If the U.S. Supreme Court had allowed Florida's courts to finish their abortive recount of last year's deadlocked presidential election, President Bush probably still would have won by several hundred votes, a comprehensive study of the uncounted ballots has found.
But if the recount had been conducted under new vote-counting rules that Florida and other states now are adopting -- rules aimed at recording the intentions of as many voters as possible -- Democratic candidate Al Gore probably would have won, although by an even thinner margin, the study found.
The study provides evidence that more Florida voters attempted to vote for Gore than for George W. Bush -- but so many Gore voters marked their ballots improperly that Bush received more valid votes. As a result, under rules devised by the Florida Supreme Court and accepted by the Gore campaign at the time, Bush probably would have won a recount, the study found.
Since the study was launched, the nation's debate over the Florida recount has cooled and Bush, whose legitimacy as president already was accepted by a large majority in January, has won massive public approval for his leadership of the war against terrorism.
The study, a painstaking inspection of 175,010 Florida ballots that were not included in the state's certified tally, found as many as 23,799 additional, potentially valid votes for Gore or Bush.
The significance of these ballots depends on what standards are used to weigh their validity. Under some recount rules, Bush wins. Under others, Gore wins.
But in almost every case, the outcome still is a virtual dead heat, with the two candidates separated by no more than a few hundred votes out of nearly 6 million cast in the state.
A little more than a year ago, after one of the most tumultuous election nights in the nation's history, Americans awoke to discover that the presidential race was -- improbably -- deadlocked.
Florida, with 25 electoral votes, was too close to call. And without Florida, neither Bush nor Gore had a majority of electoral votes.
The official results, which the state certified over Democratic protests, were: Bush 2,912,790, Gore 2,912,253. The margin of 537 votes, less than 0.01 percent of the total votes cast, triggered an automatic recount.
For 36 days, politicians and lawyers argued over whether and how to recount the state's votes. The Florida Supreme Court ordered a recount to begin; the U.S. Supreme Court ordered the recount to stop. On Dec. 13, Gore conceded. On Jan. 20, Bush took office as president.
But thousands of potentially valid votes remained uncounted. And, as a result, the Florida election's outcome remained a matter of debate.
In January, eight major news organizations commissioned a definitive examination of the uncounted ballots in an effort to answer some of the outstanding questions and see if lessons could be learned for future elections.
The review found that:
- Precincts with large numbers of black voters were measurably more likely to produce spoiled ballots than precincts with few black voters. The data cannot explain why. However, the study debunked the belief that older voters are error-prone. Across the state, precincts with younger voters had higher error rates.
- Bush probably would have won any recount of "undervotes," ballots that were rejected because they registered no clear vote for any presidential candidate. By contrast, Gore would have won most recount scenarios that included "overvotes," ballots that showed votes for more than one candidate.
However, Gore's lawyers never pressed for overvotes to be recounted.
- Ballot design was a key factor. Although the Florida fiasco initially focused on the "butterfly ballot" for punch cards in Palm Beach County, the voters' error rate was even higher in some counties that used more modern optical scanning systems but had equally confusing ballots. Most of the errors occurred in 18 counties where ballots spread the presidential candidates across two pages or two columns.
- Hand recounts can be reliable, but only if the rules are clear. The researchers who examined the ballots agreed on the marks they saw more than 97 percent of the time. The disagreements came mostly when they were asked to judge whether a voter who failed to punch a clear hole in a ballot had left a "dimple," an indentation on the card.
- Some Florida counties handled their ballots so carelessly after election night that county officials could not say with any certainty which ballots had been counted and which had not.
White House spokeswoman Nicolle Devenish, speaking on behalf of Bush, said: "The American people moved on a long time ago. This latest media recount was an expensive undertaking that turned up additional inconclusive data. The election was settled last year."
Even before the results were made public, White House Press Secretary Ari Fleischer said his boss considers the issue ancient history. "It's over," Fleischer said.
In a written statement, Gore said: "We are a nation of laws, and the presidential election of 2000 is over. And, of course, right now our country faces a great challenge as we seek to successfully combat terrorism. I fully support President Bush's efforts to achieve that goal."
Democrats long have contended that a plurality of Florida voters intended to cast their ballots for Gore but that thousands spoiled their votes because of confusing instructions, badly designed ballots or other obstacles. The study adds evidence to bolster that case.
In Duval County, which includes Jacksonville in the northeast corner of the state, a remarkable 21,855 ballots were invalidated because voters chose more than one presidential candidate. The county's official sample ballot erroneously instructed voters to "vote all pages." With 10 presidential candidates spread across two pages, following that instruction produced an overvote.
Of those voters who made the mistake of voting once on each page, the study found that 7,794 voted for Gore (plus another candidate), while 4,705 voted for Bush (plus another). That's a potential net for Gore of 3,089 votes, enough to carry the entire state.
Voters in Palm Beach County, where the butterfly ballots listed presidential candidates on facing pages, cast 19,218 overvotes. More appear to have come from Gore voters than from Bush voters.
The study found that 11,140 voters in the heavily Democratic county punched a hole for Gore and one other candidate; only 2,298 punched a hole for Bush and another candidate.
The same phenomenon appeared in smaller counties that used more modern optical scanner systems. On Florida's west coast, 991 voters in Charlotte County chose Gore, then added a vote for a second candidate; 787 voters chose Bush and another candidate.
On most overvote ballots, it's impossible to know which candidate a voter intended to choose. On punch card ballots, for example, there's no way to tell which of several perforated rectangles, called chads, the person intended to punch out to register a vote.
But in counties that use ballots counted by optical scanners, a manual recount often can determine which mark shows the voter's intent, because many voters explain their intentions on the ballot form. Some circle the name of the candidate they meant to vote for; others write the candidate's name on the ballot; others attempt to erase a mark they made in error.
The study found that Florida's uncounted optical scan ballots included as many as 3,527 such potentially valid votes. If those votes had been counted, Gore would have gained 2,206 votes and Bush 1,321 -- a swing of 885 votes for Gore.
- If Florida's ballots had been recounted using a restrictive standard that some Bush lawyers said they could accept, the study found that Gore would have won the state by 105 votes -- as long as optical scanner overvotes showing clear intent were included. But if overvotes were left out of the count, the study found that Bush would have won by 908 votes.
- If the statewide recount ordered by the Florida Supreme Court had not been interrupted by the U.S. Supreme Court, Bush would have won by 493 votes. The reason: Nine counties were including overvotes, but 58 were not. (The Times' analysis of this scenario recorded each ballot according to the standard each county said it was using or planned to use at the time.)
- If the recounts Gore initially requested had been completed in four heavily Democratic counties (Broward, Miami-Dade, Palm Beach and Volusia), Bush still would have won by 225 votes. Those recounts focused only on undervotes, not overvotes -- and the uncounted undervotes were not enough to swing the election to Gore.
- If a recount had been performed under the standards of a 1996 Texas election law signed by then-Gov. George W. Bush, Gore might have won by 42 votes. The Texas law provides that a vote should be counted if it reflects "a clearly ascertainable intent of the voter," including dimpled chads and overvotes on optically scanned ballots.
Most states that have revised their election laws in recent years consider those "clear intent" overvotes to be valid. In Florida, for example, aides to Secretary of State Katherine Harris proposed new recount rules in September that consider an improperly marked optical scan ballot valid so long as officials can see "a clear indication of voters' intent." When Gore asked for recounts in four Democratic counties last November, his aides didn't realize at first that a potentially critical trove of overvotes lay elsewhere -- in counties that used optical scanner ballots.
"We had a lingering suspicion that we would find more votes in the overvotes," said Ronald A. Klain, who ran Gore's recount operation. "But we were having trouble even getting those four counties recounted."
Eventually, Gore did ask for a statewide recount, but his lawyers never pressed for overvotes to be included.
When the Florida Supreme Court ordered a statewide recount in December, based on Gore's petition, it, too, focused only on undervotes -- drawing a dissent from Chief Justice Charles T. Wells. "How about the overvotes?" he asked.
When the U.S. Supreme Court took Bush's appeal of the case, Justice John Paul Stevens asked the same question of Gore's lawyer, David Boies.
"Nobody asked for a contest of the overvotes," Boies explained. "Once you get two votes, that ballot doesn't get counted for the presidency."
Ironically, Bush's lawyers, in their brief to the U.S. Supreme Court, said one of their objections to the Florida recount was that it didn't consider potentially valid overvotes.
The Supreme Court majority agreed that the absence of the overvotes was a flaw in the Florida court's ruling. If the high court had, instead, ordered Florida authorities to design a comprehensive recount -- one that included the overvotes -- the outcome might well have been a victory for Gore.
The study found another wrinkle that might have aided Gore.
Florida's vote was so close on election night that state law required an automatic retabulation. But officials in 16 counties using optical scanning systems never recounted their ballots; instead, they merely rechecked the electronic records of their election night machine count.
Over five weeks of recounts and court battles, Bush's unofficial lead rose and fell almost by the day, at one point dropping to just 286 votes. If the 16 counties had recounted their ballots and included overvotes in their tallies, Gore would have taken the lead, at least briefly, by 48 votes, the study shows.
The study answers another question that has lingered since December: Could Gore have won if he had asked for recounts of the undervotes in counties other than the four he picked?
Some of Gore's advisers worried that they should have sought a manual recount in Duval County, for example, which registered 5,090 undervote ballots (in addition to its 21,855 overvote ballots). But Gore would have lost ground had they done so. The study found that Bush would have gained as many as 834 additional votes in a Duval County recount, mostly from undervotes.
This study was commissioned by Tribune Co., owner of the Los Angeles Times; Associated Press; Cable News Network; The New York Times; The Palm Beach Post; The St. Petersburg Times; The Wall Street Journal; and The Washington Post.
The review was conducted by the National Opinion Research Center, a nonpartisan research organization affiliated with the University of Chicago. Researchers trained and directed by NORC inspected 175,010 ballots and recorded their characteristics in a series of computer databases.
Each media organization independently analyzed the data collected by NORC. The full NORC database will be released to the public today on the NORC Web site to enable readers to examine and analyze the data themselves.
NORC prepared separate computer studies to assess the reliability of the data. Those studies indicated that the researchers agreed more than 97 percent of the time when inspecting the ballots, "a high degree of accuracy," according to Kirk Wolter, a senior vice president of NORC and director of the project.
However, Wolter warned that the outcomes are so close that they cannot conclusively show who got the most votes. "It's too close to call," he said. "One could never know from this study alone who won the election."
Unlike a public opinion poll, the media-commissioned study has no "margin of error" because researchers inspected every available uncounted ballot, not a representative sample. But the study still is not entirely precise because Florida's counties could not locate every uncounted ballot.
Florida's state government doesn't record how many ballots were invalidated on Election Day, and county records are incomplete in some cases. The best estimate is that about 176,400 ballots were rejected as undervotes or overvotes. The NORC study thus included more than 99 percent of the total.
In some cases, county officials could not be certain which ballots had been counted because every time a punch card ballot is run through a tabulating machine, one or more of its paper "chads" can be dislodged.
One county, Volusia, posed a particularly complex problem. Volusia completed a hand recount after election night, and the results are part of the state's certified tally. County officials subsequently could not determine which ballots had been included in their manual recount and which had not. Accordingly, Volusia's certified totals were used in the study rather than any new data from those ballots.