Crimmins, Reynolds, and Saito 1999 — Trends in Health and Ability to Work Among the Older Working-Age Population

Summary

Using National Health Interview Survey (NHIS) data for ages 50–69 across 1982–1993 ( $N \approx 227{,}000$ pooled), Crimmins, Reynolds, and Saito ask whether the 1983 Social Security (SS) Amendments — which raised the Full Retirement Age (FRA) from 65 to 67 — rested on accurate assumptions about improving working-age health. The answer is yes for ages 60–69: work inability declined $\approx 24\%$ over 11 years, enough that disability levels at age 67 in 1993 matched those at age 65 in 1982. The improvement is mostly compositional (rising educational attainment of the cohort), though declining cerebrovascular/cardiovascular and arthritis prevalence contributed real biological change. No improvement occurred for adults in their 50s, and large racial disparities persist within education groups.

Key Claims

Among adults aged 60–69, self-reported inability to work declined significantly 1982–1993: annual regression coefficients $-0.38$ to $-0.71$ percentage points (pp)/year for men and $-0.52$ to $-0.71$ for women at ages 62–69 (ordinary least squares (OLS) on grouped NHIS data)
No significant improvement for ages 50–59; for some subgroups in their 50s, diabetes, musculoskeletal conditions, and mental disorders became more frequently reported as work-limiting, offsetting compositional gains
The time trend in work inability decreases by $\approx$ two-thirds when education is controlled (logistic regression), indicating compositional change is the dominant mechanism; but a significant residual remains, pointing to real underlying health improvement
Cerebro/cardiovascular disease prevalence fell $\approx 10\%$ ( $42\% \to 39\%$ ) and arthritis $\approx 10\%$ ( $31\% \to 28\%$ ) among 60–69 year olds over 1983–1993; probability of work limitation from these diseases fell $12\%$ and $25\%$ respectively
Improvement concentrated among those out of the labor force: healthier people are leaving the labor force earlier in later years, so the out-of-labor-force group appears healthier — a selection effect
Racial disparities persist within education groups: at age 65 with 10 years of education, $\approx$ 1 in 3 African American men are unable to work vs. $\approx$ 1 in 5 non-Hispanic White men; the $\approx 40\%$ Black-White gap exceeds the cross-age gap between 65 and 67
Policy conclusion: the SS retirement age increase is warranted for ages 60+, but persistent disability differentials by race mean the increase imposes disproportionate hardship on Black workers

Concepts Introduced or Extended

Entities Mentioned

Quotes

"The level of observed improvement in work ability means that the legislated rise in age of full eligibility for Social Security benefits should be more than compensated for by the improved ability to work."

"There are still differential abilities to work reported by gender, race/ethnicity, education and occupation. This means that subgroups are currently affected quite differentially by the rules for Social Security age of eligibility and this differential is unlikely to change."

My Take

A careful descriptive study with a clearly motivated policy question. Its most important finding is the compositional attribution: the working-age disability improvement of the 1980s is mostly a cohort-education effect, not a biological improvement in underlying morbidity. This is significant because the educational composition change will not continue indefinitely at the same rate, meaning the trend cannot be extrapolated. The finding of no improvement for 50s, and the racial disparity evidence, are the strongest arguments against naively using aggregate disability trends to justify raising the retirement age. The paper also establishes the empirical baseline that Crimmins et al. (2016) later extends — the 1982–1993 window corresponds to the post-1980 improvement phase documented in the longer time series.