Our poll (coincidentally, the final poll for Montana) gave a projection for Greg Gianforte on the eve of the special election to replace Ryan Zinke, President Trump’s pick for the Interior Secretary. As most had suspected, the body slamming incident happened far too late to significantly change the results. Reports indicate a contrary outcome to media opinion; the base was invigorated, as demonstrated by Gianforte posting one of his best fundraising hauls yet in the aftermath of the news. The media bashing played into a narrative that is very popular among the Republican segment of the population, and thus these voters became more motivated to turnout.
The conclusion is that high profile incidents like increases overall turnout — not just one side. If the district is heavily Republican then this results in turning out a lot more Republican voters than Democratic voters. Unfortunately, we don’t have an alternative sample universe where no body slamming event occurred with which to compare vote totals. Without a do-over election with no body slam, we will never know.
One thing that was especially interesting about this election was how scattered the polls were, especially in the final stretch. Despite a doom-and-gloom narrative playing out on both sides,, the polling in US elections has been fairly accurate at a national level. But the most recent poll going into the election had Rob Quist up 14%, while the poll before had him down 14%. Some village criers have taken that to mean that polls cannot be trusted. What needs to be re-assessed here is proper communication of what a poll actually is and how by how much they are capable of missing. Individual states are harder to poll; there still isn’t a “good” answer about what went wrong in Midwest state-wide polling for the 2016 presidential election. Weighing by college education was one proposition, but that hasn’t explained the variance fully.
Not to become our own cause-celebre, but our final result was fairly close to the final outcome. “Nailing” a polling outcome is, by any stretch, a random event. Polls capture a range; that’s why we report the margin of error in the first place. I accredit the strength of our online poll to our noise removal and electorate modeling. Our raw data had the outcome tied at 50-50, but we were able to deduce oversampling relatively quickly. When we factored for the age, race, and geography of voters involved, it became a lot clearer. A lot of people will use models generated from a national scale, or even just go off the demographics for the state as a whole. Coming up with models for each election takes time and effort, but I think the results speak for themselves.
So what does this mean for future special elections? Well it’s lost on no one that these special elections have been a lot closer than they were just a few short months ago. It’s important to note that Montana was a state that Democrats can — and do — well in. Its’ a state that elected a Democratic governor and had previously rejected Greg Gianforte for that post. All things considered, this should have been even closer than it ended up being.