{ "id": "2004-september-october-purged-voter-problems", "slug": "2004-september-october-purged-voter-problems", "url": "https://dwillis.github.io/uplink/article/2004-september-october-purged-voter-problems.html", "year": 2004, "month": "September • October", "monthNum": 9, "headline": "Purged voter problems", "kicker": "ELECTIONS", "author_name": "Chris Davis and Matthew Doig", "author_title": "", "affiliation": "Sarasota Herald-Tribune", "summary": "Reporters discovered a fatal flaw in Florida's felon voter purge list where Hispanics were systematically excluded due to inconsistent race categorization. The analysis led to the scrapping of the list, which could have significantly impacted the 2004 presidential election results.", "keywords": [ "voter purge", "Florida elections", "felon list" ], "topics": [ "Government & Politics", "Data & Databases", "Investigative Methods" ], "technologies": [ "Excel" ], "full_text": "Our analysis of a controversial database that Florida election officials were using to purge felons from the voter rolls exposed a fatal flaw that led to the list's demise. The analysis was simple, though the story was anything but ordinary. We imported the state's data into Microsoft Excel. With the help of a pivot table, we found how many of the 48,000 people on the list were white, black or Hispanic. Excel's pivot table function allows users to create cross-tabs to categorize and summarize data. The check was a basic step any journalist should take after he or she first gets new data. Immediately, the problem appeared in the results – 24,000 of the potential felons were white, 22,000 were black, and 61 were Hispanic. We knew that Hispanics in Florida make up a key Republican voting bloc that had been instrumental in getting Gov. Jeb Bush and the President elected. Roughly 17 percent of Florida's population is Hispanic, the largest minority population in the state. It didn't seem logical that there would only be 61 on the list. After we double-checked our work to make sure we hadn't made a mistake, it left us with three possibilities to explore: 1. Are Hispanics less likely to commit felonies and less likely to register to vote? Neither turned out to be true. Hispanics make up about 11 percent of the state's registered voters, about 11 percent of Florida's prison population and 11 percent of the state's voting-age population. Given those numbers, it didn't make sense that less than 1 percent of the people on the state's purge list was Hispanic. 2. Were Hispanics hidden on the list under different race classifications? To answer this, we took a sample of 200 names from the list and examined them one by one, looking for Spanish surnames. We settled on a sample of 200 after talking with Steve Doig, a journalism professor at Arizona State University and Matthew's father. The sample size gave us an acceptable error margin of just about 7 percent. We obtained our sample by dividing 200 into 47,763, the total number of names on the list. We took the result, 239, and used Excel to number each name from 1 to 239 and repeated that process all the way through the list Then we employed the highly scientific method of asking someone in the newsroom to shout out a number between 1 and 239. When we had our number, we used Excel's AutoFilter tool to display only those names with that number beside them. We found a list of common Hispanic surnames on the Internet, and we enlisted a Hispanic colleague at the newspaper to help us identify people who could be Hispanic felons but listed under another race category. We repeated this process three times to convince ourselves we hadn't made an error. Each time we determined that, at best, 3 percent of the people on the state's purge list were Hispanic. That fell far short of the 11 percent we expected to see. 3. Did Florida's election officials botch their list somehow? Given what happened during the 2000 presidential election, this was something we'd considered from the start. So when we ruled out the other potential explanations, we started investigating how the state had built their felon list so we could spot where a data flaw might have been introduced. Our Hispanic colleague helped again when he reminded us that Hispanics traditionally have trouble being placed into a particular racial or ethnic category. An official with the U.S. Census Bureau agreed with his observation. The perfect example of this problem would be a man born in Cuba whose skin is black. That person might think of himself as Hispanic, and he could identify himself that way when registering to vote. But if someone else, such as a police officer, attempted to categorize him, he could easily be considered black. We interviewed an official with Florida's Department of State who explained that they merged two data sets – one from the Florida Department of Law Enforcement and another from the Division of Elections – to create a list of people who were convicted felons registered to vote. Florida is one of a handful of states in which a felon's voting rights aren't automatically restored. If you've used Microsoft Access to join two tables, you know an identifier common in both databases is needed to pull off the feat. And because the data might not be uniform in the tables, you also know what a headache it can be to make that match. To ensure that John Smith the felon was also the same John Smith registered to vote, state officials used every identifier available to make their match. This included first and last names, date of birth, sex and race. If any of those fields didn't match across the tables, that person would not have been included in the purge list. Because state officials used race for the matching process, we knew that's what had caused so many Hispanics to be left off the list. We presented what we found to the state's elections chief, and she conceded that problems in classifying Hispanics into a single race category had probably resulted in their exclusion. Despite the admission, officials said they had no plans to fix the Hispanic felon flaw. But apparently it's a lot harder to resist something published in The New York Times. After seeing the Herald-Tribune story, Times reporter Ford Fessenden learned that there was more to the problem than ambiguity in race reporting. It turned out that the FDLE database didn't even offer Hispanic as a race category. Later, the Herald-Tribune reported that prior to 1994, many supervisors of elections did not offer Hispanic as a race category. That virtually guaranteed that Hispanic felons wouldn't end up on the purge list. And it meant that Florida election officials had botched three felon purge lists since 1998. Three days after the Herald-Tribune first reported the Hispanic voter flaw, Fessenden's story ran in The New York Times. Later that day, Florida Secretary of State Glenda Hood announced she was scrapping the list. The flaw has cost the state millions of dollars and diminished its credibility, especially with Democrats and African-Americans who claim the list is designed to minimize their political power. The current purge list had only 9,500 Republicans, compared to about 28,000 Democrats. The demise of the list could have a huge impact on the November presidential election, which polls show could be as close as the one in 2000. With another quick Excel analysis, we were able to illustrate the potential impact. We began with the premise that the presidential election in Florida would be as close in 2004 as it was when it was decided by a 537-vote margin in 2000. We wanted to see how many people on the purge list would have to vote along party lines to reverse the outcome of a hypothetical election that was as close as the Bush-Gore race. What we found is that even if every Republican on the purge list voted for Bush, less than 36 percent of the Democrats would need to vote for Kerry to give Florida's electoral votes to Democrats. And if every Republican and Democrat on the list voted for their party's candidate, the outcome would be an 18,000-vote victory for the Democrats. That little bit of math let us write this lead: 'If the 2004 presidential election ends up being as close as the last one, Gov. Jeb Bush may have cost his brother his job.'", "provenance": { "extracted_by": "pre-v2", "extracted_at": "", "verified": true }, "similar": [ { "slug": "2004-july-august-readmetxt", "score": 0.341 }, { "slug": "2001-marchapril-floridas-dead-heat", "score": 0.273 }, { "slug": "2005-january-february-politics-and-elections-an-evolving-story", "score": 0.24 }, { "slug": "2005-january-february-post-vote-politics-payoffs", "score": 0.222 }, { "slug": "2001-marchapril-car-in-the-aftermath", "score": 0.197 } ], "html_url": "https://dwillis.github.io/uplink/article/2004-september-october-purged-voter-problems.html", "json_url": "https://dwillis.github.io/uplink/article/2004-september-october-purged-voter-problems.json" }