Through all the decades of controversies about admissions and about affirmative action, there has been the idea, shimmering like the mirage of a desert oasis on the horizon, that there might be some way of selecting students for elite universities that everyone would agree is fair. It’s the ever-elusive admissions equivalent of the Holy Grail, or the universal solvent, or the Northwest Passage.
The main conservative version of this is to eliminate consideration of race in admissions. (Even the current Supreme Court is apparently not completely comfortable with the racial effects of strict “colorblindness”; that appears to explain its having declined in 2024 to hear a case challenging the admissions policies of a highly selective high school in Virginia, which substantively was almost identical to the 2023 Harvard and University of North Carolina cases.) Then there’s the idea of eliminating the admissions preference for children of alumni — something that two unsuccessful presidential candidates, Bob Dole and John Edwards, have proposed as federal legislation over the years, and that several prominent colleges, including Amherst, Johns Hopkins, and Wesleyan, have recently done. Preferences for athletes and for donors’ children could also be eliminated.
Probably the most persistent reform idea is of a race-blind socioeconomic preference. Anthony Carnevale, when he was an executive at the Educational Testing Service in the 1990s (he’s now director of Georgetown University’s Center on Education and the Workforce), produced a “Strivers Index,” which would identify SAT takers who got scores significantly higher than one would have predicted from their socioeconomic background. Some years earlier, another ETS psychometrician, Winton Manning, proposed adjusting the SAT score itself to account for a student’s socioeconomic background. Richard Kahlenberg, of the Progressive Policy Institute, has been advocating for class-based rather than race-based affirmative action for decades, including in briefs he has filed in challenges to affirmative action that have come before the Supreme Court. (Even Kahlenberg, however, has endorsed race-conscious policies meant to integrate the student bodies of public schools rather than selective universities.) During the arguments before the court in the most recent cases, several justices asked about the possibility of keeping affirmative action in admissions, but on the basis of class rather than race.
In October 2023, a few months after the Supreme Court decision banning race-based affirmative action in admissions, Opportunity Insights, a Harvard-based research institute directed by the economist Raj Chetty, issued a detailed report on admissions at the Ivy Plus schools — the most highly selective universities in the country. The report’s headline finding was that students from families in the top 1 percent of the income distribution have a much higher chance of admission than anybody else — and that isn’t because they have superior grades or test scores; the primary mechanisms by which their parents convert family money into admission are the preferences for children of alumni and for athletes. The obvious implication is that getting rid of these preferences would make superelite admissions — which, the report takes pains to stipulate, pays off handsomely in the form of access to high-paying private sector jobs — considerably fairer.
The report also showed that SAT scores are far more predictive of academic performance than high-school grades — so it seemed to be arguing for reinstituting SAT requirements, as several Ivy Plus schools did after the study was published. Chetty and his coauthors, David Deming and John Friedman, avoided confronting what has always been the main issue with the SAT and admissions — the strong negative effect the test has on the Black presence at elite schools — by not presenting any data disaggregated by race. (Data from the College Board shows that in 2023, the gap between average Asian and Black scores on the SAT was more than 300 points, and that nationally, fewer than 2,300 Black students got combined scores of 1400 or above, which is generally considered what a student needs to be admitted to an Ivy Plus school.)
Chetty’s and his coauthors’ heavy emphasis on applicants from families in the top 1 percent also obscures the strong association between SAT scores and class. A chart that can be found in their supporting data shows the distribution of test scores among students enrolled at Ivy Plus schools by parental income. For students with combined SAT scores of 1500 or above, every parental-income decile up to 90 percent accounts for less than 1 percent of students. For students with combined scores of 1400 or above, only at the 70-to-80-percent range does the proportion rise above 1 percent. The idea that SAT requirements are the only way to find students from disadvantaged families who would do well at Ivy Plus schools doesn’t take into account how few such students there are, not to mention what the racial effects of SAT requirements would be.
It’s best to understand the various alternatives to affirmative action as testaments to the insolubility of the problem of selective-college admissions, and of meritocracy more broadly — not to how easy the current system would be to fix. Most of the highly selective private universities have kept their preferences for athletes and the children of alumni and donors for fundraising reasons, but it’s also the case that because these applicants often come from privileged backgrounds, they are not obviously less academically credentialed than the overall applicant pool. Applicants from less advantaged backgrounds already receive informal preference at selective colleges, but if one is primarily concerned about the fairness of the system to each individual applicant, this practice doesn’t pass the test. It necessarily entails denying a place to someone with superior credentials and giving it to someone else for social-justice reasons. The battle over affirmative action really boils down to whether race is the one and only offensive admissions preference out of all the preferences.
Selective admissions is a zero-sum game that gets ever more severe. Everything about making college admissions a highly consequential socioeconomic choke point generally favors students from fortunate backgrounds. It might help make selective college admissions fairer if (in the manner of many elite admissions systems in other countries) we had a national curriculum and a national system of public-school funding, and if we imposed some kind of service requirement on selective-university graduates so that admissions didn’t so much resemble a ticket to private-sector success. But none of that is going to happen in the foreseeable future.
What if we gave ourselves the project of designing a national testing system to help manage the transition between high school and college — what would we do? The situation today is radically different from what it was in the 1920s, when the Scholastic Aptitude Test was first developed, and not only because the United States is both more diverse and more conscious of diversity as a goal. Today, 90 percent of Americans finish high school and have some interaction with the higher-education system. The idea that there is a cohort of people of very high academic ability who will never enter a college or university no longer applies, though many people from disadvantaged backgrounds aim lower than they should and face severe financial and logistical obstacles as they enter the higher-education system. The idea that elite universities have no reliable means of evaluating applicants who didn’t go to a small handful of elite high schools no longer applies, either. Standardized testing, of many kinds, pervades American public education. And highly selective universities have decades of experience in evaluating the high-school records of students from a wide variety of institutions.
Selective admissions is a zero-sum game that gets ever more severe.
More fundamentally, the question of who should get the slots in highly selective colleges and universities, besides being very difficult to settle, is not the primary issue in American higher education, especially from the point of view of enhancing democracy and opportunity. It gets far more attention than it deserves. Only about 1 percent of American undergraduates attend the 25 or so colleges that accept fewer than 10 percent of their applicants. Only 3 percent attend the 50 or so colleges that accept fewer than 25 percent of their applicants. The selectivity of admission to college is not a major factor in the lives of the overwhelming majority of young Americans. The widespread administration of the SAT and ACT to millions of people in order to identify a relative handful to admissions officers at highly selective colleges, as if it were the glass slipper used to find Cinderella after the ball, creates the impression that it is more consequential to most test takers than it actually is.
The most obvious problem in American higher education today is not its inability to identify and educate top academic performers but its failure to produce a more widely successful experience for most students. Only about 40 percent of entering students get a bachelor’s degree in four years, and only about 60 percent in six years. At community colleges, only about 30 percent of entering students complete a two-year degree program within three years. Completion rates are lower for Black and Latino students than for white and Asian students, lower for low-income students than for middle-class and affluent students, and lower for men than for women.
At the same time, despite mythology to the contrary, completing a degree pays off. A 2021 study by Carnevale, Ban Cheah, and Emma Wenzinger found that completing high school increases a person’s lifetime earnings by 33 percent. A two-year associate degree adds 25 percent to lifetime earnings over a high-school diploma, and a four-year degree adds 75 percent to lifetime earnings. Graduate degrees add even more. The study takes pains to point out that the return to higher education varies significantly by field of study; Carnevale and his colleagues have been strong advocates of a “return on investment” approach to higher-education degrees that entails disaggregating the economic payoff by academic major. Their unsurprising finding is that business, engineering, mathematics, and computer-science majors earn far more than the general run of graduates. Overall, though, completing a higher-education degree is about the surest way for a young person to get on the path to a better future. Conversely, the current low degree-completion rate is a glaring gap in the American opportunity structure, and bringing it higher ought to be an urgent national priority.
The United States made an idealistic bet on universal elementary education in the 19th century, and a similar bet on universal high-school education in the early 20th century, and a similar bet on universal higher education in the late 20th century. At each stage there were skeptics who warned that the country was trying to educate the uneducable. Today, a good way into the 21st century, our progress in this long-running project of demonstrating a commitment to ordinary people’s potential has stalled, because the massive higher-education system we have built is not delivering the results it should for so many of its students. The great bulk of higher education takes place in unselective or minimally selective institutions, mainly in the public sector. What would most enhance opportunity for most Americans would be a successful passage through these institutions.
Could testing serve as an aid in that project? Yes, potentially, if it were testing of a different kind from what we have become accustomed to over all these years. The system of higher-education testing built around the SAT and similar tests — aptitude tests aimed at selection — was not designed with the primary aim of distributing educational opportunity widely. At the time when the system was introduced, there were other, competing ideas about what the main focus of American higher education should be, and their advocates had quite different kinds of tests, built for different purposes, in mind.
Testing is, of course, ubiquitous in education at all levels — it’s almost impossible to imagine schools without both formative tests (like quizzes) and summative tests (like final exams). And there have always been voices in the never-ending debates about education that express an instinctive suspicion of all forms of testing, for the way they emphasize mastering material, rather than creativity and independent thinking, as a central purpose of schooling. For our purposes here, the question is whether tests that are made to be widely usable outside the confines of a single course at a single school could serve the aim of increasing educational opportunity broadly, for most people rather than for a lucky few, and in particular whether they could help to increase degree-completion rates at community colleges and public universities.
In the 1951 book Educational Measurement, assembled by George Zook of the American Council on Education and E. F. Lindquist of the University of Iowa, the assumption running through all the chapters is that the one and only appropriate purpose of standardized testing is to enhance student learning as broadly as possible. (And indeed, that was also Alfred Binet’s assumption when he introduced the first intelligence test back in 1905 — it’s a long-running tradition.) Testing is meant to help teachers teach their students more effectively, and therefore to help students achieve more in school. The testing pioneer Ralph Tyler, of the University of Chicago, defined teaching this way in his contribution to Educational Measurement: “Basically, instruction is the process by which desirable changes are made in the behavior of students, using ‘behavior’ in the broad sense to include thinking, feeling, and acting. Instruction is not effective, therefore, unless some changes in the behavior of students have actually taken place.”
The system of higher-education testing was not designed with the primary aim of distributing educational opportunity widely.
You can see from this what the appeal of standardized tests would be. Most of the contributors to Educational Measurement believed in the importance of what we’d now call learning outcomes: They were frustrated by the way students were being passed along in the education system merely by completing course material, and tested mainly on their mastery of sets of facts rather than the ability to think. New tests would help teachers define their higher aims more precisely and impart them to students in ways that would be profoundly helpful over the long term. For a student, not doing well on a test would open the way to extra educational opportunity — more intense teaching, toward a focused and meaningful educational goal — rather than signaling a limitation of opportunities.
Richard Atkinson, the former president of the University of California system, launched a highly influential critique of the SAT in 2001, which came after half a century of testing developments that weren’t particularly well aligned with the vision of Educational Measurement. Atkinson highlighted a series of distinctions in testing that are important to keep in mind and are often ignored in testing debates that tend to treat tests as either all good or all bad. A standardized test can be designed for diagnosis or for prediction. The SAT is a predictive test, meant to forecast academic performance in college. A diagnostic test is meant to gauge a student’s level of prior learning — for example, to help determine what level of course in a subject the student should take in college — and to identify areas where the student may need special help.
A related distinction is the familiar one between an achievement test and an aptitude test. An aptitude test is designed to measure general-purpose academic ability. (Although the College Board and ETS have always stoutly insisted that SAT scores are used only in college admissions, and even there only cautiously, some employers that prize braininess, like hedge funds and technology companies, regularly ask job applicants for their SAT scores, if the applicants haven’t already listed their scores on their résumés.) An achievement test measures mastery of a subject. For an achievement test, test prep and studying the course material should be the same thing; for an aptitude test, studying is supposed to be unnecessary, so test prep is the only available form of homework.
A final distinction is between norm-referenced and criterion-referenced testing. A norm-referenced test (and again, the SAT is an example) measures where each test taker stands within the total pool of test takers, so as to let college admissions officers aim for, say, the top 5 percent of all scorers on the test. A criterion-referenced test measures a student’s understanding of a specific body of material, so that the student is not compared with other students. In the proverbial elementary-school weekly spelling test, in theory, all the students could get perfect scores, because it’s a criterion-referenced test.
The SAT — a predictive test, an aptitude test, and a norm-referenced test — was designed to help elite colleges select a small handful of students. The reason it was designed to do that was that higher-education leaders were trying to make their institutions more academic and locate talent they believed they would otherwise miss. It wound up having a far broader effect, because of its use by many more colleges and universities than the original group, and because of its large impact on the high-school experience of millions of students who don’t go to highly selective colleges, in addition to the thousands who will.
If, today, we define the problem that testing is meant to solve not as improving selection for a few elite universities but as improving the too-low graduation rates and other aspects of the student-learning experience at a large number of relatively unselective universities, we would be drawn to diagnostic rather than predictive tests, to achievement rather than aptitude tests, and to criterion-referenced rather than norm-referenced tests. And these, in order to serve the purpose they are meant to serve, would have to go along with larger structural changes: a much greater emphasis on teaching and advising in higher education, and a strengthening of the curriculum in high school. This ought to be a national project on the scale of the project that brought us the current higher-education admissions system, or on an even grander scale.
It looks as if the college-admissions tests that have been widely in place since the Second World War may be on their way out, except perhaps at a small number of highly selective schools. Fundamentally, that isn’t a sign that the country has “turned against meritocracy,” but that the tests, rather than representing an opposing force to the existing class system, by now, on the whole, reinforce it.
One sometimes hears lamentations that, if the many colleges and universities that have suspended the tests don’t resume their use, it will signal a wholesale abandonment of academic standards or of the principle of meritocratic admissions. Those fears are wildly overblown. Elite colleges and universities are overwhelmingly academic institutions, where the high-school record is the most important admissions criterion for applicants, and published academic research is almost the only criterion for the hiring and promotion of faculty. Because the standardized admissions tests pick up so much family background, the number of students who come from disadvantaged backgrounds and who went to underperforming high schools but who nevertheless get superior scores on standardized tests is tiny. It’s not worth designing a whole system just to catch a handful of people who, in the 21st century, will be going to college anyway. It would be far better to expend the country’s energies on making the higher-education system work better for the great bulk of students than on fiercely debating what amount to incremental adjustments in the apportionment of a highly limited number of spaces in highly selective institutions.
Standardized tests, rather than representing an opposing force to the existing class system, on the whole reinforce it.
The most useful next national educational project would be to focus on the broad end of the educational system, not the narrow end — to try to create through education as much meaningful opportunity for flourishing for as many people as possible. Testing could serve that project, if it were a quite different kind of testing, but the project shouldn’t be thought of as being mainly about testing. It would mainly be about adopting a different set of primary educational and social goals.
Given the history of testing, a measure of wariness about its widespread use is justified. Tests can take on a life of their own. They can be seen as measures of immutable traits, rather than as snapshots of students’ achievement at a moment in time — something that is highly mutable. They can be used to reduce educational opportunity rather than enhance it, and to disempower teachers rather than enable them to help their students learn more effectively. Because tests themselves are so visible and the purposes behind them are far less so, it’s possible, unless we’re vigilant, for the ongoing widespread use of testing to hide quite radical shifts in the social and educational ideas underlying them.
But it’s worth imagining the good that tests could do. They could give schools and teachers a better idea of exactly where a student is at the moment, which would permit a more effective, tailored approach to that student’s instruction. They could offer a sense of a student’s overall strengths and weaknesses that might shape future educational choices. They could assess a student’s level of knowledge of a body of material that has to be learned. They could serve as a check on students’ and teachers’ self-reports of how well the students are learning. All of these purposes have deep roots in the history of testing — deeper, in fact, than testing for selective admissions. The essential question is not whether to have tests; it’s what broad educational goals tests should serve.
This essay is adapted from Higher Admissions: The Rise, Decline, and Return of Standardized Testing (Princeton University Press, 2024).
