Correlation between skin color and AFQT scores within and between African American families in the NLSY 97: a re-analysis

[Updated: I redid this analysis and I updated the file below. Here were the correlations:

While the small sample sizes preclude a definitive conclusion, the results seem to be in accordance with neither the hypotheses of strong colorism nor strong plieotrophy, both of which predict equal correlations between and within families. The results, rather, seem to be more consistent with either shared family effects and/or genetic effects (the later either due to ancestral genetic differences or to cross assortative mating for human capital and lighter color). With a large enough sample size, such as that in the Add Health survey, genetic effects could be disentangled from shared family effects by comparing the association across sibling kinships. A shared family hypothesis predicts an equality of associations; more specifically, it predicts that the correlation within families will be zero and that this correlation will not vary by genetic relatedness. Alternatively, a genetic hypothesis predicts a near zero correlation between full siblings due to independent assortment and predicts a magnitude of a correlation no less than that between families between unrelated siblings. As can be seen above, the results above are equivocal. A statistically significant difference in the magnitude of the correlations was found between full and unrelated siblings but the full sibling correlation was positive, though not significantly so. Generally, the results are ambiguous because it was impossible to separate in the majority of instances full and half siblings.]

The pithy: correlation within families: zero — except between unrelated pairs. Correlation between families: weak, but significant or trending towards — and at a magnitude little different from that found in the whole sample. Too little power to establish a statistically significant difference between the correlations within and between families. N sib pairs (available)= ~200, N sin pairs (needed) = 500. Method, data source, and computations explained in the excel file below. I’ll work on Hispanics another time and then see if I can combine results to get the critical N. If anyone has advice on methodology, I’ll listen happily.

I’m leaving as pending — for now — the previous post on this.

Updated spreadsheet: Within and between family sibling correlations between color and cognitive ability for African Americans in the NLSY 97

This entry was posted in Uncategorized. Bookmark the permalink.

3 Responses to Correlation between skin color and AFQT scores within and between African American families in the NLSY 97: a re-analysis

  1. statsquatch says:

    I missed your conclusion. Since the within family % admixture should be the same and the few genes associated with skin color are not causal for IQ this observation is consistent with a heridatarian hypothesis? If a correlation within families existed this would be evidence for colorism?

    • Chuck says:


      If the IQ-color association was the product of contemporaneous color discrimination, than the association should not be any less within than between families. Were there indeed no association within families, then this would imply that the association was due not to colorism but either to genes or to shared environmental factors (perhaps themselves due to historic colorism). The reason that a genetic hypothesis would lead to little within family association is that a large percent of families are comprised of full sibs and that there is independent assortment at least when one trait is not highly polymorphic. (The situation is somewhat different when dealing with two highly polymorphic traits (e.g., IQ and brain size). Because full sibs do differ in admixture and because assortment is not completely independent and because not all families are comprised of full sib, the within family association, though, should not exactly be zero by a genetic hypothesis. The best way to disentangle the results with respect to a shared environmental hypothesis or a genetic one would be to compare the correlation between full sibs, half sibs, and unrelated sibs. But to do so with any confidence one would need a much larger sample such as the restricted use Add Health survey N= > 17,000.

      Generally, I am not confident with my analysis above — I did it rather late at night — but I am confident about the general method employed. (Were I to redo it, I would use sample weights, look at PIAT scores, look at Hispanics, include birthdates to detect MZ twins, and so on.)

  2. Pingback: Spearman’s hypothesis and the NLSY97-ASVAB, part 2 | Human Varieties

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s