Up until today, most of the polls have been illusions. Rasmussen may be the single exception to the rule.
Watching the Rasmussen internal numbers, there has been a great deal of consistency in the polling demographics, which makes for an easy time telling if there has been movement to Mitt Romney or towards Barack Obama. Rasmussen has had the poll very tight all the way. Rasmussen keeps a weekly average on their site (available only to Platinum members) that shows that since the end of January, Obama has been ahead 13 weeks and Romney 19. And last week was “even”. There were 4 other “even” weeks. Most of these weekly averages showed one candidate or the other up by 1 to 3 or 4 points. According to Rasmussen, this has been a very stable race and has been within the margin of error all the way.
Rasmussen typically polls Democrats at a couple of points higher than Republicans and they use a good demographic mix for the rest of the important factors such as race and sex.
The same is not true of the other polls.
We have been tracking the polls at Political Poll Check and documenting the data for every poll listed at Real Clear Politics since the second week of September. Most of the polls have at least 2 surveys to compare head to head.
And we have developed a process to “normalize” the polls to correct for the wildly ranging party skew in the polls that have been released to date.
In simple terms, an analogy for Normalization is this:
Say you had 3 recipes for soup and you wanted to compare the ingredients. One makes 25 servings, one makes 50 and one 100. They have exactly the same ingredients, but for some reason, when you make the recipe for 50 servings, it tastes a lot better and is everyone’s favorite. Normalization converts all the recipes to 1 serving so you can easily see that the one you like most has more butter and salt.
There are a couple of polls that, after normalization, tend to be on the high side for Romney, but most of them fall in line with Rasmussen, within the margin of error.
In 2008, Obama won by 6.5%. He enjoyed a great deal of enthusiasm and excitement and his voters turned out in masses to put the first African-American president in office. The promise of a better future glistened and Obama’s magnificent speeches inspired an almost god-like image for candidate Obama. People were more excited to elect a president than we have seen in decades.
Four years later, the thrill is gone. The lofty promises and goals were never achieved. Many of his supporters have turned on him and voter enthusiasm is way down for the Democrats. Nevertheless, most of the pollsters have been using the 2008 model for 2012. Many have predicted an even more optimistic turnout for Obama than 2008 despite their own internal polls showing something vastly different. Even Obama’s fundraising numbers tell the same story: 2012 will be far tougher than 2008 for Obama.
Yet the pollsters insisted on using these painfully optimistic turnout models. And as a result, sites like Real Clear Politics have shown Obama ahead day after day as one skewed poll after another is averaged into the mix. Having watched this average as closely as the next political junkie, it is often frustrating to see Romney claw his way back into the mix, almost managing to deplete Obama’s (badly skewed) average, only to see a poll that surveyed Democrats ahead of Republicans by double digits slam the average needle back to Obama’s advantage. Time after time this has happened.
For the first time in recent history, Obama has been knocked off of his perch atop the RCP average roost by Mitt Romney.
Now it took a few stars aligning just right to make this happen.
One was the Obama-Romney debate last week. There is no doubt Romney got a boost from his brilliant night outshining Obama on the same stage. And this has been reflected in the more recent polls, no doubt.
But another set of events has also occurred as we predicted when we started Political Poll Check.
The only polls that will matter in 2016 as far as ranking polling accuracy of individual pollsters are the last few. What they poll in the summer and before is not even considered. The last week is especially important, and the last poll prior to the election is the “money shot”. Get that one close and no matter how badly you did the rest of the election season, that is the one that will make or break your reputation for future elections. And it’s not just bragging rights, it is revenue and clout.
Today’s events may very well have marked the end of the bogus polls we have been seeing, published mostly to drive the narrative, and the beginning of the accountability and reputation phase.
First, a PEW Poll came out today showing Romney up by 4 points! This was a milestone because not a single poll (except Rasmussen 3 weeks ago) has shown Romney with any lead at all. Even more remarkable was the previously released PEW Poll (9/12 – 9/16) that showed Obama with an 8 point lead. So we had a 12 point swing in less than 30 days. But really, we didn’t. The previous poll was +6 ion favor of Democrats – using the 2008 model. The latest poll sampled +3 Republicans which is the first National poll to sample more Republicans than Democrats. And interestingly, this model lines up perfectly with Rasmussen’s party affiliation poll average for 2012. Here is my analysis of this Pew poll.
So we finally had a single, solitary poll that showed a Romney lead.
And next came a pretty remarkable poll from IBD / TIPP. Oddly, this one over sampled Democrats by +8 (39% Dems, 31% Repubs) and still showed a 2 point lead by Romney. That is pretty significant and the internal numbers behind the polls are pretty telling. The gender gap Romney had with women closed, the gender gap Obama has with men widened and Independents moved to Romney in droves. My analysis for this poll will be posted tomorrow but the breakdown for this poll is already completed at Political Poll Check and shows the “normalized” results of this poll with a 11 point advantage for Romney. (That is here.)
And when Real Clear Politics posted this one, Romney was less than a point from taking his first RCP lead.
And the final star to align was that Gallup, a respected polling firm that refuses to release their internal numbers (cross tabs) decided today would be a good time to finally stop publishing Registered Voters results and move to Likely Voters like the rest of the polling world.
It is a well known fact that Likely Voters tend to favor Republicans. And after Labor Day any polling firm that is still using Registered Voters is simply useless.
And how much of a difference did it make with Gallup letting us know the results of those who are actually going to vote on November 6th? See for yourself. This picture below is a snapshot from this morning, before the Gallup conversion to their LV model:
As you can see, Gallup this morning was reporting a Tie among Registered Voters. Among Likely Voters, Romney is up by 2%. And notice that the Margin of Error went down from 3.0 to 2.0 using the LV model. But if you notice the asterisk (*) beside the “Gallup Tracking” line in the table above, the note below the table lets us know that this Gallup poll is using a different poll than they normally use. This one is using “post debate” numbers. The picture above has gone back to the 7 day “rolling” average and switched to the Likely Voter model – although it does not specifically point this out.
Just for comparison, yesterday’s RCP average snapshot (below) showed Obama up by +3 in the 7 day rolling average of Registered Voters.
So, essentially, moving from a Registered Voters only model to a Likely Voters model saw a 5 point swing for Romney. And since Gallup does not let us know how many R’s, D’s and I’s they survey, we have absolutely no idea if this swing is meaningful or not.
And as soon as RCP updated their average to the latest Gallup numbers, Romney popped to a 0.7% lead – his first!
So, the questions remain.
- Did Gallup simply move to LV’s today because it was 4 weeks out? (And will their data be LV’s from now on?)
- Did PEW sample more Republicans to be more accurate, knowing their reputation is on the line?
- Was IBD / TIPP influenced by other pollsters who believe Romney got a bump from the debate? Do they thing the bump is going to last through November 6th?
- Are pollsters really independent, or do they play “keep up with the Jones’s?
Well, I recently had the need to ask a pollster (and Professor) at Roanoke College in Virginia why he saw fit to release two polls on the Virginia Senate race that sampled Republicans by +2 and +3 respectively and then follow it up with a poll that sampled Democrats by +9 (and remarkably showed Tim Kaine (D) with a 10 point lead over George Allen (R).
I was, frankly, astonished by his answer.
The “political demographics “ of our two most recent polls are different. Our April Poll was an outlier in terms of being more “Republican” than most other polls.
So, the Professor intentionally sampled (using ‘quotas’) a lot more people in the heavily Democratic Northern Virginia area codes.
In other words, “we wanted to look just like the other polls” so we manipulated our samples.
Well, if that is the prevailing thought these days among pollsters, it is no wonder people don’t trust the results. Manipulating the samples to achieve an expected result is not polling. It is just making it up.