Over the past week the skewed polls meme has taken over much of conservative media. According to many conservative critics, the current set polls, which show President Obama leading nationally and in key battleground states, are all wrong because they over sample Democrats. To correct this “skewing”, Dean Chambers of unskewedpolls.com has recalculated every poll with more Republican-friendly party identification samples. Chambers essentially reverses the results of every poll, showing Mitt Romney with a big lead even though the pollster reports otherwise.
However, unskewing could theoretically work both ways. Many liberals believe the polls may actually be skewed towards Republicans because they exclude cell phone users, include too many Republicans in their sample, or include too many automated polls.
So what happens if we “unskew” the polls in the Democratic direction?
- Obama leads Romney 53.78% to 46.22% nationally, larger than the 4.3 percent lead Obama has in the Real Clear Politics (RCP) average of polls.
- President Obama job approval is at 51.14% with 45.86% disapproving, greater than the 2.7 net-positive job approval Obama has in the RCP average.
- Obama lead in Florida 53.33% to 46.67% for Mitt Romney, larger than the 4.3 point lead Obama currently has in the RCP average.
- President Obama lead in Ohio 54.79% to 45.22% for Romney, larger than the 5.4 point lead President Obama currently has in the RCP average of polls.
- President Obama leads in Virginia 53.86% to 46.14% for Mitt Romney, much greater than the 3.7 percent lead Obama currently has in the RCP average.
This “unskewing” was accomplished through the following means:
- Obama numbers increased one percent to compensate for lack of cell phone sampling. Romney’s numbers are decreased by one percent. Research done by Nate Silver at the New York Times found that polls that include cell phone-only users tended to report better results for President Obama.
- Rasmussen polls are excluded entirely. First, Rasmussen is excluded because the polling outfit has a track record of favoring Republican candidates by three-to-five points. Secondly, Rasmussen is excluded because their party identification split has more Republicans than Democrats, which contradictions nearly every other polling organization.
- President Obama is given two-thirds of the undecided voters. A recent study found that undecided voters are breaking towards President Obama, and research done by Nate Silver found that the incumbent party tends to gain 4.6 point from the late September polls until the actual election.
Let me make it very clear that I do not actually believe the “unskewed polls” in either direction are accurate. Given the current state of the “skewed” polls, President Obama will probably win, but not by the margins listed below. The purpose of making the list above is simply to show how any poll can be “unskewed” in either direction, and that unskewing a poll does not necessarily make it more accurate.
What this set of unskewed polls, and the Republican version both do is to assume the best for their side. Chambers assumes that Republicans will turn out in larger numbers than Democrats and that undecided will favor Romney. The Democratic version of unskewed polls assumes cell phone users are not being sampled enough, that Rasmussen is biased against Democrats, and that
The truth actually probably lies somewhere in the middle. Rather than assuming everything will go the way of Republicans or Democrats, the pollsters simply ask questions to respondents, and then report their numbers assuming neither side wins all of the unknowns. What the pollsters are reporting is the real version of “unskewed polls.” The skewed polls are the ones that appear in this article, and on unskewedpolls.com.