportunities and resources, high-quality educational experiences, and social and cultural capital—all of which influence test scores—render tests invalid. Consideration of GPAs, particularly from elite high schools, in admissions decisions is an additional source of significant bias. Studies have shown that differences in test performance between white and Asian students on one hand, and Black and Hispanic students on the other hand, reflect systemic inequities in educational opportunities and significant cultural bias in admission testing based on factors such as language and test format.
In summary, my argument is against using tests and GPAs as the main bases for admissions decisions. If used, they must be considered alongside other personal, demographic, and sociocultural variables that can account for the over 50 percent of college performance variation that remains unexplained. The Supreme Court decision did not take such facts, based in measurement science, into account. I believe that historical and sociocultural imperatives, bolstered by measurement science, make the court’s decision a wrong decision and reinforce my view, and the views of so many other experienced researchers and educators, that affirmative action matters.
Norris M. Haynes is professor of educational leadership and policy studies at Southern Connecticut State University.
Many thanks to Prof. Haynes for his summary of research on the predictive value — or lack thereof — of SAT/ACT scores and GPAs. I’m concerned, however, that he assumes facts not in evidence, especially in regard to the admissions practices in question. He argues that admissions decisions should take into account “other personal, demographic, and sociocultural variables that can account for the over 50 percent of college performance variation that remains unexplained. The Supreme Court decision did not take such facts, based in measurement science, into account.” The SCOTUS decision was relatively silent on all other variables that could be used in admissions decisions, and focused exclusively on the use of race, for the simple reason that there is no need for the Court to have taken all of those other variables into account, when they were not the focus of the suit. Indeed, both Harvard and UNC made it clear that their holistic admissions process does in fact take into account a wide range of “other personal, demographic, and sociocultural variables….” And the opinion of the majority also includes the (somewhat awkward) proposal that race can be used in admissions decisions, but only in an indirect way. It was not for the Court to determine what were the most powerful predictors of student success in college, or even if that is the primary goal of admissions: the issue for the Court was whether race, specifically, was used in a way that — in the view of the conservative justices — violated a constitutional prohibition. (Thus, any college is free to admit only bald people, or only tall people, without violating any protections, and we can easily imagine a Jonathan Swift college that did so.)
So let me thank Prof. Haynes again for his analysis, and remind him that many colleges are dropping the SAT/ACT admissions requirement for precisely the reason he proposes.
The argument that because test scores only explain a small part of variance in college performance they shouldn’t be used is extremely confused. I mean, by that logic we shouldn’t let your grade in any class affect college admissions, nor any essay, activity or anything else since they all predict little of the variance. Yes, maybe GPA taken as a whole is a better predictor but guess what lets you make even better predictions: the pair consisting of both your GPA and your test score.
I don’t understand how anyone with an understanding of statistics could think the right question is whether test scores are more predictive or the most predictive thing rather than whether *adding* that information allows for a more accurate prediction than without.
And regarding whether test scores mispredict outcomes for certain groups that’s again not the question. The question is whether the totality of the regression performed by the college mispredicts minority performance and if removing test scores improve it. I mean of course, the same test score means something different for someone who had every advantage and someone who had a poor school and struggled with poverity — which is why any school that’s not stupid includes those factors in the score they generate.
And even with the new SCOTUS ruling if you actually had data that the regression could be rendered more accurate by adjusting by race that almost certainly wouldn’t be barred (as long as no games were played to cherry pick the other factors). That’s not a preference, it’s merely adjusting the information to cancel out for a bias introduced by the test.
However, I’m very skeptical that there would be any systemic racial bias in a regression that included factors like wealth, school quality, parental education etc etc.. Almost certainly, black kids who face a given degree of disadvantage are going to do about the same as other kids with the same level of disadvantage (maybe they are affected more by discrimination but you’ll likely pick that up if you include variables like parental incarceration).
And if the argument is the stress and annoyance of having to take the test isn’t worth the predictive benefit then I expect to see college essays and participation in sports and every other factor put under the same microscope. Essays were incredibly stressful to write and I bet they predict even less than test scores. Some of us didn’t find tests very stressful but did find the expectation we have well-rounded extra-curriculors extremely burdensome and stressful (and that takes many hours a week not a few hours once).
“The way to stop racial discrimination is to stop discriminating by race.” There is no coherent argument for denying a Korean grocer’s child admission to a selective university as a make good for slavery and Jim Crow. The plain fact is that we need to rely on competitive testing in all sorts of field of endeavor. I don’t want brain surgeons, airline pilots, or any other situation where expertise and capability are matters of life and death to be filled through any process other than the most rigorous selection. Try becoming a journeyman electrician without passing the tests. All the attacks on the testing processes are political. You have a desired outcome, and you attack anything that stands in the way of that outcome.