Concerns about voter fraud ordinarily simmer down by this point in the election cycle. But the Bush administration's U.S. attorney firings have brought them to the fore again, because some of the ousted prosecutors had declined to bring indictments for alleged voting irregularities in their states. The most familiar trope pits advocates of greater access (usually Democrats) against those (usually Republicans) worried about the "fraud" committed when ineligible votes are counted. The latter urge, among other reforms, more stringent voter-identification provisions, while the former worry about disenfranchisement, particularly for racial minorities and the poor. What has been missing from this debate, awash in anecdotes and conflicting definitions, is any data that could give a sense of the scope of either the problem of fraud or the potential for disfranchisement.
The data vacuum may exist either because fraud rarely happens or because successful fraud goes undetected. Any process that involves the participation and counting of millions of people will include the counting of some ineligible participants and the failure to count eligible ones. When it comes to voting, however, these problems are generally brought to light only when a close contest, followed by litigation, forces the issue out in the open. It's in these very close elections that small changes in vote totals affect outcomes.
Fraud is one of those things for which social scientists simply do not have the tools for systematic measurement. We know fraud happens sometimes—particularly in the context of absentee ballots, if the number of prosecutions is any indicator. But we have not figured out a way to investigate it on a national scale.
We can, however, gauge the effects of voter-identification rules and other procedures that seek to prevent fraud. How many people are asked to show identification? How many are stopped from voting? Are blacks and whites treated differently? Such questions, while often answered with hyperbole, are amenable to empirical observation.
In the 2006 election, a consortium of 37 universities joined forces to establish a very large sample survey, the 2006 Cooperative Congressional Elections Study. The survey asked more than 36,000 people an unprecedented battery of questions about their experience of voting in 2006. The initial findings are somewhat of a surprise. They give something to—and take something from—each side of the debate.
The study's most unexpected finding is that half of the survey respondents said they were required to present photo ID, even though in 2006 the laws in only Indiana and Florida required this of all voters. The data show a stunning lack of uniformity in the use of voter-identification requirements. Voters in Midwestern and Western states were asked to show photo identification at approximately the same rate (40 percent to 45 percent), but poll workers asked for such identification very frequently in the South (65 percent) and rarely in the Northeast (22 percent). Even within states, the application of the rule varied greatly from region to region.
Did requiring people to show photo identification stop them from voting once they showed up at the polls? Both the pro- and anti-ID camps suggest that they should, either because they catch illegal voters or because they discriminate against some people. The survey results suggest otherwise. Just 23 people in the entire sample—less than one-tenth of 1 percent of reported voters—said that they were asked for photo identification and were then not allowed to vote. It is hard to analyze such a small group further, but these 23 were not concentrated in the South, where poll workers most commonly request voter identification, and they came from all walks of life. Some fear that voters subjected to ID requirements are more likely to cast so-called provisional ballots, which are less likely to be counted. The data provide little support for that argument either. As for fraud, the data suggest that either identification requirements are wholly ineffective at stopping it, or that people who are ineligible to vote rarely try.
If your concern is disparate racial impact, ID requirements don't look so good, even when we account for regional differences. African-Americans were asked to present photo ID more often than whites—54 percent of the time versus 46 percent. On average, African-Americans also had to wait somewhat longer in line to vote, though the waiting times were very short for most people. Eighty-five percent of whites reported waiting less than 10 minutes to vote, compared with 75 percent of blacks and 82 percent of Hispanics. Although only 1 percent of voters said they waited over an hour to vote, black voters were more likely to be part of that group. There does not, however, appear to be racial bias in either the share of voters who were turned away from the polling place or the share that were given provisional ballots because they had ID problems.
Of course, the usual caveats with respect to survey research apply. People may misremember or lie about their voting experience—the turnout rates evident in surveys always exceed those represented by the verified vote count, for example. They may not understand the difference between provisional and regular ballots, and when asked about photo ID requirements, they may simply remember that the polling-place administrator asked them to produce some form of ID, with a photo or not. And it is possible that certain people whom the survey didn't capture are deterred from voting when they know that the lines will be long or that they will need to present ID. We do not expect lawyers who are preoccupied with the exceptional case of the harassed voter who waited eight hours to vote to find solace in national averages, however reassuring.
To return to the finding that half of survey respondents said they were required to present photo ID despite the rarity of relevant state laws, this result is instructive because of what it suggests about how rules are not followed. Both sides of the voter-fraud debate assume that registration requirements matter because poll workers will apply them as instructed. In truth, this country continues to rely on barely trained volunteers to administer the actual process of voting. Any plan to combat fraud or increase access, let alone upgrade voting technology, must take into account that it will be administered largely by amateurs.