Cliff’s Note:
Blue Amp Media has no billionaire backing us. We have you. Pieces like this take a ton of research and arguing with numbers. They exist b/c paid subscribers make ‘em exist. If you want more—the map, the math, and the receipts, without the both-sides mush—become a paid subscriber.
It’s what keeps us independent enough to say things out loud before the consultants are ready to.
—Cliff
by Cliff Schecter and Lawrence Winnerman
John Cornyn was not going to lose a primary. Four terms, thirty years in Texas politics, the Republican establishment’s establishment. Ken Paxton beat him on May 26.
Bill Cassidy was not going to lose either. He finished third in Louisiana’s jungle primary and went home.
Steve Daines was not going to walk away from a seat he had already filed to defend, with $4.9 million banked. On March 4, at roughly 5:58 in the evening, two minutes before Montana’s filing deadline closed, he withdrew from his reelection campaign.
McConnell’s seat is open. So are Joni Ernst’s, Tommy Tuberville’s, Cynthia Lummis’, Markwayne Mullin’s (it feels like there should be two seats open for that name), and Thom Tillis’. Nine of the twenty-two Republican-held seats on this year’s ballot have no incumbent running.
Every one of those sentences was near impossible to imagine saying eighteen months ago. All of them are now facts.
The Wrong Math
A map is circulating that shows fifty-seven Democrats, three independents, and forty Republicans. Sixty votes—the number that breaks a filibuster under Senate Rule XXII, without anyone abolishing anything.
Watch how people dismiss it. They take the ten Republican seats the map needs, assign each one a 20 percent chance, multiply, and arrive at something like one in ten million. Then they go back to talking about the four seats that get you to fifty-one.
That is the wrong math, and it is not a close call.
Senate races are not ten independent coin flips. They are ten measurements of the same underlying thing, and when that thing moves, they move together. Nate Silver’s midterm model is built on exactly this premise: race outcomes, he writes, “are somewhat correlated,” and some years produce “systemic bias in the forecast rather than variance in individual races.” A 2018 study in the Journal of the American Statistical Association put a number on it—average polling error of roughly 3.5 points, twice what reported margins of error imply, and shared by every poll in a race rather than scattered among them.
Which means the distribution has a fat tail, and the fat tail is where sixty lives. If the national environment runs two points better than the polls say, it runs two points better in Kansas and South Carolina and Florida at the same time.
Nobody has to win ten separate lottery tickets. One number has to be wrong in one direction—and that number is wrong, in one direction or the other, in almost every cycle. Having begun my career in this crazy business at two different polling firms, I’ve seen this happen more times than you can imagine, and I’m seeing some things you won’t believe—and that even I am having trouble processing. But they sure are consistent in states as diverse as Ohio, Idaho, Iowa, and Texas.
