Skip to content

Does Harvard discriminate against Asian Americans in college admissions?

Sharad Goel, Daniel Ho and I looked into the question, in response to a recent lawsuit. We wrote something for the Boston Review:

What Statistics Can’t Tell Us in the Fight over Affirmative Action at Harvard

Asian Americans and Academics

“Distinguishing Excellences”

Adjusting and Over-Adjusting for Differences

The Evolving Meaning of Merit

Character and Bias

A Path Forward

The Future of Affirmative Action


  1. D Kane says:

    The sins of omission in this article are many. I will split them up into separate comments.

    First, why not mention how secretive elite colleges are with this data? I agree that a hierarchical approach would be best. So, the readers asks, why don’t you get Columbia data and show us the best way to answer this question? The reason — as you know but the reader does not — is that Columbia refuses to share this data, even with you (a member of its faculty), even if you sign agreements to keep the data secret, even if they anonymize the data first. Why do you think that might be?

    The reason, obviously, is that Columbia is terrified about what you would discover and report.

    My first complaint with your article is that you fail to inform the readers about this (important!) fact.

  2. Andrew says:


    I have no idea if Columbia would share its admissions data with me, as I don’t recall having ever asked them for admissions data. So I don’t understand what you’re talking about here.

    • D Kane says:


      > I have no idea if Columbia would share its admissions data with me

      Are you really that naive? No idea? A perfectly diffuse prior?

      I mistakenly assumed that, if you were going to go to the trouble of reading 700 (!) pages of court documents, you would be more informed about the realities of the debate.

      But, good news, this provides a great out-of-sample test for my claim! Send an e-mail to Jessica Marinaccio, Dean of Undergraduate Admissions and Financial Aid at Columbia. Indeed, given your claims (which I agree with!) about the importance of this topic and the usefulness of hierarchical models, this seems like a small request.

      Ask for the data. I guarantee she will turn you down, no matter how qualified you are, no matter how trustworthy you are.

      That refusal should help you to update your priors about what is really going on.

  3. D Kane says:

    The second sin of omission comes at the beginning. You write:

    Justice Lewis F. Powell, Jr., singled out Harvard’s admissions program as an exemplar for achieving diversity and applauded the university’s own description of its policy, according to which the “race of an applicant may tip the balance in his favor just as geographic origin or a life spent on a farm may tip the balance in other candidates’ cases.” As the Supreme Court would later emphasize, such review considered race merely a “factor of a factor of a factor.”

    Your next paragraph should have informed the reader that this is utter nonsense. Your failure to do so is the equivalent of quoting Trump saying “The Moon is made of green cheese,” without then informing your readers that, No, in fact, the moon is not made of green cheese.

    There is an academic literature on this question. Good examples:

    Espenshade, Thomas J., et al. “Admission Preferences for Minority Students, Athletes, and Legacies at Elite Universities.” Social Science Quarterly, vol. 85, no. 5, 2004, pp. 1422–1446.

    Hawkins, Stacy and Arcidiacono, Peter and Espenshade, Thomas and Sander, Richard H., A Conversation on the Nature, Effects, and Future of Affirmative Action in Higher Education Admissions (February 1, 2015). University of Pennsylvania Journal of Constitutional Law, Vol. 17, Issue 3, 2015; UCLA School of Law Research Paper No. 15-25.

    No one involved in this dispute denies that race plays a major role in elite admissions. It is not a “factor of a factor of a factor” and never has been. Round numbers, well over 100 Columbia undergraduates in each class would not have been admitted if they checked a different racial box.

    There is, of course, widespread disagreement about whether or not current policies are a good idea. (Hawkins and Espenshade are my two favorite defenders of the current status quo.) But no one knowledgeable denies the magnitude of the preferences involved. But quoting nonsense about farmers, without telling your readers the facts, you are doing them a disservice.

    • yyw says:

      Fundamentally considering race as a factor is equivalent to enforcing a quota. Depending on how heavy handed a university does it, it is either a soft quota or a hard quota.

    • xys says:

      Even if it is a factor of a factor of a factor of a factor of a factor of a factor of a factor of a factor of a factor of a factor of a factor of a factor of a factor of a factor of a factor of a factor, does it matter when it decreases the admission rate of Asians from 40% to 20%?

      It is dishonest for the authors fail to mention the impact on the admission rate of Asian students.

  4. D Kane says:

    The third sin of omission in this article is a failure to report some of the relevant data. Consider this section:

    Indeed, the university receives far more applications from students with stellar high-school transcripts and SAT scores than they could possibly admit. For example, Harvard admits about 2,000 students each year, but more than 8,000 domestic applicants for the class of 2019 had perfect grade point averages . . .

    This is true, as far as it goes, but your failure to discuss the racial composite of these academic superstars leaves the reader in the dark about why SFFA has such a strong case.

    (I don’t mean to insist that Andrew at al are aware of this data and purposely chose to keep it from their readers. (Although, having gone to the trouble of reading 700 pages of court documents, I would have expected a better grasp of the debate.) But the article is still flawed for failing to mention, for example, that approximately twice as many Asian-Americans as white Americans — in raw numbers, not as a percentage of the population — score above 750 on the SAT.

    SATs are, of course, not the only thing that matters in elite admissions, but those raw numbers should suggest (just to me?) that something is going with Asian-American admissions. The article would have been better to mention facts like this.

    • yyw says:

      It would be interesting to see the percentages of admission of top SAT scorers (say >750) of different racial groups.

      • gregor says:

        It would also be interesting to see if an SAT of 750+ is equally predictive for all racial groups. If the Asian SAT outperformance doesn’t translate into elite performance in the same way as it does for other groups, I can see why a school would not want to load too much on that one dimension.

        My guess is that Asians underperform their scores. I base this on the fact that Asians do not dominate at the grad and elite levels nearly to the extent you’d expect based on their SAT scores. MCAT and GMAT show a small Asian edge. LSAT shows a small edge for Whites and in California (with the highest Asian population and a tough bar exam) Whites pass the bar exam at a higher rate than Asians. In elite mathematics departments, you see lots of Asian grad students but only a few actual professors. And we don’t see tons of Asian Fields Medalists. Based on SAT Math they should be completely dominating elite mathematics, but they don’t.

        • yyw says:

          SAT may or may not be a good predictor, depending on what you try to predict. The graduate level and professional exam results you cited may or may not cast doubt on SAT’s predictability. We can’t know without knowing more about at least the SAT score distribution of the test takers. It would be interesting to see how composition of elite math faculty and grad students evolve over time. My impression is that the high number of Asian grad students is a fairly recent phenomenon fueled partly by international students, who may not be relevant when discussing SAT’s predictability. I certainly agree with you that using SAT scores or anything for that matter to predict Fields medalists is futile.

        • A says:

          The Fields Medal isn’t a great indicator because most recipients are over sixty, while the Asian American college age population bumped up in the mid-90s. Post-graduate education is also an iffy indicator because it includes variables like household wealth. The top half of Asian-American households by income manage more assets, but show lower homeownership, than white peers. However, the bottom half is significantly less wealthy, with much lower home equity exposure, than their white peers. If Asian-American households do better on standardized tests than their income peers, you would expect to see “under-performance” in graduate school enrollment.

          • Ido says:

            Field medalists must be under 40

            • gregor says:

              Correct. He mean all the past winners. Looking at the most recent winners, of the 16 since 2006, there have been

              8 Europeans
              2 East Asians (1 Chinese-American,1 Vietnamese)
              2 Iranians (including the only woman to win)
              2 Indians
              1 Israeli (Ashkenazi)
              1 Brazilian (of French ancestry)

              This isn’t what you’d expect if East Asians truly have >0.5 SD advantage in mathematical ability over Whites as suggested by the SAT-Math.

              • Yuling says:

                hi gregor, I don’t think you need to count fields medals to falsify “East Asians truly have >0.5 SD advantage in mathematical ability” as

                a) Asian Americans only account for 7% of US population, so even a 0.5 SD difference of sample mean in the sat data is not that striking if you consider a multilevel/partial-pooling model.

                b) Why and how should I learn anything from one data point each year? It is just too noisy to reject any conclusion if you really do a hypothesis testing or predictive checking.

                c) The sample proportion of east-asian in your list is 2/16=12.5%, which is indeed larger than their population share in the western world. Sure, it can also be compared with east-asian population in the world, but then you have many other confounders such as the social structure. After all, it is just too noisy to look at one data point each year and draw some conclusions on six billion people.

                d) Fields medal only measures the top elite mathematicians. A distribution with higher mean does not necessarily have a thinker right tail.

                I am not arguing about your conclusion. But you need another more valid proof.

              • anonymous says:

                Gregor’s classification of one winner as a “Chinese-American East Asian” is a bit strange. What is N/S/E/W Asian?

              • Carlos Ungil says:

                Anonymous, it’s not so strange:

  5. May says:

    D Kane, Upton Sinclair said, “It is difficult to get a man to understand something, when his salary depends on his not understanding it.”

    • Andrew says:


      There is nothing in my salary that depends on my not understanding something. It seems to me that you are writing a comment that insults me because the article that I wrote does not express your preferred political position. If you’d like to write an article of your own regarding Harvard’s admissions policies, feel free to do so.

      • D Kane says:

        May: I agree with Andrew that yours is an unfair comment. The wonderfully thing about tenure is that Andrew’s salary depends, in no meaningful way, on his position/research with regard to elite college admissions. The rest of us, alas, are not so fortunate . . .

        Andrew: Surprisingly (?), I thought May was criticizing me, not you! Communication on the internet is hard!

        • Andrew says:


          Yes. Intonation is notoriously difficult to convey in typed speech.

          • May says:

            I will concede that my comment may have been somewhat unfair.

            But I do think as long as Andrew does not make a sincere attempt to obtain and analyze Columbia’s admissions data, then I see it as no different from a tobacco company refusing to conduct studies while making certain claims in public.

            I do not expect Andrew to express any particular political position but I do expect him to advocate for greater transparency in college admissions and earnestly investigate if racial discrimination exists, for the public.

    • Terry says:

      This is really unfair. Nothing bad has every happened to an academic just because they expressed a politically unpopular opinion.

  6. D Kane says:

    A side point is that this debate provides an occasion to revisit Don Rubin’s slogan: “No causation without manipulation.”

    Example: Was John Wang rejected from Harvard because he was Asian?

    Rubin pointed out that this question was (sort of) nonsensical because, in order to explore causation, one needs to have a (potential) experiment in mind, something in which one could manipulate the treatments. In a case like Wang, it is/was unclear what that manipulation would be. Genetic editing in vitro? Plastic surgery? Adoption?

    However, SFFA have a good answer. Their proposed manipulation is: Check a different racial box on the Common Application. Their claim is that the causal effect of John checking the Asian box is that he was rejected. If he had checked the White box, he would have been accepted.

    Just wanted to point that out.

    • yyw says:

      Remove or switch out racial markers from past applications and have the Harvard admission staff review them again. Assign them in a way that no application is assigned to someone that reviewed it in the past. Compare the decision with actual historical decisions.

  7. jrg says:

    Nice essay. I liked the phrase “included variable bias” and was persuaded by the argument that that bickering statisticians would obscure the real issue of what are the right criteria to use for admissions.

    The graph you present, however, does present an empirical issue: it appears that the personal characteristic judgement that disfavors Asian applicants is also the same metric that is favoring African American applicants. You conclude that

    “… if the courts do find that Harvard has improperly discriminated against Asian applicants, one remedy is to simply curb that practice; they need not curtail long-standing affirmative action policies that increase the representation of underrepresented minorities on college campuses.”

    But it looks from your graph that it will be difficult to remove anti-Asian bias without influencing African-American admission decisions. This is not just because it is a zero-sum game, but because the dimension under consideration affects both groups in opposite ways.

    However, it does seem that the personal characteristic judgement that is disfavoring Asian applicants is also favoring African American applicants. It will be a challenge to remedy anti-Asian bias without having a disproportionate effect on African Americans.

  8. Jonathan says:

    Nice article, though it felt tentative; I could feel you tip-toeing around the politics. Don’t blame you.

    My thinking about this has become: isn’t it humorous how much this matters? By this, I mean the concept of Harvard admissions as meaningful in some material way at the level of larger society – societies? The attention paid Harvard must be gratifying: it signals to all and sundry: see how much a Harvard degree matters! And the corollary: we matter so much we can affect the nation and the world just by whom we admit!

    Is that true? The first part is true: people act as though a Harvard degree matters one whole heck of a lot. It matters so much that arguing over a few percentage points is considered ‘cheating’ against one group and for another group.

    What if it’s not true. You realize it isn’t true, right? You realize that a power structure is not determined by the college degree but rather that the name on the degree matters because you are in the power structure, because that’s your class or position. Dr. Seuss covered this in The Sneetches: the star-bellied Sneetches looked down on the plain-bellied Sneetches and when the plain-belly Sneetches got stars on ‘thars’, the star-bellied Sneetches removed their stars and looked down on those who had them. It was only when everyone got so confused they couldn’t tell who came from what group that … and then it becomes a fairy-tale because the Sneetches actually learn but humans don’t. (Thank you, Ted Geisel, for so many lessons.)

    What people are actually arguing over is how this particular degree is valued as a badge of entrance. The idea being promulgated is that a Harvard degree is the golden ticket – though Roald Dahl was an asshole – to visit Willy Wonka, and if you don’t have that then you are lacking and you can never make that up. It’s like you’re no longer a virgin, like you’ve lost The Precious, like Max Bialystock took your blue blankie.

    Harvard was never a golden ticket: it represented the fact that you were born holding a golden ticket. Harvard, like dear old Yale, always let in a variety of the talented but it took its cachet from the fact that it represented the upper class. That upper class associated itself with other varieties of upper class. As Philip Roth wrote, they weren’t interested in the average Jew but in the developer or investment Jew who made money or the doctor Jew or the intellectual Jew or the artist Jew. And as Roth often caught, there really isn’t anything wrong with that as long as you realize the limitations: if you’re upper class in anything, why wouldn’t you want to associate with others like you? Just because you’re upper class doesn’t mean you’re an idiot. Don’t you have the right to prefer talented Jews over non-talented Jews?

    I’ve tried to analyze this issue but there’s frighteningly little data. I’m aware of a few studies that have looked at people who got into one ‘level’ of school and who went to a lower ‘level’ school, and the like. Nothing big. Even the wording of the question makes the answer obvious: insert statistical hedging, then say it’s not surprising that it’s the person not the name on the diploma.

    Then I look at issues like fame. Does Harvard have a disproportionate amount of famous alumni? Sure, but how many of those would not be famous if they went elsewhere? It’s similar to the question of wealth: you have that when you arrive, so sure it would seem there are lots of rich alumni, but then isn’t it also that you have in you ambition and intelligence, so … that’s saying they do a decent job of admitting the ambitious and intelligent. I’m not removing the name on the degree, just saying that people focus on the label so the name matters more in their estimations than it really does in life. In other words, the blinding influence of a prior.

    As a note, I use prior also to refer to prior conviction because admissibility of priors directly leads to convictions; they are strong evidence in a person’s mind, so the absence of priors whether they exist or not creates a very different trial and result. (Or even if it goes to trial because they’re massive leverage in bargaining, so estimating low admissibility shifts the power.) In this case, I use prior to refer not only to known prior belief but to the inadmissibility of the existence of prior belief by not looking at these issues at the proper model level.

    For example, take the idea that Harvard shouldn’t admit legacies. That presumes a bunch of stuff about Harvard’s role in society and thus Harvard’s right to raise money for its self perpetuation versus its social role. Those all rest on the question: how much does the bleeping degree actually matter? Note again, not if the degree matters at all but how much. How many times have you heard the Zuckerberg story versus how many times you’ve heard about the two Seans, Parker and Fanning, the former never going to college and the latter at Northeastern? How long was Bill Gates at Harvard? Maybe it’s not the degree but the actual human. If Harvard can’t act to perpetuate its endowment, then are they a public utility?

    I totally agree that Harvard shouldn’t admit athletes. Like all the other schools in the nation because, you know, college sports aren’t popular and there’s no money in it and the kids don’t apply to your school in part because you have a big block A or W or whatever on your team jerseys and the state legislature isn’t proud of visible accomplishments that show how much better we are. And why should they change? Don’t people enjoy bread and circuses? Doesn’t all work and no play make Jack a dull boy? Nope, apparently life is all about that ticket and not really about you. The old arguments that athletics rounds a person out don’t matter because that kind of rounding is cheating against someone who isn’t athletic. Isn’t letting in the more intelligent cheating against the less intelligent? So now you’re saying the school is actually designed only for this purpose and no other?

    • May says:

      Tell me Jonathan, what do you think is chance of a high school senior in 2019 could become a Supreme Court Justice if she or he becomes unable to attend an Ivy League college or Stanford for undergraduate or graduate studies?

      Currently, it appears access to certain influential positions in American society are gatekept by these institutions for better or worse.

    • Charlene says:

      I’m a graduate student at Harvard. I disagree that it doesn’t make a difference. The access to power and influence, the potential you have here is very different compared to my undergraduate, which actually another Ivy League, Columbia. Powerful, famous people pass through on the regular. Access to prestigious internships and professors at the top of their field. A Harvard degree very much matters in being able to wield power in this society, something that Asian Americans have very little of.

      And no, not everyone here was born with a golden ticket. In fact, I’m a Chinese-American student whose family is in poverty. It’s critical to get those that were not born with golden tickets to get the opportunity to realize their full potential. Your argument that we shouldn’t try to reform their processes simply because

      • Martha (Smith) says:


        It looks like the last part of your comment got chopped off somehow — perhaps you can add a comment completing it.

        Also, to comment on what you have said:
        Speaking as an older woman, I would like to caution you not to put too much emphasis on power itself. Power can be used for good, but it can also be used for ill. There is an old saying, “Power tends to corrupt; absolute power corrupts absolutely” (Lord Acton), and a more recent one, “With great power comes great responsibility” (Spider Man’s Uncle Ben). I have seen a lot of people who sought power end up using it irresponsibility, doing more ill than good. I hope you do not become one of them.

      • someone says:

        Martha (Smith):

        But if the power is spread around, doesn’t that in itself mean less power for one person or group, so it protects against the abuse of power?

        e.g. Supreme Court judges, someone else mentioned here.

        • Martha (Smith) says:

          Spreading power around could result in equal or proportionate power for groups, but it doesn’t necessarily preclude individuals having excessive power. So I’m not convinced that different apportionments of “power” would necessarily protect against abuse of power. Even small amounts of power can be abused. (e.g., a large person typically has greater power to beat up a small person than does another small person. But that doesn’t mean all large people abuse the power they happen to have. Abuse of power is an individual choice.)

          • May says:

            Are you suggesting that Asian-Americans in this country should be denied any power because you believe they 1. are seeking it? 2. May use power irresponsibly? 3. Even if they have small amounts of power, they will abuse it?

            That seems to be what you are saying with your rhetoric.

            • Martha (Smith) says:

              No, I am not at all suggesting that Asian-Americans in this country should be denied any power, nor that they will abuse it. I don’t see how you could rationally deduce that from what I’ve said. My cautions about power apply to anyone, of any race, nationality, gender, etc. I have seen too much abuse of power in my day.

          • someone says:


            The question is: what do those of us at no risk of abusing power (because we have zero power) do to fight abuse of power?

            • May says:

              From what we’ve seen from the Asian-Americans who suspect institutions of learning have placed a racial quota on them, you file lawsuits.

            • Martha (Smith) says:


              It is extremely rare for someone to have zero power. However, there are many people who, individually, have very little power. But when such people band together for a common cause, they collectively can have a considerable amount of power, through mechanisms such as communicating with those who do have more power (e.g., elected officials or people who have influence on elected officials) who might be sympathetic to the cause or who might be persuaded by convincing arguments or evidence to back the cause; or by demonstrations that are carefully designed to make their case or (as May has suggested) by lawsuits (which can cost a lot of money, so are more feasible to do collectively).

              Changes do not take place overnight, but it is not uncommon for a cause that is worthwhile to make considerable progress through sustained collective action.

              One such change that I have seen in my lifetime is the increased acceptance of women in scientific fields and in elite universities. When I was an undergraduate, the chair of the math department at my university said that he would never hire a woman, and university nepotism restrictions prevented wives of faculty members from being hired by the university. Also, I was the only woman in some of my math classes (which had also been the case my last year in high school). That has changed a lot since then.

              When I applied to graduate school, the Princeton graduate catalogue said that “admission is normally limited to adult males”, so it seemed useless to apply there. Yale didn’t have such a statement, but was all male as undergraduate, and addressed the application form that I requested to “Mr. Martin K. Smith,” so I decided not to apply there, and I decided to limit my graduate school applications to places that I knew had admitted women to their graduate programs. Fortunately, both of them accepted me, and I went to the University of Chicago.

              When I got my Ph.D. from Chicago, I got a post doc job at Rice University, which was originally all male but by then admitted some women, and had only recently started admitting black students. I was the only woman on the math faculty, and there were only about six women on the entire faculty then. The Rice faculty is now about 50-50 balanced between men and women.

              I ended up at the University of Texas at Austin. When I got tenure, I was the only tenured woman on the math faculty. That has also changed a lot since then.

  9. David says:

    California outlaws affirmative action for admissions publicly funded universities, so UC Berkeley’s admissions data contributes an important data point to this discussion.

    Fall 2018 admissions:
    Black 3.1%
    Mexican American & Other Hispanic 14.6%
    Native American & Pacific Islander 0.5%

    White 21.3%

    International & Not stated 18.6%

    Total 58.1%

    The remainder are Asian (China, Vietnam, Korea, Japan, …)
    and South Asian: 41.9%

    To me it is unethical to discriminate against someone because they belong to an ethnic group which is academically successful.

    The moment one moves away from objective indicators such as SAT scores, this is what can happen.

    Many of those who are poor within those ethnic groups may not have had opportunities that wealthier people in that ethnic group (eg. command of language or literature, etc).

    • Martha (Smith) says:

      David said, “The moment one moves away from objective indicators such as SAT scores…”

      What are your criteria for something to be an “objective indicator”? And what is it that SAT scores are designed to predict? And how well do they predict that?

      • David says:

        SAT scores (and GREs, MCATs, LSATs, GMATs, …) are some sort of indicator for cognitive ability to an approximation.

        Preparation tests and tutoring paid for by wealthy parents could sway these results, I’m not a test expert.

        The moment a school decides the move away from using SAT scores is the moment that academically successful population, for example, Jews and Asians, can be discriminated against.

        University of Chicago to stop requiring ACT and SAT scores for prospective undergraduates

        This move by U of C makes it much easier to discriminate against those in academically successful groups in the name of “diversity.” It reopens the door for Jewish quotas, a door that was closed 100 years ago at Ivy League schools.

        To be sure and to state this explicitly: when there are quotas on academically successful ethnic groups (say, in the name of “diversity”), those that stand to lose out will not be the wealthy or well-to-do among those groups, but those who are relatively poor.

        • May says:

          David, don’t believe the colleges who claim to be moving away from SAT or have become “test optional”:

          “Colleges from Bowdoin in Maine to Pitzer in California dropped the SAT entrance exam as a requirement, saying it favors the affluent, penalizes minorities, and doesn’t predict academic success. What they don’t advertise is that they find future students by buying names of those who do well on the test.

          Pitzer buys as many as 100,000 names a year based on test scores from the College Board, owner of the SAT, to search for applicants, even after the school became “test-optional’’ in the 2003-2004 year. Wake Forest University, which stopped requiring the SAT or rival ACT test for students entering in 2009, also buys names, as does Bowdoin.”


          The colleges that are conducting pernicious racial discrimination in their admissions process are likely trying to hide the evidence any way they can right now.

      • gregor says:

        He likes SAT scores because that’s what his group does well on.

        • David says:

          As imperfect as these tests are they appear an appropriate way to measure cognitive ability which is some measure of the ability to succeed in these schools.

          Discriminating against ethic groups because of their cognitive ability (as was the case with the Jewish quotas of Harvard and other schools) is a real possibility without some form of standardized, universal measure of cognitive ability. Today one might not say, “anti-Jew or Jewish quota or anti-Asian or Asian quota” but pursing a goal of “diversity” effectively can result in the same outcomes. It is those outcomes which are worrisome.

          • gregor says:

            I think the SAT is generally a decent test, especially if it’s taken without much prep. But the Asian SAT advantage, especially on math, is too large and has risen too quickly to be cognitive ability/IQ.

      • A says:

        That’s certainly a fair question, and it would be healthy to have an introspective discussion about what colleges should achieve. But race quotas are a method of skipping that conversation. At least standardized tests are performance games relevant to academic performance games.

    • gdanning says:

      I think caution is in order when looking at overall admissions data for UC Berkeley, because students do not apply to UC Berkeley. They apply to the UC Berkeley College of Engineering, or the UC Berkeley College of Letters and Sciences, etc. When I was teaching high school in CA, I would sometimes be flummoxed by the fact that student A was accepted at Berkeley, but student B was not, when it seemed to me that student B was a stronger student. Usually, it turned out that student B applied to the College of Engineering (freshman admissions rate c. 10%) while student A applied to the College of L&S (freshman admissions rate c.19%). FWIW, in my experience, Asian-American students are more likely than Hispanic or African-American students to apply to colleges at UC Berkeley which are more selective. PS: Admissions data are here and here:

  10. gregor says:

    Non-Asian minorities (NAMs) are about 30% of a Harvard freshman class. In terms of aptitude they should probably be more like 3%. If that set aside were eliminated completely Blacks would probably fall below 1%. I just don’t see that happening. Sorry Asians.

    Asians have around a 23% share. Asians want to be represented proportionally to their SAT performance which is very good (suspiciously so). Let’s say a 40% share or so. Do these Asians expect Whites to absorb the full cost of the big set aside for NAMs? GTFO

    Quota seems to be a dirty word, but I don’t see why it’s such a big deal to be explicit about what’s going on. Well, I do understand. But people need to get over it. It’s obvious Harvard is using the “personal rating” as a fudge factor to get the numbers to work out the way they want. But it’s easier and more honest just to race-norm the entrance metrics and have quotas. And 23% seems like a perfectly reasonable number for Asians.

    P.S. I think the College Board needs to take a look at the SAT scores. Why have Asian scores been going up so much? Are they cheating? If it’s largely because of the prep industry, take steps to counteract that. Make the format less predictable. Use more g-loaded items that aren’t as susceptible to practice.

    • xysname says:

      Why is 1% Black unacceptable? There is 1% Asian in football, NBA, Emmy, Oscar, Congress, Senator, etc., how are those unacceptable? If you can accept a 0% Asian Supreme Court and 0% Asian NBA, you can accept 40% Asian Harvard. It is not that hard.

    • It’s rare to see such a straight-up example of racism these days. Congratulations, Gregor.

      • gregor says:

        What part specifically were you offended by?

        • May says:

          First, the use of racial quotas in higher education are unconstitutional (Bakke).

          Second, statements such as “X% seems like a perfectly reasonable number for ” are racist and discriminatory.

          • gregor says:

            The Supreme Court approved Michigan’s use of a “points system” that was functionally not any different than a quota. And they’ve said it’s okay to have diversity “goals” and “targets.” And a private university would have even more leeway. Like I said, people prefer to pretend like it’s not a quota. Similarly, they say it’s fine to give “extra” points to Blacks and Hispanics but they probably wouldn’t allow a school to explicitly deduct points from White or Asian applicants even though these are mathematically equivalent.

            The reality is that admissions are racialized because of affirmative action and have been for a long time.

            • yyw says:

              Supreme court decision was a prime example of intellectual contortion to reach a socially desirable (as judged by 5 individuals) outcome. This kind of dishonesty to me is more corrosive than outright racism.

        • (1) Asians’ “SAT performance which is very good (suspiciously so)… Are they cheating?” Wow! Those inscrutable Asians — they’re not like us, and we should suspect they’re doing something underhanded. It’s just like writing, “those African American personality scores are really high. Are they bribing the admissions staff?” It’s repugnant and insulting.

          (2) “And 23% seems like a perfectly reasonable number for Asians.” Thanks for letting me know that there’s a “reasonable” number. Asians aren’t individuals, of course, just some lump of matter of which which one has too few or too many. (You may reply that this isn’t your view, but that of elite Universities, but that’s not what your statement conveys.) What’s the “perfectly reasonable” number of half-Asians, by the way? Jewish kids? How do you come up with these?

          I am writing a bit strongly, but I’m honestly shocked .

          • gregor says:

            Sooooo sanctimonious.

            1) The mean Asian SAT Math score is well over 600. It is perfectly reasonable to ask if these scores (especially those coming from China) reflect those students’ true underlying abilities or if the scores are inflated in some way. And yes cheating is part of that conversation. You have apparently been living under a rock and haven’t seen any of the well-documented Chinese cheating scandals. The only question is if it is widespread enough to seriously skew the statistics.

            2) You can NAIVELY say “treat everyone as an individual” but if those “individual” decisions in aggregate mean almost zero black students at Harvard or 65% men, that isn’t going to gel with current political sensibilities. But there is also a desire to lie and not be too explicit about it (“we don’t use quotas” “holistic criteria” and all that bullshit).

            • A says:

              Seems like you passed the point of “having conversations” and “asking questions” long ago. Now it’s a hunt for confirmation of suspicions.

            • Charlene says:

              Those Chinese cheating scandals you linked to were based in the country of China. The group the lawsuit is representing is Asian-American students. The fact that that you do not distingiuish between Asian American students and Chinese students in China exposes your implicit racism.

              • Martha (Smith) says:

                There are situations where I question attributing something to implicit racism — but this is not one of them. This is either implicit racism or really, really sloppy science, or (my best guess) both (e.g., implicit racism contributing to the sloppy science).

              • May says:

                I agree. To draw a relation between a scandal involving students from China and say, Korean-American students at Lowell High School or Brooklyn Technical High School is quite ridiculous and implicitly racist.

                In addition, the Wall Street Journal reported in 2016, “Public universities in the U.S. recorded 5.1 reports of alleged cheating for every 100 international students, versus one report per 100 domestic students.”


              • gregor says:

                The very first link I posted is titled “How Sophisticated Test Scams From China Are Making Their Way Into the U.S.” Please note: INTO THE US.

                I have read many reports of significant cheating scandals here in the US (not just SAT). For obvious reasons, the media is never explicit about *who* exactly is doing the cheating but some basic inference reveals it to be a disproportionately Asian phenomenon.

                Some Examples

                Mills High School in CA. The College Board canceled all of the AP results for the school because of collaboration. The school is 59% Asian.

                Stuyvesant High School in NY. Many cheating scandals. In 2013, the principal was forced out over a cheating cover-up. 74% Asian. Just a coincidence!

                You are highly offended by the suggestion that there might be some cultural continuity between Asian-Americans and their ancestral cultures. In many cases there isn’t, especially after several generations. But to say that there is none at all is frankly ridiculous. Culture is not completely erased the moment a family comes to America, and the hyper-competitive academic culture is present among a substantial subset of Asian-Americans. That sort of culture can push kids to cheat. More broadly, I suspect it leads to kids with inflated college entrance metrics relative to their raw academic talent and this eventually leads to underperformance at higher levels. Please see Amy Chua’s Tiger Mom book for details about this culture.

              • May says:

                Gregor, your stance “some basic inference reveals it to be” is implicitly racist.

                No fair-minded person would draw a similar inference on which communities are likely to doctor college applications and stories of hardship from the T.M. Landry scandal:


                Yet you are eager to claim there is a “cultural continuity between Asian-Americans and their ancestral cultures” with regards to academic dishonesty.

                Gregor, I’m starting to suspect you are actually well-known anti-Asian racist UC Davis Professor Norm Matloff posting under a pseudonym:


              • gregor says:

                @ May

                Lol, I have no idea who that guy is. This isn’t some hobbyhorse of mine. I’m a guy giving my opinion. And I daresay I think I’m more objective about this than you as evidenced by your non-stop histrionics.

                I did not say that Asians are intrinsically dishonest. I said culturally they have a very strong commitment to schooling. That is a positive trait. But the flip side of that coin is that such pressure may push more kids to cheat.

                I saw that T.M. Landry story right when it came out. It did not change my view. Note that the fraudulent applicants were more the work of the faculty and the school administration rather than the students (although they were complicit). That matches the pattern of other “black” cheating scandals like in Atlanta where it was the *teachers* that did the cheating (to do better on the federal testing).

    • someone says:

      gregor says: “Asians want to be represented proportionally to their SAT performance.”

      I doubt it. They just want to be represented proportionally to some reasonable set of measures that are not obviously racially biased.

  11. peter says:

    Might be a dumb question, but:

    Why dont people just start NOT marking Asian on their application? Not only as a personal strategy, but also as a boycott kind of thing.

    Would it be possible for someone to have their admission rescinded based on “incorrect race declaration”? If Harvard did that, wouldn’t it make completely obvious that it is using quotas?

    Of course I guess if we did achieve a critical mass of people doing this, then Harvard would probably start looking at other factors such as last name or whatever and ignoring self declared ethnicity. Still, it seems like that for some time it would be a good strategy on the personal level (while not many people do it), and then if it does blow up and forces universities to change their admission practices, it would also serve as a clear indicator of the practices they are engaging in.

  12. jrc says:


    “If personal ratings were awarded in racially discriminatory ways, it would be inappropriate to appeal to them to explain disparities in admissions.”

    I also like the “included variable bias”, but I wanted to ask about another statistical issue that could show up here and seems potentially important: selection into application/observation. For example, if we look at African Americans compared to Asians on the personal rating metric, are we seeing favoritism towards African American applicants? Or are we seeing scores for fundamentally different sub-groups of African American and Asian students (say, the top X% of Asian students and top Y% of African American students)?

    My math may be wrong, but back-of-the-envelope I am estimating that something like 4% of Asian students who graduate high school in the US end up applying to Harvard, but less than 1% of African American students. That doesn’t prove selection bias (or relative level if not bias), but it means it is a candidate explanation. So it seemed like a statistical issue worth mentioning, without me having to take a side on what can or cannot be explained by the selection. Maybe someone here can point me to some better numbers/thoughts on this.

    *From the internet, it seems there are about 3.5M graduating high school seniors each year in the US, about 5-7% are Asian and 10-15% African American. 22% of Harvard’s class is Asian, 15% African American and the Harvard admission rates were about 5% (Asian) and 7% (African American). You can change my base numbers a fair bit and still get a really large discrepancy in application rates by race.

  13. Jonathan (another one) says:

    You (or your co-authors, or both) have as your second conclusion: “Second, when legal judgments rest on complex empirical analysis, courts are ill-served by warring expert reports.” I am biased (as you know) but I don’t see how this conclusion follows from the ensuing discussion. Both experts performed analyses on various subsets of the aggregate applicant pools and both defended their own subsetting and criticized the other’s. Your answer to this is that they should have performed a multilevel model on the entire pool. OK, but that has nothing to do with whether the warring experts obfuscated any issues. Actually, this is a case where (largely owing to the publicity) the experts were unusually candid about how their conclusions flowed from their data and why they chose the data they did. To your bias point, it is unremarkable that the choices they made made the points their sides wanted to make — the question is which set of assumptions and subdivisions of data make more sense. And, rightly enough, that’s what the rest of the essay properly focuses on, and with a discussion that it seems to me is greatly informed by the filings of the two experts — this is exactly as it should be.

    The methods here were completely transparent. Both sides had access to the other side’s data and methods and programs. And the issues in the case (statistical, methodological, cultural, sociological) were (to this outside observer) well discussed by these two experts. I think one finished clearly ahead, but that’s neither here nor there. Your proposal to outsource the statistical issues to an expert appointed by the Court is not necessarily bad, but it depends a lot on who is selected and what biases they might have. Suppose an expert is hired and happens to perform exactly the analysis one side or the other happened to perform. Now this analysis, no matter how much it is criticized by the other side, will be automatically raised in status because of its perceived unbiased character, more or less by chance. What is your mechanism to get the Olympian multilevel model? Even an observer who doesn’t care where the chips fall has to make choices, and those choices are never completely uncontroversial. (And I use the adjective Olympian intentionally — those gods were pretty biased,)

    • Jonathan (another one) says:

      I should add that I have been hired by a Court to do statistical analysis and can assure you that the vitriolic response from both sides is exactly the same as if I had been hired by one side or the other. Worse, because you get various complaints from both sides!

      Most people who complain about expert witness bias seriously misunderstand the process. Experts do not, in my experience, shade their opinions and methods to favor the side they are working for. What really happens is that the people who are hired already are using methods and have established track records of analysis which happen to favor one side. So that side hires them. Everyone (including judges) understands that coming in. But the analysis itself (especially including methodological choices) is there in black and white. If the expert can’t defend his choices articulately, he’s a bad expert. Expert testimony is rhetoric.

    • yyw says:

      Maybe we should have special branches of legal practice that require competency in at least understanding complex empirical analysis.

  14. anon says:

    Some admissions officers (not at Harvard) have also made blatantly racist comments about Asian applicants. How can anyone be confident that there is no racism in university admissions when that is going on?

  15. Dan F. says:

    Harvard’s applicant pool has a kind of clustering that has been omitted. Many students apply from a few elite high schools. These can be elite private schools like Andover and Exeter and Milton and other lesser known local places, and they can be elite public schools like Stuyvesant. As a percentage of overall admissions students admitted from such schools are not trivial, I would guess somewhere in range of 10-20 percent of admitted students come from such schools. The characteristics of such students are probably different from those of students not hailing from such schools. This matters for the admission process because Harvard treats their applications differently. It knows the college counselors, it knows the teachers writing recommendation letters, and it has a body of data that allows it to guess fairly well what sort of student profile, from one of these schools, is likely to succeed.

    With a student coming from a high school that normally sends no one to Harvard, and maybe not many to the university at all, the situation is different.

    Harvard treats the application of the 15th best student at elite school X differently than the application of the number 1 student at unknown school Y, and might give preference to the latter even though “objective” factors such as SAT and GPA favor the former. Basic inference suggests that a 750 SAT from a school full of high SAT’s is not as good as a 750 SAT from a school that sends few students to the university.

    Question: are Asian students more or less likely to hail from “elite” schools (where “elite” is defined as sending 4+ students to Harvard annually)? What are the characteristics of such students? What percentage of applicants and admitted students hail from such schools?

    • Jeff says:

      “Basic inference suggests that a 750 SAT from a school full of high SAT’s is not as good as a 750 SAT from a school that sends few students to the university.”

      I question this. My daughter doesn’t go to an Andover, but her public charter school is regularly at or near the top of our state’s standardized testing averages, and boasts a lot of graduates with high SAT scores. It achieves this feat not through excellence in instruction (charter schools in our state can’t afford to pay as much as plain old public schools, so there are a lot of brand-new teachers who are bad at presenting concepts and lack even basic classroom management skills) but by scaring away those students who won’t test well with graduation requirements that exceed state standards and with “rigor,” which apparently means “anxiety-provoking quantities of homework of varying quality.”

      There are a lot of other things I like about the school, but I think my daughter’s SAT score wouldn’t have suffered had she gone somewhere else.

      • May says:

        “Students for Fair Admissions closely examined the admissions rates of students of one elite New York City public high school, Stuyvesant, where students, admitted based solely on test scores, are mostly Asian American. Asian students from Stuyvesant, the filing said, have a much lower admissions rate than Stuyvesant students of any other race.

        The filing includes an portion of a deposition from Casey Pedrick, the school’s assistant principal. She was shown statistics showing that in 2014, Asian students were half as likely to be admitted to Harvard as whites: 7 of 55 of white students were admitted, compared to 7 of 115 Asian American students.

        Looking at the numbers, the filing said, Pedrick began to cry. “These numbers make it seem like there’s discrimination, and I love these kids and know how hard they work,” she answered, when asked why she was upset. “These look like numbers to all of you guys, but I see their faces.””

        Note: Stuyvesant High School in 2017 was nearly 75% Asian and 18% white.

  16. Justin says:


    It is ironic. I think Judea Pearl, who you went back and forth with on here a few days back, would strongly agree with both your article title and the below conclusion of yours, where you emphasize the inability of statistical modeling by itself to clarify whether the causal role of a variable is as a mediator (“included variable bias”) or confounder:

    “Qualitative grounding of statistical findings is important to avoid misleading conclusions; it is not enough to control for a factor in a statistical analysis without clarifying its role in admissions goals and educational objectives.”

  17. gregor says:

    Consider the following conditional expectations:

    E( HS grades, SAT scores, etc. | “True ability”)

    E( “Success” | HS grades, SAT scores, etc.)

    Assume for sake of argument Asians systematically “try harder” for any given ability level (I think it is true). On average they will then have lower “true ability” for a given set of entrance metrics (i.e. a white with the same entrance metrics will tend to have more true ability). In college (and beyond) true ability becomes more important than in high school and consequently you’d expect underperformance relative to entrance metrics (reversion to true ability). Note that this will be true even if the level of ability is high!

    • Andrew says:


      I’m not sure what is the distinction between true ability and working harder, given that the ability to work hard is an ability too. I guess you could define true ability as the performance that you can give if you work your hardest—but I don’t know that this is clearly defined, in that there’s generally not a maximum level of work.

      • someone says:

        I’ve also noticed some academics use “work very hard” as an insult about their enemies (who apparently have Gregor’s ill-defined “lower true ability”) and a compliment about themselves (it’s good to be hard-working, isn’t it?).

        The same seems to happen to Asians. The same feature is “good” for a general applicant, but “bad” for an Asian applicant.

        • May says:

          Gregor’s distasteful behavior is not surprising. I recall reading this article with the tagline, “If diversity is so important to liberal whites, why do they keep fleeing ethnically diverse suburbia?” –

          “Sociologist Samuel H. Kye, the author of Segregation in Suburbia: Ethnoburbs and Spatial Attainment in the Urban Periphery, examined segregation patterns in 150 middle class metropolitan black, Hispanic, and Asian ethnoburbs from 1990 to 2010. By focusing on middle-class, as opposed to lower-class ethnoburbs, he hoped to eliminate poverty as a factor for white flight. In a phone interview, he relays that the relative economic prosperity of an ethnoburb does not diminish white people’s decisions to abandon it. “Across the board, any time you see a significant presence of minority residents, there is a near perfect predictor of exodus of white residents,” he says.

          While the term “model minority” substantiates a myth about how whites value Asians, Asians are only “model minorities” when they are small in number with minimal influence on a community. When Asians “set the norms of academic achievement by which whites are evaluated [and] ultimately usurp those previously in place,” once heralded Asian achievements are critiqued with suspicion. In a school district near Princeton, New Jersey, last December, parents claimed that the academic tutoring Asian students received outside of school resulted in the ‘elementary school curriculum … being sped up to accommodate them.'”

          Scott Jaschik from Inside Higher Ed writes, “Frank L. Samson, assistant professor of sociology at the University of Miami, thinks his new research findings suggest that the definition of meritocracy used by white people is far more fluid than many would admit, and that this fluidity results in white people favoring certain policies (and groups) over others.

          Specifically, he found, in a survey of white California adults, they generally favor admissions policies that place a high priority on high school grade-point averages and standardized test scores. But when these white people are focused on the success of Asian-American students, their views change.”

          • Andrew says:


            I looked at the first article you linked to, by Anjali Enjeti, and it wasn’t clear to me what is the source of the claim that “liberal whites” are doing this. I see “liberal whites” in the subtitle of the article and the phrase, “white parent’s liberal politics and progressivism,” in the body of the article, and I was puzzled. Conservative whites move too, no?

            I’ve never heard of Johns Creek so to get a sense of how liberal or conservative are the whites who lived there (or, more precisely, who used to live there but have moved away in the past few years), I searched on *”johns creek” georgia elections*. What I’d really like to see is governor’s or presidential vote broken up by city. I assume this information is available somewhere but I could not find it.

            I did find this news article from 7 Nov 2018, “Johns Creek Election Results: Blue Wave Sweeps the City . . . In what has been a traditionally red, with Republican stronghold. Johns Creek has seen a drastic change since Cityhood 12 years ago. This was evident tonight, November 6th, 2018.” But this would seem to argue against the claim that it was “liberal whites” who moved away. If the city used to be Republican and now is leaning Democratic, that suggests it was conservative Republicans who moved away as more Dem-voting nonwhites have moved in.

            I’m also skeptical about the statement about “white parent’s liberal politics.” I’m sure that some parents in Johns Creek are politically liberal. But it’s my general impression that parents are more politically conservative, on average, than non-parents. So, again, on average, I expect that many of the whites who Enjeti is criticizing are actually conservative. This doesn’t affect the main point of the article, which is about ethnic conflict, but it removes some of the puzzle aspect. Consider a rephrasing: “If diversity is not so important to conservative whites, why do they keep moving away from ethnically diverse suburbia?” (I changed “liberal” to “conservatives” and changed “fleeing” to the more neutral “moving away.”)

            • yyw says:

              Humans being humans, stereotyping and biases are always going to be there. However, I don’t blame parents choosing the educational environment that in their mind best fit their children’s interest. There is real difference in what east Asian parents prioritize and what white (conservative or liberal) parents prioritize.

          • gregor says:

            You know what I find distasteful? Your unabashed ethnocentrism. You’ve made many comments and I have not seen a single one that suggests you appreciate the trade-offs involved in this question or are interested in anything other than advocating for your own group’s ethnic interests. You feign concern for universalist values and egalitarianism yet do not seem to care about the perspectives of black or white students or anyone else and your comment here in particular drips with animosity toward whites.

            You don’t like some of the issues raised here and so you try to shout people down with cries of racism or your outrageous implication that our gracious host Andrew’s judgment has been clouded by financial motives. These issues are sensitive and they aren’t things I would normally bring up just for kicks. But once you bring a lawsuit and lob accusations of racial discrimination YOU OPEN YOURSELF UP TO THIS SORT OF SCRUTINY.

      • yyw says:

        +1 to this. True ability is not static in an individual either. Working hard most likely will enhance the true ability even if we don’t count working hard as part of ability. Even if you define true ability as a ceiling, it is likely not static in an individual either. On a lot of learning tasks, our ceilings probably go down as we go past the teenage years.

      • gregor says:

        @ Andrew

        Below is a comment by Steven Hsu on this issue. Maybe you will find his statement of it more understandable than mine.

        “I think it is possible that Asian “strength of application” overpredicts later career success*. There could be many reasons for this. For example, it could be that Asian hard work boosts test taking results and grades more than it does real world achievement.”

        Note that this is not his position, but he acknowledges it as a possibility.

        For me the key questions are

        1) Empirically, does the standard vector of entrance metrics (grades, SAT, AP, etc.) have the same predictive validity for Asians as for other groups? I don’t know that this has been settled definitively, but I think there is some reason to suspect there is underperformance (not to there absolute performance is bad). Some of the comments from the schools imply this.

        2) If there is underperformance, what might account for this?

    • Yuling says:

      An anologe version to your argument:

      A hypothesis testing with higher power are more sensitive to true signals and hence easier to get “significance”. But in reality it is the true signal that matters. Therefore a study with fewer sample size and significant result has more merit.

      No, it does not.

  18. Mikhail says:

    But can you actually claim someone is a racist without drawing DAGs and using do() operators?

Leave a Reply