Von Wachter, Song, and Manchester (2011) provide the most comprehensive longitudinal analysis of employment and earnings for allowed and rejected Disability Insurance (DI) applicants using administrative data. Linking Social Security Administration (SSA) application records (1981–1999) to 29 years of W-2 earnings data (1978–2006), they document that rejected applicants earn far less than comparable non-applicants — around $10,000/year median vs. $35,000 for matched workers who never applied — while still far above Disability Determination Services (DDS)-allowed beneficiaries (~$3,000). The paper also updates the Bound (1989) upper-bound comparison methodology, showing that the rejected pool has shifted toward younger, lower-mortality applicants over time, raising the upper bound on the DI work-disincentive across cohorts.
Employment and earnings by applicant group (Table 1, 1997 cohort, men):
| Group | Age 45–64 any earnings | Age 45–64 median earnings | Age 30–44 any earnings | Age 30–44 median earnings | 2yr mortality (45–64) |
|---|---|---|---|---|---|
| Non-applicants | 82.2% | ~$35,000 | 87.7% | ~$32,000 | 2.6% |
| Rejected | 52.6% | ~$10,000 | 69.6% | ~$8,000 | 3.4% (4yr: 6.7%) |
| Hearings-allowed | 24.5% | ~$5,000 | 36.9% | ~$5,000 | 3.0% |
| DDS-allowed | 17.9% | ~$3,000 | 26.1% | ~$2,000 | 23.1% |
The rejected group sits far closer to allowed beneficiaries than to non-applicants in earnings, but far above allowed beneficiaries in mortality — they are sick, but not as immediately life-threateningly ill as DDS-allowed applicants.
Impairment heterogeneity — rejected applicants, ages 45–64 (Table 2, any earnings 2yr post-application):
Pre-application earnings dip:
Present discounted value (PDV) calculations:
Cohort trends (1982 vs. 1997): The 1997 cohort's rejected applicants have higher any-earnings rates than the 1982 cohort (e.g., older rejected: 52.6% vs. 40.4% for the 1982 cohort), reflecting compositional shift toward younger, lower-mortality, more marginally-impaired applicants. This raises the Bound upper bound on the work-disincentive over time. The hearings-allowed group (rejected at DDS, allowed at Administrative Law Judge (ALJ)) has higher employment than DDS-allowed, consistent with Maestas, Mullen, and Strand (MMS, 2013)'s finding that the disincentive is largest at the low-severity end of the margin.
Data:
"The employment and earnings of rejected applicants is much lower than that of non-applicants with similar characteristics, suggesting that many rejected applicants face substantial barriers to work even without receiving DI benefits."
The key contribution is the non-applicant comparison. Prior work (Bound 1989, Chen and van der Klaauw 2008) used rejected applicants as the counterfactual for allowed beneficiaries. Von Wachter et al. show that even the rejected pool is severely impaired relative to a matched non-applicant group: $10,000 vs. $35,000 median earnings, 52.6% vs. 82.2% any-earnings. This directly answers the "DI is not stringent" critique.
The PDV calculation is sobering: even if every rejected applicant returned to work, aggregate earnings recovery would cover only one-third of DI's fiscal liability. Aggressive denial policy is not the fiscal solution it is often assumed to be.
The pre-application dip finding is methodologically important for the conditional applicant literature: it provides a clean observable signature of the inframarginal (acute onset) vs. conditional (chronic, employment-contingent) distinction without requiring structural modeling.
The hearings-level finding matters for reading French and Song (2014): that paper's complier population (ALJ appellants) has higher employment potential than DDS-allowed applicants, which is consistent with their somewhat smaller measured disincentive compared to MMS (2013) — though both ultimately converge at ~26–28 percentage points (pp).