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

Tea Leaves II: FantasySCOTUS renders a verdict

May 5th, 2018, 2:09am by Sam Wang

(Welcome to readers of SCOTUSblog! -Sam)

Postscript: Obviously, you can do what I did here for other cases. You are welcome to post your results in comments. Better yet, if you want to write a program to do this automatically, that would be quite welcome!

This week, Heather Gerken, election law scholar and Dean of the Yale Law School, visited to give a lecture here at Princeton on federalism as a powerful tool for both conservatives *and* progressives. During question-and-answer, she was asked what she thought would happen in the current partisan gerrymandering cases before the Supreme Court. After getting leave from me to answer (because of this, she is quite literally my lawyer!), she thought things looked uncertain. One of her sources was betting websites. As her unofficial data scientist, I thought I’d do a deeper dive.

I’ve previously used the crowd wisdom of these betting sites to estimate the outcome of the Arizona v. Arizona Independent Redistricting Commission redistricting case. There, the crowd did well – especially the most accurate punters. After doing the same thing for the current cases, it is possible to see what the smart crowd thinks will happen next month.

First, let’s admit that the crowd as a whole is divided on both Gill v. Whitford and Benisek v. Lamone. In Gill, AFFIRM upholds the Wisconsin plaintiffs, and in Benisek, REVERSE takes the side of the Maryland plaintiffs; these are the pro-reform outcomes. According to the crowd:

  • Good Judgment Open: 65% probability of good outcome (AFFIRM) in Gill
  • PredictIt: $0.43 for a $1 payoff on good outcome (AFFIRM) in Gill
  • FantasySCOTUS: 53% AFFIRM (good outcome) in Gill
  • FantasySCOTUS: 55% REVERSE (good outcome) in Benisek

This requires a bit of unpacking. At sites like PredictIt, bettors buy shares in an outcome. There, the share price is effectively an estimate of probability by the crowd. At the other sites, it’s a bit more like a survey of the people who log on to the site. For example, FantasySCOTUS, which is a project of LexPredict and Josh Blackman, is populated by enthusiasts who estimate the outcomes of cases, as well as the likely vote of each of the nine justices. Many of the people who vote are not legal scholars at all – but they do evaluate a wide variety of information, including oral argument.

At face value, all four of the above estimates are close to coin tosses. That would suggest considerable uncertainty.

However, not all members of the crowd are equally wise. It is possible to extract more value by identifying the top experts on FantasySCOTUS, and then seeing what they think. Of the 2016-2017 top ten punters, four are somewhere in the top 20 again this year: Tim Delaune, Bill Corteal, Justin Abbasi, and nbcrcc. This term, these top players have been right about case outcomes an average of 77% of the time. What do they think?

  • Gill: all four think AFFIRM.
  • Benisek: one says affirm, one says reverse, and two haven’t voted.

Now let’s look at this year’s top 10 (one of whom is again nbcrcc) who have made predictions. Their track record so far is similar to last year’s best forecasters. Their crowd wisdom is:

  • Gill: eight think AFFIRM, two think REVERSE.
  • Benisek: six think REVERSE, two think AFFIRM, two haven’t voted yet (Dana Brudvig and nbcrcc).

Combining the two groups, the aggregated estimate is 11 to 2 for a good outcome in Gill – pretty clear.

Benisek is murkier: 7 for REVERSE, three for AFFIRM, and three abstaining bidders. A bare majority, but not definitive. This reflects the apparent confusion at oral argument, though I think the two cases are likely to be coupled (i.e. if one goes in a plaintiff-favorable direction, the other one should as well). I suppose that is a two-way street.

Speculatively, a win in Gill might lead to one new Constitutional right: for members of a party to have a fair shot at representation at a statewide level. In this scenario, which FantasySCOTUS suggests is likely, statewide measures of gerrymandering would matter quite a bit (see my analysis of these measures at the Harvard Law Review Blog). A win in Benisek might lead to an additional right, the right of residents of a particular district to avoid being targeted for their views. Such a win is less certain at the moment.

I am interested in the electronic market from a practical standpoint. We at the Princeton Gerrymandering Project are forging ahead while this decision brews. No matter which way these cases go, there is much potential action at the state level in the coming year. That brings us back to the topic of Dean Gerken’s talk, which concerned federalism. The Princeton Gerrymandering Project has a two-pronged, federalist strategy: (1) helping federal courts establish guardrails on partisan gerrymandering, and (2) working with states to craft individualized, hack-resistant standards. I’ll write more about our efforts another day.

Of course we all have to wait until the Supreme Court rules – but my thoughts about how to allocate effort right now are shaped by what I think is likely to occur.

Tags: Redistricting · Supreme Court

19 Comments so far ↓

  • Jay Sekora

    Just wanted to say I very much appreciate your posts. At this point, now that everything that happens in public before the issuing of opinions is over, it may just be a pastime to speculate, but it’s a very interesting pastime, and reading what people smarter and better informed than I am have to say about it at least helps me avoid spinning my own wheels while I wait to find out what the court decides.

    (Also, I imagine anybody considering running for office in one of the districts that might be overturned or one of the districts that might be drawn to replace them would appreciate any early estimates of which way the decisions were likely to go they could get their hands on.)


  • David Anderson

    Under Miller v. Johnson, 515 U.S. 900 (1995) A district becomes an unconstitutional racial gerrymander if race was the “predominant” factor in the drawing of its lines. I do not think that it was proven that race was the “predominant” factor in the cases before SCOTUS and with decision in Alabama Legislative Black Caucus v. Alabama, No. 13-895, 575 U.S. ___, 135 S. Ct. 1257 (2015)that held racial gerrymandering claims proceed district-by-district, not against an entire plan my thoughts are that the gerrymandering cases will not have a big of impact as some think it will. Constitutional Rights should not be created out of thin air and states under the 10th Amendment should be given the deference to set the districts as they see fit… in fact it is sometimes the court drawn districts themselves that end up worse than those drawn by a party.

    • LondonYoung

      I presume you meant to say:
      “I do not think that it was proven that *party* was the ‘predominant’ factor …”

  • LondonYoung

    “(2) working with states to craft individualized, hack-resistant standards. I’ll write more about our efforts another day”

    This sounds interesting. I also wonder if anyone has released an open-source robo-districter that can be run on every state with no partisan input?

  • Pechmerle

    Another state level victory, in Ohio:

    Issue 1 — a constitutional amendment that sets up a fair, bi-partisan process for drawing new districts — overwhelmingly passed with 74% of the vote.

  • bks

    ‘When gerrymandering backfires’: Democrats go after once-safe GOP seats

    Republican redistricting maximized GOP gains in past years, but it could exacerbate the party’s losses in 2018.

    • Sam Wang

      I don’t think there is really such a thing as a backfiring gerrymander. They are levees that hold the tide for a while…but then they fail.

      Under a gerrymandering, for all plausible heights of electroal wave, the party in control does better or the same as under a fair map. Not worse.

    • Rachel Findley

      Sam, I thought bks had it right about gerrymandering backfiring in a wave election. The gerrymandering party (GP for short) creates a lot of districts in which they have a narrow margin of victory. In an electoral wave, the narrow margin flips and the GP loses all those districts. The other party wins all those gerrymandered districts, plus the districts where the other party was concentrated. Thus the GP could lose all the districts–which would generally be worse than a fair districting. Explain?

    • Sam Wang

      Rachel and bks, here is a sketch of how I look at it. The solid curves indicate a neutral districting scheme, and the dotted curves indicate a pro-Republican paritsan gerrymander:

      The top curve is a distribution of district partisanship, and the bottom curve is the integral – in other words, the relationship between statewide vote share and statewide seat share.

      In the bottom graph, the key point here is that for most levels of statewide Democratic vote, the dotted curve is below the solid curve. In other words, under most conditions, Republicans win more seats than in a neutral scheme.

      The exception to this is the case where Democrats get a very large share of the vote. In that situation, they get a similar seat share in both the neutral scheme and in the pro-Republican gerrymander. In other words, with enough Democratic votes, Democrats finally get the seat share that they deserved all along – but they do not get more.

      The only reason this looks like a “backfiring” gerrymander is that it requires the slope of the seats-votes curve to steepen. But that’s just Republicans postponing the day of reckoning until a massive wave year, as might happen this November.

      I think of a partisan gerrymander as being like a levee. It can protect against moderate swings in vote. But if there’s enough rain, then the levee overtops. When it does, the water all comes out at once. But it doesn’t rise higher than it would have without the levee.

  • Rachel Findley

    Slightly off this topic: I’m getting emails, smails and phone calls from organizations raising money for litigation regarding redistricting, and for state legislature elections to make the redistricting process less partisan. Is there any place where I could go to assess what these organizations are doing–how much of the money goes to fund fundraising? How much for redistricting concerns? How much for central office salaries and other overhead?

    • Sam Wang

      That sounds like the National Democratic Redistricting Committee. I don’t know how effective they are, or where their money goes. But they are legitimate.

      If you wanted to give to an orgamization with low overhead and high per-dollar value, you could give to the Princeton Gerrymandering Project. We should set up fundraising for that.

    • Pechmerle

      On a loosely related topic (bang for the buck), might PEC – as in past election years – give pointers as which House and Senate races would be the most efficienct ones in which to make campaign donations this year? Or suggest another site that is an accurate source of that information?

  • bks

    I was just forwarding the Politico article. I spent some time trying to construct a backfiring gerrymander without success, so barring inspiration (I haven’t had coffee yet this morning), I agree with Sam’s call.

    • LondonYoung

      Condorcet County is divided by a wide river. It elects two senators to the state senate. Half the population lives on the west bank and votes 70/30 for the GOP; the east bank votes 40/60 and thus favors the dems. For decades the river has been the districting border and Condorcet has sent one GOP and one dem to the state senate.

      In 2017 the GOP gerrymanders the county by dividing it between north and south creating two districts that favor the GOP 55/45 according to 2016 election results. But 2018 is a wave election year and there is a uniform 10 pct swing towards the dems, who take both seats 45/55.

      One GOP says to another “if we had kept our old districts we would have still won the west bank seat 60/40, our Gerrymander backfired”.

    • Sam Wang

      First…I note that your example has only two districts. Also, it goes from a statewide R+10% win to a D+10% win, a 20-point swing. That is out of the range that has occurred in the last two cycles (I think the biggest swing was 2002 to 2008). It would be more convincing to find a more realistic example.

      Second, the “backfire” claim is equivalent to the claim that a partisan gerrymander can create a seats-votes curve (i.e. cumulative histogram) that intersects the neutral curve. This is possible, but I think the intersection point is high enough that it doesn’t happen in real life. This way of thinking about it has the advantage that one can do it with continuous distributions, and explore the idea with generality.

    • LondonYoung

      Don’t get me wrong – bks’s statement sounded kinda like a challenge to construct such a thing, so I had a cup of coffee and wanted to make it clear that it was *possible*.

      I also couldn’t resist a chance to use “Condorcet County” as a shout-out to election nerds …

  • Larry in Trenton

    Prior to the oral argument in Benisek, Linda Greenhouse wrote an excellent column (all of her columns are excellent) suggesting reasons to believe that the Supreme Court would reverse both decisions.

    I think the Benisek oral argument undercuts that suggestion and that Professor Wang is correct: probably the two cases are coupled.

Leave a Comment