2012 shaping up like 2004, on the Generic Congressional Ballot level.

I spent perhaps a bit too much time this morning trying to put the spreadsheet below into graphical form:

Pollster Time R D R +/-
NPR Oct 3/4 43 43 0
NPR Sept 4/4 45 48 -3
Politico Oct 4/4 46 45 1
Politico Oct 3/4 46 46 0
Politico Oct 2/4 44 46 -2
Politico Oct 1/4 45 46 -1
Politico Sept 4/4 44 46 -2
Politico Sept 3/4 45 47 -2
Rasmussen Oct 4/4 46 43 3
Rasmussen Oct 3/4 44 43 1
Rasmussen Oct 2/4 42 43 -1
Rasmussen Oct 1/4 43 44 -1
Rasmussen Sept 4/4 45 41 4
Rasmussen Sept 3/4 44 43 1
Rasmussen Sept 2/4 44 43 1
Rasmussen Sept 1/4 42 44 -2

It shows the current pollsters checking the Generic Congressional Ballot, as per RCP.  Most of the labels are self-explanatory: “R +/-” represents the amount by which Republicans are ahead/behind on any given poll.  RCP’s current average is R+1.3, which represents a strong shift towards the Republicans in the last month among all three pollsters: 5 points for Rasmussen, 4 points for Politico, and 3 for NPR (although ‘shift’ is possibly the wrong word for the last one, given that there’s only been two polls).



Five words with a hint of DOOM in them (Plus, a plug for the Transom).

“State polls lag national ones.”

Those of you who subscribe to Ben Domenech’s Transom and read today’s (10/19/2012) edition will understand what I mean; those of you who don’t… well, subscribe already.


#rsrh Looking at/for the partisan enthusiasm gap.

The following is a fairly vivid example of why people want to see crosstabs whenever possible when it comes to polls:

Romney 37 40 46
Obama 52 50 47
Republican 37 38 44
Democrat 50 49 48

The above is from a just-released AP-GfK poll; as you can see, it shows Obama up by one (47-46) over Romney in a poll of likely voters.  Bad news for Obama, as the intent was that he was supposed to be opening up a lead at this point; not so great news for Romney either, since the D+4 sample is a lot more likely to be reflective of the actual voting electorate in November than the D+7 or higher nonsense that we’ve been seeing lately.  But that’s another post. (more…)


#rsrh This analysis of decaying Democratic percentages…

…by The New Republic (!) is all very well, but it lacks one key phrase: “voting enthusiasm.”  Which makes a certain amount of sense: dwelling on that topic might force one to come perilously close to turning a ‘the election is tied’ narrative into an ‘actually, Obama is more or less losing right now’ one.  And we can’t have that.

Here’s the real problem with the polling, honestly: we don’t know what the electorate is going to look like in November.  We just know that basing it on ’08’s numbers is a mistake… but nobody knows how much of a mistake it’ll end up being.  And while it’s easy for me to say “Well, they should just eyeball it and pick voting percentages that look right,” that’s because I’m not a professional pollster…

Moe Lane

(H/T: Hot Air Headlines)


On that potentially troublesome Cook County poll.

Well, troublesome for Barack Obama.

The Daily Caller reported on a poll today that gave a rather remarkable result: it showed Barack Obama ahead of Mitt Romney 49 to 37… in Cook County, Illinois: which of course includes the City of Chicago.  Being under 50% in what can be considered the ultimate Democratic stronghold is not good news for the President.  In 2008 Cook County went for Obama over McCain 67/32; in 2010 it went for Quinn over Brady 54/40… which helped make the difference from Obama’s blowout 62/37 win statewide in 2008, and Quinn’s squeaker 47/46 win in 2010.  Put another way: if this poll is accurate, Obama’s got trouble in Illinois.

Since this report has been making the rounds of the Internet, I thought that I’d look into it a bit.  It turns out the poll comes from McKeon & Associates, which has been polling in Illinois since at least the 1980s.  A quick call to Michael McKeon got me access to the poll itself, which I’ll be talking about after the fold.



#rsrh Today’s Quinnipac University poll.

A helpful suggestion for Quinnipac University: there was something weird – and, worse, apparently inconsistent – going on with its last round of state-level polling.  Take a look at this chart:

Q-Poll Actual
M O Diff M O Diff Shift
Florida 40 53 -13 48 51 -3 -10
Ohio 38 53 -15 47 52 -5 -10
Pennsylvania 40 54 -14 44 55 -11 -3

(Originally noticed by @NathanWurtzel) Key: M is McCain, O is Obama.  The above compares the difference between the ratio of McCain/Obama voters in Quinnipac’s latest poll (which had Obama by six in Ohio and Florida, and by eleven in Pennsylvania) by the actual ratio of McCain/Obama voters from 2008.  As you can see… yes, Quinnipac at least found Obama voters.  But it wasn’t finding McCain voters in Florida and Ohio; and while I take Sean Trende’s point that people tend to ‘forget’ that they voted for the loser I have to note that they apparently didn’t forget in Pennsylvania. I would also suggest that the fact that Ohio and Pennsylvania flipped its state government (including governorships) to the GOP in 2010, while Florida increased its Republican majorities, suggests that the electorate in none of those states has gotten more Democratic in the last four years.  (more…)


#rsrh QotD, The Partisan Breakdown Is A Feature, Not A Bug Edition.

Ed Morrissey, on the pathetic partisan (take that any way you like) breakdown (ditto) of the latest WaPo/ABC poll (46/49 Romney/Obama):

Today’s D/R/I is 32/22/38, which means this model would only be predictive for a turnout model where only 22% of voters are Republican.  Just to remind readers, the 2008 turnout split from exit polls showed a 39/32/29 split, and that was considered a nadir for Republican turnout.  In the 2010 midterms, the split was 35/35/30.

Take a close look at the Republican representation in WaPo/ABC polls this year. Starting in January, that has been 25%, 23%, 27%, 23%, and now 22%.  The pollster seems incapable of finding a representative number of Republicans for this poll series.  Perhaps that should give the two news organizations involved a hint about finding a new pollster.

Why should they?  The pollster is fulfilling said news organizations’ needs perfectly. (more…)


#rsrh Quick observation on this Hot Air post on Biden’s gay marriage gaffe.

While I see Ed Morrissey’s point (I don’t know if I agree with it, but I see it) about whether it’s a big deal right now whether or not Barack Obama comes out in favor of same-sex marriage or not…

…it’s unimaginable that enthusiasm for Obama among black voters would suffer much, even if he reversed himself on gay marriage.  Even a slight reduction in turnout would be a rather doubtful assumption.  If Obama’s hold on this demographic is that fragile, he’s already lost the election.  The swing-state vote is a more realistic concern, but the current vacillation won’t help if those voters are keying on this topic, either, and they almost certainly aren’t.



#QotD, You Don’t Say? Edition.

The Christian Science Monitor, blinking through the sudden pain:

The poll by Gallup Inc. and USA Today showed Obama with 47 percent support in the 12 states and Romney with 45 percent, well within the survey’s margin of error of 4 percentage points. That is a tighter race than in March, when it found the Democratic president with 51 percent and Romney with 42 percent.

…You don’t say?

Moe Lane

PS: Interesting contrast with this, in some ways.


#rsrh …Dagnabbit, I need better crosstabs on these polls.

I just had to throw out four hours of work on a post on Virginia’s Romney/Obama Head-to-Head situation because I don’t really have the data to answer the question What percentage of the white vote can Obama reasonably expect in Virginia in 2012? This is important because (what follows is very much back-of-the-envelope calculating, based off of the NYT exit polls) if Obama gets 40% of the white vote (the percentage he won in 2008) and nothing else changes, he wins Virginia; if Romney gets 68% (the percentage Bush won in 2004) and nothing else changes, then he wins Virginia, by about the same percentage that Obama did*. Unfortunately, it’s still too early to make any kind of confident statement about which percentage of the white vote each candidate will get.

But it’s looks like it’s going to be that demographic that’s going to be the key to VA…

Moe Lane (more…)


#rsrh A general suggestion to pollsters, particularly D-leaning ones.

Triggered by flipping through this Hot Air post… STOP MUCKING ABOUT WITH THE SAMPLE’S PARTISAN ID PERCENTAGES.  That trick doesn’t work any more.  Everybody knows by now to look to see whether the partisan breakdown of the poll matches the actual voter percentages in the last few elections; and when they don’t (like they didn’t in the last National Journal poll) the impact of the poll is thus diminished.  Also: it’s not like skewed polling makes a difference, either: there were a lot of polls in 2010 that effectively failed to downplay just how badly the Democrats were going to do that year.  Finally: don’t just not release the partisan ID numbers, either.  People WILL ask where they are.

Seriously, all that using skewed polls is doing at this point is convincing half the country that you’re not to be trusted as far as you can be thrown.  Keep that in mind.


Crunching the July 2011 Rasmussen trust numbers.

I stopped doing this in the middle of 2010, once it became clear that the real question of 2010 was how many hits to the head with the snake the Democrats were going to take before it was all over.  As the answer was “a lot,” I feel that this was a wise prioritization of my time.

But it’s a new cycle, so let’s look at the numbers – both the latest ones, and May’s.  Short version: Republicans are scoring better in nine out of ten topics that Rasmussen charts, and there’s a ten point lead on the economy.  Which, not incidentally, is the most burning issue for Americans these days. (more…)

Site by Neil Stevens | Theme by TheBuckmaker.com