Why Do Stop-and-Frisk Rates and Outcomes Differ So Dramatically Across U.S. Cities?

Abrams and Goonetilleke (2026) examined how stop-and-frisk practices vary across major U.S. cities and whether crime, poverty, demographics, or police resources explain those differences. They analyzed more than eight million pedestrian and vehicle stops recorded in 16 cities from 2019 through 2023. They found that stop and frisk rates differed by nearly two orders of magnitude across cities. Contraband discovery rates also varied substantially, ranging from 18% to 74% for pedestrian frisks in 2023, while firearm discovery rates ranged from 2% to 9%. They found no strong, consistent relationships between stop-and-frisk activity and crime, poverty, police staffing, or police spending.

Why This Article Was Selected for The Policy Scientist

Abrams and Goonetilleke’s study contributes to understanding stop and frisk, a historically contested policing practice closely associated with racial disparities, constitutional challenges, and unequal encounters between government institutions and residents. The topic remains timely as jurisdictions reconsider proactive policing and seek evidence about its effects on public safety and administrative accountability. Published in the American Economic Association’s AEA Papers and Proceedings, the article extends the authors’ previous work on police stops and expands research beyond individual-city studies. Its dataset is unusually large, covering more than eight million stops in 16 cities, although inconsistent reporting limits comparisons across jurisdictions. The correlation analysis reveals important patterns but cannot establish causation. Future research using causal inference or randomized controlled trials would strengthen the evidence.

Full Citation and Link to Article

Abrams, D. S., & Goonetilleke, P. (2026). Stop and frisk around the country. AEA Papers and Proceedings, 116, 382–387. https://www.aeaweb.org/articles?id=10.1257%2Fpandp.20261050

Central Research Question

The central research question is: Why do stop-and-frisk practices differ so dramatically across major U.S. cities, and can differences in crime, demographics, poverty, income, police resources, or other city characteristics explain that variation? The study also asks whether cities that conduct more frisks obtain more contraband, including firearms, or whether higher activity is associated with lower productivity. Rather than evaluating one city or one police department, the study places local stop-and-frisk practices in a national comparative framework.

Stop and frisk is a proactive policing tactic. Officers temporarily detain people when they suspect criminal activity and may pat them down when they believe they could be armed. During vehicle stops, officers may search the vehicle when they have probable cause to believe that evidence is present. The policy has been used as a crime-control tool, but it has also generated litigation and sustained debate about racial disparities, constitutional limits, and the proper scope of police discretion.

Previous Literature

Earlier research on stop and frisk has generally concentrated on individual cities, especially New York, Philadelphia, and other jurisdictions with extensive administrative records. That literature has examined racial disparities, the role of officer characteristics, the productivity of searches, and the effects of police deployment. Some studies have asked whether searches are more likely to uncover weapons or contraband when officers search people from particular racial or demographic groups. Other research has examined whether intensified police presence changes crime rates or affects the number of stops.

The authors’ earlier work on police frisks forms part of this developing literature. Their broader research program has focused on measuring how frequently police conduct frisks and how often those frisks produce useful results. The current study builds on that work by assembling comparable information from a much larger number of cities. The article also relates to research showing that police presence can deter crime, while recognizing that evidence is less conclusive about whether specific tactics, such as stop and frisk, produce those effects.

The major gap in previous research is geographic scope. Because police records have historically been collected and stored locally, researchers have had difficulty comparing stop-and-frisk practices across many jurisdictions. As a result, it has been unclear whether the patterns observed in highly studied cities reflect a national pattern or are unusual local cases. This study addresses that gap by comparing cities with different crime rates, demographic compositions, economic conditions, and police organizations.

Data

The study examines more than eight million pedestrian and vehicle stop records from 16 major U.S. cities between 2019 and 2023. The main analysis includes 1,815,760 pedestrian stops and 6,293,349 vehicle stops. The researchers initially sought information from the 20 largest U.S. cities. They obtained at least some data from all but Phoenix, Jacksonville, Indianapolis, Denver, and Oklahoma City. The final sample also includes several large cities outside the original group, including Washington, DC, Milwaukee, Oakland, and Cincinnati.

The records identify whether a stop occurred, whether a frisk or vehicle search followed, and whether officers discovered contraband. Depending on the city, contraband includes firearms, drugs, weapons, stolen property, or other prohibited items. The authors distinguish between gun discoveries and overall contraband discoveries. Eleven cities provide both pedestrian and vehicle data. New York City and Seattle provide pedestrian-stop data only, while Dallas, Houston, and Fort Worth provide vehicle-stop data only.

The researchers combine stop records with city-level information on population, racial composition, poverty, income, police budgets, sworn officers, violent crime, murders, rape, aggravated assault, robbery, vehicle travel, precipitation, and public sentiment toward police. These measures allow them to compare stop-and-frisk activity with conditions that might plausibly influence police policy.

The dataset is a major strength because it covers a large number of stops across several years and regions. However, the records are not perfectly standardized. Cities differ in how they define, record, and report stops and frisks. Some datasets do not provide enough detail to identify firearm discoveries separately. These limitations make exact comparisons more difficult, although the authors argue that reporting differences are too small to explain the full scale of the observed variation.

Methods

The study primarily uses descriptive comparisons and correlation analysis. The authors calculate daily pedestrian and vehicle stop rates per million residents, daily frisk rates, and contraband-hit rates. These measures allow them to compare cities of different population sizes. They also compare changes between 2019 and 2023 to examine whether stop-and-frisk activity changed over time.

To evaluate possible explanations for the differences, the authors calculate Spearman rank correlations between stop-and-frisk measures and city characteristics. This method is appropriate when observations contain extreme differences and when the relationship may not be linear. The analysis considers whether stops and frisks are associated with crime, poverty, income, police staffing, police budgets, racial composition, vehicle travel, weather, or public attitudes toward police.

The method identifies relationships, but it does not establish cause and effect. A correlation between two variables could reflect an omitted factor, reverse causation, measurement error, or local policy differences. The study does not use a randomized controlled trial, a natural experiment, an instrumental-variable strategy, or another design intended to identify causal effects. Consequently, the findings are best interpreted as evidence about patterns and associations rather than proof that a particular factor causes a city to conduct more or fewer stops.

Findings/Size Effects

The most important finding is the extraordinary variation in stop-and-frisk activity. Across cities, per capita stop rates differ by nearly two orders of magnitude. This means that the city with the highest rate may conduct roughly 100 times as many stops per resident as the city with the lowest rate. Vehicle-stop patterns are also highly dispersed. In 2019, Philadelphia recorded approximately 846 vehicle stops per million residents per day, more than ten times the rate in Chicago. Although the median number of vehicle stops declined by nearly half between 2019 and 2023, the highest and lowest cities still differed by about an order of magnitude.

Frisk productivity also varies substantially. In 2023, general contraband-hit rates for pedestrian frisks ranged from approximately 18 percent in Washington, DC, to 74 percent in San Francisco. Thus, the most effective departments recovered contraband in roughly four times as many frisks as the least effective departments. Gun-hit rates were lower, generally ranging from about 2 percent to 9 percent. These figures show that the number of frisks and the probability of finding contraband are separate dimensions of police practice.

The analysis finds no strong or consistent relationship between stop-and-frisk activity and crime rates. Cities with substantially higher violent-crime or murder rates do not consistently conduct more frisks. Poverty, income, and racial composition also do not provide a clear explanation for the differences in city-level activity. Police budgets and staffing levels are similarly weak predictors. Some cities have considerably more officers or greater spending per resident but conduct fewer frisks than cities with fewer resources.

The authors also examine whether stop-and-frisk activity is related to public sentiment toward police. The relationship is inconsistent. Frisk rates are negatively associated with the ratio of positive to negative police-related social-media posts, but stop rates show no comparable relationship. This pattern suggests that the observed association may be spurious rather than evidence that public opinion determines police activity.

The authors consider several possible explanations for the remaining variation. Differences in reporting practices may contribute to the results, but audits and standardized vehicle-stop rules suggest that reporting problems cannot account for the entire gap. Another possibility is that police departments do not systematically learn from successful practices in other cities. A final possibility is that stop and frisk has limited effects on crime, arrests, or contraband recovery, reducing the incentive for cities to converge on a common approach.

Conclusion

The study shows that stop-and-frisk policing is far less uniform across U.S. cities than conventional explanations would suggest. Crime, poverty, demographic composition, police resources, traffic, weather, and public sentiment do not consistently explain why some cities conduct many more stops and frisks than others. The differences are too large to be treated as minor administrative variation.

The study’s principal contribution is empirical: it creates a broad comparative record of more than eight million stops and documents substantial differences in both police activity and search productivity. Its findings are potentially relevant to other large cities, but generalization should be cautious because the sample excludes several major jurisdictions and relies on locally collected records. The analysis also cannot determine whether stop and frisk reduces crime or improves public safety. Future research would benefit from standardized national reporting and causal research designs that compare policy changes, exploit quasi-random deployment decisions, or use randomized experiments where ethically and legally feasible.