The Future of Exit Polling

By Warren Mitofsky

We hear more and more about the use of exit polls to report the first results of elections around the world. When there are no exit polls, as in Iraq for December’s parliamentary election, we wait days—or even weeks—for the vote results. The longer the wait, the more suspicious citizens get about the transparency of the vote count. But all exit polls are not equal. Some, I should say many, in emerging democracies are of dubious technical merit, and some are blatant attempts at partisan manipulation of the thinking about the election.

It is a different story in the United States, where there are mounting obstacles to conducting reliable exit polls. Here we have everything from increased absentee voting to new anti-exit poll laws and declining response rates.

I started conducting exit polls for CBS News in 1967. All of my early work was in the United States, but in recent years I have done many exit polls in other countries as well. In the 1960s and ’70s, exit polling was used almost exclusively for election analysis. Projections in those days came from "quick counts," which are early reports of actual vote returns collected right after the polls close from a sample of precincts. There were few problems conducting exit polls. Hardly any voting officials interfered with them, and no laws complicated the interviewing. The public was cooperative, and response rates were in the 70 to 75 percent range. 

All that changed following the 1980 election. An unanticipated early projection on television of Ronald Reagan’s victory over Jimmy Carter brought protests from Congress and the print press—even though President Carter himself had made his concession speech before the polls closed on the West Coast. A few states made laws that attempted to hinder exit polling by establishing distance requirements for interviewers of between fifty and three hundred feet from the polling place. These laws were thrown out as a violation of the First Amendment when they were challenged in the courts.

As the years passed, however, new states made new laws with the same objective. News organizations have attempted to negotiate with these states to find workable solutions without resorting to the high cost of lawsuits. In 2004 in Ohio, the National Election Pool (NEP) members did successfully fight for a ruling against the Ohio secretary of state and his unilateral attempts to enforce restrictive distance requirements on exit poll interviewing. News organizations have won all the federal suits they have filed so far, but the potential costs are high, even though the legal fees can be recovered by complainants who win.

In the early 1980s Congress held hearings about network projections, and stories, columns, and editorials in the print press claimed that elections were lost because of them. This negative publicity further increased the difficulty of conducting exit polls, even though a series of studies failed to show a connection between the projections and election outcomes.

In addition to these difficulties, a decline in response rates for surveys of all types was helped along by pressure for longer and longer exit poll questionnaires. An experiment conducted in 1981 tested three questionnaires of differing lengths. The first filled a standard 8 ½” by 11” sheet of paper. The second was half that size, and the third was half again. The response rates were approximately 50 percent, 70 percent, and 90 percent, respectively. Even with this evidence that smaller questionnaires can improve response rates, the desire of the news organizations to gather as much analytical information as possible from each respondent has kept the questionnaires at the largest of those sizes.

Another problematic trend that started in the 1990s was a growth in the use of absentee ballots and early voting prior to election day at special polling stations set up for that purpose. Absentee and early voting plays a major role in about one-fourth of the states. In 2004, it accounted for more than 30 percent of the vote in eleven states, and an additional seven states had between 20 and 30 percent of their vote cast early or by absentee ballot. But each year the size of this early vote has increased. The more people who vote this way, the smaller will be the numbers who go to their local precincts on election day. As exit polls, by definition, are conducted at polling places on election day, it has become necessary to supplement them with telephone polls of the early voters—an additional expense for an already expensive activity.