22 October 2018 — 1542 mdt
It’s finger crossing time for Democrats
Today’s New York Times story on the tightening U.S. Senate election in Montana coincides with Montana State University Professor David Parker’s Twitter announcement that the results of MSU’s poll of the Senate race will be announced at 2100 MDT tonight on CW TV stations.
The poll’s findings on the U.S. House election will be reported tomorrow evening.
The last publicly available poll, by Public Policy Polling, was released and conducted in late September. The University of Montana’s Big Sky Poll, conducted in the last half of August and released in early October, was an outlier and should be ignored (there’s actually a good case the poll should be withdrawn).
The margin of error, “statistical ties,” and ballot lead probabilities. The margin of error is a measure of how close to the population value one should expect the sample value to fall. The larger the sample, the smaller the MOE. The PEW Research Center has an excellent discussion of the MOE.
What we really want to know, of course, is the probability that Smith is leading Jones. Unless Smith and Jones are truly tied at, say, 48–48 percent, there’s a probability that the candidate in the lead is actually leading. That probability can be calculated, and a convenient way of making the calculation is online at the American Research Group. For example, if the sample is 600, and Smith is leading Jones by three points, 49 to 46 percent, the MOE is 4.0. Although the difference is within the MOE, and sometimes is called a statistical tie, it’s neither a tie nor statistically significant at the .95 level. But the probability that Smith is leading is 77 percent.
Sampling variation is not the only source of error in a poll. How the sample is weighted introduces variation, widening the error bars, as does the wording of the questions. All reputable pollsters try to get the right answer, both because of professional pride and out of their enlightened economic self-interest, but even the best polls produce results that have a certain measure of uncertainty, which tends to confuse some and infuriate others. Generally, polls in the United States are not that far off the mark. See Nate Silver’s discussion at FiveThirtyEight.com.
Flathead Memo will analyze the MSU poll and add its results to the graph above.