The following article was reprinted from Psychological Science Agenda (Vol. 14, No. 3; May/June 2001), with permission of the American Psychological Association (APA) Science Directorate

A Vote for Democracy in Support of the Decade of Behavior
by Geoff Mumford, PhD, APA Public Policy Office
 

Reading the morning paper on November 9, 2000, I was struck by the breaking story on the Palm Beach “butterfly” ballot. Popular press accounts suggested that a change in the layout of the ballot to accommodate visually impaired aging voters in Palm Beach County led to rampant confusion in the voting booth. Nineteen thousand ballots were reportedly invalidated because voters punched a hole that they thought corresponded to their chosen candidate, recognized their error, then punched a second hole to correct the first choice. A second cohort of some 3,000 voters voted for a candidate whose party had little documented support as evidenced by registration rolls. Many complained after the fact that they were confused about which hole corresponded to which candidate. Some suggested that those voters were “illiterate.” But was it really that those voters couldn’t read, or was it a poorly designed ballot?

As the 2000 vote and subsequent recount began to take on the characteristics of an accident investigation, it seemed appropriate to bring some human factors expertise to the scene. The butterfly layout, punch-card systems, and other antiquated machinery were quickly denounced, and a cottage industry of new voting technologies was born. Unfortunately, it wasn’t clear where human factors design principles and usability testing fit into that effort.

This industrious activity was set against several independent efforts to establish task forces, committees, and commissions to examine election reform. For example, the Information Technology Association of America formed a Task Force on Election Systems Reform; the Miller Center combined with the Century Foundation to form a National Commission on Federal Election Reform; and the Carnegie Corporation provided a grant to the Massachusetts and California Institutes of Technology (MIT and CalTech) to fund the Voting Technology Project. These groups assembled leading technology firms, famous former heads of state, constitutional scholars, and legions of the brightest engineers and political scientists to rethink voting behavior, but curiously left psychology out of that mix.

Privately funded efforts were soon overwhelmed by competing legislative mandates and, in January and February, a second wave of election reform washed over Capitol Hill. Dozens of bills were introduced in both the House and Senate as key members of Congress jockeyed for jurisdiction over the issue. In the Senate, the lines of authority were fairly clean, and all bills were referred to the Committee on Rules and Administration. In the House, legislation was referred to several committees including Rules, Judiciary, Government Reform, and Commerce. The House leadership tried to elevate the debate to another level with the formation of a Select Committee on Election Reform, but disagreements over the balance of the proposed committee led to separate efforts by each party with uncertain futures as this issue of PSA goes to press.

In an effort to bring psychology’s perspective to bear on both the science and technology of election reform, APA collaborated with the American Political Science Association (APSA) and the Consortium of Social Science Associations (COSSA) to present a congressional briefing entitled “The Mechanics of Election Reform: From Registration to Results.” The briefing, held on March 16, covered many issues raised in the legislative initiatives that Congress was being asked to consider and served double-duty as the lead event in a series of briefings in support of the Decade of Behavior.

Four leading social scientists-- Michael W. Traugott, David Woods, Charles Stewart III, and Rodolfo de la Garza-- presented research findings covering issues such as trust in government, implicit discriminatory practices, voter turnout, and the human-machine interface to a room packed with legislative staff, federal election officials, and government watchdogs.

Michael Traugott began by reminding the audience that the election process is continually evolving in support of two ideals: increasing participation and ensuring that the voting population accurately reflects the electorate. Traugott reviewed his longitudinal research of individual voter behavior using Oregon’s mail-in ballot. Across three broad categories-consistent voters, consistent non-voters, and intermittent voters-he noted that efforts to turn out the vote have greatest impact on retention of intermittent voters rather than mobilization of new ones.

He also disabused the audience of lay theories about how the world works, as there is a generally held belief that any mechanism that increases turnout automatically benefits Democrats. Instead, Traugott’s research suggests that a variety of factors including education, interest, age, and length of residence affect voter turnout, not partisanship. A partisan bias will be expressed only when a given state’s residents have a concentration of the demographic characteristics that are associated with high voter turnout within a given party. As examples, Traugott suggested that in Illinois, Democrats would likely be favored because they are inherently more interested in politics. In New York, turnout would favor Republicans who have lived in the state an average of 13 years longer than Democrats. And in Florida, the result would be difficult to predict because although Democrats are older (favoring increased turnout), that effect would likely be offset by Republicans who are more highly educated (also favoring increased turnout).

Alluding to future research themes, Traugott noted that voter confidence was severely shaken in the 2000 election, and he questioned what effect that would have on the electorate’s overall trust in government. Further, voters now live in one of the most mobile societies ever, and when they move from one complex voting system to another with no comprehensive voter education program, it is likely that they will continue to wonder whether their votes have been accurately counted.

David Woods carried the banner for human factors, pointing out how often the user of a badly designed system is blamed for making an error and how those errors lead to crises when there is no system in place to allow the user to recover. Woods explained that human factors studies the interaction between people and devices, and the study of interface design has a long and storied history in other high-tech domains such as space exploration, aviation, and medicine. Examining this history reveals that simple rules about how to design systems for usability and accuracy are currently more likely to be engineered into your child’s computer game than into our democratic process. However, Woods pointed out that, following the trend in health care to reduce human error, systems to prevent and troubleshoot error have evolved rapidly to establish usability testing standards with limited resources to help guide purchasing decisions. Analogous systems could quickly and easily be developed and applied to the mechanics of election reform.

Among the most critical features of any user-friendly system is feedback. Woods reminded the audience about the incessant audit trail they get in an e-commerce exchange with companies like Amazon.com. Where is that feedback, and where is the technical support when citizens have questions about their voting transactions? Those are easy problems to fix with the application of human factors design principles, but new issues will likely emerge. Woods concluded by questioning how a recount might look as new technology is introduced. He warned that the nature of a recount might change dramatically when the audit trail is embedded in the opacity of magnetic media rather than the transparency of a punch card held against the eye.

It was clear that Woods’s discussion reached not only the audience in front of the podium, but the one behind it as well. Charles Stewart III, representing the MIT half of the Voting Technology Project (a collaboration between MIT and CalTech), began by admitting that the goal of trying to produce uniform voting technology was “a silly thing to do.” The two universities’ presidents initiated the effort because they felt technology had been “embarrassed” by the 2000 election, and they felt an obligation to restore that tarnished image. But as Stewart revealed, their initial aspirations fell apart when they realized that many of the very basic issues Woods addressed simply hadn’t been applied yet to voting technology. Billed as an effort to bring together engineers and political scientists, it was clear from Stewart’s description that they quickly realized that other social scientists were missing from the equation.

Still, it is their plan to evaluate existing technology, as well as the financing and administration of elections, to find out what works and what doesn’t with the intent of making reform recommendations as early as this summer. The results so far reveal that certain technologies, including optical scans, are associated with relatively low error rates, but that newer touch screen (digital direct registry) devices are associated with high error rates. Curiously, a Florida Task Force examining optically scanned ballots revealed that those ballots were processed using two different systems. The more accurate of the two, referred to as “precinct-level” tabulation, was designed to allow voters an opportunity to correct any errors before leaving the polling station. The spoilage rate for ballots tabulated by this method was the lowest of the five voting technologies used in Florida, leading the Task Force to recommend statewide leasing of this equipment for the 2002 election. While clearly an improvement, the spoilage rate of 0.83% for “precinct-level” tabulation meant that 17,172 votes were still discarded, highlighting the need not only for enhanced technology, but also poll-worker education and training in the systems-context that Woods referred to.

Rodolfo de la Garza rounded out the panel with perhaps the most human dimension, saying it was hard to make an electorate “look like the people” in a constantly changing nation of immigrants. He described the MIT/CalTech effort as like trying to build a rocket, where in his studies of ethnic and minority voters, he’s still dealing with “drivers rather than astronauts.” He suggested it may be a laudable quest to develop a perfect voting machine, but it may result in one that the people he represents won’t use.

He noted that although some of the traditional explicit racial barriers and many language barriers have been removed, the problems now relate predominantly to social class. For example, citizenship requirements keep resident aliens in voting limbo, because an application takes at least 14 months to process and more often 3-4 years. Further, large numbers of non-union, predominantly minority employees aren’t afforded the time off work to vote that their unionized counterparts are routinely granted. In addition, polls tend to close early in cities where concentrations of ethnic minorities live. Finally, because minorities are disproportionately represented in single-parent households, schedule complications serve as yet another barrier to getting to the polls.

He reiterated what Traugott had stated earlier: (1) the voter most likely to vote is the one who voted in the last three elections, (2) modern campaigns target those voters, and (3) the campaign spin can produce dramatic effects. Even though 90% of Latinos supported Gore policies, one third still voted for Bush. What information, asked de la Garza, are those voters acting upon? He also warned that aggressive registration campaigns may have had unintended consequences. Although there has been an explosion of registration in his home of Harris County, Texas, there has not been a corresponding increase in turnout. De la Garza speculated that registration has separated citizens from the election process so completely that by allowing voters to register just about anywhere, there is not any psychological connection to the voting process.

A lively question and answer session followed the presentations, and a full transcript and edited video archive will be posted to the Decade of Behavior website soon. Background papers for the presentations are available now at www.decadeofbehavior.org.

We hope that before the 107th Congress and individual states consider adopting uniform technology standards, they will make every effort to understand why thousands of votes were shot down even where the technology appeared to be bulletproof. More generally, we hope that those responsible for making and implementing election reform policies understand that no matter how sophisticated the technology, voting requires an interaction of human with machine, and a poor fit will inevitably result in errors. Finally, they should recognize that there is a wealth of scientific knowledge available on how to achieve a good fit and minimize the disturbing error rates cited in the 2000 Presidential election.