Chetty 2005 — Why Do Unemployment Benefits Raise Unemployment Durations The Role of Borrowing Constraints and Income Effects

Summary

A foundational paper in the optimal social insurance literature. Chetty challenges the standard interpretation of the Unemployment Insurance (UI)-duration effect — that longer unemployment spells when benefits are higher reflect a distortionary substitution effect (moral hazard) — by showing that borrowing constraints imply a non-distortionary income effect channel. For constrained households, UI raises cash on hand and reduces pressure to accept any available job, but this income-smoothing does not create a wedge between private and social marginal costs. The implication is that the efficiency cost of UI (and social insurance generally) has been systematically overestimated.

Key Claims

• The canonical finding: A $10\%$ increase in UI benefits raises average unemployment durations by $4$–$8\%$ in the U.S. (Moffitt 1985, Meyer 1990).
• Standard interpretation: UI distorts the relative price of leisure vs. consumption → substitution effect → deadweight loss.
• Chetty's revision: For borrowing-constrained households, UI relaxes the liquidity constraint dollar-for-dollar, raising consumption and reservation wages as an income effect — not a substitution effect. Income effects from lump-sum transfers are non-distortionary by definition.
• Heterogeneity evidence: UI-duration effects are significantly larger for constrained households (low liquid assets, no working spouse, high mortgage burden) than for unconstrained households. If the duration effect were purely a substitution effect, liquidity status should not matter.
• Severance pay evidence: Lump-sum severance payments at job loss — which change wealth but not the relative price of leisure — significantly extend durations for constrained households, ruling out a pure substitution interpretation.
• Policy implication: Optimal UI benefit formulas that set benefits based on observed duration elasticities (Baily 1978) overstate efficiency costs when the duration response partly reflects non-distortionary income effects. True deadweight loss is smaller; optimal benefit levels are higher than previously calculated.
• Generalization: The income-effect argument applies to any insurance program where beneficiaries face borrowing constraints — including disability insurance (DI), health insurance, and workers' compensation.

Concepts Introduced or Extended

• Optimal Social Insurance
• Causal Effects of DI Receipt

Entities Mentioned

• Raj Chetty

Quotes

"Roughly stated, the standard view has been that people take longer to find a job when receiving high UI benefits because it pays less to go back to work. The evidence described here suggests instead that unemployment durations rise mainly because households have more cash on hand while unemployed, and are therefore less pressured to find work quickly."

"In an environment with borrowing constraints, many of these responses [to DI, health insurance, workers compensation, Social Security] could be partially due to a non-distortionary income effect. Decomposing behavioral responses to other programs into income and substitution effects would be a useful step in obtaining a more precise understanding of the efficiency costs and optimal design of insurance programs."

My Take

This paper launched the "sufficient statistics" approach to social insurance design. The key methodological insight — use heterogeneity by liquidity status to identify income vs. substitution channels — was directly imported into the DI literature by Autor and Duggan (2007), who show that Veterans' Disability Compensation (VDC) benefits (non-work-contingent income transfers) reduce labor supply, confirming that DI's work-disincentive effect is primarily an income effect. Gelber, Moore, and Strand (GMS, 2016) subsequently identify this via regression kink design (RKD) ($-\$0.20$ earnings per $\$1$ benefit). Taken together, Chetty 2005 + Autor-Duggan 2007 + GMS 2016 imply that substitution-targeted DI reforms (Ticket to Work, substantial gainful activity (SGA), continuing disability reviews (CDRs)) are addressing the wrong margin, and that the true efficiency cost of DI is substantially lower than naive duration-elasticity estimates suggest.