Tuesday, July 22, 2008

For smart people, does IQ matter less and social class matter more?

Via Steve Sailer, I read Half Sigma's theory about how your career success depends on your parent's socioeconomic status the most when you have an above average intelligence:

IQ is more highly correlated with life outcomes for people with below average to average IQs. Most career tracks have an IQ floor, and if your IQ isn't high enough to meet the floor level, you can't perform that job adequately. Few career tracks have IQ floors much higher than 115, so if your IQ is higher than that, your parental wealth and connections become very important.

Thus, the higher your IQ, the more important the wealth of your parents becomes (the very opposite of what most people think). People with exceptionally high IQs but inadequate parents often have poor life outcomes because of the mismatch.

I'll use the General Social Survey ten question IQ test as well as its father's socioeconomic status index (PASEI) and the respondent's income to assess the hypothesis. The 100-115 range seems to be a grey area, so I'll compare the group below that with the group above. To be precise, the two cutoffs I'll go with are 98 and 112.

My reasons for this are two-fold: 1) there are only 11 possible scores on the IQ test (0-10 out of 10 correct) so there are only 11 scores to choose from as cutoffs--98 and 112 are two of those possibilities; and 2) the highest score is only 125--the next lower score is 118, then 112, but I don't want the cutoff to be 118 since that gives very little variation within the high-IQ group.

OLS regression coefficients are shown above (N =1,335). First, the effects are surprisingly weak. I suspect that this is due in part to the truncated IQ measure--the high-end variation is all squeezed together--and the truncated measure of income (the GSS focuses on low income folks, and lumps all those making $110k or more together).

The Betas indicate which indepedent variable has the greater influence, IQ or father's SES. For the low-IQ group, IQ is close to being as strong as social class, and both effects are statistically significant. In the smarter group, IQ drops to non-significance. The effect of social class is weak, but it manages to teeter on statistical signficance at the .05 level.

So, Half Sigma's idea is supported here, but I wouldn't be surprised if the results were stronger if we used occupational prestige instead of income since it is more normally distributed. Maybe I'll look at it later if I find time.


  1. See the graph in this Bruce Sacerdote adoption study with a full range of family incomes ($10,000 - $200,000). The adopted children are the blue line. Social class of parents only correlates with your adult income when you share genes with your parents. (red line)

    At this point there is no evidence that having rich parents boosts your adult income, and there is evidence against it.

  2. JM says: At this point there is no evidence that having rich parents boosts your adult income, and there is evidence against it.

    I agree. The longitudinal studies which have records of an IQ test on children and follow them into adult life to examine their employment status show the same story: in the modern world, we live in an 'IQ meritocracy' - IQ is 'destiny', not social class.

    For example


    where the author could find no effect of social class on the outcome of children.

    Indeed there is a British Journal of Psychology study by Duff and Thomson from the early 1920s which - although it does not perform the precise analysis - suggests the same story.

    And there are apparently no ceilings or floors on the IQ effect, since you see an IQ advantage at very high levels in the gifted children follow-up studies by Lewis Terman and later by David Lubinski.

    It is interesting that people find this the primacy of IQ over social class so hard to accept - even Half Sigma.

    After 100 years of proving again and again that IQ is more important than social class, I think we should move on and *apply* the knowledge rather than continually challenging the knowledge by generating ever more specific, subtle and partial post hoc hypotheses trying to show how social class is the really important thing.

  3. HalfSigma's quite a Social Status obsessed guy - witness his bent on Law schools and prestige or Computer Science and Prestige every couple of articles.

    I do not think hes near unbiased on the issue. Still, his blog makes for an occasionally good read.


Meta-analysis of clinical trials: Eat walnuts

I am always looking for easy eating choices that are good for you. This new meta-analysis of 26 clinical trials looked to see if walnuts ma...