Jun 1 2007

The Horserace Waste

The uselessness of early polling

At the moment, polls show former New York Mayor Rudolph Giuliani with a wide lead over his Republican counterparts, and senators Hillary Clinton and Barack Obama leading the Democratic field. The media’s campaign storyline is shaped largely to conform to these findings. Time magazine (3/22/07) recently predicted that “the political rule book has been stuffed into a shredder this year,” the “conventional wisdom” shattered by the apparently inevitable victory of current front-runners like Giuliani or Hillary Clinton.

A look back at past election cycles shows such predictions are unwise. Early polling of the 2004 Democratic nominees (e.g., CBS News poll, 12/14-16/03) showed eventual nominee John Kerry in the middle of the pack, trailing Howard Dean, Wesley Clark, Richard Gephardt and Joe Lieberman. An August 2003 USA Today/Gallup poll (8/25-26/03) showed front-runner Lieberman with a 10-point lead over Gephardt. As the dynamics of the nomination race shifted, so did the polls—but not in the direction of the eventual winner; by January 2004 (Newsweek poll, 1/2-5/04), Howard Dean was leading the pack, followed closely by Wesley Clark.

The Democratic race was similarly competitive in 1992, and the polls then were equally useless in helping to predict the eventual nominee. In March 1991, Paul Tsongas, Mario Cuomo and Dick Gephardt were the frontrunners, according to one survey of New Hampshire voters (Boston Globe, 3/31/91). Bill Clinton was hardly a factor.

And four years before that, the Democratic race was just as hard to gauge by looking at the polls. Former Sen. Gary Hart led in one early ABC/Washington Post poll (5/8/87), though a Post article (8/24/87) noted that Michael Dukakis seemed to be treated as the front-runner at a candidates’ debate even though “he barely makes it out of single digits in national polls.” By December 1987, Hart was polling at 30 percent and Jesse Jackson at 20 percent, with eventual nominee Michael Dukakis in third place with 15 percent (Washington Post, 12/19/87).

The Republican nominating process has tended to be less competitive in recent years, but there are similar lessons to be drawn. In the 2000 race, Bush’s only serious competition came from Sen. John McCain, who was trailing far behind in the early polls—behind Elizabeth Dole, Dan Quayle and Steve Forbes (e.g., NBC/Wall Street Journal poll, 1/99). McCain’s surge came much later, closer to the early primary competitions and—hardly coincidentally—when McCain was receiving far more significant (and generally positive) media coverage.

Early polls are primarily measuring name recognition, so high-profile candidates tend to do better. “Winning” the polls encourages more media attention, much of it about how a given candidate is maintaining his or her lead. When actual voters intervene in the process, however, the front-runners often become also-rans.

More importantly, early polls function as a way of giving media an excuse to ignore candidates—often the majority of candidates—who are deemed outsiders undeserving of media attention (Extra!, 9-10/03).

While reporters seem aware of the problems with overplaying these early polls, it’s not clear that this awareness has any effect on their coverage. A recent Associated Press report (3/2/07) noted that an advisor to Sen. Chris Dodd (D-Conn.) called MSNBC’s Chris Matthews to remind him that Dodd was running for president, even though he isn’t polling very highly. Matthews responded by telling AP that he doesn’t want to pay attention to early polls, but “it is so hard not to.”

When ABC anchor Charles Gibson (2/27/07) noted that it seemed “very early to be polling,” his colleague George Stephanopoulos responded: “It sure is, but we’re giving voters what they want, Charlie, apparently. We asked voters how closely are you watching the campaign right now, 65 percent said they were watching it closely.”

It’s a great leap to imagine that when Americans tell pollsters they are closely following the election, they mean they’re looking for horserace polling whose predictive value, almost a year before any actual voting, is essentially nil. More likely they are looking for substantive coverage of the entire array of candidates that will help them make up their own minds about who they want to vote for.

Imagine if instead of focusing on meaningless polls, campaign journalists reported on the public records and policy proposals of the candidates—all of them, not just the few media-anointed front-runners. Now that would really shred the conventional wisdom.