For a few days in April 2001, it was hard to miss news accounts about riots in Cincinnati. Daily stories outlined how police officers chased an unarmed 19-year-old black man named Timothy Thomas. It was past midnight in the Over-The-Rhine neighborhood. A dispatcher called in officers to help pursue Thomas. The voice over the police radio offered a few details: a brief description of Thomas, and that he had 14 outstanding warrants.
An officer responding to the call spotted Thomas. He got out of his car and drew his gun, fired and killed Thomas. Cincinnati rioters protested the police shooting, capturing headlines across the nation. Fellow Dateline NBC producer Jason Samuels and I were asked to look into the circumstances surrounding these events. After collecting traffic ticketing data and going through court records, I came back to the dispatcher's brief call on the radio: 14 outstanding warrants.
How does a 19-year-old rack up so many warrants? That question led us on a lengthy investigation that culminated with an hour-long report called "A Pattern of Suspicion." With correspondent John Larson, we examined disparities in how police across the United States deal with blacks and whites. Despite its national scope, the story focused on something simple: a traffic ticket. That's because Thomas' 14 outstanding warrants all stemmed from traffic tickets.
Almost all were for the tickets themselves, and the last two were for fleeing from officers for not paying the traffic tickets. There was also something intriguing about the kinds of traffic tickets that Thomas got. They were not for the usual things, like speeding or running a stop sign. Almost every citation was for an offense that is difficult or almost impossible for an officer to know about without first stopping someone.
The list includes driving with a passenger not wearing a seatbelt and driving without a license. Only one ticket was for a moving violation. Police, prosecutors and criminologists have another way of looking at these tickets. Some, like speeding far above the speed limit or running a stop sign, are also known as low-discretion tickets. These are moving violations, and the offense is obvious even at a distance.
Police don't have to make too many decisions before chasing those drivers. Then there high-discretion stops for driving without a seatbelt or having windows tinted too dark. Notwithstanding special crackdowns, these non-moving violations are harder to detect, and are often ignored by officers. These are usually ignored unless an officer is looking for something else. In the war on drugs, police and others told us it's perfectly legal for an officer to pull over drivers for these high-discretion violations in the hope of catching them at something worse.
Advocates charge this is essentially racial profiling, violating the equal protection clause of the 14th Amendment and also inflaming already strained relations between police and minority communities. With Cincinnati's ticketing database, we were able to see Thomas was not alone in getting high-discretion non-moving violations. We obtained the data, with more than 300,000 entries, through a state open records request.
It included fields such as the time, street, violation, the driver's name and demographic information like race. By calculating odds ratios, we could see that blacks were at least three times as likely as whites to get cited for driving without a seatbelt or not having a driver's license. This kind of ratio is known as an internal benchmark. It relies on information just within the ticketing database.
You compare the rate for which blacks versus whites get a high-discretion ticket, like driving without a seatbelt, and divide that by the rate for which blacks versus whites get low-discretion tickets like speeding. For those interested in learning more about odds ratios, Sage publishes several handy books, and Phil Meyer includes a valuable critique in his latest edition of "Precision Journalism." One advantage of internal benchmarks is that it is an accepted standard in criminal justice circles, and it does not involve population estimation.
It is as precise as possible. Comparing ticketing versus overall population is known as an external benchmark, and among police, its soundness is more debatable. We looked at the data other ways. By mapping violations in ArcView, geocoding the locations where tickets were issued, we could see the disparities were not particular to Over-The-Rhine. In predominately white, low-crime neighborhoods, blacks were still ticketed far more often for non-moving violations versus moving violations.
We collected various documents and reports that were useful for starting our data shopping list. A General Accounting Office (now known as the Government Accountability Office) report, among others, outlined which cities with at least 100,000 people across the nation are collecting ticketing and stop data. We sought data from as many city police departments as possible from coast to coast. This included police departments that had never had anyone request the information – that includes journalists as well as advocates and academic researchers.
The problems we encountered are unfortunately typical: places like San Diego took months and months in replying; Denver rejected our initial request but later turned over the data, and St. Louis charged by far the most. In the end, we had more than 4 million traffic tickets and stops in more than a dozen cities. One hurdle we faced was that almost no city kept the data exactly like another. In some places, it was impossible to tease out moving versus non-moving violations; those cities were excluded from our analysis.
Other police departments recorded stops, and noted the general category of the violation, rather than the precise violation code, that led to the stop. In those cases, we could include the data because the categorizations offered at least a window into that department. As can be expected, this kind of categorization was limiting because it prevented us from examining variations within a grouping – when a police department was pulling over blacks at a higher rate for, say, equipment violations, was the rate even more disproportionate for seatbelt offenses?
In any city that included geographic information we mapped out where the violations occurred. We cross-checked both high-crime and low-crime neighborhoods. In the end, we mapped more than 1 million addresses in six cities, a process that accelerated with programming that improved standardization. We performed a logistic regression in SPSS to examine the effects of other demographic data on the disparities involving non-moving violations.
Logistic regression tests the probability of an event, when there are only two possible outcomes. The results are often reported as odds or odds ratios. This underscored how race was a key factor by allowing us to see essentially how the odds of someone being black is cited for these violations, versus the impact of a person's sex or age. We constantly re-checked the data and talked to criminal justice experts.
We gave each police department more than two months to respond to a written summary of our findings. We wanted to ensure that our work was accurate, and that every police agency had an opportunity to talk with us about what was in the numbers. The work we did along the way proved extremely useful in these discussions. When one department cited a Bureau of Justice Statistics survey, we could point out that the questionnaire categorized stops in a way that made it impossible to detect moving versus non-moving violations.
When another department cited a NHTSA study that blacks did not wear seatbelts about a quarter of the time, we noted that the rate was virtually the same for whites. Only one police department offered a geographical argument: Richmond said the ticketing disparity was a result of heavy policing in high-crime neighborhoods, which are often black. When we pointed out that the disparity passed through to low-crime, predominately white sections, it responded with "No comment." For other journalists doing this kind of story, it's important to get the data, talk with sources and get familiar with the subject of ticketing disparities.
Aside from Dateline NBC's investigation, we helped affiliates with local stories. This allowed us to get out more information than what could fit in our piece. We also gave the data and worked with The Denver Post, which has an affiliation with NBC. It published a front-page story on the day we aired. And we posted additional reporting on the Dateline Web site at www.msnbc.msn.com/id/4703833.