OH: 52-47 Obama, PPP

[ajb]

The question I have for all the blue avatars out there is as follows:

Given that the range of polls conducted in the final week of a campaign is:

TIE (Rasmussen), Obama +1 (Gravis), +2, +2, +3, +3, +3, +4, +4, +4, +6, +8 (Zogby)

and given that a new poll is released showing Obama +5,

how do you go about deciding whether or not that O+5 poll shows a D-leaning house effect, and, if it does so show, then how large that effect is?

If you simply average those polls, they show  O+3.3. If you want to throw out Zogby, and keep Gravis, that leaves you with an average of O+2.9. The new poll is therefore about two points more Democratic than the polling average, suggesting a house effect of two points.

But it seems like Republicans here are using a different method. They're "picking winners" among the polls released, deciding that polls should or should not count towards the average, based on whether or not they come up to a standard of professionalism, whose terms were not set in advance of this election campaign. As a result, there seems to be a very strong correlation between polls that Republicans think are soundly conducted, and those which happen to be most favorable to the Romney campaign.

Given this model, polls which show results different from those most favorable to Romney can be dismissed as outliers, even if that means dismissing three-quarters or more of the polls actually released. They may be entirely consistent with the average of all polls, but are deviant by comparison with the ever-shifting list of "good polls."

It's a tempting method for all of us, but it's awfully difficult to keep from letting your desires interfere with your judgement if you use such a method. Who among the Republicans talking about polls here can honestly say that they haven't let their desires influence their methodology, and who, in particular, can support that claim by citing a post from the past two months in which they've endorsed a poll whose methodology they respect, even though they didn't like its results? And if you can find such a post, are your current posts entirely in keeping with that post?