Gelber Moore and Strand 2017 — Disability Insurance Income Saves Lives

Summary

Gelber, Moore, and Strand use a Regression Kink Design (RKD) exploiting discontinuous slope changes in the disability insurance (DI) benefit formula to estimate the causal effect of benefit size on beneficiary mortality. Using administrative data on all 3.65 million new DI beneficiaries 1997–2009, they find that higher DI payments substantially reduce mortality among lower-income beneficiaries — an effect entirely absent from prior welfare analyses of social insurance programs. At the lowest-income group (lower bend point), a $1,000/year increase in annual DI payments reduces the annual mortality rate by 0.26 percentage points (pp), with an income-mortality elasticity of −0.56. The cost of saving a life-year at the lowest bend point ($58,574) is comparable to standard value-of-a-statistical-life-year (VSLY) benchmarks, implying these mortality gains are a large previously uncounted benefit of DI.

Key Claims

• Main effect: $1,000/year increase in annual DI benefits reduces annual mortality by 0.10–0.26 percentage points (range across bend points); income-mortality elasticity ≈ −0.6.
• Lower bend point (4th percentile of Average Indexed Monthly Earnings [AIME]; mean annual DI $8,543): −0.26 pp/year; elasticity −0.56. Estimate is a lower bound due to dependent-measurement attenuation.
• Family maximum bend point (30th percentile; mean $12,648 primary + $6,324 dependent): −0.09 pp/year on primary beneficiary mortality; elasticity −0.57.
• Upper bend point (84th percentile; mean $20,777): no robust effect; point estimates negative but small and insignificant.
• Effects are largest and most robust at lowest income levels, consistent with a concave income-health relationship and prior literature from Preston (1975) to Chetty et al.
• Cost-effectiveness: $58,574 per statistical life-year at lower bend point (p<0.05); $236,626 at family maximum bend point. VSLY lower boundary per expert panel: $50,000. DI passes the cost-effectiveness threshold for lowest-income beneficiaries.
• Baseline mortality context: 14% of 2006–2010 DI entrants died within four years — roughly ten times the working-age general population rate. Year-1 annual mortality rate: 7.0%; Year-4: 2.6%.
• DI income dependence: 80% of beneficiaries in households where DI > 50% of total income; 31% have no other income source.
• Mechanism (suggestive): Mortality effect is NOT mediated by labor supply (no earnings effect at lower/family maximum bend points). Survey evidence shows DI households' income-to-expenditure elasticity is 52% higher than non-DI households, concentrated in food, housing, utilities, healthcare, and transportation — basic consumption and life-sustaining inputs. Additional health expenditures are especially valuable during the 24-month Medicare waiting period.
• Effect heterogeneity: Largest mortality effects for Black beneficiaries, women, Disability Determination Services (DDS)-allowed (vs. hearing-allowed) beneficiaries, and cardiovascular/cancer conditions. Smallest for mental and musculoskeletal disorders (lower baseline mortality, harder to detect). The effect does not vary significantly by entry cohort (1997–2005 vs. 2006–2009) or age at filing.
• Identification validity: Density of AIME, all predetermined covariates (age, sex, race, disability type, path to allowance), and Supplemental Security Income (SSI) receipt rate are all smooth through the bend points. Placebo tests on non-DI workers and on beneficiaries without dependents at the family maximum bend point all show no effect. Permutation tests confirm that the largest elasticities occur precisely at the actual bend points.

Concepts Introduced or Extended

• Regression Kink Design — applies RKD to the lower and family-maximum Primary Insurance Amount (PIA)-AIME bend points; establishes that $1k/yr in DI income causally reduces mortality by 0.26 pp at cost ~$59k/life-year
• Causal Effects of DI Receipt
• DI Beneficiary Mortality — causal income-mortality effect; RKD estimates; cost-per-life-year
• Disability Insurance Program — mortality as a previously uncounted benefit; welfare analysis implications

Entities Mentioned

• Alexander Gelber
• Timothy Moore
• Zhuan Pei
• Alexander Strand
• David Weaver
• David Cutler

Quotes

"We show that higher payments from U.S. Social Security Disability Insurance (DI) reduce mortality."

"We estimate that $1,000 in annual DI payments decreases the annual mortality rate of lower-income beneficiaries by approximately 0.1 to 0.25 percentage points, implying that the elasticity of annual mortality with respect to annual DI income is around -0.6."

"The mortality effects imply large benefits that have not been taken into account in the welfare analysis of DI and other social income insurance programs."

"Our results show that the lifespan of individuals in the U.S. with disabilities and low lifetime income can benefit from additional income in ways that are similar to individuals in less developed economies or from earlier time periods."

My Take

The identification strategy is unusually clean for a health-income study: the bend points in the PIA-AIME formula are created by administrative formula design, are nearly invisible to beneficiaries (who cannot easily calculate or manipulate their AIME), and affect only benefit size — not Medicare eligibility, program rules, or anything else. The placebo tests and validity checks are thorough. The null result at the upper bend point is coherent with a concave income-health relationship and provides an important internal control. The mechanism evidence is necessarily suggestive rather than causal, but the hypothesis (basic consumption prevents premature death among very low-income, very sick people) is mechanistically plausible and consistent with the heterogeneity by disability type. This paper closes a significant gap in the DI welfare literature by putting a credible dollar value on the life-extension benefit of benefits.