This report summarizes the findings of a study conducted using data collected by the Louisville Division of Police between January 15, 2001 and December 31, 2001. These data resulted from 48,586 interactions between law enforcement officers and citizens during traffic-related contacts.
Information was collected about the driver, the officer, and the stop event. Driver demographics included race, sex, age, residency, license number, and vehicle registration. The only information collected about the officer was officer badge number. Finally, data collected about the stop event include the date, time of day, reason for stop, activities during the stop, number of passengers, and stop outcome.
Data analysis was conducted with the aid of SPSS-11.0 (Statistical Package for the Social Sciences). Analyses were conducted on two levels. First, descriptive analysis, using percentages, summarized stop patterns, stop characteristics, and driver demographics. This information is useful only to describe the existing state of affairs (what is), but not to explain them (why) or to formulate predictions about future events (what if). To address the complex relationships that exist among different variables, a program called chi-square automatic interaction detector or CHAID was used to evaluate the variables in terms of their relationships with one another (multivariate analysis).
Temporally, it was not feasible to determine which month was the most active given several problems with the data on this variable. The most active time of day for stops was between 5-6pm, with 7.4% of all stops, followed by the time period from 4-5pm with 6.7% of all stops. Overall, the 2nd shift (3-11pm) was the most active, with 46% of all stops, followed by the 3rd shift (30%), and the 1st shift (24%).
Stopped drivers were mostly white (64%), male (70%), between 24 and 40 years old (46%), and Louisville residents (63%). Drivers were mainly stopped for penal code violations (67%), were checked for outstanding warrants (78%), were not searched (84%), and were issued citations (67%). Drivers who were searched (17%) were searched incident to arrest (52%), and by consent (40%). About 1 in 5 searches (19%) were because of the odor of drugs or alcohol. Contraband was discovered in 31% of searches. In cases where there was a search and contraband was discovered, 74% resulted in an arrest.
The descriptive analysis indicated some slight percentage differences among the races in certain events (e.g., stopped for equipment/registration violations). These percentage differences, however, cannot be used to infer correlation or causation (racial profiling). To make these types of inferences, multivariate analyses using CHAID were conducted. CHAID segments the sample of traffic stops and reveals the interrelationship between the potential predictors and the events involved in the stop. The CHAID procedure generates a decision tree that identifies significant predictors of each decision in question. In effect, the procedure cross-references each event with each potential predictor.
Results from CHAID analyses resulted in five events (violation of the penal code, being asked to exit, being searched, being subject to a warrant check, and being arrested) with significant predictors. Being stopped for a penal code violation was significantly related to the race of the driver; other persons of color (72%) and whites (69%) were most likely to be stopped for this reason. Age, however, had a strong interactive effect with race. Being asked to exit, being searched, being subject to a warrant check, and being arrested all were predicted by being stopped for a misdemeanor. Driver sex also surfaced as a predictor in some situations.
These data provide no empirical evidence that the LPD is systematically engaging in discriminatory stop practices. In general, stops conducted by the department, as a whole, during the study period, do not involve the race of the driver as a significant factor related to events and outcomes. The only exception to this involves stops for penal code violations, where other persons of color and whites were most likely stopped for this reason. These types of stops involve fairly low levels of officer discretion given that penal code violations are more serious than other reasons for which a driver might be stopped. This does not mean, however, that no individual citizen ever experienced discrimination. It is always possible that individual officers may engage in racially biased practices, both in determining which drivers they will or will not stop and in determining what steps to take after the initial contact. To detect discriminatory practices at this level, however, requires constant vigilance by the community, by all the officers within the department, and by the departmental administration. Statistical analysis, while valuable, cannot substitute for community involvement and effective management.
The full report provides a discussion of the baseline dilemmaand makes recommendations for continued study to obtain a full year of data.
Edwards, T. D., Grossi, E. L., Vito, G. F., & West, A. D. (2002). Traffic Stop Practices of the Louisville Police Department: January 15-December 31, 2001.