Test-Optional is a Failure: The Return of the SATs

I saw this going viral: Harvard will require test scores for admission again (archive link)

“Critics correctly note that standardized tests are not an unbiased measure of students’ qualifications, as students from higher-income families often have greater access to test prep and other resources,” Chetty said in a statement Thursday. “But the data reveal that other measures — recommendation letters, extracurriculars, essays — are even more prone to such biases. Considering standardized test scores is likely to make the admissions process at Harvard more meritocratic while increasing socioeconomic diversity.”

This is a long way of saying that extracurriculars and ‘holistic’ admissions favors wealthier applicants and may disadvantage lower-SES applicants. Who knew. Test-optional has not worked, because new admittees who fail to submit test scores are not up to par, and test-optional hurts low-SES students who may not be able to participate in expensive extracurriculars or ‘character building’.

As it turns out, extracurriculars are a strong proxy for SES-status or parental involvement, not competence or character. Also, GPAs, transcripts, and letters of recommendation can be inflated and are teacher and school specific. Same for science fairs and ‘independent’ research projects, which also tend to have a lot of uncredited parental help; this is harder to do with the SATs. Wealthy parents will do anything, even breaking the law, to give their dull kids an advantage. Standardized tests are one of the few ways of leveling the playing field a bit.

A common argument is that the SATs only predict or measure test-taking ability, or that GPAs alone are sufficient. But top SAT scores are more predictive of outlier ability than high GPAs. A 3.0-4.0 GPA is demonstrative of proficiency of the material, but a top SAT score, which is much rarer, signifies uncommonly high intelligence, not just proficiency. Thanks to dumbing-down and grade inflation, 4.0s are handed out like Halloween candy, but math competitions and the SATs have a much higher ceiling and are more impartial and less school or teacher-specific.

Owing to recent revisions to the SATs to have a lower ceiling compared to older (circa 1995) versions, high-stakes math competitions serve as a sort of de facto ultra-high-ceiling standardized test. Same for DIY coding repositories on Github for those who aspire to a computer science track.

Regarding the common argument that the SATs only measure test-taking ability, it’s reasonable to assume that someone with a top math SAT score or top placings in math competitions has more potential for becoming a mathematician or success at STEM than someone who scores low; same for top verbal SAT scores and the humanities. (I don’t have to assume it; many competitions winners and finalists have gone on to produce important research or achieved material success (e.g. business or hedge funds) than predicted by chance alone.) So given a finite number of slots, a top-20 college that seeks to maximize the likelihood of producing important research or otherwise maximizing prestige will want to prioritize top scorers.

About test prep and coachability, despite the ubiquity of test prep, top scores are still uncommon. This is especially true for the LSATs. As linked above, the Varsity Blues scandal showed the limitations of test prep. If test prep is so effective, why did the parents indicted in the scandal pay for people to take the SATs on behalf of their children, or fabricate/doctor scores, or resort to costly bribes instead of the far cheaper and legal test prep? This is not to say that the SATs are impervious to coaching; coaching can help, but given the ubiquity of coaching and prep, still, top scores are still relatively uncommon, and is still more impartial than a holistic or test-optional admissions process.