Meritocracy: Response to Prof. Gelman on Jewish Elite Overrepresentation
One noticeable disappointment in the ongoing discussion of my Meritocracy article has been the relative lack of critical commentary. Both my previous Hispanic Crime and Race/IQ series had unleashed vast outpouring of harsh attacks, thereby assisting me in sharpening and refining my analysis. But I think that so far the overwhelming majority of the many published responses to my recent research have either been favorable or at least neutral and descriptive. Fortunately, this somewhat lopsided state of affairs has now begun to change.
Early yesterday morning, Prof. Andrew Gelman, a statistics expert, published a sharp 3,500 word critique of my Jewish results, apparently based almost entirely on the critical analysis provided him by Prof. Jane Mertz and Nurit Baytch. As it happens, their material has been floating around the Internet for at least the last couple of weeks, and one or two people had previously forwarded it to me; I also discovered that Mertz had left a couple of hostile comments on the TAC website. Since I found their work confused and specious and they never made any effort to publish it anywhere—even if only on a personal blogsite—I never bothered to directly refute it. But now that Prof. Gelman has published major portions of it backed by his own imprimatur, I will undertake to do so.
I had actually already addressed some of these issues less than two weeks ago in a previous 1900 word column defending my techniques of Jewish surname analysis, but since neither Mertz, Baytch, nor Gelman seems to have bothered reading the piece, I must apologize for being forced to partly repeat myself.
First, Weyl Analysis—the use of extremely distinctive ethnic surnames to determine prevalence—is obviously a sampling technique, and can only be applied effectively on extremely large datasets. In my own case, none of the Olympiad or other competition lists were remotely large enough for this purpose, so any estimate of Jewish numbers could only be performed by direct surname inspection, which raises obvious questions about the accuracy of the latter approach.
In some of her TAC comments Mertz seems to have labored under the serious misunderstanding that except for explicitly Jewish surnames such as Cohen and Levy, I assumed that all other European ones, notably even including “Schwartz,” indicated Gentile origin. Obviously, if I had followed such an approach, I would be a total laughingstock, so I most certainly did not. In fact, as I have previously mentioned, the huge historical over-representation of Jews on lists of top academic performers led me to generally assume that nearly all distinctively East European or Germanic names were likely or almost certainly Jewish. This over-estimation was intended to partially compensate for the substantial fraction of Jews whose surnames—such as Miller, Gordon, or Brody—would be impossible to detect. For entirely similar reasons, I tended to assume that all “Lees” were actually East Asians, even though that surname is also quite common among American whites and blacks.
As I also pointed out in that previous column, I recently discovered that J.J. Goldberg, editor of The Jewish Forward, had published a column a couple of years ago, using surnames to estimate the number of Jewish Science Talent Search winners across several years, and when I compared his total to mine, they were virtually identical: he found 100 Jews for those ten years while my total was 96. Such a close match with the editor of America’s leading Jewish newspaper would tend to indicate that my methods are not wildly inaccurate.
However, the best independent check of direct inspection methodology would be to compare it the results from the much more precise Weyl Analysis, and fortunately the combined total of the dozens of NMS lists I located provided well over 20,000 names, a dataset sufficient for that purpose. Across all those lists, my estimate of the national total of Jewish NMS semifinalists was 5.95% based on direct inspection, and 5.92% and 6.03% based on two separate Weyl Analyses. These results tend to validate the approximate accuracy of my direct inspection methodology, which strengthens the case for its use on the STS, Olympiad, and other lists, for which Weyl Analysis is inapplicable.
One source of severe confusion on the part of Mertz/Gelman is their listing of eight or nine states in which their estimate of Jewish percentages derived from Weyl Analysis is larger—sometimes significantly larger—than my listed estimates of Jewish percentages based on direct inspection. But this is exactly what we would expect from applying a sampling technique to several small subgroups: in some cases it will overestimate the true figure. 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. Indeed, given the nature of small-size sampling, there are eight states whose Jewish estimates are zero under Weyl Analysis, but often quite significant in my table based on direct inspection. The key point is that once we aggregate all this sampling across all the different states, the overall statistical results match quite closely, as I described above. In effect, Mertz is claiming my sampling technique must be wrong because roughly half the time it tends to underestimate the result, without noticing that the other half of the time it tends to overestimate the result. I suspect that if she or anyone else made such a ridiculous logical error in one of Prof. Gelman’s own introductory statistics courses, she would surely be flunked.
One source of estimation error I had emphasized in my Appendix E was that the NMS lists I located covered only 25 states. But these states included the eight largest, and also contained over 80% of both the Jewish and the Asian populations, providing reasonable grounds for national extrapolation. After publication, I also managed to locate an NMS list for Massachusetts, and the percentage of Jewish names there was considerably above my extrapolated result. But MA is a rather small state, containing just 4% of American Jews and 2% of the NMS total, so the impact upon the national Jewish NMS average is negligible.
The next issue which Mertz/Gelman raises is the accuracy of the Jewish undergraduate percentages enrolled at the Ivy League and other major American universities as reported by Hillel, the national Jewish campus organization. I have myself repeatedly emphasized that the Hillel estimates might certainly be somewhat inaccurate, but that they are the only figures available, and are regularly used by The New York Times and all other elite MSM outlets, while also constituting the basis for Prof. Jerome Karabel’s award-winning scholarship. So I feel very comfortable in following the lead of every reputable organization and using the Hillel figures, though certainly with some caution.
Let me cite an example from my original article which underscores the credibility of the Hillel figures in elite circles. During 1999 it was discovered that Hillel’s estimate of the percentage of Jews enrolled at Princeton had dropped from 16% to about 10% over the previous 15 years, and this resulted in a huge national media firestorm, with articles appearing in the NYT, The Chronicle of Higher Education, The New York Observer, as well as four front-page stories in the Daily Princetonian, all with mention of anti-Semitism bandied about. Princeton’s administration profusely apologized for the decline in Jewish numbers and agreed to completely overhaul its admissions policy as a consequence. At that point in time Princeton’s president was Jewish, Princeton’s provost was Jewish, and Princeton had just finished building a multi-million-dollar Jewish activities center on campus; but Hillel’s report of a substantial decline in Jewish enrollment was regarded as near-prima facie proof of anti-Jewish bias at the university, especially since the figure was much lower than the figures reported by Hillel for Harvard, Yale, and Columbia.
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.
Next, consider the Mertz/Gelman complaint regarding my estimates of Jewish names on the Math Olympiad and the Putnam, both exams for which long historical datasets exist. For most of these, the number of winners listed each year is quite small, generally five or six, and obviously surname analysis or any other sort of inspection technique can easily produce errors for a given year, which is why I grouped the results by decade in hopes of minimizing these problems.
In particular, Mertz sharply criticizes me for suggesting a large decline in likely Jewish names, and cites my failure to realize that winner Brian Lawrence had a Jewish mother or that winner Daniel Kane’s family had Anglicized their name generations earlier. But such criticism is nonsensical, since just as I claimed, neither of those particular names is Jewish. Such identification error will always be a problem in any small sample analysis, but my argument is that the large decade-by-decade decline in likely Jewish names across every major competition category is probably real rather than merely spurious.
An important part of Mertz’s Jewish identification data was apparently drawn from her lengthy 2008 AMS journal article, which was repeatedly referenced by Mertz/Baytch and which I skimmed. The overwhelming focus of that article was to rebut the controversial speculation of former Harvard President Larry Summers that men might be somewhat better at math than women. Mertz and her co-authors demonstrated that generally some 5-10% of America’s top math students have been female, and claimed this refuted Summers, concluding that “the myth that females cannot excel in mathematics must be put to rest” (p. 1258, bold-italics in original). Frankly, I hardly think that Summers had meant to imply that no female could possibly do well in math, and I doubt he would be shocked that 10% or more of America’s top math students were female. I mention these points merely to provide some indication of the strongly ideological tendencies that seem to be driving one of my critics.
Although the article focused on gender issues, for reasons not entirely clear Mertz and her co-authors decided to undertake extensive background research (p. 1249) to determine the precise number of full-Jews and part-Jews among the math competition winners, summarizing those findings in exactly the same sort of Jewish/non-Jewish white/Asian tables that appeared in my own article. As it happens, their exhaustive biographical research determined that 26 of the 1988-2007 American Math Olympiad winners were Jewish or part-Jewish (p. 1253), while my own very casual surname analysis had estimated a figure of 23 for those same years. All things considered, I hardly view the difference between 23 and 26 as a gigantic discrepancy, nor evidence that my simple surname analysis tends to be wildly inaccurate.
Now let us combine these separate results.
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.
Next, the aggregate NMS semifinalist lists, though certainly imperfect, seem the best national dataset of high academic ability students, with the total numbers large enough to allow Weyl Analysis to be performed to determine ethnic distributions. Jewish surnames are hardly as distinctive as East Asian ones, but the almost exact national match between the results from direct inspection and those produced by two different Weyl Analyses lend reasonable confidence to the result. Under these estimates, the current Jewish share of high-ability American students seems likely to be around 6%, the Asian one at 25-30%, and the white Gentile total at 65-70%. I would strongly argue that the burden of proof shifts to anyone who argues for a substantially different set of figures.
Both these underlying estimates certainly contain a substantial nimbus of error, and for any result obtained by combining them, such errors would obviously be compounded. Indeed, if any of the statistical anomalies I found regarding Jewish enrollments had merely been in the 50% or even the 100% range, I would have simply discarded them as quite possibly due to measurement error. But my actual findings were in an entirely different range.
In one of the scholarly books cited in my original article, there is an extended discussion of the claims of Ivy League discrimination against Asian applicants during the 1980s. Hsia (1988) pp. 94-119 notes that after the admissions rate for Asians dropped a level substantially below the general rate, many observers viewed this as strong evidence of racial discrimination. The result was a federal investigation, and the determination of all the Ivy League schools to henceforth keep all their admissions rate figures absolutely secret so as to avoid similar problems in the future. Apparently an admissions rate anomaly of 20% or more was considered extremely suspicious, and this provides us with a useful benchmark.
Now let us return to the Jewish enrollment figures which were the subject of Prof. Gelman’s lengthy posting. If we combine the Hillel data with the officially reported racial data, we discover that college-age Jews in America are approximately 3,000% more likely to be enrolled in the Ivy League than their non-Jewish white counterparts. Even if the Hillel figures are indeed somewhat inaccurate, I strongly doubt that correcting for any such error would reduce the anomaly below the 20% threshold discussed above. Even correcting for the fact that Jews are more likely to live in the Northeast and Northeasterners are somewhat more likely to attend the Ivy League would have only a small effect.
A natural suggestion would be to normalize these enrollment figures based on the estimated totals of high performing students; but doing so is not entirely obvious. As I mentioned in my original article, the recent book published by a former Harvard Senior Admissions Officer claimed that these days Ivy League schools select only 5% of their students based solely on demonstrated academic ability, with the remaining 95% selected based on a holistic weighing of individual personal traits using a complex or subjective metric known only to the admissions office. So if nearly all Ivy League students are selected on amorphous, holistic grounds—rather than objective academic merit—perhaps the Jewish enrollment anomaly of the previous paragraph is actually the correct figure to investigate and somehow attempt to justify.
On the other hand, if we do choose to simply disregard the actual admissions policies practiced in today’s Ivy League and adjust our enrollment figures for high ability students, the anomaly of Jewish over-representation shrinks considerably but still remains rather large. We are now faced with the additional potential errors inherent in our surname analysis of the NMS lists, but the data indicates there are around ten or twelve high-ability white Gentiles in America for every such high ability Jew. Meanwhile, the Hillel figures indicate that the number of Jews and non-Jewish whites are approximately the same across the Ivy League. Thus, adjusting for the number of high ability students lowers the level of apparent Jewish over-representation to roughly 1,000%, a figure still comfortably above the 20% discrepancy threshold that helped spark a federal investigation in the late 1980s regarding Asian students.
As mentioned earlier, one irritating aspect of responding to this lengthy critique was that neither Profs. Mertz nor Gelman had apparently noticed that less than two weeks ago I had already published a similar column on Jewish surname analysis, which dealt with many of these same issues.
Still, they are hardly alone in such carelessness. By a remarkable coincidence, their critique was published almost simultaneously with that of a critical column by Prof. Kevin MacDonald, whose focus of greatest interest seems very similar to that of Prof. Mertz. In Prof. MacDonald’s case, he chided me for no longer discussing the Jewish aspects of my analysis. Apparently he, too, had failed to notice the same column of mine missed by Prof. Mertz.
Given that Profs. MacDonald and Mertz share such a strong commonality of personal interests, yet are both sometimes a bit prone to carelessness, perhaps they should join forces and henceforth work closely together to ensure that none of my future columns accidentally slips by them.
Finally, I should mention that I’ve been invited to give a Chicago Law School presentation next week, focusing on my recent NRO column. With the talk having the provocative title of “Asian Admissions Quotas: Was Bakke Based on Fraud?” I suspect the turnout will be reasonably good.