Traffic Stops Research Articles

Mitigating Bias Due to Race and Gender in Machine Learning Predictions of Traffic Stop Outcomes

Traffic stops represent a crucial point of interaction between citizens and law enforcement, with potential implications for bias and discrimination. This study performs a rigorously validated comparative machine learning model analysis, creating artificial intelligence (AI) technologies to predict the results of traffic stops using a dataset sourced from the Montgomery County Maryland Data Centre, focusing on variables such as driver demographics, violation types, and stop outcomes. We repeated our rigorous validation of AI for the creation of models that predict outcomes with and without race and with and without gender informing the model. Feature selection employed regularly selects for gender and race as a predictor variable. We also observed correlations between model performance and both race and gender. While these findings imply the existence of discrimination based on race and gender, our large-scale analysis (>600,000 samples) demonstrates the ability to produce top performing models that are gender and race agnostic, implying the potential to create technology that can help mitigate bias in traffic stops. The findings encourage the need for unbiased data and robust algorithms to address biases in law enforcement practices and enhance public trust in AI technologies deployed in this domain.

Leveraging body-worn camera footage to assess the effects of training on officer communication during traffic stops.

Can training police officers on how to best interact with the public actually improve their interactions with community members? This has been a challenging question to answer. Interpersonal aspects of policing are consequential but largely invisible in administrative records commonly used for evaluation. In this study, we offer a solution: body-worn camera footage captures police-community interactions and how they might change as a function of training. Using this footage-as-data approach, we consider changes in officers' communication following procedural justice training in Oakland, CA, USA, one module of which sought to increase officer-communicated respect during traffic stops. We applied natural language processing tools and expert annotations of traffic stop recordings to detect whether officers enacted the five behaviors recommended in this module. Compared with recordings of stops that occurred prior to the training, we find that officers employed more of these techniques in posttraining stops; officers were more likely to express concern for drivers' safety, offer reassurance, and provide explicit reasons for the stop. These methods demonstrate the promise of a footage-as-data approach to capture and affect change in police-community interactions.

The impact of weather on the racial composition of traffic stop and citation issuance in the United States

ABSTRACT We analyze how adverse weather influences the racial composition of traffic stops and the issuance of citations within the context of administrative discretion. The behavior of police officers and drivers may be affected by weather, leading to a change in the composition of stopped drivers. If changes in racial composition are due to police officers choosing to forego traffic stops because those are more costly in wet conditions, it could indicate racial bias. Based on data from 37 city and 11 state police departments, our analysis reveals that the racial composition of drivers stopped and the probability of a citation being issued are not independent of adverse weather conditions, especially for state agencies. These findings contribute to the existing literature on factors influencing police–citizen interactions and provide a foundation for investigating the potential connections among weather, administrative discretion, and racial bias.

Purchasing privilege? How status cues affect police suspicion in routine traffic stops

ABSTRACT A police officer’s decision to search a driver’s car during a routine traffic stop is based on many variables and indicates that the officer views the driver with suspicion. In this paper, we ask whether driving a luxury-brand car reduces police suspicion during a traffic stop. We find significant reductions in rates of search for minority drivers of luxury cars, though these benefits fade away as the car grows older. We further explore the interactions between personal identity and vehicle type and find powerful effects associated with whether the vehicle indicates occupational status. Our study is based on more than 10 million traffic stops conducted by the Texas Highway Patrol. These findings add status cues to the long list of factors that appear to influence how police treat drivers during routine traffic stops.

Facilitating Interaction with Police During Routine Traffic Stops for Persons with ASD.

A product of ASD socio-communicative-deficits and insufficient police training related to autism, ASD-police interactions have the potential to become problematic, with negative outcomes for ASD individuals and police alike. Thus, the combination of police acclimation, simulatory experience for drivers with ASD and the introduction of Connecticut's recent Blue Envelope could improve overall experiences for drivers. A simulated routine traffic stop practice event, utilising pre- and post- measures was conducted in an effort to quantify drivers' feelings about current and future interactions with police. Our prediction that participants would experience a statistically significant improvement in anxiety, comfort and self-perceived knowledge levels about future ASD-police interactions immediately following the intervention was confirmed. Our prediction that initially significant disparities between participants with and without police experience - those with previous police encounters versus those that don't - would become not significant immediately following the intervention was disconfirmed. While the longitudinal data suggested that improved post-intervention ASD psychological measures remained statistically significant in the long-term, the sample responses to our long-term questionnaire were too few in number to make any definitive conclusions. It is suggested that practice traffic stops such as these could benefit both drivers with ASD and law enforcement nationwide. It is further suggested that police officer curriculums should include additional training regarding special populations.

Does the public want the police to stop, stopping? An experimental look at the impact of outcome data on public perceptions of police discretionary traffic stops

Abstract Objectives To assess whether data on traffic stop outcomes causally impacts public approval of discretionary traffic stops as a crime control strategy. Methods We distributed an original online survey experiment randomly assigning respondents (N = 4740) into either the (1) contraband condition, (2) disparity condition, or (3) the control condition. Results In comparison to the control condition, the contraband condition significantly increased respondent support of discretionary traffic stops as a crime-fighting tool. Black respondents assigned the contraband treatment were significantly more likely to support the use of discretionary traffic stops compared to Black respondents assigned the control condition. Conclusions Although scrutiny exists regarding the efficacy of discretionary traffic stops, public opinion may be shifted if they are provided with information on the outcomes of such stops. Police agencies should consider coupling evidence-based strategies with data on the outcomes of crime control strategies, which may also address community desire for more transparency. Minority civilians may support crime control strategies if presented with data on such strategies’ effectiveness.

The effect of formal de‐policing on police traffic stop behavior and crime: Early evidence from LAPD's policy to restrict discretionary traffic stops

AbstractResearch SummaryOn March 1, 2022, correspondence from the Los Angeles Police Department (LAPD) headquarters stated that officers can no longer use minor infractions (e.g., equipment violations) as a pretextual reason to further investigate drivers for criminal behavior. If LAPD officers are to execute a discretionary stop, they must activate their body‐worn cameras and reasonably articulate to the civilian why they are being investigated. The intent is to reduce racial/ethnic disparities in stops and build trust of the police within the community. Critics of the policy argue that elevated crime rates will result due to the crime suppression effect of such stops. This study examines racial differences in stops before and after the policy change, as well as whether Part 1 violent and property crimes increased. Descriptive findings show that while the counts of stops, arrests, and contraband seizures during stops decreased, the percentage of non‐White civilians stopped decreased only minimally following the intervention. Results from using Bayesian synthetic control methods indicate an increase in both violent and property crimes post‐intervention compared with the synthetic counterfactual. The increase in violent crimes has a low probability of being different from the counterfactual, whereas the increase in property crimes has a high probability, suggesting that the intervention led to a real change in property crimes compared with what would have been expected under the counterfactual.Policy ImplicationsThis study provides preliminary evidence that emerging policies intended to restrict discretionary stops may not have the intended effect on racial disparities in police traffic stops. Instead, agencies passing similar policies may expect to see the potential unintended consequence of a spike in jurisdictional property crimes. We argue that such types of stops across a jurisdiction may have a general deterrent effect on more common and calculative crimes such as property crimes. To combat violent crimes, agencies should consider focusing discretionary stops in known hot spots. Further, our findings offer implications for research on the effects of de‐policing on crime. Police agencies should monitor potential unintended impacts of these policies if enacted and be prepared to deal with such consequences.

Machine Learning Classification Algorithms for Traffic Stops—A Comparative Study

The application of machine learning algorithms across various fields is gaining momentum, and the results increasingly emphasize the need for further testing and implementation. This is driven by the potential to streamline and expedite numerous processes. In this paper, we have employed five algorithms: KNN, Decision Tree, Random Forest, Logistic Regression, and Naive Bayes, and these algorithms have been tested in three large datasets. On average, their performance ranges from a minimum of 80% to a maximum of 90%. Data preprocessing has been completed, and concurrently, we have implemented the SMOTE algorithm to address the challenge of unbalanced data in this research. Simultaneously, the Naïve Bayes algorithm yields the most favorable results of Accuracy, Precision, Recall, and F1 Score, for the “is_arrested” class. Furthermore, to assess the performance of each algorithm, we employed metrics including Accuracy, Precision, Recall, and F1 Score. These metrics allowed us to decide which algorithm achieved the most effective classification.

Pulling back the veil of darkness: A proposed road map to disentangle racial disparities in traffic stops, a research note

AbstractThe veil of darkness (VOD) is a practical and rigorous methodology for examining racial disparities in police traffic stop behavior. Past research, however, has been littered with methodological inconsistencies inhibiting cross‐study comparison and decisions regarding policy. Accordingly, we clarify four aspects of its implementation: 1) coding daylight, our treatment condition; 2) constructing an intertwilight period; 3) accounting for seasonal differences in driving or patrol patterns; and 4) modeling VOD multivariable regression equations. We discuss the theoretical and practical implications of methodological decisions as they pertain to the method's functionality as a natural experiment. Furthermore, we propose a novel weighting procedure to account for seasonal driving population differences. We examined more than 50,000 traffic stops conducted by Michigan State Police during 2021 to demonstrate our suggested framework for future analyses.

Attempting to reduce traffic stop racial disparities: An experimental evaluation of an internal dashboard intervention

AbstractResearch SummaryWe conducted a group randomized‐controlled trial of an internal dashboard system deployed by the Michigan State Police to determine its effectiveness in reducing traffic stop racial disparities. Informed by a difference‐in‐differences design, analyses of traffic stop data from 2019–2022 indicated that the dashboard had no impact on traffic stop racial disparities. Additional analyses of traffic stops, crashes, and crime revealed that the dashboard had no “de‐policing” effect on traffic patrols, nor were there any significant changes in traffic safety or crime in treatment patrol areas relative to control patrol areas. Qualitative analyses of interview data from more than 40 troopers in the agency revealed unique barriers to program implementation and opportunities for future improvement.Policy ImplicationsIn an era of policing where the capacity and demand for data‐driven decision‐making is on the rise, evidence‐based policy and practice can provide police agencies with informed solutions for addressing traffic stop racial disparities. Yet, the increased demand for evidence‐based reform is fueled by a relatively low supply of evidence‐based research. This study adds to this evidence base by providing unique insights into the effectiveness of a program built specifically to reduce racial disparities in traffic stops, while also highlighting implementation challenges.

Race, local context, and traffic stops: an examination of racial profiling in traffic stops in rural and non-rural places

PurposeThe current study examines the effects of race, agency and environment on traffic stops in rural and non-rural spaces.Design/methodology/approachUsing traffic stop data collected in a Midwest US County from January 1, 2020 to December 31, 2021, the current study uses logistic regression to examine racial disparities in traffic stops.FindingsThe results indicate that police decision-making in traffic stops may be influenced by other factors besides a driver’s race or ethnicity. In other words, the police officer’s decision making in a traffic stop varies between small and large agencies as well as rural and non-rural places.Originality/valueThis study provides one of the few examinations of racial disparities in traffic stops in rural places.

Light Him Up: Addressing the Dangerous Intersection of Traffic Stops and Consent

Light Him Up: Addressing the Dangerous Intersection of Traffic Stops and Consent

Education, (re)training, and traffic stops: Felonious law enforcement officer deaths in the United States

Traffic stops continue to be the primary contact between law enforcement and the public, yet little priority is given to traffic stop education and training. A systematic review of felonious traffic stop-related law enforcement officer (LEO) deaths from 1990 to June 2021 revealed the average LEO killed during and after traffic stops was male, in their late 30s, at the rank of officer, with an average tenure of 9.6 years, and killed by a firearm. A synthesis of state and local law enforcement regulations in states with the most officers killed showed the average hours spent in the academy on traffic stop-related training was 20.3 h (3.6%). The overall annual training requirement averaged just 13 h, and only one state required traffic stop-related training, which was once every four years. The lack of education, training, and retraining may precipitate deadly circumstances for a routine law enforcement function.

Associations of cannabis use with motor vehicle crashes and traffic stops among older drivers: AAA LongROAD study

Background Acute cannabis use is associated with a higher risk of motor vehicle crashes (MVC). This study aimed to determine if self-reported past-year cannabis use is associated with MVC or traffic stops among older drivers. Methods This cross-sectional analysis used data from a multi-center study enrolling active drivers aged 65-79 years. Data regarding cannabis use, MVC, and traffic stops (i.e., being pulled over by police, whether ticketed or not) within the previous 12 months were collected through participant interviews. Log-binomial regression models examined associations of past-year cannabis use with MVC and traffic stops, adjusting for site and sociodemographic and mental health characteristics. Results Of 2,095 participating older drivers, 186 (8.88%) used cannabis in the past year but only 10 (<0.5%) within an hour before driving in the last 30 days; 11.41% reported an MVC and 9.45% reported a traffic stop. Past-year cannabis users had a higher prevalence of MVC (adjusted prevalence ratio [aPR] = 1.38; 95%CI: 0.96, 2.00; p = 0.086) and traffic stops (aPR = 1.58; 1.06, 2.35; p = 0.024). Conclusions Past-year cannabis use was associated with increased traffic stops, which are correlated modestly with increased MVC in past studies and may indicate impaired driving performance. We did not find a statistically significant association of past-year cannabis use with MVC, which may indicate limited sustained effects on driving performance from periodic use among older adults, who report rarely driving immediately after use.

The influence of traffic stop policy changes on assaults against officers: A quasi-experimental approach

AbstractThis study investigates whether the Fayetteville (NC) Police Department’s (FPD) prioritization of safety traffic stops and de-prioritization of investigatory stops from 2013 to 2016 impacted assaults on the police officers. Time series models found a statistically significant decrease in assaults on officers within the FPD following their traffic stop policy changes. The comparison police departments did not experience a significant effect of assaults on police officers during the same timeframe. Although the FPD increased the aggregate number of traffic stops from 2013 to 2016, their prioritization of safety stops, and subsequent de-prioritization of investigatory traffic stops resulted in a reduction in assaults on police officers. Since investigatory traffic stops are agued to be more confrontational and have relatively low ‘hit rates,’ de-prioritizing (yet not completely withdrawing from) such type of stops may impact assaults on officers in the aggregate. Several policy implications are discussed.

Social Biases in Traffic Stops in the United States from 2016: An Empirical Application to 'Springfield'

Social Biases in Traffic Stops in the United States from 2016: An Empirical Application to 'Springfield'

A Replication of “Do Female Officers Police Differently? Evidence from Traffic Stops” (American Journal of Political Science, 2021)

A Replication of “Do Female Officers Police Differently? Evidence from Traffic Stops” (American Journal of Political Science, 2021)

The Road to Driving Equality: A Blueprint for Cities to Reduce Traffic Stops

The Road to Driving Equality: A Blueprint for Cities to Reduce Traffic Stops

The Effect of COVID-19 on Police Activities: Traffic Stops, Arrests, and Use of Force

COVID-19 may affect police activities (i.e., traffic stops, arrests, and use of force) due to public compliance with COVID-19 restrictions, changes in individuals’ daily activities, and health threats posed by COVID-19. The purpose of this research is to examine the effects of COVID-19 on police activities and the trends in these activities in Burlington, Vermont. The data that measured arrests, traffic stops, and use of force were obtained from the Burlington Police Department website and covered the period from December 10, 2018 through June 14, 2021. Then, the daily time-series data for arrest rate, use of force rate, and traffic stop rate were created (N = 918; 459 days pre-COVID-19 and 459 days during COVID-19). The results showed that COVID-19 statistically significantly decreased traffic stops and arrests, but not use of force on the day immediately following the implementation of the COVID-19 restrictions and then remained stable over time. However, no significant change was detected in the trends of arrests, traffic stops, and use of force before and during the COVID-19 period. COVID-19 had a significant immediate and lasting effect on traffic stops and arrests, but not use of force.

Can Light Contact with the Police Motivate Political Participation? Evidence from Traffic Stops

AbstractHarsh, highly intrusive, personal contact with the criminal justice system has been shown to politically demobilize, but it is unclear whether less intrusive forms of police contact have any political effects. As the modal type of involuntary police–citizen contact is less invasive and more routine (e.g., a traffic stop), it is critical to understand the ramifications of lighter forms of contact. We argue that, unlike harsh police contact, light, personal, police contact can mobilize individuals, under certain circumstances. When a negative encounter with the police—even if it is minor—runs counter to prior expectations, people experiencing the contact are mobilized to take political action. Using 3 years of observational data and an original survey experiment, we demonstrate that individuals who receive tickets or are stopped by the police are more likely to participate in politics. These effects are most pronounced for individuals with positive evaluations of the police, often White respondents.