September 16, 2021 - 10:31am

Many US universities use standardised testing — known as SATs and ACTs — to (partly) determine who they give a place to. They also pay attention to “grade point averages”, GPAs — how well you did at school, basically — and to essays.

There’s been a move recently to drop standardised testing. The University of California colleges have scrapped it. And this week, Ron DeSantis, the governor of Florida, announced that Florida’s universities will do the same. The concern is that standardised testing disproportionately favours wealthy, white applicants.

I think that’s likely true. But scrapping standardised testing is still probably a bad idea, because all the other things you can replace it with are worse.

As Freddie de Boer points out, yes, SAT scores are correlated with socioeconomic status (SES), albeit weakly. But so are grade point averages, to almost exactly the same degree. And essay grades are more strongly correlated with household income, at least according to this big study from earlier this year. If you remove SATs, you’ll end up putting more emphasis on essay scores, and as Sanjay Srivastava of the University of Oregon says here, there’s then a strong risk that you end up making university admissions more directly linked to SES.

And the early evidence is that scrapping SATs doesn’t work. It’s only one year’s data, but the percentage of University of California applications from underrepresented groups “remained essentially flat as a proportion of the applicant pool”, while the percentage of applicants with low family income “decreased to 41.5 percent of applicants from 43.5 percent last year”. 

It’s worth remembering what the point of these tests is. The idea is to find talented students who will do well at university. And using a combination of standardised testing and GPA is a strong predictor of how well a student will do. If you remove testing, the predictive ability falls: GPA alone predict students’ success, but only about half as well as SATs plus GPA. Moving to “holistic” admissions — looking at essays, extracurricular activities, how good a “fit” the applicant is, to predict success — simply does not happen. But letting professors judge who they think best “fits” the university is almost inevitably going to end up with those professors picking the people most like themselves: people from disproportionately wealthy, educated backgrounds. 

Yes, black and Hispanic students do worse on SATs than white and Asian ones. But that is not primarily a problem with the SATs. As progressives rightly say, black and Hispanic students have faced generations of prejudice and inequality which make it harder for them to progress in the education system. SATs are measuring the impact of that inequality. Bridging that gap will be difficult; stopping people from measuring it is easier, but won’t solve the fundamental problem.

If we want to increase the number of poor and minority students in universities, removing SATs is unlikely to do it, at least if we then solely focus on the existing, equally if not more biased, alternatives. You could do other things — run it as a complete lottery, so that there is no attempt to find the smartest students, for instance. But that would be a huge change to the university system. In the US specifically, Jay Caspian Kang suggests hugely increasing the pipeline from (mainly working-class) community colleges to the Ivies and other top universities.

Removing standardised testing is not the answer, because it will simply put more weight on the opinions of flawed humans. As so often, the worry about bias in algorithms forgets just how biased human judgment is.

Tom Chivers is a science writer. His second book, How to Read Numbers, is out now.