Tuesday, October 28, AD 2008
What if the polls this year are wrong, or more wrong than they tend to be in most presidential election years?
Jay Cost of Real Clear Politics wrote this article today in which he noted that the different results of current presidential polls, “are disagreeing with one another in ways that can’t be chalked up to statistical “noise.” That gives me great pause.”
I believe that there are two main factors causing problems in the polls: partisan id and turnout models. Most polls are designed to count a certain percentage of Democrats and a certain percentage of Republicans to mirror the strength of each party. Most polls give an advantage to the Democrats, but how much of an advantage is up in the air. Nationally the Democrats in 2006, a bad year for the Republicans, had a two percent, or two point advantage. How much of an advantage to give the Dems this year? That varies all over the lot. Zogby gives a two point advantage. Rasmussen gives a 7 point advantage to the Dems. Most of the major media polls have given Dems an advantage of 9-15 points. My own opinion is that the Dem advantage is really 3-4 points this year. What is true of most national polls is also true of most state polls, a great variance in the partisan advantage given to the Democrats.
As for turnout models, the Gallup poll today graphically demonstrates the problem: 2 point difference for traditional turnout and 7 point difference if a lot more people go to the polls than in a normal presidential election year. Polls are all over the place in regard to turnout models and this greatly skews the results from poll to poll. Against current popular wisdom, I doubt if there will be much of an increase over turnout either for 2000 or 2004.
A good site for those who wish to learn quite a bit about polling and how the current polls may be very wrong is The Hedgehog Report. The site is pro-McCain, but they have a broad interest in polling and statistics in general and have been around since 2003. One commenter to look for in the comboxes is Polaris. His knowledge of polling is encyclopedic and his analysis of current polls is always well worth reading.
For a bit of perspective on the current polls, here are the polls for Bush v. Gore on October 28, 2000. Gore of course went on to win the popular vote while losing in the electoral vote.