Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

OH-12 is ominous for GOP in House…and the Senate

August 7th, 2018, 11:36pm by Sam Wang

At the moment, in the special election in Ohio’s gerrymandered 12th district, Republican Troy Balderson leads Democrat Danny O’Connor by 0.9 percentage point (with 0.6% of the vote going to the Green Party candidate). This is a loss for Democrats, but it’s a 10-point swing from the Clinton-Trump margin in 2016. That’s very much in line with special elections all year, and it has implications for November.

In 46 special elections in 2018, the overall swing from 2016 has averaged 12 points toward Democrats. Hillary Clinton won the national popular vote by a little over 2 percentage points. If this swing were to hold up in November 2018, it would mean a 14-point win in the national House popular vote. I estimate that a 6-point win would be just enough to flip control. A 14-point win is massive, enough for a gain of over 50 seats.

Another piece of evidence points toward a building Democratic wave: Harry Enten of CNN has pointed out that Democratic pollsters are reporting their internal polls far more often than Republican pollsters. This observation corresponds to a minimum gain of 30 seats for Democrats. It’s hard to tell the ceiling because the D/R ratio has not been this high before.

As for the generic Congressional poll? Mostly ignore it. Even in the home stretch it misses some aspect of public opinion, probably because midterm turnout is hard to estimate. However, I will note one thing; it’s currently showing a lead of about 7.5 percentage points for Democrats – much less than implied by the two indicators above. If polls are underestimating election-day Democratic support, any discrepancy is important for close Senate races in Tennessee, Texas, and elsewhere. Recall that Senate polls in 2014 were off: in that case, Republicans overperformed against a Democratic incumbent president. Now the shoe is on the other foot.

What can you do this year? Whether you are a Democrat or a Republican, my advice is the same:

  • Near you, find a competitive House district and get out the vote there. Use the PEC-sponsored district finder!
  • Elsewhere in the country, donate to the most competitive Senate races.¬†I curate these¬†here.

And what if you’re a third-party voter, like the 0.6% who voted for the Green Party in OH-12? I’d say you lack situational awareness of what it means to live in a two-party system. Might be time to rethink that attitude.

Tags: 2018 Election

9 Comments so far ↓

  • LondonYoung

    So, here is another data-driven idea.

    If you look through a table of, say, senate polls on RealClearPolitics you will see a disproportionate number of “blue” leads.
    This makes me think that pollsters’ aren’t “centered” on the most competitive races, but are looking at those favoring the blue team.
    This suggests there are “dark horse” races where the blue team has a real shot but aren’t getting much attention.

  • Amitabh Lath

    Turnout is difficult (impossible?) to model and can change quickly (see 2016, pre/post Comey’s announcement).

    A lot of the “Dems are shoo-ins to take the house” narrative is based on tea leaves and wishful thinking. I don’t blame reporters, boring articles about power of incumbency and midterms skewing old and white probably don’t get many clicks. Democratic steamroller driven by enthusiastic young people is a story you can sell!

    But all it takes is one turnout depressing event. In other words, Republicans are just one early-November thunderstorm away from keeping the house.

    • LondonYoung

      The gambling sites estimate about a 2/3 chance of the dems taking the house.
      I don’t think the shoe-in stories are wishful thinking, they are just the “round to 0 or 100%” tendency that makes for the good stories you mention – and is also reflective of human nature.

      In the spring of 2010 Stu Rothenberg said the GOP had 0% chance of taking the house in the fall. After they did he said that 0% was an accurate estimate at the time.
      What does that even mean?

    • LondonYoung

      Sam’s has now put down his marker at 90% chance the dems take the house.

    • Sam Wang

      Well, I already said it other places. And conditions can change. I’m just saying where I’d put donations if I were in a donatin’ mood.

    • LondonYoung

      Yeah, so I get confused with: “conditions can change”.
      I kinda interpret the 90% as meaning that the dems are gonna take the house, but there is a 10% chance that conditions change enough to allow the GOP to hang on.
      But maybe it means “absent any big game-changing news, the current conditions will give the dems control 90% of the time”.
      As a geek, I kinda want 90% to mean: if you take a collection of all the times in my life that I said something was 90% likely and look back on them, you will find that 9 out of 10 times that thing went on to happen.
      My point about Rothenberg is that he clearly does not use probability in this last sense.

    • Sam Wang

      There’s sigma1, representing drift from now until November; and there’s also sigma2, a systematic uncertainty on our current variables (special elections, generic Congressional, and other indicators). I myself am trying to figure out which sigma is larger. I suspect sigma2 is pretty big, trying to get a sense of that these days.

    • LondonYoung

      That sounds right to me.
      And I think a lot of punditry obscures whether or not sigma1 is set to zero.
      Rothenberg seems to think sigma1 = 0 is implied.

  • Amitabh Lath

    Since we are all Bayesians now I worry a lot about my priors. I live in a liberal town, I work in a ridiculously left wing profession, I tend to read news from sources where anti-R dudgeon is maximized. So it is very easy for me to fall into believing the “blue wave is coming” narrative, so it’s worthwhile to examine some assumptions.

    1) Special elections are trending +10D. Yes, but they lack (by definition) an incumbent.

    2) Trump has antagonized women, minorities. Yes, but had done so before Nov 2016 and still won with women. And minorities are numerically significant mostly in safe-D districts (thank you gerrymanderning).

    3) Young people. Maybe. But it would be unusual for this cohort to make a difference in midterms.

    4) History says president’s party does poorly in midterms. There is a nice XKCD cartoon about political truisms that are true until they aren’t.

    I’m holding at 60% likelihood of a D house takeover. Turnout will be key, and turnout probability is a gaussian with a large sigma.

    It really doesn’t matter how well you know all your other variables, when you convolve a bunch of skinny gaussians with one fat gaussian, the result is an even fatter gaussian.

    PS: I am going against Sam’s advice and giving to challenger Andy Kim in NJ-03.

Leave a Comment