Why Respond to Exit Polls?

By Robert Y. Shapiro and Lisa Ferraro Parmelee

Election day voters’ polls are commonly called “exit polls,” since the data are collected from voters as they “exit” their places of voting. The person best known for developing exit polls was a statistician named Warren Mitofsky, who died last September 1. According to Richard Morin, then polling director for The Washington Post, Mitofsky “introduced exit polling to the world, including to countries new to democracy and free elections. His reputation for accuracy and independence—a reputation he fiercely guarded—made his exit polls the gold standard with which election results were compared and confirmed.” And yet exit polls have been surrounded by controversy in recent years, with some questioning the accuracy—and hence the value—of the data they yield.

The lead article in the November 2004 preview issue of Public Opinion Pros was entitled, “Why Respond to Polls? Public Opinion Polling and Democracy.” In such discussions of the importance of responding to polls and the need for high response rates to improve the quality of the data, exit polls are included—if they are included at all—as an afterthought. We tend to assume that exit poll data will be better than data from other polls anyway, since they are surveys of actual voters leaving their voting places, and they have response rates of 50 percent or higher. Well, it turns out these 50 percent response rates alone are not high enough.

After enjoying for many years a reputation for “uncanny accuracy”—a term Jim Lampley used to describe social research in The Huffington Post—the exit polls were first cast into doubt by the chaos surrounding the presidential election of 2000, when the media twice declared the winner prematurely, based on exit poll data, before declaring the race too close to call.

In 2004, there was all sorts of confusion about whether it looked like John Kerry would beat George W. Bush. The real confusion was in the reporting of the early exit polls, which everyone should have known were not likely to be accurate since they were warned by Edison Media Research and Mitofsky International, who were running them. But by election night, when the third wave of the exit polls came in and the data were weighted for certain sampling characteristics—but not yet for the actual counted vote in each state—the exit polls were still off in a way that very consistently overestimated the Kerry vote and the number of Democrats voting nationally. We see this net Democratic bias in a great many states, which casts doubt, prima facie, on claims of errors in vote counts in any particular states. Mark Lindeman goes into greater detail on these discrepancies in this issue of Public Opinion Pros in his article, “Condemned to Repetition: The 2006 Exit Poll Controversy,” and discusses how controversy on these grounds continues to haunt the exit polls from the national elections of 2006.

The main credible claim that the 2004 exit polls were not correct came from the exit pollsters themselves, who reported there were problems having to do with interviewer characteristics (young, Democrat-friendly?) and interviewer training, plus speculation that Democratic voters were more favorably predisposed than Republicans to respond to the polls. Given that it would have been in the pollsters’ interest to claim their polls were right and the vote counts wrong, their admitting that the polls were basically, and quite frankly, sloppier than they should have been lends credibility to others' claims of inaccuracy.

In the future, the exit pollsters have, of course, said they will do better, by learning from their mistakes. But there probably would not have been these problems for estimating the election results if the exit polls had a higher response rate, like 70-80 percent or more, instead of just over 50 percent.

So, why should voters respond to exit polls? First and foremost, they provide data to explain, not predict, election outcomes. Making early predictions on election day is entertainment and pseudo-news. In fact, as Paul Lavrakas argues in “Why Our Democracy Needs Accurate National Exit Polls,” though exit polls are done by and for the news media, their purpose and importance lie less in helping to project election results early and more in helping to explain the results afterward—that is, explaining who voted for whom, and why.

Another reason to respond to exit polls was, up until recently, more relevant in other countries than here. In newly developing democracies that have histories of authoritarianism and corruption, freely conducted exit polls have provided a way to check and to validate the official election results—a way to keep the vote counts honest, or to reveal fraud in cases where fraud has occurred. This happened during the political transition in Yugoslavia where fraud was found, and in recent elections in former Soviet states. Debates have also surrounded both national and local elections in Mexico. Granted, there can be problems with biased exit polls, as controversy in Venezuela has shown, and as we saw front and center in the 2004 election in the United States. In general, however, such independent vote counts are all for the good; indeed, in “Warren in Mexico: Elections Become Citizens' Events,” Ulises Beltrán describes how Mitofsky’s work in that country contributed greatly to the construction of a reliable electoral system there.

Finally, exit poll data are utilized by a very wide range of scholars, teachers, and other researchers and analysts. This information is used very broadly and extensively by political scientists in the study of public opinion, voting, elections, and the political process. These data have also been used by sociologists in the study of American society, by public policy specialists in analyses of policymaking, and by journalists in informing the public and its leaders about the opinions of the citizenry and the choices made by voters. They are widely used by students in preparing dissertations, theses, research papers for courses, conferences papers, and publications in the study of politics and society.

The exit polls are crucial to this research for two reasons. First, they provide data on the opinions and behavior of actual voters, not "likely" voters as identified from polls that ask their respondents to report—often unreliably—whether they will vote or have actually voted (depending on whether the poll is conducted before or after a particular election). Second, the exit polls constitute the only standardized and regular election polling done in every state of the United States, plus the District of Columbia, in presidential election years. This is particularly important for studying voting behavior in the context of competition for states' votes in the Electoral College.

Here are just a few examples of articles that have been published in the journal Political Science Quarterly, reporting on the exit poll data gathered in connection with national elections in 1988, 1992, 1996, 2000, and 2004. They help illustrate the range and depth of research made possible by exit polls:

• "The 1988 Elections: Continuation of the Post-New Deal System" was written by Everett Carll Ladd. The late Dr. Ladd was the world-renowned director of the Roper Center for Public Opinion Research who wrote widely on American politics. Among several major articles he wrote on elections that were published in Political Science Quarterly was this 1989 analysis of the 1988 election, which used exit poll data to compare the presidential choices of whites, blacks, and Hispanics in specific southern states. Ladd also used exit poll data to examine how other social groups voted, and how these groups differed in their party identification and their ideological self-identification. 

• In "The 1992 Vote for President Clinton: Another Brittle Mandate?" also by Dr. Ladd, the author used exit poll data from the 1992 election to focus on patterns of voting by ethnic and religious group and by gender, and to examine the possible effect of other factors, such as education level and general outlook toward government.

• A third article by Dr. Ladd, "1996 Vote: The 'No Majority' Realignment Continues," provided his analysis of issues illuminated by exit poll data gathered from voters on election day in 1996. Here Ladd used exit poll data to see how the publics in different states differed in their attitudes toward calls for "less government." He also used them to examine how religious service attendance was related to voters' presidential choices, their partisan identification, and their self-described liberal/conservative ideologies. Other extensive exit poll data on social groups and presidential and congressional voting were reported as well.

• "The 2000 Presidential Election: Why Gore Lost" is a 2001 article by widely respected Rutgers University political science professor Gerald M. Pomper, who provided an analysis of the 2000 presidential election. Dr. Pomper used exit poll data to compare presidential voting by different social groups in the 1992, 1996, and 2000 elections, as well as to examine how perceptions of national economic conditions and voters' financial situations influenced voting, and how other issues and evaluations of the candidates' personal traits affected voting behavior.

• Three widely respected political science professors from Michigan State University and Duke University were the authors of "The 2004 Presidential Election: The Emergence of a Permanent Majority?" Paul R. Abramson, John H. Aldrich, and David W. Rohde applied exit poll data to conclude that voter decisions were based largely upon retrospective evaluations in which President Bush was perceived by a narrow margin to have been a successful president as well as a leader who would be successful in combating terrorism. They also argued that pundits may have exaggerated the role of social divisions in the election.

Many more articles, as well as a number of excellent books, report and analyze exit poll results and apply them to a wide variety of political, social, historical, and media-related subjects.

In sum, exit polling and the regular archiving and dissemination of exit poll data are extraordinarily useful for academic scholarship, teaching, and all manner of research on American politics and society. In addition to developing and conducting exit polls, Warren Mitofsky took a particular interest in their archiving and use. For many years, the organizations involved in exit polling have provided these polling data without charge to the Interuniversity Consortium for Political and Social Research (ICPSR) and to the Roper Center for Public Opinion Research, on whose board of directors Warren was serving as chair at the time of his death. For many years from now, exit poll data will continue to provide important insights into our society and the body politic.

Why respond to exit polls? In a democracy, voting is, or at least it should be, the ultimate expression of public opinion. Thanks to the work of Warren Mitofsky and the other exit pollsters, we have a means of having our votes be counted twice—once in the voting booth and once on the way out—and to have our opinions resonate well into the future. The more of us who participate in the exit polls, the stronger our voices will be. It is a great opportunity, and, for a good citizen, one not to be missed.

Robert Y. Shapiro is a professor of political science at Columbia University. Lisa Ferraro Parmelee is the editor of Public Opinion Pros. The passages above on the scholarly use of exit poll data were adapted from an affidavit prepared by Dr. Shapiro in September 2006 for the plaintiffs in the case of American Broadcasting Companies Inc., The Associated Press, Cable News Network LP, LLLP, CBS Broadcasting Inc., Fox News Network LLC, and NBC Universal Inc. v. Dean Heller, in his official capacity as the Secretary of State of Nevada, in the United States District Court for the District of Nevada.