{ "id": "2000-december-check-sheets", "slug": "2000-december-check-sheets", "url": "https://dwillis.github.io/uplink/article/2000-december-check-sheets.html", "year": 2000, "month": "December", "monthNum": 12, "headline": "Check sheets", "kicker": "RACIAL PROFILING", "author_name": "Jim Adams", "author_title": "Reporter", "affiliation": "The Courier-Journal", "summary": "A Louisville news organization investigated whether police were disproportionately stopping African-American drivers by matching police dispatch records with handwritten dispatcher logs containing driver demographic information. The journalists developed a novel data-matching methodology using unit numbers and timestamps to cross-reference traffic stop records with license checks that included racial identifiers.", "keywords": [ "profiling", "investigation", "data" ], "topics": [ "Investigative Methods", "Crime & Justice", "Demographics & Census" ], "technologies": [ "SPSS", "Excel", "floppy disk" ], "full_text": "When Louisville's mayor suggested last spring that some city police officers were stopping African-American drivers unfairly, he was giving voice to a suspicion - but he had no hard facts to support his belief.\n\nWe promptly went looking to see if there were any. And at first, we came up as empty as the mayor's assertion.\n\nWe were told, for example, that the Louisville Police Department does not enter any data on traffic citations into its computer system, instead sending all citations directly to the state courts system. The courts system does enter the data. But that turned out to be useless to us because of the impossibly complex search required to separate traffic offenses from other types of citations.\n\nOn the verge of abandoning the idea, we turned one last time to some helpful veterans inside the police department for guidance. Could we get a detailed listing of all traffic stops? Yes, we were told, every reported stop is stored in the department's computer-assisted dispatching (CAD) system.\n\nOn the verge of abandoning the idea, we turned one last time to some helpful veterans inside the police department for guidance.\n\nIs the driver identified in those records in any way? No, we were told. Is there any record anywhere identifying drivers stopped? Well, actually, there is one place, one officer said: The \"check sheets.\"\n\n\"Check sheets\" are handwritten logs kept by dispatchers. When an officer stops a car in Louisville and then radios information from the driver's license, a dispatcher writes that information down, including race, before checking to see if the license is valid and if there are any arrest warrants outstanding for the driver. The department retains those sheets for a year.\n\nMatching information\n\nThe Louisville police CAD reports and the \"check sheets,\" we learned, held two pieces of common information that would enable us to match a traffic stop with a license-and-warrant check: police car \"unit number\" and time of day.\n\nFor example, if the CAD report showed that unit 315 had a car stopped between 10 a.m. and 10:20 a.m. on a given date, and the \"check sheet\" for that date showed that unit 315 requested a license-and-warrant check on a white female at 10:07 a.m., we would have a match.\n\nBut even if we could develop this information, would it be meaningful to a community wondering about the attitudes within its police department? Experts in racial-profiling research we consulted thought it probably would be - not necessarily as proof that racial profiling does or does not exist, but at least as a worthwhile indicator.\n\nThis approach also would yield information about innocent drivers who were stopped but not cited, they said, making it in one way broader than studies drawn solely from records of arrests or citations.\n\nUsing the state's open-records act, we requested the CAD reports of all traffic stops and all \"check sheets\" for 30 separate days. We selected the 30 days randomly from a 12-month period using Statistical Package for the Social Sciences (SPSS).\n\nThe police department eventually handed us a six-inch-high stack of \"check sheets,\" and another six-inch-high stack of computer printouts of traffic stops covering all our requested days. The city also was able to give us the CAD information covering 13 of our 30 requested days on a floppy disk.\n\nUnfortunately, since then the police department has no longer decided that \"check sheets\" are public records.\n\nDoing the legwork\n\nThe job of cross-referencing these two information banks was fundamentally a manual task, as tedious as a Florida recount although it was considerably easier for the 13 days the city gave us on the floppy disk. Mark Schaver, The Courier-Journal's CAR director, transferred that information to an Excel spreadsheet, which enabled me to search the CAD records of traffic stops for individual patrol car unit numbers. That was far more efficient than my eyeball scan of paper for the other 17 days. The entire job took me, working alone, about three weeks. City officers made 6,211 traffic stops on those 30 days, and viewing the information in the most conservative way, we believed we had reliably identified the race of 1,647 of those drivers, about one-fourth. In the end, we found that 44 percent of those \"race-identified\" drivers stopped and checked for warrants were African American, in a city where an estimated 27 percent of the driving-age population is black. Experts say that justifies a conclusion that blacks in Louisville were stopped and checked for warrants at a rate twice that of whites. We also found that in the city's police district with the highest percentage of whites, 23 percent of the drivers stopped were African American four-and-a-half times the district's white population. Which raised another question: Regardless of the residential population, what if 23 percent of the district's actual drivers are African Americans?\n\nCounting traffic\n\nThe experts we consulted thought it was imperative that we confront that question by actually counting traffic in that district. Which raised yet another difficult question: Where, exactly, do we go to do that?\n\nMapInfo software showed us the answer at a glance. In our sample of 1,647 \"race-identified\" stops, 267 were within that district - and one-third of those were along one major surface street that bisects the district. Mirroring the district numbers overall, 22 percent of our \"race-identified\" stops along that street were of African Americans.\n\nWe dispatched four teams of reporters to count traffic along a four-mile stretch of that street over five days during the hours when most of the stops in our sample occurred.\n\nThe result: Of more than 3,800 drivers we observed, 7.6 percent were black - one-third the percentage of blacks stopped and checked for warrants in the district. This was important confirming information, the experts said. It was a real-world observation that tended to reinforce a major question raised by the overall numbers. Given the nuances inherent in racial-profiling research, these numbers as a whole proved nothing, but they were persuasive enough to raise a reasonable question, our experts said.\n\nWhen an editor and I arrived in the police chief's office to interview him, the chief had a guest waiting with him - an assistant professor in justice administration from the University of Louisville.\n\nThe Louisville Police Department, however, didn't have any interest in addressing that question. When an editor and I arrived in the police chief's office to interview him, the chief had a guest waiting with him - an assistant professor in justice administration from the University of Louisville.\n\nShe proceeded to explain why our research was invalid. Her primary criticism concerned our sample. Rather than drawing 30 days at random from a year, she said, we should have drawn randomly from all 75,000 traffic stops made during that year. Ideally, she argued, we would have drawn 10 percent of them about 7,500.\n\nOur experts disagreed. So did another expert we called at the suggestion of the Louisville professor herself.", "provenance": { "extracted_by": "anthropic/claude-sonnet-4-6", "extracted_at": "2026-07-05", "verified": true }, "similar": [ { "slug": "2000-december-stories", "score": 0.195 }, { "slug": "1997-november-one-to-many-how-to", "score": 0.19 }, { "slug": "2000-december-race-and-tickets", "score": 0.185 }, { "slug": "2003-may-june-ticket-data-shows-search-rate-gaps", "score": 0.172 }, { "slug": "1999-october-story-update", "score": 0.163 } ], "html_url": "https://dwillis.github.io/uplink/article/2000-december-check-sheets.html", "json_url": "https://dwillis.github.io/uplink/article/2000-december-check-sheets.json" }