Narrowing the Genetic-Environmental uncertainty — if you wish

In a recent post I concluded:

So far everything Ron has presented is utterly dismissible.

That said, a strong case against global hereditarianism of even a modest amount can be made (forthcoming, perhaps). Ron just isn’t making it.

As for the first part, I think we, Ron excepting, can agree. Ron has offered a slew of specious arguments. Were it not obvious that he had little intent in crafting a logically sound, empirically grounded case –that he was intentionally engaging in rhetoric of a type befitting of Gould — we would be forced to question his analytic capacities. But Ron’s no buffoon, rhetoric was the game. Moreover, he was not entirely disingenuous. He is, no doubt, sincere in his belief that only a weak hereditarian position is tenable — and he probably sees his sophistry justified in light of this conviction. Nothing new here. As it is, he recognized the critical defect of the racial hereditarian case — certain migrant performance. And he astutely grasped the limits imposed by the phenomena of regression to the mean on selection-based hereditarian explanations for anomalous, from the evolutionary genetic perspective, migrant performance. His major failing was his Procrustean attempt to fit the migrant argument to the American scene. And because it doesn’t fit — for whatever reason — he has failed miserably, at least from a logical, empirical perspective. Though, perhaps the argument has been a rhetorical success amongst the quarters it was was directed to.

Now, some have argued that this issue of racial hereditarianism versus environmentalism is best resolved through allele counting. I disagree for the following reason: The magnitude of the plausible genotypic differences seriously under contention is relatively small (on average about 5 to 20 IQ points, equivalent to about 3 to 30% of the within population phenotype variance. And a significant portion of the heritability of IQ is non-additive (about 20%). And this non-additive portion is riddled with gene-gene interactions and so forth. As such, determining the underlying genetics of this portion of intelligence will be long in coming, as behavioral geneticists agree. So, even if it is found that the additive allele frequency differs between populations, it can always be argued, given the percent of possible genotypic variance under question, that the non-additive frequencies are such that the net difference is zero (or non-zero, depending on the results and one’s position). One will only be able to make a probabilistic argument for population differences based on allele differences alone (i.e., given the known additive differences, the chance that the non-additive differences…). So, unless I am missing something, gene-counting, of the type possible with the next two decades, would not suffice to convince the masses of determined environmentalism (or minority of determined hereditarians, again results depending).

Assuming that the above analysis is correct, the genetic-environmental uncertainty will not be eliminated by gene counting. Additional evidence will be needed. The most obvious such evidence is that which Ron has been discussing, migrant performance. Racial hereditarians themselves maintain that migrant performance is a strong test of their (and my) hypothesis, at least the evolutionary version. As noted prior, one of the main proponents of this hypothesis has stated flat out:

The evolutionary theory does however predict that when different races occupy approximately similar environments, such as for instance in the United States, Britain and the Netherlands, the intelligence differences will remain. This prediction has been examined in twenty three societies worldwide in Lynn (2008) and has been confirmed in every case. If a multiracial society is found where these race differences in intelligence are absent, the evolutionary and genetic theory of these differences would be falsified. Those who maintain that there are no genetic differences in intelligence between the races are urged to attempt this task. (Consistency of race differences in intelligence over millennia: A comment on Wicherts, Borsboom and Dolan).

Now, as with Ron, I have tried to decisively falsify the racial evolutionary hypothesis on the basis of migrant performance, but have, as yet, failed. My main problem has been the lack of good data and more critically, literally a chronic shortage of energy, as a result of my own own dysgenics.

Some, though, have recently offered to help with this endeavor. In light of the interest shown,
I propose, for those interested, a collaborative project, where we attempt to determine the correlation between National IQs, as indexed by international test performance, and second and third generate migrant IQs across nations. To this end, we can use the PISA 2000, 2003, 2006, and 2009 and other international or national data banks to make this determination. Previously, I showed a modest correlation between general migrant performance in Europe and National IQ, but I did not limit results to 2nd and 3rd generation immigrants — let alone to country of destination language speakers — nor did I adjust for selection. The latter is important because migrant performance on cognitive tests, not surprisingly, has been shown to be a function of immigrant selectivity. As for the latter, we can make rough adjustments using educational attainment figures or less rough adjustments using country of origin/destination cross comparisons — see, for example, here and here (note that the authors’ discussion in these papers is riddled with the sociologist fallacy)– for those countries which the data permits such. Here, for example, were unweighted PISA 2006 results for migrants to Portugal and Belgium broken down by region of origin and generation based on one math plausible value. Aggregating scores across surveys and supplementing with additional sources with increase the sample sizes. For example, the UK Pakistan sample is only n=26, but aggregating results across four studies will provide a reportable n=100. Aggregating results across countries will cancel out sampling bias, due to idiosyncratic factors (e.g., Chinese in the UK being extra motivated by policy X.)

To note, such a project would not be without worth outside the fringe HBD sphere. Richard Lynn’s claims, concerning immigrant performance, have not been completely ignored. Here, for example, is an excerpt from a just published paper co-authored by prominent intelligence researcher Roberto Colom:


O capital humano e as riquezas das nações

Although Lynn and Vanhanen (2002, 2006) maintained, correctly, that human intelligence is significantly influenced by genetic factors, and that, therefore, the national intelligence differences possibly can be explained, at least in part, by these genetic factors The reality of the Flynn effect shows that the intellectual capacity of the populations is sensitive to certain environmental factors or non-genetic…

Lynn (2006) sets out a number of evidences to be optimistic about the aim of improving the IQ of populations:

1. – A neglected nutrition can reduce a population’s IQ at least 15 points.
2. – The SSA have an average IQ of 67 when living in Africa, but 85 if residing in the United States.
3. – The SSA have an average IQ of 67 when living in Africa, but 86 when living in the UK.
4. – The SSA have an average IQ of 67 when living in Africa, but 85 if residing in the Netherlands.

I imagine that Colom would be interested in a more accurate assessment of comparative migrant performance.

As Ron has correctly pointed out, much of the reticence concerning discussions, in academia and out, of National IQ differences, which are of nontrivial importance, stems from concerns about possible genetic differences of non-trivial magnitudes. Such an analysis — which I am sure that Ron would be willing to cite (if the results favored his position) — could alleviate much of such concern (or alternatively that of those of us who take the contrary position).

Anyways, the analysis itself should be fairly simple. One can download the survey data online. And one merely needs to do a three layer mean comparison. Knock off versions of SPSS or SAS can also be downloaded online as can free trial versions which allow such simple computations. This analysis will most likely not be dispositive, as information for some migrant origin regions will be sparse (e.g., Africa), but it won’t be uninformative. At least we will get a sense about for which regions a genetic hypothesis, of a modest magnitude, is tenable. Again, I would do it myself, but I am shy of energy and to do it well — by the standards of dilettante online scientific racism — requires more than I have at the present.

If interested, drop a comment. And we can begin discussing if this is a worthwhile endeavor, and, if so, methodology and so on.

Advertisements
This entry was posted in Uncategorized. Bookmark the permalink.

37 Responses to Narrowing the Genetic-Environmental uncertainty — if you wish

  1. x says:

    OT a bit i know, but it’s interesting that you bring up Colom – I read a paper of his a while back on an experiment he did on the effects of working memory training (a fad spawned by that highly publicised / hyped Jaeggi paper a while back)

  2. x says:

    his conclusion was that working memory training doesn’t improve fluid intelligence – though the study’s methodology wasn’t perfect. attempts to replicate jaeggi’s original finding (if you were aware of it) have had mixed results. i tried doing the dual n back myself and didn’t notice any improvements after extensive ‘training’ on it over a few weeks.

    i think a meta-analysis came out of norway or sweden a while ago concluding that the WM training boosts gF thing is a false lead. maybe there is an effect in there but if there is i doubt it’ll be substantial – the search for methods of cognitive training that provide ‘transfer’ continue.

    wish i could help with the proposed project in post but i don’t know how to do stats. maybe i can help you guys find data if needed.

    • JL says:

      Two recent studies failed to reproduce the dual n-back effect.

      • Chuck says:

        “Here, for example, were PISA 2006 results for migrants to Portugal and Belgium broken down”

        Notice that neither the tentative Belgium nor Portugal results support an European-African evolutionary hypothesis….

        • gestur says:

          According to this paper:

          http://www.ceg.ul.pt/finisterra/numeros/2004-77/77_04.pdf

          “the level of school failure and drop-out is much higher among the students of the Angolan community – and the other five PALOP communities – than the average for the Portugese student community as a whole”

        • gestur says:

          The majority of Africans in Belgium are Moroccons, who I believe have Arab ancestry. Here is a paper on how they(and other immigrants) do in school:

          http://ugent.academia.edu/OrhanAgirdag/Papers/377056/Why_Does_the_Ethnic_and_Socio-economic_Composition_of_Schools_Influence_Math_Achievement_The_Role_of_Sense_of_Futility_and_Futility_Culture

          • Chuck says:

            I was specifically talking about the performance of sub-Saharan Africans. Click on the “Belgium” link, above. You will see that scores are dived by generation. And that African scores are dived into north African scores and into Sub-Saharan African score. I noted that the sample sizes are small and that I only looked at 1 out of 15, or so, probabilistic values. This one could, in this instance, be entirely uncharacteristic (though I doubt it since others have reported similar scores).

            Were you familiar with this blog, you would know that I don’t put much weight on singular scores, let alone singular “plausible values.” I am interested in meta-analytic scores, which is why, for every topic I cover (e.g., colorism, biracial performance, Hispanic third generation performance, Black UK performance, African national IQs), I attempt to replicate results across as many studies as I can and aggregate findings.

            As such, I am not asking you to take that one Belgium score as definitive evidence of anything. Rather, I am offering to help you and others conduct a comprehensive meta-analysis of migrant scores across the OECD — if you want. I recognize that local population scores may vary for idiosyncratic reasons: immigration laws in country X, yob performance in country y, and so on. In response to my Black UK posts, numerous such idiosyncratic explanations were lobbed at me. Refer here: http://occidentalascent.wordpress.com/2012/02/12/a-little-bit-of-difference-goes-a-long-way/ You yourself have been proposing your own. And I agree that these are not wholly implausible. Indeed, if an environmental hypothesis is true then IQ differences are clearly malleable; if so, then malleable differences don’t mean that a genetic hypothesis is false. Likewise, if two groups could not differ in genotype despite differencing in phenotype for unknown environmental reasons (environmental hypothesis in the US), then, it logically follows that two groups could differ in genetic potential despite not differing in phenotype for unknown environmental reasons (potential genetic hypothesis in country y). Environmentally inclined commentators such as Ambiguous have to accept this logic. And we reach an impasse, due to conflicting results. .

            Now I am simply offering a possible route out: aggregating findings across numerous countries. We will treat between country variance in the magnitude of the within country gaps as sampling error and look at the average tendencies. After, we can look at specific cases and see how plausible it is to attribute differences to “sampling error.”

      • JL says:

        What does “native” mean in those tables?

        • Chuck says:

          Presumably, “parents born in country” — since that’s the standard usage.

          1st gen = born outside country, parents born outside
          2nd gen =born inside, parents born outside
          3rd gen or native = parents born in country

          Here’s the student codebook:
          http://pisa2006.acer.edu.au/downloads/Codebook_Stu06_Dec07.pdf

          IMMIG (327) Immigration status
          1 Native
          2 Second-Generation
          3 First-Generation
          7 N/A
          8 Invalid
          9 Missing

          …..

          The groups, I should note, included mixed race kids — in the case of SS Africans, they were partially driving the high scores.

          Don’t make much of this singular result. I just pointed it out in hopes of perking your interest.

          • statsquatch says:

            OK, what weights to you use? There are country weights and student weights.

          • Chuck says:

            That might depend on our methodology.

            I would argue that we should average intra-national d’s. We would average the Turk-German, Turk-Spanish, and so on, d’s.– instead of averaging the Turk and non-Turk scores across the OECD and calculating an average d. Doing so avoids destination confounds (i.e., Turks having low Pisa scores relative to Indians because they tended to migrate to countries with poor school systems.) Of course, it creates others. For example, African might perform no worse than Portuguese (d=0), but if our Portuguese underperformed relative to the UK, then our averaged ‘true’ d might be underestimated.

            Methodology is something we might want to discuss before estimating scores, no?

            Anyways, If we were to concern ourselves with intra-national ds, we would, I think, use the student level weights as country level differences shouldn’t affect intra-country level ones. Emphasis on “I think.”

          • statsquatch says:

            Did you get my version of this analysis? I posted it as a comment yesterday.

  3. nooffensebut says:

    “One will only be able to make a probabilistic argument for population differences based on allele differences alone”

    I think you are underestimating the power of the probabilistic argument, and there is an implicit argument in the use of a non-additive defense of environmentalism. Imagine writing a book with a chapter for each phenotype being compared across populations. I am currently working on such a “chapter” (blog post) for obesity. If the reader agrees that the GWAS alleles constitute a representative sample of alleles that supports conclusions that also match sociological data and personal experience in the case of a series of less controversial phenotypes, the reader might be persuaded to accept such evidence in a late chapter of intelligence or another controversial phenotype. As you say, some might lean upon plausible deniability that a low level of explained variance would allow. However, the strictest category of environmentalism would be disproved by the adherent’s reliance upon population-specific genetic architectures or epistasis because it emphasizes racial differences to buttress sameness. We have already seen this in (poor) attempts to argue that MAOA and OXTR have different effects in different populations. There are other examples of better-supported population-specific gene effects, but strict environmentalists might not be aware of that, and having to make that argument for IQ would cause dissonance with the race-as-social-construct canard.

    • statsquatchs says:

      I agree with nooffense here. Genetic data is hard to refute since there is little chance of reverse causality and there are auxilary methods (e.g., “knock out” mice) to establish function. 60 years of environmentalism has given us what specific in root causes of low IQ? Lead poisoning (actually weaker than thought)? Iodized salt?

      Still, your endeavor seems worth while. What is the three level mean: immigrant, country of origin, and generation? And which of the five data sets do you need. They are huge…

      • Chuck says:

        “Lead poisoning (actually weaker than thought)? Iodized salt?”

        In developing countries IQ is affected by nutrition and cultural deprivation associated with rural living. Refer to the Colom paper cited above among others.

        You can see the latter effect in developed countries too. Heritability is depressed in less populous regions. Take a look at the TEDs map:
        http://sgdp.iop.kcl.ac.uk/davis/teds/geocoding/

        This doesn’t help with the gaps in the US, as h^2 is about the same across populations and so on — but it’s plausible that h^2 is lower in other regions, allowing for more e^2 malleability — and of course that the IQ scores are not measure invariant — that’s a real bugger there.

        Generally, I think HBDers might be in for a rude awakening, when it comes to racial differences. The issue, IMO, will turn of the effect of immigrant selectivity — as many immigrant populations across Europe and elsewhere are not under performing to the degree they “should,” when at all (e.g., South Asians, South East Asians, and Black Africans).

        I’m digging though the research on selectivity and IQ now.

        ……
        No, it’s really simple:

        Just download:

        SAS syntax to read in student questionnaire data file
        Student questionnaire data file

        For 2000, 2003, 2006, 2009

        (TIMSS and PIRLS don’t have origin data — though you can infer it from the US set using the race variables. )

        (1) Split file by country
        (2) Do a simple 3 layer mean comparison:

        a.Mother country of birth
        b.Father country of birth
        c.Nativity

        (Codebook for student questionnaire data file)

        Right?

        The difficulty is working with the plausible values. As noted before there are 5 for each subject area. For what I have read you can get away with randomly picking one.

        Refer back here: https://occidentalascent.wordpress.com/2012/04/15/data-please/#comment-1643

        • JL says:

          Generally, I think HBDers might be in for a rude awakening, when it comes to racial differences. The issue, IMO, will turn of the effect of immigrant selectivity — as many immigrant populations across Europe and elsewhere are not under performing to the degree they “should,” when at all (e.g., South Asians, South East Asians, and Black Africans).

          Well, here’s results from the 2009 PISA math test for some Western countries. The first column is the native population (i.e., mostly whites) while the second column is 2nd generation immigrants (regardless of national/racial origin):

          Australia 513 532
          Austria 507 450
          Belgium 529 459
          Canada 531 519
          Denmark 510 447
          France 507 443
          Germany 527 469
          Italy 487 450
          Netherlands 534 477
          Norway 502 463
          Portugal 490 450
          Spain 491 456
          Sweden 504 447
          Switzerland 550 494
          United Kingdom 497 486

          The results are about what you’d expect based on HBD principles. 2nd gen immigrants in countries with substantial black and brown immigrant populations generally score clearly below the natives, while the largely Asian immigrant populations of Australia and Canada are doing better. If we assume that 2nd generation white immigrants score similarly to native whites while East Asians and some South Asian groups are doing well, too, then the lower mean scores of 2nd gen immigrants can only be explained by poor performance by Africans, Middle Easterners, Latin Americans, and/or some Asian groups.

          So at least on first approximation HBD is not falsified. Of course, it’s possible that when you get down to the details, the patterns will turn out to be quite different.

          • Chuck says:

            “So at least on first approximation HBD is not falsified. Of course, it’s possible that when you get down to the details, the patterns will turn out to be quite different.”

            devil(‘)s in the detail.

            Let’s try correlating National IQs with 2nd+ gen National migrant performance.

            I’ve been under the weather of late, so won’t be commenting on this much in the near term.

            JL, Stats, and all

            (1) One important issue is selectivity for IQ. We need a fair method of estimating that. Unfortunately, I haven’t been able to find any IQ score for emigrating parents/adults. That is, IQ scores of soon to be emigrating adults compared to non-emigrating adults, based on country of origin tests. There’s only data on immigrating adults — and these scores are, naturally, to one degree or another, confounded by cultural/linguistic bias.

            There is, however, excellent data on migrant educational attainment (EA), measured in terms of years of education::
            http://perso.uclouvain.be/frederic.docquier/filePDF/DM_ozdenschiff.pdf

            I’m not sure, though, how to turn it into an EA quotient normed on the non-emigrating population mean. (I’ve devised a number of conflicting methods.)

            Could you take a look at Appendix A.1.1 and see if you could come up with a sound methodology. For example at the bottom of the second page of A1.1 numbers are given for the Congo:

            Skill level by Working age Pop
            Low 11303000
            Medium 1267000
            High 101000
            Average 12671000

            Skill level by Emigrants

            Emigrant
            Low 27788
            Medium 20235
            High 26901

            Average 74924

            How could we estimate the mean EA difference in standard deviations between emigrating and non-emigrating Congolese?

            (2) On that note, do either of you know of a good meta-analytic estimate of the correlation between EA measured in terms of years in school and IQ? My 5 minute Google search turned up conflicting results. Often, EA was indexed by GPA or some other index.

            (3) Another issue is assortative mating. What’s the typical correlation found between spousal IQs. Based on the references listed in “Spousal concordance in academic achievements and IQ: a principal component analysis” I am getting 0.3. Does that sound about right?

            I ask because in some instances we might only have EA /IQ estimates for only one immigrating/ emigrating parent — and we might have to estimate the mid parent IQ from this.

            (4) Another issue is the relation between EA and IQ for emigrants. Let’s assume that the correlation between IQ and years of schooling is 0.6. And that a particular emigrating population is 1 SD units in EA above the home population. If the emigrants were selected for EA alone, then this would entail that they were 0.6 SD above in IQ. But we don’t always know what they are directly being selected for. Theoretically, in some instances, they could be selected for IQ (emigrating as an IQ test) and this IQ selection could be driving the EA selection. Were this the case, an emigrating population that was 1 SD above the home in EA, would be 1.7 SD above in IQ!

            The difference between 0.6 SD (EA–> IQ) and 1.7 SD (IQ –>EA) is to large to tolerate. We need a better estimate of the relation between EA and IQ in migrating populations.

            The best I could find so far was Borjas (1992). He looked at the relation between AFQT scores and year of education, among other things, among US internal (state to state) migrants. Among this group, IQ and EA was correlated at 0.5 and the IQ, EA differences relative to the non-migrating populations (“stayers”) was commensurate; “Movers” were selected 0.2 SD in EA and 0.2 SD in IQ.

            This suggests that IQ and EA are, to some extent, being independently selected for. And that the magnitude of selection is about the same. If so EA, would index IQ.

            So, is anyone else here aware of any other papers with give both (internal, if not external) )migrant IQ and EA scores? I looked though and couldn’t find anything — though I was able to replicate the Borjas results using GSS and the “Mobile16” variable. I might have to check the NLSY 97 survey.

            This seem to be a non-trivially important issue when it comes to adjusting for migrant selection.

            (5) Another issue is the relation between migrant IQ and immigration:
            http://occidentalascent.wordpress.com/2012/02/07/partially-falsified/#comment-1313

            If the act of migration is an IQ test, the IQ of migrants might vary as a function National IQs independent of EA.

            There will be “disparate impact” such that more people from low IQ countries than high IQ countries who want and try to emigrate will fail the test.

            I’m not sure how to evaluate the plausibility of this possibility or how to factor it into the selection estimates —

            If anyone has any ideas, let me know.

            Thanks.

          • Chuck says:

            “Of course, it’s possible that when you get down to the details, the patterns will turn out to be quite different.”

            The detail would be to look at scores by country of origin. And to look at 3rd gen scores.

            Refer to the references cited here:

            http://occidentalascent.wordpress.com/2012/08/21/do-national-iqs-predict-immigrant-performance-across-europe/

            Also, I’m seeing a lot of skimpy differences. What’s the average within OECD country SD? Let’s assume 85 as a minimal (for reference the within US SD is 70)::

            Australia 0.22
            Austria 0.67
            Belgium 0.82
            Canada 0.14
            Denmark .9
            France 0.75
            Germany 0.68
            Italy 0.44
            Netherlands 0.67
            Norway 0.46
            Portugal 0.47
            Spain 0.41
            Sweden 0.67
            Switzerland 0.66
            United Kingdom 0.13

            Average 0.54

            Not exactly going to sink Multiculturaldumb

  4. Anonymous says:

    What is the three level mean?

  5. Ambiguous says:

    “Now, as with Ron, I have tried to decisively falsify the racial evolutionary hypothesis on the basis of migrant performance, but have, as yet, failed.”

    I wouldn’t say you’ve failed, you’ve assembled some very challenging evidence for the global hereditarian view. For example Lynn estimates a “genotypic IQ” of Africans of 80, in light of the UK data especially that is not tenable anymore. UK Black-White seems to be in region of around 0.5 SD. So even if Blacks aren’t yet as intelligent as Whites they are not nearly as dull as they’ve been made out to be. You’ve also shown data that shows representative samples of South Asians doing much better than Lynn’s claim of a genotypic IQ of 90, even poor ones like Pakistanis and Bangladeshis. My own analysis of Thailand shows it to have an IQ of 97, way higher than the score claimed by Lynn.

    “One can download the survey data online. And one merely needs to do a three layer mean comparison. Knock off versions of SPSS or SAS can also be downloaded online as can free trial versions which allow such simple computations.”

    I’ll give it a go when I can, and I’ll keep looking for new data.

    Chuck is there anyone else in the hbd community doing what you are doing?

    • Chuck says:

      The bugger is immigrant selection. Read through the discussion in the comment section here:
      http://occidentalascent.wordpress.com/2010/02/10/super-duper-selection/

      It’s really hard to find results that are not possibly confounded by such selection (and therefore sampling bias). But (re)read:

      http://occidentalascent.wordpress.com/2012/04/01/holland-white-black-gap/

      • Ambiguous says:

        UK Black Africans are significantly more selected than Black Caribbeans were (most of who’s children are 3rd generation now) but they get a similar IQ. I just don’t think immigrant selection can explain all that much especially when you factor in not 1 but 2 regressions to the mean in the case of most Black Caribbean children.

        Pakistanis and Bangladeshis were a bunch of rural peasants for the most part, especially the latter and they aren’t doing too badly at all. Most Bangladeshi mothers had (and have) no qualifications, they just got in on family reunification or spousal visas. As for Indians they were mostly just regular people, not the “cognitive elites” that went to America.

        But yeah it is hard to get actual numbers on immigrant selection. One way might be to look at the type of visa granted. Most ancestors of South Asians and Black Caribbeans in the UK arrived when immigration to the UK was fairly easy.

        • Chuck says:

          (1) Regression to the mean (RM) occurs because parents pass down mostly only additive genetic differences and because additive genetic differences account for only 60% of the IQ variance (by adulthood). (The magnitude of regression should vary by age, since heritability does.) Parents don’t pass down non-additive genetic differences which accounts for maybe 10% of the variance by adulthood nor do they pass down unshared environmental variance, which is due mostly to random environmental effects (and typical accounts for 20% of the variance). They might, though, pass down some shared environmental effects (maybe 10% by adulthood)– and this varies with age.

          From a between populiction perspective, there would only be one genetic regression. Were emigrants (“movers”) selected 1 phenotype SD relative to non-emigrants (“stayers”), and if the narrow heritability in the full population was 0.6, it follows that the movers, relative to stayers, would be selected for by 0.6 SD in additive genetic differences. This difference would be passed on throughout the subsequent generations..

          (2) Black Caribbeans in the UK are, in fact, a selected population relative to West Indians in the Caribbean.

          http://occidentalascent.wordpress.com/2012/02/07/partially-falsified/#comment-1308

          It’s plausible that Caribbean movers were selected 1 phenotype SD relative to stayers (i.e., that on average they hailed from the upper third of the IQ distribution).

          I’m open to being convinced otherwise, though.

          http://unpan1.un.org/intradoc/groups/public/documents/apcity/unpan022366.pdf

          Granting this, It’s not implausible that their offspring were genetically selected by 0.6 or so SD.

          But this whole immigrant selection Idea really needs to be thought over. I’m open to ideas.

          (3) African immigrant were indeed more selected for than Caribbeanians and so should have higher IQs. I agree that this is just one of the many bizarre inconsistencies that riddle the immigrant selection hypothesis. (More problematic is that East Asians are about as highly immigrant selected as Africans and yet the former’s IQs are not off the charts). Somehow, we are supposed to believe that Africans perform well above their national IQs because they are super duper selected for educational attainment. And, at the same time, that East Asians, despite being super duper selected for the same, perform at their National IQs because …?

          I tried playing around with some ideas here:
          http://occidentalascent.wordpress.com/2011/12/09/2nd-gen-black-success/

          …but no one seemed interested, so I dropped the idea — despite “immigration selection” most probably being much more important — in the age of global immigration — than race.

          (4) South Asians seemed to perform well all across Europe based on my cursory inspection of the evidence. So, if a South Asian -Europe genetic hypothesis is to hold, the explanation must be with country of origin factors — factor in South Asian that lead more brighter ones to be selected for, irregardless of the destination countries’ immigration policies. I came up with a possible queer theory for this. Maybe you have some thoughts on it?

          http://occidentalascent.wordpress.com/2012/02/07/partially-falsified/#comment-1313

          (5) I think the way to go is to use the readily available educational attainment data. Refer to my comment above. 9:14 pm

          …..

          “I wouldn’t say you’ve failed, you’ve assembled some very challenging evidence for the global hereditarian view.”

          An African genotypic IQ of 80 is untenable. I’m not sure why Lynn argues that. Jensen, who is a far superior researcher, has voluminously argued that the US Black-White gap of I SD could be no more than 50-75% genetic. And he is correct on this point. Adjusting for admixture, this would put the African-European gap at about 0.7 to 1 SD (mean: 0.85). So 1 SD is the maximum difference under contention. (This difference is small but, surprisingly to some, small IQ differences amplify on the population level to produce large effect (like Detroit).

          http://lesacreduprintemps19.files.wordpress.com/2011/07/gordon-1997-everyday-life-as-an-intelligence-test-effects-of-intelligence-and-intelligence-context.pdf

          .

          • Ambiguous says:

            “(2) Black Caribbeans in the UK are, in fact, a selected population relative to West Indians in the Caribbean.”

            There may have been some selection but I didn’t see any data specifically on Black Caribbeans who migrated to the UK between the 1940s and 1960s which is when the majority of the ancestors of UK Black Caribbeans arrived. Immigrant data for 1990 and 2000 is of limited usefulness in that regard. I had a look at the Suzanne Model link and came up blank regarding data for UK Caribbean immigrants.

            I had a look at some of those tables on immigrant data and it seems that the US, Canada and Australia take up over 70% of all tertiary educated immigrants. Unfortunately it doesn’t actually give data by country on how educated the immigrants entering each country are, though it’s clear that immigrants to Europe are much less educated than immigrants to the 3 main destinations for educated migrants. The US alone takes half of them. This isn’t anything we don’t already know, some countries are a lot harder to immigrate to than others. As above, this isn’t much use for estimating selection on those particular migrants who arrived in the UK between the 1940s and 1970s. Until 1962 all Commonwealth citizens could migrate to the UK without restriction. There still may have been some selection though, it did cost money to migrate after all.

            ” Somehow, we are supposed to believe that Africans perform well above their national IQs because they are super duper selected for educational attainment. And, at the same time, that East Asians, despite being super duper selected for the same, perform at their National IQs because …? ”

            Yeah I really doubt selection plays anything like as big of a role as some people make out it does.

            “I came up with a possible queer theory for this. Maybe you have some thoughts on it?”

            Are you applying this to Indians in the US or the UK? The former are indeed highly selected, as are most immigrants to the US from outside Latin America. I don’t have any decent IQ data specifically on US Indians though. The 112 score for Indian immigrant children was just based on reverse digit span.

          • Chuck says:

            “There may have been some selection but I didn’t see any data specifically on Black Caribbeans who migrated to the UK between the 1940s and 1960s”

            I’ll have to look into the UK selectivity issue some more. Early Caribbean immigrant to the UK are generally said to have been unskilled by UK standards, but I have been unable to find good info on the average skill level of Blacks at that time in the West Indies. Selection depends on the emigrant skill level relative to the non-emigrant skill level.

            “Unfortunately it doesn’t actually give data by country on how educated the immigrants entering each country are, though it’s clear that immigrants to Europe are much less educated than immigrants to the 3 main destinations for educated migrants. The US alone takes half of them.”

            It’s not clear, however, if immigrants to Europe are less selected, since selection is based, as noted above, on the relative performance of emigrants to non-emigrant. Imagine this situation: immigrant to the US hail from countries in which the average years of education is 14. The emigrants have, on average, 16 years of education; immigrants to Western Europe hail from countries in which the average years of education is 12. The emigrants have, on average, 15 years of education. Now, in this scenario, the European immigrant will be more selected despite having lower average levels of education.

            If you can dig up data on percents of immigrant to the US and Europe by region of origin, we could determine this.

            “This isn’t anything we don’t already know, some countries are a lot harder to immigrate to than others.”

            It’s even more complex than this. For example, immigrants will be less selected, controlling for immigration laws, if there is a large immigrant community. So Hispanics in the US are now less selected, because they have connections.

            Because of the various destination effect, I was suggesting in the post that we average scores across countries. And then of course look at anomalous situations in particular, such as the UK.. This would be quite a bit easier than conducting a detailed examination of the immigration patterns for each country. I try not to get lost in particulars, at the present, because I have been entertaining a more complex model of genetic differences:

            http://occidentalascent.wordpress.com/2012/02/12/a-little-bit-of-difference-goes-a-long-way/

            I think the Flynn/Dickens model is not inherently implausible. And the model is widely cited — mostly for ascientific reasons — elsewhere I noted:

            “Environmentalists, naturally, have seized on the active rGE explanation. Accordingly, genetic differences in intelligence result from cognitive shaping environments, environments which individuals select on the basis of their genetically predisposed predilections [5]. By this model, the heritability of intelligence increases with age because with age individuals become freer to follow their dispositions and select their environments. Quasi-environmentalists are not the only ones that find this model appealing. For example, in the discussion section of their ground breaking study of 11,000 twin pairs across four counties, Haworth et al (2009) state…:”

            If the model is correct, it would imply that group differences could be locally malleable. This possibility, of course, rather complicates the debate.

            “Yeah I really doubt selection plays anything like as big of a role as some people make out it does. ”

            Agreed, but it may play a more important role than you think. And it probably plays a more important role that race.

            Refer below.

            “Are you applying this to Indians in the US or the UK?”

            Both. I have a double selection model in mind.

            (a) selection due to relative emigrant non-emigrant differences.

            Example:

            Assume Low= -1SD, Medium = 0 SD, High = 1 SD

            Congo

            Numbers by skill level for Natives
            Low 11303000
            Medium 1267000
            High 101000

            Numbers by skill level for emigrants
            Low 27788
            Medium 20235
            High 26901

            Estimated difference in SD = 0.88

            China

            Numbers by skill level for Natives
            Low 371500000
            Medium 200917000
            High 11682000

            Numbers by skill level for emigrants
            Low 343898
            Medium 208825
            High 371249

            Estimated difference in SD = 0.65

            So in educational attainment (EA) Congolese and Chinese emigrants are, respectively, 0.88 and 0.65 SD selected

            As for the relation between EA and IQ refer to (4) in my
            September 4 9:14 pm comment.

            (b) Selection due to the cognitive complexity of emigrating .

            I have in mind here the research on adverse impact. If we treat emigration as an IQ test, then different people will have different rates of passage for a given selection rate.

            Let’s assume that “National IQs” are somewhat accurate and that the Global average IQ is 90 (as said). Now let’s assume that on average for every 100 people that want to emigrate only 85% can make it on average on account of general intelligence differences. Our selection ratio would be 85%.

            So on average global immigrants would be selected 0.19 SD, net of other factors — (the lower 15% won’t make it and so the mean of emigrants will be at the 58th percentile. (If I’m thinking about this correctly.)

            Now since countries differ in National IQ, they will be differentially selected.

            National IQ/Selection in SD
            105/.03
            100/.08
            95/ 0.14
            90/ 0.19
            85/0.24
            80/0.3
            75/0.36
            70/0.41

            So…

            The national IQ of Congo is 75 (my estimate) and the national IQ of China is 105 (Lynn’s estimate :O) ), so the former will be selected an extra 0.36 SD and the latter an extra 0.03 SD.

            Taking both together — assuming IQ differences are commensurate with EA differences as discussed in (4), we get:

            Congo selection: 1.24 SD
            China: 0.67 SD.

            Factor in regression to the mean — and the offspring of the respective emigrants, on the account of selection, should perform, in aggregate, 0.74 and 0.41 SD above the UK mean.

            Chinese seem to perform maybe 0.3-0.5 SD above the mean in the UK, US, Canada, Australia, Italy, and the Netherlands and Congolese perform maybe 0 SD above the mean (in Belgium) — just guessing.

            Now we can then backwards calculate the genotype IQ of Native Chinese and Congolese by subtracting the found performance advantage form the predicted one.

            China = 0.41 – 0.4 = IQ of 100
            Congo = 0.74 – 0 = IQ of 89.
            ..
            This is just an example of a possible method. Looking at all countries it might generate outlandish values and have to be rejected. But to do that, we need: (1) better national IQ data for some regions (e.g., South Asia and the Caribbean); (2) better estimates of migrant performance; (3) a better understanding of the relation between EA and IQ; some way to determine the selectivity of the act of emigrating — and so one.

          • Ambiguous says:

            Immigrant selection, this might be useful:

            African men are indeed highly selected compared to everyone else. Bear in mind some of these parents are 2nd generation themselves who would have higher educational attainment than their own parents had.

  6. x says:

    didn’t susanne model have data on african-carib selection in the uk?

  7. statsquatch says:

    I need to crawl before I walk with this data. Here is the mean (weighted) by country and immigration status for pv1math. The sample sizes are low for the immigrant question and the STDs are naive, you need to use one of their macros to get right one. Thus will not be a simple exercise. Also, did I miss the Race variable? I see none in student.

    Country code Immigration
    3-character status N Obs Mean Std Dev
    ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ
    Belgium Native 7656 514.3926677 103.6797262

    Second-Generation 518 433.3567676 105.2282304

    First-Generation 569 411.3843152 109.7322838

    Germany Native 3946 509.8864549 103.4059767

    Second-Generation 355 427.5933307 116.8786753

    First-Generation 302 436.8428932 129.3104831

    Denmark Native 4145 500.6887281 85.4523917

    Second-Generation 190 434.8149469 97.9876328

    First-Generation 158 421.6867803 96.6536636

    France Native 3988 494.6093037 101.1519652

    Second-Generation 436 458.4169289 105.3450446

    First-Generation 151 451.3396841 108.8991870

    United Kingdom Native 12131 499.4980190 100.0148299

    Second-Generation 341 495.0108673 99.3928719

    First-Generation 279 455.4382295 112.7880011
    ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ

    • Chuck says:

      Thanks. There are no race variables in the international file. Just country of origin ones. They’re right before the plausible values. See page 74 here: http://pisa2006.acer.edu.au/downloads/Codebook_Stu06_Dec07.pdf

      As noted for most countries this produces very small samples sizes, so I suggested aggregating results across countries as other have. See, for example (Pisa 2003, 2006, 2009):

      (a) Dronkers and de Heus. The Educational Performance of Children
      of Immigrants in Sixteen OECD Countries

      (b) Levels et al. Immigrant Children’s Educational Achievement in Western Countries: Origin, Destination, and Community Effects on Mathematical Performance.

      (c) Kornder and Dronkers. Do Migrant Girls Always Perform Better?

      …And then aggregating scores across data sets as has not been done.

      I mostly wanted scores for 2nd+ generation kids by Origin. The papers above included first gen. And this inflated the d.

      Why not try just Africans in Belgium and Portugal as the samples sizes are reportable n=100 or so.


      Can’t you use the within country SDs? If anything I can get them from international explorer.

      • statsquatch says:

        I can get the SEs from the Macro. To do the immigrant thing I think you need a multilevel model to aggregate scores. They have a macro for that too. May take a while though…Ron called me out…so I have to reply. Speaking Ron, any comparison between the US and Mexican PISA scores? You could control for parental SES?

        • Chuck says:

          (1) I should have noted that you can readily and easily get scores by nativity using NAEP’s international explorer (at least for 2006 and 2009). It’s because the 2nd gen students performed so well — compared to my expectations — that I though it important to break scores down by country of origin.

          (2) Why can’t you aggregate scores by hand? Just average either the ds or the n-weighted scores. This is what Dronkers et alia did. They just list the scores of immigrants by origin by country of destination and sum the scores across the countries. Take a look at the table 1.

          https://lesacreduprintemps19.files.wordpress.com/2012/08/the-educational-performance-of-children-of-immigrants-in-sixteen-oecd-countries.pdf

          (3) I’m not in agreement with you on the point about the MCV. It seems to work ok when it’s combined with the power of meta-analysis. Refer to te Nijenhius’s comments in the last paragraph:

          http://analyseeconomique.files.wordpress.com/2012/08/the-flynn-effect-group-differences-and-g-loadings.pdf

          Or:

          “The fact that our meta-analytical value of r=−1.06 is virtually identical to the theoretically expected correlation between g and d of −1.00 holds some promise that a psychometric meta-analysis of studies using MCV is a powerful way of reducing some of the limitations of MCV…Additional meta-analyses of studies employing MCV are necessary to establish the validity of the combination of MCV and psychometric meta-analysis. Most likely, many would agree that a high positive meta-analytical correlation between measures of g and measures of another construct implies that g plays a major role, and that a meta-analytical correlation of −1.00 implies that g plays no role. However, it is not clear what value of the meta-analytical correlation to expect from MCV when g plays only a modest role.” (Score gains on g-loaded tests: No g, 2007)”

          I’m not saying that it’s great but it’s easy to do and not uninformative when using meta-analysis.

          MFGA might be gold standard but it has a number of restrictive requirements. Elsewise it would be done all the time.

          Here was Wai and Putallaz’s discussion:

          “This demonstrates that both the ACT and EXPLORE are not factorially invariant with respect to cohort which aligns with the findings of Wicherts et al. (2004) investigating multiple samples from the general ability distribution. Following Wicherts et al. (2004, p. 529), “This implies that the gains in intelligence test scores are not simply manifestations of increases in the constructs that the tests purport to measure (i.e., the common factors).” In other words, gains may still be due to g in part but due to the lack of full measurement invariance, exact estimates of changes in the g distribution depend heavily on complex partial measurement invariance assumptions that are difficult to test. Overall the EXPLORE showed stronger evidence of potential g gains than did the ACT.”

          With MGFA it’s difficult to determine the magnitude of g differences or determine definitively if they exist. MGFA is better at showing if scores are MI. But the absence of MI does not imply the presence of g as differences could, in principle, be due to lower stratum factors.

          (4) “Speaking Ron, any comparison between the US and Mexican PISA scores? You could control for parental SES?”

          I did these comparisons looking at generation and using International explorer. I also looked at differences by SES level and found that the differences were no less (and often greater) at the upper SES levels. This, of course, isn’t the same as controlling. But then we run into the sociologist fallacy with that — anyways, there’s so much research on this that it’s easier to just look it up. — I’m sure that I’ve commented on it before. Generally controlling nearly eliminates the W-H (at least at younger ages)
          gap:http://www.economics.harvard.edu/faculty/fryer/files/Fryer_Racial_Inequality.pdf

          Or do you mean that by looking at Mexican SES and scores I could control for immigrant Mexican SES. How? wouldn’t you need to know the SES specifically of emigrants?

          (5) Anyways, I tried to find data to replicate Jensen’s study that Ron cited. But I’m aware of none. Jensen mistakenly that that PPVT and cultural loaded tests had lower heritabilities than cultural reduced tests — but this isn’t necessarily true, for whatever reason as discussed in my Ron rebuttal.

          What one would need to do is to correlate H/W differences on tests or test items for which the heritabilities are known, I tried this with PIAT and AFQT in the NLSY only to discover after reading though the lit that heritabilities are not to different. Whatever, this is a worthless test because environmentalists just dismiss positive results.

  8. statsquatch says:

    My problem with MCA is that when you correlate pairs of items from factor analysis the different pairs of data are no longer necessarily independent from each other. I don’t know how to interpret the correlation statistics in that case. You are better off with a complicated factor analysis model.

    You can get the SES of the immigrants from the PISA data. They have the parents profession. You could show, theoretically, that a brick layers son in the US does as well as teachers son in Mexico.

    I need to set this down for awhile though.

    • Chuck says:

      “You can get the SES of the immigrants from the PISA data. They have the parents profession. You could show”

      This assumes that the parents’ profession in the destination country matches that in the origin country. How robust of an assumption do you think that is?

      “I need to set this down for awhile though.”

      Take your time.

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s