Thursday, September 23, 2010

Pollster Skew

Since last fall the liberal refrain has been that Rasmussen is really a Republican pollster and that they were lying about polls to help Republicans. The evidence of this was that Scott Rasmussen had once worked for a Republican. Rasmussen was polling "likely voter," while all the other pollsters were polling "registered voter." That accounted for some of the difference. Now that everybody has switched to "likely voter" I decided to check to see how skewed the polls were.

I took any gubernatorial or senate election where two or more pollsters had polls within a week of each other during the month of September. I used August for a few that haven't been polled lately. I then compared the poll results of two pollsters, comparing them in pairs. I then combined them to determine how far apart each pollster was usually from each other. I only used the five most prolific pollsters due to sample size.

These numbers are relative to each other and shouldn't be read as whether any have a Republican or Democratic skew. I present the numbers as how Republican the pollster is, but, if I used Quinnipiac as the floor it'd show how Democratic the pollster is. We don't know which pollster is correct, as we'll only know that when the election comes. You can look at a Survey USA poll and determine how the poll would've come out if PPP had done it.

It turns out that Rasmussen doesn't skew Republican compared to the other polls. Democratic pollster PPP has almost identical results. Sorry, Democrats. Turns out you should've been bashing the others and complimenting Rasmussen. It's shocking to see that Quinnipiac is so far off the others. They don't have a pro-Republican reputation.

In the last few days, Rasmussen, Quinnipiac, and Survey USA have polled the New York Senate-B race. Rasmussen has Gillibrand by 10. Quinnipiac has Gillibrand by 6. Survey USA has Gillibrand by 1. Quinnipiac usually differs by around 6% of Rasmussen. So these results aren't surprising for those two. Survey USA skews a lot more Republican than you'd expect.

So next time you see a poll, decide which pollster you like and then adjust the other numbers toward them.

No comments:

Post a Comment