ELECTION 2004
Berkeley study scrutinizes Florida tally for Bush
Unexplained boost linked to e-voting; some not so sure

Wyatt Buchanan, San Francisco Chronicle Staff Writer

Friday, November 19, 2004
 
Researchers at UC Berkeley released a statistical analysis Thursday that shows, they say, that President Bush may have received at least 130,000 extra and unexplained votes in Florida counties that used electronic voting machines.

A professor and three graduate students from the university's Survey Research Center conducted the study and said they have been able "to explain away all other factors" that might have increased Bush's support.

But some political scientists dismissed the analysis, pointing out that researchers did not and probably could not account for massive Republican get- out-the-vote efforts, differences in money spent or differences in amount of advertising by candidates, as well as other political intricacies.

"(E-voting) is not the only factor left because the model is so incomplete. How do you control for the fact that churches and gun groups were out there pumping out people; how would you measure that?" asked Bruce Cain, a political science professor and director of UC Berkeley's Institute of Governmental studies.

"Until you can disprove what Republicans claim was the biggest factor, you don't have a case," he said.

Even 130,000 fewer votes simply would have trimmed Bush's margin of victory in Florida, and the study only shows that something went wrong, but not what that the problem may have been, the researchers said.

Several faculty members reviewed the analysis and could not find major flaws, said Michael Hout, a sociology professor who made Thursday's announcement.

"It's held up to our skepticism and our scrutiny, and I've concluded something went awry with e-voting in Florida," Hout said.

The researchers said the counties with the largest discrepancies, which they attribute to electronic voting machines, were Democrat-heavy Miami-Dade, Broward and Palm Beach counties. Counties where Bush won decisively did not show such dramatic shifts.

In the statewide study, researchers used a technique called multiple regression analysis to account for differences in county size, change in overall voter turnout (not by party), past voting history of residents and demographic factors of income and race.

They also looked at whether a county used an electronic voting machine or another ballot counting system. That distinction was the only one of the six factors that showed a statistically significant relationship with the increase in the Bush vote from the 2000 election, Hout said.

The researchers conducted an identical analysis of Ohio counties and found no distinction between electronic voting machines and other types of systems. Only seven counties in Ohio use electronic machines, and 70 percent of the state still uses paper punch ballots.

Florida uses the electronic machines in 15 counties that represent a majority of the state's residents.

Hout said he doubted Florida residents in larger counties simply changed their minds and voted for Bush over Kerry.

"If they did, they're acting very much out of the pattern that is established in the other 64 counties in Florida," he said.

The researchers said their work was not politically motivated and does not provide a type of smoking gun that shows fraud or an attempted stolen election.

The results could mean embedded software glitches, smudged touch screens that misread votes or machines that counted backward, as happened in at least two counties in other states, Hout said.

But academic scrutiny like the Berkeley study can lead to all types of incorrect conclusions and conjectures, said Jack Pitney, associate professor of government at Claremont-McKenna College.

"There's a big difference between social science and reality," said Pitney, who questioned the results of Hout's study. "The difference is social scientists like to confine things to formulas, and reality is a lot more complicated than a formula."

The company that manufactures the electronic machines used by Broward and Miami-Dade counties, Election Systems & Software of Omaha, Neb., had their own concerns about the study.

In a statement released Thursday, they called the Berkeley analysis a "hypothetical statistical study" and said ES&S officials had not yet reviewed the results.

"If you consider real-world experience, we know that ES&S' touch screen voting system has been proven in thousands of elections across the country. ... Based on this solid track record as well as the extensive testing process that is required before equipment may be used in an election we are confident in the security, reliability and accuracy of all our voting systems," the statement read.

There were problems nationwide with electronic voting machines from a variety of manufacturers. The Verified Voting Foundation has logged 695 separate complaints about e-voting, some that account for thousands of lost votes, said Will Doherty, executive director of the foundation.

"What we're finding is the more you look at the election, the more problems you see,'' he said. "We don't have any way to say if this election has more problems than prior elections because we've never looked at an election before as much as we've looked at this one."