No Quick Fix for Exit Polls

The media group that oversees these surveys analyzed past mistakes, but the proposed changes look difficult to implement

By Peter Coy

Exit polls that were either flawed or misused caused problems in each of the last three national elections. On Election Night 2000, the networks picked Al Gore to win Florida, even though the actual results were a virtual dead heat. In 2002, computer snafus prevented the exit pollsters from reporting data to subscribers. In 2004, exit polls produced sometimes-unreliable results -- and were leaked to the Internet, where overeager bloggers concluded that John Kerry was poised to win the Presidency.

The media consortium that oversees exit polls -- named the National Election Pool -- described a plan on May 14 to do better in 2006 and beyond. But consortium members and the pollsters themselves admitted that there's no simple solution. Case in point: Last fall, the exit polls overestimated the support for John Kerry because the interviewers -- who tend to be young and liberal -- were more likely to succeed in interviewing Kerry supporters than Bush ones on their way out of the polling places.


  Luckily, the faulty polls didn't result in any inaccurate calls because election-night analysts sensibly weighed them against other indicators, such as preelection polls and early vote tabulations. The biggest problem was with the bloggers, whose interpretation of the results spread like wildfire and may even have driven down the stock market, where some investors feared the consequences of a Kerry win.

On May 14 in Miami Beach, the pollsters laid out their plan of action in a forum at the annual meeting of the American Association for Public Opinion Research. They said they hope to prevent leaks of exit-poll data, shrink the exit-poll questionnaire so more voters will be willing to complete it, get permission for interviewers to stand closer to the exits to catch more voters on their way out, and improve the recruiting and training of the interviewers.

Sounds good. But every part of the plan is fraught with difficulties. Leaks are extremely hard to plug in open organizations like television networks. While the exit-poll questionnaire could be shrunk, that would mean losing key facts about the demographics and political attitudes of voters that are used to analyze who voted which way and why. And getting local election officials to let pollsters to stand closer to polling-place exits will be an uphill struggle, given widespread public skepticism toward the media.


  Still, all those problems pale in comparison with the difficulty of recruiting and training reliable interviewers. To ensure a random selection of voters, interviewers are supposed to approach and interview every, say, fifth voter who emerges from the polling place. But if that fifth voter happens to look unwilling to chat, the interviewer may pick someone else -- spoiling the randomness that's crucial to accurate polling.

One solution is to change the demographic mix of interviewers so they more closely match the ages of the voting public. But older interviewers are more expensive to hire, Joe Lenski, co-founder and executive vice-president of Edison Media Research, tells BusinessWeek Online. Lenski's firm is one of those hired by the National Election Pool. Also, he says, the demographic mismatches that caused problems in 2004 might not be the same ones that cause problems in 2006 -- because a lot depends on who the candidates are.

The organizations that constitute the National Election Pool are among the biggest names in media: ABC, Associated Press, CBS, CNN, Fox, and NBC. They're trying to get it right. But it seems like every two years, some new problem reaches up and bites them.

In Miami Beach this past weekend, they came clean about what went wrong last time and vowed to do better in 2006. The world will be watching.

Coy is economics editor for BusinessWeek

Before it's here, it's on the Bloomberg Terminal.