Embedded Files
I'm an economist by training—and a collaborator at heart—focused on using data and rigorous methods to drive evidence-based policy. After earning a Ph.D. in Economics from Georgetown University in May 2022 and dual bachelor’s degrees in Physics and Economics from Rutgers University–Camden, I joined the U.S. Census Bureau’s Center for Economic Studies in the Evidence Group.
Currently, I serve as a Research Coordinator at the University of Michigan Law School, where I assist in creating infrastructure to conduct empirical research and leverage data to enhance the effectiveness of the Law School.
I maintain active collaborations with the Justice Innovation Lab and Free Our Vote, employing rigorous causal inference methods to better understand and improve the criminal justice system through evidence-based solutions. Additionally, I have previously taught as an adjunct instructor at Rutgers University–Camden, bringing cutting-edge research and real-world policy insights into the classroom.
Table of Contents
Table of Contents
Peer Reviewed Papers
Peer Reviewed Papers
De-prosecution and death: A comment on Hogan (2022)
De-prosecution and death: A comment on Hogan (2022)
(joint with Jacob Kaplan and Tom Scott)
R&R with conditional acceptance at Economic Inquiry
Abstract:
In a manuscript published in Criminology & Public Policy, Hogan (2022) presents results from a synthetic control model that suggests de-prosecution in Philadelphia in the mid to late 2010s resulted in a large increase in the number of homicides that occurred in the city. In this comment, we point out several potential problems with the analysis and re-estimate the relationship between de-prosecution and homicide under different model specifications. Our primary concerns include the short pre-intervention period, a failure to correct for imbalance over covariates in the synthetic control models, the use of homicide counts instead of rates as an outcome, an inaccurate description of the data used, and an inadequate explanation of data cleaning procedures including missing data. We reproduce the author’s results after addressing these issues and find that the effect presented in Hogan (2022) occurs only in certain model specifications and other decisions that maximize the reported effect of de-prosecution on homicide counts. Based on our findings, we conclude that the study should not be used to inform criminal justice policy. We call for a greater dedication to open science and reproduction/replication in economics and related fields.
Published as De-prosecution and death: A comment on the fatal flaws in Hogan (2022) on Aletheia
Felony Disenfranchisement and Voter Turnout: Randomized Trials in Iowa and Washington
Felony Disenfranchisement and Voter Turnout: Randomized Trials in Iowa and Washington
(joint with Alexander Billy and Neel U. Sukhatme)
AEA RCT Registry Entries: AEARCTR-0010141 and AEARCTR-0010199
Published in Northwestern University Law Review
Abstract:
Prior to the 2022 midterm elections, we conducted large-scale randomized controlled trials in Iowa and Washington aimed at increasing voter turnout among newly enfranchised individuals with past felony convictions. Alongside national and grassroots partners, we designed and implemented experiments to ascertain the effectiveness of alternative outreach mechanisms, including targeted mailers and digital ads. We did not detect statistically significant or economically meaningful effects on voter registration or turnout; most observed effects were precise nulls. The absence of measured impact is likely attributed to low digital engagement with our online ads as well as extensive voter outreach already conducted by our local partners prior to the study. Our evidence highlights the importance of context in voter outreach efforts, as the political and legal environment in Iowa and Washington differed significantly from other regions where similar interventions had previously shown success.
Prosecutors in the Passing Lane: Racial Disparities, Public Safety, and Prosecutorial Declinations of Pretextual Stops
Prosecutors in the Passing Lane: Racial Disparities, Public Safety, and Prosecutorial Declinations of Pretextual Stops
(joint with Donald Braman, Jared Fishman, Lily Grier, Kevin Himberger, Jarvis Idowu, Rory Pulvino, Jess Sorensen, Joanie Weaver)
Published in San Diego Law Review
Abstract:
In response to a growing set of empirical studies demonstrating their widespread discriminatory effects, pretextual stops have been subjected to decades of criticism from scholars, the public, and jurists. However, pretextual stops have been defended by some as a necessary public safety measure, particularly in the fight against violent gun crimes. Following a series of highly publicized police shootings of unarmed Black drivers during pretextual stops, and in the absence of substantial judicial or legislative guidance, a growing number of prosecutors have developed policies deprecating the prosecution of pretextual stops absent a clear public safety benefit. Without empirical evaluations of pretextual stops, however, it has been difficult for practitioners or justice advocates to rebut complaints that these new policies remove an important deterrent to crime and the circulation of illegal firearms. This Article reports the results of the first empirical evaluation of the impact of pretextual stops on crime and gun seizures, made possible by the Ramsey County Attorney’s decision to both decline prosecution of non-public-safety stops and to share data about those stops. We find that reduced stops have led to decreased racial disparities without affecting crime rates. Notably, the most common justification for pretextual stops—the recovery of illegal firearms—remained constant in the largest police department aligned with the new policy. We urge prosecutors to review the growing body of evidence we describe and contribute to this evidence base by gathering and sharing data about their policies with researchers. To that end, this Article provides a comprehensive review of relevant empirical evidence, reports the findings of the Ramsey County evaluation, and outlines how prosecutors and police departments in other jurisdictions can utilize the Ramsey County model to engage in evidence-based reform.
Cognitive Load and Mixed Strategies: On Brains and Minimax
Cognitive Load and Mixed Strategies: On Brains and Minimax
(joint with David Owens and John Smith)
Published in International Game Theory Review
Abstract:
It is well known that laboratory subjects often do not play mixed strategy equilibria games according to the theoretical predictions. However, little is known about the role of cognition in these strategic settings. We therefore conduct an experiment where subjects play a repeated hide and seek game against a computer opponent. Subjects play with either fewer available cognitive resources (high cognitive load treatment) or with more available cognitive resources (low cognitive load treatment). Surprisingly, we find some evidence that subjects in the high load treatment earn more than subjects in the low treatment. However, we also find that subjects in the low treatment exhibit a greater rate of increase in earnings across rounds, thus suggesting more learning. Our evidence is consistent with subjects in the low load treatment over-experimenting. Further, while we observe that subjects do not mix in the predicted proportions and that their actions exhibit serial correlation, we do not find strong evidence these are related to their available cognitive resources. This suggests that the standard laboratory deviations from the theoretical predictions are not associated with the availability of cognitive resources. Our results shed light on the extent to which cognitive resources affect (and do not affect) behavior in games with mixed strategy equilibria.
Working Papers
Working Papers
The First and the Worst: Evidence of Peer Effects in a Large U.S. Jail
The First and the Worst: Evidence of Peer Effects in a Large U.S. Jail
(joint with Angelo Diaz and Neel U. Sukhatme)
Abstract:
Do brief social interactions shape long-term behavior? Using daily cell-level data from a large U.S. jail---where individuals are quasi-randomly assigned to cellmates---we causally examine how short-term incarceration impacts future criminal activity through peer exposure. Our novel approach employs a standardized index that captures both the frequency and severity of peers' criminal histories, extending beyond traditional linear-in-mean models to reveal nuanced peer dynamics. We find that exposure to peers with serious criminal backgrounds significantly increases recidivism. Specifically, a one standard deviation increase in exposure to peers with severe criminal histories elevates rearrest likelihood within one year by up to 0.72 percentage points (1.64%). Strikingly, this influence is most pronounced from an individual's first cellmate and the most severe peer encountered, rather than cumulative exposure alone. We further validate these findings using a new structural estimation approach (Boucher et al. 2024), applied for the first time to an exogenously assigned peer network and expanded to estimate the temporal dimension of peer influence. Beyond direct peer effects, we document indirect spillovers: individuals are influenced not just by their immediate peers, but also by those peers' prior networks. This suggests a critical path dependence in social interactions during incarceration, reinforcing the disproportionate impact of initial peer exposure. Our findings have important policy and economic implications. Given that most jail detainees are legally presumed innocent awaiting trial, our evidence of negative peer influence calls attention to the harms of unnecessary pretrial detention and poorly structured initial cell assignments. Policies that minimize exposure to high-risk peers or reduce pretrial detention duration could significantly lower recidivism rates and correctional expenditures. Broadly, our work contributes to understanding social spillovers in constrained settings, with potential insights for labor markets, schools, and other institutional settings.
Reducing Pretextual Stops: Evidence from a Policy Reform in St. Paul, Minnesota
Reducing Pretextual Stops: Evidence from a Policy Reform in St. Paul, Minnesota
(joint with Rory Pulvino)
Abstract:
Traffic stops account for most police-initiated contacts in the United States, often serving as a broad gateway into the criminal legal system. In response to rising concerns about the harms of such stops—and the need to maintain public trust—this article examines a novel policy, jointly crafted by the Ramsey County Attorney’s Office and the St. Paul Police Department, that sharply curtailed traffic stops for minor vehicle violations. Alongside reducing stops, the policy replaces punitive measures with supportive interventions---aiming to foster cooperation and trust between law enforcement and the community. Using an interrupted time series framework, we find that the policy virtually eliminated these stops—particularly for Black motorists—without adversely affecting long-term crime rates, gun-related offenses, or 911 response times. In short, it relieved a significant source of discretionary police contact, lowered racial disparities, and showed no evidence of weakening public safety---highlighting a promising strategy for balancing efficiency, fairness, and community trust.
How Do Students Value an Elite Education? Evidence from NYC Specialized High Schools
How Do Students Value an Elite Education? Evidence from NYC Specialized High Schools
(joint with Lawrence Costa)
Revise & resubmit at Real Estate Economics
Abstract:
Are students willing to brave long commutes for access to good schools? Using New York City Department of Education administrative data matched with Google transit directions, we find that longer commutes from home markedly deter students from applying to even the most elite high schools. For the top public school in New York State, a student with a 20 minute commute is 73% more likely to apply than one who lives 40 minutes away. For two other schools above the 99th percentile of performance, the differences are 232% and 138%. We also find that eighth grade exam scores relate to how well students understand the admissions process.
Race, Criminal History, and Prosecutor Case Selection: Evidence from a Southern U.S. Jurisdiction
Race, Criminal History, and Prosecutor Case Selection: Evidence from a Southern U.S. Jurisdiction
Abstract:
State prosecutors are powerful actors who handle the vast majority of criminal cases in the United States. Furthermore, they are located at the core of the criminal justice pipeline, which runs from arrest to sentencing, and produces large persistent racial disparities. Given their centrality and influence, it is important to understand how prosecutors exacerbate or mitigate these disparities. This paper uses administrative data developed in collaboration with a prosecutor’s office located in a medium-sized jurisdiction in the southern United States. Using both descriptive and causal methods, I find that prosecutors are complex actors that are aware of upstream biases that taint signals they receive and act to reduce racial disparities. Specifically, after controlling for a rich set of covariates, I find that Black individuals receive shorter sentences and are more likely to have their charges dismissed relative to similarly situated White individuals. To learn more about what is happening at the margin, I leverage quasi-random assignment of cases to prosecutors and a simple—yet novel—model to show that prosecutors discount how prior convictions map into punishment for Black individuals relative to White individuals. My results suggest policies aimed at removing prosecutorial discretion or “blinding” prosecutors from knowing the race of an individual may have unintended consequences.
It's Always Sunny in Politics
It's Always Sunny in Politics
Abstract:
A desirable property of democratic elections is that they should not be influenced by forces that reveal no information about the candidate. However, the extant literature suggests that precipitation has a significant impact on electoral outcomes. This paper investigates an understudied dimension of weather—sunshine. Using novel daily weather measurements from satellites, linked to county level U.S. Presidential electoral returns from 19482016, we document how sunshine affects the decision making of voters. We find that election day exposure to sunshine increases support for the Democratic party on average. Additionally, we show that—contrary to prior findings that do not control for sunshine—precipitation has no detectable impact on partisan support, but universally depresses turnout. To rationalize our results we propose a mechanism whereby sunshine modulates voter mood which causes a change in voter choice, while precipitation only impacts turnout through increasing the cost of voting. We then build a theoretical model, which features this mechanism, and generates additional tests that find support in the data. Our main result—that election day sunshine noticeably impacts voter choice—highlights the need to reduce the effect of election day shocks (e.g. by allowing early voting). Furthermore, our results regarding precipitation suggest that reducing costs to voting does not confer partisan benefits—a potentially policy relevant finding for the current vote by mail discussion.
Using Rain for Electoral Gain: Evidence from FEMA’s Public Assistance Program
Using Rain for Electoral Gain: Evidence from FEMA’s Public Assistance Program
Abstract:
Distributive politics is the study of “who gets what, when, and how”. In this paper I find empirical support for predictions produced by a model of political competition at the state-level. To perform my analysis, I construct a novel county level dataset that merges fine-grain physical measures of large destructive storms, satellite data on existing infrastructure, demographic information from Census, and multiple types of relief spending. I find robust evidence that political parties target public spending to counties with higher historic turnout relative to their political neighbors. To give more credence to these results I confirm that each model passes multiple placebo tests. Additionally, to reduce the possibility that my results are driven by some systematic bias, I propose two credible instrumental variables and explicitly model the selection process that determines a counties eligibility for relief aid.
Works in Progress
Works in Progress
Hiding in Plain Sight: Default Biases in Applications of Generative AI
Hiding in Plain Sight: Default Biases in Applications of Generative AI
We tested ChatGPT’s legal reasoning using real police reports and attorney-style prompts to simulate how lawyers might use generative AI.
While we found no significant racial bias, ChatGPT consistently leaned toward prosecution, even in weak or legally flawed cases.
This “prosecutorial default” held across prompt types, roles (prosecutor vs. defense), and case facts—suggesting certain prompt features dominate outputs.
The model often failed to identify clear legal issues, favoring punitive outcomes over constitutional scrutiny or diversion.
These findings raise concerns about domain-specific AI biases and the risks of deploying LLMs in high-stakes legal settings without proper audits.
The Effects of Prosecutorial Screening on the Timing and Disposition of Cases
The Effects of Prosecutorial Screening on the Timing and Disposition of Cases
We evaluate a randomized rollout of a prosecutorial screening unit in South Carolina’s Ninth Circuit, designed to rapidly assess low-level cases using just the police report and criminal history.
Cases randomly assigned to screening were more likely to be dismissed quickly, with large increases in dismissal within 60 days.
Screening also led to shorter time-to-disposition, suggesting prosecutors focused more efficiently on serious cases.
The intervention used only existing staff and minimal additional resources, indicating low-cost, high-impact improvements to prosecutorial workflows.
These results highlight the value of early legal triage—screening helps prevent unnecessary system involvement without increasing risk to public safety.
Ranking Crime Types Using Combinatorial Hodge Theory and Survey Data
Ranking Crime Types Using Combinatorial Hodge Theory and Survey Data
(joint with Kevin Himberger)
Developed and implemented a survey to approximately 30 assistant prosecutors in medium-sized jurisdiction regarding the prioritization of crime types
Coded HodgeRank algorithm to get back global consistent ranking
Built python package rankpy to implement HodgeRank algorithm
Unionization, Labor Market Power, and Wage Inequality
Unionization, Labor Market Power, and Wage Inequality
Grants
Grants
A Tale of Three States: Bond Court Decisions and Outcomes in Illinois, Wisconsin, and Colorado
A Tale of Three States: Bond Court Decisions and Outcomes in Illinois, Wisconsin, and Colorado
Funded by Arnold Ventures
(joint with Don Stemen and Branden DuPont)
(joint with Don Stemen and Branden DuPont)
Reports
Reports
Traffic stop policy in Ramsey County, MN
Traffic stop policy in Ramsey County, MN
This report is the result of a collaborative effort between the Ramsey County Attorney's Office (RCAO), the Saint Paul Police Department (SPPD), the Ramsey County Emergency Communications Center (ECC), and Justice Innovation Lab (JIL) to assess the impact of RCAO and police department policies and practices, implemented in 2021, aimed at reducing traffic stops unrelated to public safety concerns. Code for cleaning and analyzing the data used in this report can be found in the JIL GitHub repository for public materials.
The analyses documented in the report provide evidence that reducing use of non-public-safety traffic stops reduces racial disproportionalities without negatively impacting crime rates. Overall, the data analyses indicate that the new traffic stop policy was successful in reducing minor, non-safety-related vehicle violation stops, that this reduction resulted in a narrowing of racial differences in traffic stops and searches, and that the policy had no discernible effect on crime rates.
The Case for Screening
The Case for Screening
Justice Innovation Lab, after analyzing data provided by the Ninth Circuit Solicitor’s Office in Charleston, South Carolina (SOL 9), released its findings in The Case for Screening report. JIL and SOL9 worked together to analyze case trends and develop a data-informed approach to improving outcomes through prosecutorial screening. This report summarizes major findings from JIL’s analyses, discusses results from a pilot screening process, and provides recommendations for future case screening work.
Disparity and Prosecution in Charleston, SC
Disparity and Prosecution in Charleston, SC
Press Coverage [1] [2] [3] [4]
This report was prepared for the Ninth Circuit Solicitor’s Office (SOL9) in Charleston County, South Carolina. It focuses on the outcomes of prosecutorial decision-making for felony and misdemeanor offenses in SOL9. It assesses the existence and extent of racial and ethnic disparities across the following four decision points: (1) Complete case dismissals; (2) Plea negotiations; (3) Changes in charges from referral to disposition; and (4) Sentencing.
Aging and the Labor Market in Thailand
Aging and the Labor Market in Thailand
In this project:
Assisted Harry Moroz in cleaning, harmonizing, and analyzing Thailand's Labor Force Survey
Produced a report on the effects of an aging labor market on the economy in Thailand
Report abuse