Polls That Count

News media and campaigns long have relied on exit polls - information gleaned from voters, who, after they leave their polling places, voluntarily answer questions that help provide a sense of not only how an election is trending but the reasons behind their choices.

Still, the poll data continue to have limited value. Their reliability is less than perfect. Unfortunately, it's taken the last three national elections to confirm those limitations.

The now defunct exit polling firm, Voter News Service, incorrectly called Al Gore the winner in Florida during the 2000 election. Then, in 2002, the VNS system broke down and couldn't give its media clients any exit polling results. VNS was history at that point.

Then, the two polling companies replacing VNS for the 2004 presidential election admitted last week that their information was, in fact, the most inaccurate of any provided in the past five presidential elections. They produced estimates that mistakenly predicted John Kerry beating George W. Bush.

To their credit, major TV networks subscribing to the poll results didn't use that information to call an outcome, but enough leaked out to Internet bloggers to throw Democrats and Republicans onto quite an election day roller coaster. The mistake also fueled conspiracy theories that GOP supporters had rigged the election.

The pollsters said the mistake was "...most likely due to Kerry voters participating in the exit polls at a higher rate than Bush voters." That large numbers of Republicans decided not to talk to pollsters after voting was a phenomenon they simply didn't notice or explain.

Further, the pollsters noted their system "overstated the proportion of women in the electorate" - a problem caused by a programming error. And they acknowledged half the exit-polling interviewers were 34 years of age or younger, which, of course, could have helped skew the results.

Presumably, programming errors are fixable. And the pollsters also acknowledge the need to do better recruiting and training of pollsters. They also now say no information will be released to its media clients until 6 p.m. EST to help prevent potential mistakes from leaking out.

But a rethink on exit polling clearly is needed, one that may benefit from public input. Until these pollsters address errors they can control, the public and the media should remain skeptical of them.