You are currently viewing Wealth gains and IQ gains since Ron Unz’ 2012 post

Wealth gains and IQ gains since Ron Unz’ 2012 post

In 2012, Ron Unz (of wrote a long article with the title Race, IQ, and Wealth published in The American Conservative (follow-up also in 2012 Race/IQ: Irish IQ & Chinese IQ). Unz did not shy away from the central race and intelligence debate. In fact, he relied on data from the 2002 book by Richard Lynn and Tatu Vanhanen about national IQs (IQ and the wealth of nations). In his words (closing his article):

None of these findings would have been possible without the great scholarly effort Richard Lynn and Tatu Vanhanen put into locating and properly presenting an enormous quantity of international IQ data in their books and research papers, as well as their courage in focusing attention on such highly controversial topics. Although I would argue that a close examination of the Lynn/Vanhanen data tend to convincingly refute their own “Strong IQ Hypothesis,” I would be the first to acknowledge my gratitude to the scholars whose efforts made my own analysis possible. Meanwhile, individuals such as Stephen Jay Gould, who commit outright academic fraud in support of their ideological positions, do enormous damage to the credibility of their own camp.

A lot of Unz’ arguments center on taking all results at face value. Consider his take on German IQ results:

Consider, for example, the results from Germany obtained prior to its 1991 reunification. Lynn and Vanhanen present four separate IQ studies from the former West Germany, all quite sizable, which indicate mean IQs in the range 99–107, with the oldest 1970 sample providing the low end of that range. Meanwhile, a 1967 sample of East German children produced a score of just 90, while two later East German studies in 1978 and 1984 came in at 97–99, much closer to the West German numbers.

These results seem anomalous from the perspective of strong genetic determinism for IQ. To a very good approximation, East Germans and West Germans are genetically indistinguishable, and an IQ gap as wide as 17 points between the two groups seems inexplicable, while the recorded rise in East German scores of 7–9 points in just half a generation seems even more difficult to explain.

A more prudent reader here would probably think that the result from the 1967 study was inaccurate somehow: sampling error, sampling bias (i.e. unrepresentative sample), analytic error, or maybe even test bias. Unz instead thinks the average intelligence of east Germans improved quickly and there was a large gap between West and East at some point, and this probably has to do with the later increases in East German wealth after reunification. Looking up the German section in the 2002 book, we find:


Buj’s (1981) sample of 1,572 adults in West Germany (the Federal Republic, as it was called at that time) were tested with the Cattell Culture Fair test and obtained a mean IQ of 109. To calibrate this figure against a British mean of 100, it needs to be reduced to 107.

Around 1970, the Coloured Progressive Matrices Test was standardized in West Germany by Winkelman (1972) on a sample of 563 5- to 7-year-olds. The mean IQ in relation to the British standardization sample of the Standard Progressive Matrices is 97. Because of the 9-year interval between the two standardizations, this figure needs to be raised to 99.

In 1978, the Coloured Progressive Matrices Test was standardized again in West Germany on a sample of 3,607 6- to 10-year-olds. The norms are given by Raven, Court, and Raven (1995). In relation to the 1979 British standardization sample of the Standard Progressive Matrices, the IQ of the German sample is 101.

In 1978, norms for the Standard Progressive Matrices were collected for a sample of 2,068 11- to 15-year-olds in West Germany. The data are given by Raven (1981). In relation to the 1979 British standardization sample, the IQ was 105.

Three studies have been made of intelligence in East Germany. In 1967, the Coloured Progressive Matrices Test was standardized in the city of Rostock by Kurth (1969) on a sample of 454 7- to 11-year-olds. Their mean IQ in relation to the 1979 British standardization sample of the Progressive Matrices is 87. Because of the 12-year interval between the two standardizations, this figure needs to be raised to 90.

In 1984, further norms for East Germany were obtained by Guthke and the data are given by Raven, Court, and Raven (1995). The mean IQ of the sample in relation to the 1979 British standardization of the Standard Progressive Matrices is 98. To adjust for the 5-year interval between the two standardizations, this figure needs to be reduced to 97.

Around 1978, norms for the Standard Progressive Matrices were obtained for approximately 1,000 11- to 15-year-olds by Mehlhorn. The data are given by Raven (1981). In relation to the 1979 British standardization sample, their IQ is 99.

The average of the results for West Germany is an IQ of 103 and for East Germany an IQ of 95. For united Germany, weighting these figures by the numbers of the populations of West and East (59.5 million and 16.6 million, respectively), the IQ of united Germany is 102.

The study is: Kurth, E. von. 1969. Erhöhung der leistungsnormen bei den farbigen Progressiv Matrizen. Zeitschrift für Psychologie, 177: 85–90. I can’t find a copy of this paper on the current publisher’s website, nor anywhere else. It is not part of David Becker’s updated compilation, so probably he couldn’t find it either. I wrote to him to ask, and he’s looking into it. I also requested a scan of this via a colleague who has access to interlibrary loans. I will update if I get anything. In the meanwhile, the details of the study are mentioned in the 2012 update by Lynn and Vanhanen, which gives for Germany:

A quick glance shows that sampling errors are not a likely explanation, as the smallest sample had 200 7 year olds. In the case of Kurth, it s 454. The test is given as SPM but the text from the 2002 book gives it as CPM, which is also what the title of the paper says (in German, farbigen = colored). Either way, it’s a language free test, no likely to have serious test bias. Sampling bias might explain the results, as the city of sampling, Rostock, is kind of a low ranking German city. A 2006 study of the largest 50 cities found that Rostock came in last in their overall index of city development (I could not find data from the 1960s). OK, in the 2021 results, it is only rank 44 of 70 (Berlin is rank 41), but the 2006 is closer in time, and I can’t find anything older on their website or or IA. Anyway, so we have reason to think the children sampling there were below the German ethnic average and probably East German average. It’s like sampling data from whites in West Virginia and Massachusetts, seeing a gap of 10 IQ and concluding that it must be due to their wealth. I looked at some Danish IQ results from the army test, and one can find a 10 IQ gap between the Lolland-Falster Danes and those in Frederiksberg enclave in Copenhagen. This gap is probably not much to do with wealth, but with genetics. It would be a mistake to conclude that just because one can see a large gap between two groups of the same ethnicity, Germans, American Europeans, or Danes, that this gap is due to non-genetic factors. This error goes back to the 1930s, named Klineberg’s fallacy by Arthur Jensen.

Returning to Unz, the other in the list is the 107 result from Buj 1981. That study is commonly believed to be fraudulent or otherwise unreliable, so this result is less surprising with that in mind. Since not many people are familiar with the case, I will summarize it. It’s done by a relatively unknown scholar from Croatia living in West Germany, and the total article runs 1.5 pages despite supposedly testing 10,000+ persons! If someone really did a massive study like this, one would surely expect them to detail the methods, and do follow-up publications. The German sample supposedly comes from Hamburg (ranked 8 of 50 in 2008, if you are wondering why I keep using different years, it’s because the magazine (WirtschaftsWoche) does not provide the old rankings on their website (without paywall), and I had to rely on newspaper summaries that don’t give all the ranks for each year). In fact, according to maybe reliable online IQ test results, there is a 14 IQ gap between the state that has Rostock and the Hamburg city state. Back to Buj 1981, his table looks like this:

(Lynn has the wrong sample size, having apparently used the value for France (1320) in the bottom row instead of the Hamburg one (1572).)

Here we notice that the standard deviations vary a lot more than expected by chance. In fact, Buj notes:

The most curious feature of the table is the very great divergence of standard deviations from their mean. ranging from 11.6 for Norway to 34.7 for Bulgaria and Spain. It is difficult to explain these differences; they must certainly cast some doubt on the comparability of sample choice in the different countries. It is exceedingly difficult to standardize conditions. instructions. and motivational factors over a large number of different testers. organisers, and subjects, and any differences along these lines may contribute materially to observed differences in the results. In spite of these well known difficulties it seems worthwhile to publish the results obtained, without claiming a high degree of accuracy. They may serve as a baseline against which future investigations may be carried out. and with which future results may be compared.

With this in mind, it is odd that Ron Unz takes these results at face value to both index the German ethnic group. I mean, OK, Richard Lynn does the same, but he used the median across studies to essentially ignore outliers like these. The next largest gap by comparison are both from Raven himself (the test developer), with means of 97 and 105 based on two large samples (8 IQ gap). I didn’t investigate the discrepancy but presumably it involves similar sampling biases. Ron Unz’ error here is essentially the same as that committed many decades ago in Industrial and Organizational Psychology, called situational specificity. I covered that in a recent book review of books on testosterone. The error consists in ignoring sampling error, sampling bias etc., and interpreting any difference across studies as reflecting reality as it is (sometimes called naive realism). Since studies produce varying results for varying reasons, mostly sampling error, this realistic interpretation then leads to the conclusion that there are tons of interactions out there, or that mean values of traits vary strongly between time and place.

Finally, Unz is wrong that to a “very good approximation, East Germans and West Germans are genetically indistinguishable”. Multiple studies show that one can fine point ancestry within country, even very homogeneous Danes, using genetic markers. They are not indistinguishable. Since at least 2019, 23andme provides such results for their customers already, for many countries. There is also evidence that polygenic scores for education vary within British people by geography, supporting the very likely genetic differences in intelligence within an ethnicity that varies by geography. In Unz’ defense, his post is from 2012, so he couldn’t know about these findings.

There is even a 2012 (December, so after Unz’ post, submitted in April, so could not be inspired by Unz’ post either) study about German regional IQs and wealth gains:

Lynn and Vanhanen (2012) have convincingly established that national IQs correlate positively with GDP, education, and many other social and economic factors. The direction of causality remains debatable. The present study re-examines data from military psychological assessments of the German federal army that show strong IQ gains of 0.5 IQ point per annum for East German conscripts in the 1990s, after the reunification of the country. An analysis of IQ, GDP, and educational gains in 16 German federal states between 1990 and 1998 shows that IQ gains had a .89 correlation with GDP gains and a .78 correlation with educational gains. The short time frame excludes significant effects of biological or genetic factors on IQ gains. These observations suggest a causal direction from GDP and education to IQ.

So it shows a maximum IQ gap in 1992 of 102-93 = 9 IQ, compared to the 17 IQ gap Unz used. The author of this paper basically agrees with Unz’ thesis that IQ gap closings are due to environmental factors. They could be. That doesn’t mean that factor is wealth (a very distal cause). I would guess a more plausible factor is the change in schooling following reunification. Schooling increases IQ scores (but maybe not much general intelligence itself), especially on crystallized tests (prior learning tests), which are included in their test battery. These data do not rule out genetic changes whereby below average East Germans moved to West Germany, or above average West Germans moved to East Germany.

This discussion of the German data aside, there doesn’t appear to be any in depth academic reply to Ron Unz. But there are a number of academic papers citing his article. Google Scholar finds 11. These include Kevin Mitchell’s 2018 book Innate, which cites Unz on the Irish IQ changes. The quick rise in measured IQs of south and east Europeans is a favorite topic for African conservative culturalists like Thomas Sowell and Wilfred Reilly, so the topic will surely come up again. It should really receive more in depth study.

While there were no formal academic replies to Unz piece, there were a couple of replies by researchers published in magazines and on blogs:

I was unable to find any replies from later than 2012, either as blogposts or in academic publications. There is however some academic work relevant to Unz’ general thesis about IQ gains and wealth gains. Some researchers have examined whether countries with larger Flynn effects (IQ gains) also have more economic growth (wealth gains). I am aware of 2 studies:

According to the cognitive human capital theory, cognitive ability furthers at the individual, institutional and societal level productivity, production, income and wealth. Cross-sectional and longitudinal studies using various indicators (psychometric IQs, student assessment tests, education vs. GDP per capita, growth), different methods (correlations, regressions, path models) and different controls have supported this theory in two research paradigms (psychology, economics). An especially revealing test is, whether historical increases in IQ within countries would lead to later economic growth, i.e. about 10 to 20 years later. This design can exclude national differences being associated with human capital and growth (e.g., in culture, economic freedom and politics) which may bias the results. We used a data set of national IQ changes (“FLynn effect”) from Pietschnig and Voracek (2015). For a maximum of 28 nations and 262 periods between 1909 and 2013 IQ development was related to concurrent or lagged GDP per capita development (growth; 5, 10, 15, 20 years). In a second analysis with at least three IQ-GDP periods per country the single within-country correlations for concurrent and later intervals were estimated (13 nations). Finally, we controlled for previous wealth (advantages of backwardness). All analyses show substantial relationships between increases in IQ and GDP, the highest were found for the 5 to 15 years lagged economic growth (r = .25 to .44 resp. .46 to .77). The results back the theory that cognitive ability contributes to wealth.

The largest impact is that of Japan (dark red). I don’t know why they don’t use proper fixed effects regression either, but this pseudo-fixed effects regression supports the idea behind Unz’ piece. Also in line with Flynn effect and East Asian growth from 2018:

Data are reported for intelligence of children in China assessed by the Combined Raven’s Test in 1988, 1996 and 2006. The IQ of the samples increased by 15.0 IQ points over 18-year period. The British IQ of China in 1988 and 2006 is estimated as 94.8 and 109.8, respectively.

Curiously, this is the exact opposite finding of what Unz predicted, namely, that Asians were genetically immune to this particular effect.

If we accept the findings from Rindermann and Becker, we still have the issue that Unz claims the causation is wealth gains -> IQ gains, and not IQ gains -> wealth gains. This could be tested with a proper model with lagged variables. One can try lags in both directions, and see which works best. I leave this work to the future (I managed to get one economist interested in this question). Richard Lynn’s position is not really in trouble because he always said he thought the causal direction was both ways. This was stated both in the 2002 book (chapter 10, which mainly argues that malnutrition is lowering the IQs of poor countries), the 2006 follow up book (chapter 9, sections 9-10; Unz never referred to this for some reason), and in his reply to Unz:

Mr. Unz misunderstands our position. We do not propose that the IQs of individuals or of nations are genetically fixed. On the contrary, we are well aware that IQs have been increasing in many countries during the last 70 or so years and that these increases are a result of improvements in the environment, such as better nutrition and longer education. We propose a positive feedback relation from genetically based national IQs to per capita income, and from per capita income on national IQs.

So to conclude:

  • Wealth gains and IQ gains seem related based on multiple studies, but the direction is unclear, and may differ by context. More research indeed.
  • Large IQ gaps within an ethnic group does not imply environmental causation (Klineberg’s fallacy), it could be genetic too.
  • More research is needed on the apparent IQ gains by various East and South Europeans in USA over time. These may reflect:
    • true gains in intelligence due to environmental factors
    • selective back migration, e.g. duller Irish people moved back, increasing the intelligence of the ones who stayed in the USA
    • ethnic mixing. i.e., later generation Irish are not 100% Irish ancestry, but 50% or less
    • test bias, i.e. IQ gaps did not reflect intelligence gaps in the old studies