1. What Is the Incidence of Voter-ID Requests?

Half of all voters are asked to show photographic identification at the polls. In the 2006 sample, 49% of respondents reported that the poll workers asked them to show photo ID when they voted. In the 2008 sample that figure had risen to 56%. This rate is strikingly high. In 2006 only two states actually required photo identifications; the other Voter ID states allowed poll workers to request ID. In other words, poll workers are using their discretion and asking voters to show photo ID.

The incidence of requests for ID varies considerably across states, with requests exceeding 90% of all voters in some states and below 10% in others. The main reason for the variation is state law. In the Voter ID states, nearly 80% of voters on average are asked to show ID. In the Non-ID states, less than 20% are asked to show ID (see Ansolabehere Reference Ansolabehere2008). This pattern suggests that simply allowing poll workers to request voter ID triggers a very high rate of requests for photographic identification at the polls. The 2008 Super Tuesday survey probed whether voters showed photo ID because that was convenient or because that was what the poll worker requested. Approximately half of those who showed ID said that they did so because photo ID was convenient, but half said that the poll worker asked for photo ID.

The incidence of requests for photo ID at the polls also varies across region. In the Northeast 22% reported that poll workers asked for photo ID in the general election, compared with 65% in the South. Approximately 45% were asked for ID in the West and Midwest.

Regional variation in requests for photo identification at the polls mainly reflects the laws. Southern states tend to have more restrictive voter-ID laws. Only Mississippi and North Carolina do not ask for voter ID at the polls. The Northeast is the opposite. Only Connecticut does allow poll workers to request ID. As a result, relatively few voters are asked for ID in the North, but most voters are asked for ID in the South. The pattern of ID laws is less uniform among the Western states and among the midwestern states.

For voting-rights advocates, racial differences in these data are of greatest concern. Differential application of election laws at the polling places was the target of the Voting Rights Act and civil rights litigation from the 1940s onward. The Voting Rights Act expressly forbids the use of “tests” at the polls, and the U.S. Supreme Court has struck down the use of poll taxes as violations of equal protection. Literacy tests, poll taxes, and other administrative procedures were widely used before the 1960s to keep certain groups, especially southern blacks, from voting. There is the very real possibility that voter-ID laws provide an opening for the reemergence of such practices.

Both the 2006 and 2008 surveys show considerable racial differences. In the 2006 general election, 47% of white voters reported being asked to show photo identification at the polls, compared with 54% of Hispanics and 55% of African Americans. In the 2008 Super Tuesday primary states, 53% of whites were asked to show photo ID, compared with 58% of Hispanics and a staggering 73% of African Americans. These racial differences persisted upon holding constant income, education, party identification, age, region, state laws, and other factors (see Ansolabehere and Persily Reference Ansolabehere and Persily2008 for details). Opponents of voter-identification laws charge that they amount to a new form of “test” or “tax.” These surveys provide the first individual-level data that poll workers commonly ask voters for photo identification, even in places where they are not allowed to. The data further show that poll workers do not administer this procedure fairly or without regard to race, which raises the important possibility that in practice voter-identification procedures violate the Voting Rights Act.

2. What Is the Effect of Voter ID Requests on Access?

The immediate voting-rights concern with photo-identification laws is that they prevent people from voting and affect the access to the polls of different groups or classes of voters. How many people were denied the vote as a result of voter-identification requests?

The answer is—very few. If respondents reported that they were asked to show photo identification, the 2006 and 2008 surveys probed whether the respondents were then allowed to vote. In the 2006 survey, out of 22,211 voters only 25 said that they were asked for identification and, then, disallowed from voting—that is one-tenth of 1% of the sample of voters. In the 2008 survey, three out of 2,564 respondents said that they tried to vote but were not allowed because of voter ID, a fraction of a percent.

This is an exceptionally low rate of denial of access to the vote. Some of these denials may have been legitimate, and some may have been erroneous. But the actual denials of the vote in these two surveys suggest that photo-ID laws may prevent almost no one from voting.

One rejoinder to these findings is that the very presence of ID laws may discourage some voters from even attempting to vote. To examine this possibility we asked registered non-voters in the 2008 survey why they did not vote. This question parallels the question on the Current Population Survey with the addition of “I did not have proper identification.” Of the 1,113 non-voters in the survey, four cited this as a reason, and these individuals cited other reasons as well—“bad weather” and “forgot to vote.” All told, then, only seven out of 4,000 people (less than two-tenths of 1% of the electorate) could be considered non-voters at least in part because of voter identification.

Voter ID does not appear to present a significant barrier to voting. Although poll workers widely request ID, such requests rarely result in voters denied the franchise. Moreover, very few people chose not vote in the 2008 primaries for lack of identification. Although the debate over this issue is often draped in the language of the civil and voting rights movements, voter ID appears to present no real barrier to access. An important caveat accompanies these findings. These surveys covered a midterm election and presidential primary elections. Although the particular elections drew relatively high numbers of voters, the turnout was not nearly as high as in presidential general elections. High turnout is widely thought to put additional strains on the election-administration system. Some have argued that the denial of access is most likely to occur in those circumstances. Whether that is true is the subject for further study. At the time of this writing, the 2008 Cooperative Congressional Election Survey (CCES) and a separate survey sponsored by the Pew Charitable Trusts are planned for the 2008 general election and will replicate these questions and methods.

3. Voter ID and Confidence

The justification for photo-identification requirements rests on concerns about voter fraud. Requiring all voters to show photo identification at the polls may be justified in order to prevent people from impersonating actual voters or committing other sorts of voter fraud. Voter fraud has been perhaps been the most elusive election-related phenomenon on which to get hard facts. Large majorities believe that fraud occurs at least somewhat often in elections, but social scientists have been unable to develop unambiguous measures of the incidence of fraud, and legal cases find very little hard evidence on the matter. In fact, the case in Crawford v. Marion County Board of Election (553 US [2008]) produced no actual instances of voter fraud occurring. Even still the courts upheld the law.

The justification for the law in the majority opinion of the Supreme Court arises not necessarily from the fact of fraud but from the perception of or belief in fraud. The opinion in Purcell, and later Crawford, argues that the government's interest in limiting corruption or perceived corruption of the electoral process must be weighed against the constitutionally guaranteed right to vote.Footnote ²

The Court's concern with the perception of corruption certainly receives support from the public opinion surveys. In a follow-up survey to the 2006 CCES, the 2007 CCES asked about people's perceptions of fraud. Over half of respondents felt that voter fraud, as occurs when ineligible people vote, occurs “somewhat often” or “very often.” Similar majorities felt that election fraud, such as occurs when ballots are tampered with, also occurs somewhat or very often. Only about one in 10 respondents believe that such fraud occurs “rarely” (See Ansolabehere and Persily Reference Ansolabehere and Persily2008, 1754).

The opinion of the Court, however, asserts that a widely held belief that fraud occurs often will erode the legitimacy of elections. People will come to view elections as illegal or not reflecting the will of the people. This will discourage people from voting, further weakening the democratic process. Stronger identification laws offer a cure, as they can reassure voters that only legitimate votes are cast.

These claims are testable. First, those who believe that fraud is common ought to be less likely to vote. Second, those in states with stricter ID laws ought to perceive less fraud in their elections. Neither appears true.

The first claim asserts a mechanism through which fraud lowers voting. Those who see fraud as occurring often in elections, it is argued, will view elections as less legitimate and their votes as less effective. As a result, those who believe that fraud is common ought to be less likely to vote. In the 2007 survey, of those who thought fraud a very common occurrence, 47% voted, and of those who thought fraud rare, 44% reported voting.Footnote ³ Controlling for education, income, partisanship, and other factors did not change this non-finding. Belief in the frequency of election fraud is uncorrelated with propensity to vote.

The second claim asserts a solution—stricter identification will shore up confidence in the process and, hence, turnout. This claim also lacks empirical support. Those voters living in states with stricter identification laws did not report higher levels of confidence or higher rates of voting than those living in states with relatively weak identification rules. In states with the weakest ID rules, 26% think fraud occurs very often and 10% think it occurs rarely. In states with the strictest ID rules, 29% think fraud occurs very often and 9% think it occurs rarely. Moreover, individual voters who were asked to show ID at the polls in 2006 were not more likely to assert higher levels of confidence in the electoral process or higher intentions to vote than those who were not asked to show ID (Ansolabehere and Persily Reference Ansolabehere and Persily2008).

These twin findings reveal that ID laws will have little or no effect on the confidence in the electoral system or the belief in the incidence of fraud. Those beliefs, wherever they come from, are no different when a stricter ID law is in place and enforced than when less invasive voter-authentication methods are used. These findings also call into question the assumptions underlying the majority's opinion in Crawford. People may think voter fraud occurs often, but that belief appears disconnected from the likelihood that someone engages with politics and votes.