Autor and Duggan 2003 — The Rise in the Disability Rolls and the Decline in Unemployment

Summary

Between 1984 and 2001 the share of nonelderly adults receiving disability insurance (DI) rose $60\%$ to $5.3$ million beneficiaries, despite improving aggregate health. Autor and Duggan argue this is explained by three simultaneous forces: reduced screening stringency after the 1984 Reform Act, declining labor demand for less-skilled workers, and an unforeseen mechanical increase in the DI earnings replacement rate for low-wage workers driven by the progressive benefit formula interacting with stagnating low-skill wages. Using a Bartik industry-mix demand instrument and state-level panel data, they estimate the DI expansion lowered measured U.S. unemployment by approximately half a percentage point and doubled the labor force exit propensity of displaced high school dropouts after 1984.

Key Claims

• The 1984 inflection point: The Social Security Disability Benefits Reform Act of 1984 reversed the Reagan-era continuing disability review (CDR) clampdown (which had terminated $380{,}000$ beneficiaries), required benefits to continue during appeals, and liberalized mental impairment listing criteria. DI awards rose by one-third from the 1982 trough to a 1986 peak. Annual DI rolls then grew at $4.2\%$ per year.
• The replacement rate mechanism: The DI benefit formula is progressive: it replaces a higher share of earnings for low-wage workers. Because the Social Security Administration (SSA) computes benefits against a national average wage index that rose in real terms ($+21.6$ percentage points (pp), 1979–1995), while real wages of male high school dropouts fell sharply ($-19.1$ pp), the effective replacement rate soared for low-wage workers. Cash replacement rate for a 10th-percentile male aged 55–61: $52\%$ in 1979 $\to$ $74\%$ in 1999; including Medicare: $67\% \to 104\%$. At the 75th–90th percentiles, the rise was only $4$–$8$ pp.
• Who expanded: DI receipt more than doubled for male high school dropouts aged 40–54 post-1984. DI receipt among males 25–54 rose $50\%$, vs. only $25\%$ for males 55+. By 1999, a 40–54 year old male dropout was $4$–$5\times$ as likely to receive DI as an equivalent male with at least a high school degree.
• Diagnostic composition shift: The share of DI awardees with mental or musculoskeletal disorders (the lowest-mortality diagnostic categories) rose $60\%$ between 1983 and 1999; the shares for neoplasms and circulatory diseases fell $40\%$. Annual mortality among DI recipients fell $35\%$; exit to retirement declined $40\%$. The program increasingly covers conditions that are chronic but not imminently fatal.
• Demand shock → DI applications: Using the Bartik instrument (predicted employment growth from national industry trends applied to local industry mix), the authors show that post-1984 DI application rates in response to a given demand shock are $2$–$3\times$ higher than pre-reform levels. The reform permanently increased the labor force exit propensity of displaced high school dropouts by $\approx 60\%$.
• DI as hidden unemployment: The estimated effect of the DI expansion on aggregate unemployment among adults aged 25–64: approximately $-0.5$ percentage points ($13\%$) by 1998. Displaced workers who would otherwise be counted as unemployed exit to DI instead.
• Steady-state projection: Even absent further policy changes, DI recipiency was projected to rise a further $40\%$ from 2003 levels due to the mechanical replacement rate effect.

Concepts Introduced or Extended

• Causal Effects of DI Receipt
• DI Replacement Rate
• Conditional DI Applicants
• DI Growth Decomposition
• Disability Insurance Chronology
• DI Beneficiary Mortality

Entities Mentioned

• David Autor
• Mark Duggan
• Social Security Administration
• Disability Insurance Program

Quotes

"The sum of these forces doubled the labor force exit propensity of displaced high school dropouts after 1984, lowering measured U.S. unemployment by one-half a percentage point."

"Including the Medicare benefit, the total replacement rate for a male in the 10th percentile of the earnings distribution aged 55 to 61 rose from 67 to 104 percent."

My Take

This is the most-cited paper on DI growth, and its core mechanical insight — that a progressive benefit formula interacting with declining low-skill wages produces rising replacement rates — is undisputed and important. The replacement rate arithmetic is real; so is the diagnostic composition shift toward mental and musculoskeletal conditions.

The quantitative claims about the scale of the behavioral response are, however, subject to serious methodological critique from Pattison and Waldron (2013) and Liebman (2015):

The denominator problem (Pattison-Waldron): Autor and Duggan study the beneficiary-to-population ratio, which excludes working-age population growth from the denominator. Because the working-age population grew $\approx 53\%$ over 1972–2008, this exclusion mechanically inflates the apparent explanatory power of incidence-side factors (like the replacement rate). Pattison and Waldron decompose absolute entitlement growth and find that demographics — population growth, women entering the insured pool, baby boom aging into peak disability years — explain $\approx 90\%$ of the 1972–2008 increase in new disabled-worker entitlements. The adjusted disability incidence rate explains only $10\%$. Autor-Duggan's residual-after-demographics is large partly because they never subtract the population-growth component.

The base-year problem (Liebman): Autor-Duggan's empirical period effectively begins near the 1984–1985 CDR trough — the lowest DI incidence point in the modern era, immediately after the Reagan administration had terminated $380{,}000$ beneficiaries. Measuring growth from a local minimum inflates the apparent magnitude of subsequent incidence increases. Liebman (2015) shows that using a 1977 base (near the prior incidence peak), demographics dominate and incidence is negative over the full period. The strong 1985–1993 bounce-back that drives much of Autor-Duggan's result is a one-time policy correction, not a structural behavioral shift.

The steady-state prediction failed: DI rolls peaked around 2014–2015 and then declined, falsifying the $40\%$ further-growth projection. This is consistent with the Pattison-Waldron/Liebman view that the baby boom aging into disability years was a temporary force — once the boomers aged through the 45–64 peak risk window, incidence fell.

What remains: The replacement rate mechanism is a real and theoretically coherent explanation for who DI growth was concentrated among (low-education, low-wage workers) and why the diagnostic mix shifted toward subjective conditions. It is less defensible as an explanation for the aggregate magnitude of program growth, which is better explained by demographic transition. The Bartik demand-shock evidence that application rates are now more responsive to unemployment shocks is well-identified and survives these critiques. See DI Replacement Rate and DI Growth Decomposition.