Re-weighting Polls.

Eli,

This should scare you. Make sure you go to unskewedpolls.com, mentioned at the end of this article.
Bill

http://pjmedia.com/tatler/2012/09/23/skewed-and-unskewed-polls/

Skewed and Unskewed Polls

by Charlie Martin

In most all things, I try to follow Hanlon’s (or Heinlein’s) Razor: “Never attribute to malice that which is adequately explained by stupidity.”

This is particularly important to remember when looking at polls, Sometimes, however, one must wonder.

As I pointed out yesterday, the result of Romney’s “really bad week” was that Romney had gone from 5 or 6 points behind in Gallup, to essentially tied. Even so, a number of people have noted that there are some odd assumptions in that poll, and others. Pat Caddell and Doug Schoen talked about it recently. Asked if the polls were, in his opinion, a fair representation of the electorate, Schoen said:

“The simple answer is no John. The bottom line is there were seven percent more Democrats in the electorate in 2008 than there were Republicans. That’s from the exit polls and that’s about as accurate as you can get….President Obama won by about seven points. Given 90 percent of Democrats vote for the Democrat and 90 percent of Republicans vote for the Republican, every time you reduce the margin between the parties by one point, roughly it’s about one point off the margin.”

Schoen pointed out that the Pew poll was based on Democrats sampled for having an 11 percent voters registration edge over Republicans. He further added, “saying that America has gotten more Democratic than 2008, which is a questionable assumption.”

In fact, Rasmussen keeps a running monthly poll of party identification. In the latest poll, released September 1, they found:

During August, 37.6% of Americans considered themselves Republicans. That’s up from 34.9% in July and 35.4% in June. It’s also the largest number of Republicans ever recorded by Rasmussen Report since monthly tracking began in November 2002.

Other polls — including Gallup — apparently have similar assumptions (called “turnout models”) in their polls.

There is a new website, called unskewedpolls.com, that basically reweights the data to fit the Rasmussen party identification. Their results are quite different, giving Romney somewhere between a five and eleven point lead.

Now, this should also be taken with a grain of salt. Basically, they claim (by the site name) to be an unskewed poll. In fact, they’re just a differently skewed take on existing polls. Instead of taking their numbers over, say, Gallup, though, what it should tell us is that even if the polls are being heavily weighted to Obama, Romney’s still essentially tied. Any difference in Romney’s direction in real turnout from the pollster’s assumptions would bring Romney into a lead.