22 October 2014

Latest U.S. Senate & House polls — and limitations thereof

Montana’s elections for the U.S. House and Senate have been sparsely polled this year, and some of the polls raise red flags on methodology and/or competence. This is the natural consequence of our having a Senate election that the Democratic Party botched, and botched big-time, and a much closer House election whose outcome won’t decide control of the House.

The publicly available polls are graphed below, but some caveats are in order.

1. All pollsters try to get it right. Not all do, but all try.
2. Out of economic the samples are small, usually around 600 persons. That yields a margin of error of ±4 percent at the .95 confidence level (19 out of 20 times, the poll’s results will be the reported value, plus or minus four percent).
3. We cannot determine whether an isolated poll is an outlier.
4. Samples are weighted to approximate the population’s demographic composition. Weighting is not easy and errors do occur.
5. That a poll leaned one way or the other in the last election does not allow us to derive a correction we can apply to current polls by that pollster. Why? Because pollsters that identify a bias adjust their methods to try to correct that bias.
6. The lower the turnout, the harder it is to poll accurately.

The Yougov poll essentially puts out a call for input, which is received through the Internet. When the size and demographics of the sample match Yougov’s targets, the poll is processed and the results released. It’s more complicated than that, but that will give you a sense of how Yougov operates. Both polls displayed in the graphs below report what seems to be a severe undersampling of the age 18–29 cohort. Whether the weighting of the sample corrects for that is unknown.

The Montana State University at Billings’ poll is a school project that probably isn’t as sophisticated as polling companies that conduct polls on a continual basis. The sample is small, and the number of undecideds is big — too big, in my opinion. There could be that many undecideds, but I suspect the more likely explanation is MSU-B’s inability to sufficient push the people polled to make a choice. Below the graphs are tables displaying the MSU-B’s Senate and House numbers for 2012 and 2014.

Daines leads Curtis by double-digits in all of the polls. That’s not a surprise. Max Baucus could have been re-elected. Brian Schweitzer could have beaten Steve Daines. Instead, Baucus is in China, Schweitzer is in Playboy, and any reasonable chance Democrats had to retain the seat expired when the plagiarism that led to John Walsh’s defrocking by the Army’s war college was exposed. Curtis’ future is bright, but her prospects on 4 November are dim.

John Lewis trails Ryan Zinke in the MSU-B poll by single digits. The difference is within the margin of error, but the probability that Zinke is leading is 92 percent. That could change. If Lewis were running in a Presidential year, I think he would be leading Zinke by a few points. So even if Zinke wins, he’ll be vulnerable in 2016.

PDF for printing
Download spreadsheet

PDF for printing

Montana State University at Billings Poll