The Older Paper: g and the Social Gradient in Health

It seems it is Health&Medicine week here at the CoR (yes, there's another post on that coming up), which is surprising given that I'm not exactly an expert on the topic. Anyway, this post introduces a new feature: The Older Paper. One often reads about brandnew papers on blogs (if one reads the blogs I read, that is), but why not write something about older research if it's interesting? This is part of my general crusade against the blog norm to only link to stuff that is new. Next thing you know, I'm going to tell you about Pavlov's research on dogs.

As you probably know, lower-status people tend to be less healthy than higher-status people. Researchers call this the "social gradient" in health. The big question is why this is so. Linda Gottfredson offers her theory in a 2004 paper. Here's the abstract:

Virtually all indicators of physical health and mental competence favor persons of higher socioeconomic status (SES). Conventional theories in the social sciences assume that the material disadvantages of lower SES are primarily responsible for these inequalities, either directly or by inducing psychosocial harm. These theories cannot explain, however, why the relation between SES and health outcomes (knowledge, behavior, morbidity, and mortality) is not only remarkably general across time, place, disease, and kind of health system but also so finely graded up the entire SES continuum. Epidemiologists have therefore posited, but not yet identified, a more general “fundamental cause” of health inequalities. This article concatenates various bodies of evidence to demonstrate that differences in general intelligence (g) may be that fundamental cause.
Technically speaking, g is the first factor extracted from the results of an IQ test. Her argument is that people low in general intelligence are less well able to understand their doctor's instructions, have less health-relevant knowledge, etc. and hence display behaviours that are less healthy. The evidence is scant - it is primarily a theory paper, but I find this explanation more attractive than proposing some vague psychosomatic factor like "sense of control over one's life".

In principle, this isn't hard to test. Measure SES, general intelligence, relevant confounders (be mindful of overcontrol), regress both at the same time on a health outcome of interest, hope for low collinearity, and see whether the coefficient for SES is still significant. (If you do a cross-sectional study, make sure your dependent variable is not an illness that influences general intelligence.) If you want to, you can even develop a "sense of control" scale and include the results. (If you don't know how to do that, the psychologist down the hall does.)

My own guess is that g mediates the SES-health relationship partially, but not fully. I would think that other psychological factors include time discounting, risk-taking and conscientiousness. There is some evidence that lower-SES people show stronger time discounting, and I guess that there's a moderate correlation with the other two as well.

Anyway, interesting paper. Worth reading if you have an hour to spare - it's pretty long.


pj said...

Will read that when I get a chance. It seems to me that it is unlikely that IQ could explain the relationship between health outcomes and SES completely because things like income inequalities are also predictive of health outomes and these are unlikely to vary with IQ.

LemmusLemmus said...

Concerning income inequality, I would fully expect that to be negatively related to IQ: Poorer nations tend to have both higher inequality and lower measured IQs.

As for the effect on health, I once read an article in the Journal of Health and Social Behavior (2004?) that called the results of Wilkinson's cross-national studies into question. If I remember correctly, the author simply included nation fixed effects in the regression and the inequality effect disappeared. As inequality moves very slowly, one wonders about collinearity, though.

LemmusLemmus said...

Oh, look, here's the article; the same issue has more on the topic.

pj said...

I've never been convinced by the cross-national studies for precisely that reason, but, that study you link to is also pretty unconvincing. The question of fixed versus random effects is an interesting one, but the moderator variables in that regression are doing far too much work for me to believe that the result is anything other than overfitting - the use of moderators reflecting methodological issues makes drawing any conclusions impossible because he doesn't report or explore how these may be confounded with other factors.

The recent study into the role of parental education and MMR vaccination was very interesting (see the Lay Scientist), where they found that more educated parents were more likely to adopt the view that MMR was harmful (against the medical consensus), and hypothesised that this was due to them being more open to new medical information.

LemmusLemmus said...

I'm not familiar with the term "overfitting"; do you mean overcontrol? If so, I don't remember the article well enough to have an opinion on it and I don't currently have access to it.

Your link is interesting. I would also guess that more educated people use homeopathy more often. That's in part explained by cost. (In this country, most health insurances don't pay for homeopathy.) But I don't think that's all. Besides the point you mention, I think your typical middle-class person puts a high value on "being critical" and "not trusting authorities", whereas your average construction worker will just think ill > doctor.

Problem is, the people don't know how to be critical. Ask a random sample of university graduates whether they can explain Popper's views on falsification or why scientists perform randomized experiments, and you're in for a disappointment.

I've long thought that kind of thing should be taught in school. It's not that hard. Thankfully, I was taught this stuff informally by my father between ages 13-15 (ca.), and I had no problems understanding it.

Educated people also seem to put more weight on the feelie-goodie component, although I have no idea why.

With regards to Gottfredson's theory, the question of course is what the net effect of IQ differences is. She specifically says, as seems reasonable, that it will become more important the more cognitive load an illness entails (e.g., with some diseases patients have to monitor themselves almost constantly).

By the way, I've heard about this whole MMR-autism thing exactly once in Germany, I think. It doesn't seem to be a big topic over here.