Deshpande and Lockwood 2022 — Beyond Health Nonhealth Risk and the Value of Disability Insurance

Summary

Deshpande and Lockwood (2022) develop a sufficient-statistics framework for measuring the social value of U.S. disability insurance (DI) and apply it using Panel Study of Income Dynamics (PSID) consumption data merged with Social Security Administration (SSA) administrative records. Their central finding is that DI generates $8,700 in surplus per recipient — 64% more than a cost-equivalent tax cut — and that 63% of this surplus derives from insurance against nonhealth financial risks (mass layoffs, evictions, foreclosures, bankruptcies) rather than from the program's targeting of high-severity health conditions. The value of DI depends not on the SSA award process selecting impaired applicants but on the application process itself: the Substantial Gainful Activity (SGA) earnings limit screens out higher-earning, lower-need workers before they ever apply, so that the pool who choose to apply is disproportionately composed of financially vulnerable individuals regardless of their health severity.

Key Claims

• Four groups. The paper partitions the population of DI-eligible individuals into: L-DI (less-severe recipients), M-DI (more-severe recipients), M-NDI (more-severe nonrecipients), and L-NDI (less-severe nonrecipients). Health severity is classified by the SSA's Step 3/5 determination; DI status is from SSA 831 records.
• Pre-entry similarity of L-DI and M-DI. Less-severe recipients are not notably less financially vulnerable than more-severe recipients before program entry: L-DI and M-DI have nearly identical pre-entry consumption ($17,000 vs.$16,400), similar rates of adverse events (57% vs. 50% had at least one mass layoff, eviction, foreclosure, or bankruptcy in the prior four years), and similar counterfactual earnings ($6,600 vs.$3,600/year estimated from French and Song (2014)). Their marginal utility markups are also close: 0.88 (L-DI) vs. 0.95 (M-DI).
• M-NDI are better off. More-severe nonrecipients have 10–30% fewer adverse nonhealth events than either recipient group, are more likely to be married and educated, and have substantially higher counterfactual earnings (~$57,500/year estimated from French and Song). This directly refutes the assumption that health severity reliably predicts financial need within the applicant pool.
• Surplus values. Welfare surplus per recipient (ex-ante willingness-to-pay (EAWTP) minus program cost): All U.S. Disability Program (USDP) $8,700 (64$9,900; L-DI $7,700; M-NDI −$2,200 (negative: providing DI to this group would reduce welfare).
• Nonhealth risk dominates. Decomposing the insurance markup into health risk (between-health-category targeting) and nonhealth risk (within-health-category financial precarity sharing): 63% is nonhealth risk, 37% is health risk. Robust to a 30-category health classification (nonhealth share remains >50%).
• Selective application, not selective awards, drives value. Counterfactual experiment: randomly assigning DI to all individuals in the applicant categories (Random-App-DI) yields −$6,100 surplus — negative. Actual USDP produces$8,700 surplus. The difference is entirely attributable to who chooses to apply, not to how SSA adjudicates among applicants.
• SGA as the beneficial screen. Simulating a program that retains the SGA earnings limit but eliminates all medical screening (Earnings-Test-DI) produces a $5,900 surplus — positive, and recovering 68% of actual USDP's value. The SGA limit is the mechanism creating beneficial self-selection into the applicant pool.
• Comparison to perfect health targeting. A hypothetical program that targets only the most severely impaired applicants (perfect health targeting) would generate only $5,000 in surplus — less than the$8,700 from USDP — because it eliminates insurance coverage against the nonhealth risks that make less-severe applicants high-value recipients.
• MVPF. Marginal value of public funds (MVPF) for USDP reforms ≈ 1.42 vs. unemployment insurance (UI) reforms ≈ 0.61. DI generates substantially more surplus per marginal dollar than unemployment insurance because its applicant pool is more financially vulnerable and its self-selection mechanism is stronger.

Concepts Introduced or Extended

• Nonhealth Risk and DI Insurance Value
• DI Denied Population
• DI Application Costs and Take-Up
• Causal Effects of DI Receipt
• Disability Insurance Program

Entities Mentioned

• Manasi Deshpande
• Lee M. Lockwood
• Eric French
• Jae Song
• Nicole Maestas
• Kathleen Mullen
• Alexander Strand

Quotes

"The value of disability insurance stems primarily from its insurance against nonhealth financial risks, not its targeting of individuals with severe health conditions."

"Selective application, not selective awards, is the key mechanism generating the program's insurance value."

My Take

The paper's central contribution is a welfare decomposition that challenges the standard health-targeting rationale for DI. The finding that M-NDI are better off than recipients is particularly striking — it implies that the program is not just failing to reach some financially needy individuals, but that the individuals it misses are systematically less needy than those it serves. The policy implication — that the SGA earnings limit is a beneficial rather than harmful screen — inverts the usual framing (in which strict standards reduce access to deserving beneficiaries) and connects productively to the Deshpande-Li (2019) result (in which field office closings worsen targeting). The key distinction: SGA screens before application on work capacity (good screen), while distance screens before application on administrative navigation ability (bad screen).