The big topic of discussion over the last week or so in elite-academic circles is the announcement that Amherst College is eliminating legacy preferences in admissions. If you’re not up on the jargon, that means they’re eliminating the boost given to applications from children of alumni. As a Williams grad, I am of course obliged to express performative disappointment whenever they do anything good, but this is the right thing to do.
I say that because the legacy preference has more than a slight whiff of obnoxious class bias to it. By making it easier for the children of elite-college alumni to attend the same schools their parents did, it can work to perpetuate the dominance of a particular social class of people who have traditionally attended these schools. It’s a vestige of the era when these were just finishing schools for the idle rich, and removing that preference is a welcome move toward something more based on individual characteristics, which is a step toward a more just and equitable system.
That said, I think this is largely a PR move, and unlikely to lead to any dramatic change in the composition of the next several classes of accepted students stuck going to Amherst, or any other college that formally drops a legacy preference. It’ll shift the numbers a bit, but I’d be very surprised if this change alone led to any substantial shift in the demographics of admitted students.
(Please note that I am not speaking in any kind of official capacity here, or drawing on deep insider knowledge of the admissions processes of elite higher education. In terms of actual data I have more or less the same public information everybody has access to. That’s supplemented by plural anecdotes from my time in the elite-college world, both as a student and as faculty, so assign those whatever weight you feel appropriate.)
That may seem like an odd claim, since a lot of the image most people outside the elite-college bubble have of the world inside involves a substantial number of indolent children of rich alumni, who got admitted only because their surname adorns multiple buildings at the school. Eliminating legacy preferences would throw those 80’s-teen-movie villains out on their ears and open space for virtuous and hard-working children of families with lower socio-economic status.
It’s a nice story, but the numbers don’t really work out. Most stories about the Amherst decision give the percentage of students there who are legacies as about 11%; that’s about 200 students total, for all four classes. A few years back, Williams was promoting the figure that 1 in 7 students were the first in their families to go to college, noting in passing that this was the same as the legacy percentage; that’d be about 290 total students over four classes (Williams has a slightly larger total enrollment than Amherst, as is right and proper). The most detailed data available on these preferences comes from the lawsuit about Harvard admissions practices (available in this working paper from 2019); using numbers from their Table 1, the legacy percentage in the admitted class is around 14%. (This is, it should be noted, a slightly different figure than the other two, which refer to students who are actually enrolled.) These are all in the same ballpark— not insignificant, but nowhere near a majority.
And it’s not like eliminating the legacy preference would see all of these students denied admission— the popular picture of the wholly undeserving rich kid getting in on parental money is largely a myth. I knew a bunch of folks at Williams whose parents and even grandparents were alumni, and they weren’t notably different from the rest of the students. They were as smart and hard-working as everybody else there, in my experience. I know a bit less about the backgrounds of the students I’ve taught at Union (since I’m generally not hanging out with them socially in ways that lead to asking, “So, what do your parents do?”), but I don’t have any direct experience with anyone coasting on their name (admittedly, the indolent children of the idle rich probably aren’t going out of their way to take physics courses…).
In both places I’ve heard stories about students who fit the popular image of undeserving legacies, but those are both hearsay and rare. They’re passed around precisely because they’re unusual— most legacy admits aren’t like that.
You can also see this reflected in the Harvard data, most compactly in their Table 5 (reproduced below), which shows the percentage of students in different groups receiving one of the top two ratings in each of the components they use to make admissions decisions:
“LDC” here is a composite category for Legacies, Dean’s interest list (that is, children of big (potential) donors), and Children of Harvard employees (“ALDC” adds recruited athletes to the mix). These are overlapping categories— lots of legacies are also athletes or big donors, some are children of faculty, etc.)— so the comparison isn’t completely clean, but it’s the best data available. You can see that, for every category, the percentage of applicants getting a high academic rating is slightly higher for LDC than non-ALDC applicants— these are, on average, well-qualified applicants.
(The percentage of admitted students with high academic ratings is slightly lower than for non-ALDC students, which shows the effect of the preferences given to LDC students. LDC students with lower ratings are more likely to get in than non-LDC students, so the highly-rated percentage in the admitted class is slightly lower.)
The same pattern holds for the non-academic rating components as well: LDC applicants are more likely to be in the top two tiers than the non-ALDC applicants. Across the board, this is a strong pool of students (setting aside the subjectivity inherent in a lot of these).
If you think about it, this is pretty much what you would expect for these students: they’re coming into the admissions process with a lot of advantages. To whatever extent heredity plays a role, they should be in good shape; they’re more likely to be well-off financially (sort of by definition for the big-donor category); and they’re more likely to “know how to play the game,” by virtue of having parents who went through the system. These are students who you would expect to have a better-than-average shot of getting in even without the benefit of an explicit preference.
So, the fundamental problem here, as often happens with social-science data, is that it’s really hard to know what the rate “should” be. That is, if you removed the preference entirely, how many legacy admits would you expect to have based on their own merit?
The Harvard data allow at least an attempt at putting a lower bound on this number, using Table 2 from the working paper linked above, which shows application data by academic rating and preference status. If you look at each of the ratings, and imagine removing the LDC benefit so that the same total number of students are admitted in each, but at the same rate for LDC and non-ALDC students (putting the athletic preferences aside), that would reduce the number of LDC admits by about 1600 students, to roughly 4.5% of the total admitted class (from a bit under 21%, using the numbers from the same columns of that table; that’s higher than the legacy percentage I quoted earlier, because it includes additional categories).
I suspect that’s going to dramatically overestimate the possible effects, though, because that’s just looking at the academic rating, not any of the others, where the LDC students are likely to score highly as well. And there’s also the issue of overlap with the athlete category (which I left out of the above; those preferences are a whole other keg of worms). There’s even a “how to play the game” component here— looking at the table, you see that the percentage of LDC applicants drops as you move to lower academic ratings, which suggests that students from those groups are less likely to be doing a “What the hell, it’s worth a shot…” when they have little chance of being admitted. All in all, the admission probability for the LDC students as a group is likely to be a bit higher than the average at a given academic rank even without an explicit preference.
This is further complicated by the final step of the process— the Harvard data are for admissions, and some of those admitted students will decide to go elsewhere. I don’t know anything about their yield for different categories of students, but it’s very plausible to me that legacy admits would be more likely to accept an offer of admission than students without a family connection to the place. Put it all together, and I think the expected shift from dropping the legacy preference is likely to take you from a low-double-digit percentage to a high-single-digit one. Which isn’t nothing, but it’s not going to substantially change the character of any of these schools.
So, again, I think this is ultimately a matter of symbolism and messaging, with a relatively limited practical effect. It’s the right symbolism and the right messaging, mind (and I hope Williams and Union will follow suit), but you shouldn’t expect it to have a transformational impact on higher education as a whole.
This got a bit wonkier than I had originally thought. That’s what happens when you give me tables of numbers to play with… Anyway, if you like this sort of thing, here are some buttons you can click:
And if you want to quibble with my back-of-the-envelope math, the comments will be open.