Princeton Election Consortium

A first draft of electoral history. Since 2004

Meta-Margins for control: House D+1.0% Senate R+4.2% Find key elections near you!

Electoral maps based on 2018 results

November 7th, 2018, 12:28pm by Sam Wang


(revised Friday November 9th to correct an error in Maine Senate)

The election turned out approximately as expected from advance information, a narrowly-Democratic House and a Republican Senate. I thought it might be good to look at the results from the perspective of 2020.

Above is what an electoral map based on the state-by-state House results would look like. It was calculated by averaging the district vote share percentage for each state and seeing which side gets over 50%. Uncontested races were treated as 75%-25% for the winner. (See the comment thread for further discussion.)

The outcome is 324 D, 214 R (popular margin estimated to be 7%, subject to change). Compare that with the 2012 outcome of Obama 332 electoral votes, Romney 206 (popular margin, 3.9%), or the 2008 outcome of Obama 365, McCain 173 (popular margin, 7.2%).

However, the House election is measured not by people but districts. In terms of districts, the outcome is quite close. This arises from a combination of population clustering and gerrymandering. Without gerrymandering, Democrats would have had about ten more seats, both before and after the elec

Here is what a Senate results-based map would look like:

In this case the outcome is 269 D, 93 R, 29 unresolved (Florida), 147 with no Senate race in 2018.

This may look a bit different from the media narrative that it was a victory for Republicans – which it was, certainly in terms of retaining control. But large voter populations in safe states – California, Pennsylvania, ad so on – contribute to this larger picture.

This idea was sparked by an exchange with E.J. Chichilnisky.

Tags: 2018 Election · House · Senate

11 Comments so far ↓

  • LondonYoung

    “This is done by averaging the district vote share for each state and seeing which side gets over 50%.”

    Beware of this methodology – because dem districts usually have fewer voters than GOP districts.

    By the way, note that this maps (all else equal) to an advantage for dems in winning house seats relative to their popular vote share. Of course, a major theme of this blog is that all else is not equal.

    • Sam Wang

      No, the relationship between popular vote and electoral vote is a nonlinear function. It’s approximately sigmoidal.

      Sure, one could do it with actual vote totals. Or normalize by district to 2016 turnout.

    • LondonYoung

      I think it flips Florida.

    • David Driscoll

      Yes, using actual House vote totals does flip Florida for the Republicans, at least according to the data here: https://floridaelectionwatch.gov/Downloads.

      The vote totals are 3661700 Rep, 3274981 Dem: https://docs.google.com/spreadsheets/d/1lqIPvhw8irHZeGKDNL4FVaFIxBJLtnXG9RIGle2ek7o/edit?usp=sharing

    • Sam Wang

      That’s interesting, thanks. However, your calculation also contains a bias: 5 districts with Democratic incumbents were uncontested, so their votes are not reflected in your approach.

      The difficulty with any hypothetical calculation is that it will contain assumptions that may or may not be appropriate.

      I am thinking that a better approach would be to normalize percentage results to actual turnout in the last several elections. So let’s do that. There’s still a wrinkle: FL-24 not contested either year.Have to make some assumption, let’s say it’s the same as the average of all the other districts.

      Under these assumptions the Democratic two-party vote share in Florida is 50.9%.

    • LondonYoung

      My suggestion what that we shouldn’t assume all CD’s have the average number of voters because heavily dem districts have fewer voters.

      CD 24 is heavily dem.

      What I suspect we want to do is use the number of voters in the contested 2016 Clinton/Trump election as the proxy.
      In that election CD 24 had 265k voters, while the average Florida district had 349k voters.

      So assume CD 24 has 75% (=265/349) of the Florida average.
      Repeat process for the other uncontested CD’s.
      This flips it back to red.

    • Sam Wang

      I do not get that result.

      When I normalize district-by-district 2018 Congressional vote by 2016 Presidential turnout, I get 50.6% D, 49.4% R.

      I’ve now done this calculation three ways – averaging district % share (52.4% D), normalizing by 2016 House turnout (50.9% D), and normalizing by 2016 Presidential turnout (50.6% D). You can inspect the three calculations here.

      I’m willing to agree that Florida is a closely contested state. But when it comes to Congressional sentiment, Florida is slightly Democratic-leaning.

      I suggest that we stop torturing the data, which did nothing to hurt anyone!

  • LondonYoung

    “This may look a bit different from the media narrative that it was a victory for Republicans”

    Remember this: There were 35 senate seats up yesterday, and the dems won them 23 to 12. Put that way, very hard to think of it as a gop victory!

  • Rachel Findley

    Unrelated question: is the number of closely contested races unusual? In US House races, gerrymandered cracking might generate a lot of close races in Republican-gerrymandered states, if a Democratic tide rises above the low lying cracked districts.
    In Senate races, and other statewide races, gerrymandering is not a factor. I’d ask: statistically, does the number of knife-edge races match what polling indicated? If not, is that just a fluke, or does it indicate that one or both parties might be adding or subtracting votes through voter suppression or during the counting process, trying to just push the election over the edge?
    Or perhaps someone can tell me it’s all normal, or in accordance with whatever statistical distribution we use for elections.

    • LondonYoung

      Some data:
      I looked at the wikipedia history of the 2014 house elections page. In 2014, one week after the election, 428/435 races were listed as called. Right now for 2018 it is 425/435.

  • Phil.

    Sam, I see that Michigan voted for an independent districting commission, when is that supposed to start, after the census?

Leave a Comment