16 July 2018 — 1621 mdt
There’s a high probability that Tester is leading Rosendale
President Trump spoke in Great Falls on 5 July, denouncing Democratic Sen. Jon Tester and praising Republican senatorial candidate Matt Rosendale, who currently serves as Montana’s auditor. Three days later, the Remington Research Group began a massive robopoll that ended on 10 July. The result? Tester had a three-point lead, down from an eight-point lead a month ago.
The key statistic is the probability of a ballot lead, which can be determined by the American Research Group’s online calculator. In the Gravis poll, conducted in early June, the probability that Tester was leading was 99 percent. In the Remington poll, Tester’s lead was down to 94 percent.
Is Rosendale closing the gap? That’s not an unreasonable conclusion. The Remington poll suggests Rosendale may have gotten a boost from Trump’s rally in Great Falls. And it’s possible Rosendale may get another boost from Vice President Mike Pence’s rally scheduled for Billings on 25 July.
Both Tester and Rosendale, and perhaps third parties, undoubtedly are polling heavily before and after event’s such as Trump’s, either as a part of or in addition to their tracking polls. The large sample for the Remington poll suggests both a close race and an effort to precisely calibrate the effect of a presidential visit to Montana.
Some caveats:
- The margin of error is a standard statistic that describes how close the mean of the sampled population will be to the mean of the actual population 95 percent of the time. Other sources of error, which can be additive as well as canceling, include how the pollster weights the responses to bring the demographics of the sample into agreement with the population sampled.
- The method of contacting the voter varies with the pollster, and can have an effect on the poll. According to FiveThirtyEight, Remington uses Interactive Voice Response (IVR) — robopolls — and live calls. Gravis uses IVR and online queries. FiveThirtyEight gives Remington a "C" and Gravis a "B" for overall quality.
- Campaigns tend to make their polling results available publicly when the results are good news for their candidate. The Gravis poll was paid for by Gravis. I'm pretty sure the Remington poll was paid for by Rosendale or a third party supporting Rosendale. Adverse polling results are almost never released by a candidate's partisans.
- After a poll is released, the campaign for the candidate for whom the poll is bad news begins spinning the poll as flawed, unreliable, biased, probably unAmerican, and a distraction to be disregarded. Sometimes the criticism has merit, but usually it's proof that the poll got it right.
Finally, the polls in 2006 and 2012 were close, usually reporting that Tester’s lead was less than five points and often just one or two points. Compared to his previous campaigns, he’s doing at least as well and probably a bit better. But it’s going to be close.