We aren't getting rid of SATs because they don't work

As Jazz pointed out yesterday morning, the UC system in California has decided that it will no longer consider SAT or ACT scores in either admission or scholarship decisions. The reasons for this move weren’t hard to discern:

“Today’s settlement ensures that the university will not revert to its planned use of the SAT and ACT — which its own regents have admitted are racist metrics,” said Amanda Mangaser Savage, a lawyer representing the students…

The settlement resolves a 2019 lawsuit brought by a coalition of students, advocacy groups and the Compton Unified School District, a largely Black and Hispanic district in Los Angeles County. The plaintiffs said that the college entrance tests are biased against poor and mainly Black and Hispanic students — and that by basing admissions decisions on those tests, the system illegally discriminates against applicants on the basis of their race, wealth and disability.

But writing at Substack, Freddie deBoer has written a piece arguing that the reason people on the left hate the SAT isn’t because the tests are bad at predicting college success but in spite of the fact that they are actually pretty good at predicting such success.

Liberals repeat several types of myths about the SAT/ACT with such utter confidence and repetition that they’ve become a kind of holy writ. But myths they are.

SATs/ACTs don’t predict college success. They do, indeed. This one is clung to so desperately by liberals that you’d think there was some sort of compelling empirical basis to believe this. There isn’t. There never has been. They’re making it up. They want it to be true, and so they believe it to be true.

The predictive validity of the SAT is typically understated because the comparison we’re making has an inherent range restriction problem. If you ask “how well do the SATs predict college performance?,” you are necessarily restricting your independent variable to those who took the SAT and then went to college. But many take the SAT and do not go to college. By leaving out their data, you’re cropping the potential strength of correlation and underselling the predictive power of the SAT. When we correct statistically for this range restriction, which is not difficult, the predictive validity of the SAT and similar tests becomes remarkably strong.

He then moves on to a second, related argument.

The SATs only tell you how well a student takes the SAT. This is perhaps a corollary to 1., and is equally wrong. They tell us what they were designed to tell us: how well students are likely to perform in college. But the SATs tell us about much more than college success. Let me run this graphic again.

The SAT doesn’t just predict college performance, though it does that very well. It predicts all manner of major life events we associate with higher intelligence. And since the SAT and ACT are proxy IQ tests, they almost certainly provide useful information about all manner of other outcomes as well.

The chart comes from this 2010 study. So if the SAT does work, why refuse to use it? As mentioned above, the objection is largely based on group differences by race.

If you refer to a racial achievement gap in a lot of liberal or left contexts now, you’ll find that people clam up fast and get visibly uncomfortable, even if you take pains to point out that an academic achievement gap does not imply an academic potential gap. People just don’t want to acknowledge that gaps exist at all; our racial discourse appears to have become such a blunt instrument that the acknowledgement of racial difference is controversial even when you preface discussion with the belief (that I hold) that the gap is the product of innumerable environmental and sociocultural factors rather than genetics or other inherent differences. Simply saying “Black students consistently score lower on tests like the SATs, have lower average GPAs, and have worse metrics on ancillary concerns like truancy” – again, Barack Obama’s position, Kamala Harris’s position, Cory Booker’s position – is enough for people to start launching into harangues about the inherent violence of those comparisons. People just do not want to talk about this stuff.

The rush to rid the world of the SAT is based on this dynamic. Because Black and white students are not equal in academic preparedness, and because we have failed to close the gap in a half-century of concerted policy effort to do so, we must eliminate the tools that reveal it, such as the SAT.

The problem of course is that eliminating the test that reveals the gap won’t do anything to close it. Or as deBoer puts it. “The SATs don’t create inequality. They reveal inequality.” But while getting rid of the tests is seen as a way to help those who aren’t performing as well on them, it also necessarily hurts those who are. In this case, that means the Asian students who typically outperform all other groups on SATs are losing an advantage in school admissions.

All of this goes back to that argument over admissions at Harvard where Asian students routinely outscore every other group in academics but then curiously do worse on the individual personality score than other races. What happens when the SAT score is removed from the process? Do Asian applicants suddenly do better on the personality section? Or do fewer Asians get admitted? The outcome will obviously be a significant one to those students and their parents.