The data and charts are posted here on Political Poll Check.
The Short Version:
After a few weeks of collecting and compiling data for Political Poll Check, we finally saw our first poll come in for a second time around (since we began collecting data). This gives us the first opportunity to compare the same poll head to head.
We also have collected enough data to begin seeing “trends” in movement in the normalized polls.
And the movement is not in favor of Mitt Romney. After normalizing the polls to account for “skew” Romney still leads in the average, but the lead has dropped from +7 points to +5. Of course, the accuracy of the +5 depends on the turnout. If Republicans outnumber Democrats at the polls by 3%, the number should be pretty close (if the election were held today). If the turnout is about even, then the Romney totals will be down a bit, probably +2.
If more Democrats actually turn out than Republicans, a lot will depend on the Independents, as it always does. The normalized numbers also count the Independents based on the ratio found in each poll.
The Fox News poll was also not good news for Republicans. While the poll showed the same +5 margin for Obama in the new and older poll, the newest poll was less skewed towards Democrats. After normalization, the older poll – 2 1/2 weeks ago – showed Romney up by +3. This one, after normalization, has Obama up by +2.
Based on the data, Romney is definitely losing ground.
A few weeks ago, Virginia Right began a new web site Political Poll Check with a mission of finding the truth buried within the polls that all skew to the left. One of the first things we needed was, obviously, data. And a method to do an apples to apples comparison of these polls. It is our belief that a poll that over samples Democrats by double digits cannot then be compared to a poll that gives Democrats a 3 point advantage. And while a poll with a heavy Democrat skew may show Obama with a 8 or 9 point lead, it can’t really be concluded that a poll with a less favorable Democratic skew and a smaller Obama lead is showing an erosion of support for Obama or a gain in support for Romney.
A lot of folks in the media believe that Real Clear Politics, a site that averages all of the polls, is one viable option to “smooth” out the poll swings due to different sampling rates. And we believe that there is some validity to this belief, and the averages at RCP have some value in telling us what is happening, we have come to believe that averaging alone cannot make up for the illogical skewing that is happening with almost all of the polls. If there were an occasional poll that was “out of line” with the majority, averaging would be a valuable tool.
Unfortunately, Rasmussen seems to be the only polling outfit that uses common sense and does not systematically use completely ridiculous demographic models that predict a much greater Democratic turnout in 2012 than in 2008.
There is simply no logical justification for that. Enthusiasm for Obama is down by a large margin. Voter registration in deeply blue areas is down, the new batch of college age kids are not registering in the numbers they did in 2008 and Obama is having trouble raising both money and crowds.
Real Clear Politics does a good job of averaging. But their average is not really telling us how the voters are trending as much as it tells us the average of how the polls are being skewed. When you average badly skewed data, the output is still badly skewed. Garbage in – garbage out. And this is not the fault of the folks at RCP. They average what is out there. They are not responsible for the political agenda built into each poll.
Political Poll Check took a new approach. First we grade each poll. The poll is heavily weighted to show the trustworthiness of the poll’s data. If the pollsters hide the demographics from the public, they can’t really be trusted. But more importantly, we can’t determine the makeup of the poll to adjust for the skew. The grading system is designed to weed out the polls that are impossible to “normalize”. A low grade does not necessarily mean the poll is no good. As long as it passes – even with a “D” – it will tell us a lot after we normalize the data to a standard.
the standard we decided on was an average of the 2012 Rasmussen Poll of Political Party Affiliation. They publish a monthly breakdown of Republicans, Democrats and Independents. This is the same poll, the same methodology and Rasmussen is very good with this kind of polling. There is no “skew”.
The current Party breakdown according to the Rasmussen poll is Republicans 36%, Democrats 33% and Independents 31%.
A number of people disagree with this polling data and believe more Democrats will turn out that Republicans. And that may be true. But for the purposes of normalization, that does not really matter. What matters is the trend after normalization. Once we eliminate the skew, all the polls are equal. An average of “normalized” polls is going to be more accurate than using the polls numbers right out of the box. The only question is the voter turnout.
But after the polls are adjusted to a standard baseline, it is simple to adjust the numbers based on the turnout. We could have used 33% for all of the party data and if more from one party turn out they will win. The exact voter turnout and makeup of the turnout will always be a wild card until after the polls close.
But after a poll is normalized, two things become far more accurate. And they both have to do with trending.
First, our average of normalized polls started at Romney +7. Since our normalization process “over samples” Republicans at 36% (compared to 33% for Democrats) if the actual numbers are different, then the +7 will actually be different. The value of this number is not as important as the movement of this number. As we add in more normalized polls, the trend of positive or negative movement will be an accurate measure of actual movement for or against Romney.
Over the past few days, as more polls are processed, the Romney +7 went down to Romney +6 and is now Romney +5.
The Republican campaign is trending downward. There is no doubt about that. You could also conclude that Obama is trending upwards, or there is negative movement for Romney and positive movement for Obama.
This analysis also allows us to see trends in head to head polls. Today’s Fox News poll is the second from Fox in the past 2 1/2 weeks. Fox’s numbers showed that Obama has held a +5 point lead in both polls. However, the latest poll sampled a different mix of party demographics than the older one. The new poll still samples more Democrats than Republicans, but the margin is closer. The old poll had a skew of _9 Democrats and the new one only has Democrats favored by +6.
Normalization after the older poll actually showed Romney ahead by +3. And after normalization this time , Obama has a +2 advantage. That is a fairly large swing, but the margin of error is + or – 3.0, so both are still within the margin.