• Zombie life expectancy arguments

    Paul Krugman wrote a nice piece yesterday on life expectancy, specifically on how it hasn’t been going up like “everyone” says. It’s a zombie idea; it won’t die.

    I thought I’d add to his post by bringing back some of my favorite charts on the subject.

    1) Life expectancy at 65 has not gone up nearly as much as life expectancy at birth. That’s what matters. That’s what determines how many years of Medicare or social security you might get:

    2) Life expectancy at 65 has not gone up for blacks as much as it has for whites.

    3) Life expectancy at 65 has gone up much more for people in the top half of earners than in the bottom half.

    4) While the richest 25% of Americans compare favorably to any other country, those in the bottom 50% of earners don’t. They will live fewer years after age 65 than the average person in most OECD countries. I don’t dispute that disparities might exist in other countries, but they definitely exist here. We’re not all living longer equally:

    5) Speaking of which, there are a number of places in the country where life expectancy is dropping:

    It’s important to think of all of these things when you claim we need to raise the retirement age because “we’re all living so much longer than we used to”.


    • Ok, this wasn’t referring to the type of “zombie life expectancy” I thought it would. I clicked expecting humor, and instead found education, the horror! 🙂

    • Is it possible to produce the last chart using “expectancy from age 65”. (I assume that because it doesn’t say, that it is using expectancy from birth).

      It seems that if we’re going to compare each of these charts, they should all be displaying the same data.

      • Wish I could, but I don’t have the data.

        • Health, United States does give us change in infant mortality rates by state, in table 18 on pg 101.


          Glancing at the table, infant mortality has declined in most states since 1989. I didn’t go through for every single state, but KY, TN, MO and GA, which have a lot of red spots in the map above, have definitely registered declines. OK is flat, oddly enough. Either way, if infant mortality is declining or constant, and overall life expectancy is decreasing in large swathes of some of those states, then it seems likely that life expectancy at age 65 is the driver. Dr. Carroll can correct me if this is incorrect.

          That said, I grant that there are probably some counties where infant mortality has risen.

          In an actual zombie apocalypse we would be running for our lives and life expectancy would plummet. It seems like health services researchers don’t tend to have military experience, which I assume would be the largest predictor of survival if anyone was alive to run a regression.

          • @Weiwen (hopefully this reply attaches correctly)

            That was more or less exactly the kind of information I was looking for.

            I was mostly curious because it would be disappointing if the reason that LE was going down in those county/states was because of infant mortality going up, and then we used that as a reason to lower retirement age 🙂

            Is there any correlation between the counties with lower LE and obesity?

          • My wife’s a demographer and prof at UCLA SPubHealth. I have had mortality stats thrown at me for years, so these came as no surprise. But the amalgam of staff running regressions while running for their lives from the hallowed and zombie-infested halls of the school sent us both laughing, while she upgraded Stata and I checked our apocalypse supplies. Btw, where do Zombie Infants fall on the mortality charts?

    • The graphs here seem to reinforce the notion that people are different–by race, income, and geography. On the other hand, government “solutions” tend to be uniform–Social Security works the same for everyone, I believe it’s only dependent on income. The funny thing is, the higher your income, the more Social Security you receive.

      Isn’t this series of charts a great argument to radically reform Social Security and make it more targeted? The debate shouldn’t be about whether we should raise the retirement age, it should be whether to make this more means-tested.

      • There could be a argument to make Social Security means tested, but you have to ask at what cost in administrative complexity and what cost in politics.

        One, most Americans don’t have a traditional pension. We don’t tend to save enough, and there’s no guarantee that we won’t outlive our savings. However, you won’t outlive Social Security. You might die before collecting many benefits, but that’s a risk you have to take. People are pretty resistant to being forced to purchase annuities with their 401k savings. If the price is that we keep paying high earners their Social Security benefits, that’s acceptable.

        Two, Social Security is proportionately less generous to high earners. They gain the insurance benefits (i.e. they won’t outlive their Social Security benefits). However, they get back proportionately less than they paid in. They cross-subsidize low earners, who get back more than what they paid in.

        Three, what cut-offs do you select? How do you administer this? You do add complexity to the system. And it’s not a given that you’ll save very much – maybe only the top 20 or 10 percent of Americans arguably could live without Social Security benefits.

        Four, if you make high earners pay in to the system, but they don’t get to benefit at all from it, they will be less likely to support it politically.

        In fact, Social Security is NOT in crisis. If we hiked payroll taxes by two percentage points we would probably take care of the payroll tax deficit. Yes, that’s a regressive tax and we should be cautious about doing it. But our long-run budget problems are mostly in health care. Why the unearthly fascination with cutting Social Security?

        • 1. For starters if all retired people got the same amount of money from SS we could save money and provide better for low income earners.

          It might be even better to replace SS, AFDC, Food stamps, subsidized housing with a basic income guarantee.

          1) Life expectancy at 65 has not gone up nearly as much as life expectancy at birth. That’s what matters. That’s what determines how many years of Medicare or social security you might get:

          Both matter for SS and Medicare (if someone dies at 64 for example). Surely life expectancy at birth matters more overall that is apart from SS and Medicare.

          3. I bet that the areas with falling LE are experiencing significant out migration of their most capable people.

        • The right in the U.S. opposed Social Security when it was proposed by Roosevelt, calling it “Socialism.” The American Enterprise Association was founded in 1937 (becoming the American Enterprise Institute in 1943) by libertarian businessmen who wanted to roll back the New Deal, including Social Security. AEI and its allies have been attempting to kill all the social welfare programs in the U.S. since then. Recall the attempts to “privatize” social security under Bush.

          Secondly, the Tip O’Neil/Ronald Reagan deal that “saved” social security in the 80’s essentially raised FICA taxes to produce a surplus in the federal budget that allowed Reagan to reduce taxes, especially on high incomes. In effect, higher taxes on labor funded tax reductions on capital gains, dividends, etc.

          In the claims of some right-wingers, the debt owed to the Social Security Trust Fund is not the same as debts owed to other creditors, but can simply be considered as a promise that can;t be kept. Eliminating or reducing the “entitlement” of Social Security furthers their long-sought objective of rolling back the New Deal.

    • Does any of this take into account things like murder rate and auto accidents? Most of these charts can’t really inform healthcare policy if what we’re really looking at is higher murder rates/crime rates/accident rates for minorities and lower income groups, right? There are social implications there, but you’d need to strip out all of that to isolate health care.

      • That’s a decent argument for young people, but homicide isn’t that big a problem in the over-65 crowd…

      • Not all public health issues are health care issues. And while increased life expectancy no doubt is linked to better health care outcomes, it is also no doubt linked to declining murder rates (which are significantly lower than two decades ago), increased auto safety standards and the like…

    • That last chart gives me the willies. The cross-sectional presentation, I would argue, tells us little about the time-series. What impact does migration of older, lower-life expectancy from their cohort, people have on this? Those patterns look eerily like migration patterns.

      Further, I still don’t see how any of the above is not an argument against making the retirement age higher. How about this: supporters of Medicare and SS tell us how many years of guaranteed retirement “we” should be paying for?




      Remember that we’ve pushed the time later and later for when people enter the labor force – say into the mid-20s, and we provide education in massive quantities to most of this time. So we get 20+ years “free” at the front end and 20+ years at the back end. That is not affordable. Or reasonable.

      Or why not life-expectancy “means-tes” the benefits. Surely we know which groups are more or less likely to live longer. So raise the retirement age for really rich white college graduates, and lower it for inner city women. What could be the objection to that?

      • OK. How do you do this? By income or assets? OK, what they fluctuate? If you had high income and assets during your working years, but you lost most of it in retirement, what do we do?

        Would it even be legal to consider race, geographic location, education, etc in assigning benefits? Is it worth the administrative complexity? And the food fights?

        And what do you do for rich White folks whose degrees end in D but they have an accident or a severe illness and have to end work early?

        Social Security’s modest shortfall could be solved by hiking the payroll tax, or making it apply to more income, or changing to the chained CPI-U (which produces a modest benefit cut over time, as it grows slightly slower than the regular CPI-U). Why are people obsessed with raising the retirement age? We could fix Social Security without too much fuss but for all the food fights (which imo are stupid). It is Medicare we should be worried about.

      • We’ve pushed the time later because good jobs are not available for young people….

      • “Remember that we’ve pushed the time later and later for when people enter the labor force – say into the mid-20s, and we provide education in massive quantities to most of this time. So we get 20+ years “free” at the front end ”

        This is an old chestnut that needs to be crushed. Going into higher education (i.e. postponing entry in to the labor market) is primarily an economic decision. People – both individual and society – calculate that the earnings lost during the education years are (more than) outweighed by the increased productivity once work starts. So the longer time before work does NOT reduce the affordability of any given year of retirement.

        Ah, you say, but there isn’t any productive work for these new graduates so the investment is lost. PLEASE. If there is no productive work for people in their 20s (which is not really true) then why do you want to keep the oldies at their desks? Surely they should be moving out to make space for the new, energetic, generations.

        By the way, this is pretty much ‘lump of labor’ theory/fallacy so you don’t need to believe the above – either side – but if you do you should take both sides, as it were.

        • Not so. Many young people both work and go to school, and that’s more true now than it was when I was in school 35 years ago. Most young people have to work more hours, as those who don’t have family support have to work more hours to pay rent etc.

      • One good way to means-test by life expectancy would be to index the retirement age to lifetime earnings. If you have been a low earner you could get full benefits at 65; high earners would have to wait till 70. If you’re a healthy and happy hedge fund manager who loves his cushy desk job, then you can stay at work for years and years. If you want to retire earlier, then you can finance it yourself. But if you’re a cashier at Walmart and your knees are shot, you deserve a much earlier retirement.

    • You know, if you looked at that last map w/o any context, you would think it was voting by county. The presidential voting by county map looks almost identical. These areas are typically poor and rural and they tend to get more in federal government money than they pay in taxes (e.g., see . Yet, they are majority conservative areas, where I suspect most people oppose the ACA and more redistributive tax policies (hence their votes for Romney).

      • Chris: Incredible detection of irony Chris! Or, is it hypocrisy? Did everyone see his maneuver? The Red states oppose “big government” but, on the other hand, they take more government funds than anyone. The hypocrites! Oh, wait . With a nuanced view, we see that the Red states typically have large numbers of Blacks and Hispanics, and it is these groups, not the anti-government population, taking the funds. Good ol’ fashioned nuance to the rescue!

        • If no looks at the red areas of the map, for the most part they don’t correspond to places where there are high minority populations. For example, southern Ohio and WV, Oklahoma, or eastern Tennessee. OTOH, areas of red states with high, say, black populations – South Carolina, the Mississippi – are not so red. So declining life expectancy appears to be concentrated among poor, rural whites, particularly in Appalachia.

    • What the heck is going on with southern Oklahoma? You cross the Red River and suddenly everybody starts dying? Are there other statistics that show that pattern? I wonder if it’s an artifact of different data sets. I’m finding it hard to imagine a policy difference that could make such a startling difference, and even harder to imagine cultural differences that change so much at the state line.

    • A couple of thoughts:

      1. Social Security: remove the top cap on wages. It’s currently like $100,000 – sounds like a lot – so everything over $100K is not taxed. Remove the cap and you won’t have to raise the nominal rate. Heck, you might even be able to drop it a bit. And you’ll keep that bitch solvent a lot longer.

      2. Note that the scaling on map 5 shifts. Towards the high end, green-blue, the max is 5 and the difference between colors is 0.5. At the low end, the reds, the absolute minimum is less than zero – which encompasses a lot of negative numbers – and the difference in scale values is 0.25, half of the ‘high’ end. Add, the difference between color values is not constant. What thes tells me is that Oklahoma is about central in approaching the average for the nation. It’s reminiscent of that old joke that more than half of all drivers think their driving is above average. Somewhere has to about average, Oklahoma is it.

      3. Also note that this map (5) is for women only. The next question to ask why has women’s health not changed in what is arguably the Bible Belt. Think about how this correlates, at least visually, with a map of the election results. I don’t know if there is a true causal relationship; I rather doubt it.

    • How do you meaningfully compute life expectancy by location, given that people move around so much? Is the increase in south Florida, for example, just lots of old people moving in?

      • Some people move around .’so much’. Most people do not; millions spend their whole lives near their childhood homes, surrounded by family which also did not move around ‘so much’. And since this data is aggregated, you would think that, if trends stay somewhat similar from year to year (or even decade to decade) then the comparisons would be valid. For instance, yes, older people are moving to Florida. But then they were doing that in the 1990”s too, wen’t they? And the 1980’s. And 1970’s. therefore you can look at average age at death and make some meaningful, if not perfect comparisons.

        • It doesn’t matter if most people don’t move. It won’t take very much to skew the results, because the demographics of the people who move are quite different from those of the people who don’t.

          That is, it matters how moving is accounted for. Let me put it this way: if I was born in state A and die in state B, how is it shown in the map above? Without knowing this, I don’t know how to read the map.

          If preferentially old people move out of an area, that could make the life expectancy drop in an area, even if nothing else is happening. (Because the people who die in the area are now younger.) Or it might *not* have an effect, if the calculation is with respect to the age profile of the population in the area.

          So, to repeat the question, more broadly, what does that map mean and how is it computed?

          • Three more maps:

            This page has a US map showing tobacco use by state:


            This page has a US map showing rates of obesity by state:


            This page has a US map showing population shifts (growth/decline) by county:


            They are not perfect matches, but there appear to be pretty strong correlations with the red areas on the map that show declines in life expectancy. The correlations might look even stronger if the smoking and obesity maps had breakdowns by county.

            The effect of population decline wouldn’t have to be just the most prosperous (and thus probably healthiest) old people moving out in larger numbers than the poorer old folks, and taking their longer life expectancy with them. Places with declining populations lose their brightest, most ambitious young people. Those who stay behind are likely more prone to health problems, and less able (financially and psychologically) to deal with them when they get sick. Thus one long-term effect of generation after generation of the healthiest young people leaving would be to ultimately lower average life expectancy of old people in the region.

            I couldn’t find a map for alcoholism, but I suspect that it would correlate pretty strongly too.

    • Thanks for the graphs. But I do have a question. Life expectancy numbers are an average, should we be looking at median instead?

    • Changes in life expectancy for workers just entering the work force (e.g., age 20) are also important. It is likely that (a) a higher percentage of workers live to 65 and start collecting the benefits they have paid in, and (b) those survivors now live a bit longer past age 65.

      Fewer workers paying into the system all their life and then keeling over at their retirement party is bad for Social Security finances.

    • If everything in America was held constant, except the percentage of the U.S. population that was Black or Hispanic, you’d expect a general worsening in all outcomes–crime, healthcare, infant mortality, education, use of welfare, etc. This is non-faith-based reality-oriented data-driven fact. Sorry if I’ve offended your religious beliefs. You can blame systems, policies, and so on, or you can just accept that “diversity” can sometimes adversely affect outcomes. Again, sorry to offend any religious nuts in the audience.

    • What happened to the life expectancy of black Americans in 1978? There’s a sudden drop of about 0.5 to 0.75 years there. Anyone know the cause?

    • Is there any reason you don’t show female life expectancy?