Rep. Trey Gowdy (S.C.) on Wednesday became the latest GOP member of the House to announce his decision not to run for reelection. He’s the 34th sitting Republican to do so, according to DailyKos’s tracker of open seats, and the 22nd whose decision to leave is not based on running for another office. (In Gowdy’s case, he’s aiming to enter the judiciary.)

As of this writing — and this seems to be changing with regularity — 17 percent of the seats with which the Republicans began the 115th Congress will have no incumbent in November. That’s the highest percentage since 2008, when 17.4 percent of GOP seats were abandoned before the election. But of course, the Republicans hold a lot more seats now than they did after the 2006 Democratic rout.

We were curious about whether there was a pattern to the retirements. So we took data on the 2016 results in each House district (again via DailyKos) and overlaid partisanship data from VoteView. (This score, called NOMINATE, ranks Congress members’ votes on a scale running liberal to conservative. A minus-1 is very liberal; plus-1 is very conservative.)

This is the result.

Anyone in the top half of the chart represents a district that Donald Trump won in 2016. Anyone in the right half of the chart tends to vote conservatively.

Here’s where we find Gowdy.

He is a conservative-voting representative who represents a district that backed Trump by a decent margin. He’s not, in other words, the sort of person you might assume was looking to head for the exit.

Whom do you expect to head for the exit? Well, people who represent districts that the other party’s presidential candidate won by a wide margin, for one.

Of 12 districts that are represented by Democrats and that Trump won, three members are retiring. On the Republican side, six of 23 districts won by Democratic presidential nominee Hillary Clinton are seeing retirements. That’s 25 percent and 26.1 percent of the seats, respectively.

Most of the retirees from both parties, though, represent districts won by their own party’s presidential candidates. (That’s in part because we included people backing out of House races to seek higher office.)

Seven percent of Democrats in districts Clinton won aren’t seeking reelection. Twelve percent of Republicans in districts Trump won aren’t. So Republicans in districts that voted for Clinton are twice as likely to be planning to retire — but Democrats in Trump districts are 3.5 times as likely to retire.

There’s a bigger gulf in the 2016 vote and partisanship of Democrats who are and are not retiring than between Republicans who are and aren’t retiring. In other words, the profile of a retiring Republican looks more like a nonretiring Republican than the profile of a retiring Democrat looks like a nonretiring one. Retirees in both parties are slightly more likely to be moderate, because they’re slightly more likely to represent districts that are less partisan in their favor. But those differences are smaller among members of the GOP.

There are a few possible ways to interpret this.

One is that Republicans are retiring for reasons unrelated to electoral politics. Gowdy, for instance, wants a different job.

Another is that 2016 results aren’t adequately capturing the threat to Republican candidates — meaning that maybe it isn’t only how Trump fared in a district that tells us how vulnerable an incumbent Republican is.