October 29, 2012

Dem Polls, Dem Polls...

Hatched by Dafydd

I was skimming yesterday's AP story that confidently prophesies Barack Obama will cruise to a narrow victory in November, and I was arrested by two claims, shifty predictions on which it bases its conclusion:

President Barack Obama is poised to eke out a victory in the race for the 270 electoral votes needed to win re-election, having beaten back Republican Mitt Romney's attempts to convert momentum from the debates into support in all-important Ohio, according to an Associated Press analysis a week before Election Day....

Without Ohio's 18 electoral votes, Romney would need last-minute victories in nearly all the remaining up-for-grabs states and manage to pick off key states now leaning Obama's way, such as Iowa or Wisconsin.

I found both suppositions puzzling; is Obama really so far ahead in those three states that AP can declare victory now, even as momentum surges towards Mitt Romney? Or is the press organization just whistling past the gravy train? Let's take a look.

Start with Ohio. This 18-EV state is a biggie. It's more and more likely that Romney is going to win Florida, North Carolina, and Virginia; that would give him 248 electoral votes (which you can see by playing with RealClearPolitics' interactive electoral-college map).

Now toss in the Buckeye state. If Romney also wins Ohio, he will have 266 electoral votes (evs) of the 270 required, and every other "toss-up" state has at least 4 evs. So if Romney wins Ohio, Obama would have to "run the table" of all remaining toss-ups, else Romney wins. He could not afford to lose a single state.

As it's manifestly unlikely that Obama would lose Ohio but win every other toss-up, the president cannot allow Ohio to go to Romney in the first place. We used to say that, since Republicans have never won the presidency without winning Ohio, that makes Ohio a "must win" for the GOP. But in 2012, it's actually a must-win for Obama; Romney has many paths to victory that don't roll through Ohio, but Obama has only one: total victory everywhere else.

So let's take a look at the polling in Ohio. The RCP average is currently Obama + 2.1, well within the margin of error; but that includes the Democratic poll by Public Policy Polling (PPP). This poll is simply a joke. Its turnout model is D+8, meaning 8% more Democrats than Republicans (D - 43%, R - 35%, I - 21%), and it gives women an 8-point advantage as well. Just toss this one in the fiery furnace; the actual average is thus Obama + 1.9.

Additionally, the Ohio average includes the most recent Rasmussen poll, which shows a dead heat -- as do the most recent Cincinnati Enquirer/Ohio News poll and the Suffolk poll (all had polling within the last week). Of the remaining six polls in the RCP average, four have Obama up by less than the margin of error (MoE) of each poll.

That leaves just two of them that show Obama ahead by more than the MoE: The CNN/Opinion Research poll (Obama + 4) and the Time Magazine poll (Obama + 5)... both of which have turnout models showing unreasonably large Democrat advantages.

The CNN poll has a turnout model of 35% Democratic, 33% Independent, and 32% Republican; these are self-reports from the respondents (I think Ohio doesn't have official party registration). CNN sees a D+3 turnout model -- and they come up with Obama + 4... hardly surprising.

But the Time poll just beats the cake. Its turnout model (page 26) predicts a Democrat advantage of plus 9! (D - 37%, I - 29%, R - 28%. The unweighted sample the pollsters obtained was even worse at plus 10.) That's even sillier than the PPP poll.

Bear in mind that in the 2008 election, Barack Obama only won Ohio by 5 points, 52 to 47. Does Time really believe that Obama is going to do just as well this year as four years ago? Yeesh!

In addition, they oversampled women (as did PPP) -- 47% male, 53% female; and they found a staggering gender gap of 10 points: Men favored Romney by 9 points, while women favored Obama by 19!

With a turnout model of D+9, the only surprise is that Obama is ahead only by five points. This poll is also clearly an outlier, in my opinion.

As a reality check, let's compare to the most recent statewide elections in Ohio. In 2010, Republican challenger John Kasich defeated incumbent Democrat Ted Strickland; while in the open-seat Senate race, Republican Rob Portman utterly crushed Democrat Lee Fisher by more than 18 percent... hard to square with a supposed D + 9 (or even D + 3) turnout just two years later.

So it seems a bit unreasonable, on its face, to assume that Democrats will outnumber Republicans by 9 points this year, or even 3, for that matter. If I had to guess, I'd say the turnout will be pretty even.

If we knock out the wonky Time poll and one of the polls where Obama and Romney are tied in Ohio (the "Bulgarian Olympics judge" rule, KOing the best and worst poll numbers), then the polling looks much dicier for the president; he is only ahead in Ohio by 1.5 points, not 1.9. Not only is that inside the MoE, it's easily inside the general Democratic poll-bias. More than likely, if the rest of the polling was more realistic about turnout, the polling average would show a tie or even an edge to Romney.

This hardly comports with AP's airy claim that Obama has "beaten back Republican Mitt Romney's attempts to convert momentum from the debates into support in all-important Ohio."

The debates started on October 3rd; in the week before the first debate, there were three polls, averaging Obama + 7 points. Now it's Obama + 1.9 (counting all the polls except the PPP Democratic poll); and if we toss the risible Time Magazing poll, it's Obama + 1.5. I'd say a five and a half point movement towards Mitt Romney qualifies as "converting momentum from the debates into support." Does anybody seriously dispute that, other than the DisAssociated Press?

It's clear that the momentum in Ohio is moving strongly towards Romney in the closing days, meaning he has an excellent chance of taking the state. So let's turn our attention to Wisconsin -- which AP also pooh-poohed as a potential Romney pickup.

Wisconsin polls -- all but one -- are likewise in a virtual dead heat, which means it will all depend upon turnout (unlike, say, California, where an Obama victory is predestined by God). And I believe that Romney's ground game will beat Obama's this year.

Measuring "enthusiasm" is tricky, because nearly everybody who expresses a preference for a candidate claims he is at least somewhat enthusiastic about said candidate; but that doesn't mean he will actually vote for the guy or gal. It's just the thing to say when someone asks you that stupid question.

There's really no good way to make an objective measurement before the fact; and after the fact, we will have the vote results themselves. But I don't think anyone in his right mind imagines that enthusiasm for reelecting Barack "Leading from behind" Obama will be anywhere near as high as it was the first time in 2008.

Since Romney's enthusiasm is huge and still growing, I see an excellent chance that Romney's "get out the vote" (GOTV) push will swamp Obama's. If that is true, then Romney will outperform his polling, while Obama will underperform his. In the case of Ohio and Wisconsin, that means Romney takes both states and is assured of being the next President of the United States.

But let's look at the Cheesy-Badger state polling in more depth. RCP has four polls up: Rasmussen has it a tie, Marquette University has it Obama + 1, and Mason-Dixon has it Obama + 2, all so close it's a statistical tie. The only bubbles in the swimming pool come from the Marist poll -- which incongruously has Obama up 6 points.

Is Marist an outlier? Let's take a look at the infernals. Marist's turnout model has D+5 (Independent 38%, Democrat 33%, Republican 28%). They don't give us cross-tabs; but we know that Mitt Romney is strongly winning independents and doing just about as well among GOP respondents as Obama is among Democrats, both nationally and in the swing states. That makes me highly skeptical that, with so many indies, Romney could possibly be so far behind.

Looking back to the most recent regular statewide elections, Republican Scott Walker beat Democrat Tom Barrett by five points, and Republican newcomer Ron Johnson beat long-time incumbent Russ Feingold, also by five points. That doesn't sound much like a deep-blue state, does it?

In any event, the Marist poll is clearly an outlier, because it's so out of whack with the other polls taken recently. In fact, out of whack with polls all the way back to the first debates, when the RCP Wisconsin average dropped from Obama + 11.5 just before the debates to Obama + 2.5 just after.

Finally, what about Iowa? Another swinger, the Trapper John state's polling is kind of all over the map: Two polls (Rasmussen and ARG) find a tie; WeAskAmerica finds Obama + 3; Gravis has Obama + 4; and PPP, a Democratic polling firm, actually finds Romney + 1, funnily enough. But the weirdest poll is, once again, Marist; despite having almost equal numbers of Democrats and Republicans, they nevertheless find Obama to be up 8 points! It all adds up to an average of Obama + 2.3 -- still eminently possible.

I don't know what to make of the scattershot spread for the Hotlips Houlihan state; but applying the Bulgarian Olympics judge rule, we get an average of Obama + 1.8. Again, even a small enthusiasm advantage for Romney will turn Iowa into a victory. And for the reality check, in 2010, Republican challenger (and former governor) Terry Branstad pounded the incumbent Democratic Gov. Chet Culver by nearly 10 points; while long-time Republican incumbent Sen. Chuck Grassley cruised to reelection against Democrat Roxanne Conlin (R + 30 -- yes, thirty points).

So much for "key states now leaning Obama's way, such as Iowa or Wisconsin;" they're only "leaning" in the most facile and technical sense, where one or two points is called a "lean" instead of a "toss-up." Each of these so-called leads can evaporate in the blink of a GOTV campaign.

The reality is that this election is "Romney's to lose" -- and let's hope he doesn't! If he can keep up the momentum he already has, then not only is Mitt Romney going to win, but he's going to win big... a lot bigger than Republicans are willing to suggest, and a lot bigger than Democrats are able to fathom.

Courage, Camille!

Hatched by Dafydd on this day, October 29, 2012, at the time of 2:01 AM


The following hissed in response by: seePea

"Don't get cocky"

The above hissed in response by: seePea [TypeKey Profile Page] at October 30, 2012 7:15 PM

The following hissed in response by: Beldar

I don't understand why -- given Scott Walker's repeated successes -- Wisconsin is considered to be leaning Democrat. Yeah, it was in 2008. But not in 2010, and not in 2012. Last time I checked, it's still 2012.

The above hissed in response by: Beldar [TypeKey Profile Page] at October 31, 2012 10:23 PM

The following hissed in response by: Dafydd ab Hugh


I started to write a post about this very point; then I discovered that Hindrocket already did it (more or less -- less specificity); then I heard Huge Hewgitt opine upon the same topic, so I let the post slide.

Simply put, the huge majority of state polling is still using the 2008 turnout model -- in fact, predicting an even greater Democratic advantage.

I used yesterday's Quinnipiac polls as exhibit A: They polled in Florida, Ohio, and Virginia... and their respondent pool (after weighting) was D+7 in Florida and D+8 in the other two!

Despite that, they had Obama up by only one point in Florida and by only two points in Virginia; which means Obama is underperforming his turnout-model advantage (and underperforming his previous election). He was up by five in Ohio, but even there the risible Democratic advantage in the turnout model ruins the poll.

If we correct the turnout model to a more reasonable even-steven, those same polls -- keeping the same relative vote within each party and holding the number of independents steady -- suddenly turn ominous for Obama:

  • FL: Romney = 48.0, Obama = 44.6; Romney + 3.4
  • OH: Romney = 46.7, Obama = 45.6; Romney + 1.1
  • VA: Romney = 49.7, Obama = 44.2; Romney + 5.5

And of course, depending upon the ground games, Romney's actual turnout might be better than even with the Democrats... in which case the election could turn out to be much more resounding than currently envisioned by most "pundants."

Each of those leads is outside the margin of cheating, if not (in Ohio's case) the margin of error.

I find the same surreal turnout model in so much of state polling this year that I can only draw one of two conclusions:

  1. The pollsters are lying about what they found because they just can't credit the idea that a far-right, whitebread cult leader could possibly beat the Lightbringer in a reelection bid;
  2. The pollsters are lying about what they found because they're desperate for Barack "Leading from behind" Obama to win for their own tendentious, mean-spirited, and disreputable reasons.

Either way, however...


The above hissed in response by: Dafydd ab Hugh [TypeKey Profile Page] at November 1, 2012 1:01 PM

Post a comment

Thanks for hissing in, . Now you can slither in with a comment, o wise. (sign out)

(If you haven't hissed a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Hang loose; don't shed your skin!)

Remember me unto the end of days?

© 2005-2013 by Dafydd ab Hugh - All Rights Reserved