Friday, October 26, 2012

Poll Excitement

Just to inject the proper note of caution to all those excited Republicans outs there.  Rasmussen had the best results of the pollsters in 2002, 2004, and 2006.  But they were only average in 2008, and in 2010, below average. In all those years, Rasmussen tended, though this was not true of every poll, to overestimate Republican strength.  The numbers that have conservatives so giddy might be a few points high.

As in all elections, however - and this seldom comes up in all the criticism of which polls are better than others - turnout is hard to predict.  The polling experts like to talk about cell phones and windows and sampling - because they all have to lower their voices when it comes to turnout.

6 comments:

james said...

If poll response rates are really as low as 9% (we're in the 91% who refuse to answer), trying to estimate the biases in the results is a daunting job. I think the error bars on the approval ratings should be a lot higher than merely statistical. My gut reaction is that the systematic error is at least a factor of 3 larger than the statistical, and maybe more.

I figure the parties' jobs are to persuade me to support their candidates. If one succeeds, I don't care about the polls. If nobody succeeds (or I haven't been paying attention) the polls tell how persuasive the candidates are when they can pull out all the stops and promise the moon. Then the polls potentially convey information, but not very useful information.

Sam L. said...

Polls are interesting, but not dispositive. Only the one where all the votes are counted (or miscounted), counts.

james said...

There is one class of polls that is amusing, if presumptuous: how Europeans or Middle Easterners want the election to come out.

Dubbahdee said...

And excellent discussion of polling on the Diane Rehm show on NPR last week. Much discussion on the factors, such as turnout, that affect polling accuracy. I was not able to listen to the entire program but what I did hear was very useful.

http://thedianerehmshow.org/shows/2012-10-22/understanding-political-polls

Roy Lofquist said...

In re the 9% poll response rate:

Two notable examples of self-selection bias are the 1994 presidential race exit polls and the 2012 Wisconsin gubernatorial recall exit polls. In both cases he exit polls showed a significant biad towards the Democrat.

Gringo said...

Roy Lofquist
Two notable examples of self-selection bias are the 1994 presidential race exit polls and the 2012 Wisconsin gubernatorial recall exit polls. In both cases he exit polls showed a significant biad towards the Democrat.

I doubt that such a bias was shown in the 1994 presidential race, as that was an off year.