Chen 2012 — Spousal Labor Supply Responses to Government Programs Evidence from the Disability Insurance Program

Summary

Using Survey of Income and Program Participation (SIPP) panels matched to Social Security Administration (SSA) administrative earnings records and Disability Insurance (DI) award files (1980–2008; $N = 39{,}519$ wives), Chen estimates the crowd-out effect of husband's DI receipt on wife's labor supply via a fixed-effects event study that compares wives of accepted vs. rejected DI applicants. Both groups have disabled husbands, so the disability shock is common; the post-award divergence in wives' labor supply identifies the crowd-out of DI income on the predicted Added Worker Effect (AWE). DI crowds out wife's labor force participation (LFP) by $6\%$ starting one year after award, growing to $8.4\%$ by year 5 — but wives of beneficiaries paradoxically earn more than wives of rejected applicants, because the wives who drop out are lower earners. Crowd-out is largest for younger wives and for wives of marginal (Stage 5) or musculoskeletal-impaired husbands.

Key Claims

• Added Worker Effect (AWE): Disability is a permanent income shock. Standard theory predicts wife should increase labor supply (income effect: leisure is a normal good). DI income replaces $\approx 57\%$ of median beneficiary's pre-application earnings, potentially crowding out this response.
• Identification: Fixed-effects Jacobson-LaLonde-Sullivan event study design. Comparison group = wives of rejected DI applicants (same disabled husbands; DI award is the only difference). Pre-award parallel trends confirmed (no statistically significant pre-award divergence). Estimated effect is an upper bound on crowd-out (some rejected applicants eventually receive DI via re-application or appeal).
• Main result — LFP: DI crowds out wife's labor force participation by $-6\%$ at year 1, growing to $-8.4\%$ by year 5. Effects significant through year 5.
• Main result — Earnings: No crowd-out on earnings; wives of beneficiaries earn $\approx \$1{,}644$ more at year 2 and $\approx \$2{,}479$ more at year 5. Paradox explained compositionally: lower-earning wives exit the labor force (LFP crowd-out); the conditional mean of remaining earners is higher.
• Pre-award dip: Both groups of wives reduce labor supply $\approx 1$ year before the husband's award decision — direct evidence of an AWE response to the disability shock itself, before DI income arrives.
• Age cohort heterogeneity: Young cohort (born 1946–1976): crowd-out $-8.8\%$ LFP at year 1, $-\$1{,}830$ earnings, but short-lived (1 year). Old cohort (born before 1946): no significant crowd-out. Counterintuitive: DI replaces more income for older (higher-earning) husbands, but their wives have lower labor force attachment, leaving less AWE to crowd out.
• Stage 5 subsample: "Marginal" applicants assessed on both medical and vocational criteria ($\approx 40\%$ of all applicants); health severity more homogeneous within group. Crowd-out: $-8$ to $-11\%$ LFP through year 5. Cleanest causal estimate.
• Musculoskeletal impairment: All women: $-8.5$ to $-14.2\%$ LFP crowd-out; young women: up to $-21\%$ in year 1. Musculoskeletal is the fastest-growing DI impairment category ($\approx 28\%$ of beneficiaries).
• Data: SIPP panels (1990–93, 1996, 2001, 2004) matched to SSA 831 file, SSA Master Beneficiary Record, and SSA earnings histories. Panel spans 10 years pre-award to 8 years post-award; unit of observation is individual-year; $N = 39{,}519$ women.
• Sample profile: Average wife age at award: $49$; $23\%$ nonwhite; $33\%$ more than high school (HS) education; $64\%$ working at award date; average earnings $\$21{,}031$ (2004 $). Average time from filing to final award: $\approx 1$ year.
• Comparison to prior work: Coile (2004): $15\%$ LFP crowd-out (larger; uses Health and Retirement Study (HRS), different sample); Cullen and Gruber (2000): $30\%$ Unemployment Insurance (UI) crowd-out of wives' hours; Bound, Burkhauser, and Nichols (2003): no effect of DI on spousal earnings (self-reported SIPP data, shorter time window).

Concepts Introduced or Extended

• Added Worker Effect
• Earnings-Participation Paradox
• Causal Effects of DI Receipt
• Nonhealth Risk and DI Insurance Value

Entities Mentioned

• Susan Chen
• Disability Insurance Program
• Social Security Administration

Quotes

"DI crowds out spousal labor force participation by 6 percent and the displacement spans multiple years."

"The estimated crowd-out effects are also larger for younger wife cohorts and cohorts with particular types of impairments such as musculoskeletal disease."

"The causal effects that I identify can be interpreted as upper bounds on the crowd out effect of DI."

My Take

The earnings/participation paradox is the most important finding and deserves more attention than Chen gives it: $6\%$ fewer wives work but working wives earn $\$2{,}479$ more by year 5 — the full distributional effect is a compositional selection story, not a simple income-transfer story. The old-cohort null is puzzling and theoretically important: if DI replaces more income for older husbands (higher earners), the income effect on wives should be larger, not smaller. The resolution is probably that older wives had weaker pre-shock labor force attachment, so the AWE they could exhibit was smaller to begin with — making the crowd-out look like zero even if the crowd-out rate is similar. The Stage 5 and musculoskeletal estimates are the cleanest because they hold health severity roughly constant across accepted/rejected pairs; the $-21\%$ result for young wives with musculoskeletal husbands is startling and has direct policy relevance given musculoskeletal disease's growing share of DI caseloads. The paper is a working paper dated January 2012; not all estimates may survive peer review.