With the midterm elections now just three and a half weeks away, recent polling has favored the GOP in most close Senate races — and most forecasting models are coming into agreement:
Andrew Prokop is a senior politics correspondent at Vox, covering the White House, elections, and political scandals and investigations. He’s worked at Vox since the site’s launch in 2014, and before that, he worked as a research assistant at the New Yorker’s Washington, DC, bureau.
Five of the models we’re tracking at our forecast hub now give the GOP a narrow edge to retake the chamber, of between 51 percent and 67 percent. Though Nate Silver of FiveThirtyEight and Sam Wang of the Princeton Election Consortium have been publicly feuding over methodology lately, as of Friday morning their actual estimates were only 7 points apart. Wang’s estimate, which previously gave a sizable advantage to Democrats, has now moved much closer to where Silver has been all along.
Yet now there’s a new outlier in the forecast, in the other direction. The Washington Post’s model, created by political scientists John Sides, Ben Highton, and Eric McGhee gives the GOP an extremely high 95 percent chance of taking the chamber. The other forecasts all suggest Democrats should be worried. The Post’s forecast suggests they should lose all hope. Why the disagreement?
The GOP candidates lead recent polls in most key races
With Election Day so close, the fundamental factors used by both FiveThirtyEight and the Post don't matter very much anymore — both models are now based overwhelmingly on the polls. Accordingly, the table shows that the two models agree on who has the better chance of winning the seven key races above — the Republican. (There's one seeming exception, Kansas, that I've left out and will describe below.)
For a couple weeks now, GOP candidates have led in most new polls in most of the races they need to win the chamber. Those include races already thought to lean toward them (Arkansas, Louisiana, Kentucky, and Georgia) as well as several races where the Democrats had until recently led most polls (Alaska, Iowa, and Colorado). The significance of the poll shift in those latter three states is crucial. If the GOP wins two of those three, they'll win at least 51 Senate seats regardless of what happens in Kansas (though a shocker in South Dakota could still scramble their plans). And they're currently ahead in all three.
But how certain are those leads?
However, many of these GOP polling leads are quite small, or are based on just a few polls. So, how certain should we be that the candidate with a small lead in a few polls will actually win? That’s the heart of the disagreement here.
The Colorado race is a case in point. Throughout the summer and into early September, Democrat Mark Udall led in seven straight polls of the race. Since then, Cory Gardner led six out of seven polls taken. The FiveThirtyEight model views this change with caution, giving Gardner a 56 percent chance of winning — a very slight edge. But the Post’s model more confidently projects that Gardner has taken the lead, giving him a 78 percent chance of victory.
This greater confidence in narrow but consistent leads has been a feature of the Post’s model for some time, and it doesn’t advantage Republicans in every case. In North Carolina, Democrat Kay Hagan has maintained a small lead in the vast majority of polls. FiveThirtyEight gives her a 79 percent chance of winning, but the Post puts her chances all the way up at 97 percent. The problem for Democrats, though, is that there are more races Republicans now lead in.
Silver, though, maintains that it’s a mistake to project the outcome with extreme confidence when the margin is still pretty narrow. “Whether a 4-point lead translates into a 90 percent chance of victory or a 60 percent chance of victory is something we’ve spent a lot of time looking at,” he told me in September. “Some people that have crazily high odds in one state or another, I really want to kinda wager them.”
And then there’s Kansas
It may appear that the models vastly disagree on the Kansas race, with FiveThirtyEight giving Republican Pat Roberts a 40 percent chance of winning, and the Post giving the GOP a 97 percent chance of holding the seat. But that's because the models are calculating different things. FiveThirtyEight is looking at the chance independent Greg Orman will win, while the Post is calculating the chance the GOP ends up with the seat — whether through a Roberts victory or Orman choosing to caucus with them. FiveThirtyEight works the possibility that Orman will caucus with Republicans into its broader forecast, but not in its individual race estimate.
Kansas has been a bright spot for Democrats recently, as independent candidate Greg Orman has been leading Pat Roberts. But this week, two polls have come out showing Roberts back on top. If this trend holds up in other polls, tipping this race toward the GOP too, Democrats are in deep, deep trouble.
