Ben-Shalom et al 2012 — Longitudinal Statistics for New Supplemental Security Income Beneficiaries

Summary

Full Mathematica Policy Research report underlying the 2012 policy brief, tracking new adult Supplemental Security Income (SSI) awardees longitudinally through Section 1619(b) work-incentive use, benefit suspension, and employment outcomes. Covers the 1996 award cohort over 11 years and compares experiences across cohorts (1996–2006), states, and the disability insurance (DI) program. The central finding is that longitudinal rates of benefit suspension for work are $4$–$6\times$ the contemporaneous cross-sectional rates, and that the two statistics are mathematically consistent — they measure the same phenomenon from different angles.

Key Claims

• 1996 cohort, 11-year track: $11.9\%$ were in 1619(b) status for at least one month by the end of 2007, compared to $1.9\%$ cross-sectional in December 2009 — a $6\times$ gap. Of these, only $7.4\%$ were in 1619(b) without concurrent DI payments, meaning the gross 1619(b) figure double-counts concurrent beneficiaries.
• Concurrent DI penetration: $31\%$ of the 1996 SSI award cohort received DI benefits in their award year; by 2007, $\approx 40\%$ had received DI in at least one month. Only $44\%$ of those who ever achieved net-suspended-or-terminated-for-work (NSTW) were doing so free of DI payments.
• Employment: $28.1\%$ of 1996 SSI awardees were employed in at least one year from 1998–2006 — nearly identical to the 1996 DI cohort ($28.0\%$), despite structural differences between the two programs.
• Age concentration of 1619(b) activity: Under-40 awardees accounted for $78.1\%$ of all 1619(b) months accumulated by the 1996 cohort. The DI analogue is $62\%$, showing SSI's even steeper age concentration.
• 2001 cohort summary (6-year): $9.8\%$ achieved NSTW in $\geq 1$ month; $8.4\%$ achieved 1619(b) in $\geq 1$ month; $5.5\%$ achieved NSTW with no concurrent DI; $18.6\%$ employed at some point 2001–2007. Ages 18–19: $61\%$ ever employed, $24\%$ ever in 1619(b). Ages 50–61: $9\%$ ever employed, $3\%$ ever in 1619(b).
• Cross-state variation: 2001 cohort cumulative employment ranged from $12.2\%$ (West Virginia) to $29.3\%$ (Wyoming). Cumulative NSTW from $6.4\%$ (WV) to $19.9\%$ (Alaska). Service enrollment from $6.8\%$ (WV, MS) to $22.8\%$ (Vermont). Strong positive cross-state correlation between service enrollment rate and NSTW years per $1{,}000$ awardees ($r = 0.75$); high-NSTW states nearly identical for SSI and DI.
• Cross-cohort trends: 2001–2003 award cohorts had notably lower cumulative employment and 1619(b) stats than 1997–2000 cohorts at the same post-award age, consistent with the 2001 recession. Recovery was partial; later cohorts did not reach 1999 cohort levels before the end of observation. The gross NSTW rate rose continuously for cohorts from 2001 onward, but the DI-overlap-free NSTW showed a weaker trend, suggesting the gross increase was partly driven by the rising share of concurrent beneficiaries.
• DI-SSI comparison (Exhibit VIII.1): 2001 SSI cohort NSTW rate ($9.8\%$) substantially higher than 2001 DI cohort suspension rate ($3.9\%$), consistent with Section 1619's superior incentive structure (no substantial gainful activity (SGA) cliff, Medicaid continuation). In both programs, $77\%$ of eventual milestone activity occurs within the first 5 years after award.
• First-five-years rule: For the 1996 SSI cohort, the cumulative 1619(b) percentage at year 5 ($8.9\%$) was $77\%$ of the value at year 10 ($11.5\%$). Stapleton et al. (2010) found the identical $77\%$ figure for DI awardees, suggesting a universal dynamic of rapid early action followed by sharply diminishing returns.

Causal Mechanisms Discussed

• Section 1619(b) allows SSI beneficiaries to keep Medicaid coverage even when earnings exceed SSI payment threshold, removing the coverage-loss cliff that would otherwise deter work; concurrent beneficiaries face the opposite incentive — to stay below SGA to preserve DI and Medicare.
• Service enrollment is positively correlated with NSTW outcomes ($r = 0.75$ cross-state), but causality cannot be established from observational data; both may reflect latent demand for work.
• Business cycle effects operate at the cohort level: cohorts entering during contractions show persistently lower longitudinal outcomes, not just contemporaneous suppression.
• Age gradient: most work activity by young awardees occurs early, creating a "critical window" in which interventions could permanently change the trajectory.

Concepts Introduced or Extended

• DI Return-to-Work Milestones
• Disability Insurance Program

Entities Mentioned

• Yonatan Ben-Shalom
• David Stapleton
• Dawn Phelps
• Maura Bardos
• Social Security Administration

Quotes

"Although 1.9 percent of working-age SSI beneficiaries had their benefits suspended for work under the 1619(b) work-incentive program in the month of December 2009, by the 11th year after award (2007), 11.9 percent of those in the 1996 SSI award cohort had been in 1619(b) status for at least one month."

"Our finding that 28.1 percent of the 1996 SSI award cohort was employed in at least one year from 1998 to 2006 is almost identical to the corresponding percentage for the 1996 DI cohort (28.0 percent)."

My Take

The full report's most important addition over the 2012 brief is the demonstration that SSI and DI have roughly parallel longitudinal employment profiles despite different program structures — suggesting that the determinants of return-to-work are primarily beneficiary-level (age, impairment, prior work attachment) rather than program-structural. The DI-overlap correction is analytically crucial: without it, the SSI NSTW rate ($9.8\%$) overstates the number of beneficiaries who genuinely exit the public benefit system for work, since $44\%$ of NSTW months still involve DI payments. The cross-state correlation ($r = 0.75$) between service enrollment and NSTW years is striking but underpowered for causal inference — it could reflect services causing better outcomes, or both reflecting unmeasured state economic or demographic factors.