Behind the Controversy: A Primer on U.S. Presidential Exit Polls

By Mark Lindeman and Rick Brady

Part I. A Brief History of U.S. Exit Polls

The internet buzzed before noon on election day 2004 with news that exit polls showed Democrat John F. Kerry leading Republican incumbent George W. Bush in several crucial battleground states. Although warned midafternoon by the exit pollsters Edison Media Research and Mitofsky International of an apparent bias in the early waves of exit poll results, news anchors and commentators hinted at an impending Kerry victory.

When it became clear that Bush had won in the official vote count, dark allegations proliferated. On one hand, Republican consultant Dick Morris contended that the exit polls had been tampered with to give Kerry election day momentum. On the other, various observers argued—and some continue to argue—that the exit polls indicated massive vote fraud. Steve Freeman asserts that the exit poll data “indicate . . . a discrepancy on the order of 10 million votes” and “point sharply toward a corrupted count.”

Like many others, we found ourselves raising questions about exit polls that had never previously occurred to us. In many cases the answers were not readily available. One common question—how accurate have exit polls been in the past?—turned out to be very difficult to answer, although we know that one answer cited by Wikipedia (“‘usually accurate within a fraction of a point’”) is wrong. In this three-part article, we strive to provide a concise overview of exit poll history and practice (primarily in the United States), with the particular purpose of equipping readers to assess the debates about 2004. We begin here with a brief history of exit polls.

Exit polls have at least two purposes—to help project election outcomes, and to inform analysis by, in the words of the Edison/Mitofsky website, “deliver[ing] rich details of who voted and why.” Broadcast media sponsors have put considerable emphasis on speedy projection, allowing them to project or “call” state outcomes earlier than if they relied upon vote “quick counts” alone. Early projections also help the media to plan their coverage.

Analysis is at least as important as projection. All media sponsors, as well as academics and other observers, eagerly exploit the ability to make assertions about (say) what percentage of Catholics voted for the Democratic candidate, or the role of “moral values” in the outcome, based in part on exit poll data.

Some people have urged a third purpose for exit polls—as an independent check on the accuracy of election counts—and, certainly, in 2004, exit polls figured in debates about counts in the Venezuelan Chavez recall referendum, the Ukrainian presidential runoff, and the parliamentary elections in Azerbaijan. In the United States, major exit polls have never been designed to verify vote counts, although recently some local groups have experimented with “parallel elections” for this purpose.

The first exit poll was conducted in 1940, but that early experiment in Denver was limited and did not catch on. The exit poll era really began in 1967, when CBS attempted its first large-scale exit poll in the Kentucky gubernatorial race and achieved results that were reportedly quite accurate. CBS expanded exit polls to twenty states for the national presidential election of 1968, for which data files apparently still exist. NBC apparently conducted its first exit poll in 1973; ABC, in 1980. Initially, exit polls were used only for on-air and post-election analysis.