Uncertainties Loom as Midterms Enter Final Stretch
The 2018 midterms are working their way toward a conclusion. In a week, we should finally know who will control the House (although if the outcome is close, it could take weeks to finish counting ballots in California).
The consensus view is that Democrats are favored to take the lower chamber. Analysts disagree on just how large a majority they are likely to win, and how likely that majority is. If you split the RealClearPolitics tossups in half, it results in Democrats gaining about 25 seats, for a narrow 220-215 majority. I tend to think that the tossups will break disproportionately toward Democrats, and see something more on the order of a 225-210 Democratic majority, but this is hair-splitting to a certain degree.
At the same time, this isn’t the only way to read the data. The Democratic pickups could be larger, but they could also be substantially smaller. Here are some things that should bother any sober-minded elections analyst in the final week of the election:
Will there be a late break? As we learned in 2016, and in a less dramatic fashion in 2014, a late break in the races can alter the landscape substantially. This year, there are a lot of undecided voters remaining. Look at the most recent House polling from the New York Times/Siena: In some races, the leading candidates are at 43 percent (IA-03), 45 percent (UT-04, PA-10, NJ-03), and 44 percent (NY-11). Moreover, the trailing candidate is typically only behind by a point or two.
This adds up to a situation where a slight break in the undecideds toward one party or the other could be the difference between a healthy Democratic majority and a slim Republican one. My standard answer to how these races will break is to assume that, on average, they will converge on the president’s job approval, but there still is enough “wiggle room” in that political tendency to move things either way.
Given the closeness of many races, and the president’s improving overall job approval, it is also unclear how to evaluate this in the current cycle. In New York’s 22nd Congressional District, Siena College places the president’s job approval at 53 percent, so Republican Rep. Claudia Tenney is possibly in better shape than her polling numbers suggest. But what about the 19th District, where Rep. John Faso is up by a point, but the president’s job approval is at a 46/49 split? That suggests a very close race, and there are likely a lot of districts with splits like this one.
How many votes are Democrats wasting? One of the biggest news stories this cycle has been the massive influx of cash into Democratic campaigns. This initially showed up in marquee races, such as Beto O’Rourke’s challenge to Ted Cruz, but it has since filtered down into more marginal ones. Consider Indiana’s 9th District, where GOP incumbent Trey Hollingsworth has been outspent 2 to 1 by his challenger. Or California’s 1st, where Doug LaMalfa has been matched by his opponent. Both districts are heavily Republican and unlikely to flip. At the same time, the Democrats can probably use this cash to mobilize “low-hanging fruit” in these districts. In other words, Democrats may not win there, but will probably substantially overperform their typical showing.
This is relevant because Democrats right now have roughly a 7 1/2-point lead on the generic ballot, which asks which party people would prefer to have control Congress. Normally this would be enough to flip the majority. But if Democrats run up the score in districts that they are unlikely to win, suddenly that seven-point margin translates into a closer-than-expected seat share.
Ultimately, this illustrates a problem with using proportional representation metrics (how closely does the share of seats a party wins hew to vote share) to measure and predict first-past-the-post systems (where the candidate who receives the most votes wins a seat). While voters might have national feelings about parties, the actual campaigns are between individuals in local races. This is not to suggest that national factors or partisan identification are irrelevant – they very much are relevant, and indeed dominate such races. But factors such as fundraising, candidate quality, and third parties really do skew the local results, and this battle is still close enough that such a skew could alter the overall result.
How correlated are the errors? As noted above, RCP has about 30 races characterized as tossups, and if they split evenly Republicans would find themselves just shy of a majority. But is it reasonable to assume that these races will split evenly? Perhaps not. I’ve long assumed, for example, that out of the six most competitive House districts in California, a 5-1 split one way or the other is much more likely than a 3-3 tie. The districts are similar enough that if our underlying assumptions about how they behave are off – and they almost certainly are, even if we don’t quite know how – these races will move in one direction.
Are we relying too much on one pollster? This is a pretty straightforward concern. So far, most of our House race polling has come from the New York Times, in partnership with Siena. On the one hand, this is good, because without them we would have virtually no polling. On the other hand, polling is both an art and a science, and having a disproportionate amount of our data coming from a single view of what the electorate looks like increases the chances of No. 3 (above) occurring.
What is early voting telling us? The pat response is “nothing,” but that is too simple. Early voting returns usually tell us something, but unfortunately they can only be deciphered in retrospect. More importantly, because these returns can suggest multiple things about the electorate, interpretations of them are susceptible to bias confirmation, where we tend to accept suggestions that reinforce our previously held vision of what will happen. In 2016, many analysts convinced themselves that the early vote was good news for Democrats, when they probably should have been paying attention to the drop-off in African-American early voting participation and the fact that the Republican presidential nominee was actively encouraging his voters to avoid early voting.
This cycle is no exception. Early voting returns have not been particularly favorable to Democrats so far, and look older, whiter and more Republican than we might expect in a Democratic wave scenario. Analysis of early returns in Nevada – one of the only places I think analysis of early returns is sensible, since such a large portion of the electorate votes early – suggest a much closer Senate race than we should probably expect in a big Democratic year. From this, it is easy to conclude, as some have done, that the wave is not building.
This possibility alone should keep analysts awake at night, because it is a straight-faced suggestion that the conventional wisdom is badly off, as it was in 2016. At the same time, there is much we don’t know. For example, are stepped up Republican absentee ballot efforts in Nevada simply taking traditional Election Day voters and turning them into early voters? Or, since Democrats would probably rely upon infrequent voters to create a wave, are these voters likely to show up on Election Day rather than working through the early vote mechanism?
Perhaps most importantly, while early voting gives us insight into party identification, it does not reveal much about vote choice. If a large number of upper-middle-class whites who are registered Republicans flock to the polls to vote Democrat this time – the standard “suburban wave” narrative – we would see a surge in Republican participation that does not result in a surge in Republican vote share. A case in point: In the special election for Arizona’s 8th District, registered Republicans outnumbered Democrats by just over 20 points. The Republican won by five, which was viewed as a disastrous result for the GOP. As of today, the Republicans’ margin is only slightly larger, at 23 points.
Even this can be interpreted multiple ways – perhaps wavering Republicans re-registered as independents, or Democratic Election Day voters switched to early voting. This difficulty with reading the early voting tea leaves is exactly the point.
What else is lurking out there? This is probably the biggest concern. We’ve gotten good polling on the canonical competitive districts. But take argument two, and then realize that in some of these districts, Republicans might be caught napping. We’ve seen some examples of this, such as the open seat in FL-15 (Ross Spano vs. Kristen Carlson) or the incumbent, Rob Woodall, in Georgia’s 7th District. We don’t, however, know how big of a problem this is right now (or if it is a problem at all). I suspect we’ll have a surprise or two on Election Night, but what if there are 10 surprises? Regardless of whether that happens, we are well set up for such an eventuality.