Upon Closer Inspection, Another Gallup Poll Is Suspect
Last Friday, I raised the possibility that Gallup's national polls were suspect due to questionable sampling that misrepresented the party identification of the electorate. At the time, I suspected that this problem with Gallup's sampling and its oversampling of GOP voters may also exist with their state polls, which have been used by electoral vote predicting websites to show a national tilt towards Bush and away from Kerry since the GOP convention. In the case of Wisconsin, I now can say that this problem exists again.
Last week, in a story that the USA Today touted with the headline "Bush Widens Wisconsin Lead Over Kerry”, Gallup reported a Bush lead over Kerry of eight points (52%-44%) among likely voters in the state. This poll was based on a sample which contained a party ID breakdown of
Yet an October 2003 study of the Wisconsin electorate by Harris Interactive, found the following party ID for the state:
In other words, there is a relatively even split between the GOP, the Democrats, and independent voters in Wisconsin. Yet Gallup trots out a poll trumpeting an 8-point Bush lead amongst likely voters that is based on a sample containing a six-point GOP advantage in the state that seemingly doesn't exist.
I will look at other Gallup state polls later in the week. But it is clear so far that Gallup's likely voter sample methodology is presenting a false impression of Bush's relative strength in the minds of the voters and the lazy media.