SSI Children's Program

Definition

The Supplemental Security Income (SSI) children's program provides monthly cash payments to the parents of children under 18 who have a qualifying disability and limited household income and assets. Unlike Social Security Disability Insurance (SSDI/DI), which requires a work history, SSI is means-tested and funded from general revenues. In most states, SSI automatically qualifies children for Medicaid. As of 2014, the maximum federal monthly payment was $\$721$ per child.

Key Ideas

SSI is the largest federal means-tested cash program in the U.S. ( $\$49$ billion to $7.8$ million recipients in 2011), more than double Temporary Assistance for Needy Families (TANF) ( $\$18$ billion).
Enrollment tripled since 1980. Children's enrollment reached $1.3$ million by 2012 (up $40\%$ from 2002 alone), driven almost entirely by a doubling of mental-diagnosis cases (Attention Deficit Hyperactivity Disorder (ADHD), autism, other mental) while intellectual disability cases declined $45\%$ and physical disability grew only $24\%$ . Mental diagnoses rose from $37\%$ to $57\%$ of the total caseload (Aizer, Gordon, and Kearney 2013).
The program is means-tested: the child's SSI benefit is reduced as parental earnings rise, phasing out completely at roughly $\approx \$3{,}000$ /month for a typical single-parent family. Benefits are effectively zero-taxed up to $\approx \$2{,}000$ /month in parental earnings.
SSI children face periodic medical reviews to verify continued eligibility; an unfavorable review results in removal from the program. Children who are not reviewed cannot be removed for medical reasons.
At age 18, children must requalify under adult standards — a far stricter screen that terminates benefits for roughly half of age-18 redetermination cases (the basis of the companion Deshpande American Economic Review (AER) paper).

How It Works

Eligibility

The child must have a physical or mental impairment that "results in marked and severe functional limitations" and is expected to last at least 12 months or result in death.
Family income and assets must fall below Social Security Administration (SSA) thresholds. A portion of parental income is "deemed" to the child and counts against the benefit.
The child's benefit is calculated after deeming and is paid monthly to the parent.

Medical Reviews (Continuing Disability Reviews for Children)

Unlike adults who may receive mailer-based reviews, children receive only Full Medical Reviews (FMRs) — in-person examinations by Disability Determination Services (DDS) examiners. Reviews are scheduled at the time of award based on the examiner's assessment of the child's prognosis ("medical improvement expected," "medical improvement possible," or "medical improvement not expected"). The standard for removal is medical improvement review standard — the examiner must show that the child has experienced medical improvement since the last review, not merely that the child does not currently meet the Listings.

SSA's budget for childhood medical reviews has varied substantially over time, creating quasi-random variation in review rates that Deshpande (2016) exploits:

Fiscal Year	Review rate (% of due)
FY2002	$\approx 100\%$
FY2004	$\approx 100\%$
FY2005	$\approx 40\%$
FY2006	$\approx 1.2\%$

Children who receive an unfavorable review and do not prevail on appeal are removed from the program. Of those due for review in FY2002, $\approx 14\%$ were removed. Removal is not permanent — children can reapply.

Institutional History: Sullivan v. Zebley (1990) and 1996 Welfare Reform

Before 1990, the disability screen for non-working children used only the Listings of Impairments (same as adults), with no comparable vocational factors screening-in mechanism. The 1990 Supreme Court decision Sullivan v. Zebley found that SSA's "Listings-Only" approach violated the "comparable severity" standard of the Social Security Act; it ordered a redetermination of all childhood denials since 1980 and directed SSA to create an Individual Functional Assessment (IFA) evaluating whether a child could "function independently, appropriately, and effectively in an age-appropriate manner." The resulting eligibility expansion caused a surge in the child SSI caseload. The 1996 welfare reform legislation (Personal Responsibility and Work Opportunity Reconciliation Act, PRWORA; P.L. 104-193) eliminated the IFA, removed references to "maladaptive behaviors" from the Listings, and required age-18 redeterminations using adult criteria. More than 100,000 children were removed from the rolls; 45% of the first cohort of age-18 redeterminations resulted in benefit cessation. The caseload initially dropped, then resumed growth, eventually surpassing its pre-reform level. (Davies, Rupp, and Wittenburg 2009)

Age-18 Redetermination

When an SSI child turns 18, SSA automatically redetermines eligibility under adult DI/SSI standards (the five-step sequential determination process). The adult standard does not use functional domains but instead uses the Listings, Residual Functional Capacity (RFC), and vocational grid. Roughly half of SSI children are terminated at age 18. Those who lose benefits at 18 had a median SSI tenure of $\approx 7$ years — far longer than the younger children in Deshpande (2016).

PRWORA 1996 created a sharp discontinuity: children born on or after August 22, 1996 were subject to mandatory age-18 reviews under the stricter adult standard; those born before were grandfathered. Deshpande (2016b) exploits this birthdate regression discontinuity (RD) design with $81{,}800$ SSI children within $\pm 37$ weeks of the cutoff. First stage: $-10$ percentage points (pp) SSI enrollment pooled; unfavorable review = $+39$ pp removal. See Causal Effects of DI Receipt.

SSI Children: Descriptive Profile (December 2010)

Stegman Bailey and Hemmeter (2014), using 2008 Survey of Income and Program Participation (SIPP) matched to SSA administrative records weighted to December 2010 SSA totals, document the profile of all $1{,}193{,}848$ noninstitutionalized SSI child recipients under 18. This represents $15.9\%$ of total SSI enrollment ( $7{,}497{,}490$ ), with the remainder split between working-age adults ( $58.3\%$ ) and the aged 65+ ( $25.8\%$ ).

The SSI recipient population differs sharply from DI on demographic composition: $56.7\%$ female (vs. $51.3\%$ male for DI); $30.5\%$ Black (vs. $12.7\%$ for DI); $20.5\%$ Hispanic. SSI's means-tested design — requiring no work history — draws more heavily from populations with weak prior labor market attachment.

Income levels and program dependence: $58\%$ of all SSI recipients had 4-month personal income under $\$3{,}000$ (under $\$750$ /month); $60\%$ had 4-month family income under $\$7{,}500$ (under $\$1{,}875$ /month). The 2010 federal SSI benefit rate was $\$674$ /month for an individual. Income composition: SSI itself = $43.9\%$ of recipient income; Social Security (Old-Age and Survivors Insurance (OASI) or DI) = $22.0\%$ ; earnings = $22.5\%$ .

Poverty and anti-poverty impact: $42.9\%$ of SSI recipients fall below the federal poverty threshold. Without SSI, $65.3\%$ would be in poverty; SSI reduces the aggregate poverty gap by $71.3\%$ . Despite this, the post-SSI poverty rate remains high — the $\$674$ /month benefit level is below the poverty line for most household configurations, so SSI reaches the poverty threshold for some but leaves nearly half of all recipients still below it.

Health insurance: $94.9\%$ of SSI recipients are enrolled in Medicaid (SSI confers automatic Medicaid eligibility in most states); only $2.2\%$ are uninsured. This near-universal Medicaid coverage is SSI's most distinctive feature relative to DI (where $7.7\%$ are uninsured and Medicare access requires a $24$ -month wait).

Housing and material resources: $34.4\%$ own homes; $61.8\%$ receive Supplemental Nutrition Assistance Program (SNAP); $4.9\%$ receive housing assistance.

Parental Labor Supply Response (Deshpande 2016a)

Deshpande (2016) is the first study to estimate the causal effect of child SSI removal on household economic outcomes using large-scale administrative data and a credible identification strategy (the FY2004/05 budget cut RD and difference-in-differences (DiD)).

Key findings:

Full income replacement, intensive margin only. A $\$1{,}000$ reduction in annual child SSI raises parental earnings by $\$700$ – $\$1{,}400$ (elasticity $\geq 1$ ). Working parents earn more; employment rate does not change. Total household income is maintained.
No substitution to other disability income. Parent and sibling disability applications fall $50$ – $100\%$ after child removal, but disability receipt does not fall — the deterred applications are marginal (would not have been approved).
Asymmetric entry/exit response. The entry-margin earnings elasticity (examiner instrumental variable (IV) $\approx 0.29$ ) is far smaller than the exit-margin elasticity ( $\geq 1$ ), consistent with loss aversion, habit formation, or target-income behavior.
Household application clustering. $65\%$ of SSI children have a parent or sibling who ever applies for disability; $15\%$ have a family member apply within $60$ days of the child's own application. Job loss or information about program availability triggers simultaneous multi-member applications.

Long-Run Effects of Age-18 Removal (Deshpande 2016b)

Deshpande (2016b) provides the first causal estimates of the long-run consequences of age-18 SSI removal using the PRWORA 1996 birthdate RD.

Income and earnings:

Removed youth earn $+\$825$ /year more than retained controls — recovering only $\frac{1}{3}$ of the $\$2{,}170$ /year SSI loss
Total observed income falls $\$1{,}570$ /year; present discounted value (PDV) over $16$ years = $-\$21{,}000$ ( $-19\%$ )
Youth who remain off SSI earn $\$4{,}400$ /year at age $30$ ; PDV loss $-\$76{,}000$
Comparison benchmarks: removed youth ( $\$6{,}300$ ) earn far less than denied SSI applicants ( $\$14{,}000$ ) or National Longitudinal Survey of Youth 1997 (NLSY97)–Aid to Families with Dependent Children (AFDC) youth ( $\$17{,}800$ )

Income volatility:

Coefficient of variation quadruples ( $0.682 \to 1.010$ )
Average removed youth is at the 90th percentile of the control group's volatility distribution
For $\gamma=2$ , $8$ – $23\%$ of total welfare loss is attributable to increased volatility, not just lower income levels

Household and siblings:

Parents do not offset the income loss (in contrast with 2016a, where parental earnings offset is complete and immediate)
Household PDV falls $\$16{,}400$
Siblings lose $-\$580$ /year in earnings (significant); mechanism unclear

No human capital channel:

No significant effects on education, incarceration, marriage rates, or next-generation SSI receipt
SSI removal does not appear to create a "poverty trap" through reduced investment; the constraint is earnings capacity, not incentives

Welfare implications — Marginal Value of Public Funds (MVPF):

Program	MVPF
SSI (risk-neutral)	$0.90$
SSI ( $\gamma=2$ , risk-averse)	$1.03$
Earned Income Tax Credit (EITC)	$0.88$
Food stamps	$0.53$ – $0.66$
Housing vouchers	$0.79$

SSI's high MVPF reflects the low fiscal externality: removed youth earn so little that the tax revenue gained from their additional earnings barely offsets the SSI savings.

Criminal Justice Outcomes of Age-18 Removal (Deshpande and Mueller-Smith 2022)

Deshpande and Mueller-Smith (2022) extend the Deshpande (2016b) birthdate RD to criminal justice outcomes using Criminal Justice Administrative Records System (CJARS) linked records. The central finding is that SSI removal causes large, persistent increases in criminal activity — specifically income-generating crime — which reframes the SSI children's program as a crime-prevention tool.

Main IV estimates (effect of unfavorable age-18 review):

Total charges: $+0.464$ (control mean $2.041$ ; $+22.7\%$ )
Income-generating charges (theft, burglary, fraud/forgery, robbery, drug distribution, prostitution): $+0.380$ (control mean $0.625$ ; $+60.8\%$ )
Non-income-generating charges: $+0.085$ (not significant)
Annual incarceration likelihood: $+2.9$ pp (control mean $4.7\%$ ; $+61.7\%$ )

Why income substitution, not idleness: The absence of any effect on non-income-generating crimes (assault, vandalism, driving under the influence (DUI)) rules out substance abuse or "idle hands" as mechanisms. The effect is exclusively concentrated in crimes that generate income for the perpetrator, consistent with SSI functioning as an income floor that, when removed, causes youth to substitute criminal for legal income.

Path dependence: Crime effects grow over the 16-year follow-up even as the contemporaneous SSI income gap narrows. The mechanism is absorbing-state entry: criminal records, incarceration experience, and post-release labor market stigma progressively foreclose legal employment options, amplifying the initial crime effect over time.

Resolves 2016b null incarceration finding: Deshpande (2016b) found no significant incarceration effect. This was a data limitation — SSA administrative incarceration records are far less complete than CJARS. With CJARS the incarceration effect is large and precisely estimated.

Cost-benefit: SSI savings = $\$49{,}121$ net present value (NPV) per removed youth. Government administration + incarceration costs = $\$40{,}934$ — nearly wiping out the fiscal savings. Victim costs (McCollister et al. 2010) add another $\$85{,}628$ . MVPF = $5.6$ (no victim costs), $16.1$ (with victim costs). Every dollar spent on SSI for this population generates $\$16.10$ in social value.

Unmet Health Needs After Age-18 Exit (Hemmeter 2011)

Hemmeter (2011), using the 2001–2002 National Survey of SSI Children and Families (NSCF) ( $n=869$ , ages $19$ – $23$ ), documents that the insurance shock of SSI exit is the primary driver of post-redetermination health care gaps:

Gap magnitude: $59\%$ of SSI exiters report any unmet health need vs. $32\%$ of continuing SSI recipients ( $27$ pp gap); medical $36\%$ vs. $12\%$ ; prescription drugs $36\%$ vs. $6\%$ .
Insurance mechanism: $96\%$ of continuing SSI recipients have Medicaid; only $25\%$ of exiters do, and $61\%$ of exiters are uninsured vs. $2\%$ of continuers. When insurance is controlled for in logistic regressions, the SSI status coefficient for any unmet medical need becomes statistically insignificant. Controlling for health status leaves the coefficient virtually unchanged.
Magnitude of insurance effect: Being uninsured vs. Medicaid raises the likelihood of any unmet need by $+48$ pp, medical by $+42$ pp, dental by $+33$ pp, prescription drug by $+27$ pp (all $p<0.01$ ).
Dental paradox: In the fully adjusted model, continuing SSI recipients are $9$ pp more likely to have unmet dental needs than exiters — Medicaid dental coverage for adults is less comprehensive than other Medicaid benefits, and SSI continuers have more severe underlying conditions.
Prescription drug residual: A $9$ pp gap in unmet prescription drug needs persists even after insurance controls, concentrated in those with behavioral/mental disorders who most need medication management and are most likely to be ceased at the age-18 redetermination.
Policy implication: Insurance expansion — not disability-specific interventions — is the more effective remedy for this population. The health consequences of age-18 removal complement the income and criminal-justice consequences documented by Deshpande (2016b) and Deshpande and Mueller-Smith (2022).

Household Financial Outcomes and Credit Access (Deshpande, Gross, and Wang 2017)

Deshpande, Gross, and Wang (2017) apply the same PRWORA birthdate RD-DiD design as Deshpande (2016b) but link the SSI sample to near-census Public Access to Court Electronic Records (PACER) bankruptcy records (Gross, Notowidigdo, and Wang 2014), yielding the first causal estimates of government cash assistance effects on household financial outcomes.

Core finding: SSI removal reduces household bankruptcy rates, a counterintuitive result.

Outcome	IV estimate	Control mean	% change
Parent ever files for bankruptcy	$-12$ pp	$14.6\%$	$-80\%$
Parent number of filings	$-0.26$	$0.22$	$-100\%$
SSI youth ever files	not sig.	$3.0\%$	—
Household (parent + child) ever files	$-9.2$ pp	$16.5\%$	$-55\%$

Effect is driven entirely by Chapter 7 (low-income) bankruptcies; no significant effect on Chapter 13. Results are robust across polynomial orders and with/without covariates.

Proposed mechanism — credit access channel: SSI income is stable and predictable relative to earned income, making recipient households creditworthy. Losing SSI means losing credit access. With no credit, there is nothing to default on, so bankruptcy rates collapse mechanically. As of September 2017 this mechanism was hypothesized but untested — credit bureau data were pending.

Interpretation: The result complements Deshpande (2016b)'s finding that removed youth lose substantial income that parents do not offset. Together, the papers paint a picture of households that lose both income and the ability to smooth consumption through credit — a double financial shock. The paper is preliminary and should be read alongside Deshpande, Gross, and Su (2021), which rigorously establishes the DI–financial distress connection for adult applicants.

Pre-Transition Health Care Profile (DeCesaro and Hemmeter 2009)

Using the same NSCF dataset (N=3,155; 797,958 weighted), DeCesaro and Hemmeter document the health care needs and out-of-pocket expenses of SSI children before the transition to adulthood.

Coverage and unmet needs:

93% of SSI children are covered by Medicaid; only 7% have non-Medicaid insurance; only 35 unweighted cases (1.1%) are uninsured.
9.5% report delayed or foregone health care in the prior 12 months. This exceeds the national average for children (1.1–1.6%) but is lower than estimates for disabled children broadly (~12%).
24% report any medical out-of-pocket (MOOP) expenses; population-average MOOP = $155/year; conditional average = $657/year.

Medicaid as the dominant protector:

Non-Medicaid insured SSI children have 1.50× higher odds of unmet needs, 4.67× higher odds of any MOOP expense, and 86% higher MOOP amounts when they have expenses.
Among non-Medicaid insured SSI children with an unmet need, 69% have MOOP expenses vs. 34% of Medicaid-insured.
81% of Medicaid-insured recipients have MOOP < 10% of their SSI payment; only 59% of non-Medicaid insured do.

MOOP is modest relative to SSI income: 75%+ of children with any MOOP have expenses below 10% of their annual SSI payment. 95% have MOOP fully covered by SSI. The SSI payment ($5,500/year on average) dwarfs the average MOOP burden.

Administrative barriers exceed financial barriers: Of those with unmet needs, 68% cite administrative reasons (can't locate provider, can't schedule appointment); 38% cite cost. Medicaid eliminates financial barriers more effectively than it eliminates access barriers.

SSI income is fungible: When asked about a hypothetical $100/month income increase, families primarily indicate spending on food (48%), personal items (40%), and debt (15%). Only 3–5% would increase disability-related spending. SSI functions as general income support, not earmarked medical reimbursement.

Transition warning: This pre-transition baseline directly precedes the Hemmeter (2011) finding that SSI exiters at age 18 experience 2× the unmet health needs of continuers (59% vs. 32%), driven entirely by Medicaid loss (61% uninsured vs. 2%). The good pre-transition coverage makes the post-exit deterioration all the more striking.

Longitudinal Outcomes of Childhood Awardees (Davies, Rupp, and Wittenburg 2009)

Using SSA administrative records for 1980 and 1997 award cohorts (N≈600,000 total across six cohorts), Davies et al. document the long-run SSI participation and earnings of individuals first awarded SSI as children.

SSI participation — lifetime reliance is common:

For the 1980 cohort (age 17 at award), participation declines monotonically but slowly to ~40% at age 45 — the last observation point. Among the 1980 cohort (age 0 at award), ~60% are still on SSI by age 10 after the initial first-5-years drop.
Rupp and Scott (1998) estimated the mean SSI duration (including multiple spells) for 1974–1982 childhood awardees at ~27 years before reaching age 65 — meaning lifetime reliance is the modal trajectory.
The 1997 cohort shows meaningfully faster SSI exit across all age-at-award groups, consistent with the 1996 welfare reform's structural shift toward stricter standards and shorter durations.
Exits are driven primarily by excess income (earnings above the means test), not death. Death accounts for fewer than 7% of first SSI spells ending; excess income/resources account for 50–60% of childhood suspensions in 1999–2007.

Employment and earnings — far below the general population, flat trajectory:

Childhood SSI awardees: ~40% have positive earnings in adulthood vs. ~80% of the general noninstitutional population in the same age groups.
Average earnings conditional on employment: $10,000–$13,000 for ages 24–38 vs. $33,000–$48,000 for the general population — consistently less than half.
Critically, the age-earnings profile is flat for SSI awardees. Unlike the steep wage growth seen in the general population from ages 19–38, SSI awardees show essentially no wage growth: the same $10,000–$13,000 ceiling applies at age 24–28 as at age 34–38. SSI awardees' earnings are consistently below the full-time, full-year minimum wage average across age groups.
Employment peaks around early-mid 20s and slowly declines thereafter through age 45 for the 1980 cohort.
The 1997 cohort shows higher employment (positive earnings at ~~50% around age 20) than the 1980 cohort (~~20–28% for comparable ages), consistent with the more rapid SSI exit.

Why It Matters

Income Dependence and Household Behavior

The full earnings offset shows that low-income households treat SSI as a reliable income stream and respond strongly to its removal — more strongly than standard neoclassical theory predicts. This suggests that SSI has a larger behavioral footprint than its dollar value implies, and that its removal imposes costs (labor supply substitution, child care time tradeoffs) beyond the direct income loss.

Program Access and Marginal Applicants

The finding that applications fall after SSI removal but receipt doesn't — because deterred applicants would have been denied anyway — mirrors the Deshpande-Li (2019) finding from field office closings. Both results point to the same mechanism: marginal (likely-to-be-denied) applicants are most sensitive to costs and information, while inframarginal (clearly eligible) applicants persist regardless. The implication for program integrity is counterintuitive: administrative barriers and removal events do not improve targeting — they reduce the signal value of the application process by deterring the wrong people.

Household vs. Individual Model of Disability Take-Up

The application clustering evidence challenges the standard individual-health model of disability take-up. Households appear to apply for disability collectively in response to income shocks, not just in response to individual health shocks. This means application rate trends partially reflect household economic conditions (employment, divorce, job loss) rather than only population health trends. It also implies that reducing barriers for one family member may increase take-up for others via information and familiarity spillovers.

Open Questions

Why do removed SSI youth fail to recover more of the income loss in earnings? Deshpande (2016b) identifies two candidate mechanisms — genuine disability limiting productivity, and poverty-driven low human capital — but cannot disentangle them cleanly. The earnings response is increasing in parental earnings, pointing to a role for the poverty channel.
What happens to health outcomes of removed youth? Income volatility quadruples and Medicaid eligibility is lost; the long-run health consequences are not yet estimated.
Does the sibling spillover ( $-\$580$ /year earnings) operate through income shocks (family resources constrained) or expectation-updating about disability program generosity? Administrative data cannot distinguish these.
Has the shift in SSI children's diagnosis toward mental/behavioral conditions (ADHD, autism) changed household dynamics? Mental-condition parents in Deshpande (2016a) show smaller earnings elasticity ( $0.52$ vs. $0.90$ for non-mental), possibly reflecting greater child care demands.

Caseload Composition Shift: Mental vs. Physical (Aizer, Gordon, and Kearney 2013)

Aizer et al. (2013) document the national mental-diagnosis surge using SSA state-level data for 2002–2012. The $40\%$ overall growth masks entirely different dynamics by diagnosis type:

Diagnosis type	2002 count	2012 count	Change
Mental (other)	$\approx 340{,}000$	$\approx 700{,}000$	$+106\%$
Physical	$\approx 337{,}000$	$\approx 416{,}000$	$+24\%$
Intellectual disability	$\approx 240{,}000$	$\approx 133{,}000$	$-45\%$

The simultaneous decline in intellectual disability and rise in "other mental" diagnoses partly reflects reclassification: borderline cases previously coded as intellectual disability are now coded as autism spectrum or ADHD. But the magnitude of the mental increase exceeds the intellectual decline by $360{,}000$ cases, so reclassification alone cannot explain the growth.

State-panel regressions ( $2002$ – $2012$ , $N=456$ state-years) find that special education prevalence is the only robust predictor of mental-diagnosis allowances after controlling for poverty, health insurance, demographics, and total SSI benefit levels. The special education effect operates through allowances, not applications — suggesting that special education enrollment increases the probability that an application is accepted (perhaps by providing medical documentation or credibility to the disability claim), not just the number of applications filed.

A critical unexamined factor is the decline in Continuing Disability Reviews (CDRs): from $\approx 100\%$ of due reviews in FY2002–2004 to under $2\%$ by FY2006. Mental impairments, which are harder to verify and more likely to be classified as "medical improvement possible," are most vulnerable to CDR removal — and most insulated from it when reviews don't occur. The CDR mechanism is consistent with the data but was not formally tested.