Four years ago, long before "hanging chad" became a household phrase, Citrus County's new elections supervisor made an announcement: I want to scrap our punch card voting system and modernize.
"I think it's much easier for the voter. That's the most important thing," Susan Gill said of the optical scanning system she coveted.
Gill's purchase cost $280,000 in 1997, but it sure saved a lot of grief in 2000.
On Election Day and during the subsequent machine recount, the new system performed much better than the old punch card equipment some counties still used.
In Citrus, voters feed their optical scan sheets into a vote-counting machine at the polling place. When the polls close, poll workers, using telephone and computer technology, dial up the elections office and download the vote totals.
The optical system also produced a relatively low number of fouled ballots: of 57,000-plus vote cards, only 163 registered undervotes and 54 registered overvotes in the presidential race.
If the canvassing board had continued finding and manually tallying undervotes, it would have added 23 more votes to the state presidential vote count, according to a ballot review sponsored by the St. Petersburg Times and a group of other media companies.
Thirteen of those votes would have gone to former vice president Al Gore, according to the standard that county officials planned to use during the manual recount, which the Florida Supreme Court had set in motion Dec. 8.
President George W. Bush would have added 10 to his vote total.
A net gain of three votes for Gore would have been negligible, said Wes Stow, chairman of the Citrus County Republican Executive Committee, noting Bush's overwhelming victory in the county.
"Bush won Florida and that's the end of the story," Stow said. "The clear loser is not Al Gore but the voters in those counties where they were not using modern voting machines."
His Democratic counterpart, Joe Cino, said every vote mattered in the closest presidential election in generations. He noted that Gore could have picked up many more votes in areas with significant ballot problems.
"I'm trying my best to support Mr. Bush," Cino said, "but the stigma of the election still sits with me."
Canvassing board members said they wanted to accept ballots that showed the voter's intent. They would have accepted a circled candidate's name, affirmative marks on or near a candidate's name, a partially filled-in bubble or use of an incorrect writing instrument, such as a red pen.
If the board had used a more generous standard than it planned, Gore would have picked up 18 more votes to Bush's 11, the media review showed.
Of course, neither the overvotes nor the undervotes were ever included in the certified vote totals. The U.S. Supreme Court stopped the manual recount Dec. 9, just as election workers had started sifting through the ballots seeking the 163 undervotes.
Even if the undervotes and overvotes were registered, Bush would have remained the winner in Citrus: He took the county by 4,270 ballots, according to the certified total.
A nonprofit group hired by the Times and other media examined 208 ballots in Citrus on Feb. 14. It was part of a review of 175,010 ballots statewide that were cast but did not register a vote during machine counts and recounts.
The media analyzed a number of recount scenarios, including one that used the standards that individual counties would have employed for the court-ordered recount.
Gill said the Citrus results weren't unexpected.
Her staff could find only 154 of the 163 undervotes. They couldn't find the other nine _ six from absentee ballots, three from ballots cast at polling places _ because the ballots weren't obviously undervotes.
Of those 154, ballot reviewers found only 23 where the voter's intent could be discerned. Most of those ballots were obvious, with the voter either checking next to the candidate's name or circling the oval next to the candidate's name.
Of the remainder, Gill said, most clearly were undervotes. For whatever reason, people simply skipped the race or forgot to mark a choice.
As for the 54 overvotes, most clearly were overvotes, Gill said. Some voters marked all the choices; many chose two or even three candidates.
No matter the reasons, here's the bottom line: Of 57,513 presidential ballots, only 217 _ or 0.38 percent _ were undervotes or overvotes.
"I feel comfortable with that," Gill said last week. "As much as we try, we're never going to be perfect."
The optical scan system prevents most undervotes because it's sensitive enough to pick up on most marks made within the oval that voters are supposed to darken.
The system prevents overvotes by notifying the voter immediately of the problem. The overvotes that made it through were submitted only because voters chose not to correct their ballots, Gill said.
Of the 163 undervotes, 40 were among the county's 10,254 absentee ballots. Those are ballots cast from the voter's home or at the elections office on or before Election Day.
The other undervotes were distributed fairly evenly among the precincts, with two exceptions: Central Beverly Hills and South Inverness each had 12 undervotes.
The latter precinct had more registered voters, 5,796, than any other Citrus precinct on Election Day. The former had a greater concentration of voters age 65 and older than all but two precincts.
Of the 54 overvotes, 23 were found among the 10,000-plus absentee ballots.
The remainder were distributed among the polling places, with many having no overvotes and no precinct registering more than four.
Winner of certified vote: Bush by 4,270 votes.
Outcome under Citrus standard: Gore nets 3 votes.
Registered voters: Democrats, 33,715; Republicans, 33,144.
Voters' race: white, 76,575; black, 1,335; other, 1,206; unknown, 2,262.
Voting machine: Optical (precinct).
Reported undervotes: 163 (0.28 percent, state rank 50).
Reported overvotes: 54 (0.09 percent, state rank 56).
Undervotes reviewed: 154.
Overvotes reviewed: 54.