“The comparison of IQ scores of different nationalities is, at best, a hazardous enterprise, to be undertaken with caution and humility, and at worst, a nonsensical and mischievous waste of time” – N. J. Mackintosh.
From time to time writers on GNXP and elsewhere compare the IQ of populations in different countries, even those with widely different economic and cultural circumstances. As a notable example, Richard Lynn has produced a table setting out (alleged) comparative IQs for dozens of countries.
In comments on previous posts I have questioned these comparisons, but it may be useful to offer a more systematic critique for discussion. My main point is that differences in average IQ scores between developed and underdeveloped countries cannot be taken as reliably indicating any difference in innate capacity.
First, there are great technical difficulties in finding tests that are suitable for international comparisons. This should be obvious enough in the case of tests with a large verbal content, but is also true of non-verbal tests such as Raven’s Matrices. Such tests are highly artificial, and bear no resemblance to anything that the testees encounter in their daily life. In attempting to test illiterate or uneducated people (whether children or adults), it is often difficult to get them to see the point of the questions (Mackintosh, p.181-2, gives some amusing examples). Philip Vernon, who had huge experience of testing around the world, pointed out that sometimes ‘a sizeable proportion of the testees turn out to be ‘non-starters’… They are willing to try, and we have reason to believe that the test discriminates down to their level of ability, yet they just don’t get the hang of what they are supposed to do’ (Vernon, [1], p.101). Testees in this category either fail to answer the questions at all (in which case their zero scores will drag down the average), or they make random or near-random guesses. The latter is more dangerous, because in tests with multiple-choice format, and a scoring system that does not penalise wrong answers, even random guessing may produce an apparent IQ of 50 or 60. Jensen ([3], p.416), has pointed out that some African test results may be invalid for this reason. It is also well-known that ‘unsophisticated’ testees improve their scores substantially with greater test-familiarity. Vernon ([2], p24), concludes that ‘children (or adults) who are sophisticated or trained in tackling multiple-choice items, following instructions, and working at speed have on average about a 10-point advantage… over those who are unfamiliar with objective tests’.
However, even assuming that the technical difficulties can be overcome, my main point is that even a large difference (say, 20 points) in mean IQ between a developed country like the USA, and an underdeveloped country, cannot safely be taken as evidence of an innate, genetic difference.
Even the staunchest hereditarians do not deny that environment has some influence on IQ. Estimates of heritability (the proportion of total variance attributable to genetic factors) within modern western populations vary widely according to the methods used. In his recent survey Jensen ([3], p.446) concludes that reputable studies give estimates ‘that range mostly between .40 and .60 for children and adolescents, and between .60 and .80 for adults’. The evidence for higher heritability among adults is debatable (see Mackintosh, p.92-3), but for the sake of argument let us assume that after allowing for random error variance (correcting for attenuation) heritability is as high as .80, leaving environment to account for the remaining .20 of the variance. This still leaves room for a large influence of environment – larger than the figure of .20 may suggest to the unwary. If we conceptualise the quality of environment (QE) as a single continuous variable (which is admittedly somewhat artificial, but no more so than the IQ scale itself), the correlation between IQ and QE will be root-.20, or about .45. (See Jensen, [1], p.401, or [2], p.137.) This means that for every difference of one standard deviation in quality of environment, we will expect to find a difference of nearly 7 IQ points (assuming an IQ s.d. of 15 points), on purely environmental grounds. Thus, if we take a ‘very bad’ environment as 2 sigmas below the mean, and a ‘very good’ environment as 2 sigmas above the mean, then we will expect someone of average genetic quality, but a very bad environment, to have an IQ about 13 points below the population mean, and about 26 points below someone of equal genetic quality in a very good environment.
Of course, heritability estimates derived from one population cannot be automatically extended to different populations, or to a combination or more than one population. However, as Jensen has cogently argued, it is possible to draw plausible inferences from within-group to between-group heritability (e.g. Jensen [2] p.133-148). It seems reasonable to suppose that whatever elements enter into ‘quality of environment’, the QE in underdeveloped countries (e.g. Africa or the illiterate rural populations of many countries in Asia and Latin America) is usually at least 2 sigmas below the mean of modern western countries. Certainly this would be the case if we take such indicators as income per head, life expectancy, infant mortality, nutrition, parasite infection, or parental education and literacy. As Jensen himself has said ([3], p. 460), ‘on a scale of environmental quality with respect to mental development, these adverse environmental conditions [in third-world countries] probably fall more than 2 sigma below the average environment experienced by the majority of whites and very many blacks in America’. On this basis, we would expect average IQ in a third-world country to be at least 13 points below the western average, even if the genetic endowment is equal. And of course this is assuming a high heritability estimate of .80. If we take a more modest estimate of .60 (which is more appropriate for children or adolescents, who are the usual test subjects in these comparisons), then the correlation between IQ and QE would be root-.40, or approximately .63. This would give an expected IQ difference of about 19 points for 2 sigmas difference in QE, which is close to the difference actually observed between western and most third-world populations.
If this seems a very tortuous argument, there is a simpler and more direct way of appreciating the possible effect of environment. It is well-known that average IQ in most western countries has been increasing at a rate of at least 3 points per decade for at least 5 decades (the so-called Flynn Effect). The cumulative increase since the 1930s is probably around 20 IQ points. It is also generally agreed that the causes underlying the increase are mainly environmental (I’m aware that increased heterosis may also have contributed, but this can’t have been a major factor). It also seems reasonable to suppose that environmental conditions in the US or Europe in the 1930s were already at least as good as in many third-world countries today. It follows that a difference of about 20 points can be accounted for by environmental circumstances.
Of course, none of this implies that differences in IQ (if they are genuine indicators of relative intelligence, and not just the results of inappropriate tests), are unimportant. If people are stupid, this has implications for politics, economics, and society, regardless of why they are stupid. But if the causes are primarily environmental, this does at least raise the hope that economic growth and wider education will improve the situation.
References:
A. R. Jensen: [1] Educational Differences, 1973
A. R. Jensen: [2] Educability and Group Differences, 1973
A. R. Jensen: [3] The G Factor, 1998
N. J. Mackintosh: IQ and Human Intelligence, 1998
P. E. Vernon: [1] Intelligence and Cultural Environment, 1969
P. E. Vernon: [2] Intelligence: Heredity and Environment, 1979.
[Note: this is just a small selection of references. I have cited Jensen frequently because no-one, presumably, will argue that he is biased towards an ‘environmentalist’ position.]
Comments are closed.