Tuesday, August 31, 2010

GPI update, another weird Humphrey poll

With the release today of a new MPR/Humphrey poll I can once again update the GPI. Below is a list of the current polls included in the index followed by the current numbers.

Image Hosted by ImageShack.us

Image Hosted by ImageShack.us

Joe posted an excellent analysis of the Humphrey poll earlier today pointing out some of the issues with it. These are not the kinds of issues that disqualify a poll from being included in the index and is in fact the main reason for wanting to get a composite number from a combination of polls rather then relying on a single poll.

Having said all of that, partisan sampling problems do seem to be something of a trend for MPR/Humphrey institute polls. I posted this about their governor's race poll from May:

...the MPR/Humphrey Institute poll's sampling was 42% DFL, 42% GOP and 15% Independent. I honestly don't know where they got these numbers from, perhaps their rectum, but they certainly are not representative of Minnesota's partisan breakdown. First of all 15% is an incredibly low number for Independent's in Minnesota.

This polls breakdown is similar to the Humphrey poll from May in that the partisan breakdowns in both polls look nothing like the partisan breakdowns being used by anybody else who's polled this state. Now, maybe Humphrey knows something that all of the other pollsters don't, but judging by their 3.15 pollster rating, the sixth worst score of any pollster in Nate's database, my guess is that their polling suffers from a flaw in the methodology somewhere.

From MPR's article on the poll:

The survey data has also been weighted to accomodate for factors such as the number of telephone lines, cell phone usage, gender, age, race and ethnicity to approximate the demographic characteristics of the state's population according to the Census.

There is no mention of weighting the partisan numbers, so they probably don't, which would explain why they ended up with the split they did and probably also explains why they have such a poor track record.

Putting that aside, and as Joe pointed out, no matter how you slice it this is bad news for Tom Emmer. In a poll that consists of 46% republicans he could only manage to get the support of 34% of them. In what has to be considered a best case scenario for Tom Emmer he's only at parity with Mark Dayton. Don't let the media reports over the next few days fool you, this race is not "deadlocked" right now, Mark Dayton is clearly in the drivers seat.

No comments:

Post a Comment