By Thomas Hargrove
Scripps Howard News Service

The Series Part 1: Analysis says vote system failing Part 2: How the vote study was done Part 3: 12 states don't even tally voters

A new analysis of the way America conducts elections shows at least 1.7 million ballots did not count in the presidential race four years ago, revealing fundamental flaws far deeper than the current problems in Florida.

Thousands of voters - probably hundreds of thousands - intended to support a candidate in 1996 but lost their franchise because of errors in the marking, tabulating and recording of ballots, according to a Scripps Howard News Service analysis of presidential voting in 1,825 counties and voting precincts totaling more than half of the country.

The hidden side of democracy in the United States is that large numbers of votes disappear in every major election because of complicated ballot procedures that confuse or intimidate voters, faulty voting machines and a wide variety of clerical errors by local and state election officials.

The study found serious mistakes by election officials in which thousands of votes were lost or inaccurately recorded. The error rates are significantly higher in counties that use 1960s-vintage computer punch cards rather than newer voting technologies such as optically scanned balloting.

At least 146 counties have suspiciously high drop-off rates - a term election observers use to describe the difference between the number of people counted at the polls on Election Day and the number of presidential votes actually tabulated.

Election supervisors in some counties are at a loss to explain why their certified vote tallies suggest that nearly a fifth of the electorate did not cast a presidential ballot. "I guess they just didn't vote," asserted one Florida official.

Yet experts agree such explanations of the drop-off rates are dubious, if not ludicrous.

The Scripps Howard study re-examined the 1996 presidential election in all 50 states and the District of Columbia, a contest that drew little scrutiny because President Clinton got 8.2 million more votes than former Senate GOP leader Bob Dole. Even if all the undercounts went for Dole, he still would not have won.

The study calculated the presidential drop-off rate for every county in the 35 states that reported, county-by-county, the number of voters who went to the polls four years ago. At least 98 million Americans went to the polls nationwide in 1996, but less than 96.3 million votes were counted for president, a shortfall of 1.7 million.

The actual shortfall probably was larger, but 1,391 counties or voting precincts did not count the number of voters who went to the polls in 1996.

Tyler County in West Virginia reported to state authorities that 3,872 people cast ballots in the 1996 general election, but only 2,780 votes were recorded for the presidential race. Five months ago, officials discovered they had erroneously reported that Republican Bob Dole had received 734 votes rather than the 1,734 he actually got.

"We don't know how it happened but it was a clerical error. Someone typed the wrong number," said Lora Thomas, the county clerk. "As a result of this, we are now double checking all figures before they go out."

No one can explain why it took four years to notice that more than a fourth of Tyler County's presidential votes had been lost.

"Something like this happens at least once in every election. There have been clerical errors both at the county level and at the state level," said West Virginia Deputy Secretary of State Mary Ratliff.

The median drop-off rate in the 1996 election for president was 2.08 percent.

Eighteen counties had election returns that seem to indicate at least 10 percent of their voters went to the polls but decided not to cast a ballot for president - a finding election experts say should warn local authorities that something may be wrong.

In several counties, errors were made in the reported number of ballots cast.

"The state changed our canvassing numbers and really jacked them up," said Robin Vannatta of the Eddy County Bureau of Elections in New Mexico. "But we caught the mistake when the report came back to us. We've since corrected this."

Other counties could not account for the discrepancies.

"I don't have any idea how that happened," said Wells County, Ind., Clerk Betsey Noe when asked why 15,239 ballots were reported cast but only 11,294 presidential votes counted. "It's not in our elections book. There is not even an accumulated total. I don't have it and I can't explain it."

Officials in two counties with high drop-off rates were insistent that no mistakes have been made.

Dixie County, Fla., reported that 4,690 voters went to the polls in 1996 but only 3,805 votes were counted in the presidential race. Nearly 20 percent of the voters did not seem to cast a ballot for president.

"I don't see no problem. Nothing is wrong here. I guess they just didn't vote," said Supervisor of Elections Mae Beville. "Maybe they voted for more than one person for president and so that race was not counted. But we can't determine that from how the machine counts the votes."

Seventeen percent of the voters in Wheeler County, Ga., apparently did not cast a ballot for president.

"All I know is what the machine tells us," said Wheeler County Probate Court Clerk Nina Price. "We had some local candidates in 1996 and that might have meant that folks voted for them and didn't vote for president. It happens."

Curtis Gans, director of the Committee for the Study of the American Electorate, laughed when told some counties were defending double-digit drop-off rates.

"My assumption would be that this is mostly the result of ballots that do not conform with the voting machine's reading program, rather than people not actually voting for president," Gans said.

Kimball Brace, president of Election Data Services, which advises state and local polling officials, said authorities should aggressively re-examine the vote when they see a large drop-off for a major race.

Every year hundreds of thousands of votes are disqualified because voters apparently have selected more than one candidate in the same race. This is known as an over-vote and occurs from voter confusion, poorly designed ballots, or mistakes in the mechanical counting systems used.

"A drop-off rate of 20 percent or so is very large. It could be an indication that some of the machines are screwed up," Brace said. "The fear I have is that we probably have had more over-votes than we thought was the case in the past. And most counties have not run reports on over-votes."

The study found that the drop-off rate varies considerably according to the kinds of voting machines used.

The median drop-off rate for the 1996 presidential race was less than 2 percent in small counties that use paper ballots counted by hand, 2.5 percent in counties using computer punch cards, and 2.8 percent for the multiple punch-card system known as Datavote.

The drop-off rate was less than 1.5 percent in counties using lever voting machines and 1.8 percent in areas using optical scanning machines. Both of these voting systems generally prevent voters from selecting more than one candidate for the same office.

Counties using older punch-card voting systems have a drop-off rate generally 1 percentage point higher than those using voting machines or the newer optical scanned balloting. Election experts agree this difference almost certainly represents a problem with the method of voting rather than a decision by voters not to cast presidential ballots.

About 35.8 million presidential ballots were cast using punch cards four years ago. If the 1 percentage point difference is the result of problems with punch card voting, then at least 350,000 votes were lost that year - more than all the votes cast that year in Delaware or Wyoming.

"My guess is that four years from now, we won't have a punch card system in use anywhere in America," said Gans. "From what we've seen in the mess down in Florida, hand ballots may be more reliable than are punch cards."

Florida, center of the balloting challenges raised by Vice President Al Gore, did have a higher than average drop-off rate four years ago. But at 2.8 percent, Florida's discrepancies between the number of ballots cast and the number of presidential votes recorded is relatively mild.

Much higher were the drop-off rates of 4.9 percent in Utah, 4.5 percent in Arizona, 4.1 percent in South Carolina, 3.9 percent in Georgia, 3.6 percent in Idaho, and 3.0 percent in Kentucky.

In addition to the 18 counties that reported double-digit drop-off rates, another 128 counties had rates of more than 5 percent. The study found that rural, Southern and Western counties appear more prone to suspiciously high drop-off rates.

Sloppy record keeping may also contribute to the problem. There are 190 counties that have registered voter lists so out-of-date and laden with the names of deceased voters that they seem to have more registered voters than actual voting-age populations. These more-than-100-percent registered counties had a median drop-off rate of 2.2 percent, slightly higher than the national average.

"We just don't know how much of this are intentional non-voting for president. It seems that something like 2.2 million to 2.8 million people did not vote in 2000," Gans said. "All we can go on is that those votes, one way or another, were not counted. But with really good voting systems, we could ferret out the intentional non-votes from the mistakes."