Lindner and Nichols 2012 — The Impact of Temporary Assistance Programs on Disability Rolls and Re-Employment

Summary

Lindner and Nichols (2012) ask whether participation in four temporary assistance (TA) programs — unemployment insurance (UI), the Supplemental Nutrition Assistance Program (SNAP), Temporary Assistance for Needy Families (TANF), and temporary disability insurance (TDI) — causally affects re-employment and applications for disability insurance (DI) and Supplemental Security Income (SSI) among job losers. Using Survey of Income and Program Participation (SIPP) data 1996–2010 matched to Social Security Administration (SSA) administrative records and instrumenting for program participation with state policy rules (eligibility provisions, benefit formulas, biometric requirements), they find that UI participation deters DI applications while SNAP participation may increase SSI applications, with no robust causal effect on re-employment. The paper introduces the principle that the sign of cross-program effects is predicted by whether the TA program and the disability program share the same target population (work-history-based vs. means-tested).

Key Claims

UI → DI (negative, causal): Instrumental variables (IV) estimates show UI participation significantly reduces DI applications. The substitution effect dominates: UI income serves as a substitute for DI cash income, reducing the marginal value of applying. Consistent with Lindner (2011) and Rutledge (2012).
SNAP → SSI (positive, mixed): IV estimates show SNAP participation increases SSI applications in monthly regressions (significant), but not at the spell level. Both programs are means-tested with overlapping target populations; caseworker referrals and institutional complementarities may contribute.
TANF, TDI (null causal effects): No significant causal effects in IV. TDI's positive ordinary least squares (OLS) coefficient on DI applications — plausibly reflecting sicker workers selecting into TDI — flips negative (though insignificant) in IV, suggesting selection explains the OLS result.
Re-employment (null causal effects): OLS shows all four programs negatively associated with re-employment, but IV produces large positive and insignificant coefficients. Selection — not causal program effects — explains the raw negative correlation.
Target population overlap principle: The sign of the cross-program effect is determined by whether the TA program and the disability program share the same eligibility structure: UI and DI are both work-history-based (UI income deters DI application); SNAP and SSI are both means-tested (SNAP facilitates SSI application or at least does not deter it).
Effect magnitude: Where significant, effects are large relative to base rates — UI→DI coefficient ≈ 5–6× the mean DI application rate per spell; SNAP→SSI ≈ 5× the mean SSI application rate. Consistent with large effects for the marginal complier subpopulation.
Weak instruments: Kleibergen-Paap Wald F = 3.3 (monthly) and 2.5 (spell level), both far below the Stock-Yogo rule-of-thumb of 10. Results should be treated as directionally suggestive rather than precisely identified causal estimates.
Data: SIPP 1996–2010 matched to SSA 831 records; n=32,208 job losers ages 20–57. TA program receipt: UI 16.15%, SNAP 16.56%, TANF 3.39%, TDI 0.53%. Average spell 6.86 months; 70.67% end in re-employment, 0.74% in DI application, 0.87% in SSI application.

Concepts Introduced or Extended

Entities Mentioned

Quotes

"The target population for UI overlaps with the one for DI, but not with SSI as a means-tested program. Conversely, SNAP and SSI are both means-tested programs, but because these people tend to be poor and have a weak labor force attachment, they often do not qualify for DI benefits."

"The negative effect of UI participation on applications for DI suggests that the substitution effect dominates the income effect, but the positive effect of SNAP participation on applications for SSI suggests the opposite."

My Take

The paper's central contribution — the target-population-overlap principle for predicting the sign of cross-program social insurance interactions — is a genuinely useful organizing framework for the fragmented U.S. social insurance landscape. The UI→DI finding corroborates Lindner (2011) with a more general multi-program design. The SNAP→SSI finding is novel and policy-relevant: if means-tested program receipt facilitates disability application, then SNAP expansion may have second-order effects on SSI rolls. However, weak instruments (F ≈ 3) seriously undermine confidence in the causal interpretation — especially since the SNAP→SSI result does not survive the spell-level analysis where F is even lower. The instrument set (SNAP biometric requirements, outreach spending per non-participant, noncitizen eligibility interactions) is creative but potentially correlated with local labor market conditions or population composition. The re-employment null is consistent with the selection story but the large standard errors preclude ruling out meaningful effects. As a Center for Retirement Research working paper (later published in Research in Labor Economics 39, 2014), the results may be preliminary. Most useful as a directional guide and comparative framework rather than as a source of precise effect estimates.