Börsch-Supan and Jürges 2012 — Disability, Pension Reform, and Early Retirement in Germany

Summary

Börsch-Supan and Jürges (2012) use Germany's five-phase pension reform history (1957–2008) as a sequence of quasi-natural experiments to show that Disability Insurance (DI) uptake and early retirement are driven by institutional incentive structures — not by health or mortality trends. Despite dramatic, monotonic improvements in German life expectancy over 50 years ($+7$–$8$ years at age 60–65), DI enrollment and early retirement rates followed the contours of legislative changes rather than health improvements. The paper's summary verdict: "disability insurance appears to be mostly a train on its own track, and it is largely unrelated to mortality or measures of health status over time and in a cross-national comparison."

Key Claims

• No cross-temporal health-DI correlation: DI uptake rose sharply after the generous 1972 reform and fell sharply after the restrictive 1984 reform, while mortality declined smoothly throughout. Changes in health cannot explain sudden discontinuities that track reform dates.
• No cross-national health-DI correlation: Survey of Health, Ageing and Retirement in Europe (SHARE), English Longitudinal Study of Ageing (ELSA), and Health and Retirement Study (HRS) data show cross-sectional health differences do not explain the wide variation in DI recipiency rates across European countries ($6.5\%$ in Germany, $<2\%$ in France, $>12\%$ in Sweden/Denmark/Netherlands among ages 50–64).
• Health explains within-country cross-sectional variation: At a given point in time, within a country, health status does explain who receives DI — health is not irrelevant; it is the institutional setting that overrides it in time-series and cross-national comparisons.
• Cohort birth-date discontinuities confirm institutional causation: The 1972, 1978, and 1984 reforms each generated sharp, predictable jumps in retirement hazard rates at specific ages for specific cohorts — changes too sudden and targeted to be explained by health shifts.
• Diagnosis composition shift (1983–2008):
  • Cardiovascular: $37$–$38\% \to 6$–$14\%$ (dramatic decline)
  • Mental illness: $\approx 8\% \to 31\%$ men / $43\%$ women ($4$–$5\times$ increase; #1 diagnosis by 2008)
  • Musculoskeletal: hump-shaped; peaked $\approx 33\%$ early 1990s, fell to $\approx 15\%$
  • Cancer: $7\% \to 15\%$ (nearly doubled)
  • Net: $\approx 50\%$ of men and $\approx 60\%$ of women entering DI have non-lethal conditions
• LFP vs. health (conditional on mortality risk): A German man with $1\%$ annual mortality risk (roughly age 60 in 1970) had $91\%$ labor force participation (LFP) probability in 1970 but only $57\%$ in 2000 — same health, dramatically different labor force attachment, driven by institutional liberalization.
• WWI cohort health shock: Cohorts born 1915–18 (widespread civilian hunger) show very-early-retirement rates (DI before 55) rising from $\approx 12$–$13\%$ to $17.2\%$ for the 1918 cohort ($+36\%$), then falling for postwar cohorts. Birth during wartime — an exogenous health shock — does affect DI, showing health matters when measurable through genuinely exogenous variation.
• Riester reform (2001) and Hartz reforms (2004): Key sustainability reforms that phased out the "bridge to retirement" (generous Unemployment Insurance (UI) for age 55–59 at $63$–$68\%$ of prior wages, no job-search requirements) and introduced actuarial adjustments ($3.6\%$/yr, capped at $10.8\%$ for disability pensioners).

Concepts Introduced or Extended

• DI Uptake and Health Decoupling
• Causal Effects of DI Receipt
• Morbidity-Mortality Distinction
• DI Growth Decomposition

Entities Mentioned

• Axel Börsch-Supan
• Hendrik Jürges

Quotes

"Disability insurance appears to be mostly a train on its own track, and it is largely unrelated to mortality or measures of health status over time and in a cross-national comparison."

"Specifically, there is hardly any cross-temporal correlation between DI uptake rates (or, more generally, old-age labor force participation) and available measures of health. We also know from other studies … that there is hardly any cross-national correlation between DI uptake rates and measures of health. This result holds notwithstanding that health does explain within-country variation at a specific point in time."

My Take

This paper's greatest value is the historical narrative — the five-phase German reform chronology is a clean demonstration that DI enrollment moves with policy, not health. The quasi-natural experiments from the 1972/1978/1984 reforms are compelling despite being visual/graphical rather than regression-based: the cohort-level hazard-rate discontinuities are large, correctly timed, and directionally consistent with theory. The cross-national finding (health cannot explain country-level DI rate variation) complements and extends the Austrian (Staubli 2011) and US Bartik IV (Autor-Duggan 2003) evidence. The main limitation is that the identification is largely descriptive — the "case studies" lack formal standard errors, and the smoothness of health trends is asserted rather than tested against a null. The WWI cohort case study is the most structurally credible piece, because birth year during the hunger years is genuinely exogenous to pension policy.