[an error occurred while processing this directive]

REVISITING THE U.S. VOTING SYSTEM: A RESEARCH INVENTORY

November 27-28, 2006

Convened by the American Association for the Advancement of Science

Paul S. Herrnson (Michael W. Traugott)

A SYNOPSIS OF A RESEARCH PROJECT ON VOTING TECHNOLOGY AND BALLOT DESIGN

A major lesson derived from the 2000 presidential election is that the manner in which voters cast their ballots is important. Many polling places nationwide employed outdated technology, including unreliable punch-card ballots and mechanical lever machines; only about 40 percent f the electorate used more modern computerized technology, such as optical scanning systems or direct recording electronic (DRE) systems with "ATM-style" touch-screen voting or other types of voter interfaces.

Congress responded to the problems associated with the 2000 election passing the Help America Vote Act (HAVA), which required (among other things) the adoption of more modern technology by every state. Although most of the attention of researchers and the media has focused on issues related to ballot security, one of the major problems occurring in recent American elections thus far have involved the interface between voters and voting systems.

The Project to Assess Voting Technology and Ballot Design, funded by the National Science Foundation and the Carnegie Corporation of New York, involves political scientists, human factors psychologists, and computer scientists who are conducting a multifaceted assessment of voting systems. We tested six voting systems and three ballot designs. These were tested because, collectively, they represent the broad array of design principles incorporated into contemporary voting interfaces. The voting systems tested were: the ES&S Model 100 (a paper ballot/optical scan system), the Diebold AccuVote-TS (touch screen), the Avante Vote-Trakker (touch screen with automatic advance mechanism and a voter verifiable paper record), the Zoomable prototype (touch screen with a unique zooming interface), the Hart InterCivic eSlate (mechanical buttons and dial interface), the Nedap LibertyVote (a full-face ballot) (see figures 1a-1f). The ballots were: a standard office-bloc ballot, and office bloc-ballot with a straight-party feature, and, in the case of the LibertyVote system, a party-column ballot (as this system could not accommodate a straight-party feature). The systems were analyzed using three research approaches: review by 12 human-computer interaction (HCI) experts, usability experiments conducted with 42 voters in a laboratory setting, and usability experiments conducted with 1,540 voters under field conditions similar to those faced by voters on Election Day. In the research with voters, we constructed a simulated election, in which individuals cast their votes on each voting system device in random order. A final component of the project used natural experiments to examine aggregate voting statistics.

Our major findings are concerned with voter satisfaction and the need for help, the ability to vote as intended, and the impact of voter characteristics on those factors.

Voter Satisfaction and the Need for Help

The HCI experts and voters in the laboratory and field experiments evaluated most of the voting systems favorably, but identified significant weaknesses.
Voters had more confidence in the abilities of paperless touch screen systems to record their votes accurately than they did in other systems, including the system using a paper ballot.
Voters often felt the need to ask for help in completing the voting process.
Voters are more comfortable with voting systems that let them exercise more control over the voting process than those with automatic advance mechanisms (e.g. Avante).
Most voters are not more intimidated by DRE voting systems with computerized interfaces (e.g., Diebold, Avante, and Zoomable) than those with mechanical interfaces (Hart InterCivic and Nedap)
The more physical movements it takes to vote, the more time voting takes, and the less satisfied are voters.
Voting systems that received the lowest ratings are the same ones on which voters required the most assistance when voting.
Voters have higher levels of satisfaction and need less help when voting with a standard office-bloc ballot than with a ballot that has a straight-party feature or a party-column ballot.
Voters felt it was less difficult to cast a write-in vote on the paper ballot/optical scan system than on the DRE systems.

The Ability to Vote as Intended

When using an office bloc ballot, voters cast 97% of the simple (one-office, one candidate) votes as they intended.
Most voter errors were the result of selecting an unintended candidate, particularly a candidate proximate to the intended vote choice. They did not result from undervoting or overvoting.
Voters’ abilities to cast their votes as intended declined when they used a ballot with a straight-party feature or sought to change a vote from the candidate they initially selected. The decline in accuracy varied by voting system.
The most frequent mistake when casting a write-in vote was failing to fill in the oval on the paper ballot/optical scan system (which signals that a write-in has been cast).
Voters often failed to verify a printed record of their touch screen choices (paper trail) because they had trouble coordinating between the touch screen and the printed record.

The Impact of Voter Characteristics

Factors related to the digital divide are not consistently related to voter confidence or satisfaction using the voting systems.
Voters with little computer experience, the least educated, senior citizens, and individuals whose first language is not English had a greater need for help on most of the systems.
Older, less educated, minority voters, and infrequent computer users were more likely to cast votes in ways they had not intended.

In summary, voting systems and ballot design have an impact on voter satisfaction, the need for help, and the ability to cast a vote as intended. These outcomes vary in accordance with voter characteristics, including those related to the digital divide.

Team members are Paul S. Herrnson (PI), Richard G. Niemi, Michael J. Hanmer, Benjamin B. Bederson, Frederick G. Conrad, and Michael W. Traugott.