Summary
Mueller, Rothstein, and von Wachter (2016) ask whether Unemployment Insurance (UI) extensions during the Great Recession reduced Social Security Disability Insurance (SSDI) uptake, exploiting the dramatic variation in UI duration across states and time (26 to 99 weeks) as identification. Using three empirical strategies — national time-series regressions, state-by-month panel with fixed effects (FE), and event studies around UI extensions — they find no evidence that UI exhaustion causes SSDI applications. The preferred panel estimate rules out elasticities larger than 0.005 — far too small to be quantitatively meaningful. A supplementary Current Population Survey Annual Social and Economic Supplement (CPS-ASEC) analysis reveals why: only 28% of new SSDI awardees had any labor force attachment in the prior calendar year, and of those, only 4% received UI. The UI and SSDI populations are fundamentally distinct, ruling out program substitution as an explanation for SSDI's countercyclical pattern.
Key Claims
- Main null result: No evidence that UI exhaustion causes SSDI applications across three identification strategies. Panel FE analysis rules out contemporaneous application elasticities with respect to UI exhaustion larger than 0.005 at 5% confidence.
- Theoretical prediction that fails: A simple dynamic model (extending Autor-Duggan 2003) predicts a steady-state elasticity of ~0.5 if 17% of displaced workers are the "UI-before-SSDI" type. The empirical estimates are ~100× smaller.
- Population distinctness: In the year before SSDI award, only 28% of new awardees had any labor force (LF) attachment; only 6% spent any time looking for work; only 3% received UI benefits. For comparison, 92% of new UI recipients had LF attachment.
- The 28% subgroup: Even among SSDI awardees with LF attachment (avg. 36 weeks of work, avg. $641/week earnings — annual ~$23k), only 4% received UI in the prior year (10% if same-year UI is counted). These workers would be eligible for UI if displaced, yet they don't use it.
- Model reconciliation: The gap between predicted (
0.5) and estimated (0.005) elasticities closes by approximately a factor of four when the model's implied rate of UI-before-SSDI applicants is scaled by the empirical estimate that only 28% of awardees had any LF attachment.
- Event study mixed result: Studying weekly SSDI applications surrounding UI extensions shows a 2.5% decline in the weeks immediately following large extensions — but only for overlapping extensions that have no immediate effect on UI exhaustions. Nonoverlapping extensions (more interpretable as policy experiments) show no effect.
- Award rates unaffected: No evidence that UI extensions affect SSDI award rates, as would be expected if marginal recipients (lower work capacity) were being diverted.
- Cost-benefit calculation: Present value (PV) of a single SSDI award is ~$300k; weekly UI costs ~$300. Even four diversions per 1,000 UI extensions would break even. The data rule out even this minimal level of diversion.
- Implication for cyclicality mechanisms: Since UI exhaustion is not driving the countercyclical SSDI surge, other mechanisms must explain it — wage cyclicality (Davis and von Wachter 2011), employers' reduced willingness to accommodate disabilities in weak markets, the relative generosity effect (real wages fall in recessions while SSDI benefits do not), or examiners adjusting assessments to local labor market conditions.
- Consistency with Maestas, Mullen, and Strand (MMS, 2013): Maestas, Mullen, and Strand (2013) find ~23% of SSDI awardees are marginal workers capable of some employment. Mueller et al.'s finding that only ~3% of awardees had UI in the prior year is consistent — marginal workers are not using UI as a bridge to SSDI, regardless of how many are capable of work.
Concepts Introduced or Extended
- Conditional DI Applicants — directly constrains the UI-before-SSDI sub-type: only 28% of awardees have prior LF attachment; the conditional applicant pool is smaller and less UI-connected than the Autor-Duggan model assumed
- DI Growth Decomposition — rules out UI exhaustion as the mechanism driving cyclical SSDI applications; alternative mechanisms (wage cyclicality, employer accommodation, examiner discretion) must explain the pattern
- DI Application Costs and Take-Up — null interaction between UI exhaustion and SSDI applications is an important negative result for program interaction design
Entities Mentioned
Quotes
"Only 28% [of new SSDI recipients] spent even a single week working or looking for a job (compared to 92% for new UI recipients). Only 20% of these — about 6% of all new SSDI beneficiaries — reported even a single week of job search. Only 3% of new SSDI beneficiaries reported any UI income in the prior year."
"Taken together, the panel data analyses in tables 2 and 3 offer no sign that SSDI applications or awards respond to UI exhaustions. We can always rule out contemporaneous application elasticities larger than 0.005."
My Take
The paper's null result is important but should be interpreted carefully. The identification — UI duration variation during the Great Recession — is plausibly exogenous and exploited with appropriate state FE controls. The event-study approach is particularly credible for nonoverlapping extensions, which cleanly identify the effect of an actual reduction in UI exhaustion.
The key substantive contribution is the population distinctness finding. The 28% LF attachment statistic fundamentally reframes the SSDI-cyclicality debate: the rise in SSDI applications during recessions is not primarily driven by recently-displaced UI recipients exhausting their benefits and turning to SSDI. The populations are different from the start — most SSDI applicants were already labor-force-detached before the recession hit. This doesn't rule out cyclical mechanisms, but it eliminates the "UI bridge" story as a major one.
One limitation: the paper focuses on SSDI initiation during the Great Recession specifically. The UI-SSDI interaction might be stronger or weaker in other recession episodes, or when UI benefit levels (not just durations) vary. The 3% UI receipt rate is a cross-sectional CPS fact, not an experiment — we cannot rule out that some SSDI applicants strategically avoid UI receipt to preserve SSDI eligibility (explicitly noted in the paper as a possible explanation).
The data used — SSA Disability Research File (100% of 2008–10 applications) for weekly event studies, plus matched CPS ASEC for the population overlap analysis — are exemplary for this question.