{
  "id": "1995-march-when-crime-doesnt-add-up",
  "slug": "1995-march-when-crime-doesnt-add-up",
  "url": "https://dwillis.github.io/uplink/article/1995-march-when-crime-doesnt-add-up.html",
  "year": 1995,
  "month": "March",
  "monthNum": 3,
  "headline": "When crime doesn't add up",
  "kicker": "Beware examining FBI data",
  "author_name": "Gwen Carleton",
  "author_title": "NICAR staff",
  "affiliation": "NICAR",
  "summary": "Journalists increasingly analyze crime statistics themselves, but face significant challenges with data accuracy, as demonstrated by errors in FBI crime databases including missing data, definitional inconsistencies, and occasional deliberate misreporting by police agencies.",
  "keywords": [
    "statistics",
    "crime",
    "data"
  ],
  "topics": [
    "Crime & Justice",
    "Data & Databases",
    "Investigative Methods"
  ],
  "technologies": [
    "Uniform Crime Reports"
  ],
  "full_text": "Crime statistics are full of potential disasters. Just ask Shawn Macintosh of the Dallas Morning News. \"Several years ago, I found a ring of 9-year-old kiddie crack dealers in Macon, Georgia,\" she recalled. But Macon police said there was no problem. And it turned out they were right. The kids were fabrications, the result of data entry mistakes. In the past, many reporters left statistical analysis up to the experts. Today, more journalists are analyzing the data themselves, and are producing sophisticated stories about crime rates and trends as a result. But they also are finding just how difficult it can be to get the story right. The FBI's Uniform Crime Reports are among the nation's largest and most popular crime databases. The collection of six databases, compiled and released annually, provides data on the victims, offenders and circumstances of millions of crimes committed nationwide. The FBI publishes \"Crime in America,\" a book-length analysis of the year's statistics, shortly before it releases the raw data each year.\n\nThe Uniform Crime Reports have a number of strengths, among them size, detail and a 66-year history. But there are weaknesses as well. The FBI gathers its data from local police, then checks the numbers for mathematical errors and unusual fluctuations. But it doesn't catch every mistake. In addition, participation in the program is voluntary - and not every town volunteers.\n\n\"It's a 'we take your word for it' situation,\" said Carol Napolitano, computer-assisted journalism coordinator for the Munster (Ind.) Times. \"Often, there were significant discrepancies between what we got from the FBI and individual agencies.\"\n\nGarbage in\n\nThe discrepancies usually happen when federal and state agencies define crimes differently, or when police change reporting procedures. But cheating happens, said Jim Omohundro, writer/editor of the Uniform Crime Reports Program. \"There was one major city - I won't say which one - that we sanctioned because we found they were deliberately underreporting homicides,\" he said.\n\nThe FBI excludes any data it deems unreliable from its reports. Some omissions, such as rape statistics for Michigan, Illinois and Minnesota, are consistent: those states define sex crimes more broadly than the rest of the nation. Others come as surprises. Montana, for example, is missing from the 1993 Supplemental Homicide Report, and Kansas and Illinois are missing from several 1993 databases.\n\nThe new statistics have provided reporters with some more subtle surprises as well. \"Crime in the United States, 1993\" stated 70 police officers were killed in 1993, for example, while the electronic version of the data showed 51. The difference? The FBI left 19 state, federal and Puerto Rican officers out of the electronic aggregates.",
  "provenance": {
    "extracted_by": "anthropic/claude-sonnet-4-6",
    "extracted_at": "2026-07-05",
    "verified": true
  },
  "similar": [
    {
      "slug": "1997-decemberjanuary-taming-crime-statistics",
      "score": 0.386
    },
    {
      "slug": "1997-march-crime-connections",
      "score": 0.368
    },
    {
      "slug": "2002-november-december-data-boosts-crime-stories",
      "score": 0.348
    },
    {
      "slug": "2005-january-february-readmetxt-2",
      "score": 0.347
    },
    {
      "slug": "2005-january-february-readmetxt-3",
      "score": 0.337
    }
  ],
  "html_url": "https://dwillis.github.io/uplink/article/1995-march-when-crime-doesnt-add-up.html",
  "json_url": "https://dwillis.github.io/uplink/article/1995-march-when-crime-doesnt-add-up.json"
}