Coe et al 2011 — What Explains State Variation in SSDI Application Rates

Summary

State Social Security Disability Insurance (SSDI) application rates vary by a factor of three — from $0.49\%$ (Utah) to $1.65\%$ (Mississippi) — and this paper asks how much of that variation is explained by health, demographics, employment, and state policy. Using $862$ state-year observations from 1993–2009 (Behavioral Risk Factor Surveillance System (BRFSS) + March Current Population Survey (CPS) + Bureau of Labor Statistics Local Area Unemployment Statistics (BLS LAUS) + Social Security Administration (SSA) admin data), the authors find that health, demographics, and employment account for more than $70\%$ of cross-state variation, while temporary disability insurance (TDI) mandates, strict health insurance market regulation, and Republican governors each independently reduce SSDI applications at the within-state level.

Key Claims

• Mean state SSDI application rate is $0.83\%$; range from $0.49\%$ (Utah) to $1.65\%$ (Mississippi); roughly equal split between SSDI-only and SSDI–Supplemental Security Income (SSI) concurrent applicants
• Health status, demographics, and employment variables explain $>70\%$ of cross-state variation (ordinary least squares (OLS) $R^2 = 0.796$)
• Labor force participation rate is more predictive of cross-state SSDI application variation than the unemployment rate; within-state variation depends on both
• TDI mandates: states with mandatory temporary disability insurance programs have $0.117$–$0.126$ percentage points (pp) lower SSDI application rates ($p < 0.01$), driven almost entirely by SSDI-SSI concurrent applicants — income-poor workers substituting TDI coverage for SSDI
• Strict health insurance (HI) regulation: states with guaranteed issue + community rating in the non-group market have $0.054$ pp lower within-state SSDI applications ($p < 0.01$) — consistent with the employment lock mechanism: accessible non-group insurance reduces the value of SSDI as a path to Medicare
• Republican governor: associated with $0.021$ pp lower within-state applications ($p < 0.05$), concentrated in SSDI-SSI concurrent applicants; plausibly via welfare program administration and stigma norms
• Unemployment Insurance (UI) benefit duration: longer unemployment insurance duration associated with fewer SSDI-only applications, consistent with UI providing a bridge that delays or averts SSDI filing
• SSDI-only and SSDI-SSI concurrent populations are structurally distinct: poverty and bad health predict concurrent; white race and smoking predict SSDI-only; state policy and political variables affect them differently
• Residual unexplained variation: $<20\%$ between-state, $<8\%$ within-state after adding state fixed effects

Concepts Introduced or Extended

• Geographic Variation in Disability Insurance
• Employment Lock
• DI Application Costs and Take-Up

Entities Mentioned

• Norma B. Coe
• Kelly Haverstick
• Alicia H. Munnell
• Anthony Webb
• Disability Insurance Program
• Supplemental Security Income

Quotes

"States with higher shares of people in poor or fair health, higher poverty rates, and higher unemployment rates tend to have higher SSDI application rates."

"Strict regulation of the non-group insurance market is associated with lower SSDI application rates within states over time."

My Take

A useful supply-side complement to demand-side DI research. The TDI and HI-regulation findings are the most novel — they identify state-level institutional levers that affect SSDI applications independently of underlying disability rates. The HI regulation finding pre-dates the Affordable Care Act (ACA) and provides a natural forecast: ACA's insurance market reforms and Medicaid expansion should reduce SSDI applications, a prediction subsequently examined in the literature. The political variable (Republican governor) is interesting but the mechanism is underspecified; the paper acknowledges this is correlation, not causal. The SSDI-only vs. concurrent split is a valuable decomposition that anticipates later work on heterogeneous DI populations.