I found this article interesting, if for no other reason than people seem to be misunderstanding what it says and what it does not say.
The article by Leigh and Jencks for the Kennedy School of Government is entitled "Inequality and Mortality: Long-Run Evidence from a Panel of Countries." It compares the inequalities of income distribution in countries around the world with measures of health. The measure they use for income inequality is the percentage of income recieved by the top 10% of the income distribution. The measures of health that they use are infant mortality and life expectancy.
Here are the relevant graphs (click to enlarge):
Neither trend showed statistical significance, nor are the effects very large. As the studies author describes it:
...not only are our coefficients close to zero, but our standard errors are small enough that we can reject even modest detrimental impacts of inequality on health. As one participant at the NBER meetings ... put it, "it's not just zero, it's very zero".
These results are profoundly counterinituitive for people because we know from a variety of work that low income results in poor physical and mental health. For example, comparing the 25 percentile with the 75 percentile for income and looking at cardiovascular mortality risk shows that the reduction in income results in a relative risk of 1.32. With respect to mental health, the rates of new onset depression (1.26 relative risk) and for the persistence of depression (2.08 relative risk) are higher for individuals of low socioeconomic status. (There is much more data on this; this is just a sampling.)
Why the disconnect? Why is it that health would be associated with income but not income inequality? Well, first I should note that income inequality is associated with poor mental health. Pickett et al. show a clear positive and statistically significant association (click to enlarge):
There are strong, positive linear associations of GNI per capita with any mental illness (r = 0.80, p value = 0.02), and with serious mental illness (r = 0.89, p value
Second, it may be that the issue in health is not the comparison of your income to that of those around you, but rather the absolute value of income that relates to your health. Pass a certain level, and you will only have incremental improvements in health. This data, at least with respect to the end points they measure, would appear to support that notion. It goes against earlier theories that had posited that the source of inequalities in health were social stressors that fell predominantly on the poor. These social stressors required you to see people around you who were doing better. If we believe this data, it looks like the social stressors hypothesis could still be reasonable with respect to mental health but is not with respect to overall health.
It is very important to distinguish between income inequality and average income because the difference relates to the proposed remedies we might select. On the one hand, you have some liberals who would argue that the most direct method to improve absolute income is to reverse inequity of income by policies of redistribution. On the other hand, you have conservatives who would argue that increasing the average income would be just as effective if income inequality is not in and of itself unhealthy -- hence policies that promote growth should be forthcoming. They might also argue that with respect to mental health we should attack the origins of the social stress that results from inequality -- the resentment and frustration -- rather than the inequality itself.
I am not interested in debating the wisdom of either set of policies. I merely point out this article because when we consider as a society what we are going to do about the health of the nation's poor we need to consider the idea that -- counterintuitive as it may be -- income distribution may not be the problem. In contrast to the data related to mental health, this data suggests that redistribution may not be effective in improving health if it does not also raise the average income.
Something to consider the next time you argue around the dinner table.
Hat-tip: Tyler Cowen.
- Log in to post comments
Here's a graphic for a larger number of countries.
But perhaps this is a more appropriate graphic since it addresses "it's income, not inequality" with the same set of countries as the Leigh and Jencks paper.
I wasn't aware of that data, but I would add that they do argue that their data is similar to data that could be obtained with using the Gini coefficient as the measure of inequality:
As you can see they concede that measurements of inequality are not always as good as we would like them to be, but this one -- the percentage share of income owned by the top 10% -- is as good as the other ones available.
Yeah, there's a fairly high correlation among most measures of inequality so for big macro level views it doesn't matter that much which one you use -- I used the Gini coefficient simply because I already had the data.
My point was really about the second graph; if, based on the evidence given in the Leigh and Jencks graphs, you were convinced that inequality doesn't matter then you should be equally convinced by the second graph that income itself doesn't matter.
So, are you?
With respect to inequality, yes I don't think it matters for life expectancy no matter what measure you use. Granted for the graph you posted the trend line is at least in the right direction but the points are so all-over that I doubt it is significant.
I still do assert that income matters with respect to health, though I admit the life expectancy graph that you posted doesn't show it. There are a variety of measures where there is socioeconomic gradient for mortality -- one being the one I posted, cardiovascular health -- and I find them convincing. I am not certain why this wouldn't translate into a macro effect of life expectancy, but there may be the same issue of not choosing a measure of inequality that really captures the differences in life expectancy.
Jake wrote:
Well, you can doubt it but a MuD/PhuD should really check. The life expectancy data are here and the Gini coeffs are here. You can calculate the significance of the relationship yourself if you want, but when you do you're going to find that "the points that are so all-over" show a pretty significant relationship.
I, too, assert that income matters to health. My point is that when you're looking for small effects you have to use sensitive instruments and looking only at average life expectancies and income inequalities from relatively high income countries, as was done in the Leigh and Jencks paper, is a pretty blunt tool. Since that approach makes even a strong relationship (like that between income and life expectancy) hard to see, you shouldn't be particularly surprised that a weaker relationship (and everyone agrees that, even if it exists, the one between income inequality and life expectancy would be weaker) would also be hard to see.
P.S. I suspect (though I hasten to add that I do not know) that the problem isn't one of choosing the right measure of income inequality; it's that, in this context, life expectancy is a pretty blunt tool for measuring health.