A Blue Wave is still more likely than not

Women’s March, Washington, 2017 (Amanda Voisard, Washington Post)

In every election year, Democrats grow despondent over the outcome as the election nears. Sometimes it’s just because the stakes are so high, as they were in 2016, and in that case they were of course right to worry. Sometimes it’s because things clearly will go badly, as they did in 2010. This year, it’s because hopes have been high, but the polls appear to be softening. After much anticipation of a Blue Wave sweeping the Republicans from Congress, now there is doubt. Some degree of doubt is a good thing; the Democrats will not win anything this year if people don’t turn out to vote for Democrats, and any Blue Wave was predicated on a huge turnout, compared to historical midterms. After 2016, I feel no great confidence in predictions or my own expectations, and have no desire to encourage overconfidence by Democrats, in part because it is not warranted, and in part because it is counterproductive. But there are still plenty of reasons to think that the midterms will repudiate Donald Trump in a way that 2016 should have, but did not.

The polls for Democrats over the last few weeks have not been bad, but also not as good as they had been or need to be. Recall the ground conditions:

1) Republicans have a 51–49 majority in the Senate. A tie is broken by vice president Mike Pence, so Democrats need to pick up a net of two seats to take the Senate and have a reasonable chance of blocking presidential appointments, like the possible replacement of Ruth Bader Ginsburg, Jeff Sessions, or Rod Rosenstein.

2) Experts estimate that Democrats must win the House popular vote by 5–7% in order to win a majority of seats, thanks to gerrymandering and unfavorable population distribution. (This, however, is the national popular vote, which the generic ballot is meant to measure. In head-to-head polls of individual districts, the target is of course just to get more votes.)

The easiest path to the Senate has long been clear: hold every Democratic seat up for reelection, and pick up Nevada and Arizona. Tennessee and Texas were thought to be put in play by the particular match-ups this year: Phil Bredesen versus Marsha Blackburn, and Beto O’Rourke versus Ted Cruz. Those two are now looking more like the reach we originally expected, and some of the Democratic holds are looking difficult, too, notably Heidi Heitkamp in North Dakota, while Jacky Rosen in Nevada and Kyrsten Sinema in Arizona are not exactly walking away with their races. (Conversely, Bill Nelson’s reelection in Florida had once looked worrisome and now looks better — helped, apparently, by the nomination of Andrew Gillum for governor, and by the fact that wealthy current governor Rick Scott, Nelson’s opponent, is no longer spending unchallenged.)

Based on trends that others have identified (but which you are free to dispute), there are two countervailing general forces in midterms. One is a tendency to vote against the president’s party in midterm elections, especially in the first term. The other is a tendency for Democratic constituencies to skip midterms entirely. Midterm turnout is always down from presidential elections, but it is down more among those who prefer Democrats — young people, the poor and working classes, and minorities. Democrats have seen midterm waves before, though, so there is no absolute effect.

There are at least nine empirical reasons to believe that a Blue Wave is still going to happen. These reasons are of different types: two reasons it would happen, one mechanism by which it may happen, two indicators that it may happen, three mechanisms that are also indicators, and one reason to doubt that it won’t happen.

Trump’s approval rating

The midterm backlash that seems to hit presidents regularly is counterintuitive; a president was popular enough to get elected, but is turned on by the electorate just two years later. But in this case, the president was highly unpopular when he was elected, and only managed to sneak in, despite losing the popular vote, through narrow wins in a few large states. It’s easy to forget just how many things had to go wrong for someone so unpopular to win even as narrowly as he did — Russian interference, massive journalistic malpractice, and the head of the FBI sabotaging Trump’s opponent at the last minute. Everything had to go wrong at once.

In other words, people already disliked Donald Trump. They didn’t want him to have power. They should be even more likely than usual to put a check on his power.

Gerrymandering backfire

Utah Congressional districts (National Atlas)

While gerrymandering is mostly expected to benefit the Republicans, it is important to remember how gerrymandering works. In order to win more seats with the same number of voters, a party must win fewer voters per seat. While some gerrymandering involves packing the opposing party’s voters into a few districts, sacrificing those to make other districts easier to win, sometimes that is not possible, and occasionally it is not even tried. For example, Utah has four Congressional districts. Rather than leave the Democratic stronghold of Salt Lake as a likely or even safe Democratic seat (as had been done in the past), Utah Republicans have divided Salt Lake among multiple districts. This has allowed them to win all four districts, but of necessity by smaller margins than if they sacrificed one to the Democrats. An aggressive gerrymander can backfire when the buffer of votes is overwhelmed by a surge for the opposing party. In the simple example of Utah, the gerrymander works if the GOP wins all four seats, it fails if the Democrats win one seat, and it backfires if the Democrats win two or more. The chances of backfire are greater in states where a larger number of aggressive gerrymanders is attempted.

Undecided voters

Many of the polls are still showing large numbers of undecided voters. These voters tend to break one way in a wave election (and even, to some extent, in a normal election). Polls that show a close margin but with undecideds greatly outnumbering that margin should be seen as likely Democratic wins in a year that favors the Democrats. And it’s even possible that a Republican who is ahead by a significant margin, but not above 50%, will lose thanks to a strong break of undecideds for the Democrats. All of this is assuming the polls are accurate in the first place. Polls are already estimating who is likely to turn out. A wave election challenges turnout assumptions.

Special elections

Since Trump’s election, Democrats have been massively overperforming in special elections, held to fill vacant offices. Democrats have only won a couple of the high-profile races — Doug Jones in Alabama, and Conor Lamb in the Pennsylvania 18th. But the won-loss record doesn’t tell much of the story; the federal offices, for example, were mostly vacant because conservatives were tapped for jobs in the Trump administration, leaving behind conservative districts that Republicans should easily have won. What Democrats have done in these elections is greatly exceed the partisanship of their districts. That may not have been enough to win all of the special races, but on November 6, a similar overperformance will win the Democrats more than enough seats.

Daniel Donner, DailyKos

An index of special election results by Daniel Donner, using two-year periods back to 1988 and including state legislative as well as Congressional special elections (but excluding the outlier of 1998), shows a strong correlation between special election results and the subsequent House popular vote. That index is as high for Democrats as it has ever been, and as of June 7 predicted a 15% win by the Democrats. That is of course no guarantee; but if the Democrats do not win by 8–10%, the result would be an outlier.

Off-year elections

Two states, New Jersey and Virginia, have already held full statewide elections under Trump, a year ago. Democrats won both governorships by large margins, greatly beating the poll average (Virginia) or effecting a party change (New Jersey); Democrats won the other statewide offices in Virginia; Democrats increased their popular vote share for both houses in the New Jersey legislature, and there was a huge swing in the popular vote for the Virginia House of Delegates, and a sizable popular majority (but narrowly missing a majority of seats, thanks to gerrymandering and voter distribution).

Fundraising

Good fundraising won’t win an election by itself, but it does give a competent and competitive candidate the resources to execute her strategy. So perhaps it is impossible for Heidi Heitkamp or Beto O’Rourke to win; but if they lose, it won’t be because they couldn’t afford ads or gas for get-out-the-vote efforts. But good fundraising through small donors is also an indicator of enthusiasm and commitment. Giving money is concrete and literally has a cost, whereas responding to or ignoring a phone call from a pollster has no cost. Democrats are having a phenomenal cycle in fundraising.

Open seats

Republicans are leaving Congress at a record pace for recent elections. Some of them are running for higher office, but most of them are retiring, because they anticipated losing, or (rarely) because they no longer cared to participate in the Trump era. The lack of an incumbent greatly increases the likelihood of a party switch, since incumbents tend to get reelected no matter the circumstances. But the decision of so many to retire is also an indication that, for whatever reason, they expected a difficult reelection. Members of Congress are generally seasoned politicians with access to special information (such as polling), so if Republicans are telling us themselves that they were endangered, they probably were.

Recruitment

Candidate quality is not always objective, but it can be, partially. Candidates with experience in elected office, government, or prior campaigns run better campaigns than those without. (Lack of experience can be a selling point to some voters, it should be acknowledged.) By objective (and subjective) measures, Democrats have had a very good recruitment cycle. Beyond candidate quality, Democrats have done very well on quantity: they’ve successfully recruited for a very large share of seats. Surprise wins are a feature of waves, taking victories in unlikely places. (This is part of the motivation of the fifty-state strategy.) But it is impossible to take advantage of surprising events if there is no opposing candidate. The Democrats have mostly avoided that problem this year. If the electorate turns in unexpected ways, or a Republican candidate falters in some way, the Democrats will have a candidate there to seize the opportunity, and generally a good candidate.

Poll errors

Polls are not completely reliable, of course. 2016 reminded us of this, though the error was not as dramatic as many remember. (Rather, the error was concentrated in states that had disproportionate effect on the Electoral College outcome.) The current New York Times/Siena experiment, of releasing poll results as they are making phone calls, should draw attention to the weaknesses of polls. Not all poll respondents are weighted the same; individual respondents stand in for particular demographics, and if a demographic is responding to the poll at a low rate (compared to the population), then each respondent from that demographic gets a bigger voice in the final result, often making a visible difference in the numbers. Polls are meant to be random, but that means they are random: a few calls to unrepresentative people and the sample is no longer good. With lots of polls, the error might average out. With only one or two, it’s a problem.

There is also the issue of differential response. Polls require that people be willing to talk to pollsters, and speak plainly about their views and intentions. Variations in poll numbers may simply result from changes in who is willing to talk, and the going theory is that people who feel good about their side at any moment are more likely to respond, thus boosting their side’s numbers. Overall, differential response should boost the poll results of the Democrats this cycle, since the news about Trump has mostly been bad. But the Kavanaugh confirmation may have Republicans feeling more confident at the moment. This would provide a poll boost to Republican candidates — but not be reflected in the election results. (Differential turnout is a related effect, but the point here is that polls may not reflect the electorate, because even committed voters may choose to respond or not to a poll based on momentary feelings.)

And finally, polls can be wrong because they misestimate who will turn out to vote. There are many reasons above to think Democrats have an enthusiasm advantage this year, and some indications (such as the special election and off-year results) to think this is translating into better Democratic turnout. If pollsters are predicting an electorate heavier on Republican demographics and weighting to that, they will put less weight on Democratic groups. So polls may have a representative sample of, say, Hispanics or young people, but if Hispanics or young people then turn out at higher rates than they have in the past, the final poll results will not have the right numbers, because they are not based on the right shares of each group in the electorate.*


If the Democrats put together a wave election by historical standards, history suggests they will win the Senate as well; incumbents of the wave party almost always win, especially in the Senate. If the Democratic incumbents all win, the Democrats need just two pick-ups.

And all of these contests are to some extent correlated. The candidates of a party rise together; if the Democratic incumbents are all winning, then Democrats are likely doing well enough to win Nevada and Arizona as well. Very bad candidates can cost the wave party a few seats, but candidates of merely-ordinary quality will surge. Big waves lead to improbable wins. A wave party ends up winning almost everything in play. That is a clear pattern. It does not need to repeat. But there are reasons other than individual horse-race polls to think it will. The 538 generic ballot average has the Democrats ahead, 50.3%-41.8%. That’s a very favorable national environment; Democrats have an absolute majority even with 8% of voters not in the mix. If any individual candidate is competitive, but currently behind, it’s this national environment that may put her over the top.

Needless to say, as a voter you should always think that it is your vote alone that will put your candidates over the top. That’s how we get democracy to work.

* Updated with this paragraph on November 2, after an excellent reminder from Peter Hamby. Stupid not to have mentioned it in the first place.