Friday, September 28, 2012

Pennsylvania becomes 5 pt race when applying realistic turnout model

PPP has more bad news for Romney in their Pennsylvania poll (which was paid for by an environmental group known as the NRDC Action fund). The poll was completed September 17-18th, eleven days ago as of today. Apparently, after flirting with good numbers in the state on several occasions  throughout the last 2 years, Romney is now trailing by his largest margin yet (52-40%), whereas John McCain lost the state by 10.3 pts four years ago. Pennsylvania hasn't voted Republican since 1988, but has come close several times; Bush received 48.4% of the vote there in 2004, and 46.4% in 2000).

But if PPP's findings are correct, then Romney has no chance of carrying it this cycle. Why? Check out how Pa. likely voters are self-identifying in this poll: 50% are Democrats, 38% are Republicans, and 12% are Independents. In other words, partisan identification on this poll is D+12. Now PPP's numbers could be correct, but based on past turnout models, I'm doubting it. For example, 2008 is considered by many observers to have been a banner year for Democrats in terms of getting their most favorable demographics to vote. And in that year, Democrats only managed a partisan identification advantage of  D+7 (44D/37R/18I). Not only that, but Romney is winning an astounding 50% of independents, to Obama's mere 30%. That's a 20 point lead with Independents in...Pennsylvania (according to a Democratic pollster)!

So what would happen to Obama's 12 point lead in Pennsylvania if the electorate in November resembles that of 2008 (D+7), and not the findings of this poll (D+12)?

Well, Obama's still winning, and by a fair margin. But a 49-42% lead is a bit more encouraging than 52-40%. Obama's margin drops 3 points from 52 to 49%, while Romney jumps from 40 to 42%. And remember, this occurs simply by re-weighting the sample to 2008 turnout, an already great year for Democrats. So it would follow that if the race tightens up in a 2008 turnout scenario, it may be pretty close applying a more Republican electorate to the partisan identification. To do so, we'll use 2010 exit polls for the Pa. governor's race. That year, turnout in Pa. was 40D/37R/23I, or D+3.

So if the November turnout resembles turnout from two years ago, then Obama would lead Romney 47-44%, but it would likely be within the margin of error. So as you can see, the Romney camp has SOME reason to be optimistic. While PPP insists that 50% of survey respondents were Democrats, that just seems incredibly unlikely given electoral trends. Realistically, the true state of the race in Pa is likely somewhere between a 49-42% Obama lead (based on '08 electorate) and a 47-44% Obama lead (based on '10 electorate). So why not average the 2008 and 2010 exit poll party IDs, and reweight the PPP sample to reflect it? Party ID would be 42D/37R/20.5I. Here's what you get:

So if you buy PPPs cross tabs, but re-weight them to match an electorate somewhere between 2008 and 2010, then Obama's 12 point lead over Romney becomes a 5 point lead, 48-43%. Not the best position to be in, but certainly better than losing 52-40%.

No comments:

Post a Comment

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