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

Gerrymanders, Part 3: Redistricting Nerds, Geographic Compactness Does Not Solve Your Problems

May 14th, 2017, 4:27pm by Sam Wang

For general audiences, I have written quite a bit about gerrymandering in the New York Times (in 2013 and in 2015) and recently, in the Los Angeles Times. In addition, I have drilled into the technical details in the Stanford Law Review and here at the Princeton Election Consortium (part 1 and part 2). Today I want to discuss map-based reforms, such as the belief that requiring geographic “compactness” will solve the problem. Today I will explain why such an approach is flawed, potentially fatally.

A correspondent, Ira K., asks “what about a Constitutional amendment that requires that Congressional districts have boundaries that consisted only of state borders and at most n additional edges, where the additional edges were latitude and longitude lines,” where n is some reasonable limit?

This is just one example of a map-based reform. Another example would be a reform that focuses on geographic compactness, which reformers imagine will do away with gerrymandering. However, this is not true.

The short answer is that geometry- and boundary-based solutions to gerrymandering are (a) doomed to failure, (b) inconsistent with current law, and (c) in many ways undesirable. This is mostly described in my Stanford Law Review article, which I recommend to those who wish to get into the details.

To elaborate:

Doomed to failure: Despite the intuitive appeal of requiring simple boundaries, such boundaries can still favor one party over another. Well-known examples include the current Michigan Congressional map, the current North Carolina Congressional map, and even the original gerrymander of 1812 in Massachusetts. In all cases, boundaries are relatively straight and even follow county lines for the most part. In other words, the districts are largely geographically “compact.” Yet all three cases give one major party a strong and asymmetric advantage over the other party.

The reason for this is simple: it’s not really possible to limit the drawing of boundaries enough to constrain partisanship, yet still allow population equality to be maintained and communities of interest to be kept together.

Inconsistent with current law: The Supreme Court has clearly said that single-district maps can assume odd shapes. They have said a district should be “compact” — but they have also said that compactness can be metaphorical, for instance involving the the joining of related communities of interest that are separated in space. In addition, the Voting Rights act Actually requires the drawing of districts where minority communities have the ability to elect members that represent them fairly.

Undesirable: In many ways, it is undesirable to constrain the drawing boundaries in the way you describe. Putting an artificial condition on district shape, such as limiting the number of sides and vertices, also puts severe limits on what legislators can do. There’s an old political science saying that all districting is gerrymandering. Another way of saying this is that compromises have to be made in order to satisfy all the many constituencies that compose a community. Although we may not like the idea of legislators doing this, it is part of their job. It could also be done by a commission of people who will take the same trade-offs into account. But an automated algorithm would be very limiting.

Looking at this another way, I will merely point out that no legislature is likely to ever allow such a process to be automated. So in addition to being a method that will not work, it is hard to pass into law.


It is for these reasons that I have taken a statistical approach to diagnose gerrymanders. To make an analogy to medical testing, we do not need to recapitulate the biology of cancer in order to diagnose cancer; we merely need a test that detects it with high accuracy. It is the same with gerrymandering. We don’t need to draw maps to detect when an offense has occurred. Since the goal of a partisan redistricter is to pack votes in some districts and crack them in others, vote totals themselves provide a valuable guide for detecting when they have achieved their goal.

Interestingly, the fair drawing of districts could also be mandated using a statistical approach. The rule would be that one side’s winning districts should not be won by more lopsided majorities than the other side’s districts. That would pretty much solve the problem at a stroke.

Tags: Redistricting

11 Comments so far ↓

  • Ali Mckay

    How about we employ larger districts with multiple (3-4) winners and ranked choice voting? Another solution is proportional representation.

    • Sam Wang

      A three-member district would be worse, since it would usually split 2-1, which is not an improvement at representing the minority party than a single-district system. One would have to get up to five members to come close. Even so, there are court precedents that get in the way of such an approach.

      For other proposed reforms, the general difficulty is they deviate too far from current practice to be accepted by legislators.

  • Amitabh Lath

    What about maps weighted for population or income or ?

    I like density, because basically we have two tribes, one likes to clump up in high density urban areas and the other likes to spread out in rural ones.

    I am troubled by the idea of using voting patterns to choose selection criteria that is meant to affect voting patterns. It’s the moral equivalent of looking at data as you design your study, you are bound to get the result you want but difficult to say you were totally unbiased.

    • ColinMcAuliffe

      Redistricting seems inherently unfair to me since no matter what approach you take you’re bound to draw a certain number of people into districts where they have little chance of influencing the election outcome. I also think there is no way to have any redistricting process that is truly unbiased, since the effects of any redistricting criteria on a map could be more or less known in advance. For example any party blind process will tend to disadvantage democrats due to population clustering. Therefore it makes sense to use all available data to get maps that are competitive and partisan symmetric.

    • Amitabh Lath

      I agree that redistricting is unfair, and there are statistical tests one could do to see how unfair. And certainly in an equilibrium situation one can shoot for ratios of representatives that mirror the overall state’s voter ratio.

      But what happens to these metrics if party loyalties are changing? The South went from solid Democratic to solid Republican over a few decades. These transition periods give me pause.

    • Sean Patrick Santos

      Baker v. Carr is perhaps the most important redistricting case in SCOTUS’s history, and it involves a case where Tennessee simply refused to redistrict for decades, which discriminated against cities that had grown up within that time. Refusing to redistrict is inherently unfair!

      There’s never been an obviously right (by which I mean widely accepted and automatable) way of drawing districts, so we do have to deal with the fact that redistricting should be done,and somehow humans have to do it.

  • Lucy Berlin

    I would like to see links to your project on other sites that are also working to explain and reduce gerrymandering.

    For example EndGerrymandering,com, a collaborative project of the Brennan Center for Justice, FairVote, etc. A few of its links include”All About Redistricting” Professor Justin Levitt’s guide to drawing electoral lines, and to

    Shouldn’t sites like those include among their resources your guide to diagnosing gerrymanders — and your solution? Do they even know about you?

    And, separately, could you at some point write a “part x” with a table that categorizes and analyzes the approaches proposed in such sites? These include the multi-member districts, ranked choice voting, compact districts, etc., mentioned in comments on this and previous parts. It would be great to collect in one place your opinions of the + and flaws with each approach.

    • Sam Wang

      I will avoid citing specific reforms on map-drawing procedures until I understand how — and whether — they address partisan asymmetry. Any set of rules may carry hidden biases. To my knowledge, reformers have not systematically analyzed this problem. The reform movement could really use a rigorous analysis of how their proposals could be gamed by bad actors, now and in the future.

      For example, multi-member districts can easily make the partisan-asymmetry problem worse – see my other reply. Multi-member districts can also run afoul of existing election law, including court rulings.

      This blog post is already focused on the subject of compactness, which is an example of a map-based reform.

      Ranked-choice voting addresses problems that arise during the primary. However, if you want a legislator who avoids representing the extremes, RCV only accomplishes this in competitive districts. Partisan gerrymandering prevents RCV from ever becoming useful: a packed-and-cracked map is still filled with lopsided partisan districts, no matter how the representatives are chosen.

      Yes, I am in communication with other organizations, including the ones you list. Our resources, which can be found at, are under construction. By the end of the summer we plan to make the tools easier to use and more desirable to link.

  • Jay Vaidya

    I am looking forward to your analysis on the Supreme Court decision on the North Carolina redistricting case.

    The majority did not comment on the merits of allowing partisan redistricting, because they found evidence of race-based redistricting, which was enough to strike down the map.

    I noted that the dissenting opinion by Justice Alito affirms support for partisan redistricting, with no conditions. Justice Kennedy joined that opinion.

    Justice Kennedy was the key vote for your project on using objective standards to identify and strike down egregious gerrymandering. Does Justice Kennedy’s joining the Alito dissent signal that he will not even accept objective standards now?

  • LondonYoung

    On multi-member districts:
    State governments have found that clumping all their allocation of votes together is beneficial to the influence of the state. California has 55 electoral votes – when Hillary wins 62% of the vote there, she gets all 55 votes – there is no representation in the electoral college for the losing side.
    States historically thought the same thing about their congressional delegations. Before the civil war, in Sam’s home state of New Jersey (and so many others) voters chose between statewide slates comprised of just one party for congress.
    However, voters didn’t like this because they want a single representative to represent their interests at as fine a level as possible.
    Agitation for such got the congress to agree, in 1872, to pass a law requiring single representative districts – no more slates.
    Going to multi-member districts would reopen the issue.
    However, this does remind us that minimizing Gerrymandering need not be done just through the courts since congressional action would be good enough.
    In fact, we probably only need the House to decide the rules – no need for Senate and Prez:
    “Each House shall be the Judge of the Elections, Returns and Qualifications of its own Members”

  • run75441

    Mike Bishop in the mostly Republican 8th Congressional District in Michigan represents >700,000 constituents. He has yet to show face in a Town Hall meeting and will only do phone ins or talk to a select few.

    That he will not appear in a Town Hall meeting is a bone of contention with many in his district. They have done cardboard cutouts of him at meetings. It is doubtful too, he could represent 700,000 constituents and know what their views are.

    The House was never supposed to get past 60,000 constituents per Congressional Rep. I would think if you reduced the number of constituents per Representative, you could very much reduce the chance of gerrymandering and even so the party percentages would be closer in each district.

    California is vastly under represented if it were to be compared to Wyoming and its 586,000 constituents. If we used 586,000 constituents per district, California would pick up 12 more districts and representatives.

    In today’s age, does The House have to be in Congress or could it webex in on a secured link? I think reducing the size of Congressional districts would go a long way towards killing gerrymandering. My $.02.

Leave a Comment