"Legacy and Athlete Preferences at Harvard"

Peter Arcidiacono, the Duke professor who was the expert witness for Students for Fair Admissions in the Harvard case, has now (with co-authors) produced a paper on the subject entitled “Legacy and Athlete Preferences at Harvard”.
(http://public.econ.duke.edu/~psarcidi/legacyathlete.pdf). This is the abstract:

“The lawsuit Students For Fair Admissions v. Harvard University provided an unprecedented look at how an elite school makes admissions decisions. Using publicly
released reports, we examine the preferences Harvard gives for recruited athletes, legacies, those on the dean’s interest list, and children of faculty and staff (ALDCs). Among
white admits, over 43% are ALDC. Among admits who are African American, Asian
American, and Hispanic, the share is less than 16% each. Our model of admissions
shows that roughly three quarters of white ALDC admits would have been rejected if
they had been treated as white non-ALDCs. Removing preferences for athletes and
legacies would significantly alter the racial distribution of admitted students, with the
share of white admits falling and all other groups rising or remaining unchanged.”

Thanks for posting. Should be fun reading.

I am glad to see that the Harvard data is continuing to be analyzed. “Holistic” has always been a sham in the sense that its true workings are nothing like what it has been sold as.

LOL. I’d venture most folks dissing holistic have never been a direct part of holistic college reviewing.

From the paper, p. 3:

Anyone with common sense always understood that legacies are weaker than the average unhooked student, but it is nice to see confirmation. What perhaps most people fail to grasp is just how much weaker they must be.

Consider that the non-legacy group includes all the first generation students, by definition, as well as practically all of the low-SES group (obviously there are overlaps with the first generation), including all the Pell-eligible students. By contrast, it would be hard to imagine a group having enjoyed greater privilege than children of Harvard alumni and faculty, and of donors rich enough to pay the admissions price. And yet this privileged group is still weaker on average.

Interesting, considering SAT scores are highly correlated with family wealth, I would guess LDC applicants have higher (“stronger”) SAT/ACT scores as an aggregate, do they?

@makemesmart

I don’t think the linked paper specifically breaks out SAT/ACT score independent of Harvard’s internal “academic rating,” which primarily uses scores together with GPA to output a rating from 2 through 4, with pluses and minuses . (There are ratings of 1, 5 and 6, but these are special circumstances and only apply to a minute portion of applicants.)

Table 5 on page 44 of the document shows the shares of LDC and non-ALDC candidates receiving a 2- or better. LDC applicants are marginally better on academics than non-ALDC, but not by much. (I have previously looked closely at the trial exhibits and came to the conclusion that the LDC group was only slightly better on SAT/ACT, but weaker on SAT subject tests and GPA.)

LDC admits, by contrast, are weaker academically than non-ALDC admits, also shown on that table.

LDC preferences are basically affirmative action for the already privileged. I especially liked this part on page 12, which confirms a process that common sense had already suggested to anyone with functional grey matter:

The old “you are an outsider, so you must be wrong” argument?

I’d venture that most admissions officers couldn’t understand the logit regression models (used by both Harvard’s expert and Arcidiacono) if you gave them 10,000 hours. Different talents, different worlds.

Unrelated to the legacy/development/athlete question generally, the latest report contains a link to the average scores and grades of the Harvard applicant pool. This is a topic that frequently comes up on here in the context of how “qualified” the applicant pool is.

Here are the numbers (some are concorded or extracted from indices):

Average SAT score: 1460 (concorded)
Average ACT score: 31
Average SAT subject test score: 720
Average HSGPA: 3.75 (extracted from index value)

One particular poster often goes on about “750” as being the “bar” for serious candidates on the SAT. Looks like significantly more than half the applicant pool does not meet that threshold.

Likely more than half of subject test scores are less than ~710 (it is difficult to estimate exactly the median because we only have mean data).

Harvard itself rates its students on Academics from 1 through 4. The overwhelming majority of admits will have ratings of 1 or 2. Approximately 39% of the applicant pool has these credentials.

See here: https://github.com/tyleransom/SFFAvHarvard-Docs/blob/master/TrialExhibits/D730.pdf

What I see, among some on CC, is a certainty holistic is rigged. For some who back off a bit, so do I. Do adcoms really need to understand regression theory? That’s the test?

As it is, too many don’t try to understand more about a TT college’s wants. Calling it a crapshoot or relying heavily on the lawsuit can cloud things, including the openness to learing more. And, yes, from my perspective, I see the vast percent of kids who seem to wing their apps. They have the stats, founded some club or make money in some venture and think that’s the “it.” Unfortunately, this seems to leave them in the lurch. Not because holistic is rigged, but because they stop short. (I can give examples.)

The 750 comes direct from experience. Sure, some have lower totals (not that it’s totals that matter for a tippy top- it’s the individual scores. Or ACT sub scores.) But do you really want to tell ordinary average excellent, unhooked kids that they just have to hit a 1460? Are you doing them some service? You want to praise them and imply that they’re the whole deal?

Meanwhile, a small percent of those who make it past first cut get anywhere in the subsequent readings. And this is holistic, not hierarchical, not as simlpe as stats, some club titles, hours of service. Qualitative. Not a public vote, but the adcoms for that college deciding. Try to learn more. You can learn from the colleges- with an open mind.

Selectivity or admit rates aren’t that meaningful without these numbers. Wish all colleges would release these numbers voluntarily.

It is a deliberate decision not to release the data. So those without the special hooks ALDC/URM likely do need a 750+ in each test, but the substantial portion of the class which is hooked brings the average down to 730 or so.

Only if they want to publicly dismiss the work of economists who spell out all of their data sources and methodology. And swear to the accuracy of their work through affidavits.

It is not meaningful to dismiss something that is not understood.

I’d love examples. Not because I have kids applying to college (BTDT - I’m done), but you do make alot of vague pronouncements about needing to “know what it takes” and there are never really examples other than “achievement in the context of your high school isn’t enough”.

@dropbox77177 , the linked source in post #12 provides averages for applicants, not admitted or matriculating students. There are a good number of delusional “lottery” applicants. Admitted students averages are higher which you can eyeball by the CDS. When my kids were applying, we focused on trying to hit the midpoint of the 25th and 75th percentiles as their “hurdle” after which the admission decision was going to be driven by something else. They took the ACT’s so their target was a 34.

If you want to dive into admitted student objective stats, see Tables 6 and 7 which segments the subgroups by Academic Index deciles which is a measurement of GPA and test scores (SAT,ACT,SAT2). The discussion here is mostly contained in Section 3.3.3 starting on p 21.

Dang, yes, @Ohiomomof2 . How about the vast number of kids, including top performers, who cannot answer a Why Us? (I’ve mentioned this repeatedly, over the years.) They frequently get as far as “You have my major.” Duh. Or, “You have study abroad.” Well so do lots of colleges. Or kids who play almost their entire app to how much they want some major, but somehow don’t send an LoR from that subject area? Or never took available AP in that? Or look at what kids in the past 6+ weeks want to write for essays. And more. How does anyone expect half effort or sloppy effort to play at Harvard?

I’ve seen at least 5 kidsin the past week, hoping their business ventures are what will make them a shoo-in.

So, I say: read up on that college and try to learn what does matter, what does show that college how you match.

And adcoms make their decisions, based on the whole. Not what some economists observe or posit. Why wouldn’t a bright kid want to focus on what the colleges say and show-- not some economist?

Btw, I like this age group very much. But it’s a lot more than stats or how H numbers its reviews. Same as in real life, whether we’re impressed or not with new acquaintances.

And there simply aren’t enough slots to take every good kid. And geo diversity matters to top schools.

@BKSquared - Yes, of course the source I linked is the for the applicant pool, not the admit pool. Maybe I wasn’t clear enough in my post?

About “thresholds” or “hurdles,” if they exist then I have never seen a convincing explanation why there are such dramatic increases in admissions rates between the supposed threshold (or hurdle) and the upper bound. For instance, look at Harvard’s data here, which show admit rates rising monotonically with scores, literally tripling between SAT 1520 and 1600 (concorded values): http://samv91khoyt2i553a2t1s05i-wpengine.netdna-ssl.com/wp-content/uploads/2018/06/Doc-421-145-Admissions-Part-II-Report.pdf (page 7).

What is nice about that graph is that it superimposes the number of applicants achieving those scores. You can see by inspection that the number of applicants scoring above 1520 decreases monotonically such that there are only one third the number of applicants scoring 1580-1600 as 1520-40.

Usefully, that chart excludes legacies and athletes, leaving bare the true landscape for the unhooked. Many weaker legacies are of course admitted, but the charts tell the tale of the unhooked.

We see the exact phenomenon with academic index, presented in SD terms in the immediately preceding page in that document (page 6). Here the effect is more dramatic, as the components of academic index are not perfectly correlated (although obviously positive). The last half standard deviation is where all the action is!

Now, if there were a true threshold or hurdle, we would expect to see admit rates flatten beyond it. We don’t.

The alternative explanation is that Harvard is looking for many things other than academic ability, which is of course true. But this begs the question why it is finding these other combinations in such abundance at increasingly higher scores (or academic index). Especially as the number of applicants falls off so dramatically at the higher numbers. It could simply be that Harvard values scores and AI among the unhooked more than it says (that is actually probably true).

More likely, scores also capture an underlying causal factor which is correlated with excellence in the other dimensions that it is looking for (personal, ECs, etc). It’s not too hard to figure out what it is :slight_smile:

It is not difficult to figure out Harvard’s model of admissions here.

Because in the end, when they’ve got their finalists, they can cherry pick. And that includes high scores. (And geo diversity, balance in majors, gender, etc.) Doesn’t mean high scores are the primary influence.

So, in other words, @lookingforward, they use scores to cherry pick the unhooked admits when they get to the finalists. And the higher the scores are, the better the applicants fare. Makes sense.

@lookingforward I am not Ohiomomof2. Poor thing, whoever she is, gets tagged by people talking to me pretty often.

OK not being able to do a unique “why us” seems pretty common to me and of course those kids aren’t finding the unique thing(s) that make that college and that kid a good match.

Honestly, the LOR/major thing seems less important to me. A lot of kids don’t know what they will major in. My kid thought it would be Chem and that HS teacher was just clueless and missed her emails requesting a LOR - she sent those in May junior year, too. Good thing she wound up not majoring in Chem, I guess.

You do say this a lot. But I don’t really know what that means, beyond the obvious. What DOES matter to a particular elite college? You can read mission statements and such on web sites and not really come up with more than they want “leadership” or “low income kids who succeed despite obstacles” and such.

More concrete examples from your experience would be helpful.