Ashkenazi PRS

There are now a couple of surveys with Ashkenazi EA polygenic scores.  We don’t know that the populations were representative, but probably they weren’t far off. Scores correspond to an IQ of about 110. Personally, I’m still pulling for 112.

A fair number of people publicly disagreed with our Ashkenazi hypothesis.

They were wrong.



This entry was posted in Uncategorized. Bookmark the permalink.

108 Responses to Ashkenazi PRS

  1. The Right Number for AJs says:

    Maybe interesting to some, was looking at how the US General Social Survey predicts interaction between migration status (migrated as adult, child, didn’t migrate at all, US born expatriate), ethnic background and Wordsum / education as proxy for IQ:

    Using migration allows you to remove the effects of foreign birth on Wordsum as a proxy for IQ, which are substantial.

    If you use Wordsum as a proxy for IQ, then among born in the US, IQ is for Jews about 110. East Asians about 105, Blacks about 90, Natives 88, and Hispanics about 91. (Compared with general Non-Hispanic Whites at 100).

    If you use years of education as a proxy, then among born in the US, Jews about 117, East Asians 111, Blacks 93, Natives 89 and Hispanics about 95.

    Suggests there’s a smarts advantage, but also an education advantage relative to smarts for all Non-Hispanic White groups, particularly Jews. I’d guess 110 for IQ is about right for Jewish “smarts” but the advantage is higher above the general population for representation in higher education. That matters if we’re looking at higher education achievement and then trying to infer from that.

    Education PRS might find some different values for AJs than IQ PRS.

    • Lot says:

      Jewish Wordsum in GSS understates AJ IQ because of a low ceiling. In other words, a relatively high number of AJ get a score 10 but would have had an even higher score if the test had a higher ceiling. Another issue with trying to get AJ IQ from GSS Wordsum is that the question asked is “Jewish” that that would include converts, mixed breeds, and Sephardi. About 1% of GSS Jews also ID as black. Years ago I estimated the exact effect of the ceiling and black on a GSS estimate and posted it on iSteve. The effect of white Jews who aren’t pure AJ (or AJ at all) isn’t something that can be easily estimated.

      • Josh Gonik says:

        Wouldn’t the point about a low ceiling also apply to a lesser extent to East Asians?

        Also, I really do wonder how much intermarriage has reduced the Ashkenazi Jewish average IQ in the US. After all, isn’t assortative mating widespread in the US nowadays?

      • The Right Number for AJs #2 says:

        Nice challenge.

        Removing the effects of Blacks who identify as Jewish is pretty easy, just control for the race variable (as I did in my constructed variable).

        Sephardis and Mizrahis you can’t control for easily (you could use country of origin and religion together but it seems messy). But they’re about 10% of the Jewish population, so at most you lose about 10% of the difference between them and AJs, and S and Ms seem like mild overachievers relative to the White American pop, so it’s probably not that large. (If you assumed SM were only White American average of IQ 100, then AJ only would go up from IQ 111.5 to get to that same target number).

        The range restriction estimates are a bit more difficult to address.

        The closest thing you could do is look at the composition of Jewish scores and test some possible changes on the mean that could represent effects of range restriction.

        About 14% of American born Jews score in the top category of 10 words answered correctly on Wordsum. If we shift about 2/6 of those 14% into an imaginary Wordsum 11 and 1/6 into imaginary Wordsum 12, we can see the changes on the mean here:

        That gives about 0.8 of a Wordsum converted “IQ” point. Move some of the 9s into 10s and you can get another 0.8.

        If everything above applies, you might get up to 112.5. For larger changes, you’d need a change in the Jewish modal number, which isn’t currently range restricted.

        Of course, this would probably be an overestimate, as range restriction also reduces the White American SD that I’m using to convert into “IQ” scale (with a slightly larger white SD say it’s 0.05 larger, the gain to AJ from considering range restriction is lost).

        “Purity” of American Jews I can’t talk about; they could gain from boiling off intermarriage at the low end or lose from gaining non-elite White ancestors (other intermarriage effects are of smaller effect).

        In any case, there’s probably a robust Ashkenazi Jewish education advantage well beyond their Wordsum advantage. (An advantage shared in lesser degree by East Asians). The “pure” education advantage for AJs beyond their Wordsum (IQ) advantage is probably almost the same size as their Wordsum (IQ) advantage.

        (Separately, you also see on GSS that within education categories Jewish Americans make much more money! That’s rather off topic, but this is all to highlight that indirect estimates of Ashkenazi intelligence from income or education status probably inflate the intelligence gap.)

        • Lot says:

          I agree with the first part of your correction. I don’t see the exact reasoning behind “move some of the 9s into 10s and you can get another 0.8.” Are you imagining a hypothetical more precise test instead of one that has an odd very non-bell shape result for Jews (with a peak at 9 but much less than expected 10s)? Notice that all the other groups do have roughly bell shaped results.

          If specifically concerned about purebred AJ’s, the ceiling effect is even more intense than the 14% of Am self-ID’d Jews who got 10s.

          In any event, I think purebred USA AJs born between 1940 and 2000 likely have IQs a fair amount above 110, probably about 115, but it is kind of a historical point as the number of them is rapidly declining and pure AJs have gone from ~4% of US births to under 1% now. “White Jews” that are some mix of AJ, Non-J White, and Other White Jew are the actual more relevant category going forward demographically.

          • Right Number #4 says:

            It may be a badly argued point; shifting the 10s over results in a less normal distribution of scores (as you note), so anticipating that, I proposed some shift of the 9s to renormalize a little bit (there are quite possibly better ways to do this), to test the argument in a maximally sympathetic way.

            As you note, there isn’t actually a justification in the sense that there is that some of the 10s would have been 11s if they could; the 9s could have been 10s if the could but they weren’t.

            Re; pure vs non-pure AJs, it’s hard to say anything about that within the resolution of the GSS. Trying to use the GSS, one indicator to at least estimate prevalence of mixed individuals in the self declared Jewish sample is to look at parental and maternal religion within each religious category, to check outmarriage:

            (re-coding the maternal and paternal religion variables into a single variable combining by paternal/maternal Jewish religious status).

            I don’t have any confidence in the sample sizes that I start to get when I do this though, as they become a small subset of what the Wordsum / education were calculated on. In that sample, the categories with less than two Jewish parents don’t drop the Wordsum… but the sample size is small it’s hard to have much confidence.

            Even is we assume along the lines of the above that about 20% of self declared Jews are half Jewish, then that equals about 10% “virtual gentiles” in the Jewish pool. Assuming that there is no selection of these virtual gentiles and they’re about average for White wordsum (100), then that might translate into the remaining Jews having a boost of about 2 IQ points to get to that 110-111 mean, pushing them up to 112-113. But that itself would be a questionable assumption due to assortative mating, so any loss would probably be less (I’d guess probably less than a point if there is even modest assorting).

            See –

      • Bonner Tal says:

        I once corrected wordsum scores for Jews for the ceiling effect (in the context of whether jewish IQ is declining) and it wasn’t a big difference (between 0.5 and 0.7).

    • Often EA is increased not due to ability but due to membership in a group. Here’s why using education as a proxy overestimates blacks’ IQ. Also many Jews get more enrollment in universities due to being a relative of alumni.

      • The Right Number for Askhenazi Jews #3 says:

        There might be individual differences in propensity for education between groups as well, but yes, it seems possible the “Culture of Education” between groups and group membership itself can affect the outcomes here.

        If you look at the wordsum by highest degree on GSS, the advantage within each category is lower among Jewish Americans that the whole population, and you also see that the correlation between Wordsum and years education is a bit lower for JA subjects than NH Whites and even East Asians (which gets around arguments about range restriction from graduate being the highest education degree rank).

        This is all considering only American born and raised.

        See –

      • Right Number #5 says:

        One further note, I’d add, turning the Kirkegaard Jan 2019 paper that I would guess inspires this thread, we find the same greater educational than IQ gap –

        Pop == IQ == Education

        Lutheran == 101.36 == 2.39
        Catholic == 101.48 == 2.45
        Jewish: == 109.72 == 5.62

        Expressed as SD with Catholic as baseline

        Pop == IQ == Education

        Lutheran == -0.008 == -0.03
        Catholic == 0 == 0
        Jewish: == 0.57 == 1.46

        Greater magnitude of educational achievement gap than IQ gap (2-3x?).

    • The Z Blog says:

      Here’s an old post using Russian chess performance to assess AJ IQ.

      Funny how the number 112 keeps turning up.

  2. Thom says:

    Prediction: those wrong individuals will soon pivot to claiming that it was never about Ashkenazi IQ per se, but Jewish IQ in general and that includes mischlings and Ethiopians and converts.

    Because Jews are absolutely a race, except when it’s more convenient for the conspiracy narrative to label them an ethnicity, religion, or arbitrarily-defined tribe. And we must at all costs avoid any precise definition such as phenotype or genotype.

    • Josh Gonik says:

      Let’s see what the data for Israeli Ashkenazi Jews (excluding those with partial ancestry from other Jews, such as Mizrahi Jews) will say once/if anyone will do a study or two on them. I certainly wouldn’t rule out the possibility that there was some selection in favor of intelligence for Ashkenazi Jews who came to the US. A selection for a couple of extra IQ points seems like a reasonable hypothesis.

  3. Matt H says:

    Isn’t this an example of garbage in garbage out. We know the IQ score of AJs, so when you do a polygenic score, you are using the same data that makes your polygenic score effective, so you are just fitting to the curve.

    This means you aren’t measuring genetic IQ just measuring membership in the group that has high IQ. I think this is a weakness of the scoring. If you don’t understand what a gene is doing you can’t use it for prediction. This might not matter for predictive power, but have your IQ poly score might just be looking at things that make you resilient to environmental problems, or genes common to groups that culturally think education is important. Of course if this is the major source of difference, maybe it doesn’t matter. Again score might predict embryos just fine, while only slightly measuring genetic effects on the brain.

    • gcochran9 says:

      No: the polygenic scores were developed on mostly non-Jewish samples.

    • Curtis says:

      The opposite of this is true however. If you choose a group (e.g. Africans, East Asians) that have little genetic mixing with the sample group (mostly European), you will not have a valid result. In other words, the genetic component of IQ for East Asians may have little overlap with the mostly European samples we have collected. Polygenic scores could be highly misleading.

      • Dr J Thompson says:

        The point of Piffer’s paper is that on the basis of European DNA he finds he can make reasonably accurate predictions about other genetic groups’ average IQ. This would not be possible if polygenic scores were highly misleading. The Lee “2018” paper found that on the basis of European DNA you could predict 11% of European individual’s IQs and, surprisingly, 1.6% of African individual’s IQs. This latter finding, coupled with the fact that European-DNA-based predictions on African’s heights are similarly attenuated, suggests that similar SNPs may be involved, although the precise locations are probably different.

        • Curtis says:

          Yes, this what I was saying. A polygenic from a mostly European population is a poor match for Africans. Your figures imply that it is not totally useless but has low predictive value. That sound right to me.

          Since Africans have a very diverse DNA, I would expect there to be similar ratios of predictive effectiveness between African countries using a Eurocentric score (e.g. one country could be 3% predictive and another 0.4%) And it may be harder to find 11% markers for Africans as a whole due to their genetic diversity.

  4. James Thompson says:

    Yes, increasingly apparent. 112 reasonable

    • Josh Gonik says:

      Do you also have a prediction for the genetic average IQ of Israeli Ashkenazi Jews?

      • Lot says:

        I’ve never seen a high quality study. While the farmer to Israel, scholar to USA effect might have created a gap at first, mean regression would have closed some of it. Israel AJs also benefit from low parental age effects, and a boiling off of lower IQ AJs into intermarriage with non-AJ Israeli Jews.

        You can guess at all of these factors, but I wouldn’t dignify it with the word estimate.

        • gcochran9 says:

          A. there weren’t any Ashkenazi farmers.

          B. The Ashkenazim that moved to the US weren’t an elite. not financially, anyhow.

          C. The majority of ashkenazi Jews moving to Israel did so after WWII: survivors.

          • Josh Gonik says:

            Was there actually a significant Jewish financial elite in Eastern Europe in 1900, though?

            BTW, I didn’t say that the Ashkenazi Jews who came to the US were a financial elite. Rather, I speculated that they might have been a cognitive elite. I don’t know if intelligence was as correlated with life outcomes back in 1900 as it is right now due to the presumably smaller amount of opportunities that intelligent people (as in, those without connections) had back in 1900.

            • Mike1 says:

              The Jews formed an almost unimaginable financial elite in Europe generally. I assume you are asking about the high IQ Eastern Europeans of scholarly fame and my perception is that they are as financially awkward then as high IQ people are today.
              Most high IQ people tend to be proud of high credit scores (a sign of bank profitability rather than social standing as a lot of people seem to think), playing weird games with credit cards like moving balances and refinancing mortgages to lower rates while remaining unaware of fees for doing so.

            • R. says:

              Uh, iirc, Jews owned something like 80% of Hungarian industry. Or something like that. 60-80%.

          • Josh Gonik says:

            You are correct that the Ashkenazi Jews who moved to Israel after WWII were Holocaust survivors. That said, though, even for Holocaust survivors, there were a few who moved elsewhere–especially among Soviet Jews, some of whom moved to the US instead of Israel.

          • Lot says:

            I think we agree here and I should have been clearer. Lynn’s Israel IQ figures are suspect and doesn’t reflect any sort of dumb AJ migration to Israel, but poor quality testing and translation of test results to the US white = 100 standard.

            The selection for IQ of US v. Israel AJs was very small, mainly because as you note there wasn’t a choice for the vast majority of them. Mandatory/Ottoman Palestine had heavy restrictions on Jewish migration, so 1880-1948 it was mostly US or nothing. For pennyless WWII survivors, both US and Israel were open for a while, but their Israel migration was subsidized by Zionist organizations and physically closer.

            However, Jews able to flee Germany and Eastern Europe to UK/USA in the 1930s were very IQ selected. They had the good sense to leave and the resources to do so, so perhaps were like 1960s cubran migrants. And the choice for AJs 1950-1990 of the USA or Israel probably had at least a little IQ selection favoring the USA.

            Finally, there’s the issue of AJ migration between Israel and the USA in both directions. All the AJ Israeli migrants to the USA I have met over my entire life have been academics, programmers/tech execs, or grad students of some sort.

            Overall, however, these are all just too small to have a real effect on AJ population average IQs beyond maybe half a point, not worth really wondering about in detail.

            • Josh Gonik says:

              Were you replying to me here?

            • J says:

              In most of the 20th Century there was selection between AJs choosing to emigrate to Israel or the USA. The choice was (and continues to be) between a poor country in permanent war (Israel) and the world richest, most tolerant and safe country with limitless economic opportunities (America). The mass moved to America, the elite (or the dumbest?), to Israel.

  5. gwern says:

    “A fair number of people publicly disagreed with our Ashkenazi hypothesis.”

    Wasn’t your hypothesis about rare mutations and ones linked to genetic disorders…?

    • Eponymous says:

      No. Those were simply evidence of recent strong selection. But selection should work on all variants.

    • gcochran9 says:

      Only in part. I assumed that selection was mainly on standing variation, which is why I used the breeder’s equation, and why I used to that to estimate the maximum possible gene flow that would still allow selection to generate a ~1 std change. But strong selection for any trait often raises up Mendelian variants with strong heterozygote advantage ( and homozygote disadvantage) ( like sickle-cell or the myostatin null in Belgian blue cattle) and if IQ-boosting variants exist in any human population, it would likely be in the population that has the highest IQ, and likely experienced the strongest selection for intelligence – the Ashkenazim.

      Of course nobody has looked at Tay-Sachs or anything like that.

      • gwern says:

        They haven’t looked at Tay-Sachs specifically but they have looked at exomes of high-IQ samples like TIP which are of course heavily Ashkenazi Jewish and found zero IQ-boosting variants. As they are enriched samples from an ostensibly enriched ethnicity and extremely well-powered to detect any effects like the postulated multiple IQ points… Even if you fall back to the weaker version of the theory that it’s all selection on common variants (and doesn’t the Dunkel paper suggest the entire advantage is mediated through the PGS ie. common variants?), it looks like much of your theory, like all of the Tay-Sachs stuff, was simply wrong.

        • gcochran9 says:

          MAF frequency threshold is usually 1%.

          • teageegeepea says:

            What does MAF stand for?

          • gwern says:

            I said ‘exome’.

            • gcochran9 says:

              where’s the TIP article?

              • gwern says:

                “A genome-wide analysis of putative functional and exonic variation associated with extremely high intelligence”, Spain et al 2015: TIP is similar to SMPY. No exome hits, total exome heritability similar to what one gets in non-Jewish white people using GREML-KIN, hints of lower mutation load. examines the common SNPs as well. If there are any T-S-related or similar exomic variants which are functionally the same thing triggering whatever mechanisms are supposedly increasing intelligence (given polygenicity or if you prefer that buzzword, ‘omnigeneticity’), how do, despite being frequent enough to be amplified by relatively short periods of selection to make meaningful contributions to group differences, they all manage to all be in linkage equilibrium with all the exomic variants and all the common variants against a genetic architecture indicating that rare variants are uniformly damaging without any large beneficial effects having ever been discovered?

                (Why not SMPY? SMPY was used to debunk one of the earliest candidate-gene IQ hits in the 1990s but, aside from the ‘argument from silence’ that the BGI project – which recruited from SMPY among other places – didn’t find anything publishable in their WGSes before it fell apart for organizational reasons, AFAIK SMPY has not been meaningfully genotyped in any way post-2000).

              • Article says:

                N=1409. I’d like to see Gwern’s power calculation, it doesn’t look like it holds up to me:


              • gwern says:

                What is your objection to the power analysis they provide?

              • Article says:

                “A threshold-selected case–control analysis of this design provides 80% power to detect associated variants explaining >0.0015 of the trait variance assuming an additive model and α=1 × 10−7.”

                The Gaucher’s disease allele (actually multiple alleles) is present in 10% of Ashkenazim. An unknown minority of the 1409 are Ashkenazi (with unknown ancestry share).

                For the Ashkenazi, with a higher mean IQ, the people in the study population are less selected (the threshold for the study is fewer SD above their mean), so IQ+ allele frequencies will be less inflated for them.

              • gwern says:

                “An unknown minority of the 1409 are Ashkenazi (with unknown ancestry share).”

                If you calculate out the order statistics, most or all of them will be Ashkenazi and will obviously be of higher ancestry share (begging the question about there being any genetic advantage, of course. They will still be enriched for all the rare variants supposedly boosting IQ, and all of them were – somehow – not detected. What’s the joint power for ‘any hit’?

                This explaining away all of the nulls, from D-F extremes to regular GWASes to gene enrichments to per-chromosome heritabilities to exome sequencing to common-PGS-mediation, is getting into pretty desperate ‘G of the gaps’ territory. The simplest explanation is that these supposed large-IQ-boost genes just don’t exist and their hypothetical existence was never based on very good evidence in the first place.

              • gcochran9 says:

                Most or all? Not any more, I suspect.

              • gwern says:

                The Makel et al 2016 ethnic breakdown I’m going off of reports that only 22% of TIP is Asian (which is, incidentally, about what one would predict using the order statistics & thresholds). The Asians would obviously be excluded from any GWAS and as Spain says, they did exclude them: “Exome array genotyping data were available for 1759 HiQ individuals who reported their ethnicity as ‘white’.” So, who are all these white TIPers? Well, they are mostly or almost all going to be Jewish, for the exact same reason that Asians are 22% of TIP despite being only 0.7% of the sampling population at the time (1981): tail effects.

                “So why are all those particular similar Mendelian diseases so common among AJs?”

                Bottlenecks. Lots of weird neurological disorders elsewhere too.

              • Anonymous says:

                Bottlenecks are not the answer. A bottleneck for Tay-Sachs would require 100% of the population to have it ~300 years ago. It must have some positive selection to make up for the negative selection.

              • gcochran9 says:

                I’ve done something like a million simulation runs, and it was almost impossible to get the observed Tay-Sachs frequency, even with two bottlenecks. But maybe I was missing something.

              • Tanturn says:

                IQ of 170? I don’t think IQ tests are reliable in that range.

              • gwern says:

                Simulations don’t have a good track record of reproducing the complex population structure and demographic history of human populations in general, much less minorities with as weird histories and niches. The population geneticists and statisticians have been wrangling over this for decades, which is a major reason why detections of polygenic soft selection sweeps remain so contentious and disappear/reappear with various methods.

                But my point is simple: there have been lots of opportunities for hints of rare highly-beneficial variants to have popped up in various ways as of 2019, and every single one is a zilch. “When you have eliminated the impossible, whatever remains is often more improbable than your having made a mistake in one of your impossibility proofs.”…

              • gcochran9 says:

                If I looked for myostatin nulls in cattle, I would find them, but only in the breeds subject to the most extreme selection for muscle mass.

            • Rolph says:

              Those who can, publish scientific papers with theories.

              Those who can’t, publish rambling repetitive online guides and go online and take shots at exaggerated versions of prior theories?

        • F says:

          So why are all those particular similar Mendelian diseases so common among AJs?

        • Bonner Tal says:

          I wondering about that recently. Tay-Sachs is reported by 23andMe so I assumed it might generally be part of the PGS-data. Would be great if somebody specifically looked at it maybe within-family design.

        • st says:

          Gwern, the foundation of your assumptions is bad (not your fault). (“…high-IQ samples like TIP which are of course heavily Ashkenazi Jewish ..”) – Not really. My kid participated in TIP back in 2013. In a class of 22, just one of the kids was Ashkenazi. About half of the participants were Chinese Americans; there also were Greek-American kids, Irish-Americans, Yugoslavian-Americans and few Western European-Americans – like one or two. Also, the entrance test was simply General SAT . You need to rethink your entire position. You might also want to ask yourself why there was just one Ashkenazi kid only or even do some meditation on the topic.

          • gcochran9 says:

            Earlier, the SMPY was about half Jewish, or so I was told.

          • gwern says:

            You’ll notice that your anecdote contradicts the aggregate statistics about TIP: you say half were Chinese Americans, while TIP’s SMPY equivalents for Makel et al 2016 were only a quarter ‘Asian’. Which is more likely, that they’re wrong, or a sample size of 22 (with a relevant subset of half that) has a lot of sampling error…? Plus, as Cochran says, all feedback I have gotten through informal channels from SMPY participants is that they are in fact heavily Jewish, and this is consistent with all other high-IQ samples I’ve found which break out or note ethnicity (Barbe, Terman, Hollingsworth, HCES/Subotnik). So yes, really.

            • gcochran9 says:

              Those SMPY numbers were some time ago.

            • gwern says:

              Which is fine, because Spain et al was not done using the latest batch of TIP kids.

              And if you thought that myostatin mutations were the main reason that big heavy cattle were big and heavy and why big cattle breeds were big, and you sequenced the heaviest cattle you could get your hands on, which you knew were, obviously, sampled massively disproportionately from the heaviest cattle breeds (which you believe have to all sorts of special mutations), and you didn’t find any mutations (much less myostatin specific mutations) of large effect, one would think that would be troubling for proponents of myostatin theories.

              • gcochran9 says:

                I don’t think that much of the variation in intelligence is due to rare highly-beneficial alleles. But I would be somewhat surprised if none existed.

              • RD says:

                Pretty sure GC never said main reason. The point of the paper was a couple rare alleles (0.5% to 2%) that could plausibly increase intelligence seem to be selected for in Jewish populations despite their brutal or lethal effects on homozygote fitness. You don’t get to an IQ of 110-112 because 5-10% of your population carries one of these….these are an overall minor effect at best and would still leave small effect variants doing must of the work. Or these could even be small effect variants…

                It’s also possible these help with memory (someone else mentioned), something else, or don’t otherwise help with 12yo SAT… Or their presence helps by a few points generally but hurts your chances of extreme 4sd IQ.

                “While we think that IQ is the most relevant outcome measure it may be that the historical process we have discussed selected for other traits that led to success and fitness in medieval times, so a broad spectrum of outcomes should be examined.”

                In other words, seems a little early for gotchas here.

              • gwern says:

                They were hardly the first to speculate on a Jewish IQ advantage from genetics: the entire meat was that there interpreted there to be evidence from rare variants which strongly support selection & genetic advantage. But, if they’re not there, they’re not there. Why should anyone be impressed by being (partially) right for the wrong reasons, when the wrong reasons were the novelty of the paper?

              • Rd says:

                Have you presented any evidence the theory is wrong, or just that one paper N~1000 smart 12-year olds, didn’t show hardly any hits, on anything. And was obviously underpowered to pick up alleles with 0.5% or lower frequency.

                Or is this all to celebrate maybe these are 1 IQ point and not 5?

                Or something else. It’s not quite clear what you’re so happily celebrating.

              • gwern says:

                I have listed 6 different categories of evidence all of which point to the non-existence of large beneficial rare variants anywhere much less in Ashkenazi specifically, and no one has yet given a single piece of positive evidence. If you want to make a G-of-the-gaps argument that ‘there is still a chance!’, feel free, but be honest about it. And misleading descriptions of power (that’s power for individual alleles, while power for any alleles is obviously much higher) are not that.

          • Lot says:

            “In a class of 22, just one of the kids was Ashkenazi.”

            Pure AJ are maybe 1% of Americans, possibly even less, born circa 2000. So 5x population over-representation. The mixed AJs are often hard to ID as such.

        • R says:

          This paper always seemed suspect wrt IQ. 1 in 3000 is an IQ of around 151. But these are SATs with a 0.85 correlation or so, so more like 143. And they’re 12 year olds. Intelligence then is not fully 100% correlated with ultimate intelligence at 18, so the IQ cutoff is even lower. Very unclear where they’re coming up with this fantastic “170” figure. That’s 1 in 650,000, on their claimed 100/15sd scale.

          Also Gwern seems very sloppy saying “most or all” Jews. A reasonable figure is maybe one third or one quarter based on every other stat of elite Jewish representation. Where is Gwern getting this idea no white people are smart… maybe from the fake “IQ 170” claim.

          • gwern says:

            12yo IQ is pretty stable. It’s earlier than that that there’s considerable instability. (This is why Hunter Elementary doesn’t work.) You’re also doing your math wrong. Selecting top 1 in 3000 != mean IQ of 151: that’s the minimum. The average will obviously be higher than the minimum threshold. It’s a truncated normal, the mean for a threshold of 1/3000 would be 155 given simple random sampling (which this is not). The correction of the SAT is discussed in more detail in

            And I did in fact work out the order statistics, and somewhere between half to nearly all is roughly what you get when you look at the mean IQs, consistent with what is usually anecdotally reported; and further, you recover the mean IQs for Asians & whites when you work back from the reported races of those as a sanity check.

            Not that any of this establishes anything other than that there was a good chance that any rare variants would turn up and they failed to do so, just as they have failed to do so in the other 5 approaches, and there remains zero direct evidence.

    • Pincher Martin says:

      I take it from this back-and-forth between Gwen and Greg that while recent GWAS correlation studies confirm Greg’s idea that Ashkenazi Jewish IQ is higher than the European White mean by about two-thirds of a standard deviation, they do NOT show any evidence of Greg’s hypothesis that this IQ gap is caused or helped by rare alleles that are also responsible for the genetic disorders in the AJ population.

      To summarize: Ashkenazi Jews appear to be smarter than other whites only because they have more, genetically speaking, of what makes those other whites smart and not because they have anything different. If a geneticist studying these gene-IQ correlations were to isolate a group of White Goys with above-average intelligence of around 110, then from the point of view of his study that group would look indistinguishable from a group of Ashkenazi Jews with average intelligence (110).

      Is that a fair summary?

      • gcochran9 says:

        the two 110 samples would not be identical: first, the Ashkenazi samples would be more influenced by genetics, less by environmental luck. second, the ability profile is different: Ashkenazi Jews do relatively better on verbal, relatively worse on spatial visualization.

      • Jeffrey says:

        “A prediction is that Gaucher, Tay-Sachs, and Niemann-Pick heterozygotes will have higher tested IQ than control groups, probably on the order of 5 points.
        We do have strong but indirect evidence”

        By ‘they were wrong’ Greg must have meant he was mostly wrong.

  6. Eponymous says:

    What studies are you referring to? I saw the N=50 one by Emil a few months back. Have there been others?

  7. Eponymous says:

    What studies do you have in mind? I saw the N=50 one by EOK a few months ago. Have there been others?

  8. Josh Gonik says:

    I’d like to see separate studies done on Israeli Ashkenazi Jews–and excluding all Israelis who identify as Ashkenazi but who have some other Jewish ancestry as well.

    It would be interesting to see how the figures for Israeli Ashkenazi Jews compare to those for US Ashkenazi Jews. I’m not ruling out the possibility that the US got a lot of the Ashkenazi Jewish cognitive elite.

  9. Boiling Off...? says:

    Hmm… As I’d see it the main point of “The Natural Selection of Ashkenazi Jewish Intelligence” that needs testing is whether this is actually due to selection.

    The alternative is not the absence of an AJ advantage in a “forced” niche, but the Maristella Botticini related theory that Rabbinic Jews opted into literacy for religious reasons, then only those that opted into high income jobs coul afford to maintain the tradition –

    A “boiling off” theory of sorts, possibly compounded in Europe by the particular bottleneck of AJs, then followed by massive growth of Eastern European population, with Eastern AJs showing even higher demographic growth as along for the ride with the rise of the Eastern European burgs and burghers. In this model, selection within the AJ population doesn’t really matter.

    I have my doubts about this of course; I’m not sure if Rabbinic Jews were more widely literate in the manner she suggests (the Bar-Mitzvah’s inclusion of literacy, for’ex, seems like a recent invention among European AJs), and this all suggests that the Middle Eastern Rabbinic Jews should show the same advantage as they follow the same religion, yet the Israeli experience doesn’t seem to be like that so much.

    • gcochran9 says:

      Selection in europe has to matter, because the Ashkenazi Jews (mostly European) effectively came into existence there.

      • Boiling Off...? says:

        Sure, but differential reproduction within the population which “came into existence” such that more intelligent individuals within the population had higher fitness and this changed that population phenotype over time, that is what you propose and that is what requires testing. It is not the case that “Higher PRS scores, job done”.

        • gcochran9 says:

          Pretty close. The original population was half Italian chicks, and I’d bet money that they weren’t picked for their SAT scores. The original average IQ can’t have been especially high. But it is now. What could have caused that change?

          If the extra IQ came from the Middle eastern fraction… come on, do the math.

          • gcochran9 says:

            Most or all of the selection to happen in Europe because the founding groups was at least half european. If the smarts came from the Middle East, Middle Eastern Jews would be smarter, less diluted, than the Ashkenazi. But it’s the other way round. Do I have to explain everything?

            While I’m at it, the whole paper was an obvious exercise in ag science/psychometrics. The fact that nobody ever even laid out the basic conditions for something like this to happen before us …

            • Boiling Off....? says:

              Greg, I’m not sure what’s up with your blank incomprehension here. The opposed hypothesis I am presenting is not differential fertilty happened in a single Middle Eastern Jewish, it is that, pace Botticini, all the Rabbinic populations today are are remnant of a bunch of different, admixed Jewish populations that all boiled off their less intelligent members due to a religious shift to (expensive) literacy and education. The admixed population with Hebrew males and Roman females forms before the religious change, but most of that group were sloughed off into the general Italian population, except the smartest, as the new religious norms form. Pace Botticini, same thing happening in Jewish populations in ME at same time.

              • gcochran9 says:

                The mix probably does not form before the religious change. From the latest genetic reports, looks as if the formation time is close, maybe simultaneous with Jewish settlement in the Rhineland: so they must not be descended from Roman Jews in Imperial times.

                Second, since Oriental Jews in Israel score 14 points lower than Ashkenazi Jews, we may not need a special explanation of their high IQ.

              • Boiling Off....? says:

                Significant uncertainties around admixture dates, and we all know the technical reasons why if we’re at the level of talking about this (Do we ever really trust ALDER to get it closer than the nearest half millennium? Even with much more diverged populations).

                But if we’re talking after the formation of the religion (Rabbinic Judaism, which is in the 6th century), we’re talking Ashkenazi/Sephardi as purely the legacy of 7th century admixture of fresh-off-the-boat Hebrew males and Italian females? What’s bringing them to Italy during the last downphase of the Roman Empire? (Some would say the “Fall of the Roman Empire”). Great opportunities for agricultural workers during the fall of the empire?

                If we have solid evidence of the un-special IQ of Jewish Middle Eastern populations, that secures that a correlated shift to a boiling event to a high IQ subset does not happen across different Rabbinical-izing populations (as their religion demanded literate education of all male participants).

                But then what do we make of Botticini’s claim (and evidence) that a correlated shift was going on in all these different Jewish communities into high financial reward, white collar, cognitively loaded occupations, and that this has nothing to do with being forced out of agriculture by the Roman church, and was intimately related to with the skills of literacy that their religion demanded, and the commercial value that these skills had? Did the ME Jews take part in this but “get” nothing for it cognitively? Is she wrong and this never happened (Rabbinic Judaism never had literacy requirements, the ME Jewish populations were not concentrated in high status/literate occupations)?

              • gcochran9 says:

                Shai Carmi’s group looked at this problem every which way: lots of simulations to check accuracy. Not just Alder. But it’s a complex problem. ADNA should tell us more.

                Looking at the sum total of the Jews in Israel that originated in the Islamic countries, their average IQ isn’t especially high. That we know. There may be regional exceptions.

                I’ve looked hard at “boiling events”: they require very unusual circumstances.

                Second: were the Jews of Islam concentrated in white-collar jobs? No.

  10. Lior says:

    Congratulations, were you also right about indo-european lactose tolerance?
    Iv’e heard that they had more alleles that relate to lactose tolerance.

  11. Dingbat says:


    I love science

  12. Abelard Lindsey says:

    Vox Day claims that the mean IQ of Ashkenazim Jews cannot be over 106. He has been very emphatic on this point.

      • Abelard Lindsey says:

        Vox Day likes to pontificate as one of the smartest guys in the room.

        BTW, he now thinks the moon landings were faked. I think he’s gone over the top.

    • Montmorillonite says:

      He also doesn’t believe in evolution. Or was it abiogenesis?

      Anyway, yeah, a refutation of Vox Day’s claims would be nice because there are many HBD dilettantes going around saying that the higher Ashkenazi IQ is a Jewish supremacist hoax because of him. The crux of his argument, iirc, is that the Israeli mean IQ is too low for a high Ashkenazi IQ to be possible.

      • Tbomb says:

        Pretty sure the best evidence for explaining Jewish dominance of the Ivy League and billionaire charts etc. remains squarely anchored in mental ability: High level math exams and chess remain as clean and open a measure as exists. You don’t have a talent at even 2 SD below the cutoff used here without noticing….

        That speaks to the US populations, Soviet populations, and worldwide-ish for Fields Medal and who really cares if there is a tad bit of adverse selection wrt Israel.

        This is a bit like claiming there’s a Kenyan supremacist hoax in marathon running…. Because of some random messy datapoint some internet troll can’t explain while ignoring all other evidence…

      • just a lurker says:

        He also doesn’t believe in evolution. Or was it abiogenesis?
        Both. And also vaccines, holocaust, moon landing, 9/11, climate change, theory of relativity and many other things.
        But he is not completely bad – he is 100% right about the liar, fraud and charlatan Jordan Peterson. It takes a master to recognize fellow master.

        • Rfg says:

          Yes, saying vaccines might have adverse consequences (and have been shown to have them in many studies on rats and humans), saying many Jews were killed out east or in transport or from famine and disease not necessarily gassing, distrusting the now-lost moon footage and materials and the guilt ridden press conference, and noticing those buildings in free fall is all on par with doubting the science supporting natural selection.

          As to climate change, of course there’s more CO2 from fossil fuel burning, and of course it causes warming, the question is whether other factors will trump that small delta and actually result in warming, cooling, nothing over what timeline. And whether “being green” is the solution or climate engineering or some smarter solution besides using a hair less fossil fuel that’s getting used anyway…

          I don’t really read the “man”, and suspect his wife is fake, a quite plausible conspiracy theory on its own, but are those meant to be insults to him?

      • Altitude Zero says:

        Vox is a smart guy, and has done some good things, but his recent stands on things like the moon landings and the Holocaust are a good example of where being a total contrarian can take you. Some things are true, even though liberals and the US Government say them

  13. Reisen says:

    “We don’t know that the populations were representative” – Below 100 for one, below 200 for another. Yes, very representative.

    “… but probably they weren’t far off. Scores correspond to an IQ of about 110. Personally, I’m still pulling for 112.” – Maybe from a select sample, like Helmuth Nyborg’s.

    “A fair number of people publicly disagreed with our Ashkenazi hypothesis.

    They were wrong.” Like Theodore Beale? Ron Unz? Curious to see what becomes of that.

  14. blasarius says:

    What do you make of this preprint making the rounds?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s