Summary
Using a unique state-level panel (2001–2015) matched to Social Security Administration (SSA) vocational award data, Michaud, Nelson, and Wiczer decompose the 3–4× rise in vocational-stage Social Security Disability Insurance (SSDI) awards since 1985 into demographic composition effects (age, education, occupation) versus behavioral and implementation changes. Their central finding: mechanical demographic change explains at most 16.5% of the rise in vocational awards. Population aging (more workers in their 50s) would predict higher awards, but rising educational attainment more than offsets this. The residual — the majority of the trend — must reflect changes in how individuals within demographic cells behave (application propensity) or how SSA rules are implemented.
Key Claims
- Vocational-stage awards and denials have grown threefold since 1985; the award rate conditional on reaching the vocational stage roughly doubled from 1985 to 2000.
- Demographic composition (age + education + occupation) explains ≤16.5% of the rise in vocational awards and ≤18.5% of total determinations — aging's contribution is more than cancelled by rising education.
- The 55–59 age group, not 60–64, is the pivotal demographic driving applications and vocational awards; despite the grid being most generous to 60–64-year-olds, 55–59-year-olds have the largest total application and vocational acceptance effect.
- Workers aged 55–59 without a high school (HS) degree in manufacturing/production occupations (Standard Occupational Classification [SOC] 10–16) have the strongest relationship with vocational awards — the precise demographic of displaced manufacturing workers.
- A puzzle: rising educational attainment should lower the vocational award rate (more education → more job options → harder to qualify), yet the award rate rose sharply 1985–2000. This cannot be explained by the demographics the paper considers.
- Cross-state: variation in total applications explains 61% of state-level variation in vocational awards; the conditional acceptance rate at the vocational stage explains 35%. High-application states (South, Rust Belt) screen harder at earlier steps, passing a more marginal pool to the vocational stage.
- States with the highest application rates are negatively correlated with overall acceptance rates but positively correlated with vocational-stage acceptance rates — suggesting high-application states do not simply have laxer standards overall.
Concepts Introduced or Extended
- Vocational Grid — empirical test of whether the grid functions as designed across demographics; first paper to decompose vocational award trends using state-level data
- DI Growth Decomposition — parallel demographic decomposition restricted to the vocational stage; complements Liebman (2015) and Pattison/Waldron (2013) by adding education and occupation to the age-only framework
- Conditional DI Applicants — 55–59-year-old manufacturing workers without HS degrees are the demographic fingerprint of conditional applicants at the vocational stage; cross-state application-rate correlation corroborates the mechanism
Entities Mentioned
Quotes
"Our main finding is that mechanical changes in the composition of age, education, and occupation demographics of the workforce have nearly zero impact on SSDI trends."
"We find secular increases in educational attainment should work heavily against the aging of the population... It is further puzzling that the award rate at the vocational stage has doubled since the 1980s."
"Workers in their 50s in service or production sectors drove the incidence of awards with vocational considerations."
My Take
The paper is a useful but limited empirical contribution. Its main value is ruling out a simple demographic story for vocational award growth and formalizing the puzzle of rising awards despite rising education. The cross-state decomposition is credible and the demographic-nullity finding is robust. The main limitations are significant: state-level data only (not individual), regression estimated on 2010–2015 (heavily recessionary), no causal identification, and no exploration of health trends. The paper cannot say whether the residual trend reflects genuine disability insurance (DI) leniency drift, secular health deterioration among the 55–59 manufacturing demographic, or both. It frames the questions more than answers them, which the authors honestly acknowledge.