In 2020, many colleges suspended their requirements for standardized test scores as the COVID-19 pandemic forced testing centers to close. That same year, the University of California Board of Regents announced it would phase out its SAT requirement for prospective students in a bid to increase equity and diversity. At the time, Yale SOM’s Faidra Monachou was getting her doctorate in Stanford’s Management Science and Engineering department, and she and her fellow graduate students thought they could use tools from operations research to understand what trade-offs schools were making between losing the information they gained from test scores, and gaining wider accessibility for students who might face barriers to taking the test. The decision to drop test scores looked very much like “an operational problem,” Monachou says. “We could develop a mathematical framework to formally study the trade-offs between information and access.”
They knew that colleges and universities typically have three goals when they design their admissions process: maximize capacity, ensure a high level of academic merit, and build a diverse cohort. Each component of a college application—say, a test score, grades, a personal essay, and a recommendation—theoretically provides information that can help achieve those goals.
Dropping the test score could help reduce the informational gap in the application components of students from different backgrounds, or it could widen it.
For a new study, forthcoming in Management Science, Monachou and her co-authors, Cornell Tech’s Nikhil Garg and Columbia University’s Hannah Li, developed a mathematical model that accounts for all of these moving parts in order to formalize how this trade-off can ultimately influence an admitted class’s academic merit and diversity. In 2020, the policy debates playing out in op-eds and think tanks often began with the assumption that dropping the test score would lead to more diversity at the expense of merit. They found that it’s not that simple.
In fact, she says, “it is possible that dropping the test score can simultaneously improve or simultaneously worsen both merit and diversity.”
This is because each component of a student’s application can vary in how genuinely informative it is. An A+ grade at a school able to offer rigorous classes is more informative than an A+ from other schools. This can be thought of as the “differential informativeness” that surrounds each component and gives it more or less predictive power regarding a student’s academic merit. Moreover, schools and applicants act strategically with respect to their options and competitors, interactions that will also affect the applicant pool. All of these factors interact in a complex fashion to ultimately influence the outcome of eliminating the test score component.
Monachou gives the example of two students with high GPAs who apply to Yale. One is from a New Haven high school, in Yale’s backyard, so the admissions committee is familiar with the high school and knows how to evaluate those students. The other student applies from an unfamiliar school. In this case, the variation in informational context for the GPA is quite wide, presenting a lot of uncertainty around how predictive the GPA will be for a student’s ultimate academic merit. In such a situation, if test scores are similarly informative across high schools, requiring a test score component decreases the overall uncertainty in the applications.
Although the SAT is a standardized test, it too presents some uncertainty about whether it is directly predictive of a student’s academic merit. Some students receive intensive tutoring and test prep, and others take the test cold. Still others face hard barriers—perhaps no one at their high school alerts them to register, or they cannot afford the registration fee—or soft barriers, such as a weekend job or caregiving responsibilities that might make it challenging to carve out time to sit for the test. Thus, dropping the test requirement could allow a school to welcome additional academically adept students who were less inclined to take the test.
But whether that is the case again depends on whether dropping the test component reduces, or increases, the overall levels of informational uncertainty across the other components of the application. “It could help reduce the informational gap in the application components of students from different backgrounds, or it could widen it,” Monachou explained, depending on those other aspects of the application and how informative they are.
Schools do not act in a vacuum, though. When the UC system dropped its testing requirement, a move that made it more attractive to applicants who faced higher costs to taking the test, other schools followed suit. Students also weigh the costs and effort of taking the test against their chances of admission: some selective schools still require a standardized test score or offer students the option to submit one. “Middle-performing students may opt out, preferring guaranteed admission to a less selective school compared to taking the test and applying to the more selective one, even if they do have a chance of being admitted,” says Monachou, which means that more selective schools could lose applicants that they might otherwise prefer to admit.
Each of the application components can be thought of as “an important lever for schools to achieve their goals,” says Monachou. The researchers’ model can help colleges think about how to find an optimal balance of components so as to make the information they provide relatively equal—or, at least, equally uncertain—between students from more familiar backgrounds and those from underrepresented backgrounds. Or it can help them identify which levers to pull depending on whether their goals are more oriented towards, say, expanding the applicant pool in order to fill seats, increasing diversity, or trying to raise the academic skill of the admitted class.
The most important thing, Monachou emphasizes, is to understand how informative each other part of the application is relative to the test score. “The decision to drop the test cannot be made without jointly considering the interaction between the information it provided and the relative predictiveness of the other application components,” she said. “This interaction is very complex.”
“The Yale School of Management is the graduate business school of Yale University, a private research university in New Haven, Connecticut.”
Please visit the firm link to site
