Review of the debate
In making his case against Richard Lynn’s racionational hereditarian hypothesis, Ron Unz, in his much cited Race, IQ, and Wealth article, pointed to the lack of correspondence between the IQ scores of US and European Whites of the same ancestry. He reasoned that were the racionational genetic hypothesis true, then the rank order of ethnic groups in IQ should be similar in the US as in Europe and across time. Quote:
In order to minimize these extraneous factors, let us restrict our initial examination to the 60-odd IQ datapoints Lynn and Vanhanen obtained from European countries and their overseas offshoots over the last half-century. Obviously, some of these countries have at times been far poorer than others, but almost none have suffered the extreme poverty found in much of the Third World…..Similar sharp differences occur in the case of Italian populations separated historically and geographically. Today, Italian-Americans are very close to the national white average in income and education, and the limited data we have seem to put their IQ close to this average as well. This would appear consistent with the IQ figures reported for Italy by Lynn and Vanhanen, which are based on large samples and come in at just above 100. However, there is a notoriously wide economic gap between northern Italy and the south, including Sicily. The overwhelming majority of Italian-Americans trace their ancestry to the latter, quite impoverished regions, and in 2010 Lynn reported new research indicating that the present-day IQ of Italians living in those areas was as low as 89, a figure that places them almost a full standard deviation below either their Northern Italian compatriots or their separated American cousins. Although Lynn attributed this large deficit in Southern Italian IQ to substantial North African or Near Eastern genetic admixture, poverty and cultural deprivation seem more likely explanations.
Ron was correct concerning the claimed regional IQ scores of Southern Italians and concerning Dr. Lynn’s racial hypothesis for the north-south Italian difference. Indeed, Lynn and associates have defended and elaborated on this position in a number of recent articles, including:
Lynn, R. (2010). In Italy, north–south differences in IQ predict differences in income, education, infant mortality, stature, and literacy. ; Lynn, R. (2010). IQ differences between the north and south of Italy: A reply to Beraldo and Cornoldi, Belacchi, Giofre, Martini, and Tressoldi; Lynn, R. (2012). IQs in Italy are higher in the north: A reply to Felice and Giugliano; Rindermann, H., Woodley, M. A., & Stratford, J. (2012). Haplogroups as evolutionary markers of cognitive ability; Templer, D. I. (2012). Biological correlates of northern–southern Italy differences in IQ.
In the last three years, across five separate papers, hereditarian leaning researchers have brought forth evidence of a north-south regional Italian ability difference based on half a dozen of sources (national tests, international tests, IQ tests, Piagetian tests, and literacy tests) and have shown that these difference correlate with biological and genetic markers, including haplogroups, hair color, eye color, skin color, head size, and rates of neurological diseases.
To quote Lynn (2012):
The data reported in this paper provide further evidence confirming the thesis that there is a north – south gradient for IQ in Italy and that this is associated with the frequencies of genetic markers for the percentages of European and North African ancestry in the populations. The thesis that average IQs are higher in the northern than in the southern regions is now supported by seven sets of data. There are (1) the studies by Peluffo (1962, 1964, 1967) using Piagetian tests; (2) the results of several studies using the Progressive Matrices reported in Lynn (2010b); (3) the reading comprehension, mathematics and science performance of 15 year olds in the 2006 PISA study for 12 regions, for which the average scores are correlated with latitude at .97; (4) the reading comprehension, mathematics and science performance of 15 year olds in the 2009 PISA study for 20 regions, for which the average scores are correlated with latitude at .93; (5) the MT-Advanced results for math and reading abilities reported by Cornoldi et al. (2010); (6) the INVALSI results for math and reading abilities reported by Cornoldi et al. (2010); (7) the newer extensive INVALSI results for math and reading abilities reported in this paper. The data given in Lynn (2010a) for the north – south gradient of the frequencies of the percentages of European ancestry and North African ancestry in the populations, such that in northern regions the frequencies of genetic markers for the percentages of European ancestry are higher, while frequencies of genetic markers for the percentages of North African ancestry are lower, are corroboratated by three further data sets. These are, first, the high correlation between latidude and the percentage of blonds (.81) as a marker for the percentage of northern European ancestry and showing greater percentages of blonds in the north of Italy. Second, the high correlation between latitude and the frequency of the xR1a allele (.85) as another marker for the percentage of European ancestry in the populations. Third, the high negative correlation between latidude and the frequency of the haplogroup E1b1b allele (.84) as a marker for the percent- age of North African and Middle Eastern ancestry….
Hereditarians have successfully established Italian north-south cognitive ability and ancestral differences. The existence of these can no longer be reasonably denied. And Ron Unz is correct that migrant data offers a test of a casual genetic hypothesis which proposed that differences in ancestral related genes condition differences in cognitive ability. Indeed, Hereditarians have made this point time and time again. As Lynn has noted:
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)
Indeed, the view that trait differences in questions with follow populations is the primary reason that racionational traits are of interest. Showing a lack of correspondence between claimed regional differences and the outcomes of the diaspora from these regions not only empirically falsifies the racionational hereditarian hypothesis but it also undercuts its Raison d’être. Since this point is so obvious, I will elaborate no further on it. Now pertaining to this point, Ron has claimed that “the limited data we have seem to put [the average Italian-American] IQ close to [the White Average].” Imaginably, Ron was referring in part to Thomas Sowell’s collection of data, presented in Essays and data on American ethnic groups, which seemingly showed a rise in Italian American IQ scores.
A critique of Sowell’s White Ethnic American Data
Sowell’s results, though, as presented are deeply problematic for any usage and worthless for his own employment i.e., historic trends. Moreover his discussion of scores is queer. On page 2010, for example, Sowell notes that, “It must be emphasized that historic trends only can be compared among these groups — not absolute I.Q. levels — for the samples were not drawn from identical schools, nor were they identical with regards to geographic distribution or age distribution”. Sowell’s statement logically implies that groups also can not be compared against a national mean, since were they each comparable to a national mean, they could each be compared to each other, yet, in the very next sentence, he states: “Iit is clear that Italian and Polish I.Q.’s have risen significantly overtime, to a level at or above the national average.” Moreover, in the Appendix, he suggests that the data was effectively randomly sampled:
If so, we can make between group comparisons — assuming, of course, also that the averages were based on random sets of tests and test editions. Concerning Sowell’s assertion on the IQs of Italians and Poles, he references his table 1. This is shown below:
It’s tempting to dismiss Sowell’s initial warning and make group comparisons, but soon after we are told of more problems with the data:
Now we have, medians — not transformed standardized differences — based on a heterogeneous collection of unnamed tests based on maybe unrepresentative samples. One can readily see the potential problem with test heterogeneity. Imagine, for example, that the mostly German schools in the mid 1960s used the Kuhlmann-Anderson Intelligence Test form A (standardized in 1927) and that the mostly Italian schools during the same period used the Otis Quick Scoring Mental Ability Tests: New Edition (standardized in 1954). For a list of test editions and their publication dates, refer here (in the right corner, test files beginning with…). Assuming a Flynn effect of 3 points per decade, the German scores would be inflated 11.4 points ((1965-1927)/10*3) relative to the (1927) “national average” while the Italian scores would be inflated 3.3 points relative to the (1954) “national average”. Assuming that the Italian and German samples were both nationally representative and that both samples had the same distribution of latent ability, by 1950, simply owing to the Flynn Effect and differences in tests administrated, one would find an 8.1 point difference, given a 3 point per decade Flynn effect. This doesn’t, of course, mean that Sowell’s ethnic scores are incomparable (contrary to what he himself claims), but it does mean that comparisons are problematic. And owing owing to the Flynn Effect in combination with test/test edition heterogeneity, comparisons become increasingly difficult with subsequent data.
Now scores based on specific tests are reported for some ethnic groups, but these just illustrate more problems with the data. For example, below are the Italian scores:
Here, it seems as if there is a score rise, but the apparent rise is undoubtedly in part a the Flynn Effect; Sowell, obviously didn’t make Flynn Effect adjustments. It’s impossible to precisely evaluate the impact of the FE on scores because even when test are reported, test editions are not. But one can get a sense of the problem by looking at scores for which one the the same edition was probably used e..g, Alpha Otis. For these tests, correcting for a three point Flynn effect per decade abolishes the score increases. The Italian Flynn Effect corrected scores for the Otis A would be about 90 in the 1920s and about 90 by the 1970s. In fact, for all the ethnic data for which tests are noted — therefore allowing a relatively straightforward decade to decade comparison — there is no score increase, once Flynn corrections are made. This doesn’t, of course, mean that Sowell’s scores are completely uninformative. Just that, as presented, one can’t make sense of the historic trends. Since IQ tests were first standardized in the 1920s and 1930s, if one assumes that samples were nationally representative (of the 1920s and 1930s sub populations), which Sowell both argues is and is not the case, one can compare groups scores from the first two decades, since the issue of test heterogeneity and the Flynn Effect will not come (much) into play. So we return back to Sowell’s table 1, and restrict out focus to the 1920 and 1930 data:
This, Jason Malloy tells me, is the best data we have — assuming that we don’t wish to pay the $305 for the microfilm data and do a re-analysis of the 10s of thousands of scores taking into account demographics, region, the Flynn Effect, Et cetera, Et cetera . And I don’t think we do. (It will be noted that I cut Jewish scores out because Jewish-American IQ scores have been extensively reviewed by others and because other considerations come into play when dealing with this ethnoreligious population.) Based on these scores, there seems to have been appreciable ethnic differences. But what about contemporaneous difference? Since we can’t rely on Sowell’s data and since, as Ron has noted, the data is limited the issue is in need of investigation.
Post 1970 White Ethnic American Data based on Four Nationally Representative Samples
To that end, I asked Meng Hu to decompose the NLSY 79 and 97 scores of racially White individual by parental reported European ethnicity. The results were as follows (with syntax included for ease of replication):
To Meng Hu’s analysis, I added PPVT results from the Add Health for the ethnic groups with appreciable sample sizes (e.g., English, French, German, Irish, and Italian):
And GSS standardized scores differences:
And then averaged the scores of the largest groups:
In short, there are no appreciable differences between Whites of the major different European ethnic backgrounds in the US. So, it seems that Ron Unz was correct in that regards, after all.
An investigation of Peter Frost’s counter argument
Now, it has been argued, more or less, by Peter Frost that reported ethnicity, in the US, may be an unreliable index of true white ethnic heritage. Quote:
I am even more skeptical of your recent data on “Irish Americans.” There is no such population. There are simply a lot of people with varying amounts of Irish ancestry. “Irish Americans” are increasingly people who take an interest in Irish music, culture, and history, and such people tend to be more educated than average. Another factor is that people tend to identify with the branch of their family tree that has a stronger sense of ethnic identity. If a person is part English and part Irish, they tend to identify as Irish. But if a person has equal contributions of African and Irish ancestry (like Mariah Carey), they tend to identify as African American. More on Race, IQ, and Wealth
I noted, a while back, on some queer blog, that this hypothesis could be investigated. Quote:
The Add Health survey has information on height, skin color, eye color, hair color, IQ, ancestry, etc. for subjects in wave III (conducted in 2001-2002) when the individuals were 20-30 years old. If someone wanted to, they could test to see to what extent differences in reported European ancestry among Whites in the US corresponded with genetic differences between corresponding European populations (e.g., if the odds ratio of having red hair was higher for those with reported Irish ancestry than those with reported Polish ancestry…). This would test Peter Frost’s position versus Ron Unz’s. Recall, Peter argued that reported European ancestry was virtually meaningless. If it was found that Peter was wrong, and that reported European ancestry corresponded somewhat with intra-European genetic differences, one could then test Ron’s position versus Richard Lynn’s. Recall, both Ron and Richard agreed that reported ancestry, in the US, indexed true European ancestry. They merely disagreed on whether ethnic IQ differences in the US matched national IQ differences in Europe.
I have since done that analysis using height, skin color, hair color, eye color, and reported ancestry using the Add Health data. The results for English, French, German, Irish, and Italians are as below:
Not being a biological anthropologist, I don’t know how well the US ethnic-European National differences in physical traits (e.g., Hair Color) correspond. But the standardized physical traits differences did load on a common factor which explained 70% of the variance and on which IQ differences only weakly loaded.
To take an example of the differences, the German-Italian standardized difference was for Light Eye Color 0.42, for Light Hair color 0.38, for Height 0.20, for Skin color 0.33. And for PPVT 1 & 3 0.25. Yet the n-weighted cognitive ability score differences was only 0.09 SD or about 1.5 IQ points based on samples sizes of 5566 and 1230 for parental or self identified Germans and Italians, respectively. The fact that “Irish-Americans” and “Italian-Americans” deviate in propensity for red hair color by almost 0.5 standard deviations suggests to me that ethnic identify in the US is not wholly a social construct unrelated to genetic ancestry. For reference, according to the gurus over at Europedia, 10% of autochthonic Irish have red hair versus 1% of autochthonic Italians, which is equivalent to a standard deviation of 1.04. Based on the various physical differences (and lack thereof), I would hazard that the (a) reported national ancestry partially indexes true national ancestry and (b) that Whites in the US are not ethnically homogenous. Surprise!
These results, of course, should be replicated using the eye and hair color data in the NLSY 79 and the eye, hair, and skin color data in in the NLSY 97. For now, though, I will hazard, after a belated but not exactly superficial investigation of this matter, that Ron Unz’s basic argument is valid. Quote:
Being “Irish” is based on self-identification and reporting, so I’d certainly expect that most of the “American Irish” aren’t “pure Irish.”
But none of that makes a difference. If the Irish had an actual, innate, genetic IQ of 87, and this figure was not subject to rapid change under socio-economic influences, there would be *massive* evidence of this in American society.
For example, something like 15% of all the Irish in America would have IQs below 70, and would be subject to clinical mental retardation. Do you really believe that 15% of all American Irish are mentally retarded?
Okay, maybe lots of those Irish aren’t pure Irish, and are part German or Italian or something. Well, according to Lynn the (South) Italian IQ is around 89, so that wouldn’t really help much. But anyway, we’d still be seeing millions of mentally-retarded Irish-Americans. Do you believe that? (Unz quoted by Frost in More on Race, IQ, and Wealth
Ron’s Missing Argument
Valid, but not compelling in the least. In response to Ron’s point about Irish IQ, one can dismiss the claimed low Ireland Irish IQ scores or make a case for super duper Irish-American immigrant selection. The latter claim is not inherently unreasonable as the sources indicate that the Irish in the US were never found to be, on average, cognitively sub par. And the former is defensible, as far as I am aware; I myself, in my Redebunking Ron piece re-analyzed the Irish IQ, Achievement data by birth cohort and found no significant correlation between birth year and scores — a point which Mr. Unz cared not to acknowledge. Alternatively, one can grant the point about the Irish IQ and note that it is of little overall significance with regards to the overarching global hereditarian position. One can note Ron’s primary fallacy, which he makes time and time again, and which I discussed:
Above are but a few instances of where Ron has either been grossly in error or intentionally misleading. Realizing the empirical shakiness of his position he has resorted to a handbag of fallacies. His remaining argument is some permutation of:
Ron’s False Dilemma
Either Lynn/Eysenck was right about the low Irish IQ in the ‘70s and there was a massive Irish Flynn effect or he was wrong and his judgment isn’t to be trusted. If the former, (b) it’s clear that national IQs are highly malleable, and therefore Lynn’s is incorrect about his genetic hypothesis; (b) if the latter, Lynn’s national IQ data can’t be trusted.
No matador is needed to handle this bull. It can be squarely grabbed by the horns.
(a) Firstly, heritability is a population level statistic. This applies to nations just as well as individuals. As such, even if it were shown that the difference between nations X and Y were completely environmental, this would imply nothing necessarily about the difference between nations A and B.
(High heritability only allows one to make probabilistic statements on the individual level.
Richard Lynn has claimed that approximately 50% of the between national variance in IQ is heritable. For this to be correct, it’s obviously not necessary, given what was said above, for 50% of differences between all (or any) nations to be genetically conditioned. The total amount of variance explained by genes just has to add up to 50%.
Now this theoretical point has practical importance. To make the case that the difference between B and C is not genetic, it’s not sufficient to merely show that the difference between A and B is (probably) not. One has to show: That some environmental factor(s) X plausibly causally explain the difference between A and B, that these factors could causally explain the differences between B and C, and that B and C differ in these factors by a magnitude enough to explain the B,C gap….
As it is, the purported low Irish IQ doesn’t make much sense in context to Lynn’s evolutionary hypothesis which posits that differences evolved, along a north-south gradient, in response to cold weather selection. If anything, low Irish IQ, not high, challenges this theory. Ron’s case, with regards to this data point, boils down to an ad hominem: Lynn possibly made a mistake, he’s a leading hereditarian, he and hereditarians can’t be trusted.
The whole Irish IQ line of reasoning was nonsense. Absolute nonsense.
The argument Ron should have made
The argument Ron should have forcefully made is the argument that I am elaborating on here:
(1) Lynn has shown that Southern Italians have lower levels of cognitive ability than do northern Italians and Northern Europeans (Lynn, 2010; Lynn, 2010; Lynn, 2012).
(2) Importantly, Lynn and Co. have tied, though extensive correlational research, this cognitive ability gap to the overarching global hereditarian hypothesis, by showing that inter-Italian differences correlate with skin color, hair color, brain size, latitude, and African genes (Rindermann et al., 2012; Templer, 2012; Lynn, 2012).
Given (1) and (2), showing that the North European-North Italian-South Italian cognitive ability gradient isn’t congenitally conditioned undermines (a) global hereditarianism with respect to Caucasoids (i.e., North Europeans, South European, North Africans, Middle Easterners, etc.), (b) the climate, brain size, brain efficiency hypothesis, and (c) the color pleiotropy hypothesis. Also, since a Jensen Effect has been shown for the White European-Middle Eastern immigrant differences, just as for the Hispanic-White difference, the Jensen Effect argument would be challenged. That is, showing this would fundamentally damage the global hereditarian hypothesis (which in other respects isn’t fairing too well — a point which will perhaps be discussed at another time).
(3) Migrant performance is a crucial tests of the global hereditarian hypothesis; moreover, it is the Raison d’être for this hypothesis.
(4) There is no evidence that Italian immigrants to the US were cognitively selected (relative to the region). Numerous studies have shown that, as a groups, in the 1910s, 1920s, and 1930s, they and their offspring performed sub par as compared to norther European immigrants and their offspring. For example, in Characteristics of the American Negro, Otto Klineberg cited 15 studies on Italian Americans:
As Klineberg’s summary includes samples based on first generation immigrants for which cultural and linguistic bias was an issue, these results are suspect. Nonetheless, Sowell’s 1920s and 1930s results — discussed above — paint a similar picture. The reality of a substantial Italian American — Northern European American score differences, at this time, is undeniable. The relevance here is that this suggests that the Italian immigrants were not positively cognitively selected — at very least, there is no prima facie evidence of positive selection.
(5) Obviously, if the Italian American immigrants were substantially negatively selected relative to the region from which they hailed either they wouldn’t have closed the IQ gap — and based on the four nationally representative studies reviewed above, they have. — or the global hereditarian hypothesis (as formulated) must be false. So, by a global hereditarian hypothesis (as formulated), it must be said that Italian American immigrants must either have been neutrally or positively selected relative to the regions of origin.
(6) Again, the idea that they were positively selected seems odd, given the early 20s and 30s scores, but is not impossible. Since 75% of the American Italian gene pool came from South Italy, which has a reported regional IQ of 90, this could explain why the contemporaneous Italian American IQ is not 0.5 SD below the IQs of Norther European Whites.
(7) That self/parental-identifying post-1970s Italian Americans are a social construct unrelated to ethnic ancestry is an untenable position given that anthropometric results reviewed above. They are nonetheless admixed, as indicated by the intermarriage rates:
The tests subjects in the NLSY 79 (born circa 1960), NLSY 97 (born circa 1980), and Add Health (born circa 1980) were plausibly only, genealogically, 50% Italian. If so, we might only expect a 0.25 SD or so gap. This still is substantially larger than is found. But perhaps there was some positive selection and perhaps self identification attenuates differences.
(8) But both of the latter possibilities can be tested by looking at the correlation between cognitive ability scores and indexes of admixture within the Italian American population, just as within the Hispanic and African American populations. One wouldn’t expect migrant selection to abolish the intra-Italian American differences, which according to Lynn exceeds 0.66 SD (between Italians of Northern and Southern regional origin). And if admixture was attenuating differences, one would expect to find an association between indexes thereof and cognitive ability with this population. In both cases, there should be evidence of a correlation between Northern European traits (hair color, eye color, skin color) and cognitive ability. Also, in the latter case, individuals of mixed Italian – Northern European parentage (based on parental report) should perform intermediate to those of only Italian and only Norther European heritage. This is just applying the various admixture analyses to Italians and treating this group as we might any other for which a mean congenital difference is proposed.
Those results, once I am finished compiling them, will be reviewed in a subsequent post. For now, I will merely note that this post is entitled “The argument Ron should have made” for a reason.
But can we explain the substantial failure of Lynn’s hereditarian hypothesis in this case? Well, of course, there is one obvious way.
Speaking of Hereditarian hypotheses, recently, Ron Unz restated his social Darwinian model for high Han IQ. Elsewhere, in a clever comment, he also noted one of the strongest arguments against this and related models i.e., just as English don’t outperform Russians as one would expect given Clark’s model, Han don’t outperform Japanese as one would expect given Unz’s model. Nonetheless, Ron suggested that Chinese might be more apt than Japanese and Koreans and cites as possible evidence scores from California. Strangely no one seems to have bothered to take a look at the STAR achievement scores, which allow one to compare the performance of different monoracial Asian ethnic groups by grade by year on the moderately g-loaded California Achievement Test.
Generally, I don’t see any evidence in support of Ron’s conjectured ethnic North East Asian differences at least as pertaining to achievement in California. Then again, the mostly Black public school students in Bermuda score little less than the mostly White US normative sample on basically the same tests, so maybe the problem is with the CAT/Terra Nova.
On a related note, the Thai HBDer, Sagat, got around to verifying my claim concerning the Thai Nation IQ. Quote:
I looked into it further and it appears that I was wrong and the average Thai IQ is actually 98.6. I find this hard to accept, because this IQ level does not correlate well with the performance of Thai students on standardized tests and with other developmental indexes. Also, I have found nothing but articles in Thai bemoaning the results of this study rather than celebrating the rise in IQ. It doesn’t make sense to me….
…I found this Thai site that addresses the discrepancies between the earlier study cited by Lynn and this more recent study. They explain that it was due to a more representative sample size and type of tests used, with the previous study only testing students in Chonburi province.
So, Ron’s social Darwinian HBD model seems to be problematic — but, perhaps, no less than Lynn’s climatic model, as least pertaining to Eurasian differences.