Deshpande and Li 2019 — Who Is Screened Out Application Costs and the Targeting of Disability Programs

Summary

Deshpande and Li exploit 118 Social Security Administration (SSA) field office closings between 2000 and 2014 as quasi-random variation in the cost of applying for disability insurance (DI)/Supplemental Security Income (SSI). Using a difference-in-differences design with future-closing zip codes as controls, they find that office closings reduce DI/SSI applications by 10% and recipients by 16%. The fact that the recipient decline exceeds the application decline is the central result: the closings screen out applicants who would have been approved, making the program less well-targeted toward severely disabled individuals. This directly refutes the Nichols-Zeckhauser (1982) hypothesis that application hassles improve targeting by discouraging fraudulent or low-severity claimants. Medium-severity applicants (those with more residual work capacity) are most discouraged — not low-severity ones. The net social welfare cost of all 118 closings is estimated at approximately $1.2 billion.

Key Claims

• Applications −10%, recipients −16%. The gap between these two declines is statistically significant at the 1% level and constitutes direct evidence that targeting worsens: the closings differentially screen out would-be approved applicants relative to denied ones.
• Who is screened out. Medium-severity applicants are most deterred (−33.8% in medium-severity group). High school dropouts are the most deterred educational group (−14.2%). These are groups with more residual work capacity and lower education — not the fraudulent or low-severity applicants that the Nichols-Zeckhauser hypothesis predicts would be deterred.
• Refutation of Nichols-Zeckhauser (1982). The classic argument that application costs improve targeting by functioning as self-selection devices is wrong empirically: the deterred applicants are approved-quality cases, not cases that would have been denied. Hassles impose dead-weight losses without the targeting benefit.
• Channels: congestion dominates distance. The mechanism of the effect is not travel distance (which contributes only 4% of the decline) but rather congestion at neighboring offices when cases are diverted (54%) and the administrative friction of switching to a new office (42%). Closing an office pushes applicants to already-busy neighbors; the resulting processing slowdowns and wait times deter application.
• Actual closings were poorly targeted. Offices that SSA actually chose to close had net social costs more than double the lowest-cost counterfactual closings. Had SSA applied a cost-minimizing algorithm — targeting offices with fewer potential applicants and lower nearby congestion — the social cost of achieving the same number of closings would have been substantially lower.
• Net social welfare cost ≈ $1.2 billion across all 118 closings over the study period. This accounting includes foregone benefit income to deterred recipients minus avoided administrative costs.
• Application cost elasticity. A 10% increase in application cost reduces applications by roughly 10% (elasticity ≈ −1). This is substantially larger than the elasticity of take-up with respect to benefit generosity found in comparable literatures, suggesting that process costs are as important as program parameters in determining who is served.

Concepts Introduced or Extended

• DI Application Costs and Take-Up
• DI Denied Population — targeting worsens because approved-quality applicants are deterred; Nichols-Zeckhauser refutation
• SSA Sequential Determination Process — field office role in Step 1 screening and initial application intake; office congestion as a barrier at the front of the pipeline
• Causal Effects of DI Receipt — adds an aggregate-level instrumental-variable (IV) design using office closings to identify the effect of access on program participation

Entities Mentioned

• Manasi Deshpande
• Yue Li
• Social Security Administration
• Disability Insurance Program

Quotes

"We find that field office closings reduce the number of DI and SSI recipients by 16 percent, which is larger than the 10 percent reduction in applications. The larger reduction in recipients indicates that the composition of applicants worsens following a closing — those who are deterred from applying are disproportionately likely to be approved."

"Our results refute the hypothesis that application costs improve the targeting of disability benefits by screening out low-type applicants, as predicted by Nichols and Zeckhauser (1982)."

"The primary channel through which closings reduce applications is congestion at neighboring offices rather than increased travel costs."

My Take

The identification strategy is clean and the result is striking: field office closings worsen targeting by deterring approved-quality applicants. The congestion-over-distance finding is counterintuitive but mechanically plausible — closing an office increases processing loads at neighbors, lengthening wait times and increasing friction for potential applicants even in nearby zip codes that are geographically close. The refutation of Nichols-Zeckhauser is important for policy: it suggests that administrative simplification (online applications, longer office hours, more caseworkers) is not a loosening of the screen — it is a correction of an unintentional screen that falls on exactly the wrong population. The welfare calculation depends on what weight one places on denied applicants' foregone income vs. administrative savings, but even under conservative assumptions the closings fail a cost-benefit test. The finding that SSA closed offices with above-average social costs rather than below-average ones suggests the organizational decision process was not optimizing for program targeting or welfare.