http://www.herald.com/content/archive/news/elect2000/decision/10 4268.htm

If no one had ever heard of hanging chads, if the butterfly ballot had never flown, if no voter
had bungled in the booth, who would have won Florida and the presidency of the United States?

In a race so tight, it may never be known for certain. But an analysis commissioned by The
Herald of voting patterns in each of the state's 5,885 precincts suggests that Florida likely
would have gone to Al Gore -- by a slim 23,000 votes -- rather than George W. Bush, the
officially certified victor by the wispy margin of 537.

It's a hypothetical result derived from something that clearly doesn't exist in Florida or anywhere
else in the nation -- an election where every ballot is fully filled out and every one of those ballots
gets counted, an elusive ideal going these days by the buzzword "the will of the people."

But it is also as close as anyone is likely to get to the statewide manual recount that some
people say is the only way to fairly assess who should be awarded Florida's 25 Electoral
College votes. Reaction to the analysis from the two camps locked in an exhausting and
tense legal battle was radically different. The Gore campaign called it "compelling
evidence," and the Bush campaign dismissed it as "statistical voodoo."

One fundamental flaw, Republicans argued, was an assumption that every voter actually
intended to cast a vote in the presidential race. A large majority of ballots in the disputed
counties of Palm Beach and Duval didn't even have a dimple on them, said Bush
spokesman Tucker Eskew.

"If you want to divine voters' intent when there isn't even a mark on the ballot, you'd do
better to hire a palm reader than a statistical analyst," he said.

But Stephen Doig, a professor at Arizona State University who crunched the numbers for
The Herald, defended the analysis. For example, he said, even if the analysis were
adjusted to include the remote possibility that 90 percent of voters whose ballots were
discarded actually intended to skip the race, the margin still would make a decisive
difference for Gore -- about 1,400 votes.

Doig described it as a matter of analyzing extremes. In his, he started with the assumption that
every one of the 185,000 discarded ballots represented an intent to vote in the presidential race.
The other extreme, he said, is the Bush contention that none of them should count.

"That extreme is the reality that we have, that Gov. Bush won by a razor-thin 500 votes,"
Doig said. "I'm no psychic. I don't know what they really intended to do, but I do know that
almost anywhere in that margin, Gore wins. You can argue about where in the point it should be."

Political analysts were mixed on what the numbers mean. Larry Sabato, director of the
University of Virginia's Center for Governmental Studies, said he considered the analysis
open to questions.

"That is a reasonable assumption for the purposes of analysis," he said. "For the
purposes of politics, it's highly questionable. In most precincts, that may well be true, but in
some precincts it may not be, and that's a critical difference." Still, Sabato said he found
the end result îîperfectly reasonable."

"What you're providing evidence for, however speculative, is that more people showed up
on election day for Al Gore," he said. "But I'd also state that in our system, woulda,
shoulda, coulda doesn't matter. Only legal votes matter."

And all statistical and anecdotal evidence he'd seen, he said, indicated Bush probably
collected more of those -- the ones that counted.

Susan MacManus, a political science professor at the University of South Florida, said
there were too many variables in the analysis "to feel comfortable."

"Inferring what the voters' intent was, I have a real problem with people who can say they
can do that," she said.

No one, of course, can accurately assess what 185,000 voters intended to do with their
discarded ballots, but in purely statistical terms, there are consistent trends.

The results, derived from precinct returns Herald researchers collected from the state's 67
counties, were determined using this formula:

Statewide, at least 185,000 ballots were discarded, rejected as either undervotes (failing
for whatever reason to successfully mark a ballot or punch out a chad) or overvotes
(selecting more than one candidate for whatever reason). That number includes rejected
absentee ballots.

If those ballots had been included and those voters behaved like their neighbors in the
same precincts, Bush would have gotten about 78,000 (42 percent) of the uncounted
votes and Gore would have gotten more than 103,000 (56 percent). The remaining 4,000
or so would have gone to the minor candidates.

That assumption of voting patterns is based on a concept long accepted by pollsters --
that the opinions of a small percentage of people can be extrapolated to project the views
of a larger group. In this case, however, the projection uses a larger group, generally from
90 to 98 percent of the successful votes in precincts, to project the intent of a few.

The result: Gore ahead by 23,000 votes, a comfortable lead in comparison with the official
statistical toss-up, though still narrow enough to trigger the state's automatic recount,
which kicks in when elections finish closer than one-half of one percent.

The analysis also confirmed that the voters in Democratic precincts had a far greater
chance of having their ballots rejected. Only 1 in every 40 ballots were rejected in
precincts Bush won, while 1 of every 27 ballots were rejected in precincts Gore won.

In addition, Doig, a former Herald research editor who now holds the Knight chair at the
Cronkite School of Journalism specializing in computer-assisted reporting, found a
number of interesting other trends:

* Voting machinery played a large role in rejections.

Of the 51 precincts in which more than 20 percent of ballots were rejected, 45 of them
used punch cards -- 88 percent. Of the 336 precincts in which more than 10 percent were
tossed, 277 used punch cards -- 78 percent.

The overall rejection rate for the 43 optical counties was 1.4 percent. The overall rejection
rate for the 24 punch-card counties was 3.9 percent. That means that voters in punch-card
counties, which included urban Democratic strongholds such as Broward and Palm
Beach counties, were nearly three times as likely to have their ballots rejected as those in
optical counties.

* In dozens of Florida precincts, at least one out of every four ballots was discarded as
having no vote or too many votes for president.

* Nearly half of Gore's margin, more than 11,000 extra votes, would come from Palm
Beach alone. The other counties that would give him more than 1,000 new votes are
Broward, Miami-Dade, Duval and Pinellas. Of those, Bush carried only Duval in the official
tabulation.

* Palm Beach, home of the infamous butterfly ballot, and Duval, where candidate's names
were spread across two pages, had 31 percent of the uncounted ballots, but only 12
percent of the total votes cast.

* More than 11 percent of precincts statewide recorded no discarded ballots.

* Attesting to how close things were, the analysis shows only one county that would
actually switch preferences for president -- tiny Madison, which officially went to Bush, but
would go to Gore under The Herald's projections. More than 10 percent of Madison's
4,000-plus ballots were rejected.

Doug Hattaway, a spokesman for the vice president's campaign, said the results
bolstered Gore's contention that the official results did not fairly and accurately reflect the
vote.

"The outcome of the presidential election rests on determining the will of the voters of
Florida and this new evidence makes it extremely hard for the Bush forces to ignore the
people's will," he said.

Eskew, the spokesman for the Texas governor, flatly rejected it as "hocus pocus" and "an
utterly unfounded scientific process."

In addition to mistakenly assuming that voters handing in undervotes intended to vote, he
said, the analysis also ignores the notion that many of the double-punched ballots may
have been "protest votes," intentionally spoiled.

"That is a deeply flawed model that suggests statistical voodoo," he said.

There are, however, ways of analyzing the data that attempt to account for the possibility
of protest votes and deliberate non-participation in the presidential balloting. Even so,
Gore hypothetically still would have collected enough votes to change the outcome of the
election.

Historically, about 2 percent of votes in presidential races don't count -- most often
because voters either skipped the race or their marks weren't recorded by counting
machines. Florida's rejection rate this year, however, was higher -- about 2.9 percent.

The analysis tested even higher percentages of non-votes, ranging from 10 to 90 percent
of the 185,000 discarded ballots. In each instance, Gore still earned more votes.

The analysis also attempted to discard all undervotes as intentional non-votes, counting
only overvotes. That analysis was hampered by the fact that 37 counties did not
differentiate in their reports between ballots discarded as undervotes and those
discarded as overvotes.

But based on results from the 30 counties that did, 43 percent of the uncounted votes
were undervotes. If that pattern held statewide and every undervote was tossed, ignoring
the entire chad issue, Gore still would have a 13,000-vote margin.

"One thing I would note is that there were other opportunities for protest votes, one of
whom was at last widely seen as a legitimate protest vote and in fact styled himself as that
[Ralph Nader]," Doig said.

The results also would seem to challenge Bush camp assertions that the Texas governor
would prevail in a statewide recount. But Republicans and some analysts didn't think they
were strong enough to stand up.

MacManus, the USF political scientist echoed Eskew's concerns about protest and
apathetic votes, and said there were such wide variances in the size and social and
economic mix of precincts that it would be too difficult to extrapolate accurate results.

"In polls, you're used to a margin of error," she said. "Here, there's no room for margins of
error."

Others saw more validity in the analysis.

Alan Agresti, a professor of statistics at the University of Florida, reviewed the
methodology and called it "overall reasonable."

"You can always raise criticisms. You can never know for sure," he said. "But I think when
you do it at a very fine level like this, at the precinct level, it's very interesting, a good
projection of what could have happened."

Jim Kane, an independent pollster based in Fort Lauderdale, agreed the analysis
contained many uncertainties, what statisticians call "ecological inference" -- a false
assumption that voting patterns would be systematic within precincts. In reality, the small
percentage of voters whose ballots were discarded might be the most unpredictable
group of all.

But he also said "I'm not shocked that Gore would have won."

"All of the evidence points to that, if every single ballot would have been cast correctly,
Gore would have won the state," he said. "I don't think anyone with any reasonableness
would dispute that."

In fact, Kane, Agresti and Doig agreed that the formula probably was conservative,
awarding Bush too large a share of the pie. The biggest problems with rejected ballots
were in low-income, mostly minority neighborhoods statewide -- areas that voted heavily
Democratic. That could suggest that the same group, which included a larger percentage
of first-time and less educated voters, might have made similar errors in all precincts.

Stephen Hess, a senior fellow at the Brookings Institution, a Washington, D.C., think tank,
also found the numbers persuasive.

"If you did this at the county level, you'd have too many variables," Hess said. "You can't
get any smaller than precincts."

"It's perfectly scientific if it's presented in a sense as the most massive statewide poll in
Florida," he said. "Sure, it's fun and games, but it says something about what would have
happened if everybody knew how to vote. The problem is that all of those people whose
votes were not counted, they were not counted in part for perfectly good reasons. It wasn't
all shenanigans. In some cases, they didn't choose to vote, or they were too dumb, or they
just didn't follow instructions."