It’s a tough climate but it ain’t this tough.

According to new polling from NBC News and Marist, just 34 percent of Wisconsin’s registered voters say Walker should win re-election in the fall, while 61 percent say a new person should be given the chance to lead the state…

If [state schools superintendent Tony] Evers faces Walker in the general election, the poll shows the Democrat would start the contest with a significant advantage. Evers leads Walker 54 percent to 41 percent in a hypothetical matchup among registered voters.

The only other survey testing Walker versus Evers thus far this year came from the respected Marquette poll, which had Walker leading by four last month. Just last week Marquette followed up with a new poll of Walker’s job approval: 47/45, not juggernaut material but not bad for a guy running for a third term in an environment favoring Democrats with Trump’s trade war started to pinch midwestern businesses like Wisconsin’s Harley-Davidson. Having a job approval that’s slightly net positive doesn’t mean you’re a favorite for reelection but it all but guarantees you’re not trailing by double digits unless your opponent is freakishly well-known and popular. And Tony “Who?” Evers isn’t freakishly well-known and popular.

Data nerds are also skeptical of NBC’s numbers. Nate Cohn of the NYT ran through a few of the usual caveats about higher Democratic enthusiasm this year and how tricky it can be to poll midterms accurately, but 13 points? Nuh uh:

But despite all of that, I have to wave a red flag on polls that don't weight for education and show a striking Dem lead in Wisconsin (or other white Midwestern states).
It could be right! It could be sampling error! But there's a serious, plausible mechanism for bias too

— Nate Cohn (@Nate_Cohn) July 26, 2018

Harry Enten of CNN is also skeptical:

It's both possible to think Scott Walker is in trouble for reelection in a way he hasn't been since first elected, while also thinking that he may not be down double digits.

— (((Harry Enten))) (@ForecasterEnten) July 26, 2018

Where did NBC and Marist go wrong in today’s data? The most compelling theory comes from Luke Thompson:

Michigan? No party reg. Minnesota? No party reg. Wisconsin?

Wait for it.

No party reg.

— Luke Thompson (@ltthompso) July 26, 2018

It's pretty obvious these dudes used basically the same party weighting schema for all three states. pic.twitter.com/tMfraCsyX6

— Luke Thompson (@ltthompso) July 26, 2018

Lacking good data for party ID, NBC guesstimated and apparently guesstimated badly. They have the likely electorate in Wisconsin this fall shaping up as 33D/25R/41I. In 2014, when Walker won reelection by nearly six points, the Wisconsin electorate was 36D/37R/27I. Read that again: There were actually more Republicans at the polls that year than Democrats, yet NBC has Republicans lagging badly this year. Now, that might happen. 2014 was a big red wave aimed at Obama, 2018 could be a big blue wave aimed at Trump. It may also be that a certain number of Wisconsinites who considered themselves Republicans four years ago have drifted into the independent column since then for whatever reason, possibly because they dislike Trump. (One of Nate Silver’s theories for why national Republican support for Trump is nearly unanimously in favor of him is that anti-Trump Republicans are simply exiting the party.) It’s a cinch that the 2018 electorate will be bluer than the 2014 one was. But a nearly one-third decline in the share of Republicans turning out? C’monnnnn.