Wednesday, September 26, 2012

Ridiculous Party ID Aside, Quinnipiac/NYT/CBS reveal rough numbers for Romney

Upon looking at the internals for today's Quinnipiac polls (which did a great job in the 2010 elections), I had to do a double take at the Party ID section. Like many who are familiar with past exit poll data, the partisan divide of the likely voter sample, for all 3 polls really, was heavily favorable to Democrats. But applying 2008, 2010, and an average of the two elections turnout numbers to the internals of the Q-poll show that Romney is STILL trailing in BOTH Ohio and Florida.

Lets first take at look at the Ohio Q-poll. Obama leads by an astounding (and downright frightening if you're a Republican) 53-43%. Keep in mind that Obama won the state by 4.6 pts, 51.4 - 46.8% in 2008, and Bush carried it twice by 50.8% and 50.0% in 2004 and 2000, respectively. Partisan identification in Ohio in 2008, according to CNN exit polling, favored Democrats by 8 percentage points (D+8). In other words, 39% of the voters that turned out on election day that November in 2008 identified themselves as Democrats, 31% identified as Republicans, and 30% identified as Independent. However, findings from Quinnipiac University indicate Ohio likely voters are currently identifying as Democrats at a slightly higher net rate than they did on election day 2008, the peak of Obama-mania; 35% to be exact, while just 26% identified as Republicans. 35% identify as Independents, meaning the party ID for poll is D+9

So what would the Q-poll look like if you simply applied the 2008 exit poll partisan identification? Here you go:

Number in parentheses = percentage Obama/Romney received from Republicans/Democrats in the poll
So the race stays at 53-43%, even re-weighting to a 2008 style electorate. That's because the net party ID for the Q-poll (D+9) is essentially the same as 2008 Ohio turnout (D+8). Obama's 9.7 point lead over Romney in a 2008-style electorate would indicate that Romney is under-performing in comparison to John McCain's '08 Ohio numbers (who remember, only lost by 4.6 pts with big Democratic turnout).

But there are plenty who disagree that 2012 voter turnout will resemble anything close to the record-setting year Democrats had four years ago. Many of those point to the Tea Party landslide from the 2010 midterms as a better model for turnout this November. In that election, Republicans tied Democrats in party identification nationally (35% Dem, 35% Rep, 30% Ind), as well as in both Ohio and Florida Governor races. Specifically, in Ohio, partisan identification according to 2010 CBS exit polls was 36% Dem, 36% Rep, and 28% Independent. Here's what that Quinnipiac poll would look like under a 2010 turnout scenario:
The good news: the new result has Obama leading Romney 50-47%. Obama's margin over Romney drops from 9 pts using the Quinnipiac partisan sample, to 2.6 points under a 2010 sort of scenario. While the race is much tighter than the actual Quinnipiac sample would indicate, Romney is still losing to the President under even the rosiest of turnout scenarios. That's the bad news.


 The same holds true for Florida. While Quinnipiac finds likely voter identification to be D+9, both the 2008 and 2010 Florida electorates were less Dem friendly (D+3 four years ago, TIED two years ago). In either scenario, the President would still defeat Mitt Romney in Florida, using the Quinnipiac numbers.

Assuming 2008-style turnout, the President would lead Romney in Florida 51-47%, which is larger than the 2.8 pt margin Obama defeated McCain by. Only when you reweight the sample to reflect 2010 party ID does the race become essentially tied (though Obama still leads 50-49%). So exactly how is Romney winning Independents, yet still losing, even in a favorable 2010-style environment? Because in both Ohio and Florida Q-polls, Obama attracts a larger share of both his own party's vote, as well as his opponent's party. 

So the bottom line is that Romney is in trouble (again, if you buy Quinnipiac's internals), no matter what type of electorate we have in November. He trails in Ohio and Florida whether the sample spread is as large as Quinnipiac suggests, or as small as 2010 turnout. Just for fun, here are the results of the Quinnipiac poll if you average the 2008 and 2010 party identification numbers (trying to be as fair as possible here). That would make the party ID in Florida 36.5D/35R/29I and 37.5D/33.5R/29I in Ohio. The results are below, and still spell trouble for Romney:

FLORIDA PARTY ID = D+1.5, Number in parentheses = percentage Obama/Romney received from Republicans/Democrats in the poll

So  in Ohio, Obama leads in a realistic D+4 electorate, 51-45%. In a realistic D+1.5 electorate in Florida, Obama leads 50-48%. Assuming Quinnipiac's numbers are correct, Romney needs improvement no matter what the turnout is like in November.

Note: Pennsylvania wasn't include in this analysis because Quinnipiac did not provide voter break down by party identification for this particular state only.

No comments:

Post a Comment

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