Wednesday, October 17, 2012

SurveyUSA shows Ohio is no "firewall" for Obama. Romney behind only 1 pt in 2008-scenario

Paul Ryan, his wife and mother campaign in Owensville, Ohio with Governor John Kasich (R), Sen. Rob Portman, and his wife, on Sep. 12, 2012. Photo found here.
Survey USA is out with a new poll showing Obama leading Romney in Ohio, the President's "firewall," by a 45-42% margin. 3% choose someone other than Obama or Romney, and 9% are undecided. Poll findings indicate that 39% of Ohio likely voters self-identify as Democrats, 32% identify as Republican, and 26% identify as Independent. Both Obama and Romney are attracting equal portions of their base, while Romney captures Independent voters by 8 points, 41-33%. As a result, it's easy to see that we're dealing with a Democratic friendly poll sample, given that Romney leads strongly with Independents, yet still trails the President overall. In the Democratic year of 2008, Ohio partisan identification was D+8 according to CNN exit polls. But poll-nerd/analyst @NumbersMuncher points out that the actual Ohio partisan ID in 2008 was closer to D+5, or 37.5D/32.5R/30I. He explains it here:

This was a wave election and Barack Obama took Ohio by 4.6%, 51.5 to 46.9. The exit polls show a split of 39% Democrats, 31% Republicans, and 30% Independents.  When you take that breakdown and put it against the exit poll results of vote per party, Obama would win 52.8 to 45.6 – 7.2% victory.  This means that the exit polls were off a little, which is normal since they are, after all, a poll.  On the plus side, having an *actual* vote total to compare to makes it easier to work with.  If you mess around with the exit poll party % against the share of vote for each party, you come out with this breakdown: 37.5% Democrats, 32.5% Republicans, and 30% Independents.  That gives you a vote of 51.6% for Obama and 46.9% for McCain – pretty close to actuals. That means for 2008 the exit polls show an EIGHT point Dem advantage, but in reality it was likely closer to 5%. That is a huge difference when looking at the current polls.  Article here

SurveyUSA's party ID finding in Ohio could be entirely accurate when compared to actual turnout this November. But what if it isn't? What if turnout looks more like the D+5 scenario laid out above? Here's how the SurveyUSA poll would look if that were the case: 

SurveyUSA poll weighted to 2008 Ohio turnout (per @NumbersMuncher):

Party ID        Obama                    Romney

37.5D            33     (88%)             2.25     (6%)
32.5R            1.625 (5%)              28.925 (89%)
30I                9.9     (33%)            12.3     (41%)
                       44.525%                   43.475%

What was a 3 point race becomes a 1 point race in the event turnout in Ohio resembles '08 turnout. On the other end of the turnout spectrum, here is what the SurveyUSA poll would look like if the 2012 electorate resembles 2010 turnout:

SurveyUSA poll weighted to 2010 Ohio turnout (per @NumbersMuncher):

Party ID        Obama                    Romney

37.5D            33     (88%)             2.25     (6%)
36.5R            1.825 (5%)              32.485 (89%)
26I                8.58    (33%)            10.66     (41%)
                      43.405%                   45.395%

Imagine that: when you reweight the SurveyUSA poll to 2010 turnout, Romney takes a 2 point LEAD in Ohio. When you weight it to 2008 turnout, Obama leads by 1. Either way, I'd hate to have to defend calling Ohio my "firewall" state, whether you're Romney or Obama. It appears to be a toss-up.

No comments:

Post a Comment

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