{
  "id": "1998-december-first-ventures",
  "slug": "1998-december-first-ventures",
  "url": "https://dwillis.github.io/uplink/article/1998-december-first-ventures.html",
  "year": 1998,
  "month": "December",
  "monthNum": 12,
  "headline": "FIRST VENTURES",
  "kicker": "",
  "author_name": "Bill Muller",
  "author_title": "Reporter",
  "affiliation": "The Arizona Republic",
  "summary": "A journalist investigating workplace deaths in Arizona using OSHA data discovered critical gaps in the database, finding 26 undocumented deaths after verification with state officials. The analysis revealed that workplace fatality fines are often reduced by 40% through appeals and that Arizona provides minimal survivor benefits to families of deceased workers.",
  "keywords": [
    "workplace",
    "deaths",
    "fines"
  ],
  "topics": [
    "Investigative Methods",
    "Data & Databases",
    "Public Safety"
  ],
  "technologies": [
    "Access",
    "Excel"
  ],
  "full_text": "Death on the job\n\nIt was one of those moments that every reporter dreads. I woke up in the middle of the night staring at the ceiling. After analyzing an OSHA database, I had just finished writing a story about workplace deaths in Arizona. In cases where there was a serious safety violation, I had identified 108 deaths since 1992 and established the median fine per fatality: about $3,000. But now I tossed and turned. What if there were more than 108 deaths? What the data was flawed? analysis would be worthless. Luckily, though I already had written the story, its run date was still weeks away. The next day, I called the director of the state industrial commission, who had come to know well during the project, and told him of my misgivings. Without hesitation, he agreed to have a member of his staff go through all workplace deaths since 1992 and make sure we had them all. We didn't.\n\nUndocumented deaths\n\nDue to some reporting problems on OSHA's part, 26 deaths were not in our analysis. Using data provided by the commission, we added the deaths into the Excel spreadsheet that we had used to calculate the median. A funny thing happened: The median was still about $3,000.\n\nIn picking median for analysis, we had done ourselves a favor. When we added the missing cases, some came in above and some came in below, and the median stayed about the same.\n\nStill a good lesson. Some other findings:\n\nFrom 1992 until mid-1997, 79 of the 134 worker deaths investigated by OSHA in Arizona led to citations for serious safety violations. Just 12 fines were more than $10,000.\nThough OSHA fines related to workplace deaths might start high, the appeal system is designed to reduce them. In fact, an analysis of initial fines and current fines showed that they dropped by about 40 percent, saving the companies involved about half a million dollars.\n\nFor widows and widowers of workers who die on the job, Arizona ranks low for survivor benefits. State law protects companies from wrongful death lawsuits, saying essentially that employees covered by workers' compensation cannot recover damages in the event of injury or death.\n\nThough OSHA fines related to workplace deaths might start high, the appeal system is designed to reduce them.\n\nEarly experiments\n\nNaturally, getting the answers wasn't as easy as we first thought. It was my first time dealing with a fairly large database—about 3 MB, with four tables and 43,000 records in the largest table. The data weren't laid out quite the way I thought they would be.\n\nFor example, when I first calculated the fines per death (sorted in Access, then crunched in Excel), I noticed that each record included only one company name. This puzzled me because I knew several cases in which more than one company had been fined in connection with a death. To get the death cases, you had to link the accident table with the OSHA table by case ID number. Each time I asked for cases with '1' in the degree field (the death code), I got the same list and only one company name for each deceased worker.\n\nI decided to experiment. I sorted the data by date and then by address, looking for investigations in the OSHA table that occurred on the same date and at the same address as deaths in the accident table. Essentially this worked. To make sure the cases were related (and not just OSHA inspectors spotting unrelated violations on the job site), we checked the related activity numbers that appear on the case records available on the OSHA Web site. We also had the industrial commission check our work. They found one related case that didn't turn up, mostly because it was added to the case file a month after the accident.\n\nScanning for the striking\n\nAfter generating our initial list of 108 deaths, we searched the data for cases that would be interesting. We looked for small fines, large fines, and large fines that hadn't been paid (by checking the \"penalty remit\" field in the data). I also went through the newspaper library, scanning for details on various cases. We found striking cases, including a small company that had dodged a $47,000 fine by declaring bankruptcy and another that was fined just $7,500 after a worker had been decapitated. To find the relatives of these victims, many of whom had scattered, we used Track, locating one relative in Rochester, N.Y., and another by sending letters to her last two known addresses. Also, after an editor asked the question, I went back to the data to find out if we could document the total amount that OSHA fines had been lowered during the appeal process. Fortunately, the data allowed such a calculation by including \"penalty\" and \"initial penalty\" fields. Because of ongoing appeals, the fines associated with more recent cases tended to change, so we updated them using the OSHA Web site establishment search. (See http://www.osha.gov/cgi-bin/est/est1)\n\nBut 26 fatalities were still missing from the records. The reason, as far as we can tell, was that state officials had not filled out the accident summary when logging the case. So while the accident appeared in the main OSHA table, it was not in the accident table, which contained the degree field and the name of the victim.\n\nI used the records provided by state OSHA to update my chart, and the rest is history. By the way, you might want to replace that frayed electric cord.\n\nBill Muller can be reached by e-mail at mullyb@aol.com",
  "provenance": {
    "extracted_by": "anthropic/claude-sonnet-4-6",
    "extracted_at": "2026-07-05",
    "verified": true
  },
  "similar": [
    {
      "slug": "1995-november-osha-falls-flat",
      "score": 0.362
    },
    {
      "slug": "1995-november-beyond-osha",
      "score": 0.32
    },
    {
      "slug": "2002-march-april-mining-workplace-data",
      "score": 0.287
    },
    {
      "slug": "2006-july-august-making-sense-of-osha-data",
      "score": 0.272
    },
    {
      "slug": "1991-june-the-safe-and-the-sorry-working-in-ohio",
      "score": 0.23
    }
  ],
  "html_url": "https://dwillis.github.io/uplink/article/1998-december-first-ventures.html",
  "json_url": "https://dwillis.github.io/uplink/article/1998-december-first-ventures.json"
}