on my thank you money good morning good afternoon my name is luisa grion i work for the italian newspaper republic i am here to listen to the presentation of professor raj chetty professor raj chetty has been invited because really he is a new thinker one of the most outstanding representatives of the new economic thinking in the u.s he has a very interesting profile he got to the us from india when he was nine years old he had a very quick career a formidable career at the age of 23 he started studying in harvard he then he soon became professor at berkeley and then professor again in harvard the youngest tenured professor for this important university and he's now a professor of economics at stanford now the we have here professor at the festival of economics in trento and he as you know this year the the festival is devoted to growth and he has put growth at the center of his research the places of growth and also the politics that can be developed in a country both at the local regional national and central level to grow to have to make a population grow and its economics applied to the different stages or steps in life this is extremely significant especially in this time of recession which is afflicting italy and europe now his links between economic choices economic policy choices and the repercussions on the life of people attached upon several aspects for instance by studying how the level of education the beginning of education from right from the elementary school may have an impact on the role or quality of life that a young boy may have when he gets when he's 25 or 30. but the most important thing and professor will dwell on it has to do with the existing uh links between income and longevity now it's very easy to think that those who have more money can live longer they can resort to better healthcare etc but of course it's not as easy as that there is much more to it of course it's not necessarily that easy to explain has been able thanks to his studies that the play the place nowadays 5 of the richest americans live 15 years longer than the poorer americans now thinking that there is a link between income and longevity thinking that this is unavoidable would be a mystery and i think it is that all the rich are the same but all the poor are different from each other being poor may be a matter of life or death i mean there are areas where territories where the poorest and exactly the same life expectancy has the richest and in other places where the same poor people have the same life expectancy as people in developing countries for economic choices play a role and have an impact and it will be interesting to see whether this whether this approach has had an impact on the american policies and especially on the ongoing campaign for presidential elections these political decisions and these choices of course are closely linked to what is going on in europe and in italy for instance the migration waves will have an impact and change the problems in the political choices in italy and in europe and indeed looking at the most recent data of the observatory for health clearly indicated in italy for the first time we've seen a lowering of average life expectancy we live we don't live as long as in 1914 only slightly less and this slightly less is considered as a deficiency in prevention policies so this is a problem that affects us all just like inequality and in recent statistical data again you've seen that inequality is increasing less prevention unfortunately and people are discussing about new welfare policies locally and centrally obviously italian welfare the public welfare and health care is again is losing ground is decreasing and more and more the focus is on enterprise welfare local welfare somebody will have to cover these gaps we'll have to make up uh for what has been lost from the central level a colleague of mine published a survey quite recently and and found out that trade unions are play the role of associations offering services making up and covering the deficiencies and the gaps of what was previously offered by the public sector in italy so i think it will be interesting to listen to professor chetty and he will provide us with plenty of food for thought so you have the floor thank you very much it's really a pleasure to be here in trento at this very important and unique event so i'm going to talk today about the relationship between income and life expectancy in america but i think what i will uh present on as as luisa was mentioning a lot of implications for italy which i hope we'll be able to discuss further during the question and answer period so if i can bring the slides up here great so as you all know studies of inequality and growth in economics typically focus on measures of earnings or wealth so most of the talks you hear at this conference are going to be focused on thinking about income or how much wealth people have and how it's been changing over time but one of the most salient aspects of well-being for many people is health and longevity i think that's a dimension of inequality and growth that's extremely important to lots of people and as louis i was mentioning it's very well known that higher income is associated with longer life both across countries so when we compare rich countries like the united states or united kingdom or italy with poorer countries we see that those places have higher life expectancy on average but also within countries we know that richer people tend to live longer than lower income people so in this talk i'm going to essentially try to focus on two very broad questions uh first ask why is that the case why is there such a strong relationship between income and life expectancy and is that something that we can change through policy and if so what types of policies and that leads to the second question which is you know specifically how can policymakers improve longevity especially for low-income individuals who currently might have much shorter life expectancies than the rich i won't have definitive answers to either of these questions but i'll show you some new data that i think can be quite useful in thinking about these problems both in the u.s and perhaps here in italy so in a recent study the the talk that i'm going to be giving is based on a recent paper with several co-authors that was published in the journal of the american medical association a few weeks ago and in that paper we use newly available data from tax records to present a comprehensive portrait of income and life expectancy in america now these data are far larger than any prior data set that's been used to study the relationship between income and life expectancy we have 1.4 billion observations in the analysis that i'm going to show you from this data set that covers the us population over about the past 15 years from 1999 to 2014 and the key lesson that emerges from this analysis which i'll show you uh in the remainder of this talk in much more detail is that local geography seems to matter greatly in determining life expectancy especially for the poor so the relationship between income and life expectancy is not a universal relationship that is stable in all parts of the united states rather it varies greatly depending upon where you live and that simple result i think has powerful implications for how to think about this problem because it implies that discussions about how to improve the health of the poor which traditionally tend to take place at the national level how can we try to change the health outcomes of the poor how can we tackle inequality our data suggests that that conversation should be occurring at a much more local level where local mayors or community level efforts could potentially have an impact on the health and longevity of disadvantaged populations now some of you might know that in a related set of work with my colleague nathan hendren we've studied social mobility actually using the same types of data and the lessons we find there are very similar to the lessons that i'm going to talk about today in the health context in fact nathan was here at the festival last year presenting on this work on geography and social mobility and in in some ways you can think of this paper that i'm going to talk about as the health analog of that work so before i start showing you some results let me quickly tell you about the data and very briefly summarize the methods underlying the analysis those who are interested can read much more detail about this in the paper so we are going to measure mortality using death records from social security databases we measure income using tax returns at the household level so we define income at the household rather than the individual level for most of our analysis rather than looking at income in dollar terms we're always going to look at it in percentiles so in each year we're going to rank individuals in the national income distribution relative to all the other people of the same age and the same gender so for example take everybody who's currently 40 years old in the united states take say all the men and rank them from 1 to 4 million they're about 4 million people of any given age in the us we rank them based on their incomes and convert that to percentiles so i'm going to show you a lot of results that are based on income percentile and that's how they're defined now using these data on income and mortality rates we calculate mortality rates by age and income group and by gender and then we estimate life expectancy based on income percentile at age 40. so conditional on your income at age 40 we're going to construct an estimate of how long the average person will live given their uh given their income so there's quite a bit of methodological detail in terms of exactly how we do that we fit what are called comparts models statistical models to estimate mortality rates at older ages and so forth so i won't go into those details here but i'm happy to answer any questions people might have at the end now in everything i'm going to show you we're going to report estimates that are adjusted for differences in life expectancy across racial and ethnic groups it's well known from prior work that hispanic individuals and asian individuals have longer life expectancies than whites and whites tend to have longer life expectancies than blacks so especially when we compare across areas where there might be different fractions of blacks and whites and hispanics and asians we need to control for those differences in the demographic mix and so all the estimates that i'm going to show you you can interpret them as controlling for these racial differences they reflect the levels of life expectancy you'd see if every area had the same mixture of blacks whites hispanics and asians okay so with that background let me start to move into the results and i'll divide them into three sections first i'll show you the data at a national level in the united states then i'll turn to the local area estimates which are really the key focus of this analysis and then i will talk about predictors of local area variation that is what is it that seems to make the poor live much longer in some parts of the us than others and what types of lessons can we learn for policy that could potentially be applied here in italy for instance okay so let's start with the first part looking at the data at the national level and i'll start with this chart here which shows you the relationship between life expectancy and income for men in the united states so on the horizontal x-axis here is the income percentile of the individual from 1 to 100 and just to give you a sense of the magnitudes of these percentiles the 20th percentile corresponds to an income of roughly 25 000 the 40th percentile is about 47 000 and so forth and then the cutoff for being in the top one percent of the income distribution or sorry the average income in the top one percent of the income distribution in the u.s is about 2 million dollars per year okay so there's a broad spectrum of income and we're just converting that to percentiles and then on the vertical y-axis we're plotting life expectancy the expected age of death for 40 year-old people in each of these 100 percentile bins and you can see that there's a very clear relationship between income and life expectancy richer people live significantly longer lives than poorer americans and the gaps are quite large if you happen to be in the top one percent of the income distribution you can expect to live till you're about 87.3 years old for men and that compares with a life expectancy of 72.7 years for men in the bottom one percent of the income distribution now notice that that relationship is almost perfectly linear in terms of life expectancy versus income percentile and that's interesting because you might have expected that beyond a certain level of income there would be no further relationship between income and life expectancy but what you can see here is even when you're above an income of a hundred thousand dollars or two hundred thousand dollars there continues to be a strong relationship between income and life expectancy so coming back to one of the remarks lisa made initially you know is this just about the fact that richer people can afford better health care or that they are able to buy things that help them live longer this fact that the relationship is so clear even at higher income levels doesn't seem that consistent with that view because it's hard to believe that once you have an income above 100 or 200 000 a year that uh you know additional increases in income are really allowing you to buy something critical that allows you to live longer this is more consistent with the view that there's some other difference that's related to these differences in income that's driving the differences in life expectancy now just to give you a sense of the magnitudes of these differences the 15-year gap it's useful to have a benchmark to compare it to so let's put this difference that we see within the u.s in the context of differences in life expectancy across countries so the way this chart is constructed is each of the gray lines represents a different country and shows you the mean life expectancy in that country so there are 194 gray lines one for each country in the world uh in this graph and at the very bottom of the distribution you see lesotho has the lowest life expectancy and at the very top uh we have san marino or places like japan tend to have the highest life expectancies on average you can see italy is towards the top but not as high as the united kingdom or canada but well above places like china and libya now what we've done is put the data for the united states by income percentile on this chart so you can see that the men in the top 1 of the income distribution in the u.s they have higher life expectancy than men in any country in the world okay that 87.3 years it's higher than the mean life expectancy in any country in the world in contrast if we look at men in the bottom one percent of the income distribution the p1 we can see that they have life expectancy between men in sudan and pakistan so even within america there's this incredibly broad difference in rates of life expectancy where some people are living lives comparable to people in pakistan and sudan and other people are living much longer than anywhere else in the world and so that shows you the magnitude of the variation uh is enormous within america now i started out by talking about the relationship for men so now let's look at women so the red chart here shows you the same patterns for women you also see a strong increasing relationship between income and life expectancy but the gradient is flatter for women the gap between the top and the bottom is 10 years rather than 15 years and an implication of that which is interesting and i think hadn't been known in prior work uh is that the difference in life expectancy between men and women varies substantially across income groups so we all know it's a well-known fact that women tend to live longer than men on average but as you can see in these data that's much more true among low-income people than among high-income people among the rich there's only a 1.6 year advantage for women relative to men in terms of life expectancy at the bottom of the income distribution that gap is six years that's consistent with growing evidence in economics and other fields that women appear to be more resilient in the face of poverty and hardship than men so when you get to the lower part of the income distribution there seems to be something about uh the nature of women or genetic differences or something else uh that leads women to continue to do uh you know they have lower life expectancy but the drop-off is not quite as bad as it is for men at the bottom all right so those are the patterns broadly at the national level just taking a snapshot of the data so the next thing i want to turn to is to look at how these patterns are changing over time we know that inequality is growing over time lots of people are concerned about that inequality is measured by wealth or by income how does inequality look in terms of health outcomes so to look at that let's turn to this chart which plots estimates of life expectancy by year for four different income quartiles the bottom quartile shown in the blue going up to the top quartile shown in the orange and we're showing you the average life expectancy against starting with men by year from 2001 to 2014 for each of these income quartiles so you can see that for people in the bottom quartile here the lowest series in the blue there's an increase in life expectancy of only about 0.08 years annually which is small when compared with the increase in life expectancy for people we see in the top quartile where life expectancy has gone up by about 0.2 years every year so that is between 2000 and today the life expectancy of the richest americans has increased by about three years 0.2 times 15. in contrast the life expectancy of the poorer americans has increased by only 0.08 times 15 something like 1.2 years so what that means is that the differences in life expectancy that we're seeing are growing quite steadily over time that there's more and more inequality in life expectancy as time goes on now if we look at women we see a totally analogous pattern where again you can see the increases are much larger at the top in the orange series than they are at the bottom in the blue series so once again growing inequality over time now a different way to look at this gives you a more detailed picture is to plot the slopes of the lines that i was showing you in the previous graph you know like the 0.08 type of number we're now doing that estimating those slopes the trends separately for each income ventile so for 20 separate income bins by five percentile groups were plotting the annual change in life expectancy and what you see here again starting with men is that if you look at the poorest americans in the bottom five percent of the distribution there's actually been no gain in life expectancy at all over the past 15 years or so in america for the poorest men in contrast for the richest men as we were seeing before life expectancy went up by about 0.2 years annually every year translating to a three-year gain over the past 15 years similarly the same you see the same kind of pattern for women no gains at the bottom quite substantial gains at the top reinforcing the view that not only are there substantial gaps to begin with in terms of life expectancy but these gaps are growing over time you know which is quite concerning i think okay so all that shows you the picture at a broad national level but what i want to turn to next is that these pictures while they give you some information they're in a sense quite incomplete or actually somewhat misleading because it turns out that the patterns i was showing you of big gaps and growing uh inequality and life expectancy is not at all a uniform phenomenon within the united states in fact there are many parts of the country where you see very different patterns and more encouraging patterns in the sense that the poor appear to be doing quite well to illustrate let me start with this chart which once again now goes back to the relationship between the level of life expectancy and income so same kind of chart exactly like what i started out with but now we're looking at the data in four different major cities in the us new york city san francisco dallas and detroit and again we're starting with men and we're plotting life expectancy versus household income in five percentile point bins okay now you see a very striking pattern here which is if you're rich so if you're on the right side of this graph your life expectancy doesn't vary very much regardless of where you live you live pretty much to roughly 87 years or so regardless of whether you're in new york or whether you're in detroit if you come to the bottom of the income distribution however uh you can see that life expectancy for the poorest men in new york is still around 80 years or so whereas for the poorest men in detroit it's around 74 years about a six year difference now again to give you a benchmark for you know how big is a six-year difference in life expectancy uh one useful comparison is that uh the u.s uh center for disease control estimates that if we were to eliminate cancer as a cause of death if we were to win the war on cancer in a sense we would increase mean life expectancies in america by about 3.2 years so one way you can think about the six year gap is that it's twice as large as the change in life expectancy we'd see if we completely eliminated cancer as a cause of death right so it's almost as if you know the people in new york in a sense are immune from getting cancer while the people in detroit are getting cancer plus you know something else equivalent to that so these are really quite large differences in life expectancy uh at the bottom if you look at the pattern in new york and san francisco for the poor relative to the rich it's striking how much flatter it is there's just much less of a gap between the poor and the rich in new york than there is in a place like detroit where the gradient is just much much steeper now this chart shows you the analogous patterns for women and you see exactly the same kind of thing at the top of the distribution it doesn't matter very much where you live at the bottom there's a big separation and for women the pattern in new york and san francisco i find especially remarkable notice that for much of the income distribution that relationship is almost entirely flat higher income is almost not associated at all with higher life expectancy in those places which i think is uh you know remarkable given that there's a 15-year slope difference on average between the poorest and richest americans so trying to understand what is going on in new york and what is going on in san francisco that's enabling the poor to do so well is i think a great way to try to figure out how to improve longevity more broadly and that's what i'll try to turn to in the third part of this talk so these data show you the statistics for a handful of cities in the u.s what i want to do next is show you the data for the country more broadly so i'm going to focus on the bottom of the income distribution uh and this map shows you at the level of commuting zones which are analogous to metro areas that divide the country into 740 different uh metro and rural areas it shows you the average life expectancy for men in the bottom income quartile uh by commuting zone and so the the way the map is constructed is that the lighter colored areas are areas that have higher levels of life expectancy for poor men so you can see that in the lightest colored areas in the country like on the west coast san francisco for example or new york city in the northeast uh life expectancy for the poorest men is above 78.2 years in contrast in the darkest red colors life expectancy for the poor men like uh in the middle of the country places like ohio the the center uh they have life expectancy below 74.9 years so you can see there's quite a bit of broad geographic variation across the united states now one interesting feature of this map which those of you who know about the us might find somewhat surprising is that the south actually doesn't look so bad in terms of life expectancy for poor men often in many studies of health and economic outcomes social outcomes we find that the south looks much worse than the rest of the country but that doesn't appear to be true in this map so what is going on here the reason for that is that we are looking at life expectancy conditional on having a low income in this map if in contrast we were to just look at average life expectancies across all income groups which is what this next map shows you you see that here the south looks much much worse than the rest of the country so what is going on the south has many more poor people than the rest of the united states average incomes are much lower in the south than the rest of the country when you so as a consequence of that average life expectancy is lower in the south than the rest of the country because we saw initially that lower incomes are associated with shorter lives however once we control for income and look among the people who are poor in the south their life expectancy isn't particularly low so the key lesson that's coming out of that is that the low levels of life expectancy we see in the south in america are not driven by poor health conditional on income they can be attributed mainly to just very low levels of income but conditional on having a low level of income health doesn't appear to be particularly bad in the south okay and so that that illustrates why it's extremely important when looking at this regional variation to look at the variation by income level and not just pool all the income groups together because then you conflate the effects of income with differences in health conditional on income now again let me show you the same patterns for women so this chart this map shows you life expectancy for women in the bottom quartile of the income distribution and once again you can see here for women the south actually looks pretty good the colors are relatively bright in the south relative to the rest of the country but again if we were to pool all income groups and just look at mean life expectancy here we see the south looks much worse than the rest of the country so again the same exact pattern as we saw for men now in the big map it's very difficult to see the biggest cities in america so let me now show you a table that lists the 10 cities with the highest levels of life expectancy for the poor and the 10 cities on the right side with the lowest levels of life expectancy number one is new york city which has the highest life expectancy for poor americans and then you see a lot of places in california as well as miami florida and so forth on the right side here the places with the worst life expectancy for the poor are these cities in the midwest in america like detroit and cincinnati and gary indiana what used to be traditional like auto manufacturing industrial hubs of the united states these places had the lowest levels of life expectancy okay now i've been focusing on broad metro area level variation across you know new york city versus detroit and so forth but i want to show you next is that there's a lot of much more local variation than even that so if we take new york city and break it up into separate counties so we compare manhattan versus queens versus brooklyn versus bronx et cetera you continue to see quite substantial variation in levels of life expectancy where in the lightest colors in this map like manhattan for instance you have life expectancy above 80 years for poor americans but below 76 years in the outlying suburbs so the general pattern we tend to see when we look at the data at this finer level is that people who are living closer to the city center tend to have higher levels of life expectancy than the people who are living more in the outlying areas around the city now why exactly that is again i think remains to be understood but that is a very consistent pattern we find when we look at these data okay let me now next turn to trends in life expectancy across areas so i've been giving you again the snapshot average levels of life expectancy across places let's now look at how things have been changing over time in different parts of america and again let me start with a couple of examples so let's start with the city called birmingham in alabama in the southern part of the us uh birmingham is intro very interesting because if you plot annual uh life expectancy for the poorest americans this is showing you for men and women in the bottom income quartile you can see that there are quite substantial increases in life expectancy for the poorest americans in birmingham these increases of about 0.2 to 0.3 years annually you'll remember actually almost as large as the changes we saw for the richest americans so that is in birmingham the poor are gaining just as much in life expectancy as the rich in contrast if we look at tampa florida we see exactly the opposite pattern where life expectancy is actually steadily falling for the poorest americans uh by a quite substantial amount a decline of almost 0.2 years annually so what this chart shows you is that that pattern we were seeing in the nation as a whole where life expectancy was not changing significantly for the poorest americans really masks a great deal of variation across areas where there's some parts of america where the poor are doing perfectly well as well as the rich like birmingham but there are other parts of the america where things inequality is not only growing the poor are actually going backwards in terms of life expectancy and are living shorter lives over time so this is why it's so important to disaggregate the data to the local area level because the same story though you know that there's no one story that applies throughout the united states now this map shows you trends in life expectancy across states in the united states so for each of the states we calculate the annual change in life expectancy again focusing here on men in the bottom income quartile the green colored places are the places that have had more positive gains and the red colored places are the places that have had declines in life expectancy over the 2000s and you can see once again the point that i've been emphasizing there are many states that have dark green colors that is increases in life expectancy that are around three years over the past 15 years places on the east coast the west coast california and so forth but then there are a number of other states in the middle of the country and most notably florida new mexico and so forth where life expectancy is falling over time if you look at women you see the same exact pattern quite substantial declines in life expectancy in florida and in oklahoma where that hand is and so forth all right uh and once again in the big map it's hard to see the cities so just to give you a sense what are the places where we see the biggest gains and the biggest declines that that first place toms river new jersey is the new jersey shore near new york city you see really big gains the second place is birmingham alabama which i talked about and other places like cincinnati uh gary indiana and so on at the other end of the spectrum you see a lot of these cities in florida in tennessee life expectancy is falling by almost two years over the past 15 years for the poorest americans so quite substantial differences uh across these different big cities okay now in the in the remaining time uh i'm going to talk about what drives this variation in life expectancy across areas and what we can ultimately learn from it uh from a policy perspective so we're going to characterize the features of areas with high versus low levels of life expectancy conditional on income and i'm going to start by assessing measures of health behavior okay so we get data from surveys on things like rates of smoking rates of exercise rates of obesity and we ask are the places where we see the poor living longer is that associated with these differences in health behavior and the answer is that there's a very strong association between differences in life expectancy and differences in health behavior so i'm going to show you a bunch of correlations on this chart between average life expectancies averaging over men and women we find similar patterns for both genders we're going to correlate average life expectancy for four americans with various factors starting here with health behaviors and each dot shows you the magnitude of the correlation that we estimate and then the line shows you a confidence interval uh representing the statistical uncertainty on that estimate so starting with the fraction of smokers we see that there's a correlation of minus 0.7 a very high correlation places with more smokers uh higher rates of smoking tend to have much shorter life expectancy places with higher rates of obesity tend to have much shorter life expectancy and places with higher rates of exercise tend to have much longer life expectancy these correlations are so strong that you can actually just see them visually so this map here shows you smoking rates by area for people in the bottom income quartile among poor americans and notice that this map looks almost identical to the map of life expectancy that i showed you right you see the darkest red colors in the middle of the country the south actually looks pretty good if you look at the west coast california versus the adjacent state nevada you see much lower rates of smoking in california relative to nevada exactly consistent with the patterns i was showing you for life expectancy for a life expectancy for poor americans in california is much higher than nevada and so forth so you can see why there's a very strong relationship between uh rates of smoking very strong correlation between rates of smoking and life expectancy right and more broadly the variation in health behaviors is really strongly associated with the differences in longevity that we see in the data across areas now that naturally leads to the question of you know what is driving not just the variation in longevity across areas but also the variation in health behaviors apparently we need to understand why people are much more likely to smoke in some places relative to others why they have unhealthy diets why they're much less likely to exercise and so forth and so there are lots of theories that economists sociologists health researchers doctors have talked about over the years and there are many different factors you might want to consider i'm going to focus on four theory that have been widely discussed in the literature the first is differences in health care so a very natural hypothesis one that luisa mentioned in her introduction is that maybe the rich just have access to better health care than the poor and in certain parts of the country maybe there's more health insurance or better access to doctors or higher quality doctors than in other parts of the country a second hypothesis is that this is about differences in environment across areas so maybe there's more pollution in some places than the others and maybe that pollution disproportionately affects the poor who might live near factories that emit a lot of pollution for example a third very prominent hypothesis in this literature this is actually the one that has received the most attention by far is the idea that income inequality might explain differences in life expectancy so this goes back to a famous set of studies by marmot in england that showed that people who are lower down in the hierarchy of the income distribution tended to have shorter levels of life expectancy and that led to a literature which showed that across countries and across areas places that had more income inequality tended to have lower life expectancy and has led to the view that reducing inequality might be critical to improve longevity that the underlying theory here being that one of relative comparisons if you live in a society where lots of people are earning a lot more money than you do maybe you feel more stressed or you're under more pressure and that affects your health and makes you live a shorter life so that's the third theory we're going to evaluate and then a fourth theory is that this is about economic decline so in places with increasing unemployment maybe lots of layoffs as the auto industry has declined or increasing loss of jobs due to trade uh maybe we are seeing shorter life expectancy in those types of places so we are going to evaluate the explanatory power of these four theories by correlating our measures of life expectancy for poor americans with measures of each of these factors i want to emphasize before showing you the results that of course all of this analysis is correlational right we don't know what the causal effect of any of these factors are because we're not directly manipulating them experimentally or anything like that but you should think of this as just a first pass way to assess you know does it seem like there's a very strong relationship with any of these factors where should we be looking if we want to understand what's driving the differences in life expectancy across places so let's start with healthcare and i'm going to show you the data in exactly the same ways before plotting correlations between the measures of life expectancy and various proxies for the quality of healthcare in an area so the first is the fraction of people who have health insurance so america does not have universal health insurance until recently uh and so there's a lot of variation across places in the fraction of people who are uninsured so if you thought that access to health care was a critical factor in driving the variation you might have expected that higher rates of uninsurance would be negatively correlated with average life expectancy for the poor but in fact you can see there's almost essentially no correlation at all between rates of insurance coverage and life expectancy across areas if we look at other measures of access to medical care expenditures through the medicare program per person or an index for the quality of acute care based on 30-day mortality rates within hospitals or an index for the quality of preventive care which is based on the fraction of women who get mammograms or the fraction of diabetics who have their eyes checked various measures of the quality of care we see that all of these things have essentially no association with the differences in life expectancy across areas at least the associations are just dramatically weaker than they are for the first category of health behaviors so just from these first two pieces of data i think you can say quite clearly that the differences in life expectancy for the poor across areas in the u.s are much more readily explained by differences in behavior rather than differences in access to quality medical care so it's that's not to say that medical care plays no role at all but at least in explaining the spatial variation that we're seeing the first order thing seems to be behavior rather than medical care let's turn now to environmental factors so environmental factors are very difficult to measure because there might be many different types of pollution that could be affecting people different types of toxins and so forth and so rather than trying to directly measure the quality of the environment we went about this in a different way we we had the intuition that a city that is more residentially segregated would be one where the environment for the poor the physical environment for the poor might be more different from the physical environment for the rich if the poor and the rich all live in the same place there's no way that differences in environment can explain differences in life expectancy because they're all living in the same area and are exposed to the same environment in a more segregated city you might expect environmental factors to play a bigger role and typically you would expect the poor to be living in a worse environment and more segregated cities so this is a rough way to gauge whether environmental factors seem to be relevant asking whether the degree of residential segregation is correlated with the differences in life expectancy and once again we find a relatively weak association here if anything more segregated cities have higher life expectancy for the poor and smaller gaps in life expectancy between the poor and the rich so there's no direct evidence at least based on the segregation approach that differences in the physical environment are explaining why the poor are living much longer in some places than others the next factor i'm going to turn to is this inequality theory so start with the first point here the genie index the genie index is a measure of the degree of income inequality in area and you can see that we don't find a significant association between the level of inequality in an area and the average level of life expectancy for the poor so this for people in this literature is quite surprising given that there's been something like 100 papers that have documented that there is a strong relationship between levels of inequality and levels of life expectancy so why are we finding a very different result here once again it has to do for with the fact that we are controlling for the level of income we're focusing on the poorest americans when measuring life expectancy and so to illustrate why this really matters let me turn to this chart here which plots at the national level for men life expectancy versus income but now instead of putting the income percentile on the x-axis we're putting the income level the average level of income in dollars now here remember with with percentile ranks we saw an almost linear relationship right but now when you look at income in dollars on the x-axis you get a very concave relationship and why is that it's because the income distribution is very skewed so going from the median to the 60th percentile is a much smaller increase in income than going from the 90th to the 99th percentile right so now because this relationship is concave that is there are diminishing returns to having higher levels of income in terms of life expectancy if you just correlate average life expectancy with inequality which is what most prior studies have done you're going to mechanically find that places that have more inequality have lower levels of average life expectancy so that is a place with more inequality has more people on the left side of this chart and more people on the right side of this chart right and so because you lose more in terms of life expectancy when you go towards the bottom then you gain in life expectancy when you have more people at the top the average level of life expectancy is going to be lower that's just a mechanical consequence of the fact that they're diminishing returns to income what we are doing is different we're saying let's take a person who's earning fifty thousand dollars a year who's relatively low in the income distribution how does that person's life expectancy correlate with the amount of inequality in an area and we find that there's not much of an association once you control for that own person's income prior studies haven't been able to do that very effectively because they don't have good data on individual level income and so we end up concluding coming back to this chart that in fact there doesn't seem to be a strong association between inequality and life expectancy it doesn't appear to be the case that poor americans who live in cities with higher levels of inequality actually live shorter lives and that again you know insofar as you interpret that as a causal effect that is important from a policy perspective because it suggests that just trying to directly combat levels of inequality for instance by having more redistribution might not have a significant effect on the life expectancy of the poor now you can use various other measures of how socially cohesive a society is or how you know how much social inequality there is in a sense and all of these other proxies like indices of social capital or the fraction of religious individuals none of these things really correlate strongly with the differences in life expectancy the final theory that we evaluate are labor market conditions that economic decline hypothesis is it the case that life expectancy is lower in places that have a lot of unemployment places that have had steep declines in population places that have had steep changes in the fraction of working adults and again echoing the theme of this entire chart you don't find a strong relationship there either so there just really isn't much of a correlation with unemployment rates in an area in average life expectancies for the poor so all of the four theories we evaluated you know this is somewhat of a negative picture we don't really find much support for any of them in a correlational sense so what we did next is came at the data from a different perspective we said okay the the four hypotheses we had going in don't really seem to explain what's going on so what does explain the variation in life expectancy across areas what are the strongest correlations that we can actually find and this chart uh answers that question it summarizes the five strongest correlates we find between life expectancy for poor americans and area level factors so let me just read them off they're the fraction of immigrants in an area average house values so a measure of the cost of living of the place local government expenditures population density and the fraction of college graduates a measure of how educated the population is so let's come back to the examples of new york and san francisco as two cities that have very high levels of life expectancy for the poor you can see that they exactly fit this picture right there are places that have very large numbers of immigrants they're very expensive cities with very high uh housing costs they tend to be liberal cities that have a lot of local government expenditures they're very dense they're big cities big urban areas and they're very educated cities very high numbers of college graduates so this chart basically shows you that the new york and san francisco pattern is one that holds more generally it's these affluent educated urban areas that tend to have high life expectancy among the poor this is actually you might think somewhat surprising because you might have expected that living in a very expensive city like new york or san francisco might actually be harmful if you're poor because you in real terms are actually somewhat poorer but here you see actually the pattern seems to be exactly the opposite at a given income level of say 25 or 30 000 a year your life expectancy is higher if you're living in a place like san francisco or new york than if you're living in a much less expensive city like cleveland or detroit or something like that so that's the you know that's what really seems to explain the variation across areas in the data that is low-income people as i've said before in affluent educated cities live longer and have health healthier behaviors than than low-income people living in other cities so the the question is why that's the case why is it the case that we're finding that living in an urban educated area seems to be associated with longer life for the poor there we don't know the answer to that question we think that's the key question to investigate going forward uh let me just speculate on some potential mechanisms one possibility is that this is about spillovers from the rich to the poor so maybe when you live in an area with a lot of more affluent people that leads to regulations like for instance bans on smoking in offices or public places or bans on unhealthy fats in foods and that ends up affecting both the health of the rich and the poor or maybe when you live in an area with a lot of other rich people especially a liberal community that has high government expenditures there's a lot of spending on public health programs so san francisco for instance has a very well known program to try to reduce the incidence of hiv with public health efforts uh targeted at lower income uh individuals one possibility is that you're just directly benefiting from government programs in in these types of cities another possibility is that it's about peer effects maybe being around a population that's very healthy so if you think about in the bay area people are very interested in exercise and healthy eating that naturally makes it easier for others to buy healthy food or to exercise than if they lived in a place like oklahoma where the norms might be very different a third possibility is that this is not actually about the effect of the place but about differences in the type of people living in these places the poor americans living in san francisco might just be different from the poor americans living in other parts of the country and maybe that's what explains their differences in health behaviors and health outcomes so again we don't know which of these is going on we think it would be really interesting to study and discriminate between these explanations going forward so let me conclude by summarizing what we've learned here we found is that inequality and life expectancy is large and worryingly is growing over time in the u.s and i suspect similar patterns are true in italy and in other countries european countries as well however importantly and i think for on from a more positive perspective the data clearly show that these changes and these gaps are not immutable that is we can change them there are some parts of the us like san francisco and new york that have relatively small gaps in life expectancy between the poor and the rich and there are some parts of the us where those gaps are actually shrinking over time like birmingham alabama now what that that less that fact means it suggests to me that reducing health disparities is likely to require local policy interventions so rather than talking about how we improve health of the poor in the us as a whole i think we really need to be talking about how to improve the health among low-income individuals in specific cities where things are not going so well like las vegas or detroit where you see really bad outcomes for the poor now given the very strong correlations we see between these differences in life expectancy and health behaviors my guess is that any policy solution that's going to have an effect is going to have to go through changing people's health behaviors and can't just directly try to change their health outcomes so that is providing more medical care might be useful but it's likely to be medical care that changes people's health behaviors not just provides treatment conditional on having people in poor health so those are i think some of the implications the local level there are also some important implications at the national level the fact that we see such different trends for poor and rich americans and across areas imply that federal policy changes for instance indexing eligibility for social security and medicare to mean life expectancy these types of changes might have significant effects on the progressivity of social security and medicare systems so in the us a common proposal to deal with budget shortfalls is to increase the retirement age of eligibility for social security as life expectancy is going up but if you do that in proportion to average life expectancy in the population you're going to hurt the poor relative to the rich because as we've seen the gains and life expectancy are disproportionately going to the rich rather than the poor so what what this variation shows you is that we need to think hard when making any such federal policy change about how that's going to have different effects across the income distribution and potentially across areas finally our hope is that the statistics that we've constructed here which we've made publicly available on a website called healthinequality.org where anyone can download these statistics that i've been showing you at a county level by income group by gender by age and so forth we're hoping that these will be used as a tool to monitor local progress and ultimately identify uh solutions that can be used to improve health behaviors and life expectancy of the poor now here in italy i want to emphasize that while i've talked about the u.s and i think some of the lessons from the u.s data might apply here it's important to note that analogous tax and social security data actually already exist in other countries such as italy and i was just talking this morning to tito boeri about uh the existence of such data here in italy and i think would be incredibly valuable to construct analogous local area statistics like i've been showing you for the us here in italy so that you can monitor which places seem to be doing well which places need further improvement and what types of policies can make a difference thanks very much thank you professor chatty a number of the things you mentioned are really very challenging and let's start with the questions good evening research thank you so much for sharing all the information with us much appreciated my name is massimiliano i'm sales manager in a textile company in milan and i have a two quick questions for you please what are the advices would you give to the italian politics to reduce uh basically the same thing that is happening in the us because the middle class is totally disappearing no more middle class and then some another question more personal why all you indian people are so smart and thinking about the ceo of google or like all the indian people i mean i know like bangalore in india they make so many good engineer for all the world thank you so all much so let me start with the first question which i think i may be able to answer better um so shrinking middle class same kind of problem in the us as well uh i think some of the consequences of that are obviously increasing inequality but also potentially um affecting the health of the the you know the poor part of the italian population i suspect you'd see reductions in life expectancy like we do in the us perhaps in some parts of italy i think there are two types of solutions to to think about one is how do you actually combat the erosion of the middle class people often talk about increased redistribution so can we change taxes and transfers so that we provide more support to the lower part of the lower income part of the population in the us at least you find that that tends to have very limited support even though inequality is going up there isn't that much support for increasing taxation of the rich so instead in our other work which i briefly mentioned the work on social mobility uh i think a different approach to think about is how you can provide better opportunities for the poor to move up in the income distribution so thinking about how to improve equality of opportunity in a sense and i think that often has to do with the quality of schools or the quality of neighborhoods we emphasize very much like i was talking about here that local area variation seems to be extremely important in thinking about these issues of inequality and so thinking about how to integrate cities so that the poor are living near the rich i gave you the examples of places like san francisco and new york where living around higher income people seems to have positive effects on health we find similar things in terms of opportunity and moving up in the income distribution so i think thinking about the problem not so much as how do we literally bring more people into the middle class but how do we give more people opportunities to move up can be a very constructive way to solve the problem on the second question i think you're seeing a selected set of the indian population basically and so like in any country there's a lot of variation it's a lot of emphasis on math and engineering uh in india uh it's actually unusual that i'm in economics a lot of people think you know why am i not in medicine or in or in engineering yes yes so engineering and medicine are like the top things and economics is like way down on the list in the indian system another question here and then up there so first off again just to congratulations your work is fantastic and has brought us really new tools and economics that we're not uh seeing but i did want to go back and ask about the relation of your income studies which i know you must have thought about you gave us one example there because in the income studies so how things like inequality uh and social capital are correlated with intergenerational mobility and of course then the outcome variable there is higher incomes over time but here you're saying that even though life expectancy is correlated with higher incomes it's not cruel it's not the same drivers in a way except for the residential segregation i know these aren't these are correlational yeah not causal but what do you what's your thinking on that i know you must have thought about that song great question so let me just recap that for those who might not be familiar so in this other set of work we've done looking at differences in social mobility what are the chances children climb up in the income distribution and realize the american dream we find very strong correlations with levels of inequality levels of social capital segregation a lot of the factors as you correctly point out we don't seem to find strong associations with here so what is going on the answer actually has a lot to do with exactly what i was describing in regards to why the south looks like it right why the south looks different in terms of life expectancy if you pool all income groups versus just look at the poor so if we were to just look at average life expectancy we would find exactly the same types of correlations places with more inequality places with more segregation they all tend to have lower levels of life expectancy but the mechanism appears to be through income consistent with that earlier set of studies right that those places tend to have more low-income people perhaps because there's less social mobility exactly along the lines of the earlier work however in the correlations i was showing you at the end here we are focusing on life expectancy conditional on having a low level of income and that's why the way to interpret that is health condition on being poor doesn't seem to be strongly associated with those factors but of course being poor might be strongly is appears to be strongly associated yes i would have a question i think it was the first chart that you showed the correlation between uh income distribution and life expectancy what was impressive probably you referred to that is that the top 0.2 percent still uh live a little bit shorter than the top 0.1 percent and i don't think there is any i don't know health care service that the top 0. cannot buy or in terms of quality of life so i was wondering how do you whether you have any hypothesis to explain that yeah yeah i mean it's amazing that that continues to be true that's true i showed you the chart by percent it's true even within the top one percent so if you look at the people making 5 million dollars instead of one million dollars per year the people making five million dollars are still living you know longer and so clearly that can't be what would literally what you're buying i think the most plausible explanation is that there's some third factor that drives both your income and your longevity right so my colleague vic fuchs at stanford who's a famous health economist who's been studying these issues for a long time his view is that that third factor is something like stamina so there's some people who have a lot of stamina a lot of energy in some sense that allows them to achieve work a lot and achieve a high level of income and also leads to a long life i think the data are more consistent with that type of explanation then the idea that money is literally allowing you to buy a longer life any other question i have one question when you were answering a question you mentioned the the idea that it is very important to have a sort of uh recovery of this inequality for young people in italian statistics so this is one of the central part of the problem inequality is increasing but inequality is especially increasing for for kids for uh i mean for youngsters from 14 to for a young italian it is more difficult to to go up the social ladder more difficult than for a french young man and our country is characterized by a public expenditure that has gives a great priority to social security and retirement benefits how could we redefine our public expenditure a country such as italy that has a an aging of the population that is accelerating and a very low birth rate how could we work things the other way out so as to reduce the inequality in among young people so it sounds like inequality in italy is increasing among the youth much of it is it much as it is in in other countries and a related phenomena or separate phenomenon that plays into this is that there's an increasing burden of the youth to support the elderly right uh in order to finance their retirement and so forth i think it comes back to one of the the earlier questions you know what can you do to try to combat inequality if they're really two ways to think about it one is redistribution of income and one is creating more opportunities to uh move up in the income distribution i think trying to invest in things like improving the quality of public schools especially in areas that currently don't have very good schools or other types of efforts so for instance in the u.s there's a large public housing voucher program that gives families low-income families assistance to live in better neighborhoods i don't know if there's something analogous in italy but making more of an effort to help lower income people live in better areas we think there's good experimental evidence showing this that can really have profound effects on the outcomes of low-income youth and really increase their incomes and that would not only provide perhaps a better way to support the older generation as they themselves have higher incomes it would of you know deal with the problem of increasing inequality and potentially also have a significant effect on these health outcomes that we're talking about here in the presidential campaigns how are your arguments considered i've read that there was an interest for your theories both uh by hillary clinton but also by john bush at the very beginning i don't think donald trump is very keen on this arguments but considering the harshness of the american campaign where there is this very strong opposition is there still room for the argument such as yours yes so i've definitely spent a lot of time with hillary clinton and the democratic current democratic administration who are very interested in trying to figure out how to use this type of evidence to improve policy making i also have spent a good amount of time with people on the republican side paul ryan in particular who's the speaker of the house and is an influential republican is very interested in trying to use evidence to improve policymaking but the campaign has gone in a different direction and as you point out i have not spent time with donald trump i don't know who who has in fact among the economics community and so you know we'll see what what exactly happens uh i think a number of us are concerned in terms of the future of doing this type of analysis evidence-based uh you know trying to use evidence to inform economic policy how is that going to look in these two different administrations our senses if it's the clinton administration it will be very similar to the how it was under the obama administration if it's trump i think nobody has a really good sense of what what will happen well is there a risk for trump to become the next president of the us what are your your expectations so i don't have any special insight into that beyond looking at the polls that you all can look at yourselves i mean i think the statistics at the moment suggest that it would take an incredible change in terms of current polling numbers for trump to be elected but on the other hand you never know what could be dug up about hillary clinton and what can change in the next few months so we'll see but just one uh uh small point i think on the relationships that we are showing yeah there is a strong non-linearity at the bottom right yes so uh five years of life expectancy is essentially explained by the bottom one percent two or three years so when you look at the correlations among the poor uh you look at the bottom quartile right yes so if you drop the uh the volume one percent you get i suppose you did it clearly but i was uh interested more in the quantitative uh feature because i mean for a policy from a policy perspective i think it's important whether you have to target the bottom one percent or the bottom quartile you know in terms of affordability code yeah so you're absolutely right the gap between the bottom one percent in the top one percent is 15 years the gap and so you see this really sharp drop at the bottom but if you look at the gap between say the bottom the fourth percentile and the top percent it's still about 12 years or so so roughly two or three years are in that very bottom uh what we think might be going on there especially for men is things like being in prison so everybody who's in prison which is an incredibly large population in the us is in our data and people in prison or in very harsh conditions might have much shorter life expectancy then than others in order to have so at the very bottom to have an income you know less than a thousand dollars a year typically you have to be either homeless or in jail or some other extreme circumstance and so it's true that those people have the shortest life expectancy but my senses as you argue there's a different set of issues at play there relative to just low-income americans who have a house or you know they're living in kind of normal circumstances but are just have relatively low incomes i have three small questions for you the first one is why you selected the age of 40 to 42 to start with because it's a good number for those leaving 80 years old but it's quite high for just leaving only 60 or 50. and second question is uh if there is some importance uh in using instead of the location where you leave the location where you are being raised like if the first 20 years of your life has been having influence on the length of your life and the third thing is just that you spoke about a lot about household income and things and at the neighborhood where you live and if there is a correlation with criminality ratio because the detroit area are like that is i don't know perceived by me as an italian like a high criminality area so does that have an effect and has been taken out yeah off from the charge yeah uh excellent question so the uh first on the first point um why do we start at age 40. so we wanted to measure the relationship between income in adulthood and life expectancy you can't measure income too much before age 30 because a lot of people are in college and so the amount you're earning in your 20s is not a good measure of your lifetime income between age 30 and 40 you know you could have started anywhere there our data constraints make it harder for us to go further back because you lose you only have a finite amount of data right and so the mortality rate between 30 and 40 is so low that we might as well start at age 40. so that's why we did that now the second question you ask is actually somewhat related so we are looking at where you live when you're 40 basically and associate you know using that for the geographic variation another hypothesis which i think is certainly worth investigating is that maybe it's about where you grow up that that really matters childhood conditions are other works addressed on intergenerational mobility suggest that it's really a childhood environment that matters and the same could be true here we just haven't investigated that directly and it's a little bit difficult for us to do that with the data we have again because of the same limitation of the number of years we see but in order to do that uh both of your questions are somewhat related in the sense that you can't of course measure income in adulthood if you want to start in childhood you want to look at something like how it relates to your parents income or your socioeconomic status when you were brought up and so forth that's a study one could do but it requires a really long data set right because you have to be able to see kids when they're say 10 years old and then mortality rates are so low until people become 60 or 70 or older ages that you need that super long time spend which is not currently available in the u.s it'd be more countries like sweden or denmark where you have that type of data the third question on crime uh there are associations with rates of crime but crime is particularly a factor directly at younger ages so when you think about detroit and rates of homicide and murder and things like that being causes of death that's not so much true once you look at people who are 50 60 years old that's completely concentrated between the ages of 20 and 35 or something like that so in a sense it again relates back to your earlier questions why that was not a major factor we looked at any other questions presentation my question concern because you can you compare different city in in the us i'm curious to know if you or also did some some analysis between city and country if there is uh which is the relation between living in a in country or or in a general city thank you and by country let me just clarify do you mean rural rural versus urban yeah so uh the pattern you tend to see here is that people living in cities have higher life expectancy than people living in more rural areas so that's why it's places like new york city and san francisco that are showing up as the highest life expectancy places more generally high density areas appear to be places with higher life expectancy so why is that again not clear exactly if it's something about the causal effect of living in the city or is it about the types of people living in cities versus rural areas insofar as it's a causal effect it could be driven by public health programs it could also be driven by things like public transportation so in the u.s at least when you live in a more rural or more suburban area you tend to spend a lot of time in a car and you tend to be more kind of sedentary in a place like new york city or in san francisco you're much more likely to use public transit and as a result people are much more active and that is we know from studies in epidemiology and health a very highly correlated factor with health and life expectancy so you know we tend to find urban areas generate better outcomes and i think this could be one reason why that's the case thank you very much professor chetty for your availability and your uh excellent presentation and thank you very much to the audience now i hope to meet you again