Weighting It Out: Party Identification and Election 2004

By Larry Hugick and Stacy DiAngelo

Last of three parts.

The party identification distributions in preelection surveys came under unusual scrutiny in election year 2004. Poll critics who maintained that party ID should be treated like a demographic constant highlighted deviation from the “correct” party ID distribution as evidence of flawed survey methodology, leading to charges that the Newsweek poll and other major national polls had a partisan bias. In the end, however, exit poll results showing a more Republican electorate in 2004, as compared to previous elections, strongly supported the view that party ID is a fluid, not a fixed, characteristic. In the final installment of Public Opinion Pros’s three-part series, Larry Hugick and Stacy DiAngelo look at historical precedents to party ID movement around conventions, and the horserace, and offer their conclusions on the question of weighting preelection polls by party ID.

The movement in party identification toward the Republicans after the GOP convention in the Newsweek poll was confirmed by surveys of other organizations, including those that use a different version of the party ID question (the ABC News/Washington Post poll and the CBS/New York Times poll.) As noted at the time in a release by the Pew Research Center, five national media polls showed a shift toward the Republicans of between four and ten points in their early September 2004 results.

In the clamor at the time about “badly weighted” samples in post-GOP convention polls, one thing generally missing was any perspective on possible precedents to the party ID changes observed. While the political and journalistic community seemed to have no problem accepting the idea of a “post-convention bounce” in candidate support, the idea of a bounce in party ID did not seem credible to many. But polling data are available to test whether the shifts in party ID seen in September 2004 were unique or were part of an historical pattern.

Since 1984, the Newsweek poll has interviewed voters at the end of each party’s convention week. The timing of these polls allows for consistency of data, at least in the “post-convention” measure. Table 1 summarizes the finding of this analysis.

 *Popular vote

When we compare the party ID recorded in the individual polls conducted before and after the second presidential nominating convention—that is, the convention of the party holding the White House—we find a shift toward the incumbent party of similar magnitude to that of 2004 in four of the last five presidential election years. Only in 1992, when the Bush-Quayle ticket was soundly defeated in its bid for reelection, did party ID fail to move much in the incumbent party’s direction after the second convention. Therefore, the party ID movement seen after the 2004 GOP convention is not without precedent; it is a phenomenon that routinely occurs when an incumbent party’s presidential candidate goes on to win the popular vote in November.

For additional perspective, we also looked at movement in party ID around the first presidential nominating convention—the one held by the party out of power. As seen in Table 2, the historical patterns here are less well-defined. The shifts toward the nonincumbent party are typically not as large. And the size of the shift does not seem to vary depending on the nonincumbent party’s performance in November. In 1984, when Walter Mondale lost to Ronald Reagan in a landslide, the Newsweek poll showed a seven-point shift in party ID in the direction of the Democrats after their convention. But in 1992, when Bill Clinton defeated George Bush in a three-way race with Ross Perot, Clinton’s party received no boost at all in the Newsweek poll after his party’s convention.

 *Popular vote

A number of factors may explain why party ID seems to move more after the second convention than after the first. A key one is the advantage of incumbency. Just as the public rallies around the president in times of war or international crisis, people’s patriotic feelings might translate to support for the president’s party in the context of a political convention. Another major factor is the advantage of going second. Whatever movement the first convention generates for the opposition party tends to be erased by movement in the opposite direction after the second convention. Part of the second-convention bounce in party ID may be the erosion of the first-convention bounce. And, finally, for the time period we have studied, the incumbent party has generally been the more successful, winning the popular vote four out of five times.