Can Strategic Text Messages Reduce Parole Violations Among High-Risk Parolees?

Aboaba et al. (2026) ask whether a low-cost messaging intervention can deter crime among high-risk parolees. They analyze data from a randomized controlled trial in New York, using administrative records on arrests, parole violations, and neighborhood crime. The authors find no meaningful reduction in arrests or violent crime. However, they report a 3 percentage point decline in parole violations—a 15% reduction—and a 2 percentage point decline in absconding violations, a 25% decrease. These findings suggest that while the intervention did not reduce crime, it did improve compliance with parole conditions among participants.

Why This Article Was Selected for The Policy Scientist

This article addresses a central policy question: whether low-cost deterrence strategies can meaningfully reduce crime, a concern that has intensified amid fiscal constraints and declining tolerance for high-cost enforcement. The topic is timely given sustained recidivism rates and growing interest in scalable interventions that can substitute for incarceration. The authors, who have contributed extensively to the deterrence and policing literature, extend prior work rooted in Becker’s framework and modern focused deterrence strategies. The randomized controlled design represents a clear methodological strength, allowing credible causal inference.

Full Citation and Link to Article

Aboaba, O., Chalfin, A., LaForest-Tucker, M., Parker, L., & Sharkey, P. (2026). Can crime be deterred at low cost? Evidence from a randomized experiment in New York. Journal of Policy Analysis and Management, 45(2), e70092. https://doi.org/10.1002/pam.70092

Central Research Question

The article examines whether a low-cost, deterrence-based messaging intervention can meaningfully reduce criminal behavior among high-risk parolees. Specifically, it asks whether a one-time “notification forum”—in which individuals are informed of the consequences of reoffending while also being offered social services—can deter future arrests, reduce parole violations, and generate spillover effects within communities. The inquiry is grounded in a broader policy concern: whether scalable, inexpensive interventions can substitute for more resource-intensive strategies such as incarceration or intensive policing. The study also evaluates whether such interventions produce effects not only at the individual level but also at the neighborhood level, which is critical for assessing policy relevance.

Previous Literature

The study builds on a long tradition of deterrence theory originating with Becker (1968), which posits that individuals respond to incentives by weighing the expected costs and benefits of criminal activity. Empirical work has generally supported the importance of certainty of punishment over severity, with policing interventions often demonstrating measurable deterrent effects. More recent research has emphasized “focused deterrence,” which targets high-risk individuals or groups through a combination of enforcement threats and social support. Prior quasi-experimental studies have reported reductions in violence associated with such strategies, though concerns remain about selection bias and external validity.

Randomized evidence in this domain is comparatively limited. Existing experiments have produced mixed findings, with some suggesting reductions in reoffending and others lacking statistical power. The present study contributes to this literature by providing a large-scale randomized evaluation of a relatively low-intensity intervention. It also extends prior work by explicitly testing for spillover effects and neighborhood-level impacts, which are often absent from earlier evaluations.

Data

The analysis relies on administrative data from the New York State Division of Criminal Justice Services and the Department of Corrections and Community Supervision. These data include detailed criminal histories, demographic characteristics, and parole violation records for individuals in the study sample. The dataset spans multiple years, allowing for both short-term and longer-term outcome measurement. Arrest data include both overall arrests and violent felony arrests, while parole data capture violations that lead to revocation and reincarceration, including absconding and technical violations.

In addition, the study incorporates publicly available New York City Police Department data to assess neighborhood-level crime outcomes, including crime complaints, arrests, and shootings. The use of administrative data ensures high reliability and minimizes measurement error, as outcomes are recorded through official channels rather than self-reports. However, the sample is restricted to high-risk parolees in specific geographic areas, which may limit the extent to which findings generalize to broader populations or jurisdictions with different institutional contexts.

Methods

The study employs a randomized controlled trial design, representing a strong approach to causal inference. Randomization occurs at two levels: neighborhoods and individuals. In New York City sites, census tracts are first randomly assigned to treatment or control conditions, and then eligible individuals within those tracts are randomly assigned to receive an invitation to attend the notification forum or to a control group. This two-stage design allows the authors to estimate both direct treatment effects and potential spillover effects within communities.

The primary analysis estimates intent-to-treat effects, comparing outcomes for individuals assigned to treatment with those assigned to control, regardless of actual participation. Treatment-on-the-treated estimates are also calculated to account for noncompliance, though results are substantively similar. Regression models include covariates such as age, race, prior criminal history, and fixed effects for randomization blocks to improve precision. The study further conducts separate analyses to isolate individual-level effects, test for spillovers, and evaluate neighborhood-level impacts. This multi-layered design strengthens internal validity and provides a comprehensive assessment of the intervention’s effects.

Findings/Size Effects

The results indicate no statistically meaningful effect of the intervention on arrests. Approximately 23 percent of individuals in both treatment and control groups are rearrested within six months, with an estimated treatment effect of −0.6 percentage points. The confidence interval excludes large effects but cannot rule out modest reductions. Similarly, there is no detectable impact on violent felony arrests or on neighborhood-level crime, suggesting that the intervention does not produce broader public safety benefits.

In contrast, the study finds a notable effect on parole violations. Individuals assigned to the treatment group experience a reduction in violations of approximately 3 percentage points, corresponding to a 15 percent decrease relative to the control mean. This effect is driven primarily by reductions in absconding violations, which decline by about 2 percentage points, or 25 percent. These effects are concentrated in the first six months following the intervention and are consistent across multiple specifications.

The absence of spillover effects further indicates that the intervention’s impact is limited to those directly exposed to the messaging. Individuals in treated neighborhoods who were not invited to the forums show no differences in outcomes compared to those in control neighborhoods. Similarly, there is no evidence of changes in neighborhood-level crime rates, with estimated effects close to zero.

Conclusion

The findings suggest that low-cost deterrence messaging, as implemented in this study, does not reduce criminal activity but does influence compliance with parole conditions. This distinction is substantively important, as it indicates that the intervention affects behaviors that are more directly tied to supervision and monitoring rather than more complex or impulsive criminal actions. The results are consistent with a mechanism in which the intervention increases the salience of formal sanctions for rule violations without altering underlying propensities for crime.

From a methodological perspective, the use of a randomized controlled trial provides credible evidence on causal effects, addressing limitations in prior quasi-experimental work. The high-quality administrative data further enhance the reliability of the findings. However, the specificity of the sample—older, high-risk parolees in a low-crime urban context—suggests caution in generalizing results to other populations or settings. Future research could explore whether different populations, higher-intensity interventions, or repeated exposures produce stronger effects on criminal behavior.

Overall, the study contributes to the literature by clarifying the limits of low-cost deterrence strategies and by demonstrating that measurable behavioral changes can occur even in the absence of reductions in crime.