[Freddie deBoer] Please, Think Critically About College Admissions

This paper aggregates three of the most commonly-cited studies performed to answer the question of how best to quantitatively predict college performance. Again and again, we have found that SAT scores explain variance in college GPA and graduation rates not explained by GPA. In other words, they provide useful predictive information that can be utilized to reduce the number of students who are accepted into college who then fail out, a major negative event in a young person’s life due to opportunity cost and taking on student loan debt. The predictive effectiveness of the SAT is inconvenient for liberals who hate the test, but it is about as empirically well-justified as a claim about education can be, so they try to dissemble their way around it. The simple fact of the matter is that, for making the determinations that college admissions departments are meant to make, the SAT is a useful tool. Getting rid of it just makes us dumber.

How well do “holistic” factors predict college success? lol, nobody even pretends they can do that.

[...]

One of the bizarre elements of this conversation is that critics of the SAT routinely argue for the social value of replacing GPA+SATs with GPA, and criticize the SES effect in SATs while doing so. But GPA is also SES-influenced as well!

In the Zwick and Green study cited above, GPA correlates with SES at .171 - smaller than the SAT correlation, but not much smaller. As with SATs, I would call this a real but ultimately small association. GPA-SES correlations jump around a lot more than SAT-SES correlations do depending on study (and this says not-great things about GPA as a predictor, incidentally) and I invite you to investigate other datasets. But there is no question that GPA is attenuated by SES, a fact SAT critics simply ignore.

This association is real even though GPAs are generally normed, systematically or not, to show a grade distribution through a given school population. That is, educators in specific contexts will tend to assign a distribution of scores throughout their classes even though their classes are non-randomly assembled and represent a skewed performance sample regardless of context. (This means, among other things, that high school GPA is a less objective and consistent predictor, which is precisely the problem the SAT is meant to solve; the SAT’s critics have literally no answer when you point out the inherent subjectivity of GPA to them.) This is why many or most colleges have algorithms that they use to adjust raw reported high school GPAs, which they know they can’t trust - schools will go to great lengths to unduly privilege their students in the college rat race. And there’s every reason to think that the adjustments colleges make to GPA likely deepen the influence of SES. So the criterion that many SAT critics would like to focus on exclusively to the detriment of the SAT replicates or even intensifies their claimed problem with the SAT. It makes no sense. I think people are not doing a lot of thinking here.