Driving While Black in Columbia
Annual Vehicle Stops Reports (VSRs) from the Missouri attorney general’s office have prompted heated discussions at Columbia, Missouri, city council meetings, community forums, and other venues for several years. However, a new report using advanced statistical methods from MU economics Professor Jeffrey Milyo finds “there is no statistically significant evidence of systemic racial profiling in recent traffic stops and searches by the Columbia Police Department, nor any evidence that racial profiling in stops and searches is on the increase.”
Milyo says he has been interested in this topic for a number of years and has used the VSR in the classroom setting to talk about what citizens can and cannot learn from this kind of data. He hopes the data presented in his report will foster thoughtful conversations and discussions about disparities and discrimination in the region.
“As an educator, it’s a particularly apt example of data that is easily misunderstood, but with appropriate statistical tools, you can learn more from the data,” he says. “The VSR is pretty clear in that it says this report is just a measure of disparities, and it’s meant to start conversations. It’s not an attempt to test for discrimination, but people tend to read a lot more into what the report says about disparities.”
Milyo says much of the confusion surrounding the annual VSRs stems from the fact that the terms “racial disparities” and “racial discrimination” are interchangeable in many people’s minds. In this context, disparity refers only to a difference in raw descriptive statistics, while discrimination is to act toward someone with partiality or prejudice. Milyo’s report states that, “Given the sensitive and sometimes heated nature of the discussion over the existence and extent of racial bias, it is understandable if the fundamental distinction between disparities and discrimination is sometimes lost in the fray.” He says there are many reasons why disparities exist in social and economic indicators, and, in general, minority populations tend to score worse on most indicators. Milyo cites a history of racial segregation, differences in educational and employment opportunities, and possible bias in the criminal justice system as underlying factors in racial disparities, but he says to use those types of social indicators as a way to evaluate local police might not make logical sense.
“That’s going to be unsatisfying to a lot of people for whom these disparities raise red flags, and it might sound like it’s giving up and saying you can’t learn anything, but you can,” Milyo says. “Over the last 18 years, the way in which social scientists analyze disparities and look for discrimination has advanced quite a bit, so in this report I just applied some of those tools to the data for Columbia.”
In addition to analyzing the data from the annual VSR, Milyo also reviewed a wealth of incident-level data on vehicle stops from the Columbia Police Department, which he calls “particularly detailed and useful for analyzing the existence and extent of racial bias in policing.” He then conducted several tests for racial bias in both stops and searches, including the “Veil of Darkness” method, which compares the probability that a stopped driver is black in daylight or in darkness, controlling for other factors. Milyo says racial profiling in vehicle stops requires that police officers identify the race of the drivers, which he says is more easily done in daytime than in nighttime. Therefore, controlling for other variables, he says fewer black drivers would be stopped at night if police were engaging in racial profiling, all else constant, ---- but his research finds no consistent evidence of profiling.
Changing the Conversation
“Overall,” the report states, “there is no evidence of racial bias against black drivers in 2016–17 based on an extensive examination of vehicle stops and post-stop outcomes. In contrast, there is evidence consistent with racial profiling in vehicle stops for 2014–15.” Still, he says, the evidence for the years 2015 and prior is mixed, and his interpretation is that the earlier indicators do not indicate widespread, systemic bias.
Milyo’s report includes a number of recommendations for policymakers. Number one, he says, is that researchers should use statistical tools to try to test for bias and to monitor police behavior from the VSR. The other main takeaway, he says, is that the public discussion has been focused on disparities, which he calls an unproductive discussion because most people misunderstand the crucial difference between the presence of racial disparities and real evidence of bias in policing.
“I think there has been some acrimony and talking past each other, and part of that is people looking at ambiguous data with different prior beliefs about what they are going to see in this data,” Milyo says. “By providing a more objective and arms-length analysis, I would hope that this might create some common ground and show how we can learn more from all of this data that is being collected. My hope is that it would make the conversation more productive and less a source of mistrust between citizens and the police.”