New machines could mean new glitches
An analysis finds that sometimes, trying to use new technology leads to errors.

  
 
By David Damron |  Orlando Sentinel Staff Writer
Posted October 30, 2004

Whether voters punched a card or check-marked a paper ballot, Florida proved in 2000 that the machines people vote on can be as important as the candidate they pick in deciding a close presidential race.

Florida has done much since then to improve voting technology. But an Orlando Sentinel analysis of votes from the Aug. 31 primary and the 2002 governor's race shows that differences in machines and voters' experience in using them could still be a factor in Tuesday's election.

Voters in counties that have switched to new ballot equipment can expect a greater percentage of lost votes on Election Day than those using the same equipment they used in 2000, the analysis found. Problems could be worse in smaller counties, where election staff and voter-education resources are smaller.

About two-thirds of Florida's 10.3 million registered voters live in counties that have changed their election equipment since the last presidential election. About 5.5 million live in 15 counties using new automated-teller-style, touch-screen machines. An additional 1.1 million live in 26 mostly small, rural countries using new optical-scan equipment.

A state report on the 2002 election found that the ATM-style machines produced about two to three times more undervotes instances where no candidate is picked or no choice can be detected on a ballot than the optical-scan machines did.

And a Sentinel survey of Florida's 67 counties in this year's primary showed that even among optical-scan counties, new users of the equipment tend to make more errors than veterans.

"History would show you have a better shot at getting fouled elections the earlier you are in using a new piece of equipment," said Kimball Brace, director of Election Data Services, a leading Washington-based ballot-system expert.

The Sentinel analysis shows the most successful Florida voters Nov. 2, such as those in Orange, Seminole, Volusia and Polk counties, may be those who used the same superior-performing optical-scan equipment that they voted on in 2000.

Goodbye to punch cards

After the 36-day delay in choosing a president four years ago, 41 of Florida's 67 counties left behind punch cards, lever machines, centralized counting systems and hand counts. They switched to new touch-screen machines and optical-scanning systems that counted ballots in the precinct. Both are designed to cut down on mistakes.

In one key way, they worked: Overvotes, where voters pick more than one candidate in a race, nearly vanished in 2002, and again in the August primaries.

In 2000, about 113,820 presidential ballots were overvoted among the 6 million cast statewide. But there were only about 2,092 overvotes from the 2002 gubernatorial race and U.S Senate Democratic and Republican primaries combined.

Florida's overall spoiled-ballot rate also ped from about three spoiled votes for each 100 in 2000 to slightly fewer than one for each 100 in 2002.

However, not all voters and election workers master these new technologies at the same rate, Sentinel analysis shows.

Counties that will use optical-scan machines for their first presidential election Tuesday performed worse in the 2002 general election and this year's primary than counties using the same vote technology from 2000 or before.

"Familiarity certainly breeds comfort," said Diebold Election Systems official David Baer.

Voters in counties that had optical-scan machines for at least one previous presidential election had an average of 4.1 invalid votes for every 100 cast in the Aug. 31 primary. But the average rate of invalid votes was 5.4 per 100 in counties still learning to use their new optical-scan equipment.

In 2002, more than two-thirds of the worst-performing counties with optical-scan equipment were small, rural communities using the systems for the first time.

Even on touch screens, smaller counties fared worse than larger counties in the past two major elections. In the 15 counties using ATM-style touch screens for the first time in 2002, small- or medium-sized counties, such as Pasco, Sarasota and Sumter, showed the highest rate of undervotes.

In 2004, Nassau, Sumter and Indian River counties the smallest to use that equipment the rate of invalid votes averaged 5.25 per 100. In the largest touch-screen counties, such as Miami-Dade and Palm Beach, the average rate of invalid votes was 2.25, less than half as many.

Compared with optical-scan machines in 2002, touch-screen machines were less effective at preventing undervotes. These counties produced about 26,000 undervotes among 2.8 million cast on those systems for governor that year, while optical-scan counties produced only 7,800 undervotes out of about 2.4 million votes cast.

A leading proponent of touch screens, Pasco Elections Supervisor Kurt Browning, said voters are likely hitting problems "as they learn to use" the new machines.

The small difference in error rates among technologies seems trivial until factored against the millions of ballots to be cast Tuesday. Even a difference of one spoiled ballot in 100 could affect a close election.

'There is a learning curve'

The 2000 presidential race was decided by only 537 votes. Even if Florida's more-accurate systems run as smoothly as in 2002, it is likely that no clear choice for president will be detectable by machines on tens of thousands of ballots.

As Florida found out in 2002 and again in the 2004 primary, new machines do not mean clean elections.

"With anything, particularly a new voting system, there is a learning curve," said Bill Cowles, top elections official in Orange County, which is using the same optical-scan system it had in 2000. "We experienced that learning curve years ago, and now these counties with the new systems are hitting their learning curve."