When are sociological studies going to learn about this modern \"genetics\" stuff?

June, 2006

Parental Use of Sunscreen and Child Skin Color In First Grade
Friedman, Patri

OBJECTIVE: The goal was to determine the relationship between the parental use of sunscreen products and the skin color of children in first grade.

METHODS: Data from the National Longitudinal Study of Youth were analyzed. Families with complete data on parental sunscreen use and child skin color were included in the analysis. Sunscreen use was categorized into "High, Medium, Low, None" by quartiles. Skin color was a continuous variable assessed by computer analysis of skin images. Multivariate logistic regression analysis was used to evaluate the relationship between parental use of sunscreen products and child skin color in first grade, controlling for gender, race, maternal education, income/needs ratio, marital status, parental income, and child behavioral problems.

RESULTS: A total of 2742 children, 81.3% white, were included in the analysis. Children of High sunscreen-using parents (n=1652) had an increased risk of being light-skinned, compared with children of No sunscreen-using parents (n=241, odds ratio: 45.2, CI 33.4-63.8). The same association held, to a lesser degree, between less extreme sunscreen categories. Of the covariates, only race was significant, but a substantial effect between parental sunscreen use and child skin color remained.

CONCLUSIONS. Among the 4 sunscreen use categories, High sunscreen use was associated with the highest risk of being light-skinned among young children. Understanding the mechanisms through which parental use of sunscreen are associated with skin-color risk may lead to the development of more comprehensive and better-targeted interventions.


The article that inspired this is here, reported by the mainstream media here. The study found an association between parenting style and child weight levels, with the implication of causation, and that we could reduce childhood obesity by getting parents to be less strict.

This particular example is far from the worst I've seen, but it pisses me off because sociology is still using this bogus pattern which I first saw pointed out by Judith Rich Harris. A researcher studies parents and kids, and finds that parental attribute A is associated with child attribute B, and then assumes or implies that A causes B. This ignores the fact that *any* study of normal parents and children is tainted by an *incredibly powerful* covariate, namely genetic similarity, which causes all sorts of non-causative associations.

Ignoring this lets researchers say things like "Violent parenting causes kids to grow up violent" (err...or maybe violence is genetic), "Being in a home with lots of books causes kids to grow up liking to read" (err...or maybe liking to read is genetic). In cases like the parenting style/weight study I cited, it's not quite as clear what the genetic commonality is. But the influence of genetics on behavior is so widespread and diverse that unless you aggressively brainstorm about and control for confounders, your research is worthless. For example, they should have controlled for parental weight, in case the actual causation is "Weight is genetic, and overweight parents tend to have an authoritarian parenting style."

But even if you do control for such factors, in order to actually find that A causes B, you need to eliminate the genetic effect by looking at twins reared apart, adopted kids, or doing a randomized study, all of which are much more difficult than just finding another one of the never-ending string of associations which stem from genetic similarity.

Share this

LOL. So very, very true.

LOL. So very, very true. Fortunately this sort of silliness is on the retreat, and for good -- nobody in the social sciences is going to expect to be taken seriously without taking genetics into account ten years from now.

Neither GP Vogler or Bruce

Neither GP Vogler or Bruce Sacerdote found any significant correlation between children and their adoptive parents on BMI. The shared environment showed next to no effect. I wonder if they bothered to mention this "complication" to their interpretation of the data in a footnote?

And AJ Stunkard: We

And AJ Stunkard:

We conclude that genetic influences on body-mass index are substantial, whereas the childhood environment has little or no influence. These findings corroborate and extend the results of earlier studies of twins and adoptees.

An exception appears to be this study from last month, but note this is for teens - by adulthood apparently no lasting parental effects.

Sociology research

Sociology research useless?!? I'm SHOCKED!

[...] ll Tuesday June 6,

[...] ll Tuesday June 6, 2006 Patri Friedman has a wonderful post with that title over at Catallarchy. He st [...]

This Modern Genetics

This Modern Genetics Stuff
Patri Friedman has a wonderful post with that title over at Catallarchy. He starts with a lovely little spoof and then explains his concern: that social science researchers are all too eager to state that correlation implies causation. To put that into...

Strict moms, fat kids Strict

Strict moms, fat kids
Strict mothers are more likely to raise overweight children, according to a new study in the June issue of Pediatrics. Madison's Capital Times reports: Strict mothers were nearly five times more likely to raise tubby first-graders than mothers who tre...

One problem with sociologic

One problem with sociologic studies is that they are usually based on retrospective observations, which are subject to huge biases. Improved results could be obtained by prospective studies but there are difficulties. For example, you could design a study in which kids were randomly assigned to be raised by strict or lax parents and at the end of the study you could weigh them. Sure, how are you going to do that?

Maybe you think understanding genetics will solve the problem? For instance take the Mozart Effect. It was said that listening to Mozart would make babies smarter. The Governor of Georgia proposed issuing Mozart tapes to all newborns. The legislature of Florida mandated classical music in state run daycare. http://en.wikipedia.org/wiki/Mozart_effect

Bull you say? Doesn’t this just show that politicians have low IQs?

Not at all.. Why, experiments in rats show that Genetic are involved. Researchers found that the rats who listened to Mozart had increased gene expression of BDNF, a neural growth factor, CREB, a learning and memory compound, and synapsin I, a synaptic growth protein, in their hippocampus.

So what? In my opinion the Mozart Effect is still BS. I listened to one of his piano concertos today on the way to work and it has made no appreciable difference.