Maestas and Song 2011 — The Labor Supply Effects of Disability Insurance Work Disincentives

Summary

Maestas and Song analyze the natural experiment generated by the automatic conversion of disability insurance (DI) beneficiaries to the Social Security Old-Age (OA) program at Full Retirement Age (FRA), which abruptly eliminates the DI program's ~100,000% implicit marginal tax rate (MTR) on earnings above the Substantial Gainful Activity (SGA) threshold. Using the universe of primary-worker DI beneficiaries from the 1934–1942 birth cohorts (19.1 million person-year observations, 1995–2008), they find a significant work-disincentive effect concentrated among the ~12% of beneficiaries with recent labor force activity: among this group, annual earnings reverse their downward trend and rise sharply between ages 66 and 67 once the full cohort has converted to OA. Because the experiment occurs at the program exit margin — when beneficiaries are already in their mid-60s and the dominant trend in population labor force participation (LFP) is downward — the paper argues the estimates are best interpreted as a lower bound on the residual work capacity of all DI beneficiaries.

Key Claims

• Work disincentive is real but selective. In the full sample, labor supply (extensive and intensive margins) trends steadily downward with age with no detectable uptick at FRA. But among the ~12% with recent work activity (positive earnings in the prior year), an abrupt earnings reversal occurs between ages 66 and 67 once the entire cohort has converted to OA.
• Intensive margin dominates. The earnings response (conditional on positive earnings) is more pronounced than the extensive margin (LFP). Mean earnings for recent-work-activity beneficiaries fall from ~$9,300 (age 60) to ~$4,800 at age 66, then rise to ~$7,800 by age 67 — a near-63% recovery in a single year.
• Extensive margin response. Among early entrants with recent work activity, the fraction earning above the SGA threshold falls from ~24% at age 64 to ~15% at age 66, then climbs to ~25% at age 67.
• Early entrants respond more. Those who entered DI before age 59 show larger responses than those who entered at 59+, ruling out regression to the mean (biological recovery) as the explanation — those on DI longer show the largest work capacity response when the disincentive is lifted.
• All impairment categories respond. Earnings rise after conversion for all six major impairment groups, including musculoskeletal and mental disorders (the two fastest-growing diagnostic categories in the caseload since 1990). Musculoskeletal shows a ~31% earnings increase from $6,400 to $8,400 between ages 65 and 67 for late entrants.
• Robustness: birth-month placebo. December-born beneficiaries convert one year later than January-born; the earnings reversal is delayed by exactly one year for December births, confirming that the response is driven by the conversion event and not confounded by age-related health trends.
• 2000 earnings-test elimination amplified the experiment. Before 2000, DI-to-OA conversion relaxed the implicit MTR from ~100,000% to ~33% (OA Retirement Earnings Test at FRA). After 2000, the RET was eliminated, dropping the MTR to 0%. The post-2000 cohorts are therefore exposed to a stronger incentive change, making post-2000 responses a cleaner estimate of the full disincentive.
• Lower bound framing. Estimates are from the program exit margin (ages 65–67, general-population LFP declining). Any positive labor supply response detected under these conditions understates the disincentive effect on younger beneficiaries at earlier stages of their DI spell.

Concepts Introduced or Extended

• Causal Effects of DI Receipt
• Disability Insurance Program
• Vocational Grid

Entities Mentioned

• Nicole Maestas
• Jae Song
• Social Security Administration

Quotes

"Our estimates imply that the DI program depresses labor supply among even the oldest DI beneficiaries."

"Our use of quasi-experimental variation arising from the program exit margin, when individuals are already in their mid-60s and the dominant trend in labor force participation in the population at large is downward, suggests that our estimates are most appropriately viewed as a lower bound estimate of the residual work capacity of all beneficiaries."

My Take

The paper's main contribution is methodological triangulation: the entry-margin designs (Maestas, Mullen, and Strand [MMS] 2013, French-Song 2014) estimate the causal disincentive for applicants on the margin of award; this paper estimates the causal disincentive for the installed base of current beneficiaries using the FRA conversion as an exit-margin natural experiment. That both approaches find positive labor supply responses — at the margin of entry and at the margin of exit, for very different populations and ages — is strong evidence that the DI work disincentive is not a selection artifact. The lower-bound framing is conservative and credible. One limitation is the reliance on administrative earnings data, which can understate labor supply (informal work, self-employment misclassification) and is measured only annually, complicating identification of the exact conversion-month response. The finding that the response is largest for early entrants (those on DI longest) is notable: it rules out the hypothesis that the only workers who respond to incentive removal are those on the verge of natural recovery anyway.