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

Pennsylvania’s (current) 18th CD: The top of the levee

March 13th, 2018, 9:08pm by Sam Wang

Whoever wins today’s special election to fill the open Congressional seat in Pennsylvania’s 18th District, it wasn’t supposed to be close. It’s part of a statewide partisan gerrymander, a district that a generic Republican ought to win by about 20 points. In November 2016, Donald Trump got 58% to Hillary Clinton’s 39%. The closeness of the race suggests there’s a massive wave of discontent with the Republicans.

If Conor Lamb (D) wins, that doesn’t mean that gerrymanders don’t matter. Instead, it means that even the highest levee can be breached if the wave is high enough. This year’s potential Democratic wave is evidently very high indeed. If this race is near-tied in a district that Donald Trump carried by 19 points in 2016, that is a massive swing – consistent with other special elections in 2018, which have swung by a median of 23 points toward Democrats.

Follow the NYT predictive tracker here, and DailyKos Elections here.

Tags: 2018 Election

15 Comments so far ↓

  • James

    Is it possible for gerrymandering to backfire and actuality lead the party in power with fewer seats? For instance, what if Republicans create a bunch of districts that are supposed to be R+10, and Democrats outperform by 15 points, capturing most of the red districts in addition to the packed blue ones?

    • Sam Wang

      Yes, but that’s an example of the levee being breached. That’s my point.

    • Pechmerle

      Brennan Center is just out with a detailed study, with special emphasis on swing states, suggesting that to “breach the levee” nationally in 2018 the Dems would need a wave of a magnitude not seen in the past several decades. They argue that the 2010 Repub. gerrymanders were crafted well enough to be highly resistant to change in the state-wide distribution of Congressional seats. On their numbers, it often takes an extraordinary shift in the generic Congressional ballot – even into the 20 percent range – for the Dems to take one more seat than what they were getting in the particular state in 2012, -14, -16.

      Depressing reading.

    • Sam Wang

      Probably overstated. I think the national win has to be about 8 percentage points.

  • Sean Santos

    It seems like absentee ballots are likely to decide this election. I’m curious if anyone knows how often that happens? And if there’s a systematic bias (i.e. a net plus for D or R) in absentee ballots relative to the district or state they are cast for?

    My belief has been (and still is) that absentee ballots are pretty negligible compared to the margin of error of the polls in pretty much any race (e.g. in this race absentee ballots overall will be about 3% of the total I think?), but if there’s any hard data in favor of or against that general idea I’d be curious.

    • 538 Refugee

      Accounts I read last night said the outstanding ballots would have to break ‘bigly’ for the Republican candidate. Most outlets had the remaining count around 3200 except for Fox who had it at 3900. I think Lamb and the Democrats claiming victory was an indication they didn’t think there would be enough to change the outcome.

    • Sean Santos

      @538 Refugee

      My point wasn’t that the *remaining* ballots would actually *swing* the outcome, but that the margin was thin enough that the absentee ballots made a significant difference. I don’t know where to find data on absentee vs. non-absentee votes officially (if the state even has that on a public website somewhere?), but my recollection is that at one point the reports were that all the non-absentee ballots were in except for in two precincts, but none of the absentee ballots were counted yet. At that point the margin was ridiculously tight: by the AP’s count, Lamb was up by 95 votes, but the remaining two precincts were in heavily Republican-leaning Westmoreland County.

      If that’s right, and we take this tweet at face value, the non-absentee vote would have given Saccone the win by a razor-thin 109 votes:

      However, the absentee votes (some of which were reported before the last non-absentee votes) were disproportionately from Allegheny County, and those boosted Lamb enough that the absentee votes from the other counties seem unlikely to change the outcome. In other words, the non-absentee vote seem to have basically tied, and the absentee voters broke the tie in favor of Lamb.

      My question is whether that should matter to how we think about polls. My feeling is that it probably shouldn’t in most cases in the contemporary US, because:
      (a) There aren’t a lot of absentee voters.
      (b) It seems like absentee voters don’t deviate *that* much from non-absentee voters.
      (c) Presumably at least some absentee voters are included in at least some polls (e.g. because they were physically in the district at the time of the poll, but not while voting, or because they get called for a phone survey when not physically in the district).
      (d) Polls’ collective margin of error, even when aggregated well, is way bigger than any absentee effect.

      Given all this, the Pennsylvania 18th race seems like kind of a fluke, where the election happens to have been decided by the fact that absentee voters skewed a bit more Democratic in a ridiculously tight race.

      But it’s not clear that the above points will always be true, especially if you look internationally. One can imagine a region where a sizable portion of the population are absent from their primary residence for a significant part of the year (active military, seasonal workers, college students, truckers, upper-class people who have multiple homes or take frequent vacations, emigrants who still vote in their country/region of origin, etc.). So I’m curious if there are circumstances in which pollsters have had to concern themselves with absentee voters who voted in considerably different ways from non-absentees.

  • bks

    Lamb underperformed the polling. Only bad thing about the result.

  • Pechmerle

    Two more pieces of good news here:

    — Trump personally showed up and supported Saccone; that didn’t help him enough (or did it hurt? exit polls would be interesting).

    — Outside R sources spent $14 million on this race, to outside Dem sources $2 million. Couldn’t buy the result. Although Lamb directly raised $3.9 million to Saccone’s $900,00. Also, it appears that Saccone pushed positive messages, while all the outside R money was negative messages. leaving some contradictory impressions to the voters

  • Amitabh Lath

    Gerrymandered districts are almost by definition unsafe. A “perfect gerrymander” would give you large number of 51% districts and one large 99% district to your opponent. To keep with Sam’s levee breach metaphor, the R retirements we are seeing are folks abandoning homes in floodplains.

    At this point we should be arguing about the size and geographic variability of the coming wave. What metrics do you trust?

    • 538 Refugee

      What is the measure of discontent? Will it last. Will the tax cut buy votes by the mid-terms? What will Mueller’s influence be? As I remember it seems Nixon had a huge number of hard core supporters turn on him very quickly as that whole ordeal evolved. In the end they realized they had been lied to and became very vocal and very bitter. Trump is jumping down that hole head first without a parachute.

      Personally I’d like to believe that people are horrified by the erosion of checks and balances. Perhaps they remember the last time Republicans had the ‘one ring’ and see things going bad fast again? Actually, that sounds like quite the campaign strategy.

    • 538 Refugee

      “What is the measure of discontent?”

      I found this in answer to my own question. Read what this man thinks now!

      “Peters was briefly suspended in 2015 for calling President Obama a “total xxxxxx” while on the Fox Business Network.

      Even at Fox the levee has sprung a leak.

Leave a Reply to James (Cancel)