Amid a national conversation on race and justice, many law enforcement agencies in the United States are facing increased scrutiny, both internal and external, related to potential racial bias in their policing activities. But bias is not always easy to quantify. The discussion of racial bias in policing must be grounded in scientific evidence, which means gathering data on officers and their interactions with the community.
In 2015, the police department in Durham, North Carolina, took steps to better understand the basis for accusations of racial bias related to their policing practices. The department wanted to examine traffic stops, one of the most emblematic and visible types of police-community interactions. The department turned to RTI to use scientifically rigorous methods to look for patterns of racial disproportionality in its traffic stops.
We analyzed more than 150,000 traffic stops conducted by the Durham Police Department during a six-year period—focusing on stops that occurred during the intertwilight period, roughly between 5 p.m. and 9 p.m., when it is light during some months but dark during other months.
Our results uncovered a pattern of racial disproportionality. Specifically, we found that when the police stopped male drivers, the odds of the driver being black were 20 percent higher during daylight, when the race and sex of the driver are easier to see.
The DPD is using this information to facilitate discussions on race and policing with officers and the community and has taken steps to address evidence of racial disproportionality.
To extend the benefits of our work on this project to other communities, we made the software developed to conduct the analysis a publicly available tool— RTI-STAR (for Statistical Traffic Analysis Report).
Automating the Veil of Darkness Method and Correlating Key Results
Launched in 2016, RTI-STAR uses a peer-reviewed, scientifically sound method to identify racial disproportionality. Known as the “veil of darkness,” this method is based on the assumption that police officers are less able to determine the race of a motorist when it is dark out. Evidence of racial disproportionality exists if minority drivers are more likely to be stopped during the lighter periods of the intertwilight period, compared to darker periods.
It may seem simple, but the veil of darkness analysis can be difficult and time-consuming to conduct. Traffic stop records must be correlated with the time of sunset and civil twilight on the day they occurred. The data management and analysis this approach requires puts it out of the reach of many police departments, cities, and others who would like to analyze traffic stops. By automating the data processing and analysis, RTI-STAR bridges the resource barrier that would otherwise prevent agencies and stakeholders from undertaking this analysis.
Users can also explore results in different ways, including the unit assignment of the officer conducting the traffic stop and the sex of the person being stopped. This capability was key to some of our findings about the Durham Police Department. In Durham, we found that racial disproportionality in traffic stops was confined to male drivers and that officers from a unit focused on drugs and gang violence showed the greatest level of disproportionality.
Grounding a National Debate in Statistically Sound Evidence
Before RTI-STAR made its public debut, we used it to analyze traffic stops in three other large North Carolina cities. We found no evidence of racial disproportionality in traffic stops conducted by the police departments in Raleigh, Fayetteville, or Greensboro. Using open-records requests from other cities, we continue to test and revise the tool.
Undoubtedly, the tool’s greatest impact will be in facilitating the ongoing national conversation about law enforcement and race. We believe transparency is fundamental to informed discussion and progress. With RTI-STAR, police departments, governments, researchers, community leaders, and citizens now have a free, independent source for scientific analysis of the data behind the debate.
