Summary
Deshpande and Mueller-Smith (2022) extend the Deshpande (2016b) birthdate regression-discontinuity (RD) design — which exploited the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) 1996 age-18 redetermination cutoff of August 22, 1996 — to criminal justice outcomes using Criminal Justice Administrative Records System (CJARS) administrative records linked to the original Supplemental Security Income (SSI) sample. The core finding is that youth removed from SSI at age 18 are substantially more likely to be charged with crimes, especially income-generating crimes, and more likely to be incarcerated. These effects persist and grow over the 16-year follow-up period even as the direct SSI income effect shrinks. The paper concludes that SSI functions as a crime-prevention program, with a cost-benefit accounting that makes SSI positive net-present-value (NPV) even on fiscal grounds alone when victim costs are included.
Key Claims
Research Design
- RD design: Running variable is date of 18th birthday relative to August 22, 1996 (the date PRWORA was enacted). Births on/after the cutoff face mandatory age-18 redetermination under adult standards; births before are grandfathered.
- Bandwidth: 10-month window (±5 months around the cutoff); 28,843 observations in the first-stage sample.
- First stage: ~90% of those born after the cutoff receive an age-18 redetermination. The unfavorable (removal) rate jumps by +36 percentage points (pp) at the cutoff. Instrumental-variable (IV) estimates = RD reduced-form / 0.36.
- CJARS linkage: CJARS (Finlay and Mueller-Smith 2021a) covers ~50% of the US population via exact name + date-of-birth (DOB) + state matching. Approximately ⅓ of the Deshpande (2016b) sample has CJARS records (N=21,768 for charge outcomes; N=26,991 for incarceration outcomes).
Main IV Estimates (Effect of Unfavorable Age-18 Review)
- Total charges: +0.464 criminal charges (control mean 2.041; +22.7%)
- Income-generating charges: +0.380 (control mean 0.625; +60.8%). This taxonomy includes: theft, burglary, fraud/forgery, robbery, drug distribution, prostitution.
- Non-income-generating charges: +0.085 (not statistically significant)
- Annual incarceration likelihood: +2.9 pp (control mean 4.7%; +61.7%)
- Employment >$15K/year: +4.4 pp (control mean 11.3%; +38.9%)
- The crime response is entirely concentrated in income-generating offenses — consistent with SSI as an income substitute, not a behavioral/substance-use mechanism.
Mechanism: Income Substitution
- Removed youth face a $2,170/year SSI income loss and recover only ⅓ in legal earnings (+$825/year). The crime effect specifically targets income-generating offenses with no effect on non-income-generating crimes (assault, vandalism, driving under the influence [DUI]), ruling out "idleness" or substance-use as the primary channel.
- The bimodal labor market: removed youth specialize either in legal employment OR criminal income generation, not both simultaneously. Only ~11.3% of the control group earns >$15K/year in any given year, implying the legal labor market absorbs very few of the marginally removed youth.
- Path dependence: Crime entry is an absorbing state. The criminal record, incarceration experience, and post-incarceration labor market scarring progressively close off legal employment opportunities. This explains why crime effects grow over time even as the contemporaneous SSI income gap narrows (the cohort has partly aged back onto SSI by the end of the follow-up period).
Heterogeneity
- Effects are concentrated among youth who had no prior criminal history before age 18. Those with prior criminal exposure show smaller and noisier responses.
- Males drive the criminal justice response; females show primarily an incarceration response.
- Effects are larger for youth in low-employment states, consistent with the income-substitution channel (legal employment alternative is scarce).
Cost-Benefit Analysis (Table II)
| Item |
Per-removed-youth |
Direction |
| SSI savings (NPV, 16 yr) |
$49,121 |
Benefit to government |
| Administrative + incarceration costs |
−$40,934 |
Cost to government |
| Net fiscal benefit (without victim costs) |
+$8,187 |
Near zero |
| Victim costs (McCollister et al. 2010) |
−$85,628 |
Cost to society |
| Net social cost (with victim costs) |
−$77,441 |
Large net loss |
- Marginal value of public funds (MVPF) (Hendren and Sprung-Keyser 2020): 5.6 without victim costs; 16.1 with victim costs (i.e., every $1 spent on SSI generates $16.10 of social benefit when victim costs are internalized).
- The near-zero net fiscal benefit implies that even from a narrow government-budget perspective, SSI removal generates almost no savings. With victim costs, removal is deeply negative-NPV.
Comparison to Prior Literature (Deshpande 2016b)
- Deshpande (2016b) found no significant effect of SSI removal on incarceration. The current paper resolves this: the CJARS sample is substantially larger and the CJARS crime data is far more complete than SSA's administrative incarceration records used in 2016b. The null incarceration finding in 2016b was a data limitation, not a true null effect.
Concepts Introduced or Extended
Entities Mentioned
Quotes
"We find that removing youth from SSI increases criminal charges by 22.7 percent and the likelihood of incarceration by 61.7 percent."
"The effect is driven entirely by income-generating crime, consistent with an income substitution mechanism."
"We calculate that SSI has an MVPF of 16.1 when we account for the costs borne by victims, relative to an MVPF of 5.6 when we exclude victim costs."
"These findings indicate that SSI serves as a crime prevention program for this population of low-income youth."
My Take
This paper is a careful extension of a well-identified prior design, made possible by an unusually complete administrative crime dataset. The income-substitution interpretation is credible because the effect is entirely concentrated in income-generating crimes and the labor market channeling evidence is consistent with a bimodal specialization story. The most important econometric contribution is the path-dependence mechanism: the fact that crime effects grow over time even as the direct SSI income gap closes, because criminal record incapacitation forecloses legal alternatives progressively.
Two caveats: (1) The CJARS sample covers only ~50% of the US population, so the matched subsample may differ from the full Deshpande (2016b) sample — the authors address this with balance checks but it remains a generalizability concern. (2) The victim cost accounting follows McCollister et al. (2010) using jury award valuations, which are contested and likely vary substantially by crime type and victim population. The 16.1 MVPF is very sensitive to the victim cost figure.
The broader implication — that means-tested disability income for low-income youth is a crime prevention tool — reframes the political economy of SSI. Prior debate centered on work-disincentive effects; this paper shows the relevant counterfactual is not employment but crime, which inverts the welfare calculus.