WIXLIB

BOX 2.Geographic Variation in Life ExpectancyRecent research has shown that life expectancy varies widely across geographic locations, with especially large differencesamong Americans at the bottom of the income distribution (Chetty et al. 2016). For example, the life expectancy of a manreporting income in the bottom quartile was five years shorter in a so-called commuting zone (a particular definition ofgeographic area) with the lowest longevity compared with that of a man reporting income in the bottom quartile in an areawith the highest longevity. Chetty et al. examined geographic factors that might be expected to affect the life expectancyof those with the lowest incomes: access to medical care, environmental factors, income inequality, and labor marketconditions. These factors were surprisingly uncorrelated with differences in life expectancy across areas, with the exceptionof smoking rates, which were correlated with geographic differences in life expectancy among those with incomes in thebottom quartile. In terms of geographic factors, Chetty et al. concluded, “Low-income individuals tend to live longest (andhave more healthful behaviors) in cities with highly educated populations, high incomes, and high levels of governmentexpenditures, such as New York, New York, and San Francisco, California” (Chetty et al., E15).Box figure 1 uses Chetty et al.’s (2016) data to show the expected age at death for men, by household income percentile, forfour geographic areas that contain major cities. Notably, those with low incomes in New York City fare much better thantheir counterparts in Detroit, while life expectancy for those at the top of the income distribution exhibits much less variationacross locations.This research suggests that there remains much we do not know about determinants of life expectancy for low-income menand women. Further research will help illuminate the local factors that affect life expectancy, as well as the place-basedstrategies that could help struggling populations.BOX FIGURE 1.Expected Age at Death among Men by Household Income Percentile in SelectedCommuting Zones8682New YorkSan Francisco78Washington, DC74Detroit700102030405060708090100Household income percentileSource: Chetty et al. 2016.Note: Sample pools men who turn 40 during the period 2001–14. Cities in the figure refer to commuting zones that includeadjacent counties as defined by commuting patterns, in addition to urban cores. (Commuting zones are similar to metropolitanstatistical areas, but may cover rural areas as well.)The Hamilton Project Brookings5

A DIVERGENCE IN LIFE EXPECTANCYMany drivers of improved mortality—for example, moreeffective medical treatments, safer cars, and smoking cessation—are first adopted by higher-income people, and later by thebroader population. In fact, variation in the use of technologyand behavioral changes helps explain some of the divergence inlife expectancy across income groups and age cohorts.The most quantitatively significant improvements in lifeexpectancy have generally been experienced by peoplethroughout the income distribution. Inexpensive medicalinnovations like vaccines and mass-produced antibiotics, as wellas public goods like clean water and sanitation, were eventuallyextended broadly. Other factors that affect life expectancy,such as smoking behavior, have yet to change for the entirepopulation. Figures 4A and 4B use data from the NationalSurvey on Drug Use and Health to show several importantpatterns in smoking from 2000 through 2013.3 First, amongmen, older cohorts had much higher rates of smoking than didyounger cohorts. Second, the decline in smoking among theolder cohorts occurred first among higher-income individualsbefore extending to those with lower incomes. In other words,high-income Americans reduced their smoking behavior morequickly than did low-income Americans. Perhaps surprisingly,lower-income women past age 50 actually increased their rate ofsmoking. Third, younger people at all income levels have beenreducing their smoking from 2000 to 2013. Because of the timelag between smoking and its effect on mortality, it takes decadesfor declines in smoking among young adults to become apparentin reduced mortality rates. The fact that higher-income groupsquit smoking earlier likely explains some of the divergence in lifeexpectancy by income. In the future, however, we should expectto see broader impacts of smoking cessation on mortality.As has been widely reported, Case and Deaton (2015) documentthat the decline in mortality rates among middle-aged whiteshas unexpectedly stalled. As shown in figure 5B, at the sametime that there was an increase in mortality rates amongmiddle-aged white non-Hispanics, there were sharp declinesin mortality rates among middle-aged blacks—from 800 tofewer than 600 deaths per 100,000. Middle-aged Hispanics alsoexperienced steady declines in mortality from 1999 to 2014. Theincrease in mortality among whites appears to have been drivenlargely by increases in mortality for those with little education,and appears to be associated with changing rates of drug andalcohol abuse as well as mental illness and suicide (Case andDeaton 2015). As shown in figure 5A, although lung cancerdeaths fell and diabetes deaths rose only slightly, deaths fromsuicide, alcohol, and drug-use rose sharply. In 2014 the rate ofsuicide among white Americans age 45–54 actually surpassedthe rate of mortality from lung cancer, reflecting both worseningsuicide rates and falling smoking rates.Suicide mortality is related to means—fatal suicide attempts arehighly correlated with having access to a firearm (Miller, Azrael,and Barber 2012). The majority of deaths from suicides—andhomicides—are from firearms (Grinshteyn and HemenwayBOX 3.What Is a Public Good?Most goods and services—and most activities that people engage in—are a matter of concern only for the people engaged in thetransaction. That is, virtually all the costs and benefits associated with the transaction are experienced by the buyer and sellerthemselves. For instance, when an electrician fixes some wiring in a house, the consumer and the professional experience thecosts and benefits: the payment, the utility from having a working light fixture, the opportunity cost of the time spent fixing thewiring, and so on.By contrast, there are goods and services that create significant costs or benefits for third parties: individuals who are not at allinvolved in the transaction. A classic example of a public good is vaccination. When one person is vaccinated, that person becomesmuch less likely to catch and pass on the disease to others. This constitutes a positive spillover—also known as an externality—ofthe transaction to third parties. Note that a public good can also have the reverse structure. If a transaction produces a negativespillover for a third party, then minimizing the number of such transactions is a public good. The classic example here is industrialpollution: producing energy from fossil fuels has costs and benefits to the consumer and energy producer, but it also has negativespillovers in the form of public and environmental health risks related to particulate matter and carbon emissions.With both types of public good, economists typically expect that the market will settle on the “wrong” quantity of transactions,viewed from a social perspective. Because they do not experience all the costs of air pollution, industries will pollute too much;similarly, too few people will pay the cost of vaccination, given that they do not receive all the social benefits conferred byvaccination. Public goods therefore offer an opportunity for government interventions to improve on a private market outcome,usually by imposing a tax or offering a subsidy that allows market participants to take account of the spillovers to third parties.6The Changing Landscape of American Life Expectancy

FIGURE 4A.FIGURE 4B.Men Who Have Ever Smoked by Age andIncomeWomen Who Have Ever Smoked by Age andIncome70%Share of populationLow income, age 50 60%Low income, age 26–49Low income, age 26–4950%Low income, age 50 High income, age 50 40%High income, age 26–49High income, age 50 High income, age 2006200820102012Source: National Survey on Drug Use and Health 1999–2014.Note: Smoking is defined as having ever smoked at least 100 cigarettes. Income bins are terciles of family income for the entire sample, and thesample is restricted to individuals age 26 and older.FIGURE 5A.FIGURE 5B.Mortality by Cause, White Non-HispanicsAge 45–54Mortality, Age 45–541000100035 35LungLungCancerCancerPoisoningPoisoning30 30Deaths per 100,000Deaths per 100,00025 2520 cide15 15DiabetesDiabetes10 10Deaths per 100,000Deaths per 100,000800 800BlackBlack600 600WhiteWhitenon-Hispanicnon-Hispanic400 400HispanicHispanic200 200505019991999200420042009200920142014Source: CDC 1999–2014.Note: Figure is adapted from Case and Deaton (2015) figure 2. Chronic liver diseasesinclude alcoholic liver diseases and cirrhosis. Poisonings include drug and alcoholpoisoning, both accidental and with undetermined intent.0019991999200420042009200920142014Source: Centers for Disease Control and Prevention1999–2014.Note: Mortality data are for all-cause mortality. Figure isadapted from Case and Deaton (2015) figure 1.The Hamilton Project Brookings7

2016). Research has found that firearm ownership and suiciderates are robustly correlated (Hemenway 2014) even afteraccounting for rates of antidepressant prescriptions (Opoliner etal. 2014). Firearms are lethal in other scenarios: more than 100children are accidentally killed with a firearm every year in theUnited States, most of them by other children (Hemenway andSolnick 2015). Means restriction policies—a variety of measuresthat include things like erecting suicide-prevention barrierson bridges and restricting access to firearms—are effective inreducing rates of preventable death (Mann et al. 2005; Yip et al.2012).Given that Americans with low incomes have seen stagnatingor declining life expectancy, it is worth investigating howthe link between income and substance use has changedover time. In figures 6A and 6B we again use data fromthe National Survey on Drug Use and Health, this time toexamine recent changes in the rate of drug and alcohol abuseby income group. Interestingly, heavy drinking has becomemore common at all income levels; by contrast, use of illegaldrugs, excluding marijuana, has become more commonfor those with low incomes, while high-income men haveexperienced smaller increases.The recent trends have caused many to speculate about apossible link between labor market changes and the factorsthat have depressed life expectancy for lower-income people.The stagnation in real wages at lower percentiles, andincreasing real wages at upper percentiles (Bivens et al. 2014)would seem to correspond to the observed divergence in lifeexpectancy. Concurrent with this increasing wage inequalityis an increasing pessimism about future economic conditions(Uslaner 2012). Perhaps the bleak economic outlook, real orperceived, for low-income and lesser-educated workers hasled to problems—related to mental health, alcohol abuse, anddrug abuse—that depress longevity.The trends in life expectancy do not entirely align with thishypothesis, however. Life expectancy for those with highincomes diverged from life expectancy for those with lowerincomes in both the male and female populations over the past35 years, yet the labor market fortunes of men and women haveevolved differently over that period. While men—aside frommen with at least a college degree—have suffered stagnant ordeclining incomes, women have been more likely to experiencelabor market gains. Household income was roughly constant forlow-income women from 1980 to 2010 while household incomedropped considerably for low-income men (Current PopulationSurvey March Supplement and THP calculations). If labormarket stagnation was an important driver of deteriorating lifeexpectancy among those with low incomes, we might expectto see men suffering more than women. Instead, as we see inFIGURE 6A.FIGURE 6B.Illegal Drug Use and Heavy Drinkingamong Men Age 26–49, by IncomeIllegal Drug Use and Heavy Drinkingamong Men Age 50 and Older, by Income24%24%12%12%12%12%LowLowincomeincome16%16%8% 8%IllegalIllegaldrugdruguse use(dashed(dashedline)line)12%12%6% 6%LowLowincomeincome8% 8%4% 4%HighHighincomeincome4% 4%2% 2%0% 0%0% 2011201120132013Heavy drinking (share of population)Heavy drinking (share of population)Heavy drinking (share of population)Heavy drinking (share of population)10%10%HighHighincomeincome10%10%8% 8%3.0%3.0%2.4%2.4%IllegalIllegaldrugdruguse use(dashed(dashedline)line)LowLowincomeincome6% 6%1.8%1.8%LowLowincomeincome4% 4%1.2%1.2%HighHighincomeincome2% 2%Note: Illegal drug use refers to the use of hallucinogens, cocaine, or inhalants on three or more days in the past year, the use of heroin in the past year, or theuse of prescription drugs—sedatives, tranquilizers, stimulants, or analgesics (i.e., pain killers)—on three or more days for nonmedical reasons in the pastyear. By definition, this excludes any use of marijuana. Heavy drinking is defined as drinking five or more drinks on the same occasion on three or more daysin the past 30 days. Income bins are inflation-adjusted terciles of family income for the entire sample. The sample is restricted to individuals age 26 and older.The Changing Landscape of American Life Expectancy0.6%0.6%0% 200920092011201120132013Source: National Survey of Drug Use and Health 1998–2014.8Illegal drug use (share of population)Illegal drug use (share of population)Illegal drug use (share of population)Illegal drug use (share of avydrinkingdrinking(solid(solidline)line)

figure 3, life expectancy for the bottom quintile of women fallsconsiderably more than the bottom quintile of men.Moreover, the labor market stagnation narrative is difficultto square with other aspects of mortality trends in recentdecades. Currie and Schwandt (2016) find that mortalityrates for black Americans have fallen dramatically, and thatmortality inequality among children has also fallen, all whileincome inequality has been increasing. Appendix figure 1reproduces some of Currie and Schwandt’s findings, showingthat increasing inequality in mortality and life expectancy isrestricted to the population age 50 and above. The authorsargue that investments in health care for pregnant womenand children, changing smoking behavior, and other factorsmay be important for understanding divergences in mortalityand life expectancy.In understanding the economic determinants of mortality, itis helpful to distinguish three types of mortality effects: (1)the effect of poverty amelioration, (2) the effect of incomestagnation, and (3) the effect of increasing income inequality.First, there is considerable evidence that poverty amelioration(at both the family and community levels) causally improveshealth and, by implication, life expectancy. For instance,nutrition assistance (Hoynes, Schanzenbach, and Almond2016), the Earned Income Tax Credit (Hoynes, Miller, andSimon 2015), and relocation to lower-poverty neighborhoods(Chetty, Hendren, and Katz 2016) all have been shown toimprove health. Other policies aimed at poverty reduction—the minimum wage, for example—may have similar impacts.The mortality effect of recent U.S. income stagnation formany workers is much less clear. While it is certain that highincome workers live longer than low-income workers, it isnot obvious that recent changes in income for those abovethe poverty level are having an effect on mortality. Finally,researchers have generally concluded that mortality is notimportantly affected by increasing income inequality (Deaton2003), though again, poverty itself is quite important.Protecting Life Expectancy Gains forEveryoneThe policies that can help ensure future life expectancy gainsfall into two categories: (1) those that expand participationin technological advances contributing to longer lifespans, and (2) those that address social factors relevant tolife expectancy. Often these factors go hand in hand. Forexample, one enduring problem in medicine—mentionedby Hippocrates around 400 BC (Hugtenburg et al. 2013)—isthat medication cannot help the patient who does not takeit (Ho et al. 2006; Bailey et al. 2010). Differential rates ofmedical adherence (e.g., by race) may be related to disparitiesin mortality (Wu et al. 2010). Evolving technologies, such asautomatic text messages to patients to remind them to taketheir medications (Granger and Bosworth 2011), provide anavenue for further progress.IMPROVING PARTICIPATION IN TECHNOLOGICALADVANCESInvestments in clean water infrastructure contributedimportantly to life expectancy improvements before andduring the 20th century. Opportunities exist to consolidateand extend these improvements. The recent crisis in Flint,Michigan—centering on toxic levels of lead in the drinkingwater—underscored the importance of this infrastructureand the continuing problems with delivering clean water tohouseholds (Simon and Sidner 2016). Research has exploredthe sometimes subtle yet damaging effects of lead exposurefor children, as well as the role that public policy has playedin both creating and resolving this problem (Troesken 2006).As with clean water infrastructure, the benefits of antibioticsand vaccines are now largely taken for granted. However,with falling rates of vaccination in some locations (Lee,Rosenthal, and Scheffler 2013) and increasing rates of drugresistant bacteria (Andersson 2003)—including the firstcolistin-resistant bacteria in the United States (McGann etal. 2016)—protecting the modern victory over infectiousdisease has becom

dramatic: 15-year-olds in 1900 could expect to live 46.8 more years, whereas their counterparts in 2000 could expect to live 62.6 more years, an increase of almost 16 years. At the older end of the age distribution, 60-year-olds in 1900 and 2000 could expect to live 14.8 years and 21.6 mo