Unlikely Voter

Tweet

I try my best to avoid various forms of bias in my polling analysis on this site, but some polls just look so ridiculous to me that I throw out any fears of confirmation bias and just go in with the plan of tearing the poll apart.

This Braun Research poll for cn|2 of the Kentucky Senate race is one of those polls.

Firstly, if you’re a pollster and you repeatedly emphasize you are from Princeton, NJ to make people think you’re affiliated with Princeton University, your cheap tactic actually makes me think less of your polls right from the start.

But going further: Braun conducted 803 interviews, and found 82% (660 of 803) of those surveyed to be “likely voters.” We don’t get 82% turnout in Presidential elections, not in Kentucky or any other state. Why would anyone expect 82% for a midterm? This isn’t even close to a credible result.

For comparison: According to Wikipedia, In 2004 the turnout nationally was at 57%. In 2008 in Kentucky, the State Board of Elections says that 64% of registered voters turned out to vote. 82% of adults showing up for a midterm just isn’t in the realm of credibility, sorry.

Now, it’s possible that 82% of the actual sample, found using random digit dialing, wasn’t likely to vote. Braun took the random data and then fudged it using “Statistical weights” that “were designed from the United States Census Bureau statistics.” No, it doesn’t say how they were designed.

At this point I would love to compare Braun’s readings of votes for Obama or votes for parties in the Congressional Districts in 2008, but no such questions were asked. There is only one real world check I can find to test this poll’s weighting with the real world, and that is party registration.

Again turning to the 2008 statistics from the Kentucky State Board of Elections, we find that 1,076,583 Democrats voted and 688,561 Republicans voted out of 1,861,577 total votes. That gives us 58% Democrats and 37% Republicans. The poll gives us 54% Democrats and 38% Republicans. So Braun wants us to believe the 2010 electorate will look like the 2008 electorate swung 5 points to the Republicans, and in particular Republicans will keep the same proportion of the electorate.

Seriously: Braun wants us to think Republicans are not going to make up a greater proportion of the electorate in Kentucky in November 2010 than in November 2008.

I just don’t believe it.