Monday, November 8, 2010

Polling in Minnesota Governor's races; part 1

This was intended to be one post but quickly developed into a monster that just had to be segmented. While I am certainly not above splitting posts up simply for the sake of more posts, that is not the case in this instance.

The following comment from ldc sums it up:

I haven't seen a good explanation for why they were so far off, not just MPR/Humphrey but others too. At least in the Nevada senate race, they could explain the difference in pollster vs real election results by the Latino vote and whether the pollsters gave a Spanish language option. So what is the answer here??

She is right in that there doesn't appear to be an easy explanation like in some other states for the inaccuracies of some of the Minnesota Governor's race polls. I have advanced the argument before that errors in 2006 polling was due to late breaking events in that race that caused undecided voters to break heavily for Tim Pawlenty.

What about 2002 than? In that year the pollsters correctly predicted a Pawlenty win, but missed the final margin by about the same amount as in 2006 and 2010, 5 points. On the other hand polling in Minnesota Senate races, and Presidential races, has been rather accurate. Is there something about Minnesota Governor's races that makes polling them especially difficult?

I said this prior to the election in reference to Nate Silver's projecting an 86% chance of a Mark Dayton win:

That said, there is still a 14% chance that everyone is wrong. One of the data points that makes up that 14% is of course the 2006 Minnesota Governor's race. In that contest Mike Hatch had a three point lead in the simple polling average 45.5-42.5 and lost 46-47, a four point swing for Tim Pawlenty.

Let's take a look at the polling averages of all the polls conducted in the last month in the last three Governor's races.

(The number in the lower right box is the difference between the margin of the polling average and the margin of the race)

Image Hosted by

Image Hosted by

Image Hosted by

The polling has been off in Minnesota Governor's races, on average, by between 4 and 5 points in the Democratic candidates favor. This tendency is not evident in Senate and Presidential elections however where the polling has tended to be much more accurate.

Image Hosted by

Image Hosted by

Image Hosted by

The two presidential races and the Senate race in 2008 were by far the most accurate and not coincidentally they were also the races that received the most polling.

Image Hosted by

The polling in the 2006 Senate race was off by more, but only about half as much as the three Governor's contests we looked at above even though it received the least polling of all the races.

There was another factor at play in that race that wasn't in play in the others, Amy Klobuchar had a clear and commanding lead in every single poll conducted. In races like that the undecided vote tends to break for the leading candidate which goes a long way toward explaining why the polls underestimated her margin at the same time they underestimated Pawlenty's.

Having established that the polling in past Minnesota Governor's races has been consistently wrong while the polling in other races in the same time frame have been generally very accurate we can move on actually answering ldc's question from the very beginning of this post; why?

There are two possibilities; either the polls themselves are wrong, or the polls are right and the GOP candidate is receiving a wave of support at the end of the race. The exploration of these two possibilities will be the content of part two, which I'll put up tomorrow.

In the meantime here's some laid back Monday music to enjoy:

No comments:

Post a Comment