The Counter-deception Blog

Examples of scientific analysis used to unmask deceptive behavior are rare--there are few uber-analysts like the famous R. V. Jones--but this piece notes some noteworthy counter-deception techniques.

October 1, 2004 NYTimes.com OP-ED COLUMNIST
Sacrifice and Sabotage
By BOB HERBERT
...
Two professors at the University of Miami did an extensive analysis of so-called voter errors in Miami-Dade County that has not previously been reported on, and that gives us an even more troubling picture of the derailment of democracy in Florida in the 2000 presidential race.

Bonnie Levin, a professor of neurology and psychology, and Robert C. Duncan, a professor of epidemiology, said the purpose of their study was to examine the demographics associated with the uncounted votes in Miami-Dade, a county that disqualified 27,000 votes.

Professor Levin told me yesterday that the study convinced her that a much bigger problem in Miami-Dade involved "over-votes," instances in which ballots were reported to have been disqualified because individuals cast votes for more than one presidential candidate.

In their analysis, the professors factored in variables associated with increased errors, such as advanced age or lower education levels. What they found startled them. The instances of voter errors, after taking all relevant variables into account, was much higher - higher than could reasonably have been expected - in predominantly African-American precincts. And, peculiarly, there was an especially high amount of over-voting among blacks.

"Although African-American and Hispanic precincts are similar in terms of household income and education, the African-American precincts have many more over-votes and under-votes," the professors wrote. "Interestingly, they differ strongly in party affiliation (African-American predominantly Democrat, Hispanic more Republican)."

Surprise, surprise.

Dr. Levin said she did not believe these were the kinds of honest errors one would expect to find in an analysis of voting patterns. Something else was at work. "The data show that it was so specific to certain precincts," she said. "It was so targeted toward African-Americans. There was nothing random about it."

She said, "The most important finding was that education was not a predictor for African-Americans."