Touch Screen Voting A Hit; Critics Miss Mark On Security, Study Says

The study finds that these usability concerns cannot be addressed by adding paper trails to e-voting systems, and concludes that most critics have focused on the wrong issues.

"Recent history is clear: the election problem most likely to tilt a close race is not security, but the inability of voters to cast their ballots the way they intended," says Paul Herrnson, principal investigator and a University of Maryland political scientist who directs the school's Center for American Politics and Citizenship. "The hazards of poor ballot design didn't end with Florida's hanging, pregnant and dimpled chads in 2000. But tremendous improvement in voters' abilities to cast their votes accurately and without assistance can be accomplished simply by improving the way ballots are laid out on touch screen and paper-based systems."

The five-year study is the most comprehensive of its kind, focusing exclusively on usability issues and relying on data from field tests with more than 1,500 subjects, laboratory tests and expert reviews. The results and recommendations are reported in the new book, Voting Technology: The Not-So-Simple Act of Casting a Ballot, published by the Brookings Press.

Summary of Findings and Recommendations

Voter Errors

The report describes the findings as both reassuring and sobering. While voters expressed confidence in the systems, all proved vulnerable to various types of voter error, such as unintentionally failing to cast a vote in some races, or worse, actually voting for the wrong candidate.

Recommendations

The study recommends a series of needed improvements to make e-voting more user-friendly, adding that that manufacturers and election officials can readily implement these steps. Also, it calls for educational campaigns to ensure voters and poll workers know what they're doing.

"One of the things we've learned in this study is that training may be even more important than which voting system is used," says Richard Niemi, a report co-author and University of Rochester political scientist. "People don't automatically know how to vote on these or any unfamiliar machines. We saw this with incorrectly marked paper ballots, problems with straight-party voting and the number of subjects that needed assistance."

Voter Satisfaction and Confidence

All six electronic voting systems tested were judged favorably, though subjects identified strengths and weaknesses in each system. Voters responded positively to displays with a high degree of computerization, and prefer systems that give them greater navigation control. Also, voters expressed the most confidence in the paperless touch screen systems to accurately record their votes.

"Most touch screen systems were found to be easy to use, support large and clear type, and the ability to readily change a vote," says Benjamin Bederson, a report co-author and computer scientist at the University of Maryland Institute for Advanced Computer Studies. "This added up to an overall positive response by voters. But there are still glitches, and these must be fixed."

Voter Verification Systems

In separate tests of voter verification systems, the report concludes that these devices "appear to produce modest improvements in voter accuracy," though little additional voter confidence, and warns that adding paper trails to Diebold and some other electronic systems may be more problematic than helpful, in part because of their printers' tendency to jam and break down.

Switching to Paper

Policy makers considering a switch to paper ballot/optical scan voting systems should consider special security problems connected with paper. "The history of the paper ballot in the United States is checkered with ballot theft and ballot box stuffing," the report says. Tampering with touch screen systems requires greater technical skill.

Specific Accuracy Findings

The study found an overall voter accuracy rate of 97 percent. "A three percent error rate sounds good until you consider that in the 2000 presidential race the percentage of uncounted ballots was only two percent," Herrnson says. "Voters did pretty well with these new machines - probably better than they would with older technology - but it's still enough to affect the outcome of a close election."

The accuracy rate dropped to the 80 to 90 percent range as the task got more complicated, such as voting for more than a single candidate in a race, voting a straight-party ticket or making corrections before casting the ballot.

"The most common type of error made by voters was registering a vote for the wrong candidate," says Michael Hanmer, a co-author and assistant professor of government and politics at the University of Maryland and a research fellow at the Center for American Politics and Citizenship. "This is the worst kind of error because not only does a voter's preferred candidate lose a vote, but it may go to the main opponent."

Frederick Conrad, a co-author and associate research professor in the Joint Program in Survey Methodology at the University of Maryland and the Institute for Social Research at Michigan, adds: "We observed that voters can get quite lost in the voting process, and when they do, the chances are greater they will not recover, ultimately voting for no one or a candidate other than they intended."

Assistance and Voter Inequality Findings

Gender: "Women may be more likely to ask for help, but voting error rates suggest that it is men who really need it," the report says.

Certain types of voters need more help in using the machines: those with little computer experience, senior citizens, lower income and African American voters.

Voter Verification System Findings

• There was no significant difference in the performance of the four verification systems tested.
• Paper records received greater approval ratings than the other systems tested.
• The systems tested were all prototypes under development and newer versions may prove superior.
• The report concludes that voter verification systems represent a trade-off and are unlikely to meet the expectations of their advocates. While they offer some hedge against massive fraud, they will add complexity and delays as well. Also, these systems could lead to higher levels of inaccurate votes, to the extent that users have technical difficulty changing their votes.

Recommendations for Election Officials and Policymakers

• Voting systems should be purchased only after election officials have tested their usability, preferably in comparison to other systems under consideration. Small field studies with only a few dozen voters may be sufficient.
• Program ballots should be as simple and straightforward as possible.
• Review use of straight-party devices, as they are not well understood by voters.
• Allocate poll workers to reflect the number of citizens expected to vote in each polling place and with regard to the number of voters likely to be more challenged, such as precincts with a large number of senior citizens.
• Offer voters training to ease the transition to new systems or ballots through the mass media and mail, as well as hands-on demonstrations at shopping malls, county fairs, and other public venues.

Recommendations and Guidelines for Manufacturers

• "First and foremost, voting system designers and manufacturers should emphasize usability engineering in the development of their products," the report says.
• Systems should not provide too much information at once. The "full-faced" system showing the entire ballot in a single display overwhelmed voters.
• Voters need clear feedback on where they are in the voting process.
• Special care should be taken to make it clear when voting is complete.
• Review screens should display full information on a single screen, if possible, including races that the voter has skipped.

Machines Tested

The researchers conducted expert reviews, laboratory tests and field studies comparing five current electronic voting systems and one prototype:

1. paper ballot/optical scan (Election Systems and Software);
2. manual advance touch screen (Diebold AccuVote-TS);
3. auto advance touch screen with paper trail (Avante Voting Systems);
4. zoomable touch screen (research proto-type, designed by Benjamin Bederson, UM);
5. dial and buttons (Hart InterCivic);
6. full-face ballot with membrane buttons (Nedap Election Systems).

The companies and developers provided the machines for these tests.

Subsequently the researchers compared four vote verification/election audit systems: the Diebold AccuVote-TSx with AccuView Printer Module; VoteHere Sentinel; the Scytl Pnyx, VM; the MIT Audio System. The Diebold AccuVote-TS had no verification system and served as the statistical control.

Funders

Funding for the studies was provided by the National Science Foundation, the Carnegie Corporation of New York and the Maryland State Board of Elections.

Research Team

• Paul Herrnson, principal investigator, University of Maryland Center for American Politics and Citizenship;
• Richard Niemi, University of Rochester, dept. of political science;
• Michael Hanmer, University of Maryland dept. of government and politics and Center for American Politics and Citizenship;
• Benjamin Bederson, University of Maryland Human-Computer Interaction Lab, Institute for Advanced Computer Studies;
• Frederick Conrad, University of Maryland, Joint Program in Survey Methodology, University of Michigan, Institute for Social Research;
• Michael Traugott, University of Michigan, Institute for Social Research.