While the vast majority of Florida's overvotes could never have been assigned legally to any presidential candidate, experts say there are some clues in the ballots that offer at least evidence of, if not proof, for whom voters meant to vote.
And that evidence suggests Al Gore was preferred by more voters than George W. Bush.
Among the clues: 71,548 overvote ballots had a vote for Gore, but not Bush. There were 25,082 overvote ballots that had a vote for Bush, but not Gore. The difference: 46,466.
Another clue: In eight punch-card counties where computer records of the ballots could be examined, 76 percent of the voters who chose Gore and another candidate also voted for Democrat Bill Nelson in the U.S. Senate race. In the same counties, 45 percent of voters who chose Bush and another candidate voted for Republican Bill McCollum in the Senate race.
And still another possible clue: More than 4,300 overvotes occurred when voters darkened the ovals for all the candidates but Bush or Gore. Elections supervisors refer to these as ``anybody but'' voters. Of these, 3,604 were ``anybody but Bush'' votes, apparently favoring Gore, while the ``anybody but Gore'' vote, apparently favoring Bush, numbered 737. Again, the difference, 2,867, is large enough to overcome any Bush lead.
NO PROOF
To be sure, all such analyses are highly speculative and can never be proved. Under Florida law, overvotes -- in which a voter choses more candidates than are to be elected to the office -- cannot be tabulated. They also assume that most voters who picked more than one candidate did so by mistake and not on purpose -- a proposition some people disagree with.
``This sort of thing is just full of speculation and the possibility of disagreement and guesswork,'' said former Montana Gov. Marc Racicot, speaking for the Republicans. ``That should not be the standard for how you go about electing a president.''
Nonetheless, the results of the overvote review are similar to The Herald's findings in December that suggested Gore would have won a 23,000-vote margin in Florida if all discarded ballots had been countable. That analysis, by Arizona State University professor Stephen Doig, examined voting patterns in each of the state's 5,885 precincts and assigned uncounted ballots to the two candidates based on the percentage of valid votes the candidates had received in that precinct.
Another professor, Anthony Salvanto, a political science professor at the University of California at Irvine, has reached a similar conclusion in examining the overvote ballots.
``It appears as though most people who went to the polls intended to vote for Gore,'' Salvanto said, ``especially when you consider Palm Beach and Duval counties.''
MISUNDERSTANDING
By looking at how a person who overvoted in the presidential race voted in other races, Salvanto concluded that many overvoters intended to select only one presidential candidate but accidentally nullified their vote because they misunderstood the voting instructions.
Salvanto said that computer records of punch-card ballots in eight counties showed that 82 percent of presidential overvoters cast a valid Senate vote, while only 6 percent overvoted in the Senate race and 12 percent did not vote in that race. Salvanto also analyzed how many Gore overvoters supported Nelson and how many Bush overvoters favored McCollum.
``Everything else being equal, Democrats are more likely to vote for a Democratic senator and president, and Republicans are more likely to vote for a Republican president and senator,'' Salvanto said.
The most common overvote combination of all the various possible permutations -- 10,233 -- was Gore and conservative commentator Pat Buchanan, the Reform Party candidate. Almost half of these overvotes came in Palm Beach County, where the ``butterfly ballot'' prompted many Gore supporters to mistakenly choose Buchanan as well.
The second most common overvote combination -- 9,321 -- was Gore and Browne, the Libertarian Party candidate.
The third most common was Bush and Gore, 8,952 times.
Palm Beach's ``butterfly ballot'' has received worldwide attention but Salvanto says the other combinations can also be explained as attempts to cast a valid vote.
In more than 18,000 cases, voters chose two candidates who were next to each other on the ballot, possibly because they thought the instruction ``Vote for Group'' meant they should vote for more than only the presidential candidate, Salvanto said.
EIGHT COUNTIES
Examining figures in eight big counties, Salvanto found that voters selected candidates next to each other in 30 percent of overvotes. In nearly all examples, the combination was Bush-Gore or Gore-Browne.
Bush was the first candidate on the ballot, followed by Gore and then Browne.
Salvanto concluded that many of those who voted for Bush-Gore in the eight counties favored Bush because 45 percent -- 3,030 -- voted for the Republican candidate in the U.S. Senate race.
Similarly, Salvanto concluded that most of those who voted for Gore-Browne favored the former vice president because 80 percent of them -- 4,660 -- voted for the Democratic U.S. Senate candidate. Augmenting this analysis: 55 percent of the Gore-Browne overvotes voted a straight Democratic ticket in the other races, Salvanto found.
Other problems caused overvotes.
In 19 of Florida's 67 counties, the local supervisor of elections did not put all 10 presidential candidates in a single column. In 15 of the counties, the left-hand column contained eight names and the next column had two candidates -- Howard Phillips of the Constitutional Party and Monica Moorehead of the Workers World Party.
In those counties, the most common overvotes paired either Bush or Gore in the left-hand column with either Phillips or Moorehead in the next column, the review shows.
Voters in counties where the candidate list was split were three times more likely to spoil a ballot with an overvote than were voters in counties where the candidates' names appeared in an unbroken list.
``Ballot design did play a major role in overvotes,'' said Pam Iorio, Hillsborough County's supervisor of elections and president of the state association of supervisors. ``That is one of the significant aspects of the analysis from the presidential election in Florida.''
ONE CANDIDATE
``Most voters were able to understand that they were only supposed to vote for one candidate,'' said Clay Roberts, who heads the state Division of Elections. ``But others -- maybe because they hadn't voted recently or were in a hurry to pick up their kids from daycare -- voted for someone in both columns. It's a bad idea to break the races between columns.''
The election law just approved by the Legislature gives the Secretary of State -- Roberts' boss -- the authority to draw up uniform ballot guidelines statewide. But Roberts said he does not yet know whether his office will direct or suggest that the 67 supervisors of elections keep all the candidates in one column.
Salvanto said his research also supports the idea that voters who marked all the candidates except Bush or Gore favored ``anybody but'' the candidate they didn't mark.
In the eight counties he examined, Salvanto found that 80 percent of ``anybody but Bush'' voters -- over 2,000 votes -- chose the Democratic Senate candidate.
Similarly, he found that 66 percent of all ``anybody but Gore'' voters -- or 400 votes -- favored the Republican Senate candidate.
Salvanto was able to examine only the electronic records of ballots. The Herald review of overvotes did not record Senate preferences in other counties.
Another type of overvote -- where the voter darkened the oval next to the candidate's name and then darkened the write-in oval and wrote in the name of the candidate or his vice presidential candidate -- also cost Gore more votes than it did Bush.
He received 1,221 of these votes, versus 752 for Bush, a difference of 469.
Many canvassing boards -- which looked at disputed ballots on election night -- determined that a voter who chose Gore and then wrote in his name definitely favored the vice president, or Bush in a similar fashion. But others chose not to count those votes.
Herald database editor Tim Henderson contributed to this report.
