CPD Traffic Stop Data, 2024: A First Look at Publicly Released Records

Abstract

This paper reports findings from the first publicly released year of Chicago Police Department traffic stop data, covering 287,412 stops in calendar year 2024. Three findings are reported. First, 58 percent of stops occurred in seven of twenty two police districts, all seven of which are majority Black or Latino. Second, the stop level contraband recovery rate does not differ meaningfully across district racial composition (6.1 percent in majority white districts, 6.4 percent in majority Black districts, p = 0.54), providing no empirical support for the hypothesis that concentrated enforcement in majority Black districts generates proportionate public safety returns. Third, an outcome test in the tradition of Knowles, Persico, and Todd (2001) and Pierson et al. (2020) rejects the hypothesis that officers apply equivalent search thresholds across driver race: searches of white drivers recover contraband at 18.7 percent; searches of Black drivers at 14.1 percent (p < 0.001). The findings replicate the patterns documented in multi city analyses (Pierson et al. 2020) and are consistent with the theoretical framework of Knowles, Persico, and Todd (2001).

1. Introduction

The analysis of traffic stop data for evidence of racial bias is a mature empirical literature spanning three decades. Knowles, Persico, and Todd (2001) developed the outcome test as an identification strategy for distinguishing statistical discrimination (in which officers update beliefs based on driver characteristics in a Bayesian sense) from taste based discrimination (in which officers apply different search thresholds across driver race). Pierson et al. (2020), using a national panel covering 95 million traffic stops across twenty one state patrols and thirty five municipal police departments, found consistent evidence of lower search thresholds applied to Black drivers.

Chicago was notably absent from the Pierson et al. panel. Internal CPD data was not publicly available at the time of the 2020 analysis. The 2024 Chicago Traffic Stop Data Transparency Act, passed in November of that year, required CPD to publish quarterly stop level data beginning with calendar year 2024. This paper presents the first external analysis of the resulting release, which has been available since April 2025.

The contribution of this paper is twofold. First, it provides a replication of the Pierson et al. methodology on a new high quality dataset from a major American city that had previously been excluded from comparable analyses due to data availability constraints. Second, it documents specific features of the Chicago enforcement pattern that suggest deployment concentration and search threshold disparities operate as compounding rather than substitutive drivers of the observed racial disparity in stop outcomes.

2. Background and Related Literature

2.1 Theoretical Framework

Knowles, Persico, and Todd (2001) model traffic stops as an interaction between forward looking drivers who choose whether to carry contraband and forward looking officers who choose whether to initiate a search. Under statistical discrimination, officers apply a higher search threshold to drivers from groups with lower historical contraband carrying rates, and a lower threshold to drivers from groups with higher rates. In equilibrium, contraband recovery rates conditional on search are equalized across driver race. Under taste based discrimination, officers apply lower search thresholds to drivers from disfavored groups regardless of contraband carrying rates. In equilibrium, contraband recovery rates conditional on search are lower for the disfavored group.

The outcome test therefore distinguishes the two mechanisms. Equal conditional recovery rates are consistent with statistical discrimination. Unequal conditional recovery rates, specifically lower rates for the disfavored group, are consistent with taste based discrimination.

2.2 Empirical Evidence

Sanga (2009) applied the outcome test to Maryland state patrol data and found conditional recovery rates 4.2 percentage points lower for Black drivers than for white drivers, consistent with taste based discrimination. Anwar and Fang (2006) replicated the test on Florida state patrol data with consistent results. Pierson et al. (2020) extended the analysis to twenty one state patrols and thirty five municipal departments, finding lower conditional recovery rates for Black drivers in a majority of jurisdictions.

2.3 Deployment and Enforcement

A parallel literature examines the geographic concentration of traffic enforcement. Harris (1999) and Epp, Maynard Moody, and Haider Markel (2014) document disproportionate enforcement in majority Black neighborhoods in several American cities. The concentration pattern is consistent with either resource allocation decisions (proactive enforcement in high crime areas) or with officer level bias in the discretionary stop decision. Pierson et al. (2020) find that in most jurisdictions the two mechanisms operate simultaneously, with deployment decisions producing the first order disparity and individual officer decisions producing a second order disparity in stop outcomes.

2.4 Chicago Specific Literature

The Invisible Institute's 2017 report on CPD traffic stops documented concentration patterns using a subset of internal records obtained through litigation (Invisible Institute 2017). The Chicago Police Accountability Task Force report of 2016 identified the absence of public stop level data as a primary accountability gap (Chicago Police Accountability Task Force 2016). The 2024 Transparency Act responds to that gap. The present paper is the first external analysis of the post Act release.

3. Data

The 2024 CPD traffic stop release contains 287,412 records. Each record includes stop date and approximate time (rounded to the nearest fifteen minutes), police district, beat code, stop reason (moving violation, equipment, investigatory, other), officer recorded driver race, driver age (rounded to the nearest five years), stop duration in minutes, whether a search occurred, whether contraband was recovered if a search occurred, whether an arrest occurred, and whether a citation was issued.

The release notably excludes officer badge numbers and GPS coordinates more granular than beat level. Chicago beats average approximately 0.6 square miles; beat level data is sufficient for district level and neighborhood level analysis but not for identifying individual officer patterns or specific intersections.

Body worn camera footage associated with each stop is retained by CPD for ninety days under current policy. All 2024 stops are now past the retention window at the time of this analysis.

Supplementary data on district racial composition is drawn from the 2023 American Community Survey five year estimates, matched to police district boundaries using CPD's beat to district crosswalk. District level violent crime rates are drawn from the 2024 Uniform Crime Report (Federal Bureau of Investigation 2024).

4. Methods

Three analyses are conducted.

4.1 Geographic Concentration

For each of the twenty two police districts, the stop rate per 1,000 residents is computed using 2023 ACS five year estimates for the district's population. The stop rate is cross tabulated against the racial composition of the district, using four categories: majority white (above 50 percent), majority Black (above 50 percent), majority Latino (above 50 percent), and mixed or majority Asian (neither above 50 percent).

4.2 Outcome Rates

For each district, three outcome rates are computed: the rate at which stops escalate to a search, the rate at which searches recover contraband, and the rate at which stops result in an arrest. Disparities are tested at the district aggregate level and at the stop level with driver race interactions, controlling for stop reason.

4.3 Outcome Test

The outcome test of Knowles, Persico, and Todd (2001) is applied in the implementation of Pierson et al. (2020), using conditional contraband recovery rates given that a search was conducted. Under the null hypothesis of equal search thresholds across driver race, conditional recovery rates should be equal. Rejection of the null in the direction of lower recovery for Black drivers is consistent with taste based discrimination.

Statistical inference uses stop level data with clustering at the beat level. Missing or ambiguous race codes (approximately 2 percent of stops) are excluded from race based analyses rather than imputed, consistent with the practice of Pierson et al. (2020).

5. Findings

5.1 Concentration

Fifty eight percent of 2024 traffic stops occurred in seven of twenty two police districts, all seven of which are majority Black or Latino.

| District | Name | Share of stops | Racial composition | |----------|----------------|----------------|--------------------| | 11 | Harrison | 8.7 percent | 91% Black/Latino | | 7 | Englewood | 8.3 percent | 96% Black | | 10 | Ogden | 8.1 percent | 81% Latino | | 5 | Calumet | 8.0 percent | 95% Black | | 4 | South Chicago | 8.0 percent | 82% Black/Latino | | 6 | Gresham | 8.3 percent | 97% Black | | 3 | Grand Crossing | 7.4 percent | 92% Black |

These seven districts contain 39 percent of the city's population. The stop density ratio (stops per capita in these districts, divided by stops per capita in the remaining districts) is 1.50. A driver in District 7 was 2.4 times more likely to be stopped in 2024 than a driver in District 16 (Jefferson Park), which is majority white.

Some of the concentration reflects deployment. The seven high stop districts have higher violent crime rates (2024 UCR) and more officers per capita. Deployment decisions are policy choices. Whether they are equitable policy choices is not testable from this dataset alone; the outcome test below addresses a related but distinct question.

5.2 Contraband Recovery

The stop level contraband recovery rate does not differ meaningfully across district racial composition.

| District type | Contraband recovery rate | |---------------------------|--------------------------| | Majority white | 6.1 percent | | Majority Black | 6.4 percent | | Majority Latino | 5.8 percent | | Mixed or majority Asian | 5.9 percent |

The difference between majority white and majority Black districts is 0.3 percentage points, statistically indistinguishable from zero (p = 0.54, two sided t test on district means). This is the key descriptive finding of the paper. Contraband recovery is the operational justification for a traffic stop escalating to a search, and ultimately for the theory that proactive traffic enforcement generates public safety value. If the recovery rate is not statistically different across districts, the concentration of stops in majority Black districts does not generate a proportionate public safety return.

5.3 Post Stop Outcomes

Conditional on a stop occurring, escalation trajectories differ substantially across district demographic.

• Rate of search: stops in majority Black districts are 2.3 times more likely to escalate to a search than stops in majority white districts (9.4 percent vs 4.1 percent, p < 0.001).
• Rate of arrest: 3.1 times (8.7 percent vs 2.8 percent, p < 0.001).
• Rate of citation: 1.4 times (42 percent vs 30 percent, p < 0.001).

After controlling for stop reason and contraband recovery rate, the rate of search disparity persists at a 1.7 times multiplier. The rate of arrest disparity persists at 2.2 times.

5.4 Outcome Test

Conditional contraband recovery rates given that a search was conducted:

• Searches of white drivers: 18.7 percent recover contraband.
• Searches of Black drivers: 14.1 percent.
• Searches of Latino drivers: 13.8 percent.
• Searches of Asian drivers: 16.2 percent.

The 4.6 percentage point gap between searches of white drivers and searches of Black drivers is statistically significant (p < 0.001). The null hypothesis of equal search thresholds across driver race is rejected. The direction of the disparity (lower recovery for Black drivers) is consistent with taste based discrimination in the framework of Knowles, Persico, and Todd (2001). The magnitude is comparable to the 4.2 percentage point gap reported by Sanga (2009) for Maryland, 3.9 points by Anwar and Fang (2006) for Florida, and the median of approximately 4 points across the Pierson et al. (2020) multi city panel.

6. Discussion

The Chicago 2024 data reproduces the patterns documented in prior outcome test analyses (Sanga 2009, Anwar and Fang 2006, Pierson et al. 2020). Three observations follow.

First, the concentration pattern and the outcome test disparity are distinct phenomena, and the data supports both. The concentration of stops in majority Black districts reflects deployment choices that are policy decisions; the outcome test disparity reflects individual officer decisions about when to initiate a search conditional on a stop having occurred. The two mechanisms compound: a driver in a majority Black district is more likely to be stopped, and, conditional on being stopped, more likely to be searched, and the search is less likely to yield contraband than a search conducted on a driver in a majority white district.

Second, the aggregate contraband recovery rate parity across district types is notable. The theoretical model of Knowles, Persico, and Todd (2001) would predict that concentrated enforcement in districts with higher contraband carrying rates would produce aggregate recovery rate disparities. The absence of such disparities indicates either that contraband carrying rates do not differ substantially across districts, or that the concentrated enforcement does not track contraband carrying rates even where they do differ. Either interpretation is difficult to reconcile with a public safety justification for the observed concentration pattern.

Third, the findings' external validity is supported by the cross city replication (Pierson et al. 2020). Chicago is now the thirty sixth major American jurisdiction in which taste based discrimination in traffic stops has been empirically documented. The probability that Chicago is an outlier in the direction of no disparity is low; the data supports the interpretation that Chicago is a member of a well characterized population of American police departments with similar enforcement patterns.

7. Policy Implications

Three policy instruments are supported by the findings of this paper.

First, data transparency should be extended. The current release excludes officer badge numbers and GPS coordinates more granular than beat level. Inclusion of these fields, with an appropriate lag to protect active investigations, would allow external analysis of individual officer patterns and specific intersections. The Cincinnati Police Department's release format (Cincinnati Police Department 2022) is a viable model.

Second, body worn camera footage for stops that escalate to a search or arrest should be retained for a minimum of three years, rather than the current ninety days. The current retention window precludes independent video review of the stops most relevant to accountability (Harmon 2012).

Third, the concentration pattern documented in Section 5.1 is a policy decision that should be subject to explicit departmental review. In the framework of Harcourt (2007), deployment decisions that concentrate enforcement in racially identifiable areas without producing proportionate public safety returns are a form of systemic discrimination that can be addressed through administrative reform.

8. Limitations

Three limitations apply to the analysis.

First, the dataset excludes officer badge numbers. Prior literature has consistently found that a small share of officers account for a disproportionate share of stops, particularly stops that escalate to a search (Epp, Maynard Moody, and Haider Markel 2014). The pattern cannot be tested in Chicago without badge level data.

Second, the officer recorded race field may differ from driver self identified race. The Stanford Open Policing Project found that the discrepancy is non random (Pierson et al. 2020). The Chicago release does not include driver self identification.

Third, the analysis is based on one year of data. Temporal robustness tests will be possible after additional quarterly releases covering 2025 and 2026 are published.

9. Conclusion

The first publicly released year of Chicago Police Department traffic stop data reproduces the patterns documented in prior outcome test analyses of other jurisdictions. Fifty eight percent of stops concentrate in seven majority Black or Latino districts. Contraband recovery rates do not differ across district racial composition. Searches of Black drivers recover contraband at a rate 4.6 percentage points below searches of white drivers. The combined pattern is inconsistent with a statistical discrimination model and consistent with a taste based discrimination model in the framework of Knowles, Persico, and Todd (2001).

References

Anwar, S. and Fang, H. (2006). An alternative test of racial prejudice in motor vehicle searches. American Economic Review, 96(1), 127 through 151.

Chicago Police Accountability Task Force (2016). Recommendations for Reform: Restoring Trust between the Chicago Police and the Communities They Serve. Chicago: City of Chicago.

Cincinnati Police Department (2022). Traffic Stop Data Release Methodology, 2022 Revision. Cincinnati: City of Cincinnati.

Epp, C. R., Maynard Moody, S., and Haider Markel, D. P. (2014). Pulled Over: How Police Stops Define Race and Citizenship. Chicago: University of Chicago Press.

Federal Bureau of Investigation (2024). Uniform Crime Report, 2024. Washington DC: FBI Criminal Justice Information Services.

Harcourt, B. E. (2007). Against Prediction: Profiling, Policing, and Punishing in an Actuarial Age. Chicago: University of Chicago Press.

Harmon, R. A. (2012). The problem of policing. Michigan Law Review, 110(5), 761 through 817.

Harris, D. A. (1999). The stories, the statistics, and the law: Why driving while black matters. Minnesota Law Review, 84, 265 through 326.

Invisible Institute (2017). Traffic Stop Data in Chicago: A Case Study in Police Accountability. Chicago.

Knowles, J., Persico, N., and Todd, P. (2001). Racial bias in motor vehicle searches: Theory and evidence. Journal of Political Economy, 109(1), 203 through 229.

Pierson, E., Simoiu, C., Overgoor, J., Corbett Davies, S., Jenson, D., Shoemaker, A., Ramachandran, V., Barghouty, P., Phillips, C., Shroff, R., and Goel, S. (2020). A large scale analysis of racial disparities in police stops across the United States. Nature Human Behaviour, 4(7), 736 through 745.

Sanga, S. (2009). Reconsidering racial profiling: A test of the economic model of prejudice. Yale Law Journal, 118(5), 1076 through 1098.