Designing Democratic Government: Making Institutions Work

10. Accuracy and Security in Voting Systems

Chapter 10 ACCURACY AND SECURITY IN VOTING SYSTEMS  Henry E. Brady and Iris Hui “Trust in Paper,” proclaims the May 5, 2007, editorial in The New York Times congratulating Florida for getting rid of electronic voting machines, which in 2006 had “somehow lost 18,000 votes” in Sarasota County. “The new law will eliminate touch-screen voting in favor of the more trustworthy opticalscanning system. Unlike touch screens, optical-scanning machines are based on paper.” So “by next summer Floridians should be a lot more confident that whenever they vote, their votes will finally be counted correctly” (New York Times, editorial, May 5, 2007, A26). INTRODUCTION Reality is more complicated than this rhetoric suggests. Many of the 18,000 lost votes in Sarasota were most likely the casualties of poor ballot design (Frisina et al. 2007; but see Mebane and Dill 2007)—a problem that could have (and has) occurred with paper ballots, most famously with the “butterfly ballot” in Palm Beach County Florida, as well as with electronic systems. And although paper ballots may help with recounts when there are fears that computers could be programmed poorly or maliciously, paper trails do nothing to protect voters from systems that confuse them or fail to record votes in the first place. There are two basic challenges for voting machines: accurately recording people’s votes and keeping them secure once they have been cast by the voter. Both are equally important, but attention has shifted since 2000 from accuracy to security, with the result that at least half the problem is being papered over with simplistic rhetoric. Moreover, efforts to improve voting systems have been stymied by inadequate legislation, the limitations of legal action, and the sclerosis of American federalism. In this chapter we describe the sad course of voting reform, and we provide evidence showing that accuracy must be taken at least as seriously as security. The next chapter in this book, by Charles Stewart III, provides a comprehensive overview of how we might do better.1 ACCURACY AND PUNCH CARDS, SECURITY AND ELECTRONIC VOTING The extraordinarily close election in Florida in 2000 exposed many problems with America’s machinery of democracy, but voting systems, the methods for recording and counting votes, performed especially poorly.2 Confusing ballot forms in Palm Beach County and Duval County led to thousands, if not tens of thousands, of mistakes. The butterfly ballot caused at least 2,000 Democratic voters in Palm Beach to vote for the Reform Party candidate, Pat Buchanan, when they meant to vote for the Democratic presidential candidate , Al Gore (Wand et al. 2001). A voting instruction booklet that said “vote all pages” and a ballot that spread presidential candidates over two pages may have caused thousands of voters in Duval County to vote for two presidential candidates, thus spoiling their ballot for the presidential contest. In many counties in Florida, especially in minority areas, there were substantially fewer valid votes—sometimes as many as 9 percent fewer—in the presidential contest than the number of voters who went to the polls. These ballots without valid votes, called “residual votes,” seemed especially prevalent in counties using Votomatic-style punch cards, central-count optical scan systems, and Datavote punchcards.3 Finally, recounts suggested that Votomatic-style punch-card systems and central-count optical-scan systems were especially prone to varying counts depending upon the interpretation of ambiguous marks such as hanging chads, smudged ballots, and inadequately marked ballots . Many people concluded from these experiences that using different ballots and methods for recording votes led to serious problems with the accuracy and fairness of voting systems. In fact, in Bush v. Gore, the case that decided the 2000 election, the Supreme Court argued explicitly that it was unfair—it violated equal-protection guarantees—to have one standard for counting votes in one county and a different standard in another county.4 Since 2003, another problem, the security and trustworthiness of voting systems has come to the fore, driven by concerns that direct-record electronic (DRE) voting systems, similar to automatic teller machines (ATMs), might be subject to computer hacking or to malfunctions. These DRE systems were initially heralded as “solutions” to the problem of accuracy and fairness: they prevent people from “overvoting” for more than one candidate or proposition ACCURACY AND SECURITY IN VOTING SYSTEMS 249 in a contest. By providing reminders that citizens had not voted in a contest...