Every close election leads to speculation about what might have happened if events had gone another way, if the candidates had performed differently, or if the voters had been presented with other arguments and appeals. But the presidential election of November 2000 raised an entirely new set of counterfactual possibilities that had escaped notice in the past: suppose voting systems worked better, suppose felons were allowed to vote in more states, suppose registration were easier, suppose absentee ballots were counted differently? Before November 2000, these questions about the details of American electoral machinery were mostly ignored, but they mattered a great deal in 2000. The articles in this symposium explore the implications of the basic rules and the performance of the machinery of American politics.

The symposium begins with Jeff Manza and Christopher Uggen's consideration of felon disenfranchisement. They show that felon disenfranchisement laws have been at least partly driven by racial politics, that 4.7 million felons (about half of whom are African Americans) are currently not allowed to vote, and that only about one-quarter of them are in prison. They estimate that about 35 percent of these felons would vote in a presidential election if given the chance, and 73 percent of them would vote Democratic. The impact of these results for Florida in 2000 would have been a net Democratic gain of 63,079 votes and a final Gore victory margin of 62,542. Ultimately, of course, the question of whether felons should be allowed to vote is one about values, and Manza and Uggen present public opinion data suggesting that at least for felons not in prison, there is public support for allowing them to vote.

Although political scientists have neglected some aspects of voting, they have paid a good deal of attention to registration and voting. They have even been involved in formulating such legislative initiatives as “Motor Voter” to simplify the registration process. In his essay, Benjamin Highton argues that registration reform has probably gone about as far as it can to enhance turnout, and that registration requirements are no longer a primary cause of the demographic skew of the voting population toward those with more income and education. His claims rest on a theoretical analysis of the costs and benefits of voting, and empirical research that has estimated the impact of voting system changes that have affected these costs and benefits. Highton characterizes voting as a low-benefit activity that must have low costs because so many people do vote, and for which even small increases in costs can deter people from voting. The poll tax and literacy tests imposed substantial costs that reduced turnout and disenfranchised blacks and poor whites in the south, but today's registration laws impose relatively uniform costs that are smaller than at any time since the adoption of registration laws in the late nineteenth and early twentieth centuries. With the reduction in costs, the “continued nonvoting by substantial numbers of citizens suggests that for many people, voting remains an activity from which there is virtually no gratification—instrumental, expressive, or otherwise.” Increasing turnout, therefore, requires going beyond registration laws.

The first two papers deal with who votes. The last three consider what happens to the votes of those who do cast a ballot. Before 2000, political scientists had barely studied this topic, but the spectacle of Florida election officials peering at hanging chads and voters claiming that they had mistakenly voted for Pat Buchanan because of the butterfly ballot, led to a spate of work on the topic. Some of it is represented here.

Justin Buchler, Matt Jarvis, and John McNulty consider “Punch Card Technology and the Racial Gap in Residual Votes.” The number of residual votes in an electoral contest is the difference between the number of ballots cast in an election and the number of valid votes in a particular race. Residual votes are the sum of overvotes (where more than one mark appears for a race thus invalidating the vote) and undervotes (where no discernible marks appear for a race). For the “top-of-the-ticket” race, a comparison of residual votes from different voting systems can provide an indication of the performance of the voting systems. Buchler, Jarvis, and McNulty present evidence showing that votomatic-style punch card voting systems not only produce higher rates of residual voting than other technologies, but also they disproportionately affect minority voters, who are less likely to have their votes properly recorded with these kinds of punch card systems. “The result,” they say, “is a system that gives greater weight to votes cast by nonminorities, through a mechanism very similar to literacy tests.”

Most of the publicity about voting in Florida and the Supreme Court case, Bush v. Gore, which ultimately decided the 2000 presidential election focused on undervotes, and one particular kind at that—cases where Votomatic style punch cards had been only partially punched through making it ambiguous whether or not a vote had been cast. In the recount of the ballots funded by a consortium of news organizations and conducted by the National Opinion Research Center, the researchers concluded that a narrow Gore victory was possible using some standards for counting undervotes, but the more likely result using the Supreme Court's guidelines was a narrow Bush victory.

Walter Mebane takes a different tack in his paper. He uses ballot-level data—actual data on the characteristics of individual ballots—from Florida for the November 2000 election to show that of the 94,000 discarded overvotes, as many as 57,000 were “true votes” intended to go to either Bush or Gore. By making some plausible assumptions, such as that the pattern of votes found for the Senate and the presidential race on those ballots that counted should be a guide for voter intentions for those ballots with overvotes in the presidential race, Mebane concludes that “If the best type of vote tabulation system used in Florida in 2000—precinct-tabulated optical scan ballots—had been used everywhere in the state, Gore would have won by more than 30,000 votes.” Mebane also proposes that we should make a practice of using ballot image data to evaluate voting systems to determine if these systems are properly capturing voter intentions.

Kosuke Imai and Gary King consider still another source of dispute surrounding the 2000 Florida results—the consideration of overseas absentee ballots. After the election, New York Times reporters recovered the envelopes in which these ballots were mailed and found 680 cases where there were clear violations of election laws which should have disqualified those ballots. But the envelopes had been separated from the ballots months before—making it impossible to determine directly whether these ballots made the difference in an election ultimately decided by 537 votes. Imai and King were brought in to use statistical methods to estimate their impact. They conclude that discarding the invalid overseas absentee ballots would have certainly narrowed the gap between Bush and Gore, but there was only a small likelihood that Gore would have won. This likelihood increased to beyond 50 percent, however, if the results for the disqualified overseas ballots were combined with other scenarios.

What can we conclude from these papers? Everyone should come to his or her own conclusions about the events in Florida, but I think that three conclusions about the American electoral system more generally come through very clearly. First, our voting systems and the machinery of elections are shabby, in disrepair, and need work. Second, the wretched history of using electoral laws such as poll taxes and literacy tests to disenfranchise African Americans should make us think hard about how we treat the voting rights of felons who are disproportionately African American. Third, even after the voting machinery is fixed and we find a way to deal fairly with felons, we will still face a disconcertingly low rate of turnout that cannot be fixed by tinkering with voting machinery or registration laws.