Low and Pistaferri 2015 — Disability Insurance and the Dynamics of the Incentive-Insurance Tradeoff

Summary

Low and Pistaferri (2015) build and estimate a structural life-cycle model to quantify the incentive-insurance tradeoff in U.S. Disability Insurance. Using Panel Study of Income Dynamics (PSID) data on male heads with a high school education or less (1984–2008), they identify three health states (healthy, moderately work-limited L=1, severely work-limited L=2) and estimate via indirect inference that the DI system suffers far more from Type I errors (false rejections of the severely disabled) than from Type II errors (false awards to the moderately disabled). Counterfactual simulations show that the insurance value of higher generosity, looser screening, and more food stamps dominates the incentive cost across all policy experiments.

Key Claims

• Health shocks are quantitatively large: severe work limitation (L=2) reduces offered wages by 18%; moderate limitation (L=1) by 6%. Fixed costs of work are $4,301/quarter (95.2% of average wage) for the severely disabled, $2,472/quarter (54.7%) for the moderately disabled, and $0 for the healthy.
• Type I errors dominate: 67% of workers under 45 with severe work limitations who apply are rejected on first application; 37% for workers over 45. These are genuine false rejections — the severely disabled cannot return to work after denial.
• Type II errors are modest: ~17% of applications from moderately disabled (L=1) workers are awarded benefits; the award error rate (fraction of DI recipients not severely disabled) is approximately 12%.
• Application elasticity is highly heterogeneous by health state: overall DI application elasticity to benefit generosity = 0.62; for the moderately disabled (L=1) = 2.22 (very elastic, the main false-application margin); for the severely disabled (L=2) = 0.018 (near-zero — they apply regardless because work is infeasible).
• Nonemployment elasticity is at the low end of the literature: 0.056 overall, vs. the literature range of 0.06–0.93.
• Award rates match the literature: estimated initial award rate 0.42; 0.53 after two years; 0.73 after ten years — close to French and Song (2014) estimates of 0.39, 0.53, 0.67.
• False claimants are mostly recoverers: the stock of false DI claimants consists predominantly of individuals who recovered health while on DI and failed to exit, not people who were never disabled.
• Policy experiments (all revenue-neutral via wage tax adjustment):
  • More generous benefits → welfare increases; insurance value dominates false-application cost; low-productivity types benefit, high types lose.
  • Stricter screening → welfare decreases; young severely disabled already face 67% rejection and lack assets to self-insure; 1 percentage point consumption equivalent gain from reducing strictness from $\bar{S}=0.72$ to $0.60$.
  • More generous food stamps → large welfare increase (+0.9% per 10% generosity increase); food stamps substitute for moderately disabled (reduces false DI applications) but complement for severely disabled (enables application by raising the consumption floor); less distortionary than DI because does not require full labor force exit.
  • More frequent reassessment → small welfare decrease; removes some truly disabled alongside recovered false claimants.

Concepts Introduced or Extended

• DI Classification Errors
• DI Application Costs and Take-Up
• Causal Effects of DI Receipt
• Disability Insurance Program

Entities Mentioned

• Hamish Low
• Luigi Pistaferri

Quotes

"Government insurance against these shocks is incomplete: There are substantial false rejections for those in need. We estimate that 37 percent of older workers with a severe work limitation who apply for benefits are rejected on their first application, with an even greater rejection rate for younger, severely work limited individuals."

"Increasing the strictness of the DI screening process leads to a decline in welfare because the existing program already suffers from turning down large numbers of the severely disabled with little assets enabling them to self-insure."

"Increasing the generosity of food stamps leads to a fall off in false applications for DI and a reduction in misdirected insurance, leading to better targeting of DI and a substantial welfare improvement despite the extra cost of food stamps."

My Take

The paper's key contribution is quantifying that Type I errors dwarf Type II errors, and that the welfare cost of excessive strictness exceeds the cost of excessive generosity. The structural model separates the L=1 (elastic) and L=2 (inelastic) margins — not visible in reduced-form designs — showing that the severely disabled are essentially price-inelastic with respect to DI parameters while false applications are concentrated among the moderately disabled. The food stamps complement/substitute asymmetry is particularly valuable: it reframes the DI–food stamps interaction from substitution (reducing DI enrollment by giving people alternatives) to complementarity for the most vulnerable (enabling application by providing a consumption floor). The main limitation is that the paper does not model the appeal process, so it may underestimate how the severely young disabled eventually obtain coverage through repeated reapplication.