The sequential disability determination process is the structured, step-by-step framework the Social Security Administration (SSA) uses to evaluate whether an applicant is entitled to disability insurance (DI) or Supplemental Security Income (SSI) benefits. Field offices handle Step 1 (financial screens); state Disability Determination Service (DDS) agencies handle Steps 2–5 (medical and medical-vocational evaluations). The process is designed as an explicit screening strategy: handle the clear majority efficiently at early steps, reserve individualized evaluation for the residual.
Field offices deny applicants who are currently engaging in Substantial Gainful Activity (see Substantial Gainful Activity). This is an earnings-based test, not a medical test. Applicants earning above the SGA threshold (/month in 2012 for nonblind) are denied without any medical evaluation. Field offices also verify financial eligibility: insured status for DI (quarters of Federal Insurance Contributions Act (FICA) contributions), and income/resources for SSI. Note: Step 1 denials are mostly not recorded in the SSA-831 DDS data — they happen before the claim reaches the DDS.
DDS denies applicants whose impairment is "not severe" — no more than a minimal effect on the ability to work. Also applied here: the duration test — the impairment must have lasted, or be expected to last, at least 12 months, or be expected to result in death. If both conditions are met (severe + duration), the claim proceeds.
DDS allows applicants whose impairment meets or medically equals a Listing (see Listing of Impairments). This is a pure medical determination — no vocational factors are considered. Applicants allowed at Step 3 receive benefits immediately. Those whose impairments are severe but do not meet the Listings proceed to Steps 4–5.
The overall standard — inability to perform any work in the national economy — produces a threshold widely regarded as among the world's strictest. The Organisation for Economic Co-operation and Development (OECD) (2010) classifies the United States (along with Canada, Japan, and South Korea) as having "the most stringent eligibility criteria for a full disability benefit, including the most rigid reference to all jobs available in the labour market" among its 34 member countries. Among the current beneficiary population, have multiple disabling conditions and 1 in 5 male beneficiaries die within 5 years of entitlement — evidence that the population reaching benefit receipt is genuinely severely impaired (O'Leary et al. 2015; Zayatz 2011).
DDS assesses the applicant's Residual Functional Capacity (RFC) (see Residual Functional Capacity) and compares it to the skill and task requirements of jobs held in the past 15 years. If the applicant can perform past relevant work, the claim is denied.
DDS considers RFC alongside vocational factors — age, education, and work experience — to determine whether the applicant can perform any work in the national economy, not just past jobs. The Vocational Grid (medical-vocational guidelines) operationalizes this determination using age thresholds at 50 and 55. Applicants unable to perform any work are allowed; those able to perform other work are denied. An "expedited vocational assessment" procedure (piloted 1999, extended nationwide 2012) allows DDS to proceed directly to Step 5 when past work history is unclear.
Grid pathways vs. grid allowances (Strand and Messel 2019): The primary empirical effect of the age-55 vocational grid cutoff is on the pathway to allowance — not on the final allowance rate. Before age 55, the majority of grid-based allowances occur at the Administrative Law Judge (ALJ) hearing level (applicants must appeal an initial denial to receive vocational credit). After age 55, most grid allowances occur at the initial DDS level. The cutoff makes it easier to receive an initial award, but 71.3% of denied age 55–56 applicants appeal or reapply anyway — only slightly more than the 68.5% for denied ages 52–53. This means proposals to raise the grid's age thresholds are more likely to produce longer pathways (more appeals, more delays) than to increase sustained labor force participation. See Vocational Grid and Strand and Messel 2019 — The Ability of Older Workers with Impairments to Adapt to New Jobs.
Who gets denied at Step 5 — and what happens to them (Strand and Trenkamp 2015): Claimants denied at Step 5 constitute the administrative indicator of "own-occupation" impairment: their conditions are severe enough to prevent prior-occupation work but allow some work in the national economy. Strand and Trenkamp (2015) study the Step 5 denials from 2005 who did not appeal and did not reapply ( of all Step 5 denials). Post-denial outcomes: employment fell ( percentage points (pp)); median conditional earnings fell to cents on the dollar. worked at some point over the following six years — comparable to vocational rehabilitation recipients. Mental disorders produce the largest post-denial employment declines; sensory and cardiovascular/neurological conditions produce the smallest. An "expedited Step 5" procedure (piloted in prototype states 1999; extended nationwide August 2012) allows skipping Step 4 when past work history is unclear — Strand and Trenkamp exclude these cases because the RBC is uninformative in that context. See DI Denied Population for the full coverage-gap analysis.
Applicants denied at the DDS level (Steps 2–5) may appeal through: (1) DDS reconsideration, (2) ALJ, (3) Appeals Council, (4) federal court. The RBC covers DDS decisions only; ALJ and higher-level decisions are in the Case Processing and Management System.
Empirical stage statistics (Benítez-Silva et al. 1999, Health and Retirement Study (HRS) waves 1–3, ):
| Stage | Applicant action rate | Acceptance rate | Mean duration |
|---|---|---|---|
| DDS initial determination | — | months | |
| DDS Reconsideration | of denied request | months | |
| ALJ hearing | of Reconsideration-denied request | months | |
| Appeals Board | of ALJ-disposed request | months | |
| Ultimate award rate |
The appeal option thus raises the effective award rate by pp, but at substantial delay cost. First-stage awardees wait a median of months between application and receipt of benefits; appeal-stage awardees wait a median of – months. Despite the high appeal success rate, of rejected applicants do not appeal — and of these non-appealing rejectees self-report a health limitation preventing work (HLIMPW=1), meaning the delay cost deters genuinely disabled individuals. Onset-to-application median delay is months; of individuals with qualifying health limitations never apply at all ( of whom are legally ineligible). See DI Application Costs and Take-Up for the delay-deterrence implications.
Children follow a modified process with no Steps 4–5:
| Step | Outcome | N | % |
|---|---|---|---|
| Step 1 | Denied (SGA) | ||
| Step 2 | Denied (not severe) | ||
| Step 3 | Allowed (meets Listings) | ||
| Step 3 | Allowed (equals Listings) | ||
| Step 4 | Denied (past work) | ||
| Step 5 | Allowed (medical-vocational) | ||
| Step 5 | Denied (other work) | ||
| Other | Technical/procedural |
Step 5 is the modal outcome of DDS adjudication, accounting for of all decisions. Medical allowances (Step 3) produce fewer allowances than medical-vocational allowances (Step 5).
Not all steps respond equally to economic conditions. Lindner, Burdick, and Meseguer (2017), using the universe of all DI applications –, find that a pp rise in unemployment:
This step-specific pattern is the clearest empirical fingerprint of the Conditional DI Applicants mechanism: recessions do not increase the rate of sudden severe disability (Step 3); they activate the stock of people who are health-impaired but prefer work when employed (Steps 2 and 4). See Conditional DI Applicants for the full mechanism and policy implications.
Meseguer (2013) classifies the full adjudicative record into four mutually exclusive outcomes using the Disability Research File ( applications, –):
| Outcome | Share |
|---|---|
| Initial allowance (DDS allow) | |
| Initial denial not appealed | |
| Final allowance (ALJ/higher level) | |
| Final denial (all levels exhausted) |
Combined: of applicants are ultimately allowed; are ultimately denied.
This four-category breakdown is more informative than the binary allow/deny: of applicants abandon their claim after an initial denial (not captured by the binary), and obtain benefits only after appeal — these are the populations in the "gray zone" where individual and diagnosis-level heterogeneity is highest.
Using Bayesian hierarchical multinomial logit models with applicants clustered by state () and primary diagnosis ( codes), Meseguer decomposes variation in initial allowances:
Ranking of diagnoses is preserved from initial to final: Pearson correlation between diagnosis-level initial and final allowance predictions (intercepts-only model) (with individual controls). The ordinal ranking of impairment severity is highly consistent across adjudicative stages — diagnoses with higher initial allowance propensity also have higher final allowance propensity. By contrast, state-level initial/final correlation is near zero (), indicating that state-level variation is not systematic.
Model fit: The diagnosis model (Deviance Information Criterion (DIC) ; correct predictions) outperforms the state model (DIC ; correct) and the pooled model (DIC ; correct). Knowing the diagnosis alone predicts better than knowing all individual-level variables plus state.
Primary diagnosis codes alone understate diagnostic complexity at the point of application. Using a sample of 2009 DI qualified claimants (), Meseguer (2018) finds:
The appeals process — specifically ALJ assignment — has become the basis for the most precisely identified estimates of DI's causal effect on labor supply. French and Song (2014) exploit the quasi-random rotational assignment of appeal cases to ALJs (oldest cases get priority; judges vary substantially in allowance rates despite similar caseloads) to construct a judge allowance differential instrument. Key findings from million ALJ hearings, –:
The ALJ-stage IV identifies a different complier population than the DDS examiner IV in Maestas, Mullen, and Strand (MMS, 2013): ALJ cases are those that survived initial denial and pursued appeal, generally a more motivated and health-impaired subgroup. The convergence of estimates (– pp LFP reduction across both studies) strengthens the credibility of both.
The sequential determination process begins before Step 1: a potential applicant must first navigate to a field office (or online) to initiate a claim. Deshpande and Li (2019) show that this pre-application stage is not frictionless — office closings significantly reduce both applications and recipients, with recipients falling more than applications ( vs. ). The differential indicates that office availability affects not just how many people apply but which people apply: approved-quality, medium-severity applicants are most deterred; denied-quality applicants are least deterred. The dominant channel is congestion at neighboring offices ( of the decline), not increased travel distance ().
This finding has two implications for understanding the sequential process:
See DI Application Costs and Take-Up for the full mechanism and policy implications.
Mitchell and Thompson (2025) link the 2015 American Community Survey (ACS) to SSA administrative records (831 file, Detail Earnings Record (DER), Numident) to estimate racial/ethnic allowance rate differences at the initial and reconsideration stages for Title II applicants aged 25–65. This is the cleanest study to date on this question — prior work used SSA's native race variable, which has serious quality problems (enumeration at birth since 1987 only; coding scheme changed from 3 to 5 categories around 1980). The ACS linkage achieves a Protected Identification Key (PIK) match rate and provides Office of Management and Budget (OMB)-standard race/ethnicity classification.
| Group | Males (pp) | Females (pp) |
|---|---|---|
| NH American Indian/Alaska Native | ||
| NH Black | ||
| Hispanic | ||
| NH Asian |
These raw gaps are large. But they largely reflect compositional differences between racial groups in the characteristics that most strongly predict allowance — not disparate adjudication.
The dominant predictors of DI allowance at the DDS level are body system code (genito-urinary odds ratio (OR) ; malignant neoplasms OR ), age at application (– OR ), and pre-disability earnings. These factors are orders of magnitude more predictive than race/ethnicity. Racial groups differ substantially on all of them: NH American Indian/Alaska Native (AIAN) applicants have different diagnostic profiles, age distributions, and earnings than NH White applicants. After adjusting for age, body system code, earnings, education, marital status, and region using three independent methods (propensity score matching (PSM) with kernel algorithm, inverse probability weighting (IPW), and logistic regression with G-computation):
The one comparison that does not fully converge is NH Black males. Under logistic G-computation ( pp) and IPW ( pp), a small residual gap survives and is statistically significant. Under PSM, the gap is pp and not significant. This method sensitivity makes the evidence for a residual adjudication disparity fragile — it may exist, but the data cannot establish it firmly.
This paper covers only DDS initial decisions and DDS reconsideration — Stages 1 and 2 of the appeals process. It does not cover the ALJ hearing stage (Stage 3). This is consequential: the Government Accountability Office (GAO) (1992) found unexplained racial differences specifically at the ALJ hearing level ("Social Security: Racial Difference in Disability Decisions Warrants Further Investigation"), and the GAO (2004) reaffirmed concerns about accuracy and fairness at the hearings level. The Mitchell and Thompson (2025) finding of near-zero adjusted gaps at DDS stages does not resolve the ALJ-stage question. Whether racial disparities persist at the hearing level — where judge identity and case presentation play a larger role — remains open.
The paper's most striking supplementary finding is that racial groups differ dramatically in their propensity to apply relative to their self-reported disability rates. NH Black individuals apply at roughly higher rates than NH White individuals despite reporting only more disability. NH AIAN individuals apply more often despite reporting more disability. This over-application relative to self-reported health means that minority applicant pools include a higher share of applicants near the medical eligibility margin — producing lower allowance rates even under perfectly race-neutral adjudication. See DI Application Costs and Take-Up for the mechanism and implications.
Yates (2016) shows that the SSI acceptance rate is also shaped by factors outside the determination process itself. Using Massachusetts health reform (2006) as a difference-in-differences natural experiment, she finds that access to affordable health insurance raised the initial SSI acceptance rate by and the total acceptance rate by . The mechanism: SSI bundles cash benefits with Medicaid, so when non-SSI health insurance becomes available, marginal applicants who applied primarily for Medicaid drop out, leaving a purer pool of truly disabled individuals. The effect concentrates in Year 0 (immediate), consistent with this applicant pool channel rather than a slower application quality channel (where more doctor visits produce better medical records). The historical acceptance rate range of (Reagan) to (Clinton) reflects how broadly the rate responds to external political and economic context, not just medical eligibility. See SSI Allowance Rate for the full mechanism.
Low and Pistaferri (2019) exploit HRS–SSA Form 831 matched data to identify where the gender gap in DI rejection rates originates within the five-step framework. Among applicants with self-reported severe, permanent, work-preventing impairments:
The interpretation is that DDS evaluators assign women more residual functional capacity than observationally equivalent men — not because women's medical impairments are assessed differently, but because evaluators conclude women can find other work that men cannot. A structural model with disability vignettes confirms this is supply-side bias (), not demand-side misreporting or compositional differences in applicant health. Labor market outcomes validate the error: rejected work-limited women are employed at a rate in the years after rejection (vs. for rejected women without self-reported limitations), and the Rejected Disabled interaction is pp for women () but (n.s.) for men.
See DI Classification Errors for the full Type I/Type II framework and structural identification.
Using Survey of Income and Program Participation (SIPP)–SSA linked data (ages –), Contreary et al. confirm several features of the sequential determination process from the applicant side. The dominant predictor of allowance is age: applicants – have a pp higher allowance rate than applicants – — by far the largest marginal effect in their regression, confirming the vocational grid's age-threshold mechanism operates as a near-continuous gradient across the – range, not only at the formal cutoffs of 50 and 55. Employment recency matters but is secondary: intermittent employment versus consistent employment carries approximately pp allowance, and no prior employment carries to pp. Workers' compensation (WC) receipt at the 6-month pre-application window carries a pp allowance penalty, likely because recent WC activity signals residual work capacity relevant to Steps 4–5 RFC assessment. Ceased employment is statistically indistinguishable from consistently employed, suggesting that SSA examiners do not penalize applicants who recently left work as part of a health-driven exit.