Benítez-Silva and Yin 2011 — Disability Insurance Applications near Retirement Age

Summary

Benítez-Silva and Yin (2011) exploit the Early Retirement Age (ERA)–Normal Retirement Age (NRA) window (ages 62–65) as a quasi-experiment to isolate cash-benefit effects on Disability Insurance (DI) application decisions. Because DI provides Medicare coverage only after a 2-year waiting period, applicants in this age window will reach age-65 automatic Medicare eligibility before their DI Medicare benefit would activate — so their application calculus is driven purely by the cash benefit differential relative to Old-Age (OA) early retirement. Using a life-cycle structural model calibrated to Health and Retirement Study (HRS) and Social Security Administration (SSA) administrative data, the paper simulates elasticities of DI applications with respect to benefit levels and award probabilities at different age groups and compares them to earlier estimates.

Key Claims

• Benefit elasticity for ages 62–NRA ($3.2$–$3.9$) is substantially higher than for younger ages ($0.3$–$2.2$), because the concurrent availability of OA early retirement sharpens the financial trade-off in the ERA–NRA window.
• Award probability elasticity exceeds benefit elasticity at all age groups — potential applicants respond more strongly to changes in the probability of approval than to equivalent proportional changes in the benefit amount.
• Responses are asymmetric: benefit increases and reductions of equal magnitude do not produce equal-and-opposite elasticity estimates.
• Dynamic timing effects: when benefits increase, younger cohorts apply earlier, so the middle-aged group (45–59) actually shows fewer applicants than baseline in some simulations — a negative elasticity that is invisible in static or reduced-form approaches.
• Prior estimates in the literature (Leonard 1979: $\approx 0.35$; Halpern and Hausman 1986: $\approx 1.3$) likely overstate the pure cash-benefit elasticity because they conflate the Medicare incentive with the cash-benefit incentive.
• The paper's own younger-age elasticity estimates are also likely overstated because Medicare is not yet incorporated as a benefit in the structural model.

Concepts Introduced or Extended

• DI Replacement Rate
• DI Application Costs and Take-Up
• Early Retirement Safety Net Gap

Entities Mentioned

• Hugo Benítez-Silva
• Na Yin

Quotes

"The literature estimating the effect of benefit levels on the Social Security Disability Insurance (DI) application decisions has not been able to separate the effect of Medicare coverage provided by the program on application decisions from the effect of cash-benefits, and therefore previous estimates are likely to be overstated."

"Our estimated benefit elasticity (and award probability elasticity) at age 62 to NRA, when Medicare incentives are virtually non-existent, is much higher than that at younger ages (age 21 to 61)."

My Take

The Medicare-cash decomposition is the paper's key identification insight, and it is genuinely clever: the 2-year DI Medicare waiting period creates a natural window near retirement age where Medicare incentives are off the table, allowing the authors to back out a cleaner estimate of the pure cash-benefit elasticity. The finding that elasticity is higher near ERA ($3.2$–$3.9$) than at younger ages ($0.3$–$2.2$) is counterintuitive but follows directly from the OA trade-off: the foregone OA early retirement benefit is a concrete, salient alternative that sharpens the financial calculation. The award probability elasticity exceeding benefit elasticity at all ages is consistent with the Deshpande et al. (2025) null result on Administrative Law Judge (ALJ) award rates — potential applicants may be more sensitive to their perceived probability of approval than to the dollar level of the benefit, though the mechanisms differ (rational expectation formation vs. attentional salience). The model's acknowledged limitation — Medicare not yet modeled as a benefit — means all younger-age elasticities remain suspect, and the paper is best read as a partial result for the near-retirement subgroup. The 2011 working-paper status is worth noting; it is unclear whether this was subsequently published.