Criminal Background
Chaos Before Peace: The Impact of Exposure to a Prison Riot on Incarcerated Persons
Hina Usman & Jillian Turanovic
Journal of Quantitative Criminology, September 2025, Pages 387-428
Methods: We use administrative records from a state prison system (2010–2019) to construct a panel dataset of over 3600 individuals who served time in a facility where a major riot occurred in 2015. We treat the riot as an exogenous event and use a sharp RD design to assess the effects of riot exposure on recidivism (i.e., return to prison) and on prison misconduct. Identification tests confirm the suitability of the RD design in the context of the riot, thus ruling out threats to validity.
Results: Our findings show that individuals exposed to the prison riot were 7–12% less likely to return to prison than those unexposed; and that individuals exposed to the riot were less likely to engage in misconduct in the form of property, security, and defiance infractions.
Weaker the gang, harder the exit
Megan Kang
Criminology, forthcoming
Abstract:
This study draws on 95 interviews and observations with gang-affiliated individuals in Chicago to examine how gang structures shape disengagement and desistance from crime. During the last two decades, the city's gangs have experienced a decline in group closure, or their capacity to regulate membership and member behavior, and a blurring of boundaries between those active in a gang from all others. In the past, Chicago's gangs maintained closure and bright boundaries that made gang affiliations, norms, and territories clearly defined. Leaving these gangs required costly exit rituals that signaled an unambiguous departure while facilitating desistance. Today, with weaker gang structures and blurry boundaries, leaving a gang is no longer a distinct event. The ease of gang disengagement, however, makes desistance harder as inactive members struggle to knife off past ties and access turning points. In this uncertain landscape, desistance tactics can backfire, sending “blurred signals” -- behaviors intended to create distance from former affiliates and rivals but appear as wavering commitment to supporters -- that trap individuals in a liminal space between social worlds. Contrary to leading desistance theories that emphasize individual readiness, opportunity, and prosocial bonds, this study underscores how group structures critically shape pathways out of crime.
Does black and blue matter? An experimental investigation of race, perceptions of police, and legal compliance
Mackenzie Alston & Emily Owens
Journal of Public Economics, September 2025
Abstract:
Using an online experiment, we examine the role of race in perceptions of policing and willingness to violate the law. Black and White subjects were asked to imagine themselves driving in a real (but unnamed) majority White or majority non-White US city in which the police department was either majority White or majority non-White. Subjects were incentivized to drive quickly but also warned that they might receive a ticket based on their speed, their personal characteristics, and the real-life ticketing practices in that city. Expectations about the likelihood of being ticketed for speeding were higher among Black subjects than among White subjects. Black subjects also showed a higher willingness to pay for information on the racial makeup of the police department. However, Black subjects drove at similar speeds under all conditions tested, regardless of the racial demographics of the city, the racial composition of the police department, or the disparity or congruence of the two. By contrast, White subjects drove faster when police departments were not racially representative.
Linguistic cues and deception: A replication study of 911 homicide calls
Patrick Markey, Jennie Dapice & Brittany Wickham
Psychology, Public Policy, and Law, forthcoming
Abstract:
This study aimed to replicate Burns and Moffitt’s 2014 investigation into using linguistic markers to detect deception in 911 homicide calls. Linguistic analysis and machine learning models, including logistic model tree induction, naïve Bayes, neural networks, random forests, and discriminant analysis, were applied to an independent and larger sample of 148 calls. The sample was evenly divided between false allegation callers and true report callers. These analyses utilized the linguistic inquiry and word count software to assess model accuracy, sensitivity, and specificity through cross-validation techniques. Although certain linguistic markers such as first-person singular pronouns, anxiety, and negation words showed statistically significant associations with deception, many of the associations identified by Burns and Moffitt were not replicated. The five models in the current study yielded an average accuracy of 57.0%, with an average sensitivity of 57.2% and specificity of 57.4%, which are substantially lower than the original study’s reported averages of 77.6% accuracy, 80.8% sensitivity, and 74.8% specificity.
Longer-Term Impacts of a Youth Behavioral Science Intervention: Experimental Evidence from Chicago
Nour Abdul-Razzak et al.
University of Chicago Working Paper, June 2025
Abstract:
We conduct a large-scale, randomized controlled trial of a six-month intervention combining intensive mentoring and group cognitive behavioral therapy (CBT) for youth in Chicago, following study participants for up to five years. The program was designed to engage young people at higher risk of engagement with the criminal justice system, and successfully did so with a take-up rate of 62%. Over 24 months, youth offered the program experienced an 18% reduction in the probability of being arrested, with no impact on number of arrests. We find a significant impact on violence engagement, with a 23% reduction in the probability of a violent-crime arrest within 24 months. We find the program's impact in preventing any arrest persists into adulthood, up to four years post randomization. The program moderately improves school engagement in the first year as well. Sub-population analyses suggests that all youth are benefiting from the program, but that the program may be moving different outcomes for different groups of youth in ways related to baseline risk of engaging in the justice system or disengaging from school. We conclude that programs that combine CBT and mentoring can serve as a model to engaging a harder-to-reach population of youth, predominantly outside of school, and be cost effective in reducing criminal justice contact in the longer run.
Estimating the Effects of Restitution Penalties on Juvenile Recidivism
Jennifer Smith & Brandon Vick
Youth Violence and Juvenile Justice, forthcoming
Abstract:
This study investigates the potential causal relationship between the assignment of financial restitution penalties to juvenile offenders and subsequent recidivism, defined as being sentenced for a new offense within the next two years. Although the use of financial penalties on juveniles is prevalent, there is limited research on its potential link to reoffending. Using a dataset of all juvenile offenders in the state of Pennsylvania from 2015 to 2017 (n = 15,382), we implement a propensity-score matching model to estimate the effects of restitution assignment on future recidivism and test for endogeneity between the two using a control function model. Controlling for offender demographics, criminal history, and offense type and severity, we estimate that restitution assignment increases recidivism rates by 26-percentage points (p < 0.001) and do not find statistical evidence of endogeneity. Higher rates of recidivism are found for those assigned restitution across all offense grades, categories, and dispositions.
Testing a new procedure to administer lineups to witnesses in the field to eliminate law enforcement’s reliance on the use of highly suggestive showups
Mitchell Eisen et al.
Psychology, Public Policy, and Law, forthcoming
Abstract:
We partnered with law enforcement to conduct two experiments using the field-simulation paradigm, in which participants (N = 979) were immersed in what they were led to believe was an actual police investigation. We field-tested the efficacy of replacing live showups with photographic lineups administered to witnesses on a computer tablet in a squad car, and directly compared identifications obtained using showups and lineups under the same highly suggestive circumstances that showups are typically conducted: when witnesses are told that a suspect has been detained in the area shortly after the crime. As predicted, when showups were administered under field conditions, choosers were overconfident in their identification decisions. Although confidence and response time did not postdict accuracy for choosers when showups were used, both confidence and response time were well aligned with accuracy for rejectors. Indeed, high-confidence showup rejections were diagnostic of innocence. In contrast, photographic lineups conducted under field conditions yielded very few high-confidence false suspect identifications, and both confidence and response time were well aligned with accuracy for choosers but not rejectors. Results also showed that the inclusion of an additional-opportunities instruction lowered false identifications for both showups and lineups without decreasing culprit identifications. These data contribute to a growing body of field research showing how real-world field conditions differentially affect eyewitness performance when conducting showups versus lineups. Moreover, this study provides a practical demonstration of how police can capitalize on recent technological developments to move away from their continued reliance on the use of highly suggestive showups by employing some version of the field-lineup procedure tested in this effort.
What is the Best Response? Examining the Impact of Police and Their Alternatives
Bocar Ba et al.
NBER Working Paper, October 2025
Abstract:
Cities across America are adopting civilian crisis response programs as alternatives to traditional policing, yet causal evidence on their impact and cost-effectiveness is scarce. This paper evaluates Durham, North Carolina’s HEART program, which diverts nonviolent 911 calls from police. Using a difference-in-differences design, we find that HEART reduces crime reports, arrests, and response times -- primarily through civilian phone and in-person responses, rather than police-civilian co-responses. The program increases future 911 calls, which suggests it fosters public trust. Based on an original contingent valuation survey and applying the marginal value of public funds framework, we conclude that HEART is a fiscally self-sustainable intervention.
Prosecutorial Reform and Local Crime Rates
Amanda Agan et al.
NBER Working Paper, October 2025
Abstract:
Many communities across the United States have elected reform-minded prosecutors who seek to safely reduce the reach and burden of the criminal justice system. In this paper, we use variation in the timing of when these prosecutors took office across jurisdictions to empirically characterize their policy changes and estimate downstream effects on prison incarceration rates, local reported crime rates, and drug mortality rates. We find that after a reform prosecutor takes office there are consistent and often statistically significant decreases in charging and conviction rates for nonviolent misdemeanor offenses, particularly misdemeanor drug offenses, but not for violent or felony offenses. We find little to no downstream effects on prison incarceration rates and no effects on local reported crime rates or drug mortality rates. These findings suggest that the types of policies being implemented by reform prosecutors appear to be decreasing the footprint of the criminal justice system without adverse effects on public safety.
Have U.S. gun buyback programs misfired?
Toshio Ferrazares, Joseph Sabia & Mark Anderson
Journal of Policy Analysis and Management, Fall 2025, Pages 1211-1249
Abstract:
Gun buyback programs (GBPs), which use public funds to purchase civilians' privately-owned firearms, aim to reduce gun violence. However, next to nothing is known about their effects on firearm-related crime or deaths. Using data from the National Incident Based Reporting System, we find no evidence that GBPs reduce gun crime. Given our estimated null findings, with 95% confidence, we can rule out decreases in firearm-related crime of greater than 1.2% during the year following a buyback. Using data from the National Vital Statistics System, we also find no evidence that GBPs reduce suicides or homicides where a firearm was involved. These results call into question the efficacy of city gun buyback programs in their current form.