According to a recent study, non-white drivers in Ann Arbor are stopped and searched more frequently than white drivers in routine traffic stops.
The study was conducted by the Southeast Michigan Criminal Justice Policy Research Project (SMART) at Eastern Michigan University (EMU) in collaboration with the Ann Arbor Police Department (AAPD) and Ann Arbor’s Independent Community Police Oversight Commission (ICPOC). It analyzed AAPD traffic stops between January 1, 2017, and December 31, 2019.
Dr. Kevin Karpiak of EMU, who spearheaded the study, says it revealed “statistically significant disparities across every dimension examined.”
The largest disparities were found among multi-racial and African-American male drivers, who were 2.41 times more likely to be stopped than expected for stops initiated for equipment violations. The study’s “expected” number for traffic stops was based on the percentage at which each race was involved in traffic collisions in Ann Arbor, a number that should be independent of individual police officers’ judgments.
Eastern Michigan University Professor of Sociology, Anthropology, and Criminology Kevin Karpiak.
The disparities were not consistent across racial categories. For example, Middle Eastern male drivers were stopped 81% more often than would be expected, Karpiak’s team noted.
In the report, Karpiak and his team write, “every racial group except [w]hite motorists showed some evidence of being the target of police intervention at a higher rate than would be expected.”
“I’m not surprised that we found disparities,” Karpiak says. “I think when you’re doing these kinds of analyses in a racist society … any kind of analysis like this is going to involve disparities. So the question was not really, in the end, ‘Are we going to find disparities or not?’ The question for us was always, ‘How big are the disparities we’re finding?'”
AAPD did not respond to multiple requests for comment on this article.
How the numbers work
Karpiak’s study shows that 15.5% of AAPD’s traffic stops involve African-American drivers.
“The question is, is that a good number or not?” Karpiak says. “The answer is trickier than you might think.”
To solve that problem, Karpiak says, most people — including some researchers — might consult census data to compare the percentage of African-Americans currently living in Ann Arbor to the number stopped by police. According to census data, 6.7% of Ann Arbor’s population is Black or African-American.
“That’s actually not a good indicator,” Karpiak says, “[because] the people driving through Ann Arbor are not the people who live in Ann Arbor.”
Instead, Karpiak’s team looked at collision data, or “the frequency of people that are in collisions in Ann Arbor.”
“So what [is] the percentage of African-Americans that are in collisions in Ann Arbor versus the percentage of African-Americans that are in traffic stops?” Karpiak says.
Research analyst Sara Srygley.
Karpiak’s team found that while African-Americans are involved in 15.5% of traffic stops in Ann Arbor, they are involved in only 11.9% of collisions.
Those two numbers are then used to create an odds ratio “in order to get a sense of whether those [numbers] are in proportion or not,” Karpiak says.
According to Karpiak, “An odds ratio of exactly one means that the ratio of the people stopped is exactly what you would expect based on their representation in the population of people in collisions. An odds ratio of more than one suggests that they’re stopped more often [than you would expect]. An odds ratio of less than one suggests less often [than you would expect].”
Among African-American men, Karpiak’s team found an odds ratio of 1.48 across all categories of traffic stops in Ann Arbor.
But “one thing that’s unique about this study and stands out compared to other studies of Ann Arbor police traffic data in the past … is that we were able to look at gender as well,” says Sara Srygley, a research analyst on Karpiak’s team.
In other words, Karpiak’s team found disparities in gender as well as race. While African-Americans overall were involved in 15.5% of traffic stops, African-American women accounted for only 6.3% with an odds ratio of 1.11 compared to their involvement in collisions.
Problems in data tracking and transparency
Karpiak and his team encountered numerous problems as they conducted their study.
“I think what surprised me most was actually … some of the race categories,” Srygley says. “There’s no race category on your driver’s license. They don’t ask you what your racial identity is.”
Since drivers stopped by the police are not asked to identify their race, Srygley says, “the officer to some degree makes a judgment call on what race they believe the motorist to be” during the data collection process.
The racial categories used in Karpiak’s report, therefore, are the same as those used by AAPD. That’s problematic, Karpiak points out, because “they’re not the same categories as, for example, [those] that are on the census.”
In fact, Karpiak adds, “it’s become clear that no one really knows where the specific categories that Ann Arbor uses came from.”
Police officers have the option of categorizing drivers, at their own discretion, as follows: African-American, Asian, Hispanic, Middle Eastern, multi-racial, Native American, Pacific Islander, or white.
“When a police officer stops someone and has these categories that they can choose from, what makes them pick multi-racial versus [African-American] or … Hispanic or Middle Eastern?” Karpiak asks. “I don’t know — and police leadership doesn’t know — how people are making those choices.”
But AAPD’s problems in tracking data are even more extensive than that, Karpiak’s study suggests.
Ann Arbor Police cars at Larcom City Hall.
Since the ’60s, Karpiak says, AAPD has held a contract with an Oakland County vendor known as CLEMIS, which “was created for police to share data in real time with each other,” Karpiak says.
As part of their data input process, officers are required to report the outcome of each traffic stop — for example, a verbal warning, a citation, or an arrest.
But, Karpiak says, CLEMIS only allows officers to select one outcome per traffic stop.
“The problem with that is that in many stops — one might even say most stops — there’s actually multiple outcomes,” he says. “So you could have a single stop that ends in three verbal warnings, a citation, and an arrest.”
Karpiak says that at the very least, only being able to report one outcome has resulted in data being vastly under-reported.
“That made the data, as far as we were concerned, basically unusable for the purpose of analysis,” he says.
Moreover, Karpiak adds, “in discussions with Ann Arbor leadership, we found that there’s not actually even a set policy on how that one outcome is entered.”
In other words, if a police officer is involved in a traffic stop that results in a verbal warning, a citation, and an arrest, that outcome could be entered in the system simply as a verbal warning.
Moving forward
At the end of their report, Karpiak and his team have included a series of recommendations related to data management and policy development.
One recommendation involves reassessing racial categories so that they more closely align with those in the U.S. census.
“I think we’re already going some places,” Karpiak says. “As part of this process in itself, there were significant moves in sharing this data in a more transparent and regular way.”
To read the full SMART study, click here.
Natalia Holtzman is a freelance writer based in Ann Arbor. Her work has appeared in publications such as the Minneapolis Star Tribune, the Los Angeles Review of Books, Literary Hub, The Millions, and others.
All photos by Doug Coombe.