The Future of Highly Selective College Admissions

Where do we go from here?

Welcome back! In case you’ve just found this newsletter, yesterday I talked about the trouble with holistic admissions. Today I want to talk about possible solutions to the problem of highly selective college admissions.

On Tuesday, one day after we published our paper, the U.S. Department of Education announced that it had opened a civil rights investigation into Harvard’s legacy admissions policy. This wasn’t a reaction to our research, but rather a formal response to a request filed 3 weeks ago by activist groups who claim that legacy preferences by their very nature discriminate against nonwhite applicants.

While the timing was coincidental, the cumulative attention has ratcheted up public pressure on Harvard and all other Ivy-Plus colleges to end the practice of granting preferential admissions to legacy applicants.

[Interested to hear more? Listen to The Daily’s episode from yesterday about our paper’s findings.]

Yesterday the NYT Upshot published a detailed look at legacy preferences, drawing on data from our paper. They reproduce Figure 7b in the paper (shown below), which estimates the size of the legacy boost and shows that it is larger for high-income legacies. Low- and middle-income legacy applicants are about 3 times more likely to be admitted than non-legacies with the same SAT/ACT scores; the boost is about 5x for families in the 99-99.9^th percentiles, and about 7x for the top 0.1 percent. The overall admission rate for qualified legacy applicants was 36.5%, compared to 9.7% for non-legacies.

Figure 7B: Admissions Rate at Selected Ivy-Plus Colleges for Legacy Applicants by Parental Income

Figure 7c looks at what happens when legacy students at a particular college apply elsewhere. Their admission rate at other schools is 11.3%. This tell us that legacies are a bit more qualified than the average applicant (11.3% vs. 9.7%). But the difference between 11.3% and 9.7% is much smaller than the difference between 36.5% and 9.7% (quick maths). This tells us that the legacy boost is very real and very large, and that it is not explained by any qualifications that matter to admissions officers at other colleges.

Figure 7C: Admission Rate by Legacy Status

Understandably, the pressure is mounting on Ivy-Plus schools to take a hard look at legacy preferences. What would happen if they stopped favoring legacies in the admissions process? Based on the internal data we collected from several Ivy-Plus colleges, we estimate that treating legacies like any other applicant would reduce their presence by about 112 students in an average class of 1,650, a decrease of 6.8 percent. It wouldn’t eliminate the existence of legacies on campus, because not all of them needed the legacy boost to get in. But it would cut their numbers substantially.

The key question is – who gets to fill the extra spots? I have two views on this question: one somewhat pessimistic, the other somewhat optimistic. I don’t know about you, but I like to hear the bad news first.

Ending legacy preferences wouldn’t solve the larger problem

Suppose legacy preferences were outlawed, but nothing else changed. Refilling the extra 112 spots with “business as usual” would have only a modest impact on socioeconomic diversity.¹

We estimate that eliminating preferential treatment of legacies would reduce the fraction of students in the top 1% of family income by two percentage points, from 16% to 14%. This is because not all legacies are in the top 1%, and because most of the students from the top 1% are not legacies. Ending legacy preferences would also reduce the share of families from the 95^th-99^th percentile, but only by 0.7 percentage points.

This is probably an upper bound, because it assumes that future legacy applicants do not respond to the loss in status. Remember – these students are all extremely academically qualified, and many of them come from very high-income families. My guess is that losing the legacy advantage would lead them to strive even harder on other dimensions like athletics and extracurriculars, undoing some of the impact.

Eliminating legacy preferences without addressing the other advantages created by wealth and privilege would make incremental progress at best. It might feel good, but it’s a half-measure. To be clear, I’m not defending the policy and I understand the outrage. But I worry that legacies have become a convenient scapegoat for a much more systemic problem.

Suppose that colleges end preferential treatment for legacies, but that the net impact is to substitute out rich legacy students and replace them with rich non-legacy students. Would we actually have made any progress toward fairer college admissions?

The world is filled with talented, but not rich, young people

I would prefer to see highly selective colleges look comprehensively at all the ways that their admissions practices are biased toward high-income applicants. One possible cause for skepticism is on the supply side – are there enough talented low- and middle-income students out there to make big changes without compromising academic standards?

The answer is yes. In our paper we look directly at the reserve pool of talent for a policy change that creates more available slots at Ivy-Plus colleges. We use not only our internal applicant data, but also the universe of SAT and ACT test-takers in the U.S. over multiple years.

We simulate the impact of a “need affirmative” admission policy. Specifically, we ask “what if students below the 95^th percentile of family income with high academic ratings were treated like richer families with the same ratings?” This gives them an extra bump in our simulations. We then scale down admission rates from the higher income groups to preserve the same class size.²

The results are striking. Because academic ratings are so much more predictive than other attributes, Ivy-Plus colleges could achieve more socioeconomic diversity AND better average outcomes (higher earnings, more graduate school attendance, etc.). There is no equity-efficiency tradeoff. Current admissions practices are selecting on factors that don’t predict success very well, so there’s breathing room to do better on both dimensions.

Moreover, this could easily scale to all 12 Ivy-Plus institutions. Our simulated need-affirmative policy increases the number of highly academically qualified low- and middle-income students in each Ivy-Plus college by about 250 students per school, or 3,000 in total. We estimate that there are nearly 25,000 low- and middle-income students per year who would meet high enough academic standards to be admitted under our hypothetical need-affirmative policy.

The bottom line is that we are nowhere near exhausting the supply of academically talented low- and middle-income students. Ivy-Plus colleges could admit many more of them, if only they had the space to do so.

I have two ideas for how to make things better, both of which have been proposed by others at different times.

Set a high academic bar, and stick to it

As I mentioned in yesterday’s post, we find that academic qualifications predict later life success much better than extracurriculars and other non-academic factors.

Still, I would not advocate for an admissions process that just picks people with the highest test scores and grades. Instead, I’d establish a highly rigorous academic cutoff that applicants are required to meet to be admitted. It could be based only on SAT/ACT scores, or it could be a combination of test scores and GPA (like the Academic Indices used by many schools). It could even be weighted by course difficulty. The cutoff would be high enough that most applicants don’t meet it, but still low enough that there are many more eligible applicants than spots.

Above that cutoff, I’d offer admissions spots randomly.³ If you meet the bar, you get a lottery ticket.⁴ You could meet diversity goals either with formal quotas (definitely illegal), or with probabilistic quotas achieved by giving low-income or first-generation students, students of color, students with disabilities, and other worthy groups extra tickets (probably illegal, but I’m not a lawyer).

I see three big advantages to this. First, it would limit the college admissions arms race. You still need to work really hard to achieve a high enough SAT score or GPA, but once you meet the standard, there is nothing more you can do. Smart and talented kids would be able to enjoy their childhoods a little. The second advantage is that if every Ivy-Plus college randomized above a cutoff, everyone who is qualified would get in somewhere. There would be no need for early decision, early action, and other shenanigans that colleges undertake to manage yield and appear more selective.

The biggest advantage that a transparently random process would create a healthier understanding of what it means to graduate from an Ivy-Plus university. Lottery winners and losers would know that the outcome was due to luck, not to their own worth. We’d put the admissions consultant industry out of business, and people would stop wasting so much time and effort trying to game the system.

Most importantly, everyone who made it to Harvard and places like it would know in their core that they belong, because they all met the academic standard. No more poisonous conversations about who deserves it and who does not.

Even better, let’s expand

I also agree with many others who have urged highly selective colleges to expand. (These arguments have been around for a while, but here is a good version written earlier this week.) I really do believe in the value of what we provide at places like Harvard, so it seems only right that we should try to serve more people! Especially low- and middle-income kids, for whom Harvard can be truly transformative.

Since I am an economist, let me play to type and frame the issue as a tradeoff.

We estimate in our paper that there are 157 “extra” students from very rich (top 1%) families in an average Ivy-Plus college class of 1650 (let’s round up and call it 10 percent). One option is to target the three areas of advantage we identify (legacies, recruited athletes, and high non-academic ratings) for special scrutiny. We do some simulations in the paper that show how well that would work. We think eliminating legacy preferences, balancing recruited athletes on income, and ending the non-academic ratings boost for high-income students would collectively get most of the way there. Each of those changes on its own would be a huge lift. Doing them all together would be a huge challenge for any leader of a selective college to undertake.

Another option is to expand the class by 10 percent. If I were an Ivy-Plus college president, I would strongly consider announcing a phased-in 10 percent expansion of class size, and I would pledge to allocate the newly created spots to low- and middle-income students. I personally would prefer that outcome to the first option. I would allow privilege to persist as it has, if in exchange I get to enroll more worthy students from the rest of the income distribution.

The core issue is scarcity of opportunity. Expanding each of the 12 Ivy-Plus colleges by 10 percent wouldn’t solve everything, but it would lessen the competition a little bit. And maybe, just maybe, pulling it off successfully would encourage a risk-averse group of college leaders to expand even more. It may sound like a dream, but if dreams die, life is a broken-winged bird that cannot fly.

That’s all, folks. I hope you enjoyed this first week of Forked Lightning. After this week I’ll pursue a less hectic schedule. I plan to write at least weekly from here on out, with newsletters coming out on Monday mornings (I may wait until Tuesday for the next one – I need a break!). I will cease with the navel-gazing, and instead explore the vast array of interesting research and policy activity related to education, skills, labor markets, technology, and the future of work. Stay tuned - and thank you so much for reading.

1. We did not estimate impacts on racial diversity, but my best guess is that the impact would also be modest at best, for the same reasons that apply to socioeconomic diversity.

2. You can see the results of these simulations in Table 6 of the paper.

3. Colleges could announce their estimated cutoffs ahead of time, and then adjust based on application volume and yield. This is like the Boston Marathon, which announces cutoff times for eligibility based on age and gender, and then allocates spots only to people below the cutoff. In some years you have to run faster than the cutoff to earn a spot because of excess demand. For example, if the cut time for men aged 45-54 is 3:20 but they get a lot of applications from that group, they might only offer spots to people who run a 3:17 or less.

4. I am hardly the first person to propose the idea of an admissions lottery. The first reference I could find was a 1997 op ed by Lani Guinier, who also proposed the idea of extra consideration for diverse applicants. A nice paper by Dominique Baker and Michael Bastedo simulates the outcome of such an admissions lottery using national data and gives a sense for demographic representation (a key finding is that there wouldn’t be enough men, who tend to have lower grades). This is a recent review article with more perspective on the idea of admissions lotteries for selective colleges.