African Americans have complained about racial profiling in Detroit's suburbs for generations. They say suburban police unfairly target them for tickets when they venture outside of predominantly black Detroit and into the metro area's mostly white suburbs.
No one had ever been able to prove at least statistically - that Detroit's suburban police racially profile, however. To do it, the Detroit Free Press had to show a pattern of disproportionate ticketing by race - and in the process, take on a recalcitrant judge and 3,000 traffic tickets, among other challenges.
We chose to look at Harper Woods, a suburb that borders Detroit's east side, because of a good tip we had received. Plus, unlike most Detroit suburbs, Harper Woods cops track race on their tickets, making it possible to analyze their racial ticketing pattern.
The first question was how to do it. Harper Woods court data isn't computerized, so we had to manually sort the tickets. With piles of complex information to analyze, we knew we needed to use a database.
Getting the data
With our computer-assisted reporting expert Heather Newman's help, we set up Microsoft Excel spreadsheets with the categories we wanted to track: the driver's name and address, age, race, gender, location of ticket, infraction, time and date, car type and color, fine paid, and the officer's name. We decided 3,000 tickets was a good amount - about six months worth, a respectable sampling.
Beyond race, we wanted to see what the trend was in terms of location. That is, whether drivers coming from Detroit, or driving closer to Detroit, were more likely to be ticketed.
Then there was the question of access. By law, journalists can look at courthouse files without filing a FOIA. But we encountered numerous roadblocks.
They wouldn't allow us to photocopy the tickets to speed up the process, then one day the courthouse administrator kicked us out for talking to a court officer who happened to come in the room. The judge eventually conceded and allowed us back in, but with restrictions that included allowing only one person to look at files at the same time and requiring his workers to review our work. Later, he allowed us to bring in a portable copier. By then we were three weeks behind schedule.
Setting up the database
Meanwhile, Database Specialist Victoria Turk had been setting up the database queries based on a list of initial questions: the racial and gender breakdowns of people ticketed, the most common tickets for African Americans vs. whites, which intersections were most heavily ticketed, etc.
When Turk finally received the raw data, she had about 80 percent of the database queries set up. The results of Turk's ongoing analysis helped dictate some of the Arellano's reporting.
We often had mix-ups in the data, such as accidental duplicates or the wrong spellings of police officers. We also had to be careful when it came to the racial category. Officers check a driver's race based on their presumption, not by asking the individual. So when an officer didn't know, or put "O" for "Other" (they told us that sometimes happens with drivers of Hispanic or Arabic names or appearance, for example) we made sure they were counted that way, so we could track people by race as unambiguously as possible.
Most of the data inconsistencies resulted from the manual input process.
Four different individuals keyed in the data, and although they were working from the same categories and code list for offenses, there were inconsistencies where abbreviations were used for city names, spellings of street names, and virtually every data input item entered from the ticket.
These weren't necessarily typos; the ticketing officers themselves misspelled information or abbreviated creatively, or simply had poor penmanship. In retrospect, some of the data inconsistencies could have been avoided if clear and explicit directions had been given to our data keypunchers at the start of the project.
We also had some problems reaching our total sample goal of 3,000 tickets. Either through the keypunch process, or in the process of being provided the paper tickets from the court, we somehow managed to get some duplicates in the data that needed to be eliminated. If a ticket was inadvertently entered twice, it would affect our analysis to double count it. Once we eliminated the duplicates, we ended up having to return to the court to pick up a few more to meet our sample quota. This could have been avoided by keying in a few extra tickets, in case some turned out to be unusable for one reason or another.
Later, Turk saw that she could perform some calculations more easily in Excel. The calculated fields, as well as the original input data, were then exported to an Access database.
No one had ever been able to prove - at least statistically - that Detroit's suburban police racially profile, however. To do it, the Detroit Free Press had to show a pattern of disproportionate ticketing by race.
The queries performed tried to explore the role that race played in ticketing, if any. We looked at race differences for nearest intersection, offense, average dollar amount and total tickets.
Eventually, we looked at race differences by the ticketing officer. We looked at ticket dollar amounts by car model year to see if older cars were ticketed more because of defective equipment. We also took a look at the most heavily ticketed intersections, time of day, and model years to give our readers a feel for the way the Harper Woods police patrolled.
Traffic study
We had one more hurdle: whether black drivers were being disproportionately ticketed. To find this, we had to figure out the number of black drivers on Harper Woods' roads and to compare it to the ticketing pattern.
To do that, we conducted a traffic study, similar to one done in New Jersey by investigators who looked into the state police's ticketing patterns. Arellano sat at Harper Woods' top ten most ticketed intersections and documented the race of almost 8,000 drivers. A profiling expert, a Temple University professor, advised us how to do it.
It took a few hours every day for about five days. If we couldn't tell what the driver's race was - because of a dirty windshield, or someone whose race wasn't obvious - we didn't count them.
Finally, after three months of off-and-on work, we published the package June 1. The lead 1A story showed that Harper Woods Police are overwhelmingly and disproportionately ticketing African Americans.
Suggestions
If you're thinking about a similar database investigation, here are our suggestions:
Plan ahead with your editors and coworkers, especially when it comes to setting up the database, and give yourself extra time. If you run into access problems, be diplomatic but firm. In this case, only the threat of a good fight would budge the court judge. And of course, know the law. During a difficult project where you and others are encountering many hurdles, encourage and reward people throughout the process. Good guacamole, brownies ever you make well - and small gifts are appreciated and deserved. Fact check the database results as many times as necessary, and by different people. Avoid Harper Woods District Court.