There is a vibrant debate about the role of standardised test scores in US college admissions. The Covid-19 pandemic accelerated the trend of test-optional admissions, with colleges no longer requiring applicants to submit either SAT or ACT scores. Even after the pandemic’s physical disruptions receded, most colleges in the US have decided to stay test optional, at least for now.
Schools that go ‘test optional’ often claim that doing so will increase racial and socioeconomic diversity on campus. After all, students from affluent backgrounds have more access to resources, such as tutoring, to improve their scores. Study after study has shown that standardised test scores are strongly linked to family income and race (Camara and Schmidt 1999, Freedle 2003, Gurantz et al. 2018). But proponents of standardised tests counter that these tests are useful in predicting students’ success in college and that other components of applications are also subject to racial and income disparities.
Indeed, the original intent behind the SAT was to identify students with potential, regardless of background.
Many observers, including some academics (e.g. Garg et al. 2021), frame the choice between test-mandatory and test-optional policies as a trade-off between the information provided by tests with the tests’ biases or exclusionary properties. In our paper (Dessein et al. 2023), we argue that this perspective is incomplete. The monetary costs of taking a test are low (and are waived for low-income students), as is the physical cost of sitting for a test – at least outside of pandemics. If the vast majority of students can access some test score (even if only a low one) and if a college can use these scores as it would like, observing scores can only help it make better decisions. After all, colleges have always been free to admit students with low test scores if they are strong on other dimensions.
We propose, instead, that the trend of test-optional policies is better understood as a response by colleges to increasing social pressure.
Social pressure as a driver of test-optional admissions
Consider a college that admits a low-scoring student while rejecting a high-scoring student who has a similar GPA, perhaps because of legacy or diversity considerations. The college may face pressure from ‘society’ – parents, alumni, media, government, courts – that disagrees with the weight the college puts on tests versus legacy status or diversity.
Such social pressure is exemplified by two lawsuits, currently under review by the US Supreme Court, challenging the admissions policies of Harvard and the University of North Carolina.
We contend that a college can reduce social pressure by going test optional or even test blind. The idea is that by hiding score disparities among students who do not submit their test scores, observers can no longer point to as many, or as extreme, examples of individual students that it thinks were wrongfully accepted or rejected. The reduced disagreement between college and society may allow the college to admit students it likes more, based on diversity, extracurriculars, or legacy preferences. Importantly, the college can benefit both when it wishes to be less selective than society (i.e. to use a lower test-score threshold) or more selective (a higher threshold).
A theoretical framework
We model a college that has preferences for which students to admit, based on both their non-test observable characteristics (GPA, race, socioeconomic status, extracurriculars, legacy status) and test scores. Society has its own preferences. Owing to social pressure, the college places some value on minimising disagreement between its admission decisions and those that society would make.
The college chooses an admissions policy: a rule that determines admission decisions based on observables and test scores, and – in a test-optional regime – an imputed test score assigned to students who don’t submit scores (which can depend on non-test observables). Students submit their test scores only when they are higher than what the college would impute.
Society assesses the college’s decisions rationally using the available information, meaning that non-submitters are evaluated based on their expected test score, given non-test observables and submission behaviour. Whenever society disagrees with the college’s admission decision, the college incurs a cost proportional to the degree of disagreement.
Test-optional admissions with flexible imputation
When a college can freely choose how to treat students who do not submit a score, the college can’t be worse off under test optional: it can simply replicate the test-mandatory outcome by imputing such a low test score that all students submit. But the college can often do better.
To see how, consider a student with non-test observables at which the college is less selective than society. For instance, take students who excel in fencing, supposing the college values fencers more than society does. One option for the college is to impute a very high score for these fencers, with the policy of admitting all those with this score or higher. Then very few fencers submit their scores, and all fencers are admitted.
The cost for the college is that it may be admitting some very low-scoring fencers. The benefit is that since high-scoring fencers no longer submit, society (rationally) understands that the average test score of the non-submitters is not particularly low. This means that the college does not face much disagreement for admitting the fencers that the college wanted but society did not. Indeed, if society is willing to accept fencers with near-average test scores, imputing a very high score allows the college to accept all these now-undifferentiated fencers with no disagreement.
In general, when the college is free to choose its imputation, the college will admit more students with observables that it values more highly than society.
Now suppose the college is more selective than society – the college has a higher test-score bar than society for, say, applicants from New Jersey. It then benefits from imputing a moderate test score for non-submitters and rejecting non-submitters. Since only high-scoring New Jerseyans submit their score, there is no differentiation anymore between low-scoring New Jerseyans, whom both society and the college want to reject, and average-scoring New Jerseyans, whom society wants accepted but the college does not. This obfuscation of scores under test optional allows the college to reject average-scoring New Jerseyans with little or no disagreement.
In general, when the college freely chooses its imputation, it admits fewer students with observables that it values less than society.
For test-optional admissions to never harm a college, the college must choose its imputation rule judiciously. In practice, we see many schools promising that non-submitters will be treated ‘fairly’. Although these promises are ambiguous, they seemingly correspond to something like a Fifth Amendment ‘no adverse inference’ imputation rule: a non-submitter is imputed their expected score given other observables, but not conditioning on non-submission.
A more extreme imputation rule, where non-submitters are imputed very high test scores, essentially yields test-blind admissions, as all students would withhold scores. When constrained to use some fixed imputation rule, colleges may strictly prefer test mandatory over test optional.
Unintended consequences of a ban on affirmative action
Ongoing court cases are widely expected to result in the US Supreme Court severely limiting race-conscious admissions. Using our theoretical framework, we can analyse how a ban on affirmative action may result in more colleges choosing not to see test scores.
Consider a college with affirmative-action preferences. Conditional on all other characteristics (test scores and some non-test observables), it also has preferences over a student’s group membership, e.g. their race. Society has the same preferences as the college over other characteristics, but its preferences are group-neutral.
When affirmative action is allowed – admission decisions can be conditioned on group membership – the college chooses test mandatory over test blind in our model. The college can use different score thresholds for admitting students of different groups, and it values test scores enough to outweigh the disagreement cost.
If affirmative action is banned, however, the college may switch to test blind. Intuitively, if students in the college’s favoured group have lower test scores, then the college values tests less when it can no longer condition on group membership (as in Chan and Eyster 2003). Now, the reduction in disagreement costs from hiding test scores can outweigh the loss of information.
One implication is that banning affirmative action may backfire for society. Society prefers the college to use test scores but ignore group membership. When banning the use of group membership leads the college to stop using test scores too, society can be worse off.
References
Camara, W, and A Schmidt (1999), “Group differences in standardized testing and social stratification”, College Entrance Examination Board Report No. 99-5.
Chan, J, and E Eyster (2003), “Does banning affirmative action lower college student quality?”, American Economic Review 93(3): 858–72.
Dessein, W, A Frankel, and N Kartik (2023), “Test-optional admissions”, CEPR Discussion Paper 18090.
Freedle, R (2003), “Correcting the SAT’s ethnic and social-class bias: A method for reestimating SAT scores”, Harvard Educational Review 73(1): 1–43.
Garg, N, H Li, and F Monachou (2021), “Dropping standardized testing for admissions trades off information and access”, arXiv preprint arXiv:2010.04396.
Gurantz, O, J Smith, and J Goodman (2018), “Take two! SAT retaking and college enrolment gaps”, VoxEU.org, 4 November.
Pew Research Center (2022), “As courts weigh affirmative action, grades and test scores seen as top factors in college admissions”, 26 April.
Tergiman, C, R Siegel, S Lychagin, W Olszewski, and K Krishna (2022), “Reducing college admission stress”, VoxEU.org, 31 August.
