A study by UC Berkeley professor Henry Brady remains at the heart of much of the legal wrangling over California's recall election date.

The study, performed at the request of the American Civil Liberties Union, found that punch-card machines are more error-prone than other voting systems and have a disproportionate impact on minority voters.

A three-judge panel cited Brady's study last week when it decided to postpone the election until March.

By then, the six counties now using punch-card technology - Los Angeles, Solano, Mendocino, Sacramento, San Diego and Santa Clara - will have phased it out.

"The big problem is you have machines that are simply throwing away the votes," Brady said Monday. "Even people who are careful in using the system can have their vote not counted."

Brady, a professor of political science and public policy, attended Monday's 9th U.S. Circuit Court of Appeals hearing but did not speak. Instead, he watched attorneys and judges debate the merits of his analysis.

Looking at results from the 2000 presidential election in California, Brady compared punch-card counties to those using other technologies.

He found punch-card counties had a rate of disqualified ballots 1.34 percentage points higher than other counties.

Ballots can be invalidated for many reasons, including voters not punching all the way through and thus appearing not to vote, or punching more holes than they should and thus "overvoting."

Brady concluded that punch-card systems discriminate against minorities for two reasons.

He noted that the six counties now using punch-card ballots have a larger percentage of minorities (46 percent) than other counties (32 percent.)

When punch-card systems are used, minorities have higher rates of disqualified ballots than white voters, his study found.

Brady did not analyze why this occurs, but he said, "I think it's probably because they have lower education."

During the Monday hearing, Judge Alex Kozinski noted that Brady's study did not take into account votes that are "recaptured" during a recount. That would lower the rate of disqualified ballots.

"He does not look at what the rate of error is when you get done with the entire process," Kozinski said.

Other judges noted that the same type of detailed error analysis has not been performed on other voting technologies.

Los Angeles election officials have taken issue with Brady's findings, although they did not speak at Monday's hearing. The Los Angeles County elections office has filed a friend-of-the-court brief asking to have the Oct. 7 election date reinstated.

"I'm not at all convinced that the study was a fair and accurate representation of voting patterns," said Kristin Heffron, chief deputy registrar/recorder for Los Angeles County.

Heffron argued that many of the ballots Brady classified as invalid may have simply included races where people intentionally abstained.

Heffron added that Los Angeles County has gone to great lengths to ensure that its punch-card system performs well, including having demonstration machines at the polls for new voters and doing thorough maintenance after each election to make sure the equipment works properly.

The county has a "Got Chad?" educational campaign to make sure people remove hanging chads before placing their ballots in the election box.

Brady counters that his study found the problems go beyond intentional abstentions. He analyzed three counties, Fresno, Marin and San Francisco, that switched from punch-cards in 1996 to other voting systems in 2000. The rate of disqualified ballots in all three counties dropped after the change.

"We see that in county after county," he said. "Sometimes it just becomes obvious that something is going on here."