A nationally renowned expert on statistical research and a team of graduate students at UC Berkeley yesterday sounded another alarm bell about the Nov. 2 elections, releasing a study suggesting that irregularities in electronic voting machines in Florida may have awarded hundreds of thousands of “excess” votes to George W. Bush.
The report, published by the Berkeley Quantitative Methods Research Team, said Bush picked up 130,000 to 260,000 votes the group’s statistical analysis can’t explain from three heavily Democratic south Florida counties that used touchscreen voting equipment this year.
The probability of such a discrepancy arising by chance is less than one in a thousand, according to Michael Hout, professor of sociology at the campus, member of the National Academy of Science, and leader of the research team.
Bush carried Florida by a margin of 380,978 votes over the Democratic nominee, John Kerry. Had the state’s 27 electoral votes gone to Kerry, he would have an Electoral College edge of 279 to 259 for Bush.
“Our aim is not to attack the 2004 election results per se,” Hout said at a press conference announcing the study. “We’re not a political action committee, not a lobby—we’re just a bunch of researchers who happen to have something we think is important.”
“Our approach is like a smoke alarm—we’re calling on Florida officials to determine if there’s a fire,” he added. “For the sake of all future elections involving electronic voting, someone must investigate and explain the statistical anomalies in Florida.”
Ever since the polls closed, the Internet has been abuzz with reports casting doubt on the integrity of the reported vote totals, largely by pointing out what the critics say are discrepancies in some areas between those results and exit polls or party-registration data. The Berkeley team took a different approach, focusing on the increase in support for Bush in each county between 2000 and 2004.
In addition to results from the last three presidential elections, they collected data for each of Florida’s 67 counties on other variables that might have influenced the results, such as changes in turnout, median income, and Hispanic/Latino population—as well as the voting technology used. Then they performed a multiple-regression analysis, a statistical technique widely used in the social and physical sciences to measure the effects of different variables on quantitative outcomes.
Following statistical patterns in the counties that didn’t use e-voting, support for President Bush should have decreased by 28,000 in Broward County, for example, but the machines tallied an increase of 51,000 votes—a net gain of 81,000 for Bush, in the study’s terms.
“No matter how many factors and variables we took into consideration, the significant correlation in the votes for President Bush and electronic voting cannot be explained,” Hout said.
The group performed a similar analysis on Ohio’s results, but found no anomalies there.
The 130,000 figure, according to the study (available at http://ucdata.berkeley.edu, under the “Voting” heading), is the total of unexplained votes for Bush in the 15 Florida counties with e-voting. If it’s assumed that these votes should have been counted for Kerry, the net effect doubles to 260,000.
The discrepancy was most pronounced in three large and heavily Democratic counties—Broward, Palm Beach, and Miami-Dade. In fact, Hout said at yesterday’s press conference, “the size of the discrepancy was proportional to the level of support for Al Gore” in 2000.
Explaining how the apparent anomalies occurred is “beyond our powers,” Hout said, but in answer to questions at the press conference he mentioned several possibilities: “embedded software glitches,” “passive electro-mechanical problems,” or even “an accumulation of smudges on one area of the touchscreens causing misreads.”
In addition to Hout, the study was written by three first-year graduate students in sociology at UC Berkeley: Laura Mangels, Jennifer Carlson, and Rachel Best.
“Jennifer and I were hanging out in a café after the election, feeling really frustrated because all these rumors were flying and no one was presenting any really hard evidence,” said Mangels, 24. “We said, ‘Wouldn’t it be good for someone to test the results using solid statistical methods.’ So on the Saturday night after the election we went over to the computer lab and began entering data.”
When the results began to emerge, the student took them to Hout—not only because he’s “one of the leading statistical sociologists in the U.S.,” Mangels said, but also because “he’s known as one of the most skeptical people in the department.” Later the findings were reviewed by half a dozen other Berkeley professors, as well as others at Harvard and other universities, according to Mangels.
“We’ve been back to the lab dozens of times” as the professors suggested additional variables and tests, she said, but “at this point no one can come up with anything else to poke any holes” in their analysis.
But one elections expert, who asked not to be named because she hadn’t had a chance to read the Berkeley report, said she can think of one factor that’s been left out of most discussion of the Florida vote: the effect of the three major hurricanes that hit Florida this fall, the presidential visits that followed, and voters’ calculations about possible benefits to the state from having the brother of its governor leading the federal government.
Ted Selker, a former IBM Fellow who is now a professor at the Massachusetts Institute of Technology and MIT director of the Caltech-MIT Voting Technology Project, said he was unimpressed with the Berkeley paper. Noting that it includes no references and that the authors had not observed the Florida balloting first-hand, he said “I just find this paper neither interesting nor believably accurate.”
Kim Alexander, founder and president of the California Voter Foundation, said “It’s unfortunate that we can’t verify the results in those three counties, because they used equipment that produces no paper trail that can be checked. As long as we have a significant proportion of votes that can’t be verified, we’re just left to speculate.”