Summary
Deshpande (2016) exploits a large fiscal-year (FY) 2004/05 cut in the Social Security Administration's (SSA) budget for childhood medical reviews as quasi-random variation in the probability of removal from the Supplemental Security Income (SSI) children's program. Using both a regression discontinuity (award date as running variable, looking back 3 years to the FY2001/02 budget-cut frontier) and a difference-in-differences (DD) design (FY2004 vs. FY2005 vs. FY2006 review cohorts), she finds that parents fully replace the lost SSI income with earned income — entirely on the intensive margin — when their child is removed from SSI. A 1,000lossinannualchildSSIraisesparentalearningsby700–$1,400 (elasticity ≥ 1). Disability applications by other family members fall 50–100% after removal, but disability receipt does not fall — the deterred applications were from marginal applicants who would not have been approved. The paper also documents substantial asymmetry: the entry-margin earnings elasticity (from an examiner-leniency instrumental variable [IV]) is ~0.29, far smaller than the exit-margin elasticity, suggesting households respond much more strongly to losing benefits than to gaining them.
Key Claims
- Full parental earnings offset, intensive margin only. Losing 1,000inannualchildSSI→+700 to +1,400inparentalearnings(dependingonspecification).Allontheintensivemargin:existingworkersworkmorehours;employmentrates(extensivemargin)donotrise.Singlemothers:+590 (All Review group, DD), virtually identical to full sample (+$620). Total household income does not significantly decline — parental earnings make up the shortfall.
- Earnings volatility also falls. Parents not only earn more but find more stable employment — the coefficient of variation of parental earnings declines, suggesting movement into steadier jobs.
- No substitution to other disability programs. Parent and sibling disability applications fall 50–100% after the child's removal (treatment-on-treated, DD). However, disability receipt by family members does not fall — the deterred applications were marginal (would not have been approved). Same mechanism as Deshpande-Li (2019): application costs or loss of information/familiarity deters the marginal applicant, not the inframarginal one.
- Asymmetric entry/exit response. Using disability examiner leniency as an IV for the entry margin of child SSI receipt (analogous to Maestas, Mullen, and Strand (2013) for adult disability insurance (DI)), the earnings elasticity for parents is ~0.29 — far smaller than the exit-margin estimate of ≥1. Households respond much more strongly to losing benefits than to gaining them, consistent with loss aversion, habit formation (consumption commitments), or target income behavior.
- Household-level clustering of disability applications. 65% of SSI children have either a parent or sibling who ever applies for disability. 15% have a family member apply within 60 days of the child's own application. The "Ashenfelter dip" in parental earnings in the years of child SSI enrollment — below the pre-SSI trend — is consistent with a household income shock (e.g., job loss) triggering simultaneous multi-member disability applications, not just individual health shocks.
- Heterogeneity. Youngest children's parents show the largest earnings response (largest first-stage, proportional earnings response). Low-severity conditions: earnings elasticity 0.94; medium-severity: 0.68. Non-mental diagnoses: 0.90; mental: 0.52.
- Normative ambiguity. Parents replace lost income, but increased work necessarily reduces time for child care of a disabled child. The welfare implications require measuring child health, education, and long-run employment outcomes — not available in this paper.
Concepts Introduced or Extended
Entities Mentioned
Quotes
"A loss of 1,000inthechild′sSSIpaymentincreasesparentalearnings—exclusivelyontheintensivemargin—by700 to $1,400, meaning that parents fully offset the loss in SSI payment."
"The loss of a child's SSI payment leads to a large decrease in SSDI and SSI applications by the parents and siblings of the removed child, though this does not translate into a decrease in actual receipt."
"The earnings IV estimate of 0.292 from the entry margin disability examiner instrument is in fact statistically smaller than the earnings IV estimate from my exit margin medical review instrument."
My Take
The most surprising finding is the large earnings elasticity — well above 1, far exceeding lottery-winner and welfare-program estimates of ~0.1. The most credible explanation is probably asymmetry: this paper estimates behavior on the loss margin, while most of the literature estimates behavior on the gain margin. Households that have been receiving SSI for years adapt to it as a reliable income stream (habit formation, consumption commitments); when it is removed, they respond more aggressively than a neoclassical model would predict. The Deshpande "forthcoming" American Economic Review (AER) companion paper (on 18-year-olds) finds no parental earnings response to removal at age 18 — a striking contrast that reinforces the asymmetry interpretation: 18-year-olds have been on SSI for a median of 7 years, generating strong habit formation, whereas the younger children in this paper have shorter SSI tenure and parents with more labor market attachment.
The application clustering finding is underappreciated. It reframes the decision to apply for disability as a household decision, not purely an individual health-driven one. Household income shocks — job loss, information acquisition about program availability — generate simultaneous multi-member applications. This has implications for interpreting application rate trends: a single economic shock can generate correlated applications across family members, creating multiplier effects on program take-up that individual-based models miss. See DI Application Costs and Take-Up and SSI Children's Program.