Do Police Officers Discriminate Against Low-Income Motorists During Traffic Stops and Searches?

Feigenberg and Miller (2025) examine whether police officers discriminate against low-income motorists during traffic stops and searches. They analyze more than 11 million traffic stops conducted by the Texas Highway Patrol between 2009 and 2015, linking motorists to measures of household income and vehicle status. They find large class disparities in policing. Motorists in the bottom income quintile were searched in 2.5% of stops, compared with 1.1% for those in the top quintile—more than twice as often. Low-income motorists were also less likely to have contraband discovered when searched and were more likely to be stopped for discretionary infractions associated with pretext stops.

Why This Article Was Selected for The Policy Scientist

This article addresses an important question about whether economic status shapes interactions between citizens and government institutions. Traffic stops are among the most common forms of state-citizen contact, making the topic relevant to public administration, criminal justice, social mobility, and institutional trust. The study is especially timely because discussions of policing have focused heavily on race, while class-based disparities have received far less empirical attention. Feigenberg and Miller have contributed extensively to this literature, and this study builds on their earlier work examining search decisions. The data are exceptionally strong, covering more than 11 million Texas traffic stops. The authors employ a credible causal inference design using within-motorist variation in vehicle status rather than relying solely on conventional regression methods.

Full Citation and Link to Article

Feigenberg, B., & Miller, C. (2025). Class disparities and discrimination in traffic stops and searches. American Economic Journal: Applied Economics. Advance online publication. https://doi.org/10.1257/app.20250397

Central Research Question

Feigenberg and Miller investigate whether police officers treat motorists differently based on their economic class and whether observed class disparities in traffic stops and searches reflect discrimination rather than differences in conduct or criminal activity. Although a large literature has examined racial disparities in policing, considerably less attention has been devoted to socioeconomic status. The authors seek to determine whether low-income motorists are more likely to be stopped and searched and whether those additional searches are justified by higher rates of contraband possession.

A key innovation of the study is its recognition that discrimination may occur at two stages of the enforcement process. Officers first decide whether to stop a motorist and then decide whether to conduct a search. Most previous studies focus only on the search decision among already-stopped motorists. The authors argue that a complete assessment of discrimination must account for both margins. Their central question is therefore whether police officers use perceived economic status when deciding whom to stop and search and whether such decisions improve law enforcement outcomes.

Previous Literature

The article builds upon extensive research examining discrimination throughout the criminal justice system. Prior studies have documented racial disparities in traffic stops, searches, arrests, pretrial detention, sentencing, and police use of force. Influential contributions by Knowles, Persico, and Todd (2001), Anwar and Fang (2006), and Pierson et al. (2020) have focused on whether racial differences in police behavior reflect discrimination or statistical assessments of criminal risk.

Research on class-based disparities has been more limited. Sociologists and criminologists have shown that residents of economically disadvantaged neighborhoods experience greater police presence, more frequent stops, and higher arrest rates. However, these studies generally rely on neighborhood-level measures and cannot determine whether police directly respond to an individual’s socioeconomic status. Neighborhood analyses are also vulnerable to omitted variable concerns because disadvantaged and affluent areas differ along many dimensions.

The authors extend the discrimination literature by focusing explicitly on economic class. They also connect their work to broader research on class-based discrimination in labor markets, housing markets, and consumer interactions. Their contribution is important because it moves beyond documenting disparities and attempts to identify the causal effect of perceived socioeconomic status on police behavior.

Data

The study uses administrative data covering virtually every traffic stop conducted by the Texas Highway Patrol between 2009 and 2015. After applying sample restrictions, the final dataset contains approximately 11 million traffic stops. The data include information on motorists, vehicles, traffic violations, searches, contraband discoveries, stop locations, stop timing, and the identity of the trooper conducting the stop.

A distinctive feature of the dataset is the availability of motorists’ names and addresses. The authors use this information to estimate household income by combining Census block-group data with property assessment records. This procedure allows them to create an individualized measure of economic status rather than relying solely on neighborhood averages.

The researchers further merge the traffic stop records with Texas criminal history data and commercial address-history files. These additions enable them to identify repeat motorists, track individuals across multiple stops, and account for criminal history. The resulting dataset provides an unusually rich portrait of interactions between motorists and law enforcement and represents one of the most comprehensive data sources assembled for studying socioeconomic disparities in policing.

Methods

The authors begin by documenting descriptive relationships between household income and traffic stop outcomes. They estimate how search rates, contraband recovery rates, and the types of infractions leading to stops vary across the income distribution. These analyses control for race, gender, criminal history, location, timing, and trooper characteristics.

The study’s primary contribution, however, is its causal inference strategy. The authors exploit within-motorist variation in the vehicles that individuals drive. Because vehicles convey information about economic status, a person may appear wealthier or poorer depending on the vehicle being used at a particular time. Importantly, the same individual can drive different vehicles across different stops.

The research design compares the same motorist when driving vehicles that signal different levels of socioeconomic status. By holding constant the individual while changing the class signal communicated by the vehicle, the authors isolate the effect of perceived class on police behavior. This approach reduces concerns that observed disparities merely reflect differences between low-income and high-income individuals.

The authors also develop a formal model of stop and search decisions. The model demonstrates how discrimination can arise both when officers decide whether to stop motorists and when they decide whether to conduct searches. To investigate stop decisions, the authors examine whether lower-status vehicles are disproportionately stopped for discretionary violations commonly associated with pretextual policing. To investigate search decisions, they estimate how search probabilities change when the same motorist drives vehicles conveying different socioeconomic signals.

Finally, they employ instrumental-variable methods to estimate the productivity of marginal searches induced by perceived class differences. This allows them to assess whether additional searches generated by class signals actually increase contraband recovery.

Findings/Size Effects

The study reveals substantial socioeconomic disparities in police searches. Motorists in the lowest income quintile are searched in approximately 2.5% of stops, compared with only 1.1% of stops among motorists in the highest income quintile. Consequently, low-income motorists are searched more than twice as frequently as high-income motorists.

The authors estimate that a 100% increase in household income is associated with a 0.37 percentage-point decline in the probability of being searched. Given the relatively low baseline search rate, this represents a substantial effect. Importantly, the relationship remains strong after controlling for numerous demographic, geographic, and criminal history variables.

The evidence also indicates that searches of low-income motorists are less productive. Although officers search these motorists more often, they are less likely to discover contraband when conducting those searches. This finding is inconsistent with a simple efficiency-based explanation in which officers successfully identify individuals who are most likely to possess contraband.

The within-motorist analysis provides even stronger evidence. When the same individual drives a lower-status vehicle, the probability of being searched increases significantly. The authors estimate that switching to a vehicle that doubles signaled income reduces the search probability by approximately 0.53 percentage points, representing a decline of roughly 28% relative to the sample mean search rate.

Evidence also suggests that perceived class affects stop decisions. Motorists driving lower-status vehicles are more likely to be stopped for discretionary infractions that appear consistent with pretextual enforcement. The authors estimate that doubling signaled income reduces the implied stop rate by approximately 7%.

Finally, the instrumental-variable analysis shows that marginal searches induced by lower-status vehicles are less likely to recover contraband than comparable searches involving motorists perceived as wealthier. Reallocating those searches toward higher-income motorists would increase overall contraband yield. The authors also find evidence that low-income defendants are more likely to plead guilty or no contest and less likely to receive dismissals or acquittals, potentially reducing the expected costs associated with stops and searches of economically disadvantaged motorists.

Conclusion

Feigenberg and Miller provide some of the strongest evidence to date that economic class influences police behavior during traffic stops and searches. Using an innovative within-motorist design, they demonstrate that the same individual is treated differently when driving vehicles that signal different levels of socioeconomic status. This approach allows them to move beyond simple correlations and provide evidence consistent with class-based discrimination.

The findings indicate that low-income motorists are more likely to be searched, more likely to be stopped for discretionary infractions, and less likely to possess contraband when searched. These patterns suggest that socioeconomic status affects enforcement decisions independently of actual criminal behavior.

Methodologically, the study represents an important advance in the literature. Its combination of large-scale administrative data, within-individual comparisons, formal modeling, and instrumental-variable analysis provides a credible framework for identifying the effects of perceived class. More broadly, the article expands research on discrimination in criminal justice beyond race and demonstrates that socioeconomic status constitutes an important dimension through which public institutions allocate enforcement resources.