And Then You're Dead

The Autopsy / Case File · Disparities

A Decade Longer

No single diagnosis, injury, or drug explains this gap. It shows up as heart disease in one death certificate and diabetes in another, cancer in a third — different immediate causes stacked on top of one shared, underlying variable that a team of economists finally measured directly, at a scale no medical study had ever attempted: 1.4 billion tax records, linked to nearly 7 million deaths, sorted by income down to the percentile.

Published in JAMA in 2016, the study found the richest 1 percent of American men live 14.6 years longer, on average, than the poorest 1 percent; for women, the gap is 10.1 years. Between 2001 and 2014, that gap widened rather than closed: life expectancy for the top 5 percent of earners rose by 2.34 years for men and 2.91 years for women, while the bottom 5 percent gained just 0.32 years for men and a statistically negligible 0.04 years for women. And the gap for low-income Americans specifically wasn't uniform across the country — it varied by as much as 4.5 years depending on which metro area a poor person happened to live in, correlated with local smoking rates, immigrant population share, and government spending, but not correlated at all with local access to medical care. In some of those local areas, life expectancy for the poor rose more than four years across the study period even as it fell more than two years in others — evidence, the researchers wrote, that the widening national trend "is not immutable."

And Then You're Dead · September 2026 · 15 min read

Join the discussion →
Rows of white marble headstones in a national cemetery under a blue sky with a large tree in the foreground
Fort Scott National Cemetery, Kansas — where the 4.5-year life-expectancy range this study found among low-income Americans, depending on where they lived, is the difference between rows like these filling up years earlier or years later. Photo: Ser Amantio di Nicolao, CC BY-SA 4.0, via Wikimedia Commons.

The relationship between income and how long a person lives has been documented for decades, in dozens of smaller studies working from samples of thousands or, at best, a few hundred thousand people. What a team of Stanford, MIT, and Harvard economists led by Raj Chetty did differently in 2016 was scale: instead of surveying a sample and extrapolating, they obtained deidentified income and mortality data covering essentially the entire working-age and older U.S. population for sixteen years, and used it to calculate life expectancy directly, by income percentile, rather than by broad income bracket or self-reported survey category. The result is less a new discovery than a new level of precision applied to something researchers had suspected was true but had never been able to measure this exactly.

1.4 Billion Person-Years

The study linked two federal data sources under Internal Revenue Service and Harvard institutional review board authorization: deidentified income data from federal tax records covering every individual with a valid Social Security number from 1999 through 2014, and mortality data from Social Security Administration death records, which the researchers validated against National Center for Health Statistics vital records and found correlated above 0.98 across ages and years. The resulting sample covered 1,408,287,218 person-year observations for individuals aged 40 to 76, with a mean age of 53.0 and median household earnings, among working individuals, of $61,175 a year. Within that sample, 4,114,380 deaths occurred among men, a mortality rate of 596.3 per 100,000, and 2,694,808 deaths occurred among women, a mortality rate of 375.1 per 100,000 — nearly 6.8 million deaths total, sorted by household income percentile, sex, age, and geographic area, to calculate period life expectancy at age 40 for every point on the income distribution.

The researchers also built a specific safeguard into the design to guard against a subtler statistical trap: reverse causation, where a person's income falls near the end of life because they're already sick, which would make poor health look like a cause of low income rather than the other way around. For individuals 63 and younger, they measured income from two years earlier rather than the current year, and for those 63 and older, from age 61 — before the sharp rise in retirement that begins at age 62, Social Security's earliest eligibility age, when a person's income can suddenly look "poor" for reasons that have nothing to do with their long-run economic status. That lag is why the mortality analysis, despite pulling income data back to 1999, only reports life expectancy trends starting in 2001.

The Gap, and Which Way It Moved

The headline finding held throughout the income distribution: higher income was associated with greater longevity at every point measured, not just at the extremes. At the extremes specifically, the gap was stark. The difference in life expectancy between the richest 1 percent and poorest 1 percent of individuals was 14.6 years for men (95% CI, 14.4 to 14.8) and 10.1 years for women (95% CI, 9.9 to 10.3). The more consequential finding, for understanding where this is headed, was the trend over the study's fourteen-year window: between 2001 and 2014, life expectancy increased by 2.34 years for men and 2.91 years for women in the top 5 percent of earners, but by only 0.32 years for men and a statistically negligible 0.04 years for women in the bottom 5 percent — a difference the researchers found statistically significant for both sexes at P < .001. The country's overall life expectancy gains during this period, in other words, were disproportionately captured by people who were already living the longest.

0
years longer the richest 1% of American men live, on average, than the poorest 1% (10.1 years for women)
0
years of life expectancy gained by the bottom 5% of male earners from 2001 to 2014, versus 2.34 years for the top 5%

Zip Code Matters More If You're Poor

The study's third major finding complicates any simple national-average story: life expectancy for low-income individuals varied substantially depending on where they lived, in a way that life expectancy for high-income individuals largely did not. Among people in the bottom income quartile, life expectancy differed by approximately 4.5 years between the highest- and lowest-longevity local areas — and changes in that life expectancy between 2001 and 2014 ranged from gains of more than 4 years in some areas to losses of more than 2 years in others, meaning the trend for poor Americans wasn't uniformly flat nationally; it was flat on average while moving sharply in both directions locally. Nevada, Indiana, and Oklahoma had the lowest average life expectancies, under 77.9 years, for men and women in the bottom income quartile combined; eight of the ten lowest-ranked states formed a contiguous geographic belt running from Michigan to Kansas — Michigan, Ohio, Indiana, Kentucky, Tennessee, Arkansas, Oklahoma, and Kansas. California, New York, and Vermont had the highest life expectancies for the same bottom-quartile population, above 80.6 years. Life expectancy for the top income quartile showed a much weaker geographic pattern by comparison — areas that were bad for the poor and areas that were bad for the rich were often different places entirely, with Las Vegas posting the lowest life expectancy for top-quartile individuals in the study and Salt Lake City the highest.

Among high earners, where you live barely matters. Among the poorest quarter of Americans, it's worth 4.5 years of life — and the researchers found that gap correlated with smoking rates, not with how much medical care was available.

The South Isn't What the National Average Says It Is

One of the study's more counterintuitive findings only became visible because the researchers could disaggregate life expectancy by income rather than reporting a single regional average, something prior geographic-variation research hadn't been able to do. Pooling all income groups together, life expectancy in the American South runs well below the national average — a pattern documented repeatedly in prior public health research and generally attributed to the region's health profile. But among individuals specifically in the bottom income quartile, life expectancy in the South was much closer to the national mean, differing by less than a year for both men and women, a gap the researchers found was not statistically significant for women (−0.22 years, P = .47) and only marginally significant for men (−0.96 years, P = .03). The plainer reading of that result: the South's below-average life expectancy, as usually reported, is explained in large part simply by the region having more low-income residents overall, not by poor Southerners specifically living shorter lives than poor people anywhere else. Once you hold income constant, most of the regional gap disappears. What didn't disappear on a constant-income basis was the sharpest specific contrast the study turned up: a substantial life-expectancy difference between low-income individuals in Nevada and neighboring Utah, echoing an earlier, smaller finding from economist Victor Fuchs decades before this study had the data to confirm it at national scale.

What Predicts It, and What Doesn't

The study's final analysis tried to explain the local-area variation among low-income Americans by testing it against a set of area-level characteristics. Health behavior mattered broadly: differences in life expectancy across areas for people in the bottom income quartile were highly correlated with local rates of smoking, obesity, and exercise, with smoking showing the strongest single relationship (r = −0.69, P < .001) — areas with heavier smoking among the poor had shorter life expectancy for the poor, a robust finding that the researchers noted any competing theory of the income-longevity gradient would also need to explain. Three other characteristics showed a significant positive correlation: the local area's share of immigrants (r = 0.72), share of college graduates (r = 0.42), and per-capita local government expenditures (r = 0.57) — low-income people living in areas with more immigrants, more college-educated neighbors, and higher public spending tended to live longer, regardless of their own individual income. What showed no significant correlation at all is the more surprising half of the result: local access to medical care, physical environmental factors, local income inequality, and labor market conditions were not significantly associated with how long low-income people in a given area lived — a set of null findings that cuts against several of the most commonly proposed explanations for the income-longevity gradient, at least at the level of broad area-wide averages this study could measure.

And Then You're Dead

Fourteen years, sixteen data years of tax records, nearly seven million deaths, and the headline number that survives all of it is simple enough to say in one sentence: being rich in America buys, on average, more than a decade of additional life, and the country spent the years measured making that gap wider, not narrower. The zip code detail underneath it is almost worse — being poor in the wrong place costs years on top of being poor in general, for reasons that, according to the best data available, have less to do with hospitals than with who else lives nearby.

That's it. That's the whole thing.

Comments

Loading…

Comments are anonymous and moderated.

Sources