In the past four decades, national exit polling has become a standard part of the political landscape at times of major U.S. elections. Because of the methodological and statistical rigor with which these surveys were planned, implemented, and analyzed—by Warren Mitofsky, Murray Edelman, and their colleagues—the data they generated were embraced with confidence by the news media. However, starting in 2000, a series of unfortunate calamities that had little to do with the exit poll methods and much more with the way the data were used (or misused) by some in the media led to a diminishing level of confidence among many regarding the validity of these surveys. This article focuses on why it is important for American democracy that high-quality exit polling be conducted for the major elections in the United States. As criticism of exit polls grows, in particular with claims of their being biased in favor of Democrats, it is paramount that the major media organizations that make these surveys possible renew their commitment to fund them with the resources that will yield reliable and valid data.
When most Americans think of exit polls, they think of the projections of election winners that are made and announced by the major television networks on election night every other November. The most obvious of the information sources on which these projections are based are the data gathered from voters exiting their local polling places in sampled precincts across the nation.
Although many Americans do not appear to understand or appreciate it, the national exit polls contribute more than just vote projections to our election campaign coverage. Exit polls also shed light on the so-called “mandate” of the election, explaining why the various electorates voted as they did. Although the news media more heavily emphasize the former than the latter use of these data, it is the latter that has a much greater impact on political processes in the United States—and it is the greater importance of this second type of information that makes the gathering of accurate data by exit polls truly critical to American democracy.
Since their inception four decades ago, election day exit polls in the United States have come to be the most comprehensive and reliable source of information about who voted for whom in major U.S. elections and why they voted as they did. No other dataset has such large probability samples (well over 10,000 voters in the national sample and more than 100,000 voters in all the state-level samples that were conducted by Voter Research and Surveys [VRS] in 1992, Voter News Service [VNS] in 1994, 1996, 1998, 2000, and 2002, and the National Election Pool [NEP] in 2004 and 2006), and no other large survey gathers data from voters immediately after they voted.
Furthermore, all things considered, no other election-related surveying addresses sampling error, coverage error, nonresponse bias (or error), and measurement error issues as robustly as do these media-sponsored exit polls. Therefore, no other survey data can be used as reliably to break out the vote along a myriad of demographic and attitudinal factors. As Warren Mitofsky noted in 1991 in his first scholarly article on the science of exit polling,
They have become an invaluable tool for election analysis. No longer do scholars and journalists need to wonder about the shape of the election mandate. Exit polls make it possible to correlate votes for a presidential candidate with votes for other candidates in his or her party or a voter’s position on [key issues] with his or her vote for a candidate.
Granted, the emphasis placed on the accuracy of the exit poll data for the purpose of projecting election outcomes is perfectly appropriate for several reasons. First, the major news media that fund the exit polls (since the early 1990s, a consortium of major news media organizations that includes the Associated Press, ABC, CBS, CNN, FOX, and NBC) want accurate data to fulfill their sincere commitment to reporting accurate news. One of the biggest stories, if not the biggest, reported by the media every four years is coverage of who wins the presidential election in the United States. Thus, there is great motivation to be the first news organization to get this story right on the evening of the election or as soon thereafter as possible.
Second, the media want to be accurate so as to avoid the considerable embarrassment of faulty data contributing to a mistake in calling the outcome of an election. (This is all the more concerning because many people are unaware that the exit poll data gathered on election day are not the only source, and often not even the most heavily weighted source, used by the consortium that gathers and processes the data to predict winners of key elections. Other sources include preelection telephone surveys of “early voters” who voted prior to election day; an amalgam of the results from hundreds of state-level preelection polls of “likely voters” conducted by a multitude of survey organizations; actual vote counts in the sample precincts after voting ends on election day; and, eventually, the actual vote counts throughout all counties and states.) Embarrassment aside, such mistakes are dangerous in and of themselves because they are so highly visible, and they threaten the credibility of the entire exit-polling enterprise.
Third, despite a lack of credible evidence that exit poll projections in the eastern part of the United States have any effects on voters in other parts of the country, the media do not want to take any chance on inaccurate projections in any way altering the voting behavior of citizens living in the western regions of the country. Were this to happen, or even to be perceived to have happened, the efforts of the U.S. Congress to devise legislation to restrict the use of exit poll data—which is routine in many other democracies—might become unstoppable.
But even more important than these reasons for exit poll data to be accurate is that journalists and their news organizations need to be able to describe accurately the mandate of the election and not leave it to the self-serving spin of politicians, their handlers, and other partisan pundits. Imagine what might have happened after the 1980 and 1984 elections, for example, if the Reagan spin doctors had been able to persuade the nation of their highly partisan and subjective view that voters picked Reagan over the Democratic challenger because they were enthused and fully supportive of the winner’s conservative agenda. That this did not happen can be attributed, at least in part, to the solid evidence provided by the exit polls to the press, empowering them with empirical data to help characterize more objectively why voters voted as they did.
The power of such evidence is further attested to by the recent, growing claims of conservative forces—whose own agendas often are not well served by reliable surveys of public opinion—that exit polls are biased toward Democratic candidates (and liberal issues) and thus inaccurate. Were these partisan forces not threatened by the exit poll data that allow journalists to describe accurately why elections turned out as they did, they would be far less motivated to attack the polls’ methodologies.
As Dr. Mike Kagay, former pollster for the New York Times, has noted, the Times’s goal in using election survey data, including those from the national exit polls, is to “aid both its reporters and its readers in understanding how the American electorate [reacted] to personalities, the issues, and the events of the [election campaigns].” Without such a valid information source, neither the Times nor any other news organization would have a confident foundation upon which to build the editorial approach by which they characterize election outcomes and what these outcomes likely mean for the nation. More than merely allowing reporters to write interesting, informative, and accurate postelection news stories, valid exit poll data allow decision-makers at news organizations to make confident judgments about the tenor of their entire postelection news coverage. Furthermore, these data allow all parties interested in democracy, politics, and elections—be they conservatives, moderates, or liberals; Republicans, Independents, or Democrats—to understand in great depth what was on the minds of voters in the past election and thereby better plan their future political strategies and policies.
Writing in 1991 about the 1988 Bush/Dukakis election, Kagay observed that the
use of polling data by The Times culminated on the Thursday morning after Election Day when it printed a “Supertable” that was a half-page wide and a full page deep, showing how 102 subgroups of the population had voted in the presidential elections in 1988, 1984, and 1980. It helped to inform the Page-One story on November 10, 1988, by E. J. Dionne, Jr., under the headline “Voters Delay Republican Hopes of Dominance in Post-Reagan Era.”
The exit poll data made it possible for this story to be told, despite the fact that the Republican candidate won the White House. Had shoddy (unreliable) survey research methods been used to gather those data, no prominent news organization would have been able to explain the meaning of the 1988 election with the same confidence and insight.
There are relatively few who understand the science of exit polling that Warren Mitofsky and his colleagues (in particular Murray Edelman, former editorial director of the Voter News Service) developed and brought to maturity. I do not claim to understand fully the statistical underpinnings of how the exit poll data are used in the complex algorithms that have been relied upon by VRS, VNS, and now NEP for the past two decades. But from my personal experience of serving on the seven-person election night decision team at VNS from 1996 to 2002, I know of the very high standards for which Mitofsky and Edelman always strove in their data—including how the data were gathered, how they were analyzed, and how they were used by the media. These standards of excellence stemmed from what they learned under the tutelage of the legendary Morris Hansen at the Bureau of the Census in the 1960s. To this day, it is this legacy for accurate exit poll data that has become so important to the news media on election day.
But the legacy is a greater one in terms of the importance it plays far beyond the immediate postelection news coverage of winners and losers. Exactly who wins and loses an election may matter on a momentary personal level to many citizens, and it may matter in the longer term to some citizens regarding the effect of the policies the winners set in motion; but whether they realize and appreciate it or not, the lives of all citizens in the United States are affected by the way in which the mandates of the elections are framed. Thus, I have long believed that we are very fortunate that (a) it is our news media, not our politicians, that play the major role in framing these mandates, and (b) a good deal of the media’s interpretations of the mandate is based on the insights gleaned from exit poll data. For the news media to do this right they must have high-quality exit poll data. It is paramount for the well-being of our democracy that the consortium of news organizations that fund the national exit polls continue to invest in making those surveys as accurate as they reasonably can be.
That exit polling in the United States has taken on such prominence is largely due to the efforts of those involved in conceptualizing the exit polls, conducting them, and analyzing them. For those who share with me a high regard for the importance of accurate exit poll data for our democratic processes, we all owe a great deal to the intelligence and commitment of our exit poll pioneers, in particular Warren Mitofsky and Murray Edelman and their associates.
Having helped monitor the exit polling “quarantine room” from 11:00 a.m. to 5:00 p.m. EST on Election Day 2006 to prevent any early leaks of the data that day, I experienced firsthand how precarious is the confidence of the NEP media sponsors in the validity of the exit poll data. I fear that if another “disaster” were to take place in 2008 or 2010 or 2012, that rightly or wrongly was blamed on the exit polls themselves, the commitment of some or all of the media sponsors might wane. Were this to happen, and if there were no reliable national exit poll data gathered and disseminated by a reputable survey organization, our democracy surely would suffer.
Paul J. Lavrakas, Ph.D., is a former professor of journalism and communications studies at Northwestern University (1978-96) and Ohio State University (1996-2000), and founding faculty director of the Northwestern University Survey Lab (1982-96) and Ohio State University Center for Survey Research (1996-2000).
Exactly when the first exit poll was conducted is open to debate. Warren Mitofsky believed it took place in November 1967 and was conducted by CBS news for a gubernatorial election in Kentucky. In contrast, Bud Lewis, former pollster for the Los Angeles Times, believed it was in June 1964, when NBC news gathered data from California voters exiting polling places in twenty-one sampled precincts on the day of that state’s primary election.
Kagay, M. R. 1991. The use of public opinion polls by The New York Times: Some examples for the 1988 presidential election. In Polling and presidential election coverage, ed. P. J. Lavrakas and J. K. Holley. Newbury Park, CA: Sage.
———. 1995. The evolving use of public opinion polls by The New York Times: The experience in the 1992 presidential election. In Presidential election polls and the news media, ed. P. J. Lavrakas, M. W. Traugott, and P. V. Miller. Boulder, CO: Westview Press.
———. (2000). Continuing evolution in the use of public opinion polls by The New York Times: The 1996 presidential election experience. In Election polls, the news media, and democracy, ed. P. J. Lavrakas and M. W. Traugott. Chatham House Publishers/CQ Press.
Lavrakas, P. J., J. K. Holley, and P. V. Miller. 1991. Public reactions to polling news during the 1998 presidential election campaign. In Polling and presidential election coverage, ed. P. J. Lavrakas and J. K. Holley. Newbury Park, CA: Sage.
Lavrakas, P. J. and M. W. Traugott. 2000. Why election polls are important to a democracy: An American perspective. In Election polls, the news media, and democracy, ed. P. J. Lavrakas and M. W. Traugott. Chatham House Publishers/CQ Press.
Lewis, I. A. 1991. Media polls, the Los Angeles Times poll, and the 1988 presidential election. In Polling and presidential election coverage, ed. P. J. Lavrakas and J. K. Holley. Newbury Park, CA: Sage.
Mitofsky, W. J. 1991. A short history of exit polls. In Polling and presidential election coverage, ed. P. J. Lavrakas and J. K. Holley. Newbury Park, CA: Sage.
Mitofsky, W. J., and M. Edelman. 1995. A review of the 1992 VRS exit polls. In Presidential election polls and the news media, ed. P. J. Lavrakas, M. W. Traugott, and P. V. Miller. Boulder, CO: Westview Press.