That claim that Harvard admissions discriminate in favor of Jews? After seeing the statistics, I don’t see it.

A few months ago we discussed Ron Unz’s claim that Jews are massively overrepresented in Ivy League college admissions, not just in comparison to the general population of college-age Americans, but even in comparison to other white kids with comparable academic ability and preparation.

Most of Unz’s article concerns admissions of Asian-Americans, and he also has a proposal to admit certain students at random (see my discussion in the link above). In the present post, I concentrate on the statistics about Jewish students, because this is where I have learned that his statistics are particularly suspect, with various numbers being off by factors of 2 or 4 or more.

Unz’s article was discussed, largely favorably, by academic bloggers Tyler Cowen, Steve Hsu, and . . . me! Hsu writes: “Don’t miss the statistical supplement.” But a lot of our trust in those statistics seems to be misplaced. Some people have sent me some information showing serious problems with Unz’s methods and his numbers.

This post is long because, if we’re adjudicating claims based on statistics, details matter. The short story, though, is that Unz appears to be (a) overestimating the number of Jews at Harvard, and (b) underestimating the proportion of Jews among the set of high-achieving potential Harvard applicants. Put this together and I don’t see the evidence that Jews receive preferential admissions compared to other whites. (Again, Asians is another story, not the topic of the present post.)

I personally have connections both to Harvard and to Jews, so you can make of this what you will. All I can say on that account is that, when Unz’s article came out a few months ago, I had no problem presenting its claims as stated; it was only after receiving some recent emails with detailed statistics that I got the impression that Unz’s numbers were mistaken. What Unz did seems reasonable from a distance (and I can understand why he made the choices he did in making his estimate), but his conclusions don’t seem to hold up on closer inspection.

Unz’s claim

Unz’s argument has two parts, a numerator and a denominator. First, that Ivy League colleges were admitting tons and tons of Jews: 25% at Harvard, Yale, and Columbia and “this same general pattern” in the other five Ivy League schools. Second, Unz writes that the academic credentials of American Jews are not so impressive, with Jews representing “less than 6 percent” of National Merit Scholar semifinalists, a number which Unz presents as “an extreme upper bound to a more neutrally-derived total.”

This seems pretty clear. You have a group that’s 6% of the top achievers, getting 25% of the places in top colleges. A factor of 4, that’s a lot. Sure, Unz’s reasoning can be questioned on the edges: Ivy League schools draw more students from the Northeast, and Unz’s estimates are only approximate. Unz acknowledges some of these issues, writing, “any of the individual figures provided above should be treated with great caution, but the overall pattern of enrollments—statistics compiled over years and decades and across numerous different universities—seems likely to provide an accurate description of reality.” In short, a factor of 4! That would seem pretty solid.

Nope. When you look at the numbers carefully, though, that factor of 4 erodes and erodes until there’s nothing left.

What percentage of Harvard College students are Jewish?

Start with the claim that 25% of Harvard College students are Jewish. That number comes from the Hillel Foundation, the Jewish student organization. I received an email from a Harvard alum who went through the names of Harvard students from the classes of 2009-2012 and estimated the proportion of Jews using the same scale-up methods [see details below] that Unz used to validate his personal estimates of the rate of Jewish names in high-achieving groups (Unz stated here that these scale-up methods produced results within 1% of his own estimates based on direct inspection). Using the scale-up methods, you get an estimate that 10-11% of students at Harvard are Jewish, not 25%. My correspondent suspects that the scale-up estimates are too low and that Hillel’s numbers are too high.

As described by Unz, here are the two Jewish name analyses he uses:

We can perform the same population estimate using distinctly Jewish last names, such as the small set of Cohen, Kaplan, Levy, and “Gold—“ (J1) which were suggested by blogger Steve Sailer and his Jewish correspondent, or else extended to include the full set of such names (J2) utilized by Weyl by adding Berman, Bernstein, Epstein, Friedman, Greenberg, Katz, Levine, Rosenberg, and Stern. Based on the 2000 Census estimates, the first group includes approximately 1 in 20 American Jews, while the larger set raises the fraction to 1 in 12.

I’m a little miffed the list doesn’t include Rosenthal or Gelman, but hey, what can you do?

Numerator incompatible with denominator

As my correspondent writes, “If Unz wants to compare the representation of Jewish students among National Merit Scholar semifinalists to that among Harvard undergrads, he must use the same methodology for both data sets. This substantially nullifies Unz’s arguments about Harvard’s admissions preferences for unqualified Jewish students (particularly in comparison to non-Jewish whites, whose enrollment at Harvard he substantially underestimated).”

OK, so going from 25% to 10%—that’s a factor of 2.5. What about the rest? Two things: geography and counting.

First, the geography. Typically over 40% of Harvard College students come from New England and the mid-Atlantic, a group of states that includes 48% of American Jews but only 21% of the white population.

In comparison, Jewish admissions in competitive west coast colleges such as Stanford and the University of California are much lower, reflecting that they are drawing from a different geographic distribution of applicants. My correspondent writes:

Performing Weyl Analysis on Stanford’s public directory yields the result that 4-5% of Stanford’s undergrads are Jewish (half of the 9.5% Hillel figure cited by Unz), which also happens to coincide with the percentage of Jewish CA NMS semifinalists one finds via Weyl analysis. Note that this figure is below Unz’s estimate (which we shall soon argue is an underestimate) that Jews represent 6% of all NMS semifinalists. This does not suggest that Stanford is discriminating against Jews but rather reflects the fact that Stanford draws many of its students from California, where Jews are a much smaller % of the population than in the Northeast where the Ivies are located. When searching Stanford’s public directory, one can easily see the students’ majors, and it is interesting to note that relatively few of the students with presumably Jewish names are studying engineering, while performing a search for Asian names like Chen or Wang yields many engineering students. This casts doubt on Unz’s implication that the reason there are relatively few Jews at MIT and Caltech is due to their “objective/meritocratic” admissions practices; rather, relatively few Jews seem interested in pursuing engineering.

Second, the counting. The person who emailed me had gone to the trouble of replicating Unz’s calculations, that is, going through the names of National Merit Scholar semifinalists, counting the number of usually Jewish surnames from the above two specified lists, and then scaling these up using the procedure described by Unz to obtain an estimate of total % Jews. For almost every state, these replicated estimates are higher than the numbers reported by Unz. Here are some states:
Pennsylvania: replication estimates 14-21% Jewish; Unz reported 9% Jewish
Massachusetts: replication estimates 9-14%; Unz did not count
Maryland: replication estimates 12-15%; Unz reported 11%
Virginia: replication estimates 7-9%; Unz reported 6%
Ohio: replication estimates 6-7%; Unz reported 4%
Illinois: replication estimates 10%; Unz reported 8%
Florida: replication estimates 9%; Unz reported 8%
Michigan: replication estimates 4%; Unz reported 2%
New York: replication estimates 24%; Unz reported 21%.
This all suggests that Unz’s estimate of 6% Jewish National Merit Scholar semifinalists is too low, even using the Weyl method that Unz described (and, which by Unz’s report, gave results within 0.1 percentage point of what he got from direct inspection of the names). While the Weyl method may produce different results on a state-by-state basis from Unz’s “direct inspection” method, my correspondent found that the Weyl method produced higher Jewish totals than reported by Unz in these states in almost every state checked.

Recall that the Weyl method, when applied to the list of Harvard names, estimates Harvard’s student population as 10-11%, not 25%, Jewish. The point here is not that Unz was trying to get the wrong answer but rather that he was using different data sources and estimation methods in different places, and this leads to systematic errors when the data are combined.

Unz’s estimate is also low because he was missing data from some of the most populous states such as Massachusetts that account for a disproportionate percentage of the U.S. Jewish population. And that’s all before adjusting to account for Ivy students being disproportionately from the Northeast as well as PSAT cutoff scores required for NMS status being generally higher in that region of the U.S.

In summary: Unz claimed that Harvard and other Ivies massively over-admit Jews based on their academic accomplishments (in comparison to non-Jewish whites). But if you are careful with the statistics and compare comparable numbers, the differences go away.

That “strange collapse of Jewish academic achievement”? It’s not such a collapse nor is it so strange.

Here’s another comparison, this based on Harvard’s Phi Beta Kappa awardees.

Unz: “By the late 2000s and early 2010s, Jewish students had become one of the academically weakest groups at Harvard, constituting 25% or more of all students, but just 11-13% of PBKs selections. Meanwhile, during the 2010s the average Asian student was nearly 300% more likely to make PBK, with their proportion of Junior Year PBKs running even higher. And white Gentiles seemed to perform best of all, being about 400% more likely to gain PBK honors than their Jewish classmates.”

My correspondent writes: “Unz obtains these figures by classifying all non-obviously Asian and non-obviously Jewish high academic achievers as non-Jewish white, even though there are many Jews with non-obviously Jewish names (not to mention biracial students with an Asian mother and non-Asian people of color). In addition, I checked his stats for junior PBKs for the classes of 2010-13, and he inflated the Asian numbers by 7 percentage points. Perhaps he would claim that he checked the photos of all 96 students and somehow confirmed that some of the students with “white” names are actually Asian (or half-Asian), but had he done so, he would have noticed that at least 3 of the students are non-Asian students of color, whom he classified as non-Jewish white. Unz completely disregards the existence of black and Hispanic high academic achievers, thus obtaining inflated figures for the academic performance of white Gentiles.”

Unz also looks at very high-performing math students to document his claim of “the strange collapse of Jewish academic achievement.” If you look at the numbers carefully, though, it’s not such a collapse nor is it so strange. I learned this via some numbers compiled by Janet Mertz, a professor of oncology at the University of Wisconsin – Madison who has published a relevant article in the Notices of the American Mathematical Society on mathematics performance by gender and ethnicity on national and international mathematics competitions. She also happens to be the mother of one of the Jewish students with an Anglicized surname whom Unz failed to count among the Jewish 21st century US IMO team members and Putnam Fellows.

Unz: “The U.S. Math Olympiad began in 1974, and all the names of the top scoring students are easily available on the Internet. During the 1970s, well over 40 percent of the total were Jewish, and during the 1980s and 1990s, the fraction averaged about one-third. However, during the thirteen years since 2000, just two names out of 78 or 2.5 percent appear to be Jewish.” [full list is here]

Mertz: “For the 2000s (i.e., 2000-2009) row of his table, Unz claims there were only 3% Jews. Based upon my direct personal knowledge of these students, I have determined that there actually were at most 35% non-Jewish whites, 49% Asians, and at least 16% Jews, with 1/2 Jews being counted 50-50 between Jewish and non-Jewish white and 1/4 Jews being counted as 25-75 between Jewish and non-Jewish white. The 44% Jews Unz claims for the 1970s teams is probably a significant over-estimate. The “Weyl method” (whose numbers need to be multiplied by 12 to estimate total % Jews) yields only 1 obviously Jewish name out of 48 for that era while it yields 2 out of 78 for the 21st century. Unfortunately, my ethnicity data from the 1970s is too incomplete for me to give a firm maximum % Jews.”

OK, let’s say that again: Unz says the rate in the 21st century has been 2.5%, the actual number is over 12% (when taking the whole period 2000-2012). A factor of 5 makes a difference!

Enter the New York Times

In the New York Times, David Brooks wrote the following, in a column celebrating Unz’s piece as one of the best magazine articles of the year: “You’re going to want to argue with Unz’s article all the way along, especially for its narrow, math-test-driven view of merit. But it’s potentially ground-shifting. Unz’s other big point is that Jews are vastly overrepresented at elite universities and that Jewish achievement has collapsed. In the 1970s, for example, 40 percent of top scorers in the Math Olympiad had Jewish names. Now 2.5 percent do.”

Nope. Mertz contacted Brooks and the New York Times about this (not to be all narrow and math-testy, but 2.5% != 12+%), but the Times has not (yet) run a correction. I had a similar experience after pointing out a statistical error in a different NYT column by a different Brooks, but it was no go on that one too, so I could relate to Mertz’s story.

OK, back to Jewish academic performance.

The rate of Jews in the U.S. Olympiad teams has declined over these decades, maybe by a factor of 2 or 3 rather than Unz’s claimed factor of 17. What explains the factor of 2-3?

Mertz writes, “the recent modest drop off in the % Jews can be fully accounted for by their now having to compete with the recent influx to the US of high-achieving Asians for the fixed number of 6 slots per year available on the US IMO team, combined with Jews, especially non-ultra-Orthodox ones, having become a smaller percentage of the US population over the past few decades.”

In his article, Unz wrote: “today’s overwhelmingly affluent Jewish students may be far less diligent in their work habits or driven in their studies than were their parents or grandparents.” But if we accept that Asian-Americans are a high-achieving group in math, and we realize that they now greatly outnumber American Jews, it’s really no mystery that the proportion of Jews on the U.S. Olympiad team has dropped by a factor of 2 or 3. “Increased competition for a fixed number of slots,” together with demographic changes, would seem to be a sufficient explanation in and of itself.

There’s a similar problem with Unz’s analysis of winners of the Putnam math competition, of which he writes, “Over 40 percent of the Putnam winners prior to 1950 were Jewish, and during every decade from the 1950s through the 1990s, between 22 percent and 31 percent of the winners seem to have come from that same ethnic background. But since 2000, the percentage has dropped to under 10 percent, without a single likely Jewish name in the last seven years.”

But Mertz finds something quite different:

For the 2000s (i.e., 2000-2009), at least 34% of the students were foreigners who had only come to the US to matriculate to college here after having participated in the IMO as a member of their own country’s IMO team, i.e., they were born and educated prior to college outside of the US and, thus, competed for admission to elite US colleges in the foreign applicant pool, not the US one. This fact is very easy to confirm by simply searching the official IMO web site for what country’s team each student represented. Again, I say “at least” by assuming any student who did not participate in the IMO is a US citizen/resident. By subtracting out these foreign students, I then obtain at most 33% non-Jewish whites, at most 17% US Asian-Americans, and at least 15% Jews. My own son was among the Jewish Putnam Fellows in the past 7 years that Unz failed to count because he does not have an obviously Jewish surname.

Likewise, for the 2010s (i.e., 2010 and 2011 which is a total of only 10 students, at least 40% of whom were foreign), I calculate at most 40% non-Jewish whites, 10% US Asian-Americans, and at least 10% Jews. Again, the collapse Unz claims in Jewish achievement is simply not present in these data sets when Jews are properly counted by directly asking them whether they have any Jewish parents or grandparents (and counting people with partial Jewish ancestry fractionally in the total).

Mertz summarizes:

By far the biggest problem with the Unz article is his failure to consistently use the same methodology for counting Jews (or US Asian-Americans). I believe the primary reason he sees disparities is due to changing methods for counting, comparing data from methods which under-count against data from methods which over-count. If he maintained a consistent method, it wouldn’t matter much if his method were somewhat off (except in the cases of the very tiny data sets such as the Putnam and Olympiad data or when he compares the 1970s against the 21st century because of changing demographics). So, for example, the problem with his claim that Jews are being over-admitted to Ivy League colleges is that he uses the Hillel data sets (which are likely over-counts) against his Jewish names method (which likely under-counts). If he used his Jewish names method, rather than the Hillel data, to determine % Jews attending Harvard, Yale, etc., he might well find the disparity disappears. Likewise, he includes foreign whites and Asians and others who do not yet have green cards in his Putnam and Harvard Phi Beta Kappa data even though these students were in a separate applicant pool for admissions than were the U.S. students. The U.S. Asians who do not yet have green cards are also mixed in with the Asian-Americans in his National Merit Scholar semifinalist data. In addition, he assumes that all non-Jewish, non-Asian students are non-Jewish US whites, ignoring the fact that some of these students are, in fact, African-American, Hispanic, or, in the case of the Phi Beta Kappa and Putnam data, foreign.

Not a slam, not a fisk

My point here is not to slam or “fisk” Unz. Rather, I’m giving all these details because, as a statistician, I think details are important. One reason I suspect that Unz’s article was originally received so uncritically by many bloggers and journalists is that Unz presented lots of numbers and described where he got them. There were no other easily accessible numbers on the topic, so we were inclined to go with what Unz had. So I think it’s important to pick on the details here, so that it’s clear that these are not minor technical criticisms but rather get to the center of Unz’s argument. I read his article as claiming that Jewish students are overrepresented in Harvard’s admissions by roughly a factor of 4 (compared to other whites) compared to their academic achievements and abilities. But when you make the comparisons carefully, this disparity goes away. There still is arguably an underrepresentation of high-achieving Asian students, but some of Unz’s comparisons there are off too, in that he at times is lumping U.S. and foreign Asians into a single category, whereas it would seem more appropriate to focus on Asian-Americans when comparing to other college applicants.

Summary

In his article, Unz claims to have found that elite college admissions underrepresent Asian-Americans (in comparison to their academic talent achievements) and overrepresent Jews, leaving non-Jewish whites squeezed out. Looking at the statistics more carefully, we see no evidence that Jews are admitted preferentially compared to other whites. Unz’s error arose because he used different sorts of information with different biases that did not cancel out but actually reinforced each other, underestimating the proportion of high-achieving Jews and overestimating the Jewish presence among Ivy League students.

I have long argued that meritocracy can’t work (for more recent discussion, see here and here) and so I’m sympathetic to Unz’s general concerns. But it looks like he garbled the analysis for one of his main points.

The statistical message: your conclusion is only as good as the numbers that go into it.

P.S. Unz replies here. I should prepare a longer explanation, but, just briefly, it does not address the questions raised in this blog post, most obviously that he is using different and incompatible sources for his numerator and denominator and that one of his dramatic numbers is off by at least a factor of 5.

194 thoughts on “That claim that Harvard admissions discriminate in favor of Jews? After seeing the statistics, I don’t see it.

  1. Pingback: Andrew Gelman revises Ron Unz

  2. > But if you are careful with the statistics and compare comparable numbers, the differences go away.

    Likely applies to most published observational studies?

  3. Out of curiosity, I’ve checked the list of 1970s IMO participants. Out of 36 participants (excluding duplicates), I see 8 “probable Jewish” (22%) (Arenstorf, Bernoff, Bloch, Cohen, Kleiman, Pleszkoch, Shor, Reiter), 5 “German/possible Jewish” (14%) (Knierim, Weiss, Zeitz, Tschantz, Herdeg), and 2 (6%) “Polish/possible Jewish” (Modzelewski, Kaminsky).

    • From the 1970s US IMO teams list, I know that Arenstorf is Christian, Shor is only 1/4th Jewish (he told me so himself just last month!), Tschantz is Christian, Herdeg is Christian, and Modzelewski is Christian. One shouldn’t assume that anyone who MIGHT be Jewish is 100% Jewish when they could just as easily be German or Polish Christians or only partially Jewish. One also needs to include duplicates and to count students who participated twice double, making 48 total participants. Yes, I also know that some of the above folks are Jewish (e.g., Zeitz), but also missing from this list are Jews such as Lander. My point is that the direct inspection of likely Jewish names methodology that Unz used in his article is too inaccurate and subjective to yield meaningful percentages, especially on small data sets such as the US IMO team members and Putnam Fellows ones. My best guess is that the real % Jews on the 1970s US IMO teams is probably somewhere in the range of 25% – 30%, not 44% as claimed by Unz. However, we can’t know the answer for sure without more direct knowledge of these people’s heritage, the methodology I applied to the 21st century lists.

      • I don’t think that “Jewish” in this discussion refers to religion, but to ethnicity: if Modzelewski is a descendant of Polish Jews, he is counted as Jewish, it does not matter whether he prays in a synagogue or in a Catholic cathedral.

        Double counting multiple participants is unnecessary, all it does is widen the error bars.

        Name inspection isn’t best, but direct inspection of people’s heritage is usually not possible. And you seem to be trying to have it both ways. You’re saying that, in 1970s, the percentage of Jews is lower than Unz claims, because some people with Jewish surnames are non-Jewish. In 2000s, the percentage is higher because some Jews have non-Jewish surnames. There’s no apparent reason why the ratio of Jewish surnames to numbers of ethnic Jews should be different between the two sets.

        • “There’s no apparent reason why the ratio of Jewish surnames to numbers of ethnic Jews should be different between the two sets.”

          More Jew-Gentile marriages since the 1970s?

        • Start with the set of 2 Jews (male and female) and 2 Gentiles (male and female).

          If they marry within group and have 2 kids per family, in the next generation we’ll have 2 Jews and 2 Gentiles.

          If they marry across groups and have 2 kids per family, and kids take fathers’ names as customary, in the next generation we’ll have 4 half-Jews, 2 people with Jewish surnames, and 2 people with Gentile surnames.

          Counting each half-Jew as 0.5, the ratio of (full) Jews to surnames is the same in both cases.

        • Stephen Modzelewski is ethnically Polish. Unz double-counted multiple participants in his original article since he claimed “during the thirteen years since 2000, just two names out of 78 or 2.5 percent appear to be Jewish.” There are 6 people on each IMO team.

          Your list of probable Jewish surnames does not seem accurate to me. I’d never seen the names Arenstorf or Pleszkoch before, and one of the first hits when googling the name “Pleszkoch” is an obituary for someone buried in a Catholic cemetery. I’d also classify Knierim as unlikely to be Jewish:
          http://www.ancestry.com/name-origin?surname=knierim

        • OK, Arenstorf and Tschantz should be “German/possible Jewish”. And I think you’re right about Knierim and Pleszkoch. Knierim is a German surname but without apparent Jewish links. Mark Pleszkoch is at least half Eastern European, but seems to be Slavic rather than Jewish.

        • Nameless, I can’t reply below your comment, so I guess I’ll reply here. ancestry.com identifies possible Jewish names, and Tschantz is not one of them. For example, the surname of the son of Prof. Mertz is mentioned as a possible Jewish name:
          http://www.ancestry.com/name-origin?surname=kane
          Unz did not count Kane as Jewish, so I don’t see how Tschantz, which is not even identified as a possible Jewish name, could be counted as Jewish by Unz. In fact, Tschantz appears to be a Mennonite name:
          http://www.gameo.org/encyclopedia/contents/S5360ME.html

        • Yes, we are referring to ethnic background, not religious practice.

          Modzelewski is not Jewish. Polish Jews might have their names end in “sky”, but usually not in “ski”.

          Unz double counted to come up with 2.5% when he claimed there were only 2 Jewish names out of 78 participants on the 2000-2012 US IMO teams. I used the same method so his %s and mine could be directly compared.

          I am attempting to obtain as accurate a % Jews from both eras using the exact same method, i.e., directly asking them or people who know them well how many Jewish parents or grandparents they had/have. I simply pointed out here some of the likely errors Unz may have made in Jewish identification from the 1970s in both directions, e.g., missing Jewish Eric Lander as well as assuming incorrectly many of the German and Poland Christians were probably Jewish, to document the high inaccuracy in the approach Unz used to estimate % Jews. I used the same method with the 2000s names, eliminating a non-Jew whose surname was on the Weyl list as well as adding in Jews whose names were not. Some of these 21st century Jews and non-Jews told me this information in confidence with the request that I not reveal their ethnicity; I am respecting their request.

          More and more American Askenazic Jews have been Anglicizing their names over the decades since they immigrated to the US. There has also been a recent influx of Jews from Israel (e.g., Nir) who don’t have Askenazic Jewish-sounding names. Given Judaism is inherited from the mother, it is also quite possible that the children and grand-children of inter-marriages are more likely to identify as or remember their Jewish heritage if their Jewish lineage is maternal rather than paternal.

        • “-ski” is typically Polish. “-sky” is typically Russian. (There are only about ~10 “-sky”‘s in http://en.wikipedia.org/wiki/List_of_Polish_people out of hundreds of names, most with links to Russia.)

          Yes, European immigrants tend to anglicize their names over time. Americans with names like Arenstorf or Modzelewski are likely to have parents or grandparents who immigrated into the U.S. from Germany or Poland some time in the last century. Since the increase in the number of Jews in the U.S. between ~1880 and 1940 was greater than the total number of Americans who were born in Germany, Poland, and Soviet Union, combined, who were present in the country in 1940, it’s reasonable to assume that a significant fraction of Americans with overtly German or Polish names are Jews or part-Jews. One could try to make this more concrete by tracking down numbers of Jewish vs. non-Jewish immigrants from Germany, or figuring out how many Americans of German (non-Jewish) ancestry have non-anglicized names.

          It’s all an approximate technique anyway, but it’s the best you can do if you can’t actually call each person or his/her relatives and ask them about their ancestry.

        • Nameless:

          Before getting too lost in the details, don’t forget the original issue here which are the flaws used by Unz to claim that (a) Harvard preferentially admits Jews, and that (b) there has been a “strange collapse of Jewish academic achievement.” The first claim came from the use of a ratio where numerator and denominator were calculated using different sources, and the second claim was based on a highly-inaccurate counting method that was off by a factor of at least 5.

          I agree with you that the measurement of ethnicity—even the definition of ethnicity—is a challenge. In fact, I think I’ll share this example with my statistics class tomorrow. But we should be careful not to use these difficulties as an excuse to make unsupported claims.

          For me, this has been a fascinating example because I first thought that the ratios Unz reported were so large that they had to be true, even though each of the numbers was subject to error. But no. If you take a series of large errors and they happen to all go in the same direction, that can be enough to create an appearance of a large effect even if there was nothing to start with. It’s an important cautionary tale.

        • Nameless:
          There are 47 million Americans of German descent and 9.4 million Americans of Polish descent:
          http://factfinder2.census.gov/bkmk/table/1.0/en/ACS/11_1YR/S0201/0100000US/popgroup~535
          http://factfinder2.census.gov/bkmk/table/1.0/en/ACS/11_1YR/S0201/0100000US/popgroup~551
          According to Unz’s Appendix A, there are 6.6 million American Jews. As German- and Polish-Americans vastly outnumber American Jews, I don’t agree that “it’s reasonable to assume a significant fraction of Americans with overtly German or Polish names are Jews or part-Jews,” and clearly this assumption was wrong for some of the 1970s US IMO names. Furthermore, 2 of the post-2000 US IMO participants have names on Unz’s “distinctive Jewish names” list, so I assume Unz included them and them only in his [under]estimate that only 2 out of 78 US IMO team members since 2000 have Jewish names. Please note that there is a 2x IMO participant in this group with an obviously Israeli/Hebrew name. How is it reasonable for Unz to count as Jewish 1970s US IMO names that ancestry.com classifies as German (with Jewish not listed as a possibility) but not 2000s US IMO names that are Israeli/Hebrew?

          Since Unz underestimated the percentage of Jews among US IMO participants from 2000-2012 by a factor of 5, saying this is the “best you can do if you can’t actually call each person” is not good enough, especially when this questionable methodology leads to potentially dangerous conclusions.

  4. Whenever anyone asks me about the Unz(which is all the time, I promise, really, I swear, all the time), I have three responses. You covered the first one, which is the disparities in chosen field of study. There are, however, two more:

    a) Ultra-Orthodox Jews are increasing their proportionate share of American Jews.
    b) Inter-marriage has become far more common; the data used in this article is based on surnames.

    So even if you do find a disparity using proper methodology, that methodology doesn’t tell you anything a) the share of ELIGIBLE American Jews (since Ultra-Orthodox Jews are not applying to Ivy League Universities) and b) the share of actual American Jews (since the share of Jews with non-Jewish surnames is increasing over time).

    So, really, I think Unz’s article is all bupkis, as least as it applies to Chosen-Americans. :)

    • N.B. states that Weyl analysis of 4 consecutive recent Harvard College graduating classes yields 10-11% Jews, yet the real % Jews is likely in the 15-20% range. Others I know who are quite familiar with Harvard have independently mentioned to me % Jews in this latter range. These claims suggest that Weyl analysis under-counts % Jews at Harvard by 1 ½ to 2 fold. In agreement with Squarely Rooted’s comments above, I would like to state clearly the likely reason for this under-count. The US Jewish population consists of 2 very distinct groups of Jews with very different attributes: (i) the ultra-Orthodox and (ii) other. The former group strongly believes in preserving their distinct Jewish identity. Thus, most of them strictly observe their religion, send their children to religious schools where a significant part of the day is devoted to the study of their religion, culture, history and Hebrew, retain their Jewish-sounding names, marry within their faith, and have numerous children. The latter group tends to prefer assimilating, at least partially, into American society. Thus, few of them send their children to Jewish schools. Many have Anglicized their names. Their inter-marriage rate is at least 50% per generation. Their procreation rate, on average, has been below replacement levels in recent decades, accounting, in part, for their decreasing percentage of the US population. Most of the Jewish academic over-achievers in the US who strive to become NMS semi-finalists as a step toward getting admitted to an Ivy League or other elite college are members of this latter group. The multiply of 12 (or 20) used in the Weyl method to estimate total % Jews in a list from % Jews in the list having obviously Jewish names such as Cohen, Stein, or Gold was determined by comparing the % Jews with this small list of Jewish names against the census estimate of the total % Jews in the US population. However, in the 21st century, the percentage of Jews in group I who have retained one of these obviously Jewish names is probably much larger than the percentage of Jews in group ii who have done so. Thus, the Weyl method probably significantly over-estimates % Jews in group I and significantly under-estimates % Jews in group ii, while being fairly accurate overall. But, most of the Jews attending Ivy League colleges (and achieving NMS semi-finalist status) come from group ii. Given Unz claims that the % Jews in the NMS semi-finalist lists he analyzed agree within 1% with the % Jews he estimates by the Weyl method, I conclude that his direct inspection method, too, also significantly under-estimates % Jews in group ii, including among the very highest academic achievers in the US such as members of Olympiad teams and Intel STS finalists. It would be nice if someone could figure out what multiplier is the correct one to enable accurate use of the Weyl method with group ii Jews.

    • Not of National Merit Society semifinalists, Phi Beta Kappa awardees, Putnam Fellows, or Math Olympiad participants. Colleges probably ask undergraduates about religion, but they’re not going to release that data, and it wouldn’t do a lot of good if they did since no one’s surveying the comparison groups of interest, and it is likely that surveying produces different results than last name analysis. (And if it doesn’t, there’s not much point to it.)

      • Too bad none of the Ivy league are vulnerable to a Freedom of Information request. I see no harm in them releasing aggregate data other than some vague competitive advantage.

        I see a point to it: if it matches name analysis our confidence in estimates is higher. If not we get to think and decide which method we trust better.

    • For prominent people, there are a number of sources:

      Wikipedia usually discusses ethnic ancestry. I once compared several famous people’s write-ups in Wikipedia versus my 1971 edition of the Encyclopedia Britannica: Wikipedia was much less reticent about ethnicity.

      Wedding announcements in the New York Times can also be extremely informative.

      For Jews, there are numerous websites maintained by Jews that obsess over how Jewish various celebrities are.

    • Genetic analysis using single nucleotide polymorphisms can tell you if someone has significant Ashkenazi Jewish ancestry every single time: it’ll tell you if you’re 1/8th. You can go a lot farther than that if you want to: after Daniel MacArthur, a prominent young Australian geneticist, made his SNP profile public, amateurs determined that he’s a full 1.5% south Asian, evidently stemming from an ancestor of his who picked an Indian or part-Indian wife back in the 1840s, when he was serving in the Bengal Lancers. I kid you not.

  5. I attended Harvard many years ago. One of my classmates regularly got mailings from Hillel—slightly miss-addressed. His name was Name-F Name-L. (I don’t want to use his real name in this context. I’m paranoid.)

    The Hillel mail was always addressed Name-L Name-F. Now, the name sequence Name-L Name-F meets the test of sounding like a Jewish name—at least enough to for Hillel to so classify him.

    However, according to some weird website, Name-L is also one of the 25 most common Japanese family names. An, indeed, my classmate was of Japanese descent on both sides of his family.

    We always found this error by Hillel to be amusing. But, it probably did distort their statistics.

    Chuck (not my real name)

  6. Another observation.

    The post above states:
    Massachusetts: replication estimates 9-14%; Unz did not count

    As I recall, not having looked at such numbers in many years, significantly more students from Massachusetts attend Harvard than from any other state. I seem to recall that NY is a close second. At the same time Jew-starved Idaho and Montana supply few Harvard students. (The class of 1968 may be an exception—as I recall the year that class was admitted the Director of Admissions (an old Idaho boy) retired and, the joke was, any application postmarked from Idaho was accepted—I think there were two or three from Idaho in the class of 1968.

    • I would say that Massachusetts is the most over-represented state at Harvard College, but in terms of absolute numbers, I would guess that more students come from New York than from any other states. Typically over 40% of Harvard College students hail from New England and the mid-Atlantic:
      http://www.news.harvard.edu/gazette/2005/04.07/03-admission.html
      Thus, if Unz wants to compare the % of Jewish NMS semifinalists to the % of Jews at Harvard, he needs to correct for the geographic representation of Harvard students and more heavily weight New England and the mid-Atlantic. New York is the most populous state in the region where almost half of Harvard’s American undergrads live, and it also happens to have the highest % of Jewish NMS semifinalists of any other state (24% according to Weyl Analysis), so this significantly impacts the expected representation of Jewish students at Harvard College (and other Ivies). Furthermore, MA historically has the highest NMS qualification score threshold (with CT and the mid-Atlantic states following close behind), so high-ability students in the states with high Jewish representation in the NMS competition are under-represented among the total national NMS numbers. Thus, I expect the nationwide list of NMS semifinalists to understate the performance of Jewish students on the PSAT and overstate the performance of non-Jewish whites due to the fact that that the NMS semifinalists are the top ~0.5% of each state, and the states with high Jewish populations typically also have high NMS qualifying score thresholds, while states like ID and MT with negligible Jewish populations have low NMS qualifying score thresholds:
      http://voices.washingtonpost.com/answer-sheet/college-admissions/if-you-live-in-west.html
      (For the same reason, I expect the nationwide list of NMS semifinalists to understate the performance of Asian-American students on the PSAT compared to non-Jewish whites.)

  7. Details are very important but in my experience most people dont care about them.

    Unz got the headline, you’ll be lucky to get a footnote.

  8. Regarding Jewish-sounding names, Masha Gessen in her book about Grigory Perelman,”Perfect Rigor,” pointed out that we should pity poor Alexander Spivak. Anti-Semitism was rife in the Soviet Union and this fifteen year-old math whiz “was suddenly bumped off the list” for the 1982 International Mathematical Olympiad “in favor of an ethnic Ukrainian.” Spivak was an ethnic Russian and not Jewish but according to Gessen, his “Jewish-sounding last name” was enough to do him in.
    Gessen also wrote, “[T]he surnames of prize takers and honorable-mention recipients [for the IMO]included Alterman, Levin, Perelman and Tsemekhman. This was worse than just four Jewish boys; this was four obviously Jewish boys.”
    Regarding non-Jewish-sounding names, David Brooks shares the patronym with the quintessentially Jewish Mel, nee Kaminsky.
    Then there is the famous self-depricating Jewish joke about a Jew asking a Judge to change his name from Sullivan to O’Donnell so that in case someone asks him what his name was before, he could reply…

  9. re the question,
    is there a better way to establish jewishness then last name ?
    sure, just ask them the following question:

    Is your passport current ?

    (for those of you who aren’t jewish, it is so you can flee the country when they come for you….)

  10. Really appreciate the post. I had suspected on first reading that the Hillel numbers were over estimates. Also wonder if the Asian numbers are underestimates (of a smaller magnitude), if they perhaps don’t account for those who choose not to list race on their college app. (Non-listing applicants from anecdotal evidence seem more likely to be Asian–which I think is a justifiable thing to do)

  11. According to the official statistics published by Harvard College’s admissions office, the class of 2014 when they entered as freshman was 22% Asian-Americans (which may not have included the students who didn’t indicate their race/ethnicity on their applications. This number is significantly higher than the 16-17% “quota” Unz claimed the Ivy League colleges, including Harvard, seemed to be using based upon the data he presented in his article. Unz likely was under-counting the % Asian-American undergraduates attending Harvard, in part, by relying upon data published by others that included Harvard Extension School students, a group that is not highly selected. Again, one really needs to use the exact same method for comparisons between data sets for conclusions to be valid. That way, any errors in the method cancel out between the numerator and denominator. Given Unz determined % Asian-Americans on the NMS and other lists by family surnames, he should have used that same method to determine % Asian-Americans among Harvard College undergraduate students.

    • The 22% sudden increase was due to the Jian Li case which affected both Harvard and Princeton.Ron is correct in his narrow band assertion of 16-17% quota against Asians in the Ivies. .We are talking about a long period of time that the band was in continued existence. Jewish applicants inthe Ivies are eligiblefor preferences in the Ivies, while Asians are not.

  12. A few years ago, the Jewish Telegraph Agency’s philanthropy columnist (“The Fundermentalist”) took a quick look at the Forbes 400 by ethnicity. It’s hard to think of a topic more interesting, right?

    The JTA reporter made a number of glaring errors (e.g., George Lucas isn’t Jewish), but his list has since then been much debugged by blogger n/a. There still remains work to be done, but it’s fairly easy to look up information on billionaires, such as which charities they favor or who presided at their children’s weddings. I summarize n/a’s results here:

    http://isteve.blogspot.com/2012/07/forbes-400-by-ethnicity.html

    • Steve:

      These sorts of data are indeed interesting, and I think there is still space for Unz to step back, recognize the mistakes he made in his calculations regarding high school students and college admissions (recall, these were errors that happened to reinforce either other rather than cancel out), thank Janet Mertz and my other correspondent for their efforts in clarifying these problems, and safely retreat to the position that Jews are represented among the rich and in the news media establishment in numbers far beyond their proportions in the general population.

      Again, I think it’s unfortunate for Unz to view this as some sort of debate. It’s fine to debate policy but silly to debate the statistics here. I understand that debates get attention, and I also understand that by holding fast to all his claims, Unz will keep some of his supporters on his side, but it really seems unfortunate to me. Statistical analysis can be tricky, especially when you do it on your own without the benefit of peer review, and it would be no dishonor of Unz to take advantage of the free criticism he has received to move to a clearer view of the situation.

  13. Pingback: Unz on Meritocracy: Response to Prof. Gelman on Jewish Elite Overrepresentation | Ron Unz – Writings and Perspectives

    • Ron:

      My correspondents and I were indeed aware of your post of 1 Feb; I didn’t mention it in my blog because it didn’t seem to provide any relevant useful information. Also, I took a look at your new post but it does not address the questions raised in this blog post, most obviously that you are using different and incompatible sources for your numerator and denominator and that one of your dramatic numbers is off by at least a factor of 5.

      But I think you need to take a step back. I recognize that you are comfortable in debates, but this is not really a debate at all. You made some mistakes and as a result were off in many of your numbers. Your errors reinforced each other rather than canceling out.

      I respect that you have sincere goals, but sincerity does not guarantee statistical correctness. I recommend you step back, take a deep breath, acknowledge the mistakes you’ve made, and thank my correspondents for the efforts they have made to check your numbers and provide you with more accurate information.

      Sure, it’s dramatic to claim that Harvard overadmits Jews by a factor of 4. That’s a claim that got you into the New York Times? But wouldn’t it be even more satisfying to go toward the truth?

      I expect you will read this entire comment as patronizing, and I suppose there’s no way to avoid that, but as a statistician, let me assure you that it’s best to go with the numbers. Your experience—in which the initial numbers lead you to a surprising, interesting, and seemingly important result, but then revision leads to the apparent effect disappearing—is not uncommon. I understand your disappointment, but, moving forward, I think you’ll do better to stop and reassess rather than thinking of it as a debate for you to win or lose.

      • Prof. Gelman:

        I’d only come across your blog when you referenced my original article. With regard to your current 3,500 word critique, my operating assumption had been that you were a highly respectable scholar, who became a bit careless and was momentarily hoodwicked by a couple of agitated ideologues like Mertz and Baytch. But now I wonder…

        I had claimed that across the combined NMS lists, the Jewish estimates produced by the sampling technique of Weyl Analysis almost exactly matched those produced by direct inspection, thereby validating the latter approach. You devoted a major section of your column to debunking this claim by pointing out that Weyl Analysis actually produced a substantially higher Jewish estimate than my direct inspection for the 8 states you listed. However, you neglected to note that Weyl Analysis also produced a substantially *lower* estimate for the other 17 states I used. This is exactly what one would expect of any sampling technique, and is fully consistent with my claim that the overall averages converged across the very large sample of 25 states. Your blogsite does describe you as an award-winning Ivy League statistics professor, does it not?

        America’s national elites in academics, finance, media, and politics are today drawn overwhelmingly from Harvard (which rejects some 95% of all applicants) and the rest of the Ivy League. These universities publicly claim that they admit applicants less on objective academic merit than on broad “holistic” factors, which are known only to them. This policy is partly to ensure that their student body is fully “diverse” and reasonably reflects America’s overall population.

        According to Hillel, whose estimates are accepted everywhere, Ivy League undergraduates are 23% Jewish, implying that they are some 3,000% more likely to be enrolled than non-Jewish whites of a similar age. You challenge the Hillel figures, suggesting that they are probably incorrect. However, Jews constitute roughly 1.8% of the national college-age population, and unless the true enrollment figure were that low, Jews would be considerably overrepresented. Given the extremely large gap between 23% and 1.8%, I tend to doubt Hillel’s numbers are off by nearly a factor of 13.

        The least troubling possible explanation for the 3,000% overrepresentation of Jewish students is that Jewish academic performance is so enormously greater than that of white Gentiles, they are admitted by the Ivies at correspondingly greater rates, even though the Ivies publicly discount academic performance as an overriding factor in admissions. The best means of testing this “Jewish out-performance” hypothesis is to estimate the number of Jewish students ranked as NMS semifinalists. But unless an unreasonably large fraction of top-performing Jewish students actually have completely non-Jewish names, this approach fails. I would suggest that the burden of proof is upon those who argue that Jewish students are actually 3,000% more likely to be high-performing than their non-Jewish classmates.

        Let us consider the following thought-experiment. The number of college-age Mormons in America is roughly the same as the number of college-age Jews. Suppose an astonishing fraction of all top Ivy League officials were either Mormons or married to Mormons, while a leading Mormon campus organization reported that young Mormons were 3,000% more likely to be enrolled in the Ivy League than young non-Mormons. To avoid dark suspicions, one would surely attempt to locate some solid evidence that Mormon students were 3,000% more likely to be top-achievers than non-Mormons, or perhaps were 3,000% more likely to apply to the Ivy League.

        • Unz, I found your article very interesting and am hoping it spurs further research. But I’d like to concur with Gelman that we should approach this dispute as a more technical matter of statistics rather than a debate. We can be confident with a 1.8% population figure that they are overrepresented at college, but the degree to which they are overrepresented is a matter of interest and so a lot is resting on those Hillel figures. As someone who has been very critical of the mainstream in media or academia, you should know that just because something is widely accepted doesn’t mean it is accurate! I’m hoping that we get more Weyl analyses of these schools.

        • @Unz

          One contradiction in your post was the initial complaint that a Weyl analysis needs a much larger sample and then later you say that analyzing raw Ivy League graduation data would be Herculean because it is too large.

          Isn’t there a Goldilocks point somewhere between those?

        • Even if Hillel is right, it makes sense to use the same procedure on the Harvard names as you do on the national merit scholars. This would correct for the extent that the name procedure differs from Hillel’s estimates.

          There is apparently no standard procedure for Hillel’s statistics:

          http://www.jweekly.com/article/full/66619/the-top-60-schools-jews-choose-annual-list-includes-lots-of-statistics-but-/

          The trouble is, there is no standard system that Hillel: The Foundation for Jewish Campus Life uses to count Jews. Some data comes from students who self-identify, other comes from admissions offices’ extrapolation or “guesstimate” of those who identify plus those who don’t. Other campuses rely on historical data, and, most importantly, there isn’t a uniform definition of “who is a Jew.” A students’ definition of Jewishness might be different at Yeshiva U. and U.C. Santa Cruz.

        • Ron, your argument that hillel’s numbers are acceptable since they were good enough to provoke a media backlash against Princeton makes sense as a critique of the media/politicians, but its a pretty weak defense of a statistical methodology.

          Being hypocritical doesnt make your detractors wrong.

        • You claim: “However, you neglected to note that Weyl Analysis also produced a substantially *lower* estimate for the other 17 states I used. ” This is false. Weyl Analysis produces lower estimates for the % of Jewish NMS semifinalists in 9 states: Alabama, Colorado, Kansas, Louisiana, Minnesota, Texas (by a very small margin), Washington, New Mexico, and Wisconsin. You underestimated (or omitted) the Jewish % of NMS semifinalists in 16 states: New York, Massachusetts, Pennsylvania, California (Weyl analysis gives 4-5%; you claimed 4%), Maryland, Virginia, Ohio, Illinois, Florida, Michigan, Arizona, Indiana (based on 2010 and 2013; his 2012 link was broken), Iowa, Missouri, Oklahoma, and Tennessee.

          You left out Massachusetts, arguably the most disproportionately represented state at Harvard College, and underestimated New York, the most populous state in the region where almost half of Harvard’s American undergrads live and arguably the state that supplies the largest number of Harvard College students. You also underestimated Pennsylvania, another well-represented state at Harvard and the Ivies, by a wide margin.

          You underestimated (or omitted) the Jewish % of NMS semifinalists in most of the states that are highly represented at Harvard and ALL of the highest NMS qualifying score states. Thus, you substantially underestimated the % of high-ability Jewish students in the Harvard applicant pool.

        • sorry if this was unclear: Weyl Analysis produces higher estimates for % Jewish NMS semifinalists than Unz reported in 16 states and lower estimates than Unz reported in 9 states.

        • I just randomly googled “Texas National Merit semifinalists,” and I came across an interesting finding. The first link is to the 2012-13 list of TX NMS semifinalists, whereas Unz only linked to TX’s 2010 list, which I don’t even see in the first page of Google results. It just so happens that the 2012-13 list has 3 times as many of the distinctive Jewish surnames used in Weyl analysis as the 2010 list (there are 2 Goldsteins, 1 Goldman, 1 Levine, 1 Rosenberg, and 1 Stern). Thus, averaging the two lists gives the result that 3-4% of TX NMS semifinalists are Jewish (Unz reported TX NMS semifinalists as 3% Jewish).

          We are now left with the result that Unz overestimated the % of Jewish NMS semifinalists (compared to the results from Weyl Analysis) in only 8 states, all of which have fewer NMS semifinalists than MA (which Unz described as a “rather small” state) and are poorly represented at Harvard. Also, most are states with low NMS qualifying scores, meaning these states have a lower % of high-ability students than states like MA, NY, etc that Unz omitted or underestimated. These 8 states are Alabama, Colorado, Kansas, Louisiana, Minnesota, New Mexico, Washington, and Wisconsin. They contribute a total of 208 + 265 + 159 + 190 + 318 + 99 + 344 + 324 = 1,907 NMS semifinalists out of 12,163 total NMS semifinalists that Unz examined, i.e. Unz overestimated the % of Jewish NMS semifinalists compared to the Weyl Analysis results in 8 “rather small” states representing 16% of the total NMS semifinalist names examined, whereas Unz underestimated the % of Jewish NMS semifinalists in twice as many states, many of which are larger than the overestimated states, have high NMS qualifying scores, and are highly represented at Harvard. So I think it’s abundantly clear that Unz’s claim that Jews represent 6% of NMS semifinalists is indeed an underestimate and a significant underestimate of the % of high-ability Jewish students in the Harvard applicant pool (due to the fact that almost half of Harvard’s American undergrads are from New England and the mid-Atlantic – states with high NMS qualifying scores and a high % of Jewish semifinalists that Unz systematically underreported). Perhaps when I have more time I can try calculating the actual figure that Weyl Analysis yields.

          btw, here is the link to the 2012-13 TX list:
          http://www.dallasnews.com/news/education/headlines/20120912-national-merit-scholarship-semifinalists-announced-for-2012-13-school-year.ece
          If you get a message saying you need to be a subscriber, try googling “Texas National Merit semifinalists,” and clicking on the link from there. It’s the first link.

        • Ron:

          I don’t know what you mean by “But now I wonder…” As I wrote in my blog post, I had no problem linking to your original article, but of course I respond to evidence, in this case the correspondents who pointed out problems in your analysis, most notably the different numbers in your numerator and denominator.

          Regarding the nine states listed above: these are states that include many Ivy League students. I doubt your 6% number, given the results from Massachusetts, New York, Pennsylvania, etc.

          Regarding your last point, nobody has shared with me the data you discuss on Mormons. My impression is that Mormons mostly live far away from Ivy League schools and are less likely to apply to them and that Mormons are not represented in the same proportion as Jews in the various groups that you looked at in your article. But, again, I haven’t looked at these numbers.

          Recall that you were the one to bring up the Math Olympiad program and make the claim that Jewish students represented only 2.5% of the participants in recent years. Janet Mertz happened to have the actual numbers on this, and it turns out you were too low by a factor of 5. That’s fine. We all make mistakes. I recommend that, instead of labeling Mertz as “agitated” or an “ideologue,” you appreciate that she had data relevant to your question and revise your understanding accordingly. Of all of us in this conversation, Mertz is the only one, to my knowledge, who has published peer-reviewed work on this topic, and I don’t see that anyone has seriously questioned her numbers. So I think she deserves our respect.

          As I noted in my comment to Steve Sailer above, I think there is still space for you to step back, recognize the mistakes you made in his calculations regarding high school students and college admissions (recall, these were errors that happened to reinforce either other rather than cancel out), thank Janet Mertz and my other correspondent for their efforts in clarifying these problems, and safely retreat to the position that Jews are represented among the rich and in the news media establishment in numbers far beyond their proportions in the general population.

          I agree that even if Jews represent only 10-11% of Harvard undergraduates (as would be suggested by the Weyl analysis that you recommend in settings where you do not have data from the Hillel organization), that this is still much more than the 2% of Jews in the general population (or even than the somewhat more than 2% of Jews in the Northeast). You can go with that without needing to rely on dubious claims (such as your claims about Math Olympiad participants, which contradict’s Mertz’s direct data, or the Hillel numbers which, as noted above, are contradicted by the last-name analysis that you otherwise favor).

          You can also make your argument that Harvard etc. are insufficiently diverse, that they have too many students from the Northeast and that perhaps they rely too strongly on academic achievement criteria such as measured by the PSAT and Mathematical Olympiad. Again, you can make this social/political argument without having to hold fast to a discredited statistical argument that uses a numerator and denominator that are incommensurate.

          Granted, if you go with the data as they stand, it weakens your argument a bit—instead of being able to claim that Harvard overadmits Jews out of proportion to their academic preparation and achievements, you have to retreat to the claim that Harvard overadmits Jews in proportion to their academic preparation and achievements, but that the criteria being used (things like SAT scores and Math Olympiad performance) are inappropriate. But that’s why we look at data, so we can learn and understand.

          Again, I think it’s unfortunate for you to view this as some sort of debate. It’s fine to debate policy but it’s silly to debate the statistics here. I understand that debates get attention, and I also understand that by holding fast to all your claims, you will keep some of your supporters on your side, but it really seems unfortunate to me. Statistical analysis can be tricky, especially when you do it on your own without the benefit of peer review, and it would be no dishonor for you to take advantage of the free criticism he has received to move to a clearer view of the situation. Even now I think it’s not too late for you to thank Mertz and my other correspondent for the work that they’ve done.

        • I would like to provide more detail so that readers can easily see that Unz’s claims are untrue. Unz states in his recent blog post: “However, if Mertz had provided similar results for the other seventeen states I used, Gelman would have noticed that Weyl Analysis results were smaller—sometimes considerably smaller—than my direct inspection estimates, and these latter states (which Mertz omits) actually include California and Texas whose NMS totals are by far the largest.” Unz’s NMS data and his explanation of Weyl analysis is here:
          http://www.theamericanconservative.com/articles/meritocracy-appendices/#5
          He reports that California had 1,999 total NMS semifinalists, of which 4% were Jewish. Here are the 2 links to CA’s rosters that Unz provided:
          http://tinyurl.com/2foy2b7
          http://tinyurl.com/ajtpanp
          The 2010 roster contains 1 each of the following names: Cohen, Levy, Goldberg, and Kaplan. These are “J1” Weyl Analysis names, and Unz says “the small set of Cohen, Kaplan, Levy, and “Gold—“…includes approximately 1 in 20 American Jews.” So this results in the estimate that there are 4*20= 80 Jews in the 2010 roster. The 2010 roster also includes 1 each of the following names: Berman and Friedman. These are part of the “J2” Weyl Analysis names, and Unz states that the J1 list plus Berman, Bernstein, Epstein, Friedman, Greenberg, Katz, Levine, Rosenberg, and Stern represents 1 in 12 American Jews. So, that gives us the estimate that there are 6*12 = 72 Jews on the 2010 CA roster. The 2012 roster gives us 1 Cohen, 1 Levy, and 2 Golds, plus 1 Bernstein, 1 Friedman, 1 Levine, and 2 Sterns (one of which is a hyphenated name; I full-counted all hyphenated names in both the NMS and Harvard alumni names data to be consistent). So the J1 estimate is 4*20 = 80 Jews, and the J2 estimate gives 9*12 = 108. Thus, J1 gives (80+80)/(2*1,999) = 4%, while J2 gives (72+108)/(2*1,999) = 5%. i.e. Weyl Analysis yields the estimate that 4-5% of CA’s NMS semifinalists are Jewish. Unz reported 4%, so Weyl Analysis did not yield smaller results for CA, as Unz claimed. The only possible source of error I see here is that I used Unz’s reported total of 1,999 CA NMS semifinalists for both years. Still, I think it’s unlikely Weyl Analysis would produce a result under Unz’s reported estimate of 4%, as Unz claimed.

          Please note that Prof. Gelman stated in his original post that “for almost every state, these replicated estimates are higher than the numbers reported by Unz.” No one was hoodwinking anyone here – we knew that Weyl Analysis gave the estimate that 1-2% of TX semifinalists are Jewish vs Unz’s reported estimate of 3%. TX is a large state but also poorly represented at Harvard, so Unz’s underestimates (some of which were substantial) of ALL northeastern states would have a much more significant impact on estimating the % of high-ability Jewish students in the Harvard applicant pool.

          I did not check small states b/c Weyl analysis obviously produces spurious results for small states. However, I did so today to check Unz’s claim that ALL of the states I did not mention produced lower Weyl Analysis estimates than Unz’s reported estimates. It is easy to verify this is untrue. Please note that all of these states (except TX) have fewer NMS semifinalists than Massachusetts, which Unz described as a “rather small state” with a “negligible” impact, even though MA is a highly represented state at Harvard that Unz omitted from his analysis.

          I am not reporting these numbers b/c I think they are credible (obviously they are not) but rather to debunk Unz’s claim that ALL of the other states produced lower Weyl Analysis estimates for the % of Jewish NMS semifinalists than Unz reported.
          Arizona has 2 Cohens and 1 Stern; thus J1 gives (2*20)/342 = 12%, while J2 gives (3*12)/342 = 11%. Thus, Weyl Analysis gives 11-12% Jewish for AZ; Unz claimed 5%.
          http://tinyurl.com/b7e3kou

          Iowa has 1 Kaplan. Thus J1 gives 20/191 = 10%, while J2 gives 12/191 = 6%. Unz reported 4% for Iowa.
          http://tinyurl.com/bctaykq

          Missouri has 1 Gold. Thus J1 gives 20/344 = 6%, while J2 gives 12/344 = 3%. Unz reported
          2% for Missouri.
          http://tinyurl.com/azmdk9h

          Oklahoma has 1 Goldberg. Thus J1 gives 20/187 = 11%, while J2 gives 12/187 = 6%. Unz reported 3% for Oklahoma.
          http://tinyurl.com/a9a6x55

          Obviously, these results are as silly as the 6 “rather small” states that give 0% Jews from Weyl Analysis, but I think it’s important to see how erroneous Unz’s claims are. The fact remains that Unz underreported the % of Jewish NMS semifinalists in FAR more states (and in the most highly represented states at Harvard, which also have high NMS qualifying scores) than he overreported.

    • I am Prof. Gelman’s unnamed correspondent whom Unz took the liberty to name in his misleading and specious rebuttal. In my first reply, I’m going to focus on the Weyl analysis of NMS semifinalists, which presumably Unz knows was my work and not Prof. Mertz’s, based on the emails that were CCed to him. Let me first say that I originally performed Weyl Analysis on the most highly-represented states at Harvard College. Over 40% of Harvard College students hail from New England and the mid-Atlantic:
      http://www.news.harvard.edu/gazette/2005/04.07/03-admission.html
      Unz substantially underestimated the % of Jewish NMS semifinalists in these states on the basis of Weyl analysis. Unz also claimed that the Weyl Analysis results for California were lower than what Unz reported by direct inspection. This is false – Weyl Analysis gives 4-5% for CA, while Unz claimed 4%. (It is possible that I introduced some error by using his figure of 1999 total CA NMS semifinalists for both years, but Unz’s claim still appears to be false.)

      I had not previously checked states that are small (as Weyl analysis gives spurious results for such states) and/or poorly-represented at Harvard, but I just did so, and Unz continues to mislead. Unz claimed that half the time he reported an overestimate for Jewish NMS semifinalists. Again, false. Unz overestimated Alabama, Colorado, Kansas, Louisiana, Minnesota, Texas (by a very small margin), Washington, and Wisconsin (compared to the results from Weyl Analysis). Unz left out Massachusetts, arguably the most disproportionately represented state at Harvard College, and underestimated New York, the most populous state in the region where almost half of Harvard’s American undergrads live and arguably the state that supplies the largest number of Harvard College students. Unz also underestimated Pennsylvania, another well-represented state at Harvard and the Ivies, by a wide margin. Furthermore, Unz underestimated Maryland, Virginia, Ohio, Illinois, Florida, Michigan, Arizona, Indiana (based on 2010 and 2013; his 2012 link was broken), Iowa, Missouri, Oklahoma, and Tennessee.

      Unz underestimated (or omitted) the Jewish % of NMS semifinalists in 16 states, many of which are highly represented at Harvard and have high NMS qualifying scores, and overestimated the Jewish % of NMS semifinalists in 9 states, most of which are relatively small, poorly represented at Harvard, and have low NMS qualifying scores (which includes 6 states with 0% Jews according to Weyl analysis, not 8 like Unz claimed).

      If Unz wants to compare the % of Jewish NMS semifinalists to the % of Jews at Harvard, he needs to correct for the geographic representation of Harvard students and more heavily weight New England (from which Unz included NO states) and the mid-Atlantic. New York is the most populous state in the region where almost half of Harvard’s American undergrads live, and it also happens to have the highest % of Jewish NMS semifinalists of any other state (24% according to Weyl Analysis), so this significantly impacts the expected representation of Jewish students at Harvard College (and other Ivies). Furthermore, MA historically has the highest NMS qualification score threshold (with CT and the mid-Atlantic states following close behind), so high-ability students in the states with high Jewish representation in the NMS competition are under-represented among the total national NMS numbers. Thus, I expect the nationwide list of NMS semifinalists to understate the performance of Jewish students on the PSAT and overstate the performance of non-Jewish whites due to the fact that that the NMS semifinalists are the top ~0.5% of each state, and the states with high Jewish populations typically also have high NMS qualifying score thresholds.
      http://voices.washingtonpost.com/answer-sheet/college-admissions/if-you-live-in-west.html

      Unz also presumably knows that I did read his blog post dated 2/1 since I CCed him on an email where I specifically discussed it. I gave a detailed explanation of the statistical anomalies present in Hillel’s data on the Jewish enrollment at Harvard and Yale (perhaps I will discuss this in a later comment) that suggest Hillel’s figures are unreliable. That others have cited Hillel’s numbers does not mean their figures should serve as the basis of a statistical analysis comparing them to Jewish NMS semifinalists identified on the basis of an entirely different methodology. Once again, Weyl analysis (which Unz claims produced results within 0.1 percentage points of his direct inspection method) yields the result that Harvard College was 10-11% Jewish in Fall 2008, less than half of Hillel’s 25+% figures. That it’s time-consuming to perform Weyl Analysis on the enrollment data of universities is not an excuse to use Hillel’s numbers, especially when using Weyl analysis (the method that Unz claims produces results within 0.1 percentage point of his direct inspection method for Jewish NMS semifinalists) gives Jewish enrollment figures that are less than half of Hillel’s figures for Harvard and half of Hillel’s figures for Stanford (which has a public directory). If Unz can’t invest the time to do a proper statistical analysis, why publish an obviously faulty statistical analysis?

      • correction: Unz overestimated Alabama, Colorado, Kansas, Louisiana, Minnesota, *New Mexico*, Texas (by a very small margin), Washington, and Wisconsin (compared to the results from Weyl Analysis). my count of 9 overestimated states was correct.

        I urge impartial people to please check the NMS lists. The possible sources of error in my estimates are the following:
        1. I used Unz’s reported numbers for the total NMS semifinalists in each state in order to estimate the Jewish % on the basis of Weyl analysis (if I could not find the reported total for a particular state). This means that for some states, I used the same total for multiple years, which might be inaccurate, but should not introduce substantial errors.

        2. I followed the prescribed Weyl Analysis to the letter; for example, I did not include Grinberg or Greenburg (names which appeared in NMS lists) for Greenberg, or Levin for Levine. I also counted ANY name that started with Gold, since Unz did not clearly describe what he meant by Gold-[]. If he meant an abbreviated list of names, this would lower the Jewish NMS % slightly for a few states. However, I also included all names that started with Gold in the Harvard alumni directory in order to arrive at my Weyl Analysis estimates of 10-11% for Harvard’s Jewish enrollment. Thus, this should not affect the comparison between the % of Jewish NMS semifinalists and the % of Jewish Harvard College students; however, it could give slightly lower Jewish NMS results for a few states. I can re-do the calculations if Unz wants to suggest a restrictive list of Gold[] names.

        3. possible arithmetic errors.

        Please note that I do not have my own blog nor the ability to purchase my own personal soapbox, so I am limited in how I can properly respond to Unz’s specious claims.

  14. Pingback: Meritocracy: Response to Prof. Gelman on Jewish Elite Overrepresentation | The American Conservative

    • Hi steve, now that you’ve read gellman’s post and unz’s rebuttal, what are your views? I cant imagine that youre still confident in unz’s analysis. You might be controversial, but sloppy you are not.

      Unz is correct in saying that his detractors are being hypocrites for criticizing hillel’s numbers, but it hardly validates those numbers.

      • How are we hypocrites for criticizing Hillel’s numbers? Here is my critique of Hillel’s numbers:
        While I do not have the means to independently verify Hillel’s figures, the data for Harvard and Yale exhibit statistical anomalies for the years 2007-9, as you can see here:
        http://www.theamericanconservative.com/articles/meritocracy-appendices/#4
        It is virtually impossible for the TOTAL Jewish undergraduate enrollment to vary so significantly from year-to-year, which calls into question Hillel’s methodology. Hillel reports that Harvard College was 25% Jewish in 2006, 30% Jewish in 2007, and 25.5% Jewish in 2008. In fall 2006, Harvard was comprised of the Class of 2007, 2008, 2009, and 2010. Assume that each class was 25% Jewish (the most likely scenario). Then for Harvard to be 30% Jewish in 2007, that means the class of 2011 would have to be 45% Jewish [(0.25 * 3 + x)/4 = .3 => x = .45], which is virtually impossible since the Class of 2011 was 20.3 percent Asian-American, 8.7 percent African-American, 9.2 percent Latino, and 1.2 percent Native American. Then for Harvard College to be 25.5% Jewish in 2008, that would mean the Class of 2012 would have to be 7% Jewish [(0.25*2 + 0.45 + x)/4 = 0.255 => x = 0.07]. i.e. the % of Jews admitted would have to drop from 45% to 7% in one year. Yale’s numbers exhibit even more stark statistical anomalies, jumping from 22.6% in 2007 to 30% in 2008 and plummeting to 23% in 2009. Thus, Hillel’s estimates are unreliable data. And as someone who attended Harvard College circa 2000, they strike me as significant overestimates.

        Also, when I asked the Harvard Hillel how they obtained their estimates of Jewish undergraduate enrollment, they indicated that Harvard used to collect religious preferences cards from freshmen but that this practice ended ~20 years ago. Thus, Hillel’s data was at some time accurate and worthy of citation by scholars and the media. I did not receive a reply to my inquiry as to how Hillel currently estimates Jewish enrollment.

        btw, on Feb 3, I emailed the info above to Unz in response to his Feb 1 post, which he claimed today that I seem not to have bothered reading.

        • “For decades, the Hillel estimates of Jewish enrollments have been accepted as generally accurate by all media outlets, academic scholars, university administrators, and Jewish organizations; in any case, there is no other source of such data across American universities.”

          Unz could not be more wrong.
          This is partially anecdotal, so take it FWIW. I attended two SUNY schools in New York. The Hillel calculations were laughably wrong by a large margin as well. I don’t know how they reached their numbers but it certainly wasn’t done in a methodical/statistically accurate way. Even if they calculated gentile participants in Hillel sponsored interfaith events it wouldn’t reach the numbers. Unz’ claims that its valid because it’s been generally accepted by the media. Based on this line of reasoning any statistical figure that is “generally accepted by the media” is valid.

        • “How are we hypocrites for criticizing Hillel’s numbers?”

          Unz is correct that when it came to using Hillel’s numbers to advocate a more politically correct cause (i.g., fighting religious discrimination at princeton), no one in the media or academia had any problems accepting hillel’s numbers at face value.

          Now, when these same hillel numbers are being used to make a more controversial point, everyone is questioning or even outright rejecting the hillel numbers. Of course none of this actually validates hillels numbers, which appear to be wildly inflated.

          And considering how much analysis Unz did for this article, I find it hard to believe that he never even attempted to look under the hood of those hillel numbers to verify their veracity. With all his connections, he was unable to get a hold of a few recent student lists at Harvard to run through the names? Seems unlikely. My guess is that he did in fact look at a few and after running a cursory review he realized that it likely wasn’t going to give him the results he was looking for, so he stopped looking further and just figured everyone would accept the hillel numbers at face value as has been the case in the past.

        • Joe:

          I think you are making the mistake of lumping all of Unz’s critics together (and perhaps Unz is making this mistake as well). I’d never seen Hillel’s numbers before hearing of this whole discussion, and I have no reason to think that my two correspondents were at any time “accepting Hillel’s numbers at face value.” Whoever was accepting Hillel’s numbers at face value, that wasn’t me, nor do I think was it Janet Mertz or my other correspondent. There’s nothing hypocritical about my correspondent critically analyzing Hillel data which other people were accepting at face value.

        • This is a good example of the accusation of group hypocrisy — a regrettably common practice everywhere but especially in political commentary. Despite its ubiquity, I’ve come to see it as a reliable red flag.

          The way it works for most everyone (including myself) is that we generalize about the people who we perceive as our opponents, our “enemies”. This is particularly true with ideologies and politics and such. That generalization becomes a stereotype which becomes a caricature. And that caricature becomes identical in our minds with every particular instance of that class. So when one member of that class assert X and some other member of that class assert Y, and those two assertions are incompatible and we can imagine some motivation Z which accounts for it, then we can claim hypocrisy … even when this is not true about all, or many, or some, or even any individual members of this class.

          Why this happens is this generalization in conjunction with the perceived particular moral implications of hypocrisy makes it especially attractive and easy to find our opponents guilty of hypocrisy. Hypocrisy is a moral accusation, not merely an intellectual accusation. It implies both dishonesty and a kind of weakness.

          For Unz (and Unz himself makes this charge against his critics in his blog post in response to this one) it’s clear that a) his opponents are liberals and b) liberals have accepted Hillel’s numbers in the past. All critics of his in this particular context must also be liberals (because who else would be critical of him?) and therefore also implicitly have accepted Hillel’s numbers in the past. Therefore, their quibbling is opportunistic and hypocritical. That most of these assumptions are false is of little consequence; the psychological utility of finding a means to disregard his critics too useful.

          Note that this most certainly isn’t a vice exclusive to Unz or conservatives. It’s east to find political blog posts that do, and difficult to find political blog comments that don’t, take the form of “they said X in this instance but Y in that instance, they’re hypocrites” where “they” is always the undifferentiated opposition but X and Y are statements by individually distinct members of the opposition.

          It’s a badly corrosive rhetorical habit in civil discourse but, even more importantly, it’s a badly corrosive habit of mind. Groups of people aren’t hypocrites unless they can be fairly described as being composed of individuals who are each individually hypocritical. Most of these claims aren’t true on that basis and they’re often not true on even the basis of a single individual.

  15. Pingback: Did Jewish Genius Really Decline? | askblog

  16. Professor Gelman: is it possible to publish the data collected by Mertz? We could resolve this whole thing with a script that took in a name and returned out the ethnicity estimator. This wouldn’t be that hard to implement (some basic string processing) and could be run on several text files with names.

    • Yes. Unz says that Weyl only works on large samples, but he didn’t do it for lists of graduates for Ivy Leagues/CalTech because that would simply be too large. So there should be a way to resolve the matter.

    • Unfortunately, I am not at liberty to publish all of the ethnicity data I have collected since some of these US IMO team members gave me this information under the condition that I not reveal it, i.e., I promised them that I would only publish the pooled %s so that their personal information could remain private. It would be unethical for me to violate this promise of confidentiality. I had previously published the names of some of the 21st century fully Jewish and 1/2 Jewish US IMO team members (http://www.ams.org/notices/200810/fea-gallian.pdf). I got involved in this discussion about Unz’s article because my 2008 peer-reviewed publication indicated that Unz’s claim of only 2.5% Jews on the 2000-2012 teams was clearly an under-count. The US IMO team Jews already publicly revealed include the following: Oaz Nir (Israeli Jewish name; confirmed by direct contact; 2x); Daniel Kane (Anglicized surname; my son; 2x); Alison Miller (1/2 Jew; known via direct contact); Oleg Gol’berg (Weyl list; confirmed via direct contact); Brian Lawrence (1/2 Jew via mother; known via direct contact; 2x); and Zachary Abel (photos on www showing participation in MIT Hillel events; confirmed via direct contact). That sums to 7 1/2 out of 78 total participants (counting duplicates the same way Unz did) or 9.6% not including the students whose names I am not free to reveal. Some of these students also went on to become Putnam Fellows during the past 7 years during which time period Unz claimed there likely were no Jews.

      It would be really nice if there were some simple method for identifying Jews so we could easily answer the interesting question raised by Unz. Unfortunately, the disparities between our findings indicate that direct contact is the only reliable method to obtain accurate information, especially for small data sets. On the other hand, as pointed out by Prof. Gelman, one can still obtain reasonably accurate estimates with large data sets IF (and only IF) the %s in the numerator and denominator are determined by the exact same method (e.g., Weyl’s list of 13 probably Jewish names multiplied by 12) so that any systematic error that exists in the method cancels out. It IS possible to obtain complete lists of class names of students attending some of the elite colleges. These individual data sets are sufficiently large, yet still sufficiently manageable, to use the Weyl method over an entire group of 4 consecutive classes, i.e., an entire recent undergraduate student population. That is exactly what Gelman’s correspondent (who is NOT me) did for Harvard. It would be nice if some folks would be willing to do this for a few other colleges as well. Then we could at least get a better feel for how large the disparity really is between % Jews estimated by this method vs. the method the Hillel Foundation uses to come up with its numbers.

  17. Here is an example of Hillel’s falsehoods. Took me about 2 minutes to find (I won’t bother looking at more at this point)

    Boston College
    http://www.hillel.org/HillelApps/JLOC/Campus.aspx?AgencyId=17233

    It stats that there are:
    4642 graduate students
    0 Jewish students

    It’s absurd to believe that there are 0 Jewish graduate students (even for this hands off Catholic institution).
    The law school is similar to Cardozo-Yeshiva’s law school in the sense that religion plays a marginal factor.
    Boston college as a whole has 25 Jewish studies courses, an Israel program, a major in Judaic studies and kosher food.
    I wouldn’t be surprised if 5-10 percent of the graduate students (especially in law school) were Jewish. According to Unz, Hillel
    is a relatively accurate source for Jewish student population.
    Is it possible Unz was trying to write an intentionally provocative/contrarian article in order to increase his audience?

  18. Right on, NYESQ. Unz is making some fantastical, albeit incorrect claims in his self-published, non-peer reviewed articles to get publicity and invitations to debate. If his data had indicated that the Ivy League colleges were NOT discriminating against some ethnic groups (e.g., non-Jewish whites), it would not have been deemed news worthy except among the lawyers involved in the upcoming US Supreme Court case. If he were a serious social scientist, he would carefully consider each of the suggestions that have been made by the numerous commentators and revise his article accordingly as academics are forced to do prior to their manuscripts being accepted for publication in high-quality peer-reviewed journals. If he chooses to personally attack the messengers who make suggestions he doesn’t like, instead, we should respond by ignoring him from now on, denying him the publicity he seems to crave.

  19. I don’t understand why you say Unz’s numerator and denominator are “incompatible” and “incommensurate”. They are two different estimators. No more, no less. Let’s say they were mean and median. Would you still say they are “incompatible” and “incommensurate”?

    Another puzzle is why you (Andrew) write “Regarding your last point, nobody has shared with me the data you discuss on Mormons. My impression is that Mormons mostly live far away from Ivy League schools and are less likely to apply to them and that Mormons are not represented in the same proportion as Jews in the various groups that you looked at in your article. But, again, I haven’t looked at these numbers.” There seem to be two possibilities: 1. You actually don’t understand the point that Unz used a hypothetical example (involving Mormons) to make. That’s hard to believe. 2. You understand his point, but are ignoring it. If so, why?

    • Brian:

      Unz’s numerator comes from his subjective inspection or from the name analysis (which he reports as giving very close estimates). His denominator comes from Hillel. The numerator and denominator come from different data sources and are incompatible in the sense that when the two methods are used to estimate the same quantity (the percent Jewish at Harvard), they give much different results. That is why I call them incompatible. Regarding your question about the mean and median: In this case, everyone is discussing averages and we would only be using the mean, not the median. So that’s not an issue here.

      Regarding Unz’s point about Mormons, it just seemed irrelevant to me. Mormons and Jews live in different places, and I wouldn’t expect Mormons to be going to Ivy League schools in such great numbers. The role of Mormons in American society is a fascinating topic but it didn’t really seem to have much relevance to the discussion we were having here.

      • When two methods give much different results in estimating a quantity how do we judge which one is more compatible? Isn’t all we can say that “one or both of those two methods are flawed”?

      • “Unz’s numerator comes from his subjective inspection or from the name analysis (which he reports as giving very close estimates). His denominator comes from Hillel. The numerator and denominator come from different data sources and are incompatible in the sense that when the two methods are used to estimate the same quantity (the percent Jewish at Harvard), they give much different results.”

        As a general principle, you’re right on this and Unz is wrong. But we’re not talking about general principles here. We’re talking about a specific, empirical question and this particular “incompatibility” is not necessarily as deadly to Unz’s case that Harvard discriminates in favor of Jews as you seem to believe.

        Jews attending Harvard could be much less likely to have typical Jewish names without the same necessarily being true for Jewish NMS semifinalists (or high academic ability Jews in general).

        If the Jews attending Harvard were randomly drawn from high SAT scorers, this would not be possible. But to assume that is the case begs the question of which factors contribute what to Jewish overrepresentation at Harvard.

        One might suppose, for example, Jews attending Harvard disproportionately come from families very high on status striving, who may be more likely to change or have changed their names. If a phenomenon like this exists it probably would have been even stronger in the past, and there will be significant numbers of Jewish alumni admits today. I can’t say how significant, since I’ve never seen that particular statistic released. But Jews were overrepresented at Ivy League colleges relative to their proportion in the population throughout pretty much the entire 20th century, including during the era of “quotas”, and I wouldn’t be terribly surprised if alumni admits today are disproportiately Jewish even in relation to the pool of alumni children who apply. Additionally, some fraction of each class will be admitted due to connections to entertainment or politics, and I’d also expect the Jews among these to have less typically Jewish names on average.

        We’re not limited to speaking in abstractions here. nb says that in the past, Hillel’s numbers were based on religious preference cards. If this is the case, and people like nb are serious about determining the truth, I suggest they apply “Weyl analysis” to each class list (or pooled lists for 4 or 5 year intervals, or however they want to do it) going back through years when Hillel’s numbers are known to be reliable. Observe trends. Compare to Hillel numbers. Also, post class lists publicly to allow others to check your work, which would have the added benefit of opening up the lists to direct inspection. It may, for example, turn out that while direct inspection and Weyl produce similar estimates across NMS semifinalists, this is not the case for Harvard student.

        • Yes, that’s exactly correct, and a point I’d privately made to a few people.

          Each NMS semifinalist list contains roughly 16,000 American whites and Asians, while the figure for each Harvard class is under 1200, representing an enormous factor of 13x difference in raw selectivity, which might surely impact all surname distributions.

          I’d also pointed out in my article that over the last couple of decades the likely Jewish names on the Olympiad and STS lists—which are vastly more selective than NMS—have seemed disproportionately of immigrant (usually East European) background, based on first name and spelling of surname relative to the fraction of all American Jews. It seems quite plausible that such immigrant Jews might have surname distributions very different from that of the general Jewish population, and would be also disproportionately admitted to Harvard. This would also tend to invalidate Weyl Analysis.

          Let’s consider the alternate hypothesis advanced by my critics. They seem to accept the Hillel claims that until about 20 years ago, Jews were 20-25% of Harvard students, but they argue that the figure these days is just 10-11%. Thus, they claim that the number of Jews at Harvard has declined by 50-60% over the last two decades, without Hillel or anyone else noticing. Indeed, even while this huge decline was occurring Harvard Hillel had been claiming that the number of Jews was actually rising.

          I find this alleged scenario a little unlikely absent strong evidence. Just consider the media firestorm at Princeton over a far lesser decline in Jewish numbers.

        • Unz: “Each NMS semifinalist list contains roughly 16,000 American whites and Asians, while the figure for each Harvard class is under 1200, representing an enormous factor of 13x difference in raw selectivity, which might surely impact all surname distributions.”
          Wrong. Each Harvard class is over 1600 students:
          http://news.harvard.edu/gazette/story/2010/05/yielding-strong-results/
          Note further that this article reports that the Harvard College Class of 2014 is 22% Asian-American, whereas you claimed in your original meritocracy article that Harvard has a ceiling of 16-17% for Asian-Americans.

          Please note that I reported the Weyl Analysis figures for 4 Harvard classes combined, representing ~6,600 people.

          Unz: “I’d also pointed out in my article that over the last couple of decades the likely Jewish names on the Olympiad and STS lists—which are vastly more selective than NMS—have seemed disproportionately of immigrant (usually East European) background, based on first name and spelling of surname relative to the fraction of all American Jews.”
          Except the 2000s list of Math Olympiad people contains 2 names on the Weyl Analysis list (the only names you presumably counted as Jewish even though there were multiple names described as possibly Jewish by ancestry.com), while the 1970s list contained only 1. How you managed to obtain the result that 44% of the 70s US IMO team members were Jewish is an absolute mystery to me. Can you please list the names you counted as Jewish?

          I am not aware of any evidence that the Jewish surname distribution among Harvard students somehow differs substantially from that among NMS semifinalists, and I consider this unlikely to be true. Absent such evidence, the most reasonable approach is to use the same methodology (Weyl Analysis) on both data sets. I have already explained the statistical anomalies in Hillel’s Jewish enrollment data for Harvard and Yale here. Thus, I have already presented evidence indicating that Hillel’s data for Harvard and Yale are unreliable. You have yet to propose a plausible explanation (with actual numbers) for the statistical anomalies present in Hillel’s data.

          Unz: “they argue that the figure [for Harvard Jewish enrollment] these days is just 10-11%.”
          At no point in Prof. Gelman’s blog entry or in the comments have we argued that Harvard College is just 10-11% Jewish these days. You are misrepresenting our claims. We have repeatedly stated that Weyl Analysis yields the result that Harvard is 10-11% Jewish and that I believe that represents an underestimate. It also suggests that Weyl Analysis yields an underestimate for Jewish NMS semifinalists. The point is that the only way to obtain a statistically valid result is to use the same methodology on both data sets.

          n/a – I searched Harvard’s alumni directory for each of the Weyl names. As far as I know, the only published Harvard class list is the list of graduates provided at commencement.

        • Unz: “Each NMS semifinalist list contains roughly 16,000 American whites and Asians”
          You seem to be ignoring the existence of outstanding under-represented minorities, yet again, just like you did in your original article. Such students really do exist. For example, one African-American undergraduate student who graduated phi beta kappa from MIT stayed on there for graduate school, winning the prize for best 2011 Ph.D. thesis in computer science, with Harvard offering him a faculty position directly out of graduate school.

          Unz: “I’d also pointed out in my article that over the last couple of decades the likely Jewish names on the Olympiad and STS lists—which are vastly more selective than NMS—have seemed disproportionately of immigrant (usually East European) background”.
          Regarding the IMO students, I can state from direct knowledge of the kids, not just guessing from surnames, that this statement is no more true for Jews than it is for non-Jewish whites and Asians. In my 2008 article (www.ams.org/notices/200810/fea-gallian.pdf), I clearly document that, in recent years, ~1/2 of the top 12 scorers on the USA Math Olympiad exam from which the 6-member IMO teams are then selected are foreign born. While almost all of the Asian US IMO team members in the 21st century are children of recent immigrants, if not immigrants themselves, only 2 of the Jews (Nir, Gol’Berg) are in this category.

          Unz: “Let’s consider the alternate hypothesis advanced by my critics. They seem to accept the Hillel claims that until about 20 years ago, Jews were 20-25% of Harvard students, but they argue that the figure these days is just 10-11%. Thus, they claim that the number of Jews at Harvard has declined by 50-60% over the last two decades.”
          First, nb states that the real % Jews at Harvard is probably somewhere between the 10-11% Weyl-determined number and the 25% Hillel number. Second, Gelman’s post already explained why the % Jews among the very highest achieving US students may actually have declined somewhat during the past 2 decades.

        • nb,

          “I searched Harvard’s alumni directory for each of the Weyl names.”

          A potential issue: “Please note that alumni can choose to show or hide all or part of their records. [. . .] Since alumni can opt to hide records, the online directory is not an official record of who attended Harvard.”

          If you’re using official numbers for total class sizes and any significant number of recent graduates choose to hide their records, you’ll need to scale your estimates accordingly.

          “As far as I know, the only published Harvard class list is the list of graduates provided at commencement.”

          It would seem simple enough to scan and upload commencement programs then. Whoever does this can do so anonymously if they want.

        • Janet, you are correct to point out that Unz is ignoring the existence of black and Hispanic high academic achievers, thus obtaining inflated figures for the academic performance of non-Jewish whites. Unz gives the breakdown of TX NMS semifinalists as follows: 68% non-Jewish white, 28% Asian-American, 3% Jewish, whereas I found that Hispanic and African names (since most African-Americans do not have African names, I surely undercounted them) represented at least 3% of TX’s NMS semifinalists on the 2010 roster.

        • n/a: “A potential issue: “Please note that alumni can choose to show or hide all or part of their records. [. . .] Since alumni can opt to hide records, the online directory is not an official record of who attended Harvard.””
          I would guess that introduces fairly minimal error. Even Natalie Hershlag and Tatyani Ali are listed at alumni.harvard.edu.

          n/a: “It would seem simple enough to scan and upload commencement programs then.”
          I did not attend commencement, so I do not have one.

        • Unz: “I’d also pointed out in my article that over the last couple of decades the likely Jewish names on the Olympiad and STS lists—which are vastly more selective than NMS—have seemed disproportionately of immigrant (usually East European) background, based on first name and spelling of surname relative to the fraction of all American Jews. It seems quite plausible that such immigrant Jews might have surname distributions very different from that of the general Jewish population, and would be also disproportionately admitted to Harvard. This would also tend to invalidate Weyl Analysis.”
          But my analysis of US IMO team members indicates your ability to identify the Jewish members in the 21st century was low by at least 5-fold because you were mis-identifying as non-Jewish whites ALL of the Jews who did not have obviously Askenazic Jewish names, presumably the ones on the Weyl list, whom you appear to have assumed in this above-quoted analysis are recent immigrants because they haven’t yet Anglicized their names or inter-married. By your very definitions of whom you call an immigrant Jew, you are classifying, in most cases incorrectly, the 21st century Jews as either recent immigrants or non-Jews. Yet, I find by direct knowledge of the students that the vast majority of the 21st century Jewish US IMO team members are, in fact, at least 3rd-generation Americans who you simply failed to identify as Jews. Thus, the methodology you used to reach your above-stated conclusion is clearly seriously flawed at least with respect to US IMO team members. Thus, your conclusion is likely false if you used this same methodology to analyze all of the US Olympiad team members and STS finalists. The bottom line: when analyzing these tiny data sets, one really needs to do it by direct knowledge of the students even though this method is quite labor-intensive; guessing their ethnicity and immigration status by their name alone has a VERY high error rate, so high that one can’t use the data to draw valid conclusions.

        • Incidentally, somewhere among all his endless lengthy comments on this thread, I noticed that my energetic critic “NB” denounced me for incompetence or perhaps fraud for failing to have included a 2013 Texas NMS semifinalist list that he himself easily located via Google. He pointed out that his Weyl Analyis of the Jewish names on that list produced much higher results than those for 2010 Texas list which I had included, and darkly hinted at bad faith on my part.

          Naturally, I followed his link, and noted that the news article was released in mid-September, towards the end of my main round of data gathering; if the item had taken a few extra weeks to rise in the Google rankings, I might have missed it. I doubted very much that averaging the 2013 Texas results with those of the 2010 Texas list would significantly alter my national estimates based on some 43 such lists, but I planned to add it as an Addendum to my Quantitative Appendix E, for completeness.

          However, once I began inspecting the 2013 list, I noticed that it contained only 600 names, while Texas lists have over 1300. I also noticed that Houston and all other major cities outside the Dallas-Ft. Worth area were missing. Since this list ran in the Dallas Morning News, which described it as containing “local semifinalists,” it was obviously not a complete Texas list and useless for my purposes. During my extensive research, I had found and discarded dozens of similar non-statewide lists.

          I would be the first to admit that my NMS estimates might easily be off by small amounts here and there. Surname analysis is hardly precise and in examining over 23,000 names, I might also have easily overlooked a “Bernstein” here or a “Kim” there. Furthermore, a sampling technique such as Weyl Analysis is subject to large statistical fluctuations. In my paper I repeatedly emphasized that all all intermediate results should be treated with considerable caution and might contained significant error. However, since the overall anomaly I detected was in the 1,000% range, errors of 10% or 20% in the individual subcomponents seemed unlikely to completely eliminate it.

          Making estimation errors of 5% or 10% are unavoidable in this process. However, in his angry and desperate pursuit of any possible means of undermining the credibility of my research, “NB” apparently failed to notice the difference between a Texas NMS list containing 1300+ names and a “Texas” NMS list containing just 600. Given that he apparently devoted quite a bit of time to performing Weyl Analysis on that latter list and denounced me based on his findings, this would seem a major lapse on his part: the difference between 600 and 1300 is far greater than a 5% error.

          This simply underscores his ideologically-driven incompetence, which had been apparent to me from the first. He also claims to have performed Weyl Analysis upon several Harvard class rosters that are not publicly available, and thereby concluded that the standard figures provided by Hillel and Karabel are wildly incorrect. This is certainly possible, but given his factor-of-two numerical mistaken for Texas, I’d be quite cautious in accepting those claims.

        • Ron, you are indeed correct that I erred in thinking that that list was the full TX NMS semifinalists roster. I made no attempt to count the number of names and simply used your figure for the total TX NMS semifinalists on the 2010 list, so to be clear, I did not arrive at overestimated results. The fact remains that there are 3 times as many distinctive Jewish surnames from the Weyl list on an incomplete roster of the 2012-2013 TX semifinalists as compared to the 2010 TX list.

          So we’re back to your claim in your most recent blog entry:
          “However, if Mertz had provided similar results for the other seventeen states I used, Gelman would have noticed that Weyl Analysis results were smaller—sometimes considerably smaller—than my direct inspection estimates, and these latter states (which Mertz omits) actually include California and Texas whose NMS totals are by far the largest.” I found that the Weyl Analysis results were smaller in a total of 9 states, 8 of which have fewer NMS semifinalists than MA, which you called “rather small.” I went into considerable detail in this comment to disprove your claim that the Weyl Analysis results in CA were smaller than your direct inspection estimate: Weyl Analysis yields 4-5% vs your claim of 4%. I also exhibited 4 other states (which are among the 17 states you claimed above produced smaller Weyl Analysis results) in which Weyl Analysis overestimated the % of Jewish students vs your direct inspection method.

          Prior to Prof. Gelman publishing his blog post, I checked all of the major feeder states to Harvard from New England and the mid-Atlantic, where almost half of Harvard’s American undergrads live, plus all of the states larger than MA. In only one of these states (Texas) did you report an overestimate (by a very small margin) of the % of Jewish NMS semifinalists compared to Weyl Analysis, whereas you happened to underestimate all of the major feeder states to Harvard and omitted MA, arguably the most disproportionately represented state at Harvard. I did not check any of the states smaller than MA b/c I knew that Weyl Analysis would produce spurious results for them. In this set of small states, it turns out that there are a few more in which Weyl Analysis produces underestimates vs overestimates. This has little impact on the nationwide figure (which I still maintain that you underestimated), and I showed here that in 4 of these states, Weyl Analysis overestimated the % of Jewish students (there were a few in which they basically matched). Thus, your claim that Weyl Analysis produced smaller results (vs your direct inspection) in the other seventeen states is simply not true. In fact, you are off by approximately a factor of 2.

          Speaking of factors of 2, you state, “He also claims to have performed Weyl Analysis upon several Harvard class rosters that are not publicly available, and thereby concluded that the standard figures provided by Hillel and Karabel are wildly incorrect.* This is certainly possible, but given his factor-of-two numerical mistaken for Texas, I’d be quite cautious in accepting those claims.” By the same logic, readers should be quite cautious in accepting your claims. You posted a blog entry in which you overestimated by approximately a factor of 2 the number of states in which Weyl Analysis underpredicted the % of Jewish NMS semifinalists vs your direct inspection method (not to mention that you underreported the number of students in each Harvard College class by over 25% here, a significant error for someone purporting to be an expert on Harvard admissions). Let us not forget that you underestimated the % of Jews on the US IMO team since 2000 by a factor of 5. This is an egregious error, especially since that particular “statistic” was apparently considered one of your most notable data points (at least by David Brooks at the NYT). But perhaps you hold comments I make on blogs to a higher standard than your own published articles?

          *For the zillionth time, I have not stated that any figures provided by Karabel are incorrect. In fact, Karabel states that Harvard College was 21% Jewish in 2000. Hillel’s recent estimates average over 20% higher. I would have estimated Harvard College as 15-20% Jewish in 2000, depending on how one defines Jewish (i.e. if Hillel is full-counting half-Jews, I can see how one could arrive at 20%), so Karabel’s figure is actually quite close to my own personal estimate if you count half-Jews as full Jews. I find it odd that you continually try to accuse me of having disparaged Prof. Karabel when you in fact called him obtuse in your article.

          I have stated that Hillel’s RECENT Jewish enrollment figures for Harvard and Yale exhibit statistical anomalies and that Weyl Analysis produces much lower figures for Harvard’s Jewish enrollment than Hillel’s data. Please stop continually distorting my statements.

          Finally, since you took the liberty to name me in your blog, I would like to point out that anyone with a passing familiarity with Jewish names would know that my first name is a common female Israeli/Hebrew name.

    • You have to be VERY careful when mixing different definitions and trying to draw conclusions. If one asked “What proportion of your parish goes to Christmas services?” I might take the total attendance (250) and divide by the number of members on the roll (150) and conclude that we have 167% attendance.

      • Yes, zbicyclist, that could well be one of the reasons the Hillel % Jewish undergraduate students attending Ivy League colleges comes out 4-fold higher than the % Jews among NMS semi-finalists that Unz obtains by direct inspection. If colleges have stopped asking incoming 1st-year students their religious/ethnic affiliation, it is quite possible that some Hillel’s are now, instead, calculating % Jews by simply counting all individual students who have attended ANY event they sponsored during the school year (e.g., Latke-Hamantashen debate, free dinner) by the total number of undergraduate students. One problem with this way of determining % Jews is that many non-Jewish students attend at least an occasional Hillel event, especially given the high rate at which non-ultra-Orthodox Jews date students of other faiths/ethnicities. For example, my atheist Jewish son with an Anglicized surname would occasionally attend a non-religious Hillel event when he was an undergraduate student because his 1/2 Chinese – 1/2 Jewish girl friend, also with an English surname, would ask him to accompany her. Hillel would count them as 2 Jews, Unz would count them as 2 non-Jewish whites, and I would count them as 1 1/2 Jews and a 1/2 Asian-American. On the other hand, the recent Hillel numbers for some of the Ivy League colleges look statistically fishy. Quite possibly, once college administrators stopped handing them the number, they simply guessed rather than actually trying to determine it. That could explain why some Hillels have been reporting the same number, 25%, year after year in recent years. It could be that the Hillel numbers used to be fairly accurate and still are for some colleges, but can no longer be trusted at places such as Harvard and Yale.

  20. Somewhat tangential, but I recommend Karabel’s “The Chosen,” a historical look from a sociology-of-education perspective at admissions processes at HPY and, as a contrast, Columbia. IIRC, his central claim — based on archival records of correspondence among university leadership — is that the early- to mid-1900s shift toward finding “well-rounded” students who were not only high academic achievers but also showed leadership, athleticism, etc was an often explicit effort to reduce the Jewish population at HPY. This effort, in turn, was an organizational response to external pressures, including the demands of WASP alumni.

    As I recall, Karabel doesn’t make any claims that the contemporary fascination with finding “well-rounded” students reflects anti-Semitism, nor is he making any claims about current Jewish under or overrepresentation at HPY. But, it’s a fascinating piece of research into the admissions processes, and into the original rationale behind now taken-for-granted assumptions about what traits are valued in American admissions decisions.

    • Yes, indeed. Karabel’s magisterial book—with 700 pages and 3000 endnotes—was exceptionally thorough and persuasive, and represented the single most heavily cited source in the footnotes to my own Meritocracy article.

      I should mention that the Jewish enrollment data that constitutes the quantitative backbone to Karabel’s analysis was almost entirely based on exactly the same Hillel data that I myself also used but that Gelman and others now suggest may be wildly inaccurate.

      I’ll feel quite sorry for Prof. Karabel if those charges turn out to be correct, forcing him to return all the scholarly awards he received for his magnum opus, whose underlying research absorbed ten full years of his life.

      • As I already told you via email and posted above: when I asked the Harvard Hillel how they obtained their estimates of Jewish undergraduate enrollment, they indicated that Harvard used to collect religious preferences cards from freshmen but that this practice ended ~20 years ago. Thus, Hillel’s data was at some time accurate and worthy of citation by scholars and the media. I did not receive a reply to my inquiry as to how Hillel currently estimates Jewish enrollment.

        You continue to insist it’s valid for you to use Hillel’s numbers even though employing the Weyl Analysis methodology, which you used to validate your direct inspection method for Jewish NMS semifinalists (which produced results within 0.1 percentage point according to you), results in Jewish enrollment figures at Harvard less than half of the Hillel numbers. So you have 2 options: take Hillel at their word, which then calls into question the accuracy of Weyl analysis used to claim 6% of NMS semifinalists are Jewish, or use Weyl Analysis on both data sets.

        • Okay, let me summarize a few things:

          (1) Universities do not release their student rosters, which presumably is prohibited by various Federal privacy laws. I suspect that if I went around bribing college clerks to obtain them, I’d go to prison.

          (2) For 90 years, Hillel has been America’s major Jewish student organization and currently has branches at over 500 campuses. For decades, they have been compiling and releasing Jewish enrollment statistics, which are then also published annually by various Jewish magazines (whence I obtained them). These enrollment statistics are a major factor used to produce lists of the “top colleges for Jews” provided by various Jewish organizations. Each year, over 50,000 Jewish freshmen enter college, and probably a substantial fraction of those students have made their enrollment choices partly based on those rankings. Thus, over the last decade, Hillel’s published figures have determined the flow of many billions of dollars in student tuition. Yet you seem to be claiming that the Hillel data is totally erroneous, and possibly even fraudulent. Those are very, very serious charges to be making.

          (3) Berkeley Sociology Professor Jerome Karabel is one of America’s leading scholars on university admissions issues, and has published the definitive history of Jewish enrollment in the Ivy League, which received an award from the American Sociological Association as well as almost universally glowing reviews everywhere in the media. If you examine my Appendix D, you’ll notice that a large fraction of my Jewish enrollment data is drawn from Karabel, and if you consult the particular page references, you’ll find that Karabel heavily relied upon Hillel. Another major one of my major sources was Dan A. Oren’s Joining the Club: A History of Jews and Yale (1985), and Oren also relies quite heavily upon Hillel data.

          You seem to be suggesting that Karabel never made any effort to establish the accuracy of his Hillel data during the ten years of exhaustive research spent producing his book. Those are also very, very serious charges to be making.

          (4) I should mention that when I cited the Hillel figures in my own article, I pointed out that they were probably estimates, and might be somewhat inaccurate, emphasizing that such possible inaccuracy was one of the major sources of potential error in my own results. Neither Karabel, Oren, nor any of the other academic scholars who have used this same data seems to have been nearly as cautious: they generally provided the figures without comment.

          (5) Suppose I write an article in a small opinion magazine and cite a New York Times article as source for some particular statistics. If you suggest that the New York Times is wrong about those facts, you should obviously take that matter up with the Gray Lady not me. If you are right and both Hillel and Prof. Karabel are totally wrong, the resulting scandal will be enormous and should certainly reach the front pages of all the major newspapers, at which point I will be glad to admit my error in trusting unreliable sources. Good luck with your future efforts in that regard.

          (6) Taken together, the five items above may also directly relate to one of my findings. I believe you’ve claimed somewhere that you’re a Jewish recent graduate of Harvard or something like that. In my article, I made a strong circumstantial case that Harvard seems to have begun admitting large numbers of intellectually undistinguished Jewish students in recent years. Hmm…

        • Unz: “(1) Universities do not release their student rosters”
          As Prof. Gelman reported in his blog entry, Stanford has a public directory. You can search Stanford students here:
          https://stanfordwho.stanford.edu/SWApp/
          Performing Weyl Analysis on Stanford’s public directory yields the result that 4-5% of Stanford’s undergrads are Jewish (half of the 9.5% Hillel figure that you cited.)

          Unz: “Yet you seem to be claiming that the Hillel data is totally erroneous”
          I gave a detailed analysis indicating significant statistical anomalies in Hillel’s data for Harvard and Yale:
          http://statmodeling.stat.columbia.edu/2013/02/that-claim-that-harvard-admissions-discriminate-in-favor-of-jews-after-checking-the-statistics-maybe-not/#comment-139754
          You are welcome to suggest a plausible explanation (with actual numbers) as to how the TOTAL Jewish enrollment at Harvard and Yale could vary so significantly from year to year.

          I also have shown that performing Weyl Analysis on Harvard and Stanford students produces Jewish enrollment estimates that are roughly half as large as Hillel’s. I attended Harvard College circa 2000, and based on personal observation, I would guess that the true figure lies somewhere between the results from Weyl Analysis and Hillel’s figures.

          ***The critical point, which you fail to address, is that you are using two entirely different methodologies to estimate the % of Jewish NMS semifinalists and the % of Jewish Harvard students.*** You stated that your direct inspection method for identifying Jewish NMS semifinalists produces results within 0.1 percentage point of Weyl Analysis. Since these methodologies produce virtually identical results by your own admission, and I cannot replicate your subjective direct inspection method, I shall henceforth state that your NMS results were based on Weyl Analysis. As I said before, you have 2 options:

          1. Suppose Hillel’s data for Jewish enrollment at Harvard is actually correct. Well, then you have this issue to address: performing Weyl Analysis on the names of Harvard College alumni from the Classes of 2009-2012 yields the result that Harvard College was 10-11% Jewish in Fall 2008, at which time Hillel reported that Jews represented 25+% of Harvard undergraduates. Thus, this calls into question the reliability of Weyl Analysis in estimating the % of Jewish NMS semifinalists, as it gave a significant undercount of Harvard students.

          2. Use the same methodology (Weyl Analysis) on both data sets (NMS semifinalists and Harvard undergraduates) in order to obtain a statistically valid result.

          Unz: “You seem to be suggesting that Karabel never made any effort to establish the accuracy of his Hillel data”
          For the third time now: When I asked the Harvard Hillel how they obtained their estimates of Jewish undergraduate enrollment, they indicated that Harvard used to collect religious preferences cards from freshmen but that this practice ended ~20 years ago. Thus, Hillel’s data was at some time accurate and worthy of citation by scholars and the media.

          Unz: “If you are right and both Hillel and Prof. Karabel are totally wrong…”
          Again, I did not say Prof. Karabel was wrong since he was using older Hillel data. See above. I am not asserting that Weyl Analysis produces the correct figures for Harvard’s Jewish enrollment; in fact, Prof. Gelman quoted me in his blog entry as saying that I thought the Jewish enrollment figures obtained via Weyl Analysis seemed too low while Hillel’s figures seemed too high. Obviously, I am not in a position to verify the actual % of Jewish students at Harvard. I am merely asserting that you must use Weyl Analysis consistently to obtain a valid statistical result.

          Unz: “I believe you’ve claimed somewhere that you’re a Jewish recent graduate of Harvard or something like that. In my article, I made a strong circumstantial case that Harvard seems to have begun admitting large numbers of intellectually undistinguished Jewish students in recent years. Hmm…”
          Ooh, burn! …except I graduated summa cum laude with a BA in physics from Harvard College. As you know, this academic honor is restricted to approximately 75 students each year, which must be why you regretted only managing magna:
          http://www.onenation.org/0111/111201.htm
          So sorry to hear that you missed summa! :(

        • Ron:

          Nobody here is making “very, very serious charges.” My correspondent merely pointed out that your numerator and denominator came from different sources, and that when the names analysis was performed on the Harvard students, the estimate was 10-11%, not 25%. My other correspondent reported that a number that you estimated at 2.5% was, in fact, over 12%. Also other problems with your counting, as discussed above.

          Regarding your points 1, 2, 3, see NB’s comment (which explains that Hillel “indicated that Harvard used to collect religious preferences cards from freshmen but that this practice ended ~20 years ago. Thus, Hillel’s data was at some time accurate and worthy of citation by scholars and the media.”

          Regarding your points 4, 5: As I explained at length in my post above, you made several errors that went in the same direction. What you did seems reasonable from a distance (and I can understand why you made the choices you did), but his conclusions don’t seem to hold up on closer inspection. That’s just the way it goes. In statistics, sometimes you can use rough numbers and things work out, other times the errors don’t cancel.

          Regarding your point 6: Please be polite on this blog, also please use some common sense. Just because someone goes to the trouble of reanalyzing data that you have already analyzed, and then comes to data-based conclusions that diverge from your claims, that does not make them “intellectually undistinguished.”

          I understand that you feel embattled, but I urge you once again to reflect on my earlier comment. You did a lot of data analysis on your own, and you made some mistakes. These mistakes didn’t cancel each other out; rather, they all went in the same direction. That happens, especially when a person does not have a collaborator with a critical eye. I recommend that you step out of “debate mode” and into “learning mode.” You can use “debate mode” when you’re on a panel with Alan Dershowitz. On this blog, “learning mode” is more appropriate.

        • @Unz

          I see a lot of appeal to authority in your posts above. Some great scholar relied on Hillel data; hence it must be right. X is a recognized authority and X relied on this (possibly flawed) dataset and hence I can too.

          Are you fighting a legal debate here?

          Why don’t you try and address the substance of the criticisms?

        • @Unz

          Your #6 was absolutely uncalled for. Personally, I’m still on the fence as to who is right on the data: you or Andrew, Mertz et al.

          But remarks like #6 just make it harder for people to read your arguments in a neutral light. What happened to common courtesy?

        • Ron,
          I have not taken sides nor do I have a stake in this debate.
          The arguments Andrew makes, seem valid although I think that some could be adressed rather easily by you. That would be a nice follow-up and lead to somewhere.
          I assume the goal is not to back up own ideas with some data, but let the data speak.

          I must add that it is my conviction that sincere researchers do not use ad hominem remarks in their rebuttal. A pity that you resorted to such a remark in (6).

        • Ron, as I suggested on your site, why not use law school admissions – which, much like Caltech’s, is entirely numbers based – to test your theory. Law school class sizes are relatively small, and you can easily obtain recent commencement programs which lists the names of all graduates. If Jewish applicants were receiving an admissions boost at elite colleges you would expect it to catch up to them when applying to elite law schools, which only care about the LSAT.

          Here is Columbia Law’s 2011 graduating class, which had an average LSAT in the 99th percentile. Rough estimate, I count about 20% jewish. http://www.law.columbia.edu/graduation-2011/541529/2011-degree-candidates.

          And here is Penn’s class list which is over 20% Jewish (also of note, their medical school is also about 20% jewish). http://www.archives.upenn.edu/primdocs/upg/upg7/upg7_2011.pdf

        • More confusion I’m afraid…

          If you’ll carefully reread my 30,000 word article, you’ll notice that I only applied Weyl Analysis at one single point, namely to the aggregated 40-odd NMS semifinalist lists that I had managed to locate for 25 states. The combined total of 23,000 names was sufficiently large to allow Weyl Analysis sampling to produce meaningful results, and my sole use of the technique was to validate the direct inspection methodology that I had applied to all those states and also used everywhere else in my paper. Weyl Analysis sampling produces erroneous results when applied all but very large datasets, and you yourself discovered this when you noticed it over-predicted the results for about half the individual states (though you failed to notice it under-predicted the results for the other half). One would think that such matters would be covered in an elementary statistics course.

          Meanwhile, all my Jewish enrollment data was drawn from the most utterly reputable sources available, including leading scholars such as Karabel, Oren, and Synnott, sources that everyone has always accepted without question. To the best of my knowledge, no one has ever previously questioned these statistics. However, you are certainly welcome to do so yourself. For idiosyncratic reasons, you seem to be focusing your criticism on the most solidly established portions of my article.

          If you had actually bothered reading Karabel’s book or my citations from it, you’ll discover that he dramatically closed his 700pp study by emphasizing Harvard’s very high 2000 Jewish enrollment of 21%, which had actually surpassed the once-dominant white Protestants. The figure he cited was from Hillel. If you believe Karabel’s claim is totally erroneous, you should certainly bring that important fact to his attention.

          Certainly some of the other claims I made were vastly more controversial, notably the detailed accounts of corrupt admissions practices at Wesleyan, Harvard, and various other universities, in which I named names and provided specific dollar amounts, notably including payments to Harvard totalling $2.5 million. Under certain circumstances, such charges could surely be regarded as libelous. However, I was simply quoting the published accounts of Pulitzer Prize-winning former WSJ reporter Daniel Golden and NYT National Education Correspondent (and now College Admissions Editor) Jacques Steinberg. Since I wasn’t present at the events they describe, I can’t personally vouch for their accuracy, but anyone who disputes my article should take the matter up with Golden and Steinberg, not myself.

          I think I once wrote an article in which I mentioned North Dakota. But I’ve never been to North Dakota, and can’t really be sure it exists, just because “everyone says so.”

          Overall, I suggest you get in touch with Karabel, Oren, Synnott, Golden, and Steinberg, as well as the top officials at Hillel, and resolve your disagreements with them.

        • Unz: “you yourself discovered this when you noticed it [Weyl Analysis] over-predicted the results for about half the individual states (though you failed to notice it under-predicted the results for the other half).”
          No, Weyl Analysis under-predicted ONLY 8 states, all of which are “rather small” (i.e. fewer NMS semifinalists than MA, which you called “rather small”). These 8 states are Alabama, Colorado, Kansas, Louisiana, Minnesota, New Mexico, Washington, and Wisconsin. They contribute a total of 208 + 265 + 159 + 190 + 318 + 99 + 344 + 324 = 1,907 NMS semifinalists out of 12,163 total NMS semifinalists that you examined, i.e. you overestimated the % of Jewish NMS semifinalists compared to the Weyl Analysis results in 8 “rather small” states representing 16% of the NMS semifinalist names examined, whereas you underestimated the % of Jewish NMS semifinalists in about twice as many states, many of which are larger than the overestimated states, have high NMS qualifying scores, and are highly represented at Harvard. Thus, you substantially underestimated the % of high-ability Jewish students in the Harvard applicant pool.

          Your claim that “Weyl Analysis also produced a substantially *lower* estimate for the other 17 states I used” (including California) is false. I provide significant detail to disprove this claim
          here.

          I also noted here that you mysteriously omitted the most easily accessible TX list of NMS semifinalists, which just so happened to have 3 times as many Weyl Analysis names as the roster to which you linked. Thus, averaging the two lists gives the result that 3-4% of TX NMS semifinalists are Jewish (you reported TX NMS semifinalists as 3% Jewish).

          Feel free to post the details (just as I did) to support your claims.

          Unz: “my sole use of the technique was to validate the direct inspection methodology”
          I acknowledged that above. Repeating myself: “You stated that your direct inspection method for identifying Jewish NMS semifinalists produces results within 0.1 percentage point of Weyl Analysis. Since these methodologies produce virtually identical results by your own admission, and I cannot replicate your subjective direct inspection method, I shall henceforth state that your NMS results were based on Weyl Analysis.”

          You can cite all the scholars who used Harvard’s Hillel data back when it was primarily based on religious preferences cards until you’re blue in the face, but you continually fail to address that you are using two entirely different methodologies to estimate the % of Jewish NMS semifinalists and the % of Jewish Harvard students. Performing Weyl Analysis on the names of ~6,600 Harvard College alumni yielded the result that 10-11% of Harvard undergrads are Jewish vs Hillel’s 25+% figures.

        • Rob Schacter:

          Well, I don’t really know much about the LSAT or the law school admissions process, but those estimates—presumably obtained by “direct inspection”—don’t surprise me. As I mentioned in my article, psychmetricians agree that Jewish ability peaks in the Verbal subcomponent, being weaker in Math and generally mediocrate in Spatial. Combining that with the strong Jewish historical and cultural affinity for law and argumentation, you’d expect top law schools to be disproportinately Jewish. Since the East Asian profile follows an inverse pattern (high in Spatial, low in Verbal), you’d similarly expect their law numbers to be small relative to their general ability, with cultural factors reinforcing this. If anything, I would have guessed that the Jewish enrollment at top law schools was far higher than 20%.

          It would be interesting to try to track down the comparable figures for other elite grad programs such as med schools, whose students would have a different profile. I’d strongly suspect that the Jewish/Asian ratios would be substantially shifted from the law case.

          The main reason I focused on elite undergrad enrollments is that they constitute a single choke-point in the system, being the targets of the overwhelming majority of top students, regardless of particular interests. But after college, all the choices diverge, and you’d need to somehow separately examine and somehow aggregate professional schools—law, med, business, engineering—as well as all the different academic programs, and students who just go straight into business or something else. There just doesn’t seem any practical way of doing that sort of analysis.

          But if someone did manage to locate the graduation rosters of the various top professional schools and estimate their Jewish/NJW/Asian/Hispanic/black distributions, I’d bet they’d show some intriguing patterns based on type.

  21. There seems to be a suggestion here that Unz’s analysis is faulty because he’s not using the same methodology to count both the numerator and the denominator. In a post above, Professor Mertz claims that “one can still obtain reasonably accurate estimates with large data sets IF (and only IF) the %s in the numerator and denominator are determined by the exact same method.” But this surely can’t be a disqualifying condition for statistical analysis. I would imagine that numerators and denominators being derived by highly disparate means would be the rule, not the exception. For example, crime rates – crimes data (the numerator) are gathered by the FBI via local police departments (and not all that consistently). The denominator – population – are gathered by an enumeration or population estimate by the census. We also compare data internationally, where reporting and collection methods differ widely. So this particular criticism seems rather off-the-mark.

    • Ziel:

      To the extent that crime rates are measured wildly inconsistently across jurisdictions, it is indeed a problem to use them to compare cities. In the case discussed by Unz, it happened to be possible to use the two different methods to estimate the very same number, and these estimates differed by a factor of 2.3-2.5.

      I actually used this example in my statistics class yesterday! It is indeed often a good idea to combine data from different sources, but then it is important to calibrate these comparisons as well as you can. In this case the calibration is available and it would be a serious mistake not to use it.

  22. I think NYESQ’s citation to Hillel’s page claiming that Boston College has 0 graduate students, is pretty much a nail in the coffin for Unz’s argument.

  23. Now I understand what you mean by “incompatible” and “incommensurate”: the two estimators produce “much different” results when applied to one particular situation. That’s not what “incompatible” and “incommensurate” usually mean, but okay. Now I am puzzled by two things. One is that it isn’t clear that the two estimates that you call “much different” are reliably different. Yet your argument seems to depend on that. I was under the impression that people underestimate the effect of randomness. Maybe this is happening here — you are underestimating its effect. The other is that although the two estimators produce superficially-different results (which you call “much different”) when applied to one particular situation, they produce clearly consistent results in several other cases. Why do you ignore the other calibrations?

    I guess you don’t understand Unz’s hypothetical Mormon example. Here’s what I think he meant. Throughout history, powerful people in a certain group have sometimes (often?) oppressed less powerful people in another group by using admission or promotion criteria that discriminate against the less powerful. This behavior is sometimes described with words that end in -ism: racism, sexism, anti-Semitism, and so on. It is widely acknowledged that this has happened. Of course Unz is not saying that you are doing this. He is saying that this historical pattern gives his conclusions more plausibility.

  24. I misunderstood the estimators involved. I see that there are three (name analysis, Weyl analysis, Hillel) and that although the first two give similar results in various situations, the Hillel estimates have no such support. How inaccurate do you say they are? A factor of 10?

    Given how much they have been trusted, saying they are off by a factor of 10 is quite a claim.

    • Recall that Weyl Analysis is the methodology used by Unz to validate his direct inspection method for identifying Jewish NMS semifinalists. Unz claimed they produced results within 0.1 percentage point of each other. I performed Weyl Analysis on the names of Harvard College students from the Classes of 2009-12, which yielded the result that 10-11% of Harvard College students were Jewish in Fall 2008 vs Hillel’s claim of 25+%. This is a factor of ~2.5, not 10.

      Thus, our argument is that one must use the same methodology (Weyl Analysis) on both data sets (NMS semifinalists and Harvard undergraduates) in order to obtain a statistically valid result.

      We are not suggesting that only 10-11% of Harvard College students are Jewish. I would guess that the actual number lies somewhere approximately halfway between the two figures.

  25. Pingback: Links 2/15/13 | Mike the Mad Biologist

  26. Pingback: The Ivy League as gatekeepers for the elite « Phil Ebersole's Blog

  27. Pingback: Asian-Americans, Jews and Ivy League admissions « Phil Ebersole's Blog

  28. What a colossal waste of brainpower, if that is what to call this obsession with — what exactly? Obsession with Jews? It’s not as if Jews overachieve at Orientals’ expense, as if all to took to overachieve was a degree from Harvard. If that was it, why not require a degree to be conferred on every minority, except Jews of course, at taxpayer expense, for fairness? Liberals would be deprived of one issue to whine about for possibly ten minutes, before finding another example of unfairness that must be addressed.

    Imagine if you will, a university system concerned only with teaching the best students how to think, regardless of race, gender, ethnicity, or political orientation. Their minds could be filled with the wonderful achievements of Western Civilization before we had all the political subjects. Naaa – not enough victims or villians.

    • Michael:

      Nobody’s “whining” here. Unz (who, by the way, identifies as a conservative, not a liberal) is claiming that elite colleges—a powerful and important institution—discriminate. He also implicitly argues that it’s not just about teaching the best students: an Ivy League education isn’t just about education, it also gives students valuable connections they can use later in life. On the other side, Unz’s critics argue that Unz got his statistics wrong. This is not whining either; rather, it’s a recognition that this is an important issue and it’s worth getting it right. Just because something’s a topic that you happen not to care about, that doesn’t mean that the people tho do care about it are “whining.” And just because people disagree with you about the purpose of a university, that doesn’t mean that they are searching for “victims or villains”; they might just have different ideas about what a university education is about.

    • Andrew is too nice to rub your face in it, but it’s perfectly clear that your response to this is almost entirely predicated upon your assumption that Unz’s argument is motivated by liberal politics when, in truth, Unz is conservative and his article appears in a conservative magazine and he perceives his critics to be liberals.

      A part of me lives for these little moments on the web, when a political partisan takes a position on an issue and attacks his perceived enemies while getting the politics entirely backward. I’d like to think that the people who make these sort of mistakes will learn something from it (and partisans of all varieties make this mistake); but human nature being what it is, I don’t expect it.

  29. Andrew Gelman writes:
    “Using the scale-up methods, you get an estimate that 10-11% of students at Harvard are Jewish, not 25%. My correspondent suspects that the scale-up estimates are too low and that Hillel’s numbers are too high.”

    So if 10% is too low and 25% is too high, then lets not pick a number in the middle like 17.5%. No, let’s use that too-low number and build on it:

    “OK, so going from 25% to 10%—that’s a factor of 2.5. What about the rest? Two things: geography and counting.”

    Pure comedy. Hint to Gelman: Don’t preannounce that the number you are about to work with is wrong and then proceed to run with it. Be more subtle =)

    • Dan:

      I think you’re misunderstanding. If estimate #1 gives 25% and estimate #2 gives 10%, then they are discrepant by a factor of 2.5. If the true number is 17%, then estimate #1 is off by a factor of 25/17=1.5 and estimate #2 is off by a factor of 17/10=1.7, hence their ratio is off by a factor of 1.5*1.7=2.5.

      In my sentence that you quote, “the rest” refers to the difference between Harvard enrollment and National Merit Scholar semifinalists. If the Harvard enrollment number is 17%, then the method used to estimate the number for NMS semifinalists is basically off by a factor of 1.7, even before accounting for geography and counting. In the context of the discussion, Unz was talking about a discrepancy of 25%/6%, and I was explaining how a factor of 2.5 of this discrepancy can be explained by discrepancies between two different measures (as can be seen by applying them to the same unknown quantity), and how some of the rest can be explained by geography and counting.

      I recognize that statistics can be difficult, that’s one reason I run this blog and it’s one reason that I reply to commenters such as yourself who display confusion. In return, I ask for you not to be rude.

  30. Andrew:

    When you said “OK, so going from 25% to 10%” it seemed like the 10% number was supposed to have some real world significance rather than just a mental marker of discrepancy. But sorry if I seemed rude.

    The broader point is this:

    Whites who are not Jewish are 55.1% of the population in the 18-21 year old demographic while whites who are Jewish are 1.8% of the population in the same demographic. Whites in total are somewhere around 47% of Ivy league enrollment. Unz estimates Jewish enrollment in the Ivy league in the ~25% range. Thus he finds that Jews are over-represented by their numbers by 13.9x while non-Jewish whites are under-represented at only 40% of their numbers in the population. This amounts to a discrepancy of greater than 3000%.

    Clearly the demographics of the Ivy league have almost nothing to do with the demographics of America as far as whites are concerned. I graduated from Cornell in 2000 and saw this then. Things have apparently gotten worse.

    You talk as though the discrepancy you have to overcome is merely 25/6 or a factor of 4. No true. The discrepancy you really have to deal with is a factor of 30x plus, the difference between Jewish white and non-Jewish white admissions.

    By Unz’s data, non-Jewish whites account for 67% of national merit scholars, Jews account for 6% and Asians account for 26%. So even taking into account that Jews are a high-performing group, we have roughly 1000% over-representation in relation to similar-performing non-Jewish whites to contend with.

    You haven’t come close to closing this gap.

    I find it quite interesting that you have relied most heavily on some hidden data from ‘my correspondent’ while Unz openly shares a plethora of data.

    The larger part of the admissions puzzle is not why high-performing Jews get into Harvard. The larger part of the Unz’s puzzle is why similarly performing whites who are not Jewish face such dramatically longer odds. This central question is hardly addressed by you at all.

    • Dan:

      Unz’s factor of 4 was going from the asserted 25% Jews at Harvard to the asserted 6% Jews who were National Merit Scholar semifinalists. However, the 25% and 6% came from different sources. If you use the same source, you get 10-11% and 6%. And the 6% itself is too low as (a) it is a U.S. average rather than a weighted average over the states that supply more Harvard applicants and students, and (b) Unz’s percentages in those states were lower than what you get using the scale-up method.

      I think everyone agrees that Harvard admits a greater percentage of Jews and Asians than are in the general population. The dispute is whether the admissions percentages are greater, after adjusting for academic skills and achievement of the high-school-age applicants. That’s why it was also relevant that Unz’s estimate of Mathematical Olympiad participants was too low by at least a factor of 5.

      • Andrew,

        I fear we are still talking past each other. I point out that the ‘factor of 4’ is just one piece of the discrepancy that Unz pointed out. Ron Unz did not address Jewish admissions in a vacuum. It is in the context of enrollment rates that are 3000% higher than non-Jewish white rates and enrollment rates that are still 1000% higher when taking into account similar performance on the g-loaded PSAT.

        You vaguely cite geography as the source of the difference, without data. I found the data and it seems your geographical explanations fail utterly.

        http://theivycoach.com/2014-ivy-league-admissions-statistics/
        Harvard University
        “Geographical Diversity: In terms of geographic representation, 16.8% of the admitted students hail from New England, 23.3% from Mid-Atlantic states, 17.5% from the South, and 9.7% from the Mid-West, 1.7% from Central States, 3.8% from Mountain States, 17.1% from the Pacific, and another 10.1% of the admitted students are international.”

        Harvard is not a regional University, you see. It admits more from the South than New England; more from the Mid-West and West than from the Mid-Atlantic.

        You have emphasized that Jews are concentrated in the Northeast, which I cannot dispute. Harvard seeks regional diversity as shown above, which would not help but hurt Jewish enrollment, because very few Jews live in the South or Midwest. Thus, the Jewish advantage over non-Jewish whites must be starker than even Unz allowed.

        Again, I speak not as a resentful outsider but as someone who benefitted from an Ivy League education and who is himself 1/4 Jewish by ancestry, a fact I did not think to disclose at admission time.

        • Dan:

          Indeed, in my blog post I mentioned that approximately 40% of Harvard’s students are from the northeast and mid-Atlantic, which is what you also have written. Thus I believe it would be appropriate to take any U.S. averages and weight these states more heavily, given that they represent less than 40% of the U.S. population.

          Finally regarding your “3000%” and “1000%”: according to my correspondent who went to the trouble of counting all those names, if the Weyl method is applied to Harvard undergraduates, it gives an estimate of 10-11%, and if it is applied to NMS scholars, it is something close to that (whatever you get by taking the appropriate weighted average of 9-14% from Massachusetts, 24% from New York, 14-21% from Pennsylvania, etc). That’s what seems to happen if the same method is used to estimate both numbers. I have no particular reason to trust the Weyl method which I’d never heard of before reading Unz’s article, but it makes sense to use methods as similar as possible when counting a numerator and denominator.

      • For the record, I think the Jewish admission advantage is nothing sinister at all. Historically they were a genuinely oppressed minority and the desire to help such a group is quite reasonable.

        Ivy admissions have long openly assisted minorities and it is quite understandable that many admissions officials would include Jews in this. The oppression was real and it was horrible.

        But as Unz has showed, oppression is hardly a part of the present Jewish landscape in America.

        • The risk is that when perceptions eventually catch up with reality, Jews could be seen as an unfairly advantaged group. The numbers are so extreme that it seems unlikely that this perception can be held at bay for long.

        • Dan:

          As far as Ron Unz and David Brooks are concerned, perceptions have greatly outpaced reality. They report that 2.5% of 21st-century Olympiad competitors are Jewish, when the number is actually over 12%. As a statistician, I would like perceptions to equal reality.

        • White christian students need not apply to these “elite” schools unless they are legacy, offspring of the famous or recruited for athletics. Imagine a high school graduate, in the Northeast, white, Catholic, 4.444 GPA, 2360 SAT, great recommendations, varsity athlete, community service, being rejected by all of the following schools. Statistically, it is not unreasonable to believe that at least one of these schools would have offered admission:
          Harvard, Yale, Princeton, UPenn, Swarthmore, Williams, Amherst, U Chicago, Davidson (regional rep at Davidson was also the “director of multi-cultural admissions”). Unfortunately, this example is far from unique.

        • Davidson College has much less competitive admissions than the other institutions you listed, which suggests some other explanation for your anecdote. In particular, please note that Davidson College states the following about its mission:
          http://www3.davidson.edu/cms/x924.xml
          “The Christian tradition to which Davidson remains committed recognizes God as the source of all truth, and believes that Jesus Christ is the revelation of that God…”

          I knew plenty of white Christian students at Harvard who were not legacies, offspring of the famous, or recruited athletes. btw, William Fitzsimmons, the Dean of Admissions at Harvard College for the past 25+ years, attended a Catholic high school and was a member of the Catholic Students Association at Harvard:
          http://www.thecrimson.com/article/2008/3/31/the-gatekeepers-life-when-harvards-future/

        • NB,

          On the Davidson website:

          “Davidson’s admission process is highly selective and holistic in nature.”

          “holistic in nature” says it all. There is no other acceptable consideration beyond discrimination, except the possibility that Davidson believed the student, if accepted, would not attend thus hurting there US News rating stats.

          I’m glad you are aware of plenty of white christians who attended Harvard many years ago (Fitzsimmons). The admission process today is what is being questioned. Though, you are correct that white christians with mucho bucks will likely have an advantage.

        • Railroad, are you suggesting that a Christian college discriminates against Christians?

          I posted the Harvard Crimson article so you could learn more about the ethnic/religious background of the Dean of Admissions at Harvard College for the past 25+ years.

        • NB,

          Your question, “Railroad, are you suggesting that a Christian college discriminates against Christians?”

          No, but I do believe that any one student of the highest academic merit, such as the student indicated earlier, can be rejected in the name of diversity. From the Davidson website:

          “Davidson is strongly committed to diversity and inclusiveness. This commitment is driven in no small part by our grounding in the Reformed Tradition.”

        • Nb,

          On the basis of academic performance, as well as other accomplishments previously mentioned, the student clearly should have been accepted. Davidson does not get the diverse applicant pool as the ivies, so it is little surprise that they are 71% white. However, they are now clealy looking to add to diversity where they can be.

          Nb, if I am mistaken, what would you suggest was a reason for the rejection of this student at Davidson. You should know that the student did receive a merit scholarhip at another quality institution that recognized the students credentials.

        • I don’t know the whole story, so I’m not in a position to suggest the reason. I’m merely indicating that the fact that the student is a white Christian is almost surely not the reason he was rejected since he was rejected from a mostly white Christian college too.

          For example, Unz tells a story like this in his Meritocracy article, supposedly recounting an anecdote from Jacques Steinberg’s book “The Gatekeepers”: “Consider the case of T. Wang, a Chinese immigrant student…Although English was not her first language, her SAT scores were over 100 points above the Wesleyan average, and she ranked as a National Merit Scholarship semifinalist, putting her in the top 0.5 percent of high school students (not the top 2 percent as Steinberg mistakenly claims). Nevertheless, the admissions officer rated her just so-so in academics…Ultimately, he stamped her with a “Reject,”…[T. Wang] was also rejected by all her other prestigious college choices, including Yale, Penn, Duke, and Wellesley.” Sounds suspicious, doesn’t it? Based on the way Unz presents the anecdote, one is led to believe that T. Wang (whose first name I’ve redacted so that Google searches for her don’t pull up this negative info about her) was rejected on the basis of her race. However, I actually read T. Wang’s story in “The Gatekeepers,” and Unz left out two critical details: 1. T. Wang had multiple Cs on her HS transcript. 2. In one of T. Wang’s submitted recommendations, the teacher stated he was surprised that she had earned NMS semifinalist status.

          Thus, if Unz had included these highly relevant details in his profile of T. Wang, the reader would have arrived at an entirely different conclusion…

        • Nb,

          The GPA was 4.44 (5’s on 5 AP exams), no C’s on the students transcript – lowest grade was a lone B+ in 4 years, co-captain varsity athlete, jazz band for 2 years, community service, great recs, etc. There had to be some less qualified students accepted at Davidson.

    • Dan: “I find it quite interesting that you have relied most heavily on some hidden data from ‘my correspondent’ while Unz openly shares a plethora of data.”
      The only way I could use the same methodology on the data sets of both NMS semifinalists and Harvard students was to search the names at alumni.harvard.edu. I am not hiding anything – I can email you the names if you post your email address. If someone wants to post Harvard’s commencement programs and perform Weyl Analysis on them, I welcome them to. I also welcome a neutral party with access to alumni.harvard.edy to check my results.

      Furthermore, while Unz might openly share a plethora of data, it is rather a different question of whether he is accurately reporting what this data shows. I urge any neutral party to verify my claim that Unz omitted/underestimated (vs Weyl analysis) the % of Jewish NMS semifinalists in EVERY northeastern state, which also happen to have high NMS qualifying scores and to be major feeder states to the Ivies. Unz’s claim that he overestimated (vs Weyl analysis) the % of Jewish NMS semifinalists in 17 states is patently false as I demonstrated here.

      Finally, Unz does not take into account the fact that northeastern states like Massachusetts (which he omitted) supply more students to the Ivies than Texas. Check out this map that shows that there are 70% more students from MA than from TX at UPenn:

        • Also note that the UPenn class of 2016 includes 20.8% Asian-Americans not counting the 10% of US admits who did not indicate their race. Presumably, most of the latter 10% are Asian-Americans and whites who did so because they feared discrimination in admissions to Ivy League colleges from having read articles similar to this recent one by Unz. Thus, the % Asian-Americans admitted to UPenn from the US pool is probably somewhat greater than 20.8%. N.B. had stated previously that Harvard’s class of 2014 had 22% Asian-Americans in it, again not counting students who failed to indicate their race. These numbers are significantly higher than the 16-17% quota Unz concludes exists for Asian-American admissions to the Ivies, strongly suggesting that Unz’s claim that these colleges are limiting admission of Asian-Americans to 16-17% does not currently seem to be happening at these colleges.

      • @nb —

        Gelman’s post was about Harvard. As UPenn is a slightly less prestigious university than Harvard, it is just a bit more of a regional university. As my link shows, Harvard has extraordinary national balance.

        • Harvard does not have “extraordinary national balance.” Your link shows exactly what I’ve said several times, which is that over 40% of Harvard undergrads hail from New England and the mid-Atlantic. New England and the mid-Atlantic represent 23% of the American population, so students from the northeast are overrepresented by almost a factor of 2 at Harvard!

          Princeton is about/almost as prestigious at Harvard, and note the following:
          http://www.princeton.edu/main/news/archive/S30/15/00I77/index.xml?section=topstories
          “…the largest number of students admitted from California, followed by New Jersey, New York, Pennsylvania and Massachusetts.”
          Massachusetts, the 14th most populous state, is more highly represented at Princeton than Texas! Please note that Unz omitted MA from his NMS lists, claiming it had negligible impact.

          I would guess that if Harvard were to release similar info, it would read as follows: CA, NY, MA…

          Note also that the Princeton link above states:
          “Admitted students self-identified among the following racial and ethnic groups: 22 percent as Asian American…”

    • Dan,

      There are ~16,000 NMS semi-finalists in the US per year. Some of them live in states where the cutoff score is as low as 200 (out of 240); others live in states where it is as high as 220. Harvard admits ~2,000 students per year to achieve a class size of ~ 1,650. Clearly, simply being a NMS is not close to sufficient for admission to Harvard. I’m guessing that is part of the reason Unz also looked at data from the Olympiads and Intel STS, students who get readily admitted to the elite college of their choice (assuming the rest of their application is fine). When one uses this higher level of achievement, the ratio of US citizen/resident non-Jewish whites to Jews may well be only ~2:1, roughly similar to what their ratio in the Ivies may be, if we could count these 2 groups accurately. In other words, the huge disparity you are talking about largely disappears if one requires something close to this higher level of achievement at Harvard.

      • “Clearly, simply being a NMS is not close to sufficient for admission to Harvard.”

        That depends — a high proportion of blacks and hispanics who are admitted never reach this level.

        More to the point, admissions officials have long told us that above a certain level of achievement, most student they reject could have succeeded and they are trying to build a class. Ok. You cite the Intel STS as a cause for students to “get readily admitted to the elite college of their choice ” … really?

        Here’s an account in the Post dating from the last time letters were mailed out:
        http://www.washingtonpost.com/blogs/class-struggle/post/why-getting-into-harvard-is-no-longer-an-honor/2012/04/04/gIQAPhBVvS_blog.html

        “Take a friend of mine, for example. Despite the 14 Advanced Placement tests (11 top scores) and two consecutive placings in the Intel International Science and Engineering Fair under his belt, he found no welcome at any of the eight Ivy League schools, and neither did his co-founded company aid him in clinching even a position on the wait-lists of several of their peers. His great weakness? He’s an Asian applying for financial aid.”

  31. Dan,

    First, many of the Black and Hispanics at the elite colleges are terrific students; some are among the very, very best students at Harvard, even achieving PBK as juniors. Unfortunately, Unz’s tables of data fail to include columns for these ethnic groups; rather, he simply incorrectly assumes that 100% of the non-Asian, non-Jews must be non-Jewish whites, thus over-estimating this % among the high achievers.

    Yes, I agree that the very top colleges have several-fold more outstanding applicants than they can admit; they could easily put together a totally different class of students who are almost as good as the ones they create. They are looking to put together a balanced class of students interested in all the different majors they offer, playing all the different instruments needed for their musical groups, skilled in all of the various sports for which they have teams, etc. Just like they don’t want all students from the same ethnicity or region of the US, they also don’t want 500 students who play the violin or basketball.

    The Intel ISEF is NOT the Intel STS. The ISEF gives out hundreds of 1st, 2nd, and 3rd place prizes in lots of different categories every year. Tying for 3rd place in some category in the ISEF as a sophomore or junior is very nice, but it is not equivalent of the very high level achieved by being named among the 40 top in the Intel STS or being one of the 4-6 members of a team representing the US in an International Olympiad. The latter indicates the student is among the very top dozen or so students in their particular field in the entire US. This is what Unz and I are talking about. For example, someone I know was “deferred” this year to the HYP college to which she had applied early; within a week of this year’s Intel top 40 finalist list being announced, one of the HYP schools phoned her!

    Lastly, one can’t make generalizations from individual cases. I personally know an Asian-American student who applied early action to Harvard this year who needs full-ride financial aid; they admitted him. Harvard has a ~$30 billion endowment. I find it hard to believe they are making admissions decisions based upon who needs financial aid.

    • You write of elite colleges, “they could easily put together a totally different class of students who are almost as good as the ones they create”

      On this we can agree. On this basis, it is very hard to understand why Jewish students see a ten-fold enrollment advantage over non-Jewish whites among groups that have similarly excellent performance on the g-loaded PSAT.

      I’d love to see an honest attempt to evaluate this factor of 10 difference among NMS honorees, rather than some far easier factor of 4 hurdle. If you are going to criticize Unz get to the meat of it rather than dabbling at the periphery.

      High performing whites who are not Jewish clearly face enormous headwinds as compared to other groups, including Jewish applicants who apparently face no such headwinds. It is important as a moral question to ask why this is.

      • Dan,

        Conditional on Unz’s claims of a tenfold enrollment advantage, yes, this is a big deal. On the other hand, conditional on the data that show the Jewish enrollment at Harvard to be about the same percentage as among top scorers on the PSAT, there’s nothing much to explain at all. Hence the relevance of the data that dispute Unz’s claims. I agree that once you accept Unz’s statistical claims, there’s a lot to be said. But, given the data that’s been presented, I don’t accept his claims.

        • Andrew,

          Again and again and again, you sidestep Unz’s central point which is, why is the discrepancy between top-scoring Jewish students and similarly top-scoring non-Jewish white students so vast?

          You ignore the latter group completely in your analysis. Top scoring non-Jewish white students are represented at Harvard and other Ivy league schools at a mere fraction of their percentage among top scorers on the PSAT. Even if you had shown that Ron Unz’s 6% Jewish share of NMS scholars reasonably leaps into the 20%-25% enrollment rate at Harvard that is cited so exhaustively, you still skip Unz’s central thesis, by avoiding the enormous discrepancy in relation to top-scoring non-Jewish whites students.

        • Dan:

          Sorry, but conditional on the data I’ve seen (that show the Jewish enrollment at Harvard to be about the same percentage as among top scorers on the PSAT, also the data on math olympiad students etc), I don’t think there’s any scandal here.

          Unless you want to argue that Harvard shouldn’t be so strongly relying on academic preparation and abilities. That would be a valid argument to make, it’s a diversity argument that Unz has himself made, but that’s a separate question.

        • Andrew,

          Yet again you completely avoid the central question. As far as I am concerned, you haven’t even begun to engage Unz. Group A is whites who are Jewish. Group B is whites who are not Jewish.

          Unz compares A and B and notices, that *with similar high levels of achievement on the PSAT* enrollment for A is a rate that is many multiples of the enrollment of B. His central thesis is about comparisons.

          You cannot even be addressing his question if you deal only with group A.

        • Dan:

          If a group is equally represented in the university and its pool of high-achieving applicants, than I don’t see the scandal. You refer to Unz’s comparison but that is based on numbers that have been questioned, for reasons endlessly discussed above. I agree that if Unz’s numbers were all correct, there’d be something interesting going on. When I first posted on the topic, Unz’s numbers seemed reasonable to me. But given the additional information from Janet Mertz and others, his numbers no longer seem reasonable.

        • “reasons endlessly discussed above”

          This discussion wouldn’t be so endless if you addressed group A and B together rather than group A in isolation. As long as you do so, you are avoiding the central question that Unz raised, which is the comparison between groups after taking into account very high performance.

        • Dan:

          OK, thanks for clarifying. Here’s what I’m saying. If Jews represent x% of Harvard students and y% of high-achieving students in the Harvard pool (where this pool could be defined in various ways, such as actual applicants or residents in states where Harvard students come from), and if x is close to y (as appears to be the case based on the numbers I’ve seen), then I don’t see the scandal. You’re saying, even if x is close to y, it can still be a scandal if the ratio is much different for other groups of whites.

  32. Dan, Unz’s figures on non-Jewish white enrollment at Harvard College are obtained by subtracting Hillel’s claim that 25% of Harvard undergraduates are Jewish from the reported % of white students. Unz obtains his figures for the % of non-Jewish white NMS semifinalists by classifying as non-Jewish white all non-obviously Asian and non-obviously Jewish names, even though such a group would include Jews with non-obviously Jewish names, non-Asian people of color, and biracial students with an Asian mother. By doing so, he obtains inflated figures for the %age of NMS semifinalists who are white Gentiles.*

    Unz claims here that his direct inspection method produced results within 0.1 percentage point of Weyl Analysis. Unz underestimated the % of Jews among US IMO participants from 2000-2012 by a factor of 5. This suggests that Unz’s direct inspection method and thus Weyl analysis are undercounting the % of Jews who are NMS semifinalists.

    That said, since Unz claimed that Weyl Analysis and his direct inspection method produce virtually identical results, we can perform Weyl Analysis on the names of Harvard students, thus using the same methodology that Unz used to estimate the % of Jewish NMS semifinalists. I did so and came up with the result that Jewish students represented 10-11% of Harvard College students in Fall 2008. So we now have the following disparity: 6% of NMS semifinalists are Jewish and 10-11% of Harvard undergraduates are Jewish. There are two critical points to make here:
    1. Almost half of Harvard’s American undergrads come from New England and the mid-Atlantic, a region that contains less than 1/4 of the American population. Thus, to properly account for the % of high-ability Jewish students in the Harvard applicant pool, one must make geographic corrections and weight more heavily northeastern states, which have much higher % Jewish NMS semifinalists than states like Texas, which are poorly represented at Harvard.
    2. I attempted to replicate Unz’s calculations of the % of Jewish NMS semifinalists, and I found that Weyl Analysis produced higher figures than Unz claimed for the % of Jewish students in all of the northeastern states (which are highly overrepresented at Harvard and have high NMS qualifying scores). The number of states in which Weyl Analysis produced lower figures than Unz reported is much smaller. This suggests that Unz’s figure that 6% of NMS semifinalists are Jewish is an underestimate. Perhaps part of this disparity is due to the fact that Unz used a restrictive list of Gold-[] names as I mentioned here. If I were to use a more restrictive list of Gold-[] names, perhaps my NMS calculations will more closely match Unz’s (though he has declined to clarify that point); however, the Weyl Analysis % of Jewish Harvard undergrad students will likely also decrease since I counted names like “Goldhill” as part of the Weyl J1 set of names.

    *As proof, look at Unz’s list of Texas NMS semifinalists:
    http://www.lovejoyschools.com/10TXSemifinalistsNatlMeritProgram.pdf
    Unz gives the breakdown of TX NMS semifinalists as follows: 68% non-Jewish white, 28% Asian-American, 3% Jewish, whereas I found that Hispanic and African names (since most African-Americans do not have African names, I surely undercounted them) represented at least 3% of TX’s NMS semifinalists on the 2010 roster.

    • “Almost half of Harvard’s American undergrads come from New England and the mid-Atlantic”

      Not correct. The figure is 40.1%. Sixty percent, the overwhelming majority of Harvard students, come from outside of New England and the Mid-Atlantic.

      Unz’s analysis is obviously not perfect, but it does not need to be. The gap in enrollment rates for the two groups is staggering. Even if Jewish high performers on the PSAT were fully double what Unz finds, and even if Hillel overstated Jewish enrollment by a third, you still would have a 300% or so enrollment gap for Jewish students in relation to non-Jewish whites after accounting for NMS performance. I am not saying Unz has made such large errors, but I am saying that he could and the issue would prominently remain.

      By the way, this seems hardly fair (wink!) On this thread at the moment I have three brilliant Harvard defenders tag-teaming against me. Fortunately the data and the discrepancies are so enormous that I think they speak for themselves, mostly.

      • Dan, if 40% of Harvard’s undergrads are from New England and the mid-Atlantic, and 10% are international students, that means 44% of Harvard’s American undergrads are from New England and the mid-Atlantic. 44% is almost half, no? We do not have PSAT data on international students. So 44% of Harvard’s American undergrads are from a region that contains less than 1/4 of the American population; thus, we’d have to weight that region significantly more in order to obtain an accurate figure for the % of high-ability Jewish students in the Harvard applicant pool, and it just so happens that according to Weyl Analysis, Jews represent a significant % of NMS semifinalists in this region.

        • nb,

          You inadvertently raise an excellent point, and one apparently overlooked in previous analyses by all parties!

          Unz compared the percentage of National Merit Scholars who were Jewish to the percentage of Jewish students at Harvard. But (numerators and denominators again) the proper comparison would be the percentage of National Merit Scholars who were Jewish to the percentage of AMERICAN Jewish students at Harvard (since only Americans are eligible for NMS).

          This would *add* about 10% to the target value. For instance if the percentage of Jewish students at Harvard is 25% (and a hundred Google sources all give approximately this number, ranging from around 20% to 30%) then the percentage of American Jewish students is about 27.3% *. Going from Unz’s 6% to 27.3% is that much more difficult, and this is even before taking into account the apparent unfairness in relation to non-Jewish whites.

          * Only 1.7% of Harvard’s international students are from Israel, meaning perhaps 0.17% of all undergrads. And Israel accounts for the overwhelming majority of Jewish people outside of America.
          http://www.hio.harvard.edu/abouthio/statistics/studentstatistics/historicaldata/MultiAnnSummary91-12Students.pdf

        • Dan, you clearly do not accept our basic premise that Unz must use the same methodology on both data sets, since you would not be comparing 6% to 25% or 27.3% if you were. The comparison is 6% to 10-12%, and that’s assuming Unz’s figure of 6% is even accurate, which remains unclear. I’ve repeatedly try to emphasize to you the importance of correcting for geography since almost half of Harvard’s American undergrads come from New England and the mid-Atlantic, where Jews are highly represented among the % of NMS semifinalists. Thus, the % of Jewish NMS semifinalists among Harvard’s applicant pool is considerably higher than 6%.

        • To quote Unz:

          “Mertz argues that I should ignore these Hillel estimates—which everyone else always uses—and instead perform Weyl Analysis on the surnames of all of America’s major universities to determine their Jewish enrollments.

          But this is a total absurdity. To the best of my knowledge, American universities do not make their complete lists of past graduates publicly available, and even if they did, the total number of such names for the Ivies, the University of California campuses, and the various other schools I considered would run into the millions over just the few decades I considered. Counting the Jewish names among them all would be insanity.”

          And you say “the comparison is 6% to 10-12%” — but this is completely wrong. That is not the central issue at all. The central starting point is that Jewish enrollments are 1000% higher *****AS COMPARED TO***** non-Jewish white enrollments after taking into account being a NMS scholar. (Remember that the 1000% over-representation is already a reduction from 3000% over-representation to take into account that Jewish students are NMS scholars at a high rate). Even if you cut 1000% that gap in half, you still left with a gap of 500%.

          Sorry for the capital letters, but Unz’s central point is continually avoided.

        • Dan:

          To summarize your point: as I wrote elsewhere on this thread, if Jews represent x% of Harvard students and y% of high-achieving students in the Harvard pool (where this pool could be defined in various ways, such as actual applicants or residents in states where Harvard students come from), and if x is close to y (as appears to be the case based on the numbers I’ve seen), then I don’t see the scandal. You’re saying, even if x is close to y, it can still be a scandal if the ratio is much different for other groups of whites.

        • Dan,

          You and Unz are not taking into account two critical issues:
          1. Geography. Almost half of Harvard’s American undergrads are from the northeast, which is 5.2% Jewish according to Unz:
          http://www.theamericanconservative.com/articles/meritocracy-appendices/#2
          In producing a national estimate of the % of Jewish semifinalists, Unz weighted each state by its number of NMS semifinalists (essentially by population), meaning Texas was weighted many times more than MA, even though MA supplies more students to the Ivies than TX, as you can see here. By weighting TX many more times than MA, Unz is vastly overestimating the % of non-Jewish whites in the Harvard applicant pool. And this does not even address that TX’s NMS qualifying score is lower than most northeastern states and significantly lower than MA’s, which I will address further in my next point:

          2. A significant problem with the NMS semifinalist data is the varying qualifying score by state. ~16,000 NMS semifinalists are selected from ~1.5 million juniors who took the PSAT/NMSQT. But these 16,000 are not simply the top 1% of PSAT scorers – they are the top scorers per state, and the total # of NMS semifinalists designated per state is proportional to each state’s share of HS students. So, for example, states like Oklahoma and Iowa have qualifying scores under 210 (i.e. SAT score of 2100), and NMS semifinalists represent the top 2-3% of OK and IA students taking the PSAT, whereas in MA with historically the highest qualifying score (221-223), NMS semifinalists represent the top ~0.7% of MA students taking the PSAT. In Unz’s data, OK and IA combined are given more weight than MA (Unz claims his including the estimate that 19% of MA NMS semifinalists are Jewish had no significant impact on the results) even though few OK and IA NMS semifinalists are actually Harvard material, while a far more significant share of MA NMS semifinalists are Harvard material. OK and IA NMS semifinalists are mostly non-Jewish white, as is the case with most, if not all, states with low NMS qualifying scores, and few of these NMS semifinalists are Harvard material. In fact, it is often the case that in states with high NMS qualifying scores, non-Jewish non-Hispanic whites are underrepresented among NMS semifinalists in proportion to their population in that state (as is the case with MA, the state with the highest NMS qualifying score).

          Since states with the highest NMS qualifying scores are most likely to contain a disproportionate percentage of successful Harvard applicants, it is these states that are most relevant. Unz even makes this point in the appendix but only when discussing that the national pool of NMS semifinalists likely represents a significant underestimate of the true % of high-ability Asian students. The same is true for Jewish students, but Unz conveniently omits that point. He goes on to say, “Similarly, California and Texas contain the two largest populations of Hispanics, both overwhelmingly Mexican-American, but Texas Hispanics are almost three times as likely to be NMS semifinalists. These large discrepancies are probably less due California Jews or Hispanics being much dimmer or lazier, than that California’s required qualification scores are so much higher.”
          http://www.theamericanconservative.com/articles/meritocracy-appendices/#5
          Unz’s CA and TX lists overlap in one year: 2010, and in that year CA’s NMS qualifying score was 2 points higher, so we’ve established that Unz considers a 2 point qualifying score gap as “much higher.” MA’s qualifying score was 3 points higher than CA’s and 5 points higher than TX’s in 2010.

          Thus, the national pool of NMS semifinalists is not a reasonable reflection of the Harvard applicant pool, so I don’t see any evidence for this “500% gap,” esp since non-Jewish whites are underrepresented among NMS semifinalists in the states with the highest NMS qualifying scores.

        • Dan,

          You quote Unz saying, “But this is a total absurdity. To the best of my knowledge, American universities do not make their complete lists of past graduates publicly available, and even if they did, the total number of such names for the Ivies, the University of California campuses, and the various other schools I considered would run into the millions over just the few decades I considered. Counting the Jewish names among them all would be insanity.”

          Unz’s statement is little other than a poor excuse not to perform this analysis. N.B. was readily able to access and perform Weyl analysis on data sets from Stanford and the top 3 Ivies, Harvard, Yale and Princeton. Unz could do likewise. Using computerized data bases, it is very easy to perform the analysis. One simply types in each of the 13 usually Jewish names and a list of the students with those names pops out. No doubt, Unz, an expert at computer programming, could very rapidly have done this analysis for the key elite colleges if he believed it would have strengthed, rather than weakened, his case.

        • Janet

          This is where I have a problem. Could you please post the 13 Jewish surnames in question? Has anyone done independent checks to determine their accuracy in random tests? Are these the names Unz was using? As you know, I’m concerned with the model itself. I’m willing to let the chips fall where they may if the model is sound.

        • Bud Wiser, the names are listed in Prof. Gelman’s original post above. Quoting from Unz’s appendix: “Cohen, Kaplan, Levy, and “Gold—“ (J1) which were suggested by blogger Steve Sailer and his Jewish correspondent, or else extended to include the full set of such names (J2) utilized by Weyl by adding Berman, Bernstein, Epstein, Friedman, Greenberg, Katz, Levine, Rosenberg, and Stern. Based on the 2000 Census estimates, the first group includes approximately 1 in 20 American Jews, while the larger set raises the fraction to 1 in 12.” It was unclear to me exactly what Unz meant by “Gold-” but I interpreted it as Gold*. If Unz meant a restricted list of Gold- names, then that might explain why I appear to be getting higher results for the % of Jewish NMS semifinalists using Weyl analysis. Since Unz did not respond to my comments seeking clarification on this issue, I took out Weyl’s book from the library and discovered that his list of names is slightly different. Weyl’s distinctive list of Jewish surnames is as follows: Berman, Bernst*, Cohen, Epstei*, Friedm*, Gold, Goldbe*, Goldst*, Greenb*, Kaplan, Katz, Levine, Levy, Rosenb*, Stern.

          I have done no research to determine the accuracy of either method. As I’ve stated previously, I suspect that Unz’s version of Weyl analysis produces underestimates.

        • Not being a statistician, I can only guess about how one could use Weyl analysis to get reliable results. But it seems to me to be possible, providing that one first ensures that a) the distinctive names are reliably representative with regard to the reference population, and b) the reference and target populations are comparable with regard to the representativity of those distinctive names. That is, you’d first want to be sure that your distinctive names are truly distinctive before you use them as a measure of the portion an ethnicity is of the reference population (the US census or social security or similar); and then you’d want to be sure that the target ethnic population (say, jewish NMS winners or Harvard grads) has about the same composition with regard to names as the reference population. If those two conditions are met, then you should be able to reliably estimate the relative size of an ethnic population in a target using those distinctive names.

          The statistical methods I’d assume are well-understood and available would check these assumptions and requirements. You’d use other demographic information to test whether the target ethnic population has close to the same composition as the reference population. You’d use other demographic information to test whether your distinctive names are really distinctive for the reference population. And so on. And, of course, you’d use standard analytical tools to check whether, for example, your target populations are large enough to support this kind of analysis (that is to say, how likely they will be large enough such that there’s not too much noise).

          If you do all these things, then it seems to me to be a very useful and reliable analysis. It all depends upon how careful you are.

          With regard to the larger argument, even if you do these things, you know that Weyl analysis results in relatively noisy approximations — how well the distinctive names are representative of the reference population is a bit questionable and, anyway, could change over time. You’ll expect that there’ll be quite a bit of uncertainty about the results for the target populations. Even so, if you don’t have any very accurate means of determining an ethnic representation in a population — such as direct questioning of the entire population (and presuming truthful answers!) — then you will have to rely on less reliable methods such as this Weyl analysis. Or a source such as Hilel (which would be preferable, or even an example of a very accurate means, assuming that you actually know how they generate their data and that it’s reliable … it’s here where Unz make a big mistake because recently Hilel’s data became very unreliable). Given less reliable methods/sources, and because what Unz is looking for is comparing the ratios of academically high-achieving jewish people in general and jewish people at Harvard, then Gelman and others are arguing that he should have used the same method/source for both groups so that what errors exist will be in the same direction and thus not disturbing the balance/imbalance between the two, which is what Unz was looking to compare. That’s what NB has been doing. Alternatively, he could have ensured that at least one of his methods/sources was relatively highly accurate — which is what Mertz is doing when she directly polls Math Olympiad competitors. (You still would have to remain aware that any imbalance found could be an artifact of the less reliable method/source.)

          It’s clear that Unz’s results are extremely sensitive to his choice of the very unreliable (in recent years) Hilel data and his poor use of a partial Weyl analysis (his direct inspection of names was ad hoc and very error-prone) — and that an analysis of the Hilel data shows that it certainly greatly overstates the jewish proportion of Harvard students while, on the other hand, his name analysis understates the jewish proportion of these others groups he’s using as proxys for high-achieving jews. Therefore, his finding that Harvard has a disproportionate number of jewish students/graduates is at the very least greatly overstated and possibly simply wrong. It’s not that any one thing in isolation was by its essential nature a mistake — trusting a source like Hilel wasn’t unreasonable, using different methods/sources for the numerator and denominators is not inherently a mistake. His shoddy direct examination of names is arguably egregious, I guess. But it boils down to what Andrew said from the beginning — he’s been pretty forgiving of Unz, saying repeatedly that it’s easy to make these sorts of mistakes. What’s hard to forgive is Unz’s unwillingness to accept that he erred and that his findings are badly flawed. And less generously to Unz, this ex post facto intransigence calls into question these choices — it’s hard to avoid suspecting that he picked the data he needed to find his desired conclusion. On the other hand, no likes to make mistakes or be wrong and intransigence when corrected by others is a pretty common human behavior, regardless of intention.

    • Can you make the Harvard data for your calculations public? Otherwise it is hard to take this seriously as a rebuttal of Unz.

      • Unz himself suggested that I publish my counts for each surname from the alumni directories, so I will do that when I formally write up my own rebuttal of Unz’s claims.

  33. Pingback: Unz on Unz « Statistical Modeling, Causal Inference, and Social Science

  34. Pingback: Domain experts are key in the age of big data | downwithtime

  35. Pingback: My beef with Brooks: the alternative to “good statistics” is not “no statistics,” it’s “bad statistics” « Statistical Modeling, Causal Inference, and Social Science

  36. Pingback: Unz on Unz « Statistical Modeling, Causal Inference, and Social Science

  37. Pingback: Area high school rankings : 1998 & today - Page 9 - City-Data Forum

  38. Pingback: Unz on Meritocracy: Dangerous Cancer Statistics | Ron Unz – Writings and Perspectives

    • Unz is, yet again, in his March 16th post entitled “Dangerous Cancer Statistics” pre-selecting the data that agree with the claims he desires to draw, while ignoring all of the other data that disprove his claims. He did this with his origin “Meritocracy” article, and he is doing it again with my 2008 and 2012 Notices articles. This is not the way proper science is done.

      In my 2008 Notices article, I show quite clearly that large differences exist among countries in the frequency with which girls qualify for membership on their country’s IMO team. For example, prior to reunification, West Germany had NEVER had a female on their IMO team. On the other hand, East Germany had numerous girls, with their names listed in one of my tables in this article. Both countries had an average rank of ~7th in the IMO in the 1980s. East and West Germans were genetically essentially identical. Thus, the most likely explanation for the failure of West Germany to have identified outstanding female mathematicians is sociocultural factors that differed between it and East Germany. One plausible difference was accessibility of child care, plentiful in East Germany when it was a communist country yet almost non-existent in West Germany; East German women were expected to work full-time while West German women were expected to stay home to care for their children once they became mothers. Thus, East German girls grew up with different career dreams and expectations than their West German counterpoints. West Germany had 2% of their tenured math professors at universities being female in the 1980s. Communist Eastern European countries averaged ~25% of their tenured math professors being female in the late 20th century. Harvard University had zero female tenured math research professors from 1636 until 2006! This is one example of how sociocultural factors can strongly influence the identification of females who excel in mathematics.

      In my 2012 Notices article, I go on to show that a strong correlation exists between % girls on a country’s IMO teams and its gender equity index. It’s NOT true that all countries have ~5% girls on their teams. The US had ZERO girls on its teams until 1998, while some countries have had girls on their teams throughout the past 1/2 century. In the 1950s and 1960s, only 5% of math Ph.D.s awarded to US citizens went to women; in the 2000s, ~30% did. Humans don’t evolve this quickly. These large differences and fairly rapid changes must be largely due to sociocultural differences and changes.

      I also show in this latter article that some countries do not show greater male variance in their distribution of scores in math performance; in some countries, the boys and girls distributions are essentially coincident. If greater male variance in math performance were primarily due to an innate biologically determined difference between the sexes as suggested by Larry Summers, such countries should not exist.

      Lastly, I have NEVER claimed that innate differences between male and females in math ability at the very high end don’t exist. Rather, what I have claimed based upon my data is that much of the difference in math performance, where it exists, is due to a variety of sociocultural factors that vary among countries. As long as these sociocultural factors remain, one can’t measure the size, if any, of the remaining difference that could be due to differences in “intrinsic aptitude” between the sexes.

      I have published ~80 peer-reviewed primary scientific articles over my 4-decade career as a research scientist. NOBODY, including Unz, has ever claimed ANY of the data I have reported in these dozens in articles was in error. It is fine to agree to disagree on how to interpret data. However, it is crucial for the progress of science that published data are error-free. When errors are identified, the onus is on the author to publish a “correction” so wrong data do not contaminate the literature. In his “Meritocracy” article, Unz clearly published data that contain significant errors due to his direct inspection method being highly inaccurate and imprecise. It is long past due for him to acknowledge this methodological problem with this study and to correct his erroneous data. I look forward to his doing so soon and requesting that David Brooks do likewise.

      p.s. If Unz has Googled me, he would know I am a basic research scientist, not a clinician. The only vertebrates on which I have ever performed experiments (in very small numbers) are frogs, mice, and rabbits. As a Ph.D. biochemist, most of my work in cancer research involves test tubes and petri dishes. Why would Unz “Suppose Mertz had conducted a study of two different cancer treatments, tested in trials across two dozen countries around the world”? I’m not going to bother responding to such irrelevant suppositions given nobody has ever questioned the quality of any of my published data or statistics.

      • Janet:

        Yes, I think this is the key point: “NOBODY, including Unz, has ever claimed ANY of the data I have reported in these dozens in articles was in error. It is fine to agree to disagree on how to interpret data. However, it is crucial for the progress of science that published data are error-free. When errors are identified, the onus is on the author to publish a correction so wrong data do not contaminate the literature.”

        As a realistic Bayesian, I’d say it’s too much to assume error-free, but we certainly try to minimize error when we can.

      • Janet:

        Can you elaborate more on the countries which “do not show greater male variance in their distribution of scores in math performance”?

        That seems the crux of your argument.

        Also, when you say “distribution of scores in math performance” do you mean Olympiad scores or some others?

        Somehow I see this % of females on Olympiad teams as a red herring. What exactly would a low / high percentage indicate? (In the West-East German case, yes. There it is indeed beautiful evidence.)

        • Rahul,

          Thanks for your interest. I suggest that you take a look at some of the figures and tables in my 2012 Notices article (www.ams.org/notices/201201/rtx120100010p.pdf) which is publicly available to download for free. Yes, one cannot look at distributions with IMO data because the IMO only permits a maximum of 6 students per country per year to participate, and the students who participate are a very highly selected group, not a random sampling of populations. In recent years, ~100 countries throughout the world have participated. Thus, these data sets per country are quite small.

          Thus, in my 2012 article, we looked, instead, at math data from the Trends in International Math and Science Study (TIMSS) from 2003 and 2007 that were administered to thousands of unbiasedly sampled 4th and 8th graders per country from ~ 50 countries around the world. We also looked at math data from the Programme in International Student Assessment (PISA) as well to make sure any correlations we found were not flukes of any specific test, year, age, or collection of countries. We have recently further confirmed our findings with the recently released 2011 TIMSS data sets. In other words, we looked at data from other 2 million students who participated in these exams. These data sets are also publicly available to be downloaded for free so anyone can analyze them if they can figure out how to do so.

          My co-author on my 2008 and 2012 Notices articles, Jonathan Kane, is a Ph.D. Mathematics Professor who also obtained masters degrees in statistics and computer science as well. My co-author on my P.N.A.S. article, Janet Hyde, is a Professor of Psychology with secondary appointments in the School of Education and Gender Studies Program. She has been doing social science research for over 3 decades, teaches graduate courses in statistics for social scientists, and has served as my primary social scientist resource person. Cathy Kessel has also served as a social scientist resource person I routinely consult; she has a Ph.D. in mathematics and extensive experience and expertise in mathematics educations, including serving on federal governmental panels. Yes, I trained as a biochemist/molecular biologist/cancer researcher, but brought on board appropriately trained statisticians and social scientists to serve as my collaborators and consultants to ensure my research was performed in an appropriately rigorous manner. Who were Unz’s collaborators and consultants to provide the expertise he might be missing for his research?

  39. Ron:

    1. As I noted in my email to you (which you posted), if you do not like the term “sloppy” you can feel free to replace it with “cursory,” which is the term that you used. If you look carefully, I did not describe everything you do as sloppy; rather I referred to “sloppy counting.” But, again, it would be fine to do a global search and replace and change “sloppy” to “cursory” or “causal.”

    I did not intend “sloppy” to be an insult to you, but of course I think it’s better, all things equal, to be careful rather than sloppy/cursory/casual. But I understand that all things are not equal: we are all busy people and I respect your judgment on how much time to spend on various aspects of your data collection and analysis. On the particular issue of the nationalities and ethnicities of Olympiad and Putnam participants, Mertz put in much more care than you. That does not make her a better or worse person than you or even a better or worse researcher in general terms; it merely happens that on this particular issue, her numbers are more trustworthy. The fact that she went to the trouble to gather good data here is not a criticism of you, it just means that it would make sense to use her data rather than making contrary claims based on your admittedly cursory analysis.

    2. Mertz’s claims are very specific and are based on her counts of various participants in math competitions. That her work is peer reviewed is relevant to judging her qualifications. Peer-reviewed work can be wrong, of course, but the fact that a paper is peer reviewed gives some information. For me to remark that a paper is peer reviewed, it is not necessary for me to review a paper in detail.

    In any case, the relevance of Mertz’s work to this discussion comes from her numbers, not her conclusions. You dispute her conclusions but not her numbers. You write that she “wildly mischaracterize statistical data” but this is all a dispute about interpretations. You do not say her numbers are wrong, and those are what is relevant to our discussion.

    3. Again you write of “the collapse of Jewish academic achievement in recent decades.” As has been discussed already (based on numbers of Mertz that you do not dispute), there is no evidence of a “collapse” but rather of a steady decline. (Based on numbers which you had gathered based on a five-minute cursory inspection, you had seen a drop by a factor of 17 in Jewish U.S. Olympiad team members from the 1970s to the 2000s, but based the more careful provided by Mertz, the drop was a factor of 2 or 2 1/2, still a large drop for sure but reasonably explainable by the declining fraction of Jews in the population and increased competition from Asians in recent decades.)

    4. You write that I seem to be “backpedaling from his criticism of my work.” I am not backpedaling. I continue to believe that you presented some erroneous comparisons and estimates, some of which unfortunately made their way into the NYT. I also have continued to emphasize throughout that to note the mistakes you have made is not to dismiss your ideas.

    • The idolization of “peer review” #2 is somewhat ironic given the number of crappy peer-reviewed articles that have been dissected on this blog before.

      Of course, Mertz may still be right and Unz wrong; but hardly because one is peer reviewed and other not.

      PS. Does Mertz have qualifications relevant to sociological research? Just curious. Maybe she does. (Anyways, neither does Unz probably.)

      My point is, what Prior should I use for my confidence in the validity of a sociological study coming from a career oncology researcher. The findings might yet be superb, but not because of a strong prior but perhaps in spite of a weak one.

      Nothing wrong with researchers switching around fields, but the danger we ought to guard against is to allow their reputational capital from one sector to unjustifiably transfer to their work in another. e.g. If I were to read a new Gelman article on CPU-Design I ought to use a Prior different from my usual one for a Gelman-Statistics-article.

      • Rahul:

        What I wrote was, “That her work is peer reviewed is relevant to judging her qualifications. Peer-reviewed work can be wrong, of course, but the fact that a paper is peer reviewed gives some information.” This is hardly “idolization” of peer review. If you want to see some serious idolization of peer review, check out this comment (not by me) on an earlier post on an unrelated topic.

        The real point is that nobody including Unz is questioning Mertz’s actual numbers, and it is the numbers that are relevant to this discussion, not the conclusions she is drawing from them, which is the only thing he is disputing.

        The point of remarking that Mertz has published peer-reviewed articles on the specific topic of ethnicity of math competition students is not that her work is necessarily correct but that it’s ridiculous for Unz, David Brooks, and others, to simply ignore her criticisms—especially in retrospect given that Unz does not dispute her numbers and in fact states that his own numbers came from five minutes of cursory inspection.

        If I were to publish some numbers based on five minutes of cursory inspection, and then I were to learn that someone who had published peer-reviewed work in that particular subfield had numbers that were different from mine by a factor of 5 (or even a factor of 2, or a factor of 1.2), I’d take a look right away. I would not dismiss or ignore the criticisms. I’n not talking about reputational capital here. This is just common sense. If you’re guessing at numbers, and then someone who’s done more work in the area comes to you with information, you listen.

        • Yes, mostly I agree with you.

          I’d just be happier if both sides appealed to authority / status / credentials a tad less. e.g. “I have published ~80 peer-reviewed primary scientific articles over my 4-decade career as a research scientist” or “David Brooks had been so surprised and impressed with some of my findings” were both a bit unnecessary to the debate.

  40. Professor Gelman,
    I would remind your blog readers to pause and consider the primary source of Mr. Unz’s NMS-related ethnicity data. Regardless of methodology, which may or may not be suspect- I trust your judgement, and that of your readers, Steve Sailer and his Jewish correspondent (as @nb said) did not provide sufficiently reliable data upon which to base a study with such a sweeping ethno-socio-political scope.

    @Keith M. Ellis and @nb, and you, write *really* well! Ditto for Janet Mertz and Rahul too. Wow, it is a pleasure to read their comments. I wish Keith Ellis wouldn’t consistently spell “Jewish” in all lower-case letters though. He spells “Harvard” as a proper noun. For contrast purposes, compare Christian, Buddhist, Muslim, Zoroastrian, jewish. Even your blog CMS red-lines jewish but not Jewish. Also, isn’t it “Hillel”, not “Hilel”? You never make that mistake.

    • Ellie Kesselman, thanks for the criticism.

      I’m always a bit uncertain about this. Unless the context requires otherwise, I prefer to assume that discussions of jewish social identity is one of ethnicity and not religious affiliation, because the latter is more restrictive than the former. And I tend not to capitalize ethnic identifiers. My discomfort with capitalizing ethnic identifiers is that my sense is that it signals a kind of essentialism that I dislike. Of course, I don’t deny that these various self- and social-identifiers are extremely important to how most of us live our lives; but in a culture that tends to be biased toward essentialist and exclusionary thinking about these sort of things, I don’t like to encourage this.

      More specifically, in this case, and as a gentile, my sense is that it’s more likely than not that antisemites would capitalize “Jew/Jewish” as a means of emphasizing otherness. Your implicit argument is that failing to do so implies a lack of respect, which I also regret. I suppose that I feel that if I have to err in one direction or the other, I’d rather err in this particular direction.

  41. Andrew, since you are a statistician, I’m rather disturbed by the fact that you ignore a crucial point brought up multiple times in comments — namely, the importance of considering Jewish applicants in relation to the white pool in general, not just to the overall group of highly qualified potential applicants.

    You responded above: “as I wrote elsewhere on this thread, if Jews represent x% of Harvard students and y% of high-achieving students in the Harvard pool (where this pool could be defined in various ways, such as actual applicants or residents in states where Harvard students come from), and if x is close to y (as appears to be the case based on the numbers I’ve seen), then I don’t see the scandal. You’re saying, even if x is close to y, it can still be a scandal if the ratio is much different for other groups of whites.”

    You criticize Unz at various points by saying that he just needs to consider the numbers you’ve uncovered, and stop distorting them on the basis of policy argument. Yet the exact same can be said for you here. It is COMPLETELY inappropriate statistically just to say that “if x is close to y,” then there is “no scandal.” Let me clarify the reasons.

    Your argument would be valid if there were a “Jewish checkbox” on the Harvard application form, as there is for Asians, African-Americans, Hispanics, etc. In that case, we could imagine a fictional admissions committee dividing up the applications into piles: the Asian pile, the Hispanic pile, the African-American pile, the Jewish pile, and then just the generic white pile. (Obviously this doesn’t actually happen in practice, but these racial differences are clearly given some weight.) Now, the admissions committee can say, “Well, we want more African-Americans and Hispanics than would normally be represented, and we’ll lower the Asian numbers artificially a bit” (assuming Unz is right). The Jewish applicants would just be normally represented at about 100% of what they “should be” based on population distribution, and the remaining large pile of “other whites” will serve for the remainder of admissions (obviously at a lower rate than natural).

    If that were the case, I completely agree with you that you could say, “x% is close to y%” so Jewish students aren’t unduly favored.

    But that is NOT the case. There is no “Jewish” checkbox on the application. So, in our slightly exaggerated example of admissions committee behavior, we have the Asian pile, and the African-American and Hispanic, etc. piles. And then we have one pile of generic “other whites” INCLUDING Jewish applicants. Somehow, magically, out of that single pile of “other whites,” Jewish students manage to admitted at a much greater rate than other non-Jewish whites.

    Or, in other words, if we correct Unz’s numbers to make them more in line with your findings, he says Jewish students are represented at 435% of what would be “normal” (i.e., representative of the applicant natural breakdown), while non-Jewish whites are only 28% of what would be normal, with the overall white admissions rate at 61% of normal. You claim that this factor of 4.35 for Jewish students is probably more like 1.0. Okay, fine. But since all the students who checked the “white” checkbox (including Jews and non-Jews) have a collective admissions bias at 61% of normal, that means non-Jewish whites are significantly lower than that number, even if Jewish applicants are only at 100% instead of over 400%. Just doing a quick estimate, it seems like the number of non-Jewish whites is probably on the order of 50% of normal.

    If that’s true, among students who check the “white” checkbox, somehow the admissions committee manages to comb through these applications carefully and bias in favor of Jewish students so that they are admitted at a rate that is about twice that of non-Jewish whites. This is not a small matter of putting students into piles based on simple checkboxes: it potentially indicates a specific effort to figure out which students MIGHT BE Jewish so they can be favored.

    Any other statistical conclusion is simply not in line with the facts.

    Now, it’s possible that this apparent bias in favor Jewish applicants might in part be due to other factors, such as a much higher proportion of Jewish applicants among the white applicant pool (though you already factored a lot of that into your geographical argument to get the number down from 435% to about 100%). It seems rather unlikely, though, that we can get the Jewish numbers down to be on par with non-Jewish whites just due to these factors. Even if the advantage is not double, but only a 50% advantage for Jews over non-Jews, that’s still a much more significant admissions factor that most other preference factors, like legacy admissions (which I think tend to be around 30-35% advantage).

    Furthermore, you overlook another significant statistical problem that might potentially make these numbers a lot worse in terms of bias. Because there is no Jewish checkbox on the admissions form, the number that matters in calculating admissions bias is NOT actual Jewish enrollment (whether that’s the 10-12% you get from your method or the 25% or so from the Hillel method), but how many students receive an advantage in admissions because they MIGHT BE JEWISH or could be PERCEIVED as potentially Jewish by the admissions people.

    Back when the Jewish quotas were in force at Harvard in the 1920s and 30s, apparently there were various classifications made of whether a person was definitely Jewish, probably Jewish, possibly Jewish, etc. If we have clear evidence that admissions people go to the effort of selecting more Jewish applicants out of the “white” pile than non-Jewish applications by a factor of about 2:1, then they must think along similar lines, even if such classifications are not explicit. Besides last names, maybe they look at where students live, grew up, parents’ education or occupations, what schools applicants attended, etc., for hints at how likely a student is to be Jewish. (Obviously some percentage of Jewish applicants may explicitly mentioned their Jewishness in the application, but certainly not all and probably not even most… I don’t know.)

    So, even if Jewish students end up being only 10-12% of the enrolled students, the search for Jewish applicants by admissions officers may actually affect a much larger group of students. (Perhaps, in this regard, Hillel’s figures might actually be appropriate, since anecdotal evidence seems to suggest that they may base their names on a “best-guess” about who MIGHT be Jewish on a given campus.)

    To take an extreme example of the effect I’m suggesting, say that 50% of the “white pile” of applicants is flagged as having some possibility of being Jewish. Say the actual percentage of Jewish applicants is only 5%. (I realize this is unlikely, but I’m trying to show an effect by an extreme case.) Now, say that admissions chooses 90% of the white students to admit from this 50% of “potential Jews.” The remaining 50% of the “white pile” are put at a significant disadvantage, since they only get to make up 10% of final admissions.

    Looking at the actual Jewish numbers in this case, we might just see that the actual Jewish numbers in the applicant pool might be 5%, but now they are up to 9%. That’s an advantage, but you might say (as you might say to my 2x as likely effect for Jews over non-Jewish whites above) it’s not that much. But the quest for Jewish students in this case actually disadvantaged the PERCEIVED NON-JEWISH students compared to the “MIGHT BE” JEWISH students by 9:1 in this case, which is a MAJOR bias in favor of POTENTIAL Jewish students, even if actual Jewish enrollment doesn’t seem increased by that much. If that were true, the search for Jewish students could have a HUGE impact on admissions.

    Now, obviously this is effect is probably not that large. But if actual Jewish students are being admitted at a rate of at least twice their non-Jewish white counterparts, based on the likely applicant pool, the fuzzy criteria used to identify “likely Jews” must inevitably cast the net wider, thereby influencing an even larger section of the pool, and thereby disadvantaging “definite non-Jews” in the “white pile” a lot more.

    In sum, I’m surprised that a professional statistician would just dismiss the possibility of bias in favor of Jewish students when it’s clear that they are admitted at a higher-than-expected rate compared to non-Jews, and the only way that could be the case is that there is extensive Jewish “profiling,” explicitly looking for Jewish students in the applicant pool, as opposed to other racial classifications that are simply read from a checkbox on the application.

    • In order for me to reproduce Unz’s result that Jews represent only 6-7% of NMS semifinalists in the 25 state aggregate according to Weyl Analysis, I had to interpret {Gold-} as {Gold, Goldberg, Golden, Goldman, Goldstein} and either half-count or zero-count hyphenated names (where half of the hyphenated surname was one of the Weyl distinctive Jewish surnames). Using this same methodology on the Harvard College alumni directory, I obtained the result that Harvard College was 7-9% Jewish in Fall 2008.

      Bob claimed, “Jewish students manage to admitted at a much greater rate than other non-Jewish whites.” What is the evidence for this?

      You are also not accounting for the fact that non-Jewish whites are underrepresented among NMS semifinalists in states with the highest NMS qualifying scores, as I discussed here:
      http://statmodeling.stat.columbia.edu/2013/02/12/that-claim-that-harvard-admissions-discriminate-in-favor-of-jews-after-checking-the-statistics-maybe-not/#comment-142949
      Indeed, there is a negative correlation between a state’s NMS qualifying scores and its % of non-Jewish whites, indicating that non-Jewish white NMS semifinalists are disproportionately from states with low NMS qualifying scores (some of which have qualifying scores that are only 97th percentile). In contrast, such correlations are positive for Jews and Asians. I will be reporting my findings in an upcoming blog post, where I will go into a lot more detail about this.

  42. @nb –

    “Bob claimed, ‘Jewish students manage to admitted at a much greater rate than other non-Jewish whites.’ What is the evidence for this?”

    The original post by Andrew here started with Unz. Here’s what Unz originally said: “Consider the ratio of the recent 2007–2011 enrollment of Asian students at Harvard relative to their estimated share of America’s recent NMS semifinalists, a reasonable proxy for the high-ability college-age population, and compare this result to the corresponding figure for whites. The Asian ratio is 63 percent, slightly above the white ratio of 61 percent… [snip] However, if we separate out the Jewish students [from the white category in general], their ratio turns out to be 435 percent, while the residual ratio for non-Jewish whites drops to just 28 percent, less than half of even the Asian figure.” This passage is where the numbers came from in my post.

    As I read this, Unz is identifying a Jewish advantage over non-Jewish whites of about 435/28 = ~15.5 times as likely to get admitted.

    Based on various data reported above in Andrew’s post (and apparently from your analysis), there’s a convincing case that Unz overestimates Jewish enrollment at Harvard by a factor of about 2.5 (or, at least, it would be so in an apples-to-apples comparison using the Weyl method for both NMS and Harvard students).

    Based on your analysis of data that showed some problems with Unz’s numbers on state NMS percentages and the omission of some data from high population Jewish states, so far I’ve seen evidence than Unz’s projections may be off by another factor of 2 there.

    Okay, so we have an error of 2.5 times in Jewish enrollment, and an error of about 2 times in NMS stats. That gives a total error factor of about 2.5 * 2 = 5 times.

    As Andrew mentions in concluding his post, “one of [Unz’s] dramatic numbers is off by at least a factor of 5.” At other points Andrew alludes to the rough factor of 4 by which Jewish students as supposedly overrepresented beyond the NMS pool. So I’m pretty sure I’m reading this correctly, and I’m assuming that he’s referring to stats like the 435% number mentioned in the quote from Unz above.

    HOWEVER — Since Unz’s analysis estimated an advantage of about *15.5* times for Jewish applicants over non-Jewish white applicants, a error factor of about 5 times less than that is not enough to dismiss the case for pro-Jewish bias. Unless you have more stats to throw into the mix, it would seem that we still have a factor of 2-3 times as great admissions for Jewish applicants over non-Jewish white applicants, given their distribution in the NMS pool of high achievers.

    • “Unless you have more stats to throw into the mix…” I do. The correlation between a state’s NMS qualifying score and its % of non-Jewish whites is -0.41. The correlation between a state’s NMS qualifying score and its % of Jews is 0.63. i.e. non-Jewish white NMS semifinalists are disproportionately from states with low NMS qualifying scores while Jewish NMS semifinalists are disproportionately from states with high NMS qualifying scores. This suggests that the avg non-Jewish white NMS semifinalist has a lower PSAT score than the avg Jewish NMS semifinalist. So comparing the number of non-Jewish white NMS semifinalists to the number of Jewish NMS semifinalists (or Asian NMS semifinalists, who also are disproportionately from high NMS qualifying score states) is an invalid methodology to use to predict the expected ethnic composition of Harvard.

      Since states with the highest NMS qualifying scores are most likely to contain a disproportionate percentage of successful Harvard applicants, it is these states that are most relevant, and non-Jewish whites are typically underrepresented among NMS semifinalists in states with the highest NMS qualifying scores, as I discussed here:
      http://statmodeling.stat.columbia.edu/2013/02/12/that-claim-that-harvard-admissions-discriminate-in-favor-of-jews-after-checking-the-statistics-maybe-not/#comment-142949

      • Okay — thanks for pointing me to that comment again. I didn’t quite get its relevance, but now I do.

        So, if NMS qualifying scores are not a good way to estimate the potential number of highly skilled college applicants, do you have a suggestion about what we can use? Apologies if you’ve mentioned this already in other comments: I did read quickly through the entire comment thread here yesterday, but there seemed to be a lot of repetitive arguments, so I may have missed something.

        I’m happy to admit I’m wrong, but I would also like to get a better estimate.

        Also, frankly, now that I understand your point, I think it’s frankly bizarre to try to argue against Unz by making methodological arguments about the number of Jewish students at Harvard. If the NMS data is a crappy way of estimating anything, then that should be the end of the argument. Andrew’s entire post above is superfluous. Because of all the subtle arguments about methodologies, I was led into believing that we just needed to correct Unz’s numbers a bit. But if what you say is true and the NMS data just doesn’t work AT ALL for this sort of comparison, we shouldn’t even dignify his argument with the kind of subtle analysis I’ve been reading here.

  43. There are several problems with the data provided by the correspondent Nurit Baytch. The first issue is with the way he performed Weyl analysis, which lead to him overestimating the number of Jewish National Merit semifinalists. He has already described the way he did this at the following link: https://sites.google.com/site/nuritbaytch/#_edn15. What this means is the part including “Pennsylvania: replication estimates 14-21% Jewish; Unz reported 9% Jewish” was the result of what I would call a misunderstanding by Baytch.

    The second issue is with his Stanford statistic, where it is claimed that “performing Weyl Analysis on Stanford’s public directory yields that 4-5% of Stanford’s undergrads are Jewish”, which is “half of the 9.5% Hillel statistic.” When I checked Stanford’s directory in this manner the estimate I got for Jewish population was roughly 10%, which matches the Hillel statistic (9.5%) very closely. Unlike Baytch, I will list the frequency of the names I found: 11 for Cohen, 8 for Kaplan, 2 for Levy, 4 for Gold, 3 for Goldstein, 3 for Goldberg, 3 for Goldman, 1 for Berman, 3 for Bernstein, 1 for Epstein, 6 for Friedman, 3 for Greenburg, 2 for Katz, 4 for Levine, 4 for Rosenburg, and 2 for Stern. That is 60 in total. Assuming that Weyl analysis is consistent with itself, which I think it is, and that there has not been a significant shift in the Jewish population at Stanford, which I am pretty confident there hasn’t been, this implies some mistake on the part of Baytch when searching these directories (which was not present when searching semifinalists).

    Weyl analysis seems to be consistent with itself because, although I am looking at lists from 2018, I am getting very similar numbers to Unz (even when the sample size is very small). In Pennsylvania I got exactly 9% from 5 names. In Texas I got 2% from 2 names (Unz got 3% from 3 names). In New York I got 20% from 17 names. In Maryland I got 8% from 2 names. In California I got 2% from 3 names (Unz got 4%). Those are all I’m going to do, but I think that proves the point, particularly when you consider there were 60 names for Stanford. If someone can find a year from the past decade where Jews were either 10% or 40% of semifinalists in New York, or something like that, that would prove me wrong.

    So, what I am saying is that Baytch’s statistics for Stanford are very far from mine (which were very close to Hillel’s), and since the statistic acquired from the Weyl method should not have changed by much if at all, he did something wrong. This is important for looking at the Harvard figures. Baytch claims that using Weyl analysis on Harvard’s public directory yields 5-6% (in the link I gave above), which I just do not believe given how far off that would require Weyl analysis to be. He also claims that his results are verifiable because Harvard has a public directory, but that directory is actually useless for looking up student names (they don’t come up), and he performed his analysis on different lists which are not public.

    Princeton has a public directory as well, which I analyzed, and the figure I got was 11% for undergraduates. This is actually above the Hillel statistic of 9%, but still close. There were 47 names, this is the talley: Cohen 11, Kaplan 3, Levy 6, Goldberg 3, Golden 2, Goldman 3, Goldstein 1, Berman 5, Bernstein 1, Friedman 3, Greenberg 1, Katz 2, Levine 2, Stern 4. (There are 5251 undergraduates)

    In summary, I wouldn’t trust any of Baytch’s Weyl Analysis stats.

    These are the name counts from the state lists:
    Pennsylvania : 1 Kaplan, 1 Golden, 2 Goldberg, 1 Friedman, estimated 60 Jews; 680 semifinalists; 9%. Texas : 1 Cohen, 1 Kaplan; 1340 finalists; 2%. New York : Cohen 1, Kaplan 2, Goldstein 2, Goldberg 2, Goldman 1, Golden 1, Epstein 1, Friedman 3, Rosenberg 4; 1010 finalists, 20%. Maryland : Friedman 1, Rosenberg 1; 315 finalists; 8%. California : Levy 1, Gold 1, Goldberg 1; 2050 semifinalists; 2%.

    • [quote]The first issue is with the way he performed Weyl analysis, which lead to him overestimating the number of Jewish National Merit semifinalists.[/quote]
      You’re correct that I performed Weyl analysis incorrectly for the purpose of Gelman’s first blog post about this matter (when I was anonymous). I explained in considerable detail here that Unz’s description of it was ambiguous:
      https://sites.google.com/site/nuritbaytch/#Weyl
      Before writing my comprehensive piece about it, I ensured that I was able to replicate Unz’s results:
      https://sites.google.com/site/nuritbaytch/#_edn13

      [quote]Assuming that Weyl analysis is consistent with itself, which I think it is, and that there has not been a significant shift in the Jewish population at Stanford[/quote]
      You’re incorrectly assuming that Weyl analysis is accurate. I got significantly varying results for different years when performing Weyl analysis on Harvard’s alumni directory. I performed Weyl Analysis on all the directories in 2013. The results you’re getting on the 2018 directories do not prove anything about my Weyl analysis results from 2013.

      [quote]He also claims that his results are verifiable because Harvard has a public directory, but that directory is actually useless for looking up student names (they don’t come up)[/quote]
      This is incorrect, as Harvard’s public directory can still be searched exactly as I described in 2013:
      https://sites.google.com/site/nuritbaytch/#_edn15
      e.g. if you search for Epstein, there are 2 undergrads (who are listed as Harvard College)

      • I understand that you are getting inconsistent results with Harvard, but this is probably because of a flaw with the Harvard directory or the way you are searching it. Either way it takes too long for me to verify (I can’t write up a program like you suggested Unz should do).

        Weyl analysis appears to be consistent with itself because of how close my results were to Unz’s when looking at the number of semifinalists by state, and the nearness of my results to the Hillel estimates when looking at Princeton and Stanford. When the results are that close with sample sizes which should be far below those of Harvard, often with only a few Jewish names, I seriously doubt that the Harvard estimate is actually way off. If this was the case, you would also be able to find a recent year where the estimate given by Weyl analysis for Jewish semifinalists in New York would be similarly off, particularly given the smaller sample, but you can’t.

        • That’s funny. When the data blow Unz out of the water there must be a strange “flaw in the Harvard directory” that selectively removes Jews from the list. Or a mysterious error in the way the list was searched by someone who very carefully reverse engineered the details of what Unz did so as to exactly match his methodology.

          (( DB: “Weyl analysis appears to be consistent with itself” ))

          Of course that should be the case in data sets where the expected number of J1-J2 (Weyl) surnames is large enough. That was never the issue. The question is the INconsistency of Harvard Hillel figures with the Harvard student directory J1-J2 estimates and other direct measures. There’s Brandeis sociologists’ survey (Soc Sci Research Institute) that put Jews at 11-14 percent of Harvard, the Harvard Crimson surveys with numbers in the 5-10 percent range for Jews-by-religion, and anecdotal reports of 10-15 percent Jewish enrollment from people who have been to Harvard College in the past decade.

          There is another problem with Ron’s Harvard Hillel numbers, that he refused to answer when I posted it at Unz.com comments. It’s a mathematical impossibility result: there is no plausible time series of annual percentage intake of Jewish students whose 4-year moving average equals the 2007-12 Harvard Hillel data used in Ron’s article, ie, 25, 30, 25.5, 25,25,25, 25. To go up from 25 to 30 and back down requires gargantuan ups and downs in the year to year Jewish percent intake, such as 25% until 2007, then up to 45% the next year and down to 7% the year after that and then back up to 23%.

          (( DB: “you would also be able to find a recent year where the estimate given by Weyl analysis for Jewish semifinalists in New York would be similarly off, particularly given the smaller sample, but you can’t.” ))

          The opposite is true. New York will be the stablest of all states due to the large expected number of J1J2 surnames. The problem for you and Ron is not that J1-J2 estimation is particularly variable (it is OK in that respect for states with enough Jewish NMS) but that it completely disagreed with the Harvard Hillel data for the 8 years or so of data that Nurit Baytch gathered.

        • Wow, first paragraph was obnoxious.

          I never said that there is a flaw in the Harvard directory, only that it does not include all names, and I don’t mean that strictly Jewish names are excluded. And it was ME who carefully reverse engineered the details of what Unz did – Baytch was apparently not so careful because the first time he tried it, he misunderstood what Unz did and significantly overestimated the Jewish NMSFs for many states, which he later admitted, though these stats are still used in this article. I pointed this out in my first comment.

          I see that Hillel has revised its estimate down for Harvard undergrads to 11%. I guess that means that Hillel was for some reason systematically overestimating the number of Jews at Harvard, much more so than at other colleges. The stats I got from Weyl analysis on both Stanford’s and Princeton’s directories (which are complete) still stand and they back up the original analysis done by Unz. However, it is now pretty clear (unless Hillel is actually underestimating, which they could be) that Harvard is not being particularly favorable to Jews compared to the rest of the Ivy League.

          The Hillel data is an estimate, they are not actually claiming that Jewish intake went up 45% in a year… that fact on its own does not really undermine any of Unz’s arguments. The estimate being moved to 11% does though. Also, Unz gets a LOT of comments, too many to even read. It is not surprising that he has not responded to you, and he did not “refuse” to answer.

          I’m glad we agree that Weyl analysis is accurate?

          I do guess Unz relying on Hillel stats could have hurt his argument, though he didn’t really have a choice and its not his fault for assuming they would be at least somewhat accurate. Though, the Harvard stat is the only one I know of that was way off. Now I think the Hillel stats are generally underestimates, or at least below what would be found with Weyl analysis… my stats for Princeton and Stanford which I found with Weyl analysis are now both a few percentage points higher than Hillel.

    • If anyone is interested, I replicated Door Bell’s analysis on the public Stanford database. There isn’t data yet for the number of undergraduates for the 2020-21 year, which may be impacted due to COVID, but the number seems pretty stable from 2017 (around 7100, the number I presume Door Bell did their analysis with because the 2018 data set came out in October of that year) to 2019 (around 7000).

      Assuming all else equal, I searched in the public directory for those listed as enrolled undergraduates. I found 34 matching last names with the Weyl analysis, and I was generous and rounded up to 35 because there was a “Greenberg” but no “Greenburg”. Here are my results:

      Cohen: 5
      Friedman: 4
      Stern: 4
      Kaplan: 4
      Goldberg: 3
      Katz: 3
      Levine: 3
      Rosenburg: 2
      Bernstein: 2
      Gold: 1
      Goldstein: 1
      Goldman: 1
      Berman: 1
      Greenb(e)rg: 1
      Levy: 0
      Epstein: 0
      =35 total

      To calculate the percentage of the student body, I just took the fraction (35/60) of Door Bell’s estimation (~10%) that was appropriate. Using this rough technique suggests that the Stanford undergraduate student body is currently ~5.83% Jewish, which is in line with, but a tad higher than, Baytch’s analysis.

      I don’t think that either Door Bell or Baytch were lying; it was probably year-over-year fluctuation. I think we should cool it with the fraud accusations.

      As a prospective Stanford student of Jewish descent (though I have a non-suggestive last name) I do hope to God that Door Bell is correct, if only for better admissions chances.

      • Addendum: it was my fault that I didn’t double check the Weyl analysis procedure. I now understand that the analysis includes any name beginning with Gold-, which adds one Golden and one Golding to the list. This increases the Jewish fraction from ~5.83 to ~6.17, which doesn’t particularly affect the conclusion at hand.

  44. My theory is simple. Jews control most of the wealth in the USA. With wealth comes power and influence. Jewish parents (like other parents) want their children to have the best education and thus the best opportunities in life. Therefore, they use their influence to ensure that their children get into these top schools. In short, money buys a lot in the USA – and if any other ethic or religious group controlled the majority of the wealth; you would see that group disproportionally represented in the top universities. Macro – America is the country that benefits the super wealthy and corporations mostly – unlike the social democracies in the EU, Scandinavia, Australia and Canada – where middle class/poor people at least have a shot at a decent education and livelihood.

  45. It should not be surprising that at the present time, Jewish students at Harvard have average academic abilities on average than non-Jewish white students. Remember: since the 1920’s, 20%+ of Harvard students have been Jewish, at least until 2000. That’s one reason the 25% figure Unz estimated is plausible— do we really think there are fewer Jewish students at Harvard now than in 1930? As a result, 20%+ of alumni are Jewish, which means a lot of Jewish alumni kids. Since in the past, the Jewish students were better academically than the WASP students, and IQ is heritable, we should expect a large number of Jewish applicants who are pretty good academically and just need the alumni bump to get over the bar, whereas lots of the WASP alumni applicants don’t have the test scores to get in even with the alumni bump.

    A second “bump” comes from being the child of a Harvard professor. There are lots of Jewish professors, so this group will be disproportionately Jewish too. (This is one reason for the high number of Mass. students, by the way.)

    A third bump is if your father is a big financial supporter of Harvard. Probably more Jews in that group too.

    A fourth bump is being the child of a celebrity– a politician, journalist, or suchlike. More Jews in that group too– considering that, again, you need a high IQ for the bump to get you over the bar, so children of actors are going to find celebrity status insufficient.

    An interesting question is whether Jewish students have an advantage in athletics. That would not have been true in 1930, but by now, lots more Jewish applicants are coming from elite high school where they play lacrosse, squash, etc. and everybody knows that jocks have an advantage in getting into Harvard. There probably aren’t many Jewish high school football and basketball players, but at the Ivies most jock do minor, gentlemanly, sports.

    Thus, nowadays, unlike in 1930, the percentage of Jewish students who get into Harvard for pure academic reasons is lower than that of whites generally and of Asians. Note that we can predict that in 50 years the percentage of Asians who get into Harvard on pure academics will also fall drastically, as the current Harvard undergrads become alumni grandparents.

  46. I’m surprised that nobody has mentioned that Unz has fallen prey to the binning fallacy here: if you take (A) the group of Jewish students who are in the top 5%, and (B) the group of non-Jewish white students who are in the top 5%, the distributions of ability within these two groups will not be the same. In particular, if we use IQ as a proxy for ability, the top 5% would be about IQ 125 and above. But Ashkenazi Jews have an average IQ of about 112. The distributions we are comparing are

    A: normal(112, 15) truncated below at 125
    B: normal(100, 15) truncated below at 125

    16% of distribution (A) lies above 140, whereas only 8% of distribution (B) lies above 140.

  47. Even though Jews are, on average, qualified, they are such a high percentage of alumni that they would get in under legacy admissions alone. So their iq doesn’t matter.

Comments are closed.