The DI (Disability Insurance) Denied Population (or "Social Security Denied" group) consists of individuals who have applied for Social Security disability benefits at some point in their lives but have never received them. As of the 2014 Survey of Income and Program Participation (SIPP), this population is estimated at approximately 12.4 million working-age adults (ages 18–66) — roughly equal in size to the 11.3 million who were approved. This population is largely invisible to federal policy: approved applicants receive cash benefits, Medicare, and Ticket to Work support; denied applicants receive no targeted federal assistance.
From the 2014 SIPP Supplement (ever-applied, ages 18–66):
About 3.2% of the denied group were pending an initial decision at the time of the SIPP; the remainder had received a definitive denial at some point in the process.
| Measure | SS Denied | SS Beneficiaries | General Population |
|---|---|---|---|
| Fair or poor health | 52.3% | 62.5% | 6.9% |
| Hospitalized (2013) | 21.1% | 27.3% | 6.3% |
| >5 medical visits (2013) | 41.9% | 53.7% | 15.5% |
| Health insurance | 77.1% | 89.7% | 81.1% |
Health limitations (selected, ages 18–66):
| Limitation | SS Denied | SS Beneficiaries | General Population |
|---|---|---|---|
| Difficulty standing 1 hr | 51.6% | 65.7% | 4.7% |
| Difficulty stooping/kneeling | 54.4% | 69.3% | 7.5% |
| Frequent depression/anxiety | 44.2% | 46.1% | 5.9% |
| Difficulty concentrating | 33.4% | 38.0% | 2.8% |
The denied group is severely impaired on every measure. The gap between denied and approved is modest; the gap between either group and the general population is vast.
| Measure | SS Denied | SS Beneficiaries | SSI Beneficiaries | General Population |
|---|---|---|---|---|
| In poverty | 37.7% | 25.3% | 40.9% | 13.4% |
| Material hardship | 42.6% | 38.0% | 44.2% | 18.1% |
| Earn >$1,870/month | 13.5% | 6.6% | 5.9% | 52.3% |
| Earn >Substantial Gainful Activity (SGA) ($1,040) | 20.1% | 8.6% | 8.7% | 63.7% |
The denied group's poverty and hardship rates rival those of SSI recipients (a means-tested program), despite having no targeted federal cash support.
Cross-study note on poverty figures: Stegman Bailey and Hemmeter (2014), using 2008 SIPP matched to SSA administrative records for December 2010, find a DI poverty rate of 19.5% — lower than Weaver's 25.3% for "SS beneficiaries." The gap reflects different population definitions: Weaver defines "SS beneficiaries" as 2014 SIPP respondents who ever received Social Security (a broad category including Old-Age and Survivors Insurance [OASI] retirees); Stegman Bailey counts only active DI beneficiaries on the rolls in December 2010. The period difference (2010 vs. 2014) and the exclusion of OASI retirees from the Stegman Bailey DI-specific count explain most of the gap. The denial group's 37.7% poverty rate sitting between the DI active-beneficiary rate (19.5%, Stegman Bailey) and the SSI rate (42.9%, Stegman Bailey) reflects the denied population's structural position: impaired like beneficiaries, but without any program income to offset poverty risk.
Compared with the general population, the denied group is:
These characteristics mirror the approved beneficiary profile closely, with the denied group slightly more likely to be Black and slightly less likely to be married.
Why Black applicants are over-represented in the denied pool. The higher share of Black individuals among the denied population partly reflects that Black workers apply for DI at disproportionately high rates relative to their self-reported disability prevalence (Mitchell and Thompson 2025: higher application rates vs. higher disability prevalence). This over-application relative to disability burden means a higher share of Black applicants falls below the medical eligibility threshold, generating a higher denial rate even under race-neutral adjudication. The large raw Black–White allowance gap ( percentage points [pp] for males, pp for females) collapses to statistical insignificance after adjusting for body system code, age at application, earnings, education, marital status, and region — suggesting the denied pool's racial composition is driven more by applicant selection and compositional differences than by disparate adjudication at the Disability Determination Services (DDS) level. See SSA Sequential Determination Process for the full analysis.
Variation within the denied group matters for policy targeting:
Approved applicants receive: cash benefits, Medicare after 2 years (or Medicaid for SSI), and Ticket to Work employment support. Denied applicants receive: nothing federally. They may qualify for state Vocational Rehabilitation, Supplemental Nutrition Assistance Program (SNAP), or Temporary Assistance for Needy Families (TANF), but no program specifically targets their work, health, or poverty outcomes.
The SSA BOND demonstration (Benefit Offset National Demonstration) tested whether allowing DI beneficiaries to retain more benefits as they earn would encourage work. The evaluation found no effect on earnings, but higher benefit costs. The most plausible explanation: underlying health is the binding constraint, not financial incentives. This holds equally for the denied population.
Demonstrations that combine employment support with health intervention show more promise:
The denied population's health profile directly informs the debate over whether the disability standard is too loose or too strict. If the standard were too loose, denied applicants would look significantly healthier than approved ones. Instead, they look nearly identical on health measures. This is the strongest available descriptive evidence that even failed applicants face genuine and severe impairments — the standard is functioning as a medical screen, not a perfunctory gateway.
Von Wachter, Song, and Manchester (2011) link SSA application records (1981–1999) to 29 years of W-2 earnings data (1978–2006) to compare employment and earnings across four groups: non-applicants matched on demographics, rejected applicants, DDS-allowed beneficiaries, and hearings-allowed beneficiaries (those rejected at DDS but allowed at Administrative Law Judge [ALJ]).
Table 1 — Employment and earnings 2 years post-application (1997 cohort, men):
| Group | Age 45–64 any earnings | Age 45–64 median earnings | Age 30–44 any earnings | Age 30–44 median earnings |
|---|---|---|---|---|
| Non-applicants | 82.2% | ~32,000 | ||
| Rejected | 52.6% | ~8,000 | ||
| Hearings-allowed | 24.5% | ~5,000 | ||
| DDS-allowed | 17.9% | ~2,000 |
The rejected group sits far closer to allowed beneficiaries than to non-applicants: roughly half have any earnings at all (older group), and their median earnings (approximately $10,000) are less than one-third of the $35,000 for matched non-applicants. The pattern holds for younger men (ages 30–44): 69.6% any-earnings but median only $8,000 vs. $32,000 for non-applicants. A $10,000 median cannot support living expenses, let alone the elevated medical costs this population faces.
The hearings-level group: Applicants rejected at DDS but eventually allowed at ALJ hearings fall between the rejected and DDS-allowed groups — any-earnings and approximately $5,000 median for ages 45–64. This is the complier population studied by French and Song (2014), and their higher employment potential is consistent with MMS (2013)'s finding that the work-disincentive is highest among less-severely-impaired marginal entrants.
Impairment heterogeneity (rejected applicants, ages 45–64, any earnings 2yr post-application): Injuries (51.7%), musculoskeletal (43.9%), and mental/nervous (44.1%) have the highest any-earnings rates among rejected applicants. Respiratory (30.2%) and neoplasms (37.2%) have the lowest. This hierarchy tracks plausible residual work capacity by condition type.
Cohort trend: the 1997 cohort's rejected applicants have higher any-earnings rates than the 1982 cohort (52.6% vs. 40.4% for ages 45–64), reflecting compositional shift toward more marginally-impaired applicants over time. This also raises the Bound (1989) upper bound on the estimated DI work-disincentive across cohorts.
Implications for program stringency: This comparison directly addresses the "DI is not stringent" critique. Even the workers the program turns away are overwhelmingly unable to support themselves. The rejected population fails economically — just on a less extreme trajectory than allowed beneficiaries.
Present Discounted Value (PDV) calculations: If all rejected applicants returned to work, the aggregate earnings recovery would amount to roughly one-third of DI's fiscal liability (one-fifth once Medicare is included). Aggressive denial policy generates far smaller fiscal savings than the recipient count implies, because each rejected applicant has severely limited earnings capacity.
Bound, Caswell, and Waidmann (2013) use Health and Retirement Study (HRS) wave 4 (ages 50–64) linked to SSA administrative records to compare the economic well-being of DI/SSI beneficiaries with mental illness to two control groups constructed via propensity score reweighting: (a) never-applied non-beneficiaries with similar health profiles, and (b) rejected/uncertain applicants with similar profiles.
Key findings for the denied mental illness group:
Implications for targeting assessment: The economic equivalence of approved vs. rejected mental illness applicants is consistent with the program successfully screening on unobservable dimensions of need. It refutes the "marginal applicant" narrative — the claim that mental illness applicants are relatively healthy and strategic. For the policy debate over tightening DI eligibility for mental illness, the finding that rejected applicants cannot compensate with earned income implies that stricter standards would eliminate benefits without generating employment.
This complements the von Wachter, Song, and Manchester (2011) finding that even formally rejected applicants have severely impaired earnings capacity, and extends it specifically to the mental illness subgroup using an HRS-based observational design rather than administrative earnings records.
See Bound Caswell and Waidmann 2013 — Insurance Value of Disability Insurance for Individuals with Mental Health Impairments, John Bound, Kyle Caswell, and Timothy Waidmann.
Using the universe of all DI applications 1991–2008 (SSA Disability Research File), Lindner et al. document the labor market trajectory of applicants who are eventually denied:
| Measure | At application year | 2 years post-application | 5 years post-application |
|---|---|---|---|
| Employment rate | ~58.3% | ~55.3% | stabilizes |
| Mean earnings | ~8,137 | rises gradually |
These figures mask a crucial cyclical pattern: post-application earnings and employment of denied applicants are negatively related to unemployment at the time of application. A 1 pp rise in unemployment is associated with approximately \approx 1\approx 0.5$ pp by year 5). This counterintuitive result occurs because the labor market suppresses outcomes for the entire low-skill workforce during downturns — the compositional improvement from higher-work-capacity conditional applicants is more than offset by depressed labor demand.
The gap between initial acceptance (40.6%), final determination (57.5%), and including re-applications within 5 years (65.5%) implies that many denied applicants cycle through multiple rounds of appeal — a process that may itself erode employment attachment via the "decay effect" ( pp employment loss per additional month of processing time). See Conditional DI Applicants.
Maestas, Mullen, and Strand (2013) provide causal evidence about what the denial-margin population would have done without Social Security Disability Insurance (SSDI). Using examiner allowance rates as an instrumental variable (IV), they estimate that of applicants are on the actual margin of program entry — these are effectively the applicants whose outcome depends on which examiner they were assigned to. Employment of this marginal group would have been 28 pp higher without benefits (2 years post-decision).
The profile of marginal entrants from the MMS (2013) study differs strikingly from the profile of the denied population documented by Weaver (2020): marginal entrants are 43% more likely to have a mental disorder (not musculoskeletal, the most common impairment in the denied population), and 43% more likely to be very young (<29) or near retirement age (60–64). This suggests that the denial-margin cases — the contested borderline of the program — are not simply a random slice of the denied population, but rather a subset concentrated in diagnostic categories where functional impairment is harder to measure objectively.
The pp disincentive effect at the high-severity end of the marginal group and the pp effect at the low-severity end imply that the cases most likely to be incorrectly denied (those who would have been approved by a lenient examiner) are also the cases for whom SSDI most strongly substitutes for work — the program's disincentive is highest precisely where measurement error in severity assessment is highest.
Approved DI beneficiaries are known to have dramatically higher mortality than the general population (see DI Beneficiary Mortality). The denied population's health profile — only modestly better than the approved group — suggests they face substantially elevated mortality as well, though systematic mortality data on the denied population are not available. Combined with their very low income (see Income-Mortality Gradient), the denied group likely sits at the extreme low end of the income–life expectancy (LE) gradient, potentially with mortality rates approaching those of approved beneficiaries.
French and Song (2014), using the universe of 1.78 million ALJ hearings from 1990–1999, document the full appeals trajectory of denied applicants:
| Horizon after ALJ assignment | Status |
|---|---|
| 3 years | 35% of denied subsequently allowed; 40% still appealing/reapplying; 75% either allowed or still in process |
| 10 years | >60% of all ALJ-denied applicants ultimately allowed; 67% of all applicants allowed; 27% denied/process ended; 6% still in process |
The implication is stark: DI denial at the ALJ stage is rarely a permanent outcome. Most applicants cycle through reapplications and re-appeals until they eventually receive benefits. This pattern has two policy consequences: (1) denial functions primarily as a delay mechanism, not an exclusion mechanism, for the majority of denied applicants; (2) denied applicants suppress their earnings below SGA during the appeals process — holding down labor supply strategically — which means the measured work-disincentive at short horizons understates the full effect that ultimately materializes once benefits are received.
This finding complements the Lindner et al. (2017) estimate of 65.5% eventually allowed within 5 years of initial application, extending the window to 10 years and focusing specifically on the ALJ-denial stage rather than the initial determination.
A common assumption in the DI access literature is that the ~55% of more-severely impaired individuals who do not receive DI represent unmet need — the program failing to reach eligible people. Deshpande and Lockwood (2022) directly test this using Panel Study of Income Dynamics (PSID) consumption data merged with SSA administrative records, comparing four groups: less-severe recipients (L-DI), more-severe recipients (M-DI), more-severe nonrecipients (M-NDI), and less-severe nonrecipients (L-NDI).
M-NDI are better-off than both recipient groups:
The implication is that M-NDI self-select out of the applicant pool because they have adequate outside resources — household earnings, savings, or private insurance — rather than because they face barriers. Providing them DI would reduce welfare, not increase it. The ~55% non-application rate among more-severely impaired individuals is not evidence of access failure; it is evidence of successful self-selection.
How this fits with Deshpande and Li (2019). The field office closing result (closing an office deters approved-quality applicants) applies to individuals within the applicant pool who are prevented from completing their application. M-NDI are a different population: they choose not to apply, not because of administrative barriers, but because they have better outside options. The two papers target different mechanisms — deterrence vs. self-selection — and the policy implications differ. See DI Application Costs and Take-Up.
Within the denied population, Strand and Trenkamp (2015) isolate the subset most precisely analogous to the "own-occupation" impairment concept: claimants denied specifically at Step 5 of the sequential process who did not appeal, did not reapply, and had no prior claim — 37,110 cases out of 267,821 Step 5 denials in 2005 (). Because these claimants passed Step 4 (they cannot perform past work) but were denied at Step 5 (they retain capacity for some other work in the national economy), their Regulation Basis Code provides a clean administrative indicator of own-occupation impairment.
Post-denial labor market outcomes (pre-onset 2000 vs. post-decision 2008):
Heterogeneity by earnings decile: Earnings declines are regressive. The top decile fell from $87,123 to $45,374 median () with a pp employment drop. The bottom decile shows a smaller employment decline ( pp) and a slightly rising conditional earnings ratio (0.92), but the latter reflects selection bias: low-earning workers whose impairments caused large earnings declines would have fallen below SGA and qualified for DI, exiting the sample. The estimates understate losses for the full own-occupation population at the lower end.
Heterogeneity by diagnosis: Mental disorders produce the largest employment and earnings declines (anxiety disorders pp, affective mood disorders pp). Sensory and selected physical impairments have the smallest declines: blindness/low vision ( pp), carpal tunnel ( pp), chronic ischemic heart disease ( pp). Back disorders — the most common diagnosis — produce exactly average outcomes. This diagnosis ranking partially reverses what is observed among DI beneficiaries, where mental disorder beneficiaries have higher employment than physical-impairment beneficiaries.
Policy interpretation: This group documents the DI coverage gap — workers whose impairments impose real, lasting earnings losses but who do not meet the statutory "incapable of any work" standard. It also identifies the most promising targets for early vocational intervention: sensory and cardiovascular/neurological conditions, not back disorders (which are already the bulk of VR caseloads).
The 12.4 million "denied" applicants documented by Weaver (2020) represent individuals who completed an application. Deshpande and Li (2019) identify a prior-stage population: individuals who would have applied — and likely been approved — but were deterred by application costs before ever filing.
Using 118 SSA field office closings (2000–2014) as quasi-random variation in application costs, they find: applications fall 10%, but DI/SSI recipients fall 16%. The larger recipient decline is the central result — it proves that the closings did not merely block denied-quality applicants (which would leave the recipient rate unchanged) but specifically deterred applicants who would have been approved. Targeting worsens: the deterred population is drawn disproportionately from approved-quality cases.
The profile of deterred applicants mirrors the policy concern:
This is a direct empirical refutation of the Nichols-Zeckhauser (1982) targeting hypothesis, which predicted that application costs would improve targeting by screening out low-value, low-severity claimants. The data show the opposite: costs screen out the wrong people.
Implication for interpreting the denied population: The Weaver (2020) "denied" count is a lower bound on the population that tried to access the program and failed. The true gap between eligible-and-needing and eligible-and-receiving is larger still, because an unknown number of eligible individuals were deterred before applying. The Deshpande-Li elasticity (−10% applications per closure) suggests this pre-application attrition is not trivial. See DI Application Costs and Take-Up.
Among the denied population, Low and Pistaferri (2019) document a specific gender asymmetry: women with severe, permanent, work-preventing self-reported impairments are 21.5 pp more likely to be rejected than observationally equivalent men. This is a Type I error (false rejection), concentrated at the vocational assessment stage (Step 4/5 — residual functional capacity for other work), not at medical assessment. The critical evidence is the labor market outcome: rejected work-limited women are employed at only 2% in the 3 years after rejection (vs. 19% for rejected non-limited women). The Rejected × Disabled interaction is pp for women () but statistically insignificant for men (), confirming that SSA correctly assesses men's Residual Functional Capacity (RFC) but overestimates women's capacity to work.
This finding implies that the denied population contains a larger share of genuinely disabled women than men. The 12.4 million denied applicants documented by Weaver (2020) likely understates the welfare loss for women specifically: not only are they more likely to be denied, but denial does not lead to economic recovery. See DI Classification Errors for the full Type I/II framework.
Weaver (2020) and von Wachter et al. (2011) characterize the denied population after the fact. Contreary et al. (2017) add a pre-application dimension: approximately 47% of all DI applicants (ages 25–55) had no or intermittent employment in the 24 months before filing — the "Type 2" group. This subgroup is disproportionately represented in the denied pool because their allowance rates are 10–15 pp lower than applicants with recent consistent employment (Type 1). The Type 2 demographic profile — more female (53–57%), more Black (22–23%), lower household income ($46,178 vs. $62,986 for Type 1), higher Medicaid (27–31%) and SNAP (27–29%) enrollment — overlaps substantially with the Weaver (2020) profile of the denied population (older, less educated, more Black, lower earnings). The implication is that a large share of the denied pool was economically and occupationally marginal before applying, not merely because of the health condition that prompted application. Their denial rates are higher partly because they are less likely to have vocational factors working in their favor under Steps 4–5 of the determination process. See Conditional DI Applicants for the Type 1/Type 2 taxonomy and its policy implications for early-intervention program design.
The post-denial employment literature (von Wachter et al. 2011; French and Song 2014; Lindner et al. 2017) examines what denied applicants do after the process concludes. Khan (2018) identifies a distinct prior mechanism: the application decision itself — independent of receiving benefits — causally reduces denied applicants' employment.
Design and sample: HRS 1992–2012 (11 waves); 322 denied SSDI applicants aged 50–58; 347 controls (future applicants: non-applicants in their 50s who eventually filed at age 60+); 1,231 observations. IV: Full Retirement Age (FRA) variation across birth cohorts (1983 Social Security Amendments) interacted with state-level DDS allowance rates. First-stage F = 9.84; overidentification p = 0.57.
Result: Denied applicants aged 50–58 experience a −36 pp employment reduction two to three years post-application (IV/two-stage least squares [2SLS]; Ordinary Least Squares [OLS] upper bound −49 pp). Three channels: (i) voluntary labor force exit while planning reapplication; (ii) voluntary exit during appeal to strengthen the case; (iii) human capital deterioration from extended absence making re-entry hard once the process ends.
Distinction from adjacent mechanisms:
Policy implication: With 1.8 million denied applicants in 2013 and a rising denial rate (45% in 2000 → 72% in 2013), the aggregate employment welfare loss from the process itself is large and growing. Shortening the determination timeline would reduce this cost independently of any change in allowance rates. See Khan 2018 — Disability Insurance Application Decision.
Chen (2014) extends the post-denial picture from employment and earnings outcomes to program dependency, using SIPP 1990–2008 linked to SSA administrative records ( person-year observations, males only, ages 30–65 at filing). The design is a difference-in-differences (DiD) event study spanning to around the DI filing date, comparing rejected applicants to a healthy comparison group.
Main results:
Cohort heterogeneity: 2000s filers show markedly larger effects than 1990s filers (younger group: 13.8% vs. 4.0% social-support probability gap), consistent with a more severely impaired and lower-education applicant pool during the DI expansion era.
Interpretation: Denial does not reduce aggregate program dependency — it shifts it. This is direct evidence for the Program Spillovers / welfare-shifting hypothesis: tightening DI criteria displaces costs to other programs rather than eliminating them. The fiscal savings attributed to denial accrue to the Social Security trust fund while costs are absorbed by TANF, SNAP, UI, and state general assistance budgets.
Caveat: estimates are upper bounds. Because rejected applicants have worse unobserved health than the healthy comparison group, part of the social-support increase reflects health deterioration common to both rejected and approved applicants, not the denial decision alone. A cleaner IV design would be needed for precise causal attribution.
See Chen 2014 — Rejection from the Disability Insurance Program and Dependency on Social Support.