Breaking Down the Polls at the End

A voter fills out his ballot at a polling station in Brooklyn, New York, November 6, 2018. (Brendan McDermid/Reuters )
If this is worse than the usual midterm, the blame game will be on.

So, it comes to this. What will it come to, and what will it mean?

I looked earlier this month at the polling in the Senate and governor’s races, and revisited recently the Senate. Now, Election Day is upon us. The news is bad for a number of candidates in both parties. Continued Republican control of the Senate looks likely right now, but the party could suffer in the long term if it doesn’t hold on to a bunch of slim leads and capitalize on key opportunities. A lot of close governor’s races look likely to get away from the GOP. And the House is likely to go Democratic — likely, but not certain, and much as they were with the White House in 2016, the Democrats and their voters are wholly unprepared for the possibility that they might not take it. The last few days of polls have broken towards the Democrats, the opposite of what happened at the very end in 2016, and that should concern Republicans.

I’ve discussed my methodology for assessing races and laying out the charts below in earlier columns, and I won’t repeat those here. Henry Olsen has already laid out a lot of the “why.” Let’s get to business.

The Governor’s-Race Polls

Let’s start with the final RealClearPolitics poll averages, pictured in raw form above and fleshed out with additional details below. Quick notes on the following chart:

 The “DJT” column shows Trump’s approval rating according to Morning Consult’s September 2018 polling, which may not be entirely up to date; Trump’s approval rating nationally recovered to a hair over 44 percent in October, before finishing at 43.6 percent.

 “3” indicates the polling for third-party candidates; this is a big factor in Kansas (where 2014 Senate candidate Greg Orman is playing spoiler), Nevada (which has a long history of Libertarian candidates), and two New England states (Maine and Rhode Island) that have strong recent third-party traditions. Unfortunately, the latest Nevada polling ignores the third-party candidate, so we make do with the data we have.

 “Break” extrapolates what percentage of the remaining undecideds would need to support the Republican candidate to get to 50 percent. As I’ve noted before, this is an imprecise measure; some races end with both sides short of 50 percent, and as even the past month shows, candidates sometimes actually lose supporters (rather than simply failing to gather undecideds). But as a net measurement, this helps tell us how much room there is to grow left in the poll average, and how much of that room one side needs to capture to win.

 “Last poll” and “Post-10/1” look at the start date of the most recent poll in the race (most polls are conducted over multiple days) and how many of the polls in the average were begun since October 1.

Like the Senate, there are tiers of races here, though they’re not quite as clearly defined. Four races are complete tossups: Nevada (the one state that’s a true tossup in both the Senate and governor’s races), Kansas, Iowa, and South Dakota. Only one race that’s not a tossup still feels really close: Georgia, where there will be a runoff if neither candidate gets to 50 percent. Republicans will probably be happy to claim two of the four tossups at this point, and that’s bad, since all four are currently in Republican hands, and so is Georgia — South Dakota, in fact, hasn’t elected a Democratic governor since the 1970s.

Then we diverge by party. Republicans are fighting desperate rear-guard actions in four states where they currently control the governorship: Wisconsin, Florida, Ohio, and Maine. Wisconsin is the only one that doesn’t require a major polling error for Republicans to have a shot at a win, and frankly I would not consider Wisconsin a real race now if not for Scott Walker’s track record and the fact that the best pollster in the state is the one showing him tied.

Republican opportunities to pick off Democratic-held governorships are few, beyond Alaska. While Oregon is actually the closest in the polls, late momentum plus the severe unpopularity of departing incumbent Dan Malloy probably means Connecticut is the best dark-horse opportunity.

Besides the poll averages, we can also consider the trend line since September 15. The “Net” column is equivalent to the “Break” column, but looking back: It shows what percentage of the breaking undecideds have gone to the Republican since September 15, which in some cases is above 100 percent or below zero if a candidate has actually lost support. Some of these races aren’t close, so the late trends don’t matter much (partisans sometimes come home at the end of blowout races), and the percentage “Net” can be misleading in races where only a small slice of the electorate has actually moved.

If there’s ground for Republican optimism, despite a bitter round of polls the last few days, it’s that more of the Senate and governor’s races have broken in the Republican than the Democrat direction over the last six weeks of the campaign — the opposite of what you’d expect in a normal wave election (although we’ve seen waves before, notably 2010, when most of the wave was baked into the cake before mid September).

The Bottom Line: Expect Republicans to lose the governorship in seven to nine states they hold now, and gain no more than two. Overall, D+8 is the likely result. Anything worse than about D+3 is a bad result, so there’s no possibility of a good night for Republicans, but D+6 or less would be a “better than expected” result.

The Senate Polls

While a couple of Midwestern races have tightened at the end (Ohio, Michigan, Minnesota special), the polls will have to be well off the mark for any of them to escape the Democrats. Realistically, the odds are grim in Florida as well. North Dakota, Texas, and Tennessee are likewise safe for Republicans unless the facts on the ground go very differently than the polling. And overall, a number of races (Missouri, Arizona, Nevada, Florida) have gotten markedly less optimistic in the final days.

That leaves five races to decide control of the Senate, as well as Republicans’ margin of control. Those races are Nevada, Arizona, Missouri, Indiana, and Montana. Assuming all other races go as expected, Republicans need just one of the five to retain their current 51 seats, more than that to add. (Democrats, if they take the Senate, will under no circumstances have more than a one-seat majority — and that requires their winning one of North Dakota, Texas, and Tennessee.)

Arizona looks slightly better than Nevada, given the history of Nevada polls’ understating the Democratic ground game. Indiana is anyone’s guess. Missouri should have been in the bag by now, but President Trump stopping in Cape Girardeau on Monday night suggests Republican nervousness about dragging Josh Hawley over the line. (Recall that besides the national environment, Hawley had to overcome the disastrous scandal that claimed Missouri’s Republican governor, Eric Greitens.) Still, four out of five and a three-seat pickup is still not unrealistic.

When you look at momentum since mid September, you can see why there’s a lot more excitement around John James in Michigan than around Rick Scott in Florida, but the latter is still the closer race. Ohio also no longer looks like the blowout that Sherrod Brown has been banking on all year against Jim Renacci.

The Bottom Line: If Republicans are looking to run the Senate for the next two years, a D+1 or R+0 result is a narrow win, and anything from R+2 on up is good news — meaning less dependence on Lisa Murkowski and Susan Collins in future fights over nominees and any legislation that manages to make it to the floor in the next two years.

But with an outrageously favorable map, Republicans really need to make hay while the sun shines, because the map will be a lot less favorable in 2020 and 2022, and the senators elected this year will be in office for nearly the entirety of the presidential term that starts in January 2021. In other words, people casting a temporary protest vote against Trump and for divided government may be choosing a senator who will act as a force multiplier for President Kamala Harris for four years.

So, anything short of R+3 may have Republicans breathing a sigh of relief for now, but will give Democrats a leg up going forward.

Finally, remember that the Mississippi special Senate election will almost certainly go to a November 27 runoff, so this election is not over today.

The House
I have not followed the House-race polling as closely as the Senate and governor’s races; there are too many races and too little reliable polling. My own prediction is that Democrats will pick up about 32 to 35 seats, but we’re very much in the land of probabilities here: A bigger wipeout is possible, but it will not be surprising at all if the dust settles with the Democrats — or maybe the Republicans — clinging to a majority of just a few seats. Also, be prepared for the House to be in doubt for weeks if it’s close, given that a couple of the key races are in California, which counts ballots at the speed of an 18th-century ocean voyage.

All of the above assumes the polls are reasonably trustworthy this year despite systematically undercounting Trump-supporting white working-class voters in 2016 as well as other problems they’ve had intermittently over the past 16 years. Even if the polls are generally accurate, they are rarely precision tools, and for reasons discussed in my prior columns, the poll data may be shakier than usual this year. So don’t be shocked if there are a few surprises tonight.

Midterm elections are hard on parties in power, and House and governor retirements — some of them natural or term-limit-driven — meant that this would be a challenging cycle. But if this is worse than the usual midterm, and if the “blue wave” is enough to thwart Republican hopes in the Senate even in deep-red states, the blame game will be on.


The Latest