Aizer Gordon and Kearney 2013 — Exploring the Growth of the Child SSI Caseload

Summary

Aizer, Gordon, and Kearney examine the $40\%$ growth in the child Supplemental Security Income (SSI) caseload from 2002 to 2012 (roughly $900{,}000$ to $1.3$ million). Using publicly available Social Security Administration (SSA) caseload data merged with state-level panel data on economic conditions, health insurance coverage, special education enrollment, and population health, they document that the growth is driven almost entirely by a doubling of mental-diagnosis cases — primarily attention deficit hyperactivity disorder (ADHD), autism, and "other mental" conditions — while intellectual disability cases actually fell $45\%$ and physical disability grew only $24\%$. State-panel regressions show that the only factor robustly predictive of mental-diagnosis initial allowances, after controls and Texas exclusion, is special education enrollment prevalence; child poverty, health insurance, and population-level mental diagnosis rates are not predictive. The authors do not find a smoking-gun causal explanation and the results are sensitive to specification.

Key Claims

• The child SSI caseload grew from $\approx 900{,}000$ to $\approx 1.3$ million between 2002 and 2012 ($+40\%$); on a per-$1{,}000$-children basis, growth was also $\approx 38\%$, ruling out population expansion as an explanation.
• Mental diagnoses more than doubled: from $\approx 340{,}000$ to $\approx 700{,}000$ ($4.6$ to $9.8$ per $1{,}000$ children), growing from $36.9\%$ to $57\%$ of the total caseload.
• Intellectual disability cases declined $45\%$ ($240{,}000 \to 133{,}000$), possibly partly due to reclassification of borderline cases toward "other mental" (autism, ADHD), but the magnitude of the intellectual decline cannot account for the full mental increase.
• Physical disability grew only $24\%$ ($337{,}000 \to 416{,}000$), falling from $36.8\%$ to $32.6\%$ of the caseload.
• Gender composition of mental-diagnosis cases was stable throughout (boys $\approx 65\%$, girls $\approx 35\%$); growth rates were similar by gender ($+110\%$ for boys, $+116\%$ for girls in per-$1{,}000$ terms). No particular age group within the 3–17 range drove the growth.
• Large state-level heterogeneity: caseload growth ranged from $+20\%$ (Wyoming) to $+129\%$ (Texas). Texas drives results in population-weighted regressions.
• Regression result (Table 1): special education prevalence (Pctsped) is the only variable that is statistically significant for initial allowances once all controls are included and Texas is excluded. The coefficient on Pctsped is significant for allowances but not for applications, suggesting special education may affect the acceptance rate for applications rather than the volume of applications.
• Child poverty, health insurance coverage, and the share of the broader child population with mental diagnoses are not robustly predictive of caseload growth.
• Declining Continuing Disability Review (CDR) rates are hypothesized as a potentially important driver (CDR budget fell sharply after 2004; Government Accountability Office (GAO) 2012 documents the decline), but the CDR mechanism is not directly tested with regression.

Concepts Introduced or Extended

• SSI Children's Program — documents the mental-diagnosis composition shift 2002–2012; failure to find clean economic or health explanations supports the view that administrative factors (CDR decline, special education pipeline) are at least as important as underlying health trends

Entities Mentioned

• Anna Aizer
• Nora Gordon
• Melissa Kearney

Quotes

"The growth in child SSI caseloads has come from new and continuing mental health cases, in particular, 'other mental' disabilities such as ADHD and autism; there was no substantial increase in the number of cases with an intellectual disability or a physical disability."

"Once all controls are included and Texas excluded, the only factor that remains statistically significantly predictive of caseload growth is the prevalence of special education in the state."

"The role of Continuing Disability Reviews will become increasingly important in maintaining caseload sizes. But in fact, the rate of CDRs over this period declined substantially."

My Take

Useful descriptive census of the composition shift but limited as causal analysis. Three cautions:

1. Results are fragile: The regression findings hinge on Texas exclusion and control choice. The "special education predicts allowances" finding is suggestive but not strong causal evidence — it could reflect reverse causation (states with more SSI-type children also have more special ed) or confounding through unobserved state fiscal culture.

2. Reclassification vs. true incidence: The near-simultaneous $45\%$ decline in intellectual disability and doubling of "other mental" cases is consistent with diagnostic substitution — borderline intellectual disability cases being reclassified as autism spectrum or ADHD. The authors acknowledge this but cannot rule it out. If true, some of the "growth" is compositional rather than a genuine increase in impaired children.

3. CDR mechanism plausible but untested: The sharp CDR budget cuts after 2004 (from $\approx 100\%$ of due reviews to under $2\%$ by fiscal year 2006) are mentioned but not empirically exploited. Given that mental-diagnosis cases are the most review-sensitive (harder to verify, less likely to be "medically improvement not expected"), the CDR collapse could explain a substantial share of the caseload growth. This would be an important avenue for a follow-on causal paper.

The paper establishes "what grew and where" but not definitively "why." For the wiki's purposes, it establishes that the child SSI growth since 2002 is a mental-health composition shift, not a physical disability story — a fact with direct implications for understanding CDR policy and the future trajectory of the program.