What’s Behind Democrats’ Shrinking Senate Odds?

Republicans’ chances of keeping the Senate are up to about a 4 in 5 (79 percent), according to the “Classic” version of the FiveThirtyEight forecast. Republicans have always been favored to hold the Senate, but that’s nevertheless a meaningful improvement from recent weeks, when their odds were generally hovering between 2 in 3 (67 percent) and 7 in 10 (70 percent) instead.

FiveThirtyEight Senate forecast update for Oct. 10, 2018

This means Republicans have left what you might call the “Hillary Clinton zone” — the name I mentally assign to 70-ish percent chances where you’re only a normal-sized polling error away from losing the election. Less snarkily, Republicans have escaped a situation where Democrats could win the Senate merely by winning all the toss-up races. Instead, Democrats would have to win at least one of the four races — North Dakota, Tennessee, Texas and the Mississippi special election — that our model currently rates as “lean Republican” or “likely Republican.”¹ to have a mathematical shot at the Senate. So, while it’s crucial to remember that a 4 in 5 chance is a long way from 100 percent, the GOP is in a pretty good position.

Rather than get too philosophical, however, let’s pose a more specific question: Is the GOP’s improved position in the Senate because of just one or two races? Or a certain type of race? (For instance, races in red states?) Or is it more broad-based?

Here’s a comparison of the odds our model is showing for each competitive² Senate race as of now (Oct. 9), compared with where they stood on Sept. 12. Why Sept. 12? For one thing, that was the day we launched our Senate forecast. But it was also a relatively calm moment in the news cycle. There had been just enough time for some post-Labor Day polls (traditionally a time when voters become more attentive to campaigns) to filter into our polling averages. Brett Kavanaugh’s initial confirmation hearings had taken place, but Christine Blasey Ford had not yet had her name published by The Washington Post.

As you can see in the table, I’ve shown how the numbers have changed in both the Lite version of our forecast, which sticks to the polls as much as possible, as well as in the Classic version, which incorporates “fundamentals” (non-polling factors such as fundraising and historical trends that help predict the outcome).³ For some races, these tell different stories. In Tennessee, for instance, the Lite forecast shows Democrat Phil Bredesen’s chances waning, reflecting that Republican Marsha Blackburn has gained in polls. But the Classic version of our Tennessee forecast is essentially unchanged since last month because the shift toward Blackburn was expected based on the fundamentals since Tennessee is a very red state.

In the Lite version of our model, Democrats’ decline really boils down to a few states: Indiana, Missouri, Tennessee and especially North Dakota, where Democratic incumbent Heidi Heitkamp went from slightly behind in polls to way, way behind. Note that these are all very red states and that three of the four feature Democratic incumbents. (And Bredesen, as a former governor, might be thought of as a quasi-incumbent.)

Outside of these states, Democrats’ numbers have held up pretty well. In fact, they’ve improved in several more purplish states such as Ohio, Florida and Arizona. Furthermore, somewhat contrary to the conventional wisdom, Bob Menendez, the embattled Democratic incumbent in New Jersey, has seen his numbers improve in the polls, not worsen.

To summarize, the pattern in Senate polls has been as follows:

• In general, Democrats’ position has gotten worse in red states but has been steady-to-improving in purple and blue states. There are a couple of exceptions to this pattern (Joe Manchin has continued to poll well in West Virginia) but not many.
• On top of that, Democrats have had particular problems in North Dakota, where Heitkamp has seen her numbers go especially south. In fact, absent Heitkamp’s decline, Democrats’ overall position in the Lite forecast is about the same as it was a month ago. (The declines in states such as Indiana are offset by gains in states such as Florida.) Since North Dakota is rarely polled, that raises the question of whether Heitkamp’s position was this bad all along and didn’t have much to do with her decision to vote against Kavanaugh’s confirmation to the Supreme Court; notably, one of the polls showing her down double digits was conducted almost entirely before the hearing with Ford and Kavanaugh on Sept. 27.

The Classic version of our model adds more complexity to the story. Although the polls have moved in the direction the fundamentals predicted in some states, such as Florida, New Jersey and Tennessee, it’s having trouble with Democratic incumbents such as Heitkamp and Missouri’s Claire McCaskill in very red states. The basic heuristic behind the Classic version of our model is that it’s very rare for a party’s incumbents to lose when national trends favor it. So the Classic model has been expecting these Democratic incumbents to gain ground in the polls. Instead, with the exception of Manchin and arguably Montana’s Jon Tester, the Democratic incumbents have been losing ground.

Overall, this is one of those times where the “what” is easier than the “why.” The what is that Democrats’ position has worsened in the Senate as a result of declining numbers in deeply red states — where, because of their terrible Senate map, Democrats have a ton of exposure. But the why is not totally clear: It might be attributable to Kavanaugh, or it might have been baked in all along.