Someone queried about the ability of IQ differences to explain differences in criminality. In a reply, “B.B.” cited an article by Richard Lynn; Lynn argued that much of the difference in criminality is left unexplained by IQ and proposed a not implausible hypothesis based on personality differences. As I see it, Lynn and others fail to take into account the larger dynamic and so, in this context, give short shrift to IQ.
John Hopkins sociologist Robert Gordon, now a member of The Project for the Study of Intelligence and Society, and previously Linda Gottfredson’s second husband, developed a neat population-IQ-model, which showed that g differences could account for most differences in criminality and other outcomes (e.g., single motherhood, HIV infection, poverty, conspiracy rumors). The paper won Mensa’s 1999-2000 Award for Excellence in Research, if that says anything.
The second level of analysis, the contextual, illustrated how probabilities of error (or success) are further modulated by the intelligence level of one’s near context, particularly family and peers. Explanations of differences in g-related individual behavior must thus take account not only of individuals’ own levels of g, but also the levels of the individuals who form their proximal environment. Functional advantages go not only to individuals who are brighter, but also to those who are fortunate enough to live among brighter individuals and, especially, to participate in the fraternite’ of acceptable reciprocal exchange with those individuals….
…If g is not to be sold short, care must be taken, when considering its effects, to qualify certain forms of intelligence data as applying only to the individual g of probands (as distinct from data reflecting, in addition, the local intelligence text, or, ultimately, the potential full effects of g in the probands’ population on prevalence of the outcome concerned). For example, the phrase supplied in brackets needs to be added to statements such as the following, which summed up the fact that Hermstein and Murray (1994) had reported unexplained residual differences between races in certain undesirable outcomes after controlling for individual IQ: “Given that IQ was equated for Blacks and Whites, it is clear that much more is contributing to differences in societal outcomes than just IQ [of the proband” (Sternberg, 1996, p. 15). Most of the residual differences that prompted this critical comment, when portrayed as percentage differences instead of ratios of two, three, or five to one, were in fact rather small, 8% or less. The largest percentage difference (and source of the 5: 1 ratio), 41% for out-of-wedlock births, obviously involved partners, not to mention wider contexts, whose IQs were not controlled (e.g., Dearden, Hale, & Woolley, 1995).
The recognition of a relation between intelligence and intermediate group-level phenomena brought us to the third level of analysis, the population level. Two well-studied populations, U.S. Blacks and Whites, were used to demonstrate population- level effects of differences in intelligence distributions. A population-IQ outcome model was described that incorporates the effects of g at both the individual and contextual levels, including now remote contexts within the population, such as spokespersons, elites, leaders, and chance encounters. The model was then applied to a wide variety of outcomes for Blacks and Whites in order to test whether population differences in IQ distributions can explain differences in prevalences of various social outcomes. Outcomes concerning juvenile delinquency, adult criminality, single motherhood, HIV infection, poverty, conspiracy rumors, and two forms of key opinion concerning the O.J. Simpson case were found to be commensurate with differences between Blacks and Whites in IQ distributions. The model thus accounts for the presence of good as well as bad outcomes in both races, by addressing successfully the important question of the difference in their relative levels within each race, a question that race per se has never been able to answer. Racial categorization itself, although useful for applying the model, is thus seen to have rough descriptive rather than explanatory value, as IQ appears to be a more fundamental variable. Subcultural values and attitudinal toleration of socially unacceptable deviance were also found to be related to average IQ. (Gordon, 1997. Everyday Life as an Intelligence Test: Effects of Intelligence and Intelligence Context)
The basic insight here is that the correlation between IQ and some outcome between populations need not be the same as the correlation within populations, because the impact of IQ can amplify in aggregate. For example, individuals with an IQ of 85 are slightly more prone to crime than individuals with an IQ of 100. But Populations with an IQ of 85, are significantly more prone because there is an institutional aspect. Such populations not only have more criminally prone individuals but also have individuals, and with them institutions, that are less effective at discouraging this behavior.