« Look to the Yehudi | Gene Expression Front Page | Heritability, IQ, SES, etc. »
November 20, 2003

Once more into the breach...

In previous posts, here and here, I argued that international comparisons of IQ do not necessarily indicate genetic differences between populations. I also provided links to comparative tables of various aspects of demography, education, and health.

It occurred to me that it might be interesting to calculate the correlation between some indicator of environmental quality and national mean IQs, as listed by Richard Lynn. For this purpose infant mortality (IM) seemed a promising indicator. From inspection of the figures there seemed likely to be a substantial negative correlation between IQ and IM (high IQ associated with low IM and vice versa), and one might expect IM to reflect key aspects of environmental quality, such as income per head, nutrition, water supply, sewage and waste disposal, health care, and social services.

I don’t have a computer programme to calculate correlation coefficients, but hey, I know the Pearson formula, and I thought it would only take half-an-hour or so. Big mistake! Calculating correlations (for any sizeable n) with only a pocket calculator for assistance is brain-numbingly, eye-glazingly tedious.

But I’m glad I persevered, as the result was remarkable....

The ‘sample’ consisted of those countries for which Lynn provides actual IQ data (not just estimates based on data from nearby countries) and for which the UN tables provide IM. There are 77 of these. Their mean IQ is 88, and their mean IM is 30.2 (deaths per thousand in the first year). The standard deviation of the national IQs is almost exactly 12 [NB: don’t confuse this with the s.d. of individual IQs], and the s.d. of the IMs is 31.6. The covariance of IM and IQ is minus 321, the regression of IQ on IM is minus 0.32, and the regression of IM on IQ is minus 2.23. The correlation coefficient (Pearson’s r ), calculated from unrounded data, is minus 0.844.

This is an astonishingly strong negative correlation. It is stronger than Lynn’s own correlation of 0.757 between IQ and GDP per head. It accounts for over 71 per cent of total variance, as against 57 per cent for Lynn’s correlation (the squares of the respective correlation coefficients). And some of the remaining variance will be random error in both IQ and IM data.

But what does it all mean? To help interpret the results, I also drew a scattergram (which is a lot more fun!). This shows that the regressions are almost perfectly linear, but the 77 countries fall into three fairly distinct clusters. At one end is a very tight cluster of developed countries, all with near-zero IM and IQ around 100. At the other end is a looser linear scattering of African (and a few Asian) countries with low IQs (below 80) and high IMs (over 50). Between these groups is a hotch-potch of Asian, Latin American, and a few European countries (including Russia) with moderate IM. The IQ of these countries is more variable, so the correlation within this group is rather weaker, with several ‘outliers’ a long way from the regression lines. But overall, the correlation between IM and IQ is almost embarrassingly strong.

Chicken or egg? Nature or nurture?

The correlation in itself doesn’t prove whether environment is influencing IQ or vice versa. It seems reasonable to assume that infant mortality as such does not directly influence the IQ of a population, at least in the short term (in the longer term it could conceivably have eugenic or dysgenic effects, if IM is selective w.r.t. IQ). Nor is IQ likely to be the main direct influence on national levels of IM. No doubt dim parents are more likely to have dead children, but I take it that factors such as water supply are more important in international comparisons (or comparisons across time - 150 years ago European countries had higher IM than African countries now).

We therefore need to look for models of indirect causation. It is easy to envisage both genetic and environmental models. An obvious genetic model would be on the lines of ‘low genetic IQ causes poor economic development and inefficient government, which causes bad hygiene, nutrition, etc., which causes high IM’. An equally obvious environmental model would be on the lines of ‘poor economic development causes bad hygiene, nutrition, etc., which in turn causes both high IM and low IQ’.

The correlation data alone can’t decide between these hypotheses. However, I argued in earlier posts that from what we know about the heritability of IQ we would expect to find a difference of up to 20 IQ points between developed and 3rd-world countries.

Perhaps the most persuasive argument is based on the Flynn Effect. It is generally acepted that mean IQ in western countries has increased by about 20 points since the 1930s. It is also generally accepted that the increase is wholly or mainly due to improvements in environment. It follows that if there are 3rd-world countries now in which the quality of environment is equivalent to that of western countries in the 1930s, then we would expect (other things being equal) the mean IQ in those countries to be about 20 points below the current western mean.

Unfortunately, there is no consensus on what particular aspects of environmental quality are responsible for the Flynn Effect. Entire books have been written on the subject (e.g. The Rising Curve, ed. Ulric Neisser, 1998). Personally, I like the argument of Richard Lynn that the main factor is improved nutrition. However, Lynn also accepts that longer secondary schooling has had an effect on the IQ of young adults, such as the Danish army recruits studied by Teasdale and Owen. Interestingly, in view of Lynn’s heavy reliance on Raven’s Matrices for international comparions, he also thinks that Raven’s is more susceptible to schooling effects than Wechsler - see Lynn in Neisser, p.212-3.

But whatever the reasons for the Flynn Effect, it is plausible that the relevant environmental quality in 3rd-world countries now is as bad as (or worse than) in western countries in the 1930s. If we take such indicators such as IM and life expectancy this is obviously the case. I therefore believe that many of the differences in IQ recorded in Lynn’s data have environmental causes. For example, the IQs of around 80 (against a UK mean of 100) in countries like India, Egypt and Guatemala are just what we would expect. I don’t know whether the lower IQs in most African countries can be explained in the same way, but clearly their environmental quality is even worse.

As I stressed in an earlier post, I do not exclude the possibility that genetic differences are also relevant. Genetic differences could plausibly explain some of the ‘outliers’ on the scattergram - countries where IQ is substantially better (or worse) than predicted by the regression on IM. For example, China has mean IQ about 15 points higher than predicted. This is partly because the IM figures are inflated by infanticide of girl babies, but even if we exclude girls the IQ in China is much higher than in other countries with the same IM for boys.

There are also a few countries - notably Jamaica, Barbados and Qatar - with IQ much lower than predicted by the regression. A genetic hypothesis may suggest itself, but I wouldn’t jump to conclusions. IM can hardly be a perfect indicator of overall genetic quality, so there are bound to be cases where IM is lower than one would expect from the general quality of environment. In Qatar, for example, oil wealth has probably paid for better sanitation and hospitals without necessarily improving education and literacy. And it seems from the UN tables that IM in the Caribbean islands generally is lower than in comparable mainland countries like Brazil and Guatemala. In Jamaica, for example, GDP per head is about the same as in Guatemala, but IM is only half as high. Maybe the legacy of colonial government in the Caribbean included relatively good sanitation and health care, or maybe the island location and climate has something to do with it. If the 'true' environmenal quality in the Caribbean is similar to that of Guatemala, then the IQs are not badly out of line with expectation.

But I would also want to be satisfied that the IQ figures are valid. The IQ recorded by Lynn for Jamaica and Barbados (72 and 78 respectively) is much lower than one would expect from the performance of West Indian immigrants to the UK (see here.) This could perhaps be accounted for by selective migration, but this cuts both ways - migration from the West Indies since 1950 has been on a very large scale, so if it has been selective, then the population remaining on the islands would have its mean IQ artificially depressed.

Anyway I’ve now exhausted myself, if not the subject. I’m aware that expressing some of these views on GNXP is rather like farting in church, but if you can’t fart in church, where can you fart?

Posted by David B at 12:42 PM