Monday, October 15, 2012

UNSKEWING The Washington Post / ABC poll

WashingtonPost / ABC is out with a new poll showing Obama leading Romney nationwide by 3 points, 49-46%, among likely voters. Nothing seems too peculiar about the top line, other than the fact that it's the first national poll to show Obama with a modest lead post Denver-debate. But what is peculiar is the surprisingly high number of respondents identifying as Democrats as compared to Republicans. The poll's partisan identification is 35% Democrat, 26% Republican, and 33% Independent, or D+9.

If turnout actually reflects the Washington Post poll's findings, then we can expect to see the most Democratic electorate in modern history this November, even more so than the record breaking Democratic turnout we saw just 4 years ago (which was D+7).

In light of this difficult to trust finding, let's "unskew" the poll, and reweight it to 2010 and 2004 turnout (which was TIED among the two major political parties). To be specific, 2010 party ID broke down like this: 35D/35R/30I. Here's what the ABC numbers would look like reweighted to this party ID:

ABC/WashingtonPost poll weighted to 2010 CNN national exit poll:

Party ID        Obama                    Romney

35D              31.85 (91%)            2.8    (8%)
35R               2.45   (7%)             32.55 (93%)
30I                12.6  (42%)             14.4  (48%)
                       46.9%                    49.75%  

So, if the Washington Post is right and turnout in November is D+9, there is very little Romney can do to win. But if those Post numbers are reweighted to 2010 exit polling, low and behold, Romney takes a 3 point lead, 50-47%.

If you're like me and believe that turnout will fall somewhere between 2010 and 2008 levels, then it's easy to see we've still got a tied race. D+9, though, ain't going to happen.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.