Deshpande Gross and Su 2021 — Disability and Distress The Effect of Disability Programs on Financial Outcomes

Summary

Deshpande, Gross, and Su provide the first quasi-experimental evidence on disability programs' effects on financial distress. Using the universe of Social Security Administration (SSA) disability applicants (2000–2014) linked to nationwide records on bankruptcy, foreclosure, eviction, and home transactions, they document that applicants face peak financial distress at the time of application — an "Ashenfelter's peak" pattern — and that disability allowance dramatically reduces bankruptcy (−31%), foreclosure (−34%), and distressed home sales (−15%) within three years. The paper uses the same vocational grid age cutoffs as Chen and van der Klaauw (2008), augmented by a novel "office classification" strategy exploiting heterogeneity across Disability Determination Services (DDS) offices in applying the borderline age rule. The administrative data infrastructure built here is the direct empirical foundation for Deshpande and Lockwood (2022)'s ex-ante willingness-to-pay (EAWTP) welfare decomposition.

Key Claims

Ashenfelter's peak: Applicants' rates of bankruptcy, foreclosure, and eviction rise steadily before the application date, peak around the application, and fall afterward — both relative to the general population and relative to their own lifetime profile. Applicants apply when in peak financial distress.
Rates vs. general population: Conditional on homeownership, disability applicants' foreclosure rate is approximately twice the general population rate in the year of application; unconditional eviction rates are substantially higher.
Instrumental-variable (IV) estimates (within 3 years of initial allowance), in percentage points (pp):
- Bankruptcy: −0.77 pp (−31%), p<0.01; N = 2.22M
- Foreclosure (homeowners only): −1.75 pp (−34%), p<0.01; N = 0.60M
- Net home sale (homeowners only): −1.75 pp (−15%), p<0.01; N = 1.06M
- Net home purchase: +0.60 pp (+14%), p<0.05; N = 3.82M
- Eviction: estimates too imprecise to be meaningful (AIRS data covers ~40% of US residential areas)
Effects concentrated in year 1 and dissipate over time, mirroring attenuation in the first stage.
The treatment is timing, not permanence: Initial allowance rate jumps ~20 pp at age cutoffs; final allowance jumps only ~7 pp (denied applicants eventually reapply and are allowed). On average, those above the cutoff receive disability benefits for 0.9 additional months — not indefinitely.
Mechanism is a wealth effect: Cash transfer + health insurance coverage dominates. Credit access, credit demand effects, and program-rule incentive effects (bankruptcy exemptions, mortgage garnishment protections) all either work in the wrong direction or are too small to explain the results.
Total income effect near-zero: Allowance reduces annual earnings by ~ $1,150 but total income (earnings + benefits) by only ~$ 140, consistent with the disability insurance (DI) check mostly replacing foregone earned income with a more stable transfer.
Subgroup heterogeneity: Bankruptcy reduction stronger for women (−1.3 pp) than men (−0.4 pp); foreclosure and home sale reductions stronger for men and Social Security Disability Insurance (SSDI); home purchase increase concentrated in SSDI applicants. Supplemental Security Income (SSI) and SSDI applicants show similarly large bankruptcy reductions.
Welfare — tail risk: Mapping adverse events to consumption via Panel Study of Income Dynamics (PSID) event studies (foreclosure → ~ $6,300 annual drop in food+housing), incorporating tail consumption risk increases the optimal annual benefit by$ 50–240.
Welfare — spillovers: Averted foreclosures raise neighboring property values; 6.6% of the disability benefit amount accrues to neighboring homeowners. This raises the marginal value of public funds (MVPF) from 0.99 to 1.04. For context: averted-foreclosure property gains (~ $2,590 over 3 years) are 75% of the earnings reduction ($ 3,450) caused by disability allowance.
Optimal timing: The Ashenfelter's peak pattern combined with the causal estimates implies that awarding benefits sooner — shortening the 5-month statutory waiting period — would avert additional financial distress at peak need.

Concepts Introduced or Extended

Entities Mentioned

Quotes

"We provide the first evidence on the relationship between disability programs and markers of financial distress: bankruptcy, foreclosure, eviction, and home sale. Rates of these adverse financial events peak around the time of disability application."

"Disability allowance reduces the likelihood of filing for bankruptcy by a statistically significant 0.77 percentage point, or 31 percent, in the next three years. For homeowners, the likelihood of experiencing foreclosure in the three years after initial decision falls by 1.8 percentage points (34 percent)."

"Considering these extreme events increases the optimal disability benefit amount and suggests a shorter optimal waiting time."

My Take

The paper's most durable contribution is the "Ashenfelter's peak" descriptive fact, which reframes the work-disincentive debate: DI applicants are not strategically exiting employment to access benefits — they are applying after a period of escalating financial collapse. The causal estimates are credible but the identified treatment is narrow (receiving benefits ~0.9 months earlier, not indefinitely), so the welfare estimates are conservative lower bounds. The MVPF ≈ 1.0 result (rather than 1.42 as in Deshpande-Lockwood 2022) reflects this — the financial distress channel alone is not sufficient to generate a large MVPF once the labor supply cost is internalized. The paper's most important long-run contribution may be methodological: it built the administrative linkage (SSA 831 + Gross bankruptcy + CoreLogic + AIRS eviction) that powered Deshpande-Lockwood's full welfare decomposition.