Summary
Waldron uses the 1973 Current Population Survey (CPS) matched to Social Security administrative records through 1998 to examine the link between age of retired-worker benefit entitlement and subsequent mortality for 10,938 men surviving to at least age 65, born 1906–1932. Logistic regression controlling for current age, birth cohort, education, marital status, and race finds a monotone gradient: the earlier men claim, the higher their mortality, with exact-age-62 claimants facing a 38% greater odds of dying than age-65+ claimants — equivalent in magnitude to lacking a high school diploma. The discontinuity between exact-age-62 claimants and men retiring even a few months later identifies a subgroup rushing to claim at first eligibility, interpreted as quasi-disabled workers who cannot qualify for Disability Insurance (DI) benefits but who are too ill to continue working. A second mechanism — benefit optimization, in which men with below-average life expectancy (LE) rationally claim early because the actuarial reduction favors the short-lived — explains the continuous gradient across all early-retirement ages.
Key Claims
- Men claiming exactly at age 62 have 38% greater odds of dying vs. men claiming at 65+, and 12% greater odds vs. men retiring just 3–11 months later (62.25–62.92). The within–age-62-year discontinuity is the primary fingerprint of the quasi-disability subgroup.
- The mortality gradient is continuous: exact 62 (38% excess) > 62.25–62.92 (24%) > 63 (13%) > 64 (12%) > 65+ (reference). The smooth portion of the gradient is consistent with benefit optimization operating in addition to quasi-disability.
- Collapsing early retirees into a single "62–64" category mutes the age-62 spike: the odds ratio falls from 1.38 to 1.24. Models that do not separate exact-62 claimants from age-63–64 claimants understate the mortality disadvantage of the most vulnerable subgroup.
- Education gradient magnitude is comparable: less than 12 years of school → 38% greater odds of dying vs. college-educated. A greater share of age-62 claimants are low-educated, so the two risk factors compound.
- Mortality differentials by retirement age, education, and marital status all shrink at older ages — consistent with frailty/survivor selection: the frailer low-socioeconomic status (SES) and earliest-retiring members die first, leaving a progressively more resilient survivor pool. This confirms that at older ages, the measured gap understates the underlying health difference.
- Education mortality differentials widen across birth cohorts 1906–1932, foreshadowing the widening SES-mortality gradient documented at larger scale in Waldron (2007) and Chetty et al. (2016).
- Data: 1973 Exact Match (CPS matched to Master Earnings File, Master Beneficiary Records, Numident through 1998); 10,938 retired-worker males surviving to age 65, birth cohorts 1906–1932. Disability recipients excluded from sample.
Concepts Introduced or Extended
- Select and Ultimate Mortality Tables — benefit-optimization mechanism as frailty selection into early claiming; compositional sensitivity of retirement-age categories; age-interaction shrinkage of differentials consistent with survivor selection
- Income-Mortality Gradient — foundational paper in the Waldron series; education gradient within a Social Security subpopulation; widening education-mortality differentials across birth cohorts 1906–1932
- Period Mortality — age-interaction effects show differentials shrinking at older ages; frailty selection operating within retired-worker subpopulation
- Disability Insurance Program — quasi-disability mechanism: workers too ill for continued employment but too healthy to qualify for DI cluster at age 62
Entities Mentioned
Quotes
"Men taking benefits at exactly age 62 have higher mortality risk than men taking benefits in any of the other four age groups."
"Evaluations of proposals to index the normal retirement age to improvements in average longevity may want to consider how differing life expectancies could affect the distributional effects of such a proposal."
My Take
This is the foundational paper in Waldron's early-retirement mortality series, establishing the core findings that she refines in the 2004 trichotomy study and the 2007 earnings-gradient study. The key methodological contribution — that collapsing early retirees into a single category mutes the age-62 spike — applies broadly: any model using "early retirees" as a homogeneous counterfactual for DI beneficiaries or retirement policy simulations will understate effects on the most health-compromised subgroup. The benefit-optimization explanation matters directly for DI research: if below-average-LE workers self-select into age-62 claiming, the age-62 retirement population is not a healthy control group — it too contains adverse health selection on mortality. This logic reappears explicitly in Waldron (2004) as the trichotomy model and in the Select and Ultimate Mortality Tables frailty-selection framework. The paper lacks an earnings variable (education is the only SES proxy), which Waldron herself flags as a limitation likely to attenuate the retirement-age coefficients — addressed in the 2004 and 2007 follow-up papers.