Strand and Messel 2019 — The Ability of Older Workers with Impairments to Adapt to New Jobs

disability-insurancevocational-gridage-criteriaexaminer-ivlabor-supplyssdiwork-capacityinstrumental-variablesolder-workers

Summary

Strand and Messel (2019) use Social Security Administration (SSA) administrative data and an examiner instrumental-variable (IV) design — the same method developed in Maestas, Mullen, and Strand (MMS, 2013) — to estimate the residual work capacity of marginal Disability Insurance (DI) applicants at the age-55 vocational grid threshold. For applicants aged 55–56, DI allowance reduces above-substantial-gainful-activity (SGA) employment by at most 11.2 percentage points (pp), an upper bound on the share who would work if denied. Comparing this to the same IV estimate for ages 52–53 (−12.9 pp), the 1.7 pp difference is statistically indistinguishable from zero. The paper's core policy conclusion: raising the vocational grid's age thresholds would primarily change the pathway to allowance (initial award vs. Administrative Law Judge (ALJ) appeal), not the final allowance rate or sustained labor force participation.

Key Claims

Concepts Introduced or Extended

Entities Mentioned

Quotes

"The estimate of the employment effect of DI is smaller for persons aged 55 to 56 than for the full applicant population, which is consistent with an older applicant being in poorer health and thus less able to work."

"These results suggest that if the grid's age criteria were made more stringent, the effect would likely be a change in the pathway to DI, not a reduction in the labor market exit associated with DI receipt."

"Our estimates suggest that the vocational grid is more efficient at sorting persons by their capacity for work than prior literature suggested."

My Take

This is policy analysis at its most direct: SSA's own economists use the agency's administrative data and a validated IV method to test the empirical premise of the most-discussed DI cost-saving reform proposal. The null result — work-disincentive effects statistically identical on both sides of the age-55 threshold — is strong evidence against the reform's core assumption that the vocational grid misclassifies a large pool of workers who could realistically return to employment. The paper's methodological care in distinguishing the MMS method (appropriate for policy simulation of allowance-standard changes) from the Autor et al. delay method (appropriate for studying processing-time effects) is a genuine contribution to the causal-inference literature on DI. The main limitation is local average treatment effect (LATE) external validity: the estimates apply to the 17.3% marginal applicants identified by examiner assignment variation, not to the 76.7% inframarginal always-takers who likely have even less residual work capacity. The paper implicitly argues this makes the policy case for reform even weaker — the complier work capacity is an upper bound, and the inframarginal majority is almost certainly lower.