This story was printed from ZDNet UK, located at http://news.zdnet.co.uk/

Story URL: http://news.zdnet.co.uk/hardware/emergingtech/0,39020357,39174381,00.htm

Florida research raises e-voting questions
Robert Lemos
CNET News.com
November 19, 2004, 09:30 GMT

Researchers at the University of California at Berkeley on Friday published a statistical analysis of irregularities in Florida voter behaviour that contends that the voting patterns favoured President Bush to the tune of 130,000 to 260,000 votes.

The report, by four U.C. Berkeley researchers, analysed the statistical relationships between Florida's 2 November results of the election and a variety of factors, including historical trends in Florida, racial factors and county size. The analysis found that, statistically, people using electronic voting machines tended to favour President Bush in proportion to the number of registered Democrats in each county.

The group stressed that the results were not proof of any errors in counting the vote, but merely suggested that some link existed between the type of machine used to tally votes and the margin by which President Bush won.

"Without a paper trail, statistical comparisons of jurisdictions that used e-voting are the only tool available to diagnose problems with the new technology," the researchers stated in the report.

The paper was authored by Michael Hout, a professor of sociology at U.C. Berkeley, and three other researchers. The analysis found a statistical relationship between electronic voting machines and votes for President Bush, which seems to have accounted for anywhere from 130,000 votes to 260,000 votes. Hout was not immediately available for comment.

While some problems with election machines have surfaced after 2 November, none has been deemed a major issue that could affect the outcome of the election. According to the Web site of Florida's secretary of state, President Bush won that state by 380,000 votes.

That has not stopped liberal groups from taking issues with the results. Strange statistical anomalies in Florida's election results, for example, were initially highlighted by USTogether.org. That group's analysis of the Florida vote found that optical-scan voting machines were used in polling places where Republicans tended to gain a much larger portion of the vote than expected from party registrations.

However, political science professors at Harvard University, Stanford University and Cornell University discounted the theories by pointing out that optical-scan systems are used mainly in rural counties of Florida and that those counties have had registered Democrats that have voted Republican in the past four presidential elections.

"We conclude that allegation is baseless," wrote Walter Mebane, professor of political science at Cornell University.

However, Samuel Wang, an assistant professor of molecular biology at Princeton University who published extensive analysis of election data running up to the November primary, said he believed the latest analysis, unlike previous ones, does a credible job of explaining the statistically odd behaviour of Florida voters.

"I am not prone to conspiracy theories," he said in an email to CNET News.com. "For instance, I think allegations about Ohio are false, and theories based on exit polls are highly misguided. But the Berkeley group's evidence is more convincing to me."

The Berkeley analysis uses voting patterns by county from 2000 and 1996, income by county, total population and Hispanic population to try to explain voting patterns in 2004, all factors used in an attempt to explain Florida voters' propensity to vote Republican far more frequently than voter registration records might otherwise indicate.

"Their analysis indicates that even when all these variables are accounted for, a significant difference remains between counties that used electronic voting and counties that used optical scanning or paper ballots," he said.

Wang's own analysis, using different methods, estimated that e-voting machines inexplicitly favoured Bush by 270,000 votes, he said.

Neither analysis tries to explain what may account for the statistical departure from the expected results of the election. However, they do list several possibilities.

"Mechanisms that would produce this outcome include having votes electronically registered in the machine prior to any voters using the machine or after the last voter users it -- through software errors or hacking -- and other flaws that interfere with counting after some limit is reached," the report stated.