Does Inequality Kill You?

It is often argued that inequality is bad for your health. Indeed, in Imagining Australia, my coauthors and I made precisely this argument, saying that one of the reasons that policymakers should be worried about inequality is because it makes you sicker.

In theory, there are several ways this might happen. If each additional dollar does less for your health, then moving a dollar from a rich person to a poor person might boost average mortality. If inequality boosts crime, then this might also drive up mortality. Or it could just be that watching the rich get richer drives you to depression.

Unfortunately*, the empirical evidence is less persuasive than the theoretical evidence. In the most recent contribution to the literature, Harvard’s Christopher Jencks and I use data on mortality and top income shares (a proxy for inequality). In our preferred specification, we find precisely no relationship.

It’s important to point out that 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 I attended recently put it, “it’s not just zero, it’s very zero”.

If I were coauthoring Imagining Australia today, I’d omit the claim that inequality makes you sick. Of course, this doesn’t mean we shouldn’t care about inequality. Probably the best reason to worry about inequality is also the simplest: a dollar brings more happiness to a poor person than a rich person.

Our paper is entitled Inequality and Mortality: Long-Run Evidence from a Panel of Countries. It’ll be out soon in the Journal of Health Economics.

* Andrew Norton has pointed out to me that this was a rather unfortunate use of “unfortunately”. Of course, I meant it’s unfortunate that the theory doesn’t survive, rather than unfortunate that people aren’t being killed by inequality.

This entry was posted in From the Frontiers, Inequality. Bookmark the permalink.

25 Responses to Does Inequality Kill You?

  1. harry clarke says:

    Its an unexpected result for me. I recently posted on the fact that Lancet had found that half the difference between the mortality of rich and poor was due to increased smoking by the poor.

    But you are saying no difference and half of nothing is nothing so smoking cant explain anything. This in itself sounds strange given the close relation between smoking and mortality and the fact that many studies find smoking to be an inferior good.

    I’ll read your paper – and follow comments here – with interest!

  2. Ben says:

    “……that policymakers should be worried about inequality is because it makes you sicker.”

    “If I were coauthoring Imagining Australia today, I’d omit the claim that inequality makes you sick.”

    Andrew be carfeul with your language. Mortality and sickness are not the same. Sickness might be more appropriately be called morbidity.

  3. Ben says:

    “be carfeul”.

    And I’ll be more careful with my spelling.

  4. Russell Hamilton says:

    Link to the paper isn’t working.

  5. Sinclair Davidson says:

    Congratulations! Well done.

    Not sure I’d refer to “theoretical evidence”, as opposed to ‘theoretical argument’, or ‘hypothesis’.

    For a good laugh along these lines, see the following comment.
    “There is a price to be paid for cutting the taxes of the rich, and it is to be measured in terms of a loss of health to the Australian population.
    Are there any good arguments against cutting income taxes? Yes there are. Such cuts, by increasing income inequality, are most likely to lead to increased mortality and morbidity in the Australian population.
    Before the government swallows the arguments of the CIS on this front and cuts the taxes of the rich, can we have a health impact assessment please?” (

  6. I’ll have a squiz at the paper over the weekend if I can but I have a busy social calendar, a trilogy by a new crime writer I have just discovered and with a forecast of 19C tomorow I think I’ll be doing the first BBQ of the season out the back, while listening to Denise on TWANG on 3RRR.

    Did you only look at mortality or did you look at morbidity too? You can have bad health without dying. Everything I know tells me poor $ = poor health. Perhaps all you are saying is that it doesn’t make you die any younger?

  7. I’m looking forward to comments from Harry (and Sam Ward). In fact if the two of them get going early enough I’ll cancel my BBQ and move in front of the monitor.

  8. Fred Argy says:

    I have not been able to access. Until I do, Andrew, I agree that the distinction between morbidity and mortality is important.

    My interest has been in the converse relationship – the causal link going from inequalities of health access to inequalities of income and opportunity. Surely the fact that poor people have better access to health in Denmark than in USA must help in part to explain the higher inequality and lower income mobility in the USA?

  9. Damien Eldridge says:

    Hi Andrew,

    Your argument for income redistribution here is somewhat controversial. It involves interpersonal utility comparisons (that an additional dollar to person A, who is poor, yields more “utility” to person A than the loss of utility to person B, who is rich, from losing a dollar) and hence rests on a moral judgement.

    It seems reasonable to suppose that many people have diominishing marginal utility of income, but that simply says that an additional dollar to A yields more “utility” to person A when he or she is poor than it does when he or she is rich. It does not say anything about comparisons between person A and person B.

    While I agree with you about the desirability of income redistribution up to some point (but would strict equality of incomes really be a desirable outcome?), this needs to be based on an explicit moral argument. Such an argument is clearly going to be controversial to at least some extent.

    Having said that, Martin Zelder once described an interesting efficiencey motivation for income redistribution (rather than an equity motivation). Imagine that you are a foetus. The story went something like this. You do not know whether you are going to be born to a ctack addict in Chicago or a rich couple in Beverly Hills (and you are not sure which would be better!!! 😉 ). If you are risk averse, you would like to buy some insurance against a bad birth outcome. However, because you cannot communicate with other foetuses, the insurance market fails to exist. The inability to communicate works a bit like high transactions costs. In this situation, income redistribution from rich to poor would potentially correct for the market failure due to the inabilty to communicate. Of course, while this story provides an efficiency justification for some degree of income redistribution, it is not clear how much income redistribution would be optimal.



  10. Andrew Edwards says:


    Just a question – I didn’t quite follow and I may not have the time to read the actual paper.

    What precisely were you testing for? Is the conclusion:

    (a) “countries with high inequality have the same mortality rates as countries with low inequality”? Or is it

    (b) “across a number of countries, we have found no evidence that being vastly poorer than most people in your society increases mortality rates”?

    Because it strikes me that the implications would be quite different between the two conclusions.

  11. Fred Argy says:

    Damien, you are of course right that interpersonal comparisons involve value judgments. But so does the Pareto efficiency test. It assumes that if the winners from a reform are capable of compensating the losers and still remain better off than before, then the community must be better off even if there is no actual compensation to the poor. By assigning the same utility weighting to everyone (rich and poor) the Pareto approach implicitly makes a value judgment.

    In short, we economists must inevitably make value judgments when advising on a policy reform which has both winners and losers, especially if the losers are poor. This being so, I prefer to assume that the poor get more satisfaction from each additional dollar consumed (and suffer more pain from each dollar foregone) than the rich. This assumption is hard to prove but it is consistent with common sense observation and has backing from some happiness surveys. Of course quantification is impossible but in my view Andrew Leigh is right – it is one reason why we should worry about inequality. But there are many others.

  12. harry clarke says:

    Australia’s Health 2006 has just come out and it does not confirm your findings:

    ‘In 1998–2000, the death rate for males living in areas of most socioeconomic disadvantage was 80% higher overall than for males living in areas disadvantage; and correspondingly for females the rate was 50% higher’.

    I would think carefully about your conclusions. Unless I am reading them wrongly they seem to defy a lot of accumulated knowledge.

  13. Damien Eldridge says:


    I agree that in making policy assessments, we are inevitably drawn into making value judgements. As I indicated in my email, I wholeheartedly support income redistribution, up to a point. However, there is a real issue about the extent to which income should be redistributed. If we took the logic to the extreme and said that whenever Person A is richer than Person B, we should redistribute income away from A and towards B, so that the outcome was complete equality in terms of income (or perhaps wealth), then the incentive to work would be greatly diminished. Determining the appropriate extent of income redistribution is a difficult (and controversial) policy issue. Do we currently do too little income redistribution or too much? I don’t know the answer to that question!!!



  14. Damien Eldridge says:


    Just a quick comment on the Pareto efficiency test you describe. It appears to refer to the Kaldor-Hicks criterion, which is sometimes known as the potential Pareto improvement criterion. The Pareto criterion itself would require that the compensation takes place, since a Pareto improvement only takes place if at least one person gains and nobody loses. When people tal about the impact of a policy on economic efficiency, they are most likely talking about the Kaldor-Hicks criterion. However, I suspect that it is very unlikely that economic efficiency is the ONLY criterion by which policies are assessed. Equity impacts are also likely to be relevant. (This is probably the case even if we ignore political economy issues!!! 😉 ) But you are right that the Kaldor-Hicks criterion makes a particular assumption about interpersonal utility comparisons and hence involves value judgements. Indeed, the Pareto criterion itself involves value judgements. It does not focus on distributional issues. A change that improved the happiness of people who are rich and left the poor exactly the same would qualify as a Pareto improvement. While some people might consider this to be a reasonable welfare conclusion, others might not.



  15. harry – i haven’t got the latest Vic Burden of Disease here – do you have it ? Does it do mortality by region / income in VIC?

  16. Fred Argy says:

    Thanks Damien for your interesting and detailed response to my comment on Pareto.

    Yes, I was using the Kaldor-Hicks criterion. It is the only version of serious relevance to policy advisers, which is why they usually look for the impact on aggregate wellbeing (GDP!) when considering a policy reform. In practice, governments can’t compensate losers whenever they decide to introduce a significant policy reform, if only because they can’t identify them. (That said, Howard did a good job of achieving a win/outcome on the GST).

    I agree that beyond a point redistribution fails the cost-benefit test and I too do not know when that point is reached. But I make two comments. First, when evaluating policy changes we are looking only at INCREMENTAL changes in income distribution. Given that Australia has a fairly unequal income distribution (we are in the less equal half by international standards), an economic reform which further increases inequality (e.g. the recent workplace and welfare reforms) needs to be treated with more caution than one that does not.

    Second, the METHODS (instruments) of redistribution are of crucial significance in determining the economic efficiency cost. In particular, a policy package which involves making health, housing and education more accessible to the poor and early intervention policies targeted at disadvantaged children is likely to be less economically costly (to third parties) than one which relies on passive welfare. Again, taxes which are “neutral” in their impact will generate less inefficiency than those which distort choices and incentives.

  17. Andrew Leigh says:

    Sorry not to have joined the fray, I spent my weekend on the left coast. Thanks for a bunch of really thought-provoking comments. Here’s a few random responses.

    * Link to paper now fixed – Mark, thanks for posting an alternative link in the meantime.

    * Andrew E, we spend a lot of time on the panel vs cross-section issue in the paper. I think you’ll find your answer more quickly from skimming it than having me rehash it here.

    * Harry, we’re not saying that poor people aren’t sicker. They are, and there are oodles of studies to back it up. Our point is that, holding constant average income, more inequality doesn’t increase mortality.

    * Ben, Fred – it is indeed possible that inequality makes people sicker, yet has no impact on mortality. I myself find this improbable, but our findings cannot rule it out.

  18. Fred Argy says:

    If I read you correctly, Andrew, you are signing your name to two propositions:
    (a) poor people are sicker on average than rich people;
    (b) keeping average income constant, more inequality does not make people sicker.

    Perhaps I am out sorts this morning but isn’t the second proposition logically inconsistent with the first? If countries A and B have the same average income but A has a higher proportion of poor than country B then, if poor people tend to be sicker, shouldn’t A have more average sickness than B?

    There are two possible missing links. First, could it be that A does not have more sickness because its health system targets the poor more than does B’s health system? The other question that we have not addressed in the debate (although you may have in your paper) is the direction of causality: health impacts on income but income also impacts on health. Do your findings perhaps suggest the former rather than the latter?

    Is all this a red herring and should I go back to bed?

  19. Andrew Norton says:

    Fred – They seem consistent to me. For example, in Australia’s own case more market income inequality has (thanks to progressive income tax) created a huge cash flow to the state, that has in turn largely been spent on health and social security, the former of which would have directly affected the health status of the poor.

    Also, the studies that claim that inequality of itself creates illness probably aren’t that sensitive to degrees of inequality within society overall. If you are the kind of person whose well-being is tightly linked to status, you are probably not going to feel that much different if people you don’t know are very very rich as opposed to merely very rich. I haven’t re-read the literature for a while, but I suspect it is localised and resented inequality, eg co-workers or neighbours, that is the main problem. They won’t be affected much by broad movements in inequality.

  20. Andrew Leigh says:

    Fred, I must confess that this is something I puzzled on for a while too. Let me give you my intuition, and see if it convinces.

    The easiest model via which inequality might affect mortality is if $1 has less health benefit for a rich person than a poor person. In that case, a mean-preserving transfer from a rich person to a poor person will lower inequality and raise average health. This is Figure 1 in our paper. I suspect this is indeed the way the world works, but that for the sample of rich countries we’re looking at, the curve so flat that $1 actually doesn’t help the poor more than the rich.

    We address the issue of reverse causality in the paper pretty simply: we find no effect, which rules out effects in either direction. Admittedly, the effects might be equal and opposite (eg. inequality makes you sick, and sickness makes your country more equal), but I’ve never heard anyone in the literature make such a case.

  21. Sinclair Davidson says:

    Having now actually read the paper, I have a few thoughts to share. First, of all, I should acknowledge that the paper is forthcoming (well done), so my comments won’t make the paper any better (or worse), and also I’m not a health economist and am not familiar with the ‘inequality makes you sick’ literature. I can say, I always thought being sick was a function of germs, viruses, and life-styles choices. To the extent that life-style choices lead to poor health and also lead to low-income, it seems that the literature suffers from an omitted-variable bias problem. I’m sure others smarter than my good self have discussed this in the literature. The point remains that I’m not surprised by the results. There is no sound (my value judgement) basis for believing that inequality leads greater mortality, and the paper shows this to be the case. Apart from inequality impacting violent crime, the health outcomes described in section 2 appear to be psychosomatic (envy causes poor health? – there could be an argument for envy causing depression, I suppose).

    Lets accept, for argument sake, that income inequality leads to poor health. It doesn’t follow that income inequality leads to poor access to medical care. By this, I mean that the doctors and nurses attending ‘the poor’ have access to the same stock of medical knowledge, as do those attending ‘the rich’. Indeed, they may well be the same doctors and nurses. The flow of medical services might differ. The rich might be in 5-star private hospitals and the poor might be in public wards. The rich might get better quality drugs (would they?) while the poor might get generics (I really don’t know).

    The model under consideration seems to imply: Low income leads to poor health leads to poor access to medical care leads to death. My argument is that the ‘leads to poor access to medical care’ portion in the link breaks down – at least for the sample of economies considered in the paper. In order to test the Inequality – Heath hypothesis, a sample of economies where low income leads to access problems need be found. This, to my mind, implies that income inequality in the face of absolute poverty is the underlying problem, not inequality per se. (It also suggests a time-varying component to the analysis – has the relationship between inequality and health changed before and after WWII, for example, as the modern welfare state cranked up. Probably not, but perhaps in the 19th century).

    Turning to the paper itself, the share 10 variable is pre-tax income. Surely, the more appropriate measure would be post-tax income. Minor quibble. The life expectancy variable may be too aggregated to find anything. What about the standard deviation of that measure? Is it available? Why isn’t the hypothesis, the higher the income inequality, the higher the standard deviation of life expectancy?

    Income share is included in the regression analysis, but not income share – squared. Why not? Implicit in the paper is the notion that inequality is a negative externality. But, especially in health, it could be a positive externality. High-income individuals can afford very expensive treatments that over time become available for the poor. (This hypothesis may or may not be consistent with my argument above about stocks and flows of medical care, but can be easily tested). (For completeness, let me also say that I suspect that externalities are extremely rare).

    Silly econometric question: Have you tested for unit roots in the time series data? Or is this taken care of in the panel data specification?

  22. harry clarke says:

    OK understand Andrew – I should have checked your paper!

  23. Andrew Leigh says:

    Sinc, thanks for the comments. As to our ineq measure, I agree that post-tax would be better… if we had it for all countries. Unit roots are handled not by panel specifically, but by clustering standard errors at the country level. I found your discussion of hypotheses interesting, and if you’re keen to pursue the topic more, I can highly recommend Angus Deaton’s 2003 JEL survey.

  24. Pingback: Econ-Atrocity / Econ-Utopia » Blog Archive » The Inequality and Health Debate: What do we learn from the twentieth-century in the developed world?

Comments are closed.