The website is seeking to mitigate republican fears that Romney is slightly behind Obama by tweaking the partisan tilt rightward, rejecting polled data, and weighting them in favour of their candidate based on an extremely favourable formula from a Republican-leaning firm.
Now, you can believe it if you want, but you’re deluding yourself and you deserve to know why.
Here is why UnskewedPolls may paint a charming picture for a conservative, but isn’t statistically right.
The argument they make is that polls are oversampling Democrats, and that this is causing a more significant lean in the polls for Obama in the presidential election.
Last week, we looked at why the idea that polls deliberately oversample democrats is wrong, and why there are a bunch of legitimate reasons why polls will say they talked to more Democrats than Republicans.
The easiest solution to this is that (a) there are more registered Democrats than Republicans after a six year blitz of voter registration and (b) that many Romney voters identify as independents, not Republicans.
But that’s not even what makes UnskewedPolls misleading.
Here is where the weighting system is off: from what we can ascertain of their statistical method, they don’t really have one.
They just assign a certain weight to a poll to equalise the sample out to Rasmussen Reports’ cozy idea of the 35.4% Republican, 34% Democratic, 30.5% Independent split, a number disputed by nearly every other polling firm in the entire country.
Polls which dispute Rasmussen’s spread of Party Identification have been conducted by the following credible organisations, many as recently as last week:
- The Associated Press/GfK (D 31%, R 22%, I 29%)
- Pew Research (D 35%, R 24%, I 36%)
- CBS / The New York Times (D 35%, R 22%, I 36%)
- ABC / The Washington Post (D 34% R 24% I 34%)
- The Washington Post (D 32%, R 25%, I 37%)
- Bloomberg (D 32%, R 27%, I 39%)
- The National Journal (D 33% R 26%, I 26%)
Why is this? Well, as the polling aggregator FiveThirtyEight’s founder Nate Silver argues: Rasmussen Polls “were biased and wrong” in 2010.
Their information was inaccurate, and there’s reason to believe that since their polls are such significant outliers this time around that they are still statistically biased and in accurate from a mathematical perspective.
There’s nothing wrong with believing in a wrong poll, honestly, but you’re deluding yourself if you buy into it.
Essentially, UnskewedPolls are really just skewing individual polls to fit a preferred reality, when what they should be doing is looking at an aggregation of polls weighted for their quantifiable historical biases and correcting for them to give an appropriate picture.
Then again, this all comes down to what you want in a poll. If you want a political talking point, some thing to spin to favour a preferred candidate, that’s your own business, enjoy UnskewedPolls.com and enjoy life in a self selected vortex of spin.
But if you want to use a poll to gain insight on a race or to learn more about the public perception of an issue, just understand this: the methodology, the mentality, and the mathematics of UnskewedPolls.com leads to some incorrect conclusions.
Business Insider Emails & Alerts
Site highlights each day to your inbox.