A Gallup poll released yesterday indicates that the Kerry-Edwards ticket hasn't enjoyed much of a post-convention bounce. Though the Democratic duo leads President Bush and Vice President Cheney by three percentage points among registered voters, it trails by three points among likely voters. And when the independent Nader-Camejo ticket is thrown into the mix, Kerry and Edwards trail by four points among likely voters. How do pollsters decide who qualifies as a likely voter?
The exact methods differ from company to company and are often proprietary, but pollsters generally ask respondents about their past voting habits, their level of interest in the race, and the intensity of their allegiance to one candidate or another. The longer and more elaborate the questionnaire, the fewer likely voters the sampling is prone to yield.
Some pollsters, for example, will ask only whether the voter showed up for the last election and whether they plan on doing so again. Since many people feel slightly guilty about not voting, or want to appear more politically engaged than they really are, such polls tend to dramatically overstate the number of likely voters. Pollsters refer to this approach as a "soft screen."
Other pollsters ask their respondents to qualify their answers, instead of giving simple yes-or-no replies. A typical question in such a poll might be: "Are you very likely to vote, somewhat likely, or not likely?" The answers are then weighted in order to account for the assumption that those answering "somewhat likely" will probably vote in fewer numbers than those responding "very likely." Pollsters may also ask those surveyed to rank their likelihood of voting on a scale of one to 10.
The most thorough polls ask whether respondents know where their polling places are and how they plan to travel there on Election Day. Another tactic is to couch the questions about past voting habits in soothing language, so the respondents don't feel as if an "I didn't vote last election" reply is equivalent to confessing that they're bad Americans. A popular way to phrase the voting-history question, for example, is along the lines of: "In the last election, did something come up that prevented you from voting?"
Paul Perry, Gallup's chief election statistician in the 1950s and 1960s, is widely credited with formulating the most accurate likely voter questionnaire and statistical model. (It's not clear whether this is the system Gallup used for its latest poll.) In 1999, the Pew Research Center for the People and the Press tested Perry's approach during the Philadelphia mayoral race. Pew followed up with respondents after the election and concluded that it had correctly predicted the voting behavior of 73 percent of those surveyed. In the more-art-than-science world of polling, that's considered a pretty good result.
Explainer thanks Lee Miringoff and Barbara Carvalho of the Marist College Institute for Public Opinion.