Summary
Case and Deaton extend their 2015 Proceedings of the National Academy of Sciences (PNAS) paper through 2015, confirming that white non-Hispanic (WNH) midlife mortality continues rising while every peer country keeps declining. The central contribution is the shift in explanatory framework: from opioid supply as the proximate driver to cumulative disadvantage — a cohort-level deterioration in labor market entry conditions that cascades across wages, marriage, child outcomes, and social belonging, ultimately producing despair and addiction. The paper also systematically tests and rejects contemporaneous income/unemployment as the primary driver of the mortality reversal.
Key Claims
Mortality facts (updated through 2015)
- WNH all-cause mortality age 50–54 increased at +0.5%/year 1999–2015; 14 comparison countries declined at an average −1.9%/year over the same period.
- WNH deaths of despair (drugs, alcohol, suicide) age 50–54 doubled from ≈40 to ≈80 per 100,000 between 1990 and 2015; comparison countries converged to ≈40.
- Slowdown in heart disease decline: WNH age 45–54 went from −2%/year (1990–99) to −1%/year (2000–14), then essentially flat 2009–2015; other countries continued declining.
- Heart disease −1.0%/year and cancer −1.1%/year for WNH age 50–54 (1999–2015) vs. −2.7%/−2.4% for Black non-Hispanic (BNH) and −2 to −5% for most comparison countries.
The education gradient
- WNH without a BA (both sexes): mortality rising at every five-year age group 25–29 through 60–64, 1998–2015.
- WNH with BA+: mortality falling at every age group, with larger decreases at higher ages.
- Mortality divergence is not just widening but opposite in direction: men 50–54 without BA 762→867/100k; men with BA+ 349→243/100k (1998–2015).
- The education split is clean: even "some college" (without a BA) shows rising mortality.
Racial mortality crossover
- WNH high-school (HS)-or-less age 50–54: went from 30% lower than all-BNH mortality in 1999 (722 vs. 945/100k) to 30% higher by 2015 (927 vs. 703/100k).
- BNH mortality fell −2.3%/year and Hispanic mortality fell −1.9%/year 1999–2015 for ages 50–54.
- The racial gap in mortality among the least educated has nearly disappeared.
Morbidity confirms cohort-by-cohort steepening
- Self-reported excellent/very good health (Behavioral Risk Factor Surveillance System (BRFSS)): fraction reporting excellent health at age 50 fell 7 percentage points (pp) for HS-or-less WNHs 1999–2015; unchanged for BA+ WNHs.
- Reports of chronic pain, sciatic pain, mental distress, and activities of daily living (ADL) difficulties all rise with each successive birth cohort for non-BA WNHs; birth cohort profiles are parallel and flat for BA+ WNHs.
- This is not just a level shift: the slope of the age-outcome profile steepens with each birth cohort, indicating accelerating deterioration.
Income explanation fails (Section II)
- Aggregate WNH median household income co-moves with all-cause mortality (both turning down ~1999), but this breaks down on disaggregation:
- BNH and Hispanic incomes tracked WNH incomes (college-educated BNHs even experienced steeper post-1999 income declines than comparable WNHs), yet their mortality kept falling at −2 to −3%/year.
- By education group, median household incomes of HS-or-less and BA+ WNHs do not diverge in the way mortality does — there is no income analog to the mortality scissors.
- European countries hit by the Great Recession (Ireland, Spain, UK, Netherlands) showed income declines comparable to the U.S. without any slowdown in mortality.
- The aggregate income-mortality co-movement is spurious: two smooth independent trends (deaths of despair rising + heart disease slowing) together generate a U-shape that coincidentally matches the household income inverse-U.
Cumulative disadvantage hypothesis (Section III)
- Not contemporaneous income but initial labor market conditions at cohort entry — a latent factor Xβ indexed to birth cohort — drives the pattern.
- The factor has been worsening since cohorts born in the early 1970s (who entered the market after the early 1990s, the end of the "blue-collar aristocracy").
- Using a latent factor model, birth-cohort effects for suicide, chronic pain, sciatic pain, mental distress, heavy drinking, difficulty socializing, drug/alcohol poisoning, never-married rate, and male labor-force non-participation all rise in parallel for WNH non-BA cohorts born 1940–1988 — consistent with a single common latent factor.
- The latent factor maps to declining real wages at cohort entry for men without a BA; wages for BA+ cohorts rose 10% across the same birth years.
- The cascade: worsening labor market → withdrawal from labor force → declining marriageability → unstable cohabitation → erosion of community and religious structures → loss of meaning and purpose → despair, addiction, suicide.
- Opioids are an accelerant (especially for drug overdose), but the same pattern holds for suicide and alcoholic liver disease, which are unrelated to opioid supply — requiring an explanation that covers all three.
Opioids as accelerant, not cause
- Prescription opioid supply is "a proximate cause" of rising overdoses but not the underlying driver of the full deaths-of-despair cluster.
- Purdue Pharma: $31B in OxyContin sales through mid-2016. Food and Drug Administration (FDA) approval of a drug class that killed ≈200,000 is critiqued.
- As opioid prescriptions tightened (post-~2012), deaths from illegal drugs (heroin, fentanyl) became relatively more important — the demand for pain relief shifted supply, not the reverse.
Engagement with the 2015 critique
- Explicitly incorporates Gelman-Auerbach (2016) age-adjustment: age-adjusted figures still show WNH women +36/100k and WNH men +9/100k in the 45–54 group 1998–2015. The paper now uses five-year age groups throughout to mitigate the aggregation criticism.
- Main point is reframed as U.S. failure to improve rather than absolute increase: other countries improved by 2%/year; U.S. WNHs did not.
Concepts Introduced or Extended
- Deaths of Despair — extended: education gradient, racial crossover, cohort steepening, income refutation, cumulative disadvantage hypothesis
- Income-Mortality Gradient — extended: contemporaneous income as inadequate explanation for the mortality reversal; cohort-entry wages as the driving variable
- Period Mortality — extended: cohort analysis (birth-year effects on mortality and morbidity profiles)
- Morbidity-Mortality Distinction — extended: morbidity data confirm and parallel the mortality deterioration for non-BA WNHs
Entities Mentioned
Quotes
"Our main point is that other wealthy countries continued to make progress while the United States did not."
"We see our story as about the collapse of the white working class after its heyday in the early 1970s, and the pathologies that accompany this decline."
"Virtue is easier to maintain when it is rewarded."
"It is not the last 20 years that matter, but rather the long-run stagnation in wages and incomes has bred a sense of hopelessness."
My Take
The shift from Case and Deaton (2015) to (2017) is conceptually important: the 2015 paper documented a phenomenon; the 2017 paper offers a causal framework. The cumulative disadvantage hypothesis is compelling and consistent with the data — specifically, the birth-cohort latent factor analysis is harder to dismiss than aggregate income correlations. The most powerful piece of evidence is the racial comparison: Black and Hispanic incomes tracked WNH incomes (or did worse), yet their mortality kept falling. Any contemporaneous income-based explanation has to explain that anomaly. The cumulative disadvantage story offers a candidate (relative upward mobility vs. downward mobility for WNHs vs. BNHs/Hispanics, differences in community institutions), though it remains difficult to pin down econometrically. The paper's contribution to the wiki context is primarily as the theoretical foundation for understanding why the Disability Insurance (DI)/general-population mortality ratio may continue to change in ways not fully captured by the actuarial DI mortality literature.