The first poll in MN CD 1 has come out and it is quite a surprise. SurveyUSA has Walz (D) at 47% and challenger Randy Demmer (R) at 42%, which puts the two of them within the 4.1% margin of error. That would constitute a toss-up status. It would also belie the conventional wisdom in Minnesota that the Tea Party and conservative groundswell is not happening here. What’s even more astounding is the numbers they are using for this district. When I first looked into the metrics of the sampling, I had to laugh out loud. While it would still be a rather close race, Demmer should be in the lead given the demographics and skewed sampling data.
SurveyUSA decided to use RBA (Registration Based Sampling) for this poll instead of Random Digit Dialing(RDD), such as that used by Rasmussen. The theory behind this kind of polling is by measuring already established population metrics, like the number of registered Republicans, or Democrats, or independent voters, they can more accurately assess the likelihood of a candidate winning. Since people, in normal times, tend to identify with one of those categories, it is a stable way to measure public opinion. In Normal Times!!!
Testing of this method in 2002 and subsequent election cycles found this to be a decent sometimes more accurate measure of voter intent. During Red State/Blue State days of yore, this idea had merit. When people registered with a political affiliation they tended to stay within that affiliation when voting. It made sense, until today. Political affiliation was a prime indicator and so the other metrics, size of family, income, belief systems, etc. would all become rather moot. But, it has lead to problems when the electorate began to change from pro-Democrat to independent. (Still not very pro-Republican).
Random digit dialing is the methodology that takes out any bias in the sampling. It is presumed that given a truly random collection of data, given a large enough sample population, and weighting the results to follow demographic averages, a snapshot of the American body politic could be taken. RBS comes from a different angle, and one that in today’s political climate makes for very unreliable and invalid results. Since so many people are moving from a party affiliation to an independent status, suddenly the political affiliation metrics from 2008 or 2006 become worthless. As more and more voters become angry with the political establishment, they will not necessarily change their political registration but may say they are independent or leaning Republican in the phone survey. Once the percentages of “Republican” or “Independent” are fulfilled from the old registration lists, the surveyer must find more Democrats to fill the slots.
This leads to over sampling of Democrats who are still willing to admit they’re Democrats while under sampling the other two more vital categories.
That’s why the demographics are so skewed.
In this poll, the sample they used found 48% of southern Minnesotans as college graduates. Now, southern Minnesotans are incredibly smart, but they do not have a 50% college graduation rate within their population. The percentage for the state as a whole is 31% according to census data. That would make southern Minnesotans far more likely to have completed college and people in the rest of the state more likely to have just completed high school and nothing more. It just doesn’t work. The sample is not representative of the population, period.
Another incredibly skewed number has responders expressing pro-choice views versus pro-life views 49% to 48%. This number also shows just how completely ridiculous the demographics are in this sample. Southern Minnesota is decidedly pro-life and any poll that shows its responders are pro-choice is a poor sampling of the population. The numbers get even worse when looking at ideology.
Southern Minnesota has a large, proud conservative population that is more likely than not to say they are conservative. Yet, only 37% in this poll profess to be conservative while 43% consider themselves moderate and 15% liberal. Given the social climate of this area, many Democrats call themselves conservative so these numbers are absurd. Heck, Tim Walz, the incumbent Democrat, talks about himself as a fiscal conservative even as he spends money like a drunken sailor on leave.
Instead of randomly selecting people and collecting accurate raw data to process, these pollsters are taking old registration numbers, presuming they are accurate, picking specific numbers of affiliated voters, and calling them until someone answers and gives them the data they want. We don’t really know what the voters of southern Minnesota believe, but we do know what a few Republicans, a few independents, and a whole lot of Democrats believe.
This doesn’t really give Walz and the Democrats much hope. Demmer is still within the margin of error even given the skewed metrics of the poll.
I cry DOOM!!!
Crossposted at Looktruenorth.com