Quinnipiac College has a new set of polls out today, the general tenor of which is that McCain is making gains on Obama. One of the states surveyed was Colorado, in which Obama has been holding a steady, albeit small, lead throughout the general-election campaign.
Today's Q-Poll gives us a result of McCain 46, Obama 44. Looking at the internals of the poll, however, suggests that party ID may be awry...
(Continues below the fold)
This is the main chart from the Q-Poll document:
LIKELY VOTERS.................................................
COLORADO Tot Rep Dem Ind Men Wom Wht His UnionHH
Obama 44% 7% 86% 47% 37% 50% 41% 57% 50%
McCain 46 87 7 39 55 39 51 29 40
----------------------------------------------------------------------
What's wrong with this picture? Well, McCain and Obama each have virtually identical leads among members of their own respective parties (87/86 to 7). And Obama leads 47-39 among Independents.
It would seem, therefore, that the only way mathematically for McCain to be leading is if self-identified Republicans substantially outnumber Democrats in the sample. Unless I'm missing it (which is possible), the precise party ID numbers are not available (the internals document reports on four states, so I could see where information would need to be limited).
There are other signs, as well, that Republicans may have been oversampled. Namely, the Udall-Schaffer U.S. Senate race is tied 44-44 (where Udall has held a steady lead in most polls). Also, the Q-Poll has Bush's approval in Colorado at 31%; I can't find recent numbers from other pollsters for comparison, but 31% would put Colorado five points above Bush's national average, which seems unlikely.
Finally, what do other polls say about the party ID composition of Colorado? According to this November 2007 poll in the state, the Democrats are at worst even, and perhaps leading, in party ID. The difference rests with whether party ID is reported "2-3-2" (which, I believe, lumps weak Democrats and Republicans with Independents, and only counts very strong and somewhat strong party identifiers with their party) or "3-1-3" (which puts weak identifiers with their party).