Classical American political wisdom tells us that a candidate’s supporters regard polls as scientifically infallible when they show support for their candidate, and as bluntly inaccurate when they show support for the other guy. While I want to refrain from sounding like an outraged Kerry partisan, I do feel compelled to point out some rather strange aspects of the much-ballyhooed Time Magazine Poll conducted during the RNC.
The Time Poll was conducted over three days during the RNC. However, major polling houses almost never conduct electoral opinion polls during national party conventions in election season. The reasons for this should be obvious: partisans are more likely to be at home, watching the convention, and the increased media attention places the candidate in question even more prominently in respondents’ minds. It’s important to remember that the polls conducted during August 24-26 (just before the RNC) showed Bush and Kerry tied 46-46.
The Time DNC poll was of course not conducted during the DNC, but from the fifth to the seventh days after it. It showed Kerry leading Bush 51-44 among both likely and registered voters.
Time’s RNC poll, however, was conducted during the heat of the convention. What’s more, unlike the DNC poll, and in fact any polls normally conducted at this point in a presidential election season, Time’s RNC poll pushed undecided likely voters to name a candidate. Pushing leaners and undecideds is, from a methodological perspective, inadvisable this early in an election season and especially during a convention. UVs don’t yet feel like they know who they prefer, and when pushed by a pollster, they’re more likely to name either the candidate who’s been in the news a lot or the incumbent candidate, since they already are familiar with him. In this case, Bush was both featured more prominently in the news (it was the RNC, after all), and also happened (have you noticed?) to be the incumbent.
I’ve read on some unabashedly pro-Kerry blogs that the RNC poll asked to speak to male respondents first and female respondents only if a male was unavailable. This would lead to responses skewed toward Bush, since men tend to skew Republican and women tend to skew Democratic. I can’t substantiate any of this though, so it should be ignored in the final analysis.
And that analysis is that, quite simply, the methodology of this poll was flawed (even if that last bit turns out not to be true). The fact that Time conducted the poll during, not after, the convention, and pushed undecideds to name a candidate (something Time DIDN’T do during its DNC poll) is enough to make me regard the results of this poll as bunkum. There’s no doubt that Bush did receive a bounce from his convention speech, but there’s no reason to believe that he could have leaped so far so quickly. The Kerry-favorable results of the Time DNC poll (in which a majority of respondents named Kerry WITHOUT being pushed) were not paraded in the national press as proof that Kerry was “beating” Bush, but that’s just what’s happening now with this push-poll for Bush. Time’s poll creates a non-story (about Bush’s supposed superiority in the polls) that is likely to bias future stories for some time. Again, to name just another widely-known nugget of political wisdom, it’s not being popular that counts, only the perception of being popular.
[edit] Here’s a link that points to what is probably a more accurate picture in the polls. I’ve read elsewhere that it’s believed that Bush probably has a 3 or 4 point lead at the moment. [/edit]