November 12, 2001
Study of Disputed Florida Ballots Finds Justices Did Not Cast the Deciding Vote
News Analysis: Who Won Florida? The Answer Emerges, but Surely Not the Final Word (November 12, 2001)
The Patterns: Ballots Cast by Blacks and Older Voters Were Tossed in Far Greater Numbers (November 12, 2001)
The Method: How the Consortium of News Organizations Conducted the Ballot Review (November 12, 2001)
(Page 2 of 2)
A statistical analysis conducted for The Times determined that if all counties had followed state law in reviewing the absentee ballots, Mr. Gore would have picked up as many as 290 additional votes, enough to tip the election in Mr. Gore's favor in some of the situations studied in the statewide ballot review.
But Mr. Gore chose not to challenge these ballots because many were from members of the military overseas, and Mr. Gore did not want to be accused of seeking to invalidate votes of men and women in uniform.
Democrats invested heavily in get- out-the-vote programs across Florida, particularly among minorities, recent immigrants and retirees from the Northeast. But their efforts were foiled by confusing ballot designs in crucial counties that resulted in tens of thousands of Democratic voters spoiling their ballots. More than 150,000 of those spoiled ballots did not show evidence of voter intent even after independent observers closely examined them and the most inclusive definition of what constituted a valid vote was applied.
The majority of those ballots were spoiled because multiple choices were made for president, often, apparently, because voters were confused by the ballots. All were invalidated by county election officials and were excluded from the consortium count because there was no clear proof of voter intent, unless there were other clear signs of the voter's choice, like a matching name on the line for a write-in candidate.
In Duval County, for example, 20 percent of the ballots from African- American areas that went heavily for Mr. Gore were thrown out because voters followed instructions to mark a vote on every page of the ballot. In 62 precincts with black majorities in Duval County alone, nearly 3,000 people voted for Mr. Gore and a candidate whose name appeared on the second page of the ballot, thus spoiling their votes.
In Palm Beach County, 5,310 people, most of them probably confused by the infamous butterfly ballot, voted for Mr. Gore and Pat Buchanan. The confusion affected Bush voters as well, but only 2,600 voted for Mr. Bush and another candidate.
The media consortium included The Times, The Wall Street Journal, The Tribune Company, The Washington Post, The Associated Press, The St. Petersburg Times, The Palm Beach Post and CNN. The group hired the National Opinion Research Center at the University of Chicago in January to examine the ballots. The research group employed teams of three workers they called coders to examine each undervoted ballot and mark down what they saw in detail. Three coders provided a bulwark against inaccuracy or bias in the coding. For overvotes, one coder was used because there was seldom disagreement among examiners in a trial run using three coders.
The data produced by the ballot review allows scrutiny of the disputed Florida vote under a large number of situations and using a variety of different standards that might have applied in a hand recount, including the appearance of a dimple, a chad dangling by one or more corners and a cleanly punched card.
The difficulty of perceiving dimples or detached chads can be measured by the number of coders who saw them, but most of the ballot counts here are based on what a simple majority — two out of three coders — recorded.
The different standards mostly involved competing notions of what expresses voter intent on a punch card. The 29,974 ballots using optical scanning equipment were mostly interpreted using a single standard — any unambiguous mark, whether a circle or a scribble or an X, on or near the candidate name was considered evidence of voter intent.
If all the ballots had been reviewed under any of seven single standards, and combined with the results of an examination of overvotes, Mr. Gore would have won, by a very narrow margin. For example, using the most permissive "dimpled chad" standard, nearly 25,000 additional votes would have been reaped, yielding 644 net new votes for Mr. Gore and giving him a 107-vote victory margin.
But the dimple standard was also the subject of the most disagreement among coders, and Mr. Bush fought the use of this standard in recounts in Palm Beach, Broward and Miami- Dade Counties. Many dimples were so light that only one coder saw them, and hundreds that were seen by two were not seen by three. In fact, counting dimples that three people saw would have given Mr. Gore a net of just 318 additional votes and kept Mr. Bush in the lead by 219.
Using the most restrictive standard — the fully punched ballot card — 5,252 new votes would have been added to the Florida total, producing a net gain of 652 votes for Mr. Gore, and a 115-vote victory margin.
All the other combinations likewise produced additional votes for Mr. Gore, giving him a slight margin over Mr. Bush.
While these are fascinating findings, they do not represent a real- world situation. There was no set of circumstances in the fevered days after the election that would have produced a hand recount of all 175,000 overvotes and undervotes.
The Florida Supreme Court urged a statewide recount and ordered the state's 67 counties to begin a manual re-examination of the undervotes in a ruling issued Dec. 8 that left Mr. Gore and his allies elated.
The Florida court's 4-to-3 ruling rejected Mr. Gore's plea for selective recounts in four Democratic counties, but also Mr. Bush's demand for no recounts at all. Justice Barbara Pariente, in her oral arguments, asked, "Why wouldn't it be proper for any court, if they were going to order any relief, to count the undervotes in all of the counties where, at the very least, punch-card systems were operating?"
The court ultimately adopted her view, although extending it to all counties, including those using ballots marked by pen and read by optical scanning. Many counties immediately began the effort, applying different standards and, in some cases, including overvotes.
The United States Supreme Court stepped in only hours after the counting began, issuing an injunction to halt the recounts. Three days later, the justices formally overturned the Florida court's ruling, sealing Mr. Bush's election.
But what if the recounts had gone forward, as Mr. Gore and his lawyers had demanded?
The consortium asked all 67 counties what standard they would have used and what ballots they would have manually recounted. Combining that information with the detailed ballot examination found that Mr. Bush would have won the election, by 493 votes if two of the three coders agreed on what was on the ballot; by 389 counting only those ballots on which all three agreed.
The Florida Legislature earlier this year banned punch-card ballots statewide, directing counties to find a more reliable method. Many counties will use paper ballots scanned by computers at voting places that can give voters a second chance if their choices fails to register. In counties that use that technology, just 1 in 200 ballots had uncountable presidential votes, compared with 1 in 25 in punch-card counties.
Others will invest in computerized touch-screen machines that work like automated teller machines.
Kirk Wolter, who supervised the ballot review for the National Opinion Research Center, said that the study not only provided a comprehensive review of uncounted ballots in Florida but would help point the way toward more accurate and reliable voting systems. All data from the consortium recount is available on the Web at www.norc.org.
Mr. Wolter, the research center's senior vice president for statistics and methodology, said, "I hope in turn this can lead to voting reform and better ways of doing this in future elections." |
