A reality based independent journal of observation & analysis, serving the Flathead Valley & Montana since 2006. © James Conner.

23 May 2017 — 1547 mdt

Latest Gravis poll has Gianforte leading by 14 points

A Gravis poll of Montana special congressional election that was released today and apparently conducted yesterday reports Republican Greg Gianforte is leading Democrat Rob Quist 49 to 35 percent, with eight percent favoring Libertarian Mark Wicks and nine percent undecided. Gravis sampled 818 likely voters.

Earlier in May, a Google Consumer Survey of Montana reported Quist was leading Gianforte 48.5–41.9, with 95 percent opting for Wicks. The sample for this survey was small, and it was not weighted to compensate for the demographic differences between the sample and the population surveyed.

Update, 11 October 2019. I downloaded the spreadsheet for this poll today. The topline reports the results were weighted by age, gender, and region. I then reviewed the spreadsheet for the poll that I downloaded in 2017. The topline does not mention weighting, and the complete results array does not report a weight.The analysis below is wrong, but it was correct when written.

The poll’s results were based on a sample of 288 drawn from Montana’s age 18+ population, not registered voters. No data on educational attainment were collected, but I suspect that college graduates may have been oversampled. When “I am not likely to vote in this election” responses are omitted, the sample decreases to 241 (MOE 6.3%), or to 207 (MOE 6.8%) if the age or gender is unknown. The poll reports a 6.5 percent lead for Quist. He lost by 6.4 points.

I’ve never been able to identify the person or persons who conducted this poll. I do have some likely suspects, but until I have proof, I’m not mentioning names.

First, the polls presented graphically. After that, some discussion.

The Google survey sampled 339 computer users with Montana IP addresses. Almost 30 percent said they were not likely to vote in the election, leaving 241 responses that expressed choices and, one suspects, probably will vote. Along with a companion poll conducted a month earlier, this is the only poll reporting Quist in the lead.

A spreadsheet with the poll’s raw data can be downloaded from the survey’s website.

Here’s how the GCS broke down by gender:

google_gender_breakdown

Almost 13 percent of the 241 might vote responders failed to reveal their identity. That’s high. Otherwise, the breakdown does not seem unreasonable, at least not at first glance.

But the geographic distribution of the responses suggests a problem:

google_regional_distribution

The heavily Republican Flathead is under-represented, while Democratic leaning Lewis and Clark, Gallatin, and Cascade Counties are over-represented. Democratic Missoula is slightly under-represented. I strongly suspect that if the survey’s geographic and perhaps age distributions were weighted to normalize the sample, the gap between Quist and Gianforte not only would narrow, but would be inverted. Determining whether a GCS can be weighted successfully would be a good project for a graduate student in the social sciences.

The Gravis poll raises serious concerns for Quist’s camp. If the poll is correct, he’s lost a bit of ground.

Meanwhile, reports Ed Kilgore, the attorney and political writer for New York magazine who worked in Georgia and Congress, and who lived in Georgia’s Sixth Congressional District, Democrat Jon Ossoff is polling over 50 percent and may be pulling away from Republican Karen Handel.

If national dynamics are at work, the distance between Gianforte and Quist should be closing. Last week, various sources all reported the contest had tightened to single digits. Until the latest Gravis poll was released, there had been no sign that Gianforte was becoming stronger while Quist was weakening. Do not dismiss the Gravis poll, but do not be surprised it it turns out to have very wide error bars.