She finished second behind Mehmet Oz in the last four polls of Pennsylvania. In three of those polls she trailed by just two points. In the two weeks prior to yesterday’s vote she nearly doubled her polling in the RCP average, zooming from 12.5 percent on May 3 to 24.2 percent yesterday.

As of this morning, with votes still being counted in PA, she’s out of the running in third place. The race is between Oz and Dave McCormick.

Wha’ happened?

As you can see, the RCP average nailed Barnette’s share of the vote almost exactly. But in this case the average is misleading since the last three polls each had her at 27 percent, above where she finished last night. There was every reason to believe that number was her floor as she gathered momentum down the stretch and that she might even end up cracking 30. As such, it looks like she underperformed on Election Day.

I think the totals this morning tell two stories:

1. Barnette’s enemies in the “Girondin” wing of MAGA succeeded in turning undecideds away from her in the race’s final days

2. Barnette may have nevertheless retained just enough momentum to blow this race for Oz and hand the nomination to McCormick

On the first point, the fact that Barnette surged so late left Oz’s supporters panicked that there wasn’t enough time left to blunt her surge. They scrambled to cut ads, but ads need time to penetrate the public’s consciousness and there wasn’t much time left. So Team Oz brought out the big guns and made a prudential case to undecideds that, however much they were drawn to Barnette’s populism, the cold fact was that she couldn’t win the general election. Trump himself made that point repeatedly in the past several days, as did Sean Hannity on his nightly Fox show.

How many undecided Republicans were preparing to vote Barnette a week ago only to be scared off by Trump’s and Hannity’s warnings about Barnette’s electability? Despite how late Barnette’s surge came, it appears that it *was* possible after all to derail her. It just took some very big megaphones to get it done.

Trump and Hannity obviously hoped that blowing up Barnette would send the undecideds who were leaning towards her into Oz’s camp instead. And that may have happened — to a degree. As you can see in the RCP table above, Oz overperformed his polling last night. But McCormick overperformed his polling *by a lot.* Why? Let me point you back to this post from two days ago theorizing about how Oz’s “Stop Barnette” kitchen-sink effort might end up inadvertently handing the race to McCormick:

Ironically, if McCormick does pull it out, I wonder if he’ll have his frenemy Trump to thank. Oz may be too much of a squish for MAGA fans to support even with Trump’s backing, but Trump’s attack on Barnette a few days ago, alleging that she has no chance in the general election, may have given pause to anti-Oz undecideds who are trying to choose between Barnette and McCormick. How many Republicans were tilting towards Barnette, heard Trump say that they’ll be forfeiting the seat to Democrats if they support her, and are now leaning McCormick as a result?

To put that another way, I suspect a large majority of the primary electorate took the “Anyone But Oz” position. No matter how much Trump and Hannity pushed for him, they were never going to back a guy whom they deemed a phony conservative attempting to parlay his television celebrity into an undeserved Senate seat. A week ago, many of those ABO votes may have been headed to Barnette. But after a week of Trump and Hannity calling her unelectable, accusing her of loving Obama, hinting that she was aligned with the Proud Boys, etc etc, the ABOers got spooked and shifted to McCormick in the name of maximizing the GOP’s chances of holding the seat this fall.

But not all of them. That brings us to the second point: Barnette may have done just well enough yesterday to deprive Oz of the votes he needed to overcome McCormick. In a test of strength between a Trump-endorsed candidate and a true populist, Barnette held her own:

Barnette may still have had an impact. Consider Montgomery County, where she and Oz both live, and where Barnette pulled about 35% of the vote, compared to his 28.5%. In the third largest county in the state, that’s a big gap, and a lot of votes Oz would surely liked to have had for his close run with McCormick.

If Barnette’s support had completely collapsed in Oz strongholds like Montgomery County, he’d probably have enough of a lead this morning to be able to declare victory. As it is, the pro-Barnette vote was sturdy enough to make the outcome a true toss-up.

1) We don't know if the late-arriving mail ballots (Montgomery got 1,051 yesterday) will be more Oz-friendly
2) The biggest stash of uncounted GOP mail ballots is 5,172 in Lancaster County, one of the very few counties where Oz is already running ahead of McCormick w/ mail

— Steve Kornacki (@SteveKornacki) May 18, 2022

On the other hand, there’s also an argument that Barnette’s late surge ended up wrecking McCormick in the end:

If McCormick loses, decent case to be made that bad polls cost him the election.

Question: Were there a few thousand PA Republicans who preferred McCormick but voted Oz because they thought it was a 2-way race between Barnette and Oz? pic.twitter.com/kNN76qgEMM

— John McCormack (@McCormackJohn) May 18, 2022

If it’s true that most of the electorate was “Anyone But Oz,” Barnette’s share of the vote came out of McCormick’s column more so than Oz’s.

We won’t know who won until Friday(!), as Allegheny County won’t report its late-arriving mail ballots until then. It’s possible that the final margin will be under a thousand votes. It’s conceivable that it’ll be under a hundred. All because the MAGA effort to destroy Barnette at the eleventh hour was mostly — but not quite totally — successful.

Exit question via Dave Weigel: If McCormick pulls it out, will Trump cry fraud? We’re neck-deep in irony here from the fact that the eventual nominee may owe his victory to mail-in ballots.

Counting mail-in ballots is good now. https://t.co/Lr3VzxcBDy

— David Weigel (@daveweigel) May 18, 2022