Data Analysis: Gianforte’s Voters Weren’t Swayed by Assault

I was curious about whether Greg Gianforte’s assault of a reporter had a meaningful effect on Montana’s election. Would Rob Quist have won if the assault had happened before absentee ballots had been sent?

I consulted all of the social science literature on effects of a leading candidate assaulting a journalist on election eve.

Just kidding. There isn’t any.

When we heard about Gianforte’s assault, we were very disturbed by his actions, but we realized we were in a good position to see what the effect would be.

We’ve been surveying people in Montana about the election, yielding a very good snapshot of what the state looked like politically before the attack. After the assault, we resumed surveying. Our resumed survey received 911 responses between 8 PM MT election eve and 11 AM on election day.

The overall responses pre-assault and post-assault were almost identical:

This includes self-described “definite” voters only — among probable voters Gianforte had a stronger lead. But it’s worth asking some more specific questions.

Republicans were overwhelmingly supporting Gianforte before the assault. Did the assault affect which candidate they support?

Not much. Here’s a comparison of our survey data on Republican early voters and Republican election day voters:

The combined sample above is 321 voters. 89% of Republican early voters reported voting for Gianforte. 85% of Republican election day voters said they would vote for him, even after the assault. Here were the results of our original pre-assault survey, which combined the two groups:

These numbers, based on a larger sample (673 votes) look more like the “definite election day” numbers above — leading us to believe that the 3–4 point effect size implied in the first table is on the high side (noise is an issue in such small samples).

But let’s use the high estimate and assume that the assault caused Gianforte to lose 3.5% among Republicans. Multiply that 3.5% by the 40% of the Republican electorate who’d yet to vote, multiply that by approximately 55% (the Republican share of the electorate) — and you get a total of less than 1% of the electorate changing their votes.

In other words, Gianforte’s Republican assault-incurred vote loss was below 1%.

Did the assault have a significant effect on Democrats’ votes?

No.

Rob Quist had at least 98% support among Democrats before the assault. His support among Democrats was so high that there wasn’t really room to go up.

What about Libertarians?

Libertarians are a big group in Montana. Many candidates (including Mark Wicks in this race and Gary Johnson running for President) have received around 6%.

Before the assault, Libertarian preference was fairly evenly divided across the three candidates, with each of the three candidates getting at least 25% support. As a group, they seem more persuadable.

Among Libertarians, 25% of early voters voted for Gianforte. But when those who hadn’t voted were asked who they’d vote for on election day, only 7% of election day voters named Gianforte. That’s a huge difference — but also a small sample.

Nevertheless, let’s use that large estimate, and assume that Gianforte lost almost all of his Libertarian support as a result of his assault. Multiply 18% by 7% (the percentage of Libertarians) and by 40% (election day voters), and you get that Gianforte’s Libertarian vote loss might have been 0.5% of votes — not much smaller than his loss of Republican votes. This loss was likely split evenly between his two opponents.

What about turnout?

Turnout is tougher to judge with our data. However, we see no significant party-level changes in the percentages of people saying they would “definitely”, “probably”, “maybe”, or “probably not” vote. Both Democrats and Republicans said they’d show up to the polls about the same post-assault as pre-assault.

So what was the total effect of Gianforte’s assault?

Adding together the numbers above, our upper bound estimate is that the assault cost Gianforte between 1% and 1.5% of the vote. Quist likely received about two thirds of Gianforte’s lost votes.

Our upper bound estimate is that Gianforte’s margin decreased by about 2 to 2.5 points as a direct consequence of the assault. In this race, that wasn’t a game changer.

More likely, the effect was smaller — 1–1.5 points.

What if the assault had taken place earlier?

If the assault had taken place earlier, the effect we saw would have been multiplied across the entire electorate rather than just non-absentee voters. 2.5 times as many people would have been affected, so we would expect an effect 2.5x larger.

But even when we use the highest estimates, Rob Quist only gains about 4 points over Greg Gianforte.

Even if everyone could have taken Gianforte’s actions into account, the Republican still would have won the race.

How this was done

We concluded our initial survey of Montana Tuesday night (roughly 36 hours before polls opened). Then several hours after the story broke, we resumed surveying people in Montana.

When we resumed the survey, we decided not to ask about Gianforte’s assault directly. We did this both because we wanted to have a clean comparison between two groups, and because keeping it as is meant we could start getting survey responses more quickly.