Income Inequality, Race, and Population Health

post by Paul Kelleher

Last week, Kay H. Smith had a short op-ed in the Birmingham News wherein she says promoting population health...

...will require greater diligence as a citizen in demanding social policies that significantly address income inequality more than greater diligence in exercise and nutrition.

Smith is an MPH student, and she reports surprise when she learned in school that income inequality may have a greater adverse effect on population health than some other well-known culprits (e.g. unhealthful behaviors). I teach in an MPH program, and I also find that students are amazed to learn that income inequality itself causes poor health.

So you can imagine their skepticism when I--the program's link to the humanities--tell them it's not true. Ok, what I tell them is that there is very very little evidence that it's true. This is from a recent survey (pdf) on the issue written by Andrew Leigh, Christopher Jencks, and Timothy Smeeding, and which appears in the Oxford Handbook of Economic Inequality: 

[W]e argue that although there are plausible reasons for anticipating a relationship between inequality and health (in either direction), the empirical evidence for such a relationship in rich countries is weak. A few high quality studies find that inequality is negatively correlated with population health, but the preponderance of evidence suggests that the relationship between income inequality and health is either non-existent or too fragile to show up in a robustly estimated panel specification. The best cross-nation studies now uniformly fail to find a statistically reliable relationship between economic inequality and longevity. Comparisons of American states yield more equivocal evidence.

The "more equivocal" evidence referred to in the last sentence is actually utterly fascinating. It turns out that while the international correlation between income inequality and lower life-expectancy was an artifact of poor data (discussed further by Leigh, Jencks, and Smeeding, as well as by Angus Deaton here), the correlation remains for U.S. states and cities. However, according to two papers by Deaton and Darren Lubotsky, the correlation between income inequality and mortality disappears when controlling for the fraction of a metropolitan statistical area that is black:

White incomes are higher and black incomes are lower in places where there is a high fraction of blacks, and this between-race difference induces a strong positive correlation between income inequality and fraction black. Our regressions showed that, once the fraction black was controlled, income inequality as measured by the gini coefficient was no longer a risk factor for mortality. We regard this result as showing that there is no direct effect of income inequality on health.

Deaton and Lubotsky do not conclude that inequality does not have an adverse effect on population health. They deny only that income inequality is bad for health. They therefore do not take their findings to be at odds with the conclusions about status inequalities that emerge from Michael Marmot's Whitehall studies and studies on non-human primates. (Although Deaton has in the past questioned the conclusions Marmot draws from Whitehall, and at least one recent study suggests that the variations in health uncovered by Whitehall may have more to do with health behaviors than was previously thought.) 

Finally, Deaton and Lubotsky offer a tentative conjecture as to why the racial composition of a city should have an effect on its population's health:

We are more sympathetic to the idea that racial politics are important...In recent years, work by Bach, Pham, Schrag, Tate, and Hargraves (2004), as well as by the Dartmouth group (Skinner, Chandra, Staiger, Lee, & McClellan, 2005) has shown that America is running something of an apartheid healthcare system, in which most blacks are treated in hospitals or by primary care physicians that treat few or no white patients, and where most whites are treated in hospitals or by primary care physicians that treat few or no black patients. Bach et al. show that the physicians that primarily serve blacks--who may or may not be black themselves--have fewer resources and are less well-qualified. Skinner et al. show that mortality rates after an MCI are higher for all patients in hospitals that treat mostly blacks. In line with this work, our leading hypothesis is that blacks receive worse healthcare than whites, and that this spills over into mortality among whites who live in cities with a large black population and who share, at least in part, their inferior healthcare. This is indeed an important inequality, but it is not an income inequality.

If Deaton and Lubotsky are right, this could call into question two other common pieces of the MPH curriculum. First, it may be that the real racial problem in our health care system is not racial bias by providers, but rather black Americans' confinement to a health care system that's just not very good. Second, it seems that health care does make a difference to population health, contrary to some claims about the superlative importance of the social determinants of health.

(Lane Kenworthy has a related post on Wilkinson and Pickett's The Spirit Level here.)

Comments

[Daniel S. Goldberg]

There's far too much here to do justice in a comment to a blog post, so I'll just go with this one here:

I disagree with about 90 percent of this post. Maybe we can discuss sometime!

;-)

[Daniel S. Goldberg]

Can't resist:

I think it is simply false to say that there is "very little evidence" that income inequality does not cause poor health. I think there is quite a bit of evidence suggesting exactly this. What there also is is a tremendous amount of evidence suggesting the contrary, and also a tremendous amount of debate regarding precisely what the evidence does and does not demonstrate.

There are certainly legitimate grounds to question the causal relationship. But to make this claim is not equivalent to making the claim that there is very little evidence in favor of the causal link between income inequality and poor health.

Ok, I'll stop now. ;-)

[Paul Kelleher]

Sorry Daniel, but I have absolutely no reason to question these conclusions:

From Deaton's 2003 survey (linked above): "There is no robust correlation between life expectancy and income inequality among the rich countries, and the correlation across the states and cities of the United States is almost certainly the result of something that is correlated with income inequality, but that is not income inequality itself."

Leigh, Jencks, and Smeeding (quoted above): "The best cross-nation studies now uniformly fail to find a statistically reliable relationship between economic inequality and longevity."

See also the Kenworthy.

Would be delighted, obviously, the learn why they are wrong.

[Daniel S. Goldberg]

Hey Paul,

Hmm . . . the fact there are many epidemiologists who do in fact adduce significant evidence of a causal link between income inequality and poor health does IMO give one grounds to question the conclusions you endorse, even if those conclusions end up being right (which they may well be).

Sources include:

Kawachi and Subramanian (2004)
Wilkinson (2006)
Babones (2008)
Elgar (2010)

(Among many others).

You cite Kenworthy, but Kenworthy is issuing a critique of Wilkinson & Pickett, thus suggesting there is in fact evidence of a connection that is to be critiqued (if there wasn't, what would be the purpose of a critique)?

My point here is not to suggest that you must be convinced by this data. There may well be little or no association between income inequality and health. But to argue this is not equivalent to asserting that there is no evidence of such a connection. I continue to believe that the latter is simply false, and the citation of studies denying the connection does not by itself license the conclusion of its nonexistence (it might if there were no conflicting evidence, but my belief is that there is a great deal of such evidence).

That said, I think of the various points in your post, the most compelling and best supported (IMO) is the very one we are arguing, viz., that while social inequalities may determine health, income inequalities may not do so or may do so to a much lesser extent than other kinds of inequalities. So it is not this point that I disagree with as much, but rather the characterization that no evidence exists for the alternative.

[Paul Kelleher]

Daniel,

The three critiques I've relied on here all seem to me to raise serious questions about the putative evidence others have invoked to establish a correlation. Deaton, Leigh-Jencks-Smeeding, and Kenworthy all argue that much of the international data relied upon by others is actually problematic. (This is even before we get to inferences about causation.) Where the data are not questioned, the inferences from it are (--the Kenworthy strikes me as quite compelling here, and I do not understand your claim that by critiquing them, Kenworthy is conceding the validity of Wilkinson/Pickett's putative evidence).

And then there is the Deaton/Lubotsky study, which seems to me to be pretty conclusive about the US case. (I have only seen one critical article, but I find the reply to it by D/L--the second paper of theirs linked above--to be decisive).

Perhaps in the end I am biased, having found all of these authors in the past and on other issues to be quite fair-minded, without suspect agendas, and judged by their peers to be very reliable and sensible. Perhaps I'm also too much swayed by Deaton's perennial candidacy for the Nobel Prize for his work on the very methods one would use to suss out the causal links at issue.

All that said, I will happily take your advice and moderate the way I present my skepticism in the future, skeptical though I remain. I will encourage others to read the studies you and I both have identified, and I'll encourage them to judge for themselves---as I myself in fact have done and recorded in the blog post.

As always, thanks for the engagement on important issues.

[Brad F]

Paul
This abstract caught my eye in 2010, and I had a very interesting chat via email with author at that time re: findings, and further work they are doing on high income/low education folks and the converse--and of course, their outcomes.
http://content.healthaffairs.org/content/30/2/274.abstract

The exchange excluded, above thought provoking. I have seen several studies--and wish i could cite but cant recall sources--examining higher M&M rates in low SES neighborhoods in which minority and whites both suffer worse outcomes vs more affluent comparator regions.

Brad

[Brendan Saloner]

Hi Paul,

I know a lot less about this topic than I should since two of my main interests are income inequality and health, and I really should read the Deaton study you cite, but just from what you say in your post it seems wrong to control for racial composition in those state and city-level analyses. Racial inequality and income inequality are likely to be correlated, and conceptually they stem from related social processes (social exclusion, weak social cohesion, and historical injustices), so when you control for percent black your inequality coefficient will only reflect some residual inequality not soaked up by the race coefficient. I would interpret the finding as showing that most of the measurable income inequality's effect on health is mediated by racial inequality. I do agree with the medical apartheid hypothesis, but I would note that the few whites that end up going to the hospitals where blacks go tend to be poor and low educated.

What do you think?

[Paul Kelleher]

B,

I'll send you the papers and perhaps when you get a chance you can read them and let me know if Deaton and Lubotsky do not allay your concerns. As you know, I'm not the guy to be explaining this to you.

That said, you write, "I would interpret the finding as showing that most of the measurable income inequality's effect on health is mediated by racial inequality." Could that be all D/L are saying, as well? As quoted in another comment above, here is Deaton from his very comprehensive 2003 study (linked to in the post):

"There is no robust correlation between life expectancy and income inequality among the rich countries, and the correlation across the states and cities of the United States is almost certainly the result of something that is correlated with income inequality, but that is not income inequality itself."

D/L are after the causal claim, and if the correlation fails in all rich countries other than the U.S., and quote-unquote disappears when controlling for fraction black, then is it not safe to conclude that income inequality _itself_ is not the culprit?

[PlayBlue Judy]

This is a diverse and complex post and and I probably agree and disagree in equal parts but not really sure if its a racial problem in the health service or a socio-economic one.