There is no perfect case study of the effects of the Donald Trump presidency on down-ballot races. The special election in Kansas’s 4th Congressional District earlier this year was, so far, probably the purest, lacking much national attention for most of the contest and featuring two helpfully unremarkable candidates. The primary in Georgia’s 6th District pushed to a runoff, making that one less revealing.

None was less useful in evaluating the Trump Effect than the election in Montana held on Thursday.

The Republican, Greg Gianforte, beat Democrat Rob Quist by 6.5 points. This, of course, was the day after he allegedly grabbed a reporter and threw him to the ground in a fit of anger — and we say “allegedly” despite Gianforte’s post-election apology for doing so. He won in part thanks to the fact that the majority of votes were cast before Election Day, and he won in part thanks to the fact that his opponent was what one might safely call “untraditional.”

He also won the old-fashioned way: by having way more money spent on his behalf.

What role did national politics play in this? It’s very hard to say. But if we look at the shifts in voting in those three districts — and in a New York state assembly race earlier this month — a pattern definitely emerges.

We compared the final vote percentage for the Democrat and Republican in all four races with three other contests: The 2012 and 2016 presidential votes in those districts and the last time the same contest was held (in most cases, meaning the 2016 House elections). On the graph below, bars that point to the right show that a party improved its performance; a bar pointing to the left shows that the party slipped.

(Since the Georgia race was a primary, we didn’t include how the Republican runner-up fared, given that her party’s vote was split. Data on 2012 and 2016 results by district are from DailyKos.)

There’s a pattern there: The Democrats keep doing better, and the Republicans keep doing worse.

What’s important, though, is degree. Sure, Quist, the Democrat, did better in that race in Montana than did Hillary Clinton (by eight points), Barack Obama (by about two) and the last Democrat to run there (by three points). Quist’s loss was nine points smaller than the last Democrat to run in that race, a significant shift.

But on the scale of House races, it’s not that significant. In 2016, only 28 House seats were closer than the improvement the Democrats saw in Montana — but only nine of those seats were won by Republicans. Or, put another way, if we reran the 2016 House elections spotting every Democrat 9.2 percentage points, the Dems would only pick up nine seats.

The improvements in Georgia (considering how all Republicans did against the Democrat in the primary) and Kansas would yield much more significant gains for the Democrats. But, then, we’re comparing incumbent races in 2016 against non-incumbents in 2017, so it’s not exactly apples-to-apples.

(Lighter colored bars indicate uncontested races.)

There’s another complicating factor here. The Democrats have done better despite these being special elections — elections that generally see lower turnout and more voting by high-frequency voters, a group that tends to lean Republican.

In other words, there’s a lot of good news in the results of these elections so far for Democrats. But if you want to make a concrete determination of how much of that is a function of the national political environment, none of them is a foolproof example.