Weighting It Out: Party Identification and Election 2004 By Larry Hugick and Stacy DiAngelo

It started last June, when a Los Angeles Times poll of registered voters nationwide showing John Kerry opening up a significant lead over George W. Bush was harshly criticized by two prominent Republican political consultants. Matthew Dowd, chief strategist for President Bush's reelection campaign, told ABC News that the L.A. Times poll reporting a seven-point Kerry lead was "a mess" and skewed in favor of the Democrats. GOP pollster David Winston wrote a column in the Capitol Hill newspaper, Roll Call, delivering the same message.

This clash of established independent media pollsters versus partisans would be repeated over the course of the 2004 campaign, although later in the campaign the players would be quite different. At issue was the meaning of a much-misunderstood survey measure--the party identification question--and the appropriateness of using it to judge the quality of a preelection poll's sample. The L.A. Times poll with the seven-point Kerry lead reported the following partisan distribution: 38 percent Democratic, 25 percent Republican, and 24 percent Independent. To Dowd and Winston, such a party ID result was prima facie evidence of a biased sample, since the "proper" distribution should show a Democratic edge of no more than three or four points.

This polling controversy followed a pattern that is all too familiar to those of us who conduct preelection surveys for the news media. Campaign operatives react to an independent poll with unfavorable results for their candidate by seizing upon some technical aspect of the poll to use as means to try to undermine its credibility. In this case, the L.A. Times poll party ID results served the partisan critics' purposes very well.

To someone without any actual polling experience, this criticism seems very straightforward. If a poll has too many Democrats, how can it possibly be representative of the voter population? The impact of the critical salvos was enough to prompt the L.A. Times to take the unusual step of running an article about the controversy, giving its polling director, Susan Pinkus, the opportunity to respond to the critics. Pinkus noted that party affiliation is "a notoriously fluid characteristic," and that the current party ID figures were within range of those recorded in the L.A. Times surveys conducted since 2001.

Looking back, the characterization of the national L.A. Times poll with a seven-point Kerry advantage as an "outlier" with a flawed sample was off the mark. Horserace figures from two other major media polls taken at a comparable point in the campaign showed similar results. A CNN/Gallup/USA Today poll completed only days before the L.A. Times poll found Kerry with a six-point lead; a FOX News/Opinion Dynamics poll conducted about two weeks later found Kerry with a seven-point lead.

As the campaign moved to the post-convention period, the partisan attacks on the polls over the issue of party ID distributions took on a different shape. After George W. Bush surged to a double-digit lead in a number of national media polls following the GOP convention, it was Democrats, not Republicans, who most often employed party ID to try to discredit survey results that were bad news for their candidate. This time, it wasn't campaign operatives who led the charge. Instead, it was a group of activist bloggers, including Ruy Tuxiera, Chris Bowers, and Steve Soto of the websites The Emerging Democratic Majority, MyDD, and The Left Coaster. This group of poll critics obsessively tracked down the party ID distribution for every poll whose results they didn't like, and preached the gospel of "proper" party ID sample distributions on their websites.