Does Exposure to More Aggressive Field Training Officers Increase Police Use of Force?

Adger, Ross, and Sloan (2025) examine whether field training officers (FTOs) influence police recruits’ later use of force. They study administrative data from the Dallas Police Department, linking 911 calls, force reports, and officer records from 2013–2019. Using quasi-random assignment of recruits to FTOs, they find that recruits trained by higher-force FTOs are significantly more likely to use force themselves. A one standard deviation increase in an FTO’s prior force propensity raises recruit force use by about 14–18%, with effects lasting up to two years. They conclude that on-the-job training transmits behavioral norms across officers.

Why This Article Was Selected for The Policy Scientist

How do institutional practices shape consequential state behavior, particularly in policing where discretionary decisions carry high social costs? The focus on training is especially timely given ongoing reform debates that emphasize procedural fixes without strong causal evidence. Adger, Ross, and Sloan (2025), who have contributed to this emerging literature, provide rare evidence that workplace socialization can transmit behavioral norms with measurable persistence. The Dallas administrative dataset is unusually rich, linking calls, force, and personnel records, though external validity may be constrained by institutional features. The quasi-experimental design leveraging as-good-as-random assignment represents a clear strength, advancing the literature beyond prior descriptive and regression-based approaches.

Full Citation and Link to Article

Adger, C., Ross, M. B., & Sloan, C. (forthcoming). The effect of field training officers on police use of force. American Economic Review. https://doi.org/10.1257/aer.20240785 

Central Research Question

This article examines whether exposure to a field training officer (FTO) with a higher propensity to use force causally influences a police recruit’s subsequent use of force. The core inquiry is not simply whether officers differ in behavior, but whether those differences are transmitted through structured, on-the-job training relationships. The authors situate this question within broader debates about police reform, where training is frequently proposed as a mechanism to reduce excessive force but where credible causal evidence remains limited. By focusing on the formative field training period—an apprenticeship-style phase widely viewed within policing as the most consequential stage of professional development—the study seeks to determine whether supervisory influence shapes long-term enforcement behavior. More specifically, it asks whether recruits assigned to more force-prone FTOs adopt similar behavioral patterns, and whether such effects persist beyond the training period into independent policing activity. 

Previous Literature

The article contributes to several overlapping literatures, including research on police behavior, workplace training, and supervisor effects. Prior work in policing has documented variation in officer use of force and identified correlates such as situational factors, peer exposure, and institutional context. However, much of this literature relies on descriptive or regression-based approaches that cannot establish causality. A smaller set of studies has examined peer effects, showing that exposure to injured colleagues or minority peers can influence arrest and force decisions, though these effects are typically short-term or context-specific.

Beyond policing, a substantial literature in labor economics demonstrates that supervisors can influence worker productivity, behavior, and organizational culture. Studies using quasi-experimental designs—often exploiting random assignment or personnel rotations—have shown that managerial practices can have persistent effects on subordinate performance. However, most of this work focuses on routine or low-stakes tasks, leaving open the question of whether similar dynamics operate in high-discretion, high-consequence environments such as law enforcement.

The present study builds on these strands by isolating a specific mechanism—formal training relationships—and examining its impact on a consequential outcome: use of force. It extends prior work by providing causal evidence of behavioral transmission through supervision in a setting where decisions are discretionary, context-dependent, and socially salient. 

Data

The analysis draws on detailed administrative data from the Dallas Police Department covering the period from 2013 to 2019. The primary dataset consists of approximately 3.9 million 911 calls for service, which are linked to officer-level records on use of force, arrests, and personnel characteristics. The authors also incorporate data from force reporting systems, district attorney records, and internal personnel files, allowing for a comprehensive view of both officer behavior and contextual factors surrounding each incident.

A key feature of the data is the ability to reconstruct recruit–FTO pairings during the field training period. Although the department does not maintain a direct historical record of these pairings, the authors infer them using call-level co-attendance patterns, supplemented by overtime records associated with FTO evaluation duties. This approach yields a sample of 521 recruits and 251 distinct first-phase FTOs.

The richness of the dataset allows the authors to control for a wide range of confounding factors, including call characteristics (e.g., priority, type, location, timing), officer demographics, and institutional assignments. It also enables the construction of a pre-treatment measure of each FTO’s propensity to use force, based on their behavior prior to being assigned a given recruit. This measure is central to the empirical strategy and is refined using Empirical Bayes techniques to account for variation in sample size and measurement error. 

Methods

The study employs a quasi-experimental design that leverages the as-good-as-random assignment of recruits to FTOs within divisions and training cohorts. According to both institutional knowledge and empirical validation, these assignments are not systematically related to recruit characteristics, conditional on division and cohort. This feature provides the basis for causal identification.

The primary empirical model estimates the effect of an FTO’s prior force propensity on a recruit’s likelihood of using force after completing training. The key independent variable is a standardized measure of FTO force propensity, constructed from pre-assignment data and adjusted using Bayesian shrinkage. The dependent variable is an indicator of whether a given call for service results in use of force by the recruit.

The specification includes a comprehensive set of fixed effects and controls, including division-by-cohort fixed effects, call characteristics, geographic indicators, and officer attributes. Standard errors are clustered at both the recruit and FTO levels to account for within-group correlation. The authors also conduct extensive balance tests to confirm that FTO assignment is not correlated with observable recruit characteristics, reinforcing the credibility of the identification strategy.

This design represents a clear improvement over conventional multivariate regression approaches, as it approximates a natural experiment. While not a randomized controlled trial, the conditional random assignment mechanism provides strong quasi-experimental variation, allowing for credible causal inference. 

Findings/Size Effects

The central finding is that FTO behavior has a statistically and substantively significant effect on recruit use of force. A one standard deviation increase in an FTO’s prior propensity to use force leads to a 14 to 18 percent increase in the recruit’s likelihood of using force on a given call. This effect persists for up to two years after the completion of field training, indicating that the influence of early supervisory exposure is durable rather than transitory.

The results further show that the effect is driven primarily by the first FTO assignment, consistent with institutional accounts that the initial training phase is the most formative. Additional analyses suggest that the mechanism operates through the transmission of norms and expectations about appropriate use of force. Recruits trained by higher-force FTOs are more likely to engage in discretionary uses of force, respond to marginally more dangerous calls, and make arrests for lower-level offenses that are less likely to result in prosecution.

Simulation exercises indicate that replacing high-force FTOs with lower-force counterparts could reduce overall recruit use-of-force incidents by approximately 10.6 percent, corresponding to a reduction of about 68 incidents per year in the sample context. These estimates likely understate the full effect, as they do not account for dynamic feedback loops in which trained recruits later become FTOs themselves.

Overall, the magnitude, persistence, and behavioral specificity of the effects provide strong evidence that supervisory training relationships play a central role in shaping enforcement behavior. 

Conclusion

The study provides compelling causal evidence that on-the-job training transmits behavioral norms in policing, with measurable and persistent effects on use of force. By exploiting quasi-random assignment to field training officers, the authors isolate a mechanism that has been widely discussed but rarely identified with causal precision. The findings have clear implications for the design of training programs and the selection of supervisors, as they suggest that early-career exposure can shape long-term patterns of behavior in high-stakes settings.

From a methodological perspective, the use of quasi-experimental variation represents a significant advancement over prior research relying on observational correlations. The study demonstrates how administrative data, when combined with institutional knowledge and careful design, can yield credible estimates of causal effects in complex organizational environments.

At the same time, the generalizability of the findings depends on the extent to which other jurisdictions share similar training structures and assignment processes. While the Dallas setting appears broadly representative of U.S. policing practices, variation across departments may limit external validity. Nonetheless, the core insight—that supervisory influence during training can shape consequential decision-making—likely extends beyond this specific context.

In sum, the article advances both the empirical and conceptual understanding of how institutional practices influence police behavior, offering a rigorous foundation for future research on training, supervision, and organizational culture in public sector settings.