August 29, 2004

Lynn and Vanhanen’s IQ And The Wealth Of Nations Vindicated

By Steve Sailer

This century's most talked-about—but least written-about —book of underground social science has been IQ and the Wealth of Nations, by Richard Lynn and Tatu Vanhanen. (Click here for my review and here for J. Philippe Rushton’s.)

The two veteran scholars dredged through decades of psychometric literature to find 184 studies providing estimates of average IQ for 81 countries. They then showed that average IQ and per capita Gross Domestic Product [GDP] correlated—at the kind of high level that social scientists mostly only dream of finding.

Obviously, this is interesting and important information. Yet more than two years after publication, the only mention of the book in any widely-distributed print publication in the U.S. was when The Economist cited it as the source for the claim that Democrats are smarter than Republicans—data that I showed were a hoax.

The intellectual cowardice of the press in America is, of course, contemptible. But at least the police here aren't looking into levying criminal charges against the authors.

They have in Finland. Recently, Vanhanen, who is an emeritus professor of political science at two different Finnish universities and, for that matter, the father of Finland's prime minister, gave an interview to a Finnish newspaper explaining the content of the book. Result: as the Washington Times reported:

"Finland's National Bureau of Investigation is considering whether to start a criminal investigation into comments made by former political science professor emeritus Tatu Vanhanen, father of Prime Minister Matti Vanhanen." Finnish Premier apologizes for father Helsinki, Finland, Aug. 12 (UPI)

Fortunately, no charges were filed.

The heart of IQ And The Wealth Of Nations is its Appendix 1, which describes each of the 184 studies.

Lynn and Vanhanen's summary listing of mean IQ scores for the 81 countries has been available on the web for some time. (Here, on Lynn's website is his list. And here are other copies of the summary list: wordIQ, sq.4mg, Griffe, Nuenke.)

Unfortunately, everything on the web heretofore has made Lynn and Vanhanen's results look like a black box. This has had two bad effects.  

If you've actually studied Appendix 1, and seen the methodological hurdles Lynn and Vanhanen have had to deal with, you're not likely to say things like, "L&V showed that Sweden's IQ is higher than Norway's." Sure, they came up with a 100 estimate for Sweden and a 98 estimate for Norway. But the reality that's apparent in Appendix 1 is that there's way too much noise in the data for fine distinctions like that to be trustworthy.

To open up the black box, I've created a table displaying virtually all the information Lynn and Vanhanen provide on each IQ study they used—not just the overall the national results you've seen so far. 

This should prove highly useful to future researchers.

Skeptics are likely to be surprised by how robust and consistent the findings tend to be.

A large number of the studies were done by professional psychometricians "standardizing" well-known IQ tests, typically culture-fair nonverbal ones like the Raven Progressive Matrices on nationally representative samples. By attaining an accurate national average, they allow local school psychologists to determine how bright their students were compared to their peers within the country.

Just by eyeballing the data in the table, you can see that there is a relatively high degree of internal consistency within countries.

For example, there are three studies for Switzerland, for example. They came out to 101, 99, and 102.

Sometimes, the competing studies disagree significantly. For example, the two Polish studies are 14 points apart and the two Portuguese studies are 13 points apart. But those kinds of divergences are rare. 

By disaggregating the data, I've found that the overall estimates are reasonably reliable.

For example, for countries where L&V found more than one analysis, the average difference between studies and the national average was plus or minus 2.5 points. In other words, if L&V reported the average for the country, based on two studies, was 100, this would typically be based on studies reporting IQs of 97.5 and 102.5.

That's quite consistent for most useful purposes. But of course it's not sensitive enough to determine bragging rights between nearby countries.

Some national estimates are necessarily less trustworthy than others. For example, Lynn and Vanhanen's estimate of Colombia's mean IQ as 89 is barely more than guesswork. They found a single study of exactly 50 teenage white boys, who averaged an (adjusted) 95. Then they developed an estimate for the whole country based on an almanac's description of the racial makeup of the country, combined with IQ scores for mestizos in other countries. It's logical, but not much more.

On the other hand, L&V’s estimate of 107 for the highest scoring country, Hong Kong, appears to be quite solid. They have five studies, three with sample sizes of 4,500 or more, with average adjusted scores of 103, 110, 109, 107, 107—on a scale where Britain is 100 and the U.S. is 98.

Overall, the data when grouped into regions or races seems quite consistent. For example, among Northeast Asians (Koreans, Japanese, and Chinese), 24 of the 26 studies they uncovered reported scores between 100 and 110.

Across 65 studies countries populated by Europeans or of the European community within multiracial countries like South Africa, scores ranged from 87 to 107, with 49 of the 65 falling between 94 and 103.

The very low scores seen in black countries are often criticized as impossibly low by those with little familiarity with the subject. But the 33 studies cited are depressingly similar: five in the 75 to 80 range, and 28 below that. (African-Americans average higher, in the mid-80s.)

In an important and lively article in VDARE.com called “Solving The African IQ Conundrum: 'Winning Personality' Masks Low Scores," Phil Rushton recently addressed the riddle of how Africans can score so poorly, yet have such bright, lively personalities—but, then again, perform as badly as their IQ scores predict on real-world challenges like engineering and management.

Rushton cited a wonderful Muhammad Ali-Howard Cosell exchange to show the black gift for coming up with subjectively delightfully unstandardized answers—a talent that objective standardized tests, by their very nature, can't measure:

"'I’m gonna whoop him, Howard. You just watch!' Cosell responded, 'You’re feeling very truculent today, Muhammad.' Without batting an eye (or opening a dictionary), Ali uttered one of his trademark retorts, 'Truculent? If that’s good, I’m it!'"

Ali was one of the most fun personalities of the 20th Century.

Yet, while he was still young and un-punchdrunk, he recorded an IQ of 78 on the military's entrance exam. He denied intentionally trying to score low. Gerald Early, the prominent black studies professor and boxing historian who edited the Muhammad Ali Reader, commented, "He hadn't a single idea in his head, really ... I think the score was an honest reflection of Ali's mental abilities."

Yet, Early notes correctly, "He was intuitive, glib, richly gregarious, and intensely creative, like an artist." 

Unfortunately, Ali's winning personality couldn't fully compensate for his low IQ. Despite coming from an artsy, middle class home, he was also, more or less, illiterate. In a poignant scene in the documentary We Were Kings about his epic 1974 fight with George Foreman, Ali says his biggest regret was not learning how to read.

While IQ tests can't measure black interpersonal social competence adequately, they are not racially biased in predicting most levels of accomplishment related to abstract, impersonal thinking.

Lynn and Vanhanen made a variety of adjustments to the reported scores to make them more consistent. Notably, they adjusted for the phenomenon that raw test scores rose throughout much of the 20th century at a rate of a few points per decade. When tests like the Progressive Matrices were restandardized in Britain every few decades, the creators would raise the number of right answers required to score a 100.

This phenomenon was first noticed in the 1940s, but Lynn was one of the first researchers to call lasting attention to it. Later, New Zealand political scientist James Flynn did important work on the subject. It is now usually called the Lynn-Flynn Effect or simply (and somewhat unfairly to Lynn) the Flynn Effect.

Lynn and Vanhanen assumed the Effect is real and that it happens in all countries at the same rate (2 points per decade for Matrices-type test and 3 points per decade for a Wechsler-type test). So they fine-tuned scores based on when the test was taken relative to its restandardization in Britain or America.

This technical adjustment appears to make the results slightly more internally consistent, so it seems sensible.

Race-deniers, however, have far more ambitious hopes for the Lynn-Flynn Effect. They assume that it will cause racial gaps in IQ to converge out of existence.

The logic behind this argument is not implausible: if IQs are being depressed by poor nutrition, poor health, inadequate primary education, and so forth, then poor countries should catch up faster in IQ—because rich countries with plenty of food, medicine, and schooling should run into diminishing marginal returns to improvements in standards of living first.

To cite a seemingly relevant analogy, as East Asian countries emerge from poverty into affluence, the average height of their young people has been growing faster than in the already prosperous United States.

On the other hand, we don't know for sure what causes this rise in raw scores. So this rationale might not apply.

Fortunately, we can look at Lynn and Vanhanen's findings to see if there has been any convergence of IQs among the races. Or, have the smart gotten smarter?

I've plotted on this graph the average scores found in 124 studies ranging from 1914 to 1998.

In European's countries (blue dots), scores have stayed quite stable over several generations. Likewise, in black's countries (black triangles), average IQs have not gone up, and may have slightly declined. Thus, the white-black gap hasn't changed much over the decades.

My conclusion: there is no evidence that the Lynn-Flynn Effect is narrowing the white-black disparity. (Similarly, it hasn't narrowed it in the U.S.)

Which doesn't mean that a certain amount of convergence might not happen some day. I recently pointed out in VDARE.com that enriching commodities like salt and bread with crucial micronutrients could raise IQs in the Third World. If African countries became better governed, they should be able to help their people get smarter.

But there is a tragic chicken or egg conundrum at work in Africa: if the people from whom the governments are drawn aren't currently very smart, so it's hard to assemble a competent government.

In contrast, the scores of the highest scoring group, Northeast Asians (yellow triangles), may have gone up slightly relative to whites and blacks over the last half century. There's a lot of noise in the data, so it’s not proof that Northeast Asians are getting smarter. But considering how much conditions have improved in that region over that time, it would be hardly surprising.

Interestingly, the one outlying data point among the NE Asian studies—a 92.5 average IQ in China in 1986—was found for a sample of adults, most of whom had lived through the terrible famine of the early 1960s and the anti-intellectual chaos of the Cultural Revolution. But when researchers also gave the same test at the same time to a few thousand Chinese children, who had enjoyed eight years of more sane government beginning with Deng's pro-market reforms in 1978, they averaged 100.

In other words, it appears that over the last half century, the global racial gaps in IQ did not converge. Indeed, they have grown.

But we won’t know what’s going on in this or any other area of IQ research unless we can discuss the matter frankly and honestly.

Police investigations—and promiscuous accusations of “hate”—don’t help.