« DeLong's Ad Hominem Against Libertarianism | Main | Is it legitimate for your company to pressure you to be healthy? »



Feed You can follow this conversation by subscribing to the comment feed for this post.


Hey Bill, I did a whole literature review on Health Disparities last year. Mostly looking at the impact of DTC advertising on them. From what I recall David Cutler has some good stuff on this. I can also send you my works cited, a lot of articles there discuss the gradient I just can't remember the specifics.

Bill Gardner

thanks -- I would love to see it.

Daniel S. Goldberg

Hi Paul,

Thanks for posting on this important topic. I read this article with intense interest when it came out, and I think it's important to consider a few items here.

First, I admit to some confusion that I still have not been able to resolve given the finding in Whitehall I that even combining the mortality contributions associated with risk factors such as smoking, high cholesterol, and hypertension do not explain more than a third of the total mortality burden of the occupational class with the highest mortality. I confess that I do not understand how to square this finding with that reported in the 2010 article regarding the signal importance of such health behaviors in Whitehall II (cholesterol and hypertension are risk factors in their own right, but are epidemiologically linked to certain behaviorial health risks).

Beyond my confusion here, I also think it is critical to read this study alongside James Dunn's excellent editorial in the same issue, where he notes several important points:

"Perhaps most important, the study by Stringhini et al1 does not suggest that socioeconomic differences in health are reducible to socioeconomic differences in unhealthy behaviors. Accordingly, it would be incorrect to infer that there is no need to be concerned with social and economic justice, only health behavior."

As a justification for this conclusion, he notes that in fact the subject population of Whitehall may not be all that representative of British society in general, and, more importantly, in my view, rejects the dichotomy that posits as explanations for the social gradient of health behavior on the one hand vs. low SEP on the other. Rather, he suggests, what links them together is the allostatic load hypothesis and the key role of stress in creating a unified pathway that includes both risky health behaviors and low SEP on the same causal track that produces highly unequal distributions of health.

Whether he’s right or wrong, I think the editorial comment gives a lot of helpful context for the study you cite. In any event, as Dunn acknowledges, the evidence is strong that a number of risky behaviors are disproportionately prevalent among marginalized and disadvantaged social groups. So either we conclude that poorer people are lazier, more promiscuous, and more stupid than the rest of us, which is absurd to me, or we understand that something about the social and economic conditions in which poorer people live makes healthy behaviors significantly more difficult across populations of disadvantaged persons. I cite some of the evidence for this in a recent paper (http://www.ea-journal.com/art3.1/Goldberg-Population-health.pdf), but my sense from examining the literature is that it is very easy to find.

As ever, I’m very interested in hearing your perspective on any of this.

Paul Kelleher


With regard to your first "confusion" (your word), isn't this addressed by Stringhini et al? They claim that health behaviors accounted for only 1/3 of the gradient in the original Whitehall studies because the instruments used to assess the role of behavior were inadequate. Indeed, that's their main thesis (well, that combined with their claims about the true effects of health behaviors once one uses the right measures).

The other points you highlight seem to be correct. As I say in the post, nothing in the study entails that the distribution of health behaviors is not itself the result of underlying SES patterns.

Daniel S. Goldberg


Just reread the article, and yes, of course you are correct (although I'm less sure they attribute the difference to inadequate instruments as much as to inadequate design -- i.e., that one needs to use repeated measures to obtain the correct results regarding the extent to which behaviors mediate the SES-health connection). Sorry for the "confusion."

On the off chance you have not seen the follow-up study, which includes Marmot as an author, here's the link:


Paul Kelleher


I think I must've meant "statistical instruments," and I'll be the first to admit that I'd be the last person on earth who'd know how to use that phrase accurately!

THANKS for the link to the follow-up study; I had not seen it, and I'm REALLY excited to read it, especially since Marmot is a co-author. That suggest to me that the conversation I was hoping *would* happen *is* happening.

Paul Kelleher

Wait. Now we have the following situation:

1. Marmot is quoted in October 2011 as saying "[Lifestyle factors] accounted for about a third of the gradient."
2. Marmot is co-author of a Feb. 2011 paper that says, "Overall, health behaviors explained 75% of the association between occupational position and all-cause mortality in the Whitehall II study."

Isn't something amiss here?

Daniel S. Goldberg

"Confusion" abounds, I guess!

Paul Kelleher


Daniel S. Goldberg

Of further possible interest:


Brad F

Paul (and Daniel)
I remember reading the updated PLoS paper:

A excerpt that might clarify a bit, and reconcile confusion (ref 29 is the JAMA 2010 paper):

Results on the Whitehall II study reported in the present paper differ slightly from those previously reported [29], as we harmonized some measures in the present analysis to allow better comparison with the GAZEL cohort. In particular, the diet variable was modified to only include data on fruit and vegetable consumption. Our previous paper also included data on the type of bread and milk consumed, but as these were not available in the GAZEL study, the measure of diet was simplified for the present analysis. The measure of smoking was coded as current or not current smoker in the present paper (not as current/ex/never smoker as in our previous publication) to obtain an identical measurement with the GAZEL data. The harmonization procedure led to less missing data at baseline in Whitehall II (537 individuals compared to 707) compared to the previously published paper.

Whether this is on target (bread and milk?), cant say, but they acknowledge the disparate findings.



Hey all I meant to come back to this but there is a good NBER paper that appears to really hit the nail on the head of all the gradients.
Its called the economics of risky health behaviors:

The comments to this entry are closed.

My Photo

Become a Fan