This week, I am at the Metric Geometry and Gerrymandering Group’s conference/workshop in Somerville, and, given my poor level of information retention from EC earlier this summer, I wanted to take a more deliberate approach to absorbing and processing the information that’s being given to me this week. I’m taking notes during the talks and I’ve decided to write these blog posts to reflect on each day of the conference, for both my benefit and yours, dear reader.
First, a little bit of background on gerrymandering and why it’s interesting to me. I took AP US Government my junior year of high school, which was the 2010-2011 academic year. For those who may have forgotten (or blocked it out), that was the election cycle in which the Tea Party mounted a major populist movement and swung a lot of legislative seats, both at the federal and state levels, to Republican control. This was in conjunction with an effort called REDMAP, which sought to gain control of state legislatures in advance of the congressional redistricting process following the 2010 Census in order to swing even more House of Representatives seats towards being “safe Republican” districts. Since then, gerrymandering, or the art, science, and magic of drawing political districts to favor or disfavor certain outcomes, has been fascinating to me as someone who works at the intersection of mathematics, computing, and social science.
This conference brings together people from all three domains, plus others, and I am having a ridiculous amount of fun at this thing. Like, to the point that I’m kind of in disbelief that I’m here for research purposes. I have met some of the most interesting people, as this conference has brought together a wide mix of people. I met local government members, physicists, computer scientists, mathematicians, software engineers, and activists, all interesting in this problem from their own perspective. It’s a lot of fun to be able to talk about how I think about the problem, how other people thing about the problem, and how to use tools and ideas from a range of disciplines to address gerrymandering.
I’ll give a short blurb about each talk. For more details, I’ll link my notes as well as videos if Tufts has made them available. If you have a problem with the notes, tell me and I’ll fix it. If you have a problem with the video, I’m not the person to complain to.
1. Talk 1: Moon Duchin on Situating Redistricting as a Problem in Multiple Disciplines.
Many people have thoughts and feelings about redistricting, and there are things to be learned from people with different backgrounds. From law, we need to think about how drawing district lines affect voting rights, from political science we can think about how districting affects representation and electoral outcomes, and from mathematics, we can think about the shapes of the districts. Link to Zach’s Notes | Link to Video of the Talk
2. Talk 2: Steve Ansolabehere on Partisan Gerrymandering
How do we detect gerrymandering? There are some simple tests like the distortion of representation relative to the population, efficiency metrics like wasted votes, and partisan bias, but none of these are smoking gun tools for detecting gerrymandering. We are also interested in how to actually make good maps. Given technology developments in recent years, crowdsourcing may be a powerful tool for developing good maps. Finally, we need to think about how to bridge the gap between scholarship and legal arguments. Judges are experts in interpreting the law, but not necessarily mathematics, sociology, or electoral politics, so we need to find the balance between tools and techniques which are comprehensive and rigorous, but also easy to explain and interpret in a courtroom setting. Link to Zach’s Notes | Link to Video of the Talk
3. Talk 3: Kristen Clarke on Voting Rights Litigation
Voting rights legislation does not exist in a vacuum. During the Civil Rights Era, legislation was passed to correct severe problems of disenfranchisement, particularly in the South. Recent events, such as the Shelby verdict in 2013 have dismantled some of these protections, and we need to be aware of how certain policies, such as voter ID laws, disproportionately affect minority populations. With respect to redistricting, it’s clear that there isn’t a cookie cutter approach that will magically solve every problem simultaneously. Any solution must involve careful and rigorous analysis of data, laws, voting patterns, and the motivations of those drawing the lines. Link to Zach’s Notes | Link to Video of the Talk
2. Talk 4: Moon Duchin and Max Engelstein Do Math
This ended up being more of a Q&A/brainstorming session, so the amount of “actual content” is relatively low. The takeaways are that the modern mathematical aspect of gerrymandering is relatively undeveloped, and techniques from group theory, geometry, topology, and graph theory may be useful. Link to Zach’s Notes
Day Two was just as much fun as Day One. I’ll say again, the best part of this thing is that I get to meet a broad range of people with extremely diverse interests and viewpoints. Part of the mission of this conference is to work with educators on how to effectively teach and communicate the issues and challenges of gerrymandering, because as any political scientist will tell you, awareness of and engagement with technical policy issues is extremly low among the general population, and getting people to know about and understand an issue is the first step towards widespread support for reform. Being able to think and talk about redistricting from the from the viewpoint of experts in complexity theory, middle school education, law, civil rights, mathematics, physics, statistics, urban planning, and software development has been really helpful for me to be able to frame the problem of gerrymandering in different ways, and I think that when I get back into research, I’ll have a much better idea of how to navigate my work because I can take into account the multifaceted nature of the problem.
I also think I’d like to put together “An Interdisciplinary Primer on Redistricting” both for the benefit of myself and my research, but also to be a quick and simple resource for people interested in moving around throughout the very broad space that this problem sits in.
As with the previous post, I’ll give a brief blurb for each talk, along with a link to my notes and the video of the talk (if available).
1. Talk 1: Ellen Katz on Race and Redistricting
Ellen is a legal scholar, and presented a modern history of court cases related to racial issues in voting rights and redistricting. This talk is a nice companion piece to Kristen Clarke’s talk from the previous day. She really got at the legal precedent surrounding some of the subtleties of why gerrymandering is such a sticky legal issue, such as how we can discern between a plan which unfairly packs minorities and one which is fairly providing majority-minority districts. Link to Zach’s Notes | Link to Video of the Talk
2. Talk 2: Megan Gall on Geographic Issues in Redistricting
Megan is a GIS expert and works with legal scholars to do data analysis and produce maps related to redistricting. While there are a number of criteria for redistricting, such as compactness and preservation f existing political borders, these are loosely defined, vague, and not always strictly applied. Additionally, every state has different rules regarding how state and federal districts are drawn, and sometimes even within a state, the rules for drawing congresisonal districts are very different than the ones guiding the drawing of state districts. As analysts, a major hurdle is getting good data. Since election and voter data only exists down to a certain resolution, we need to use statistical methods to make inferences about how different groups of voters behave, as a claim of disenfranchisement which violates Section 2 of the VRA requires layers to be able to argue that an affected group both votes as a bloc and is not afforded proper representation. Link to Zach’s Notes | Link to Video of the Talk
3. Talk 3 and 4: Wendy Cho and Yan Liu on Computational Issues
The morning talk was the “general audience” presentation and the afternoon was the more “technical” view of Cho and Liu’s project which implements a massively parallel genetic algorithm for sampling from the space of redistricting maps. In an ideal world, we could evaluate how good or bad a gerrymander is by using statistical tests to compare the map to a distribution of maps. For example, if someone is claiming that their map is not a partisan gerrymander, but it looks a lot like a draw from the distribution over maps with a strong partisan bias, we may have evidence of intentional gerrymandering. At a high level, the two major computational issues we have to deal with are that the actual space of maps that split a state into the correct number of contiguous districts is absolutely massive, and that picking the map which optimizes against some constraints and objectives is computationally very difficult. Here we design a genetic algorithm which runs in parallel on a supercomputer to try to get a good sample of maps. Genetic algorithms are nice because they run quickly and they don’t require the designer to know a lot about what the sample space actually looks like. The downside is that you can’t prove a lot of things about how good the solution you get is, or how long it will take to find a good solution. Link to Zach’s Notes | Link to Video of the Talk
Day Three was (unsurprisingly) excellent. I made a stronger effort to interact more with other people approaching gerrymandering from a computational and mathematical approach. One thing that has been difficult with my research thus far is that I haven’t been able to pin down a nice juicy research problem. Part of that is just my own lack of expertise and not feeling comfortable attacking a problem which has aspects (law, sociology, politics) that I am unequivocally not well-versed in. The other part has been that I haven’t been able to interact with other math/CS professionals who have spent time thinking about this problem in a research capacity. Getting to do that has helped me develop a whole menu of ideas, as well as some excellent contacts and potential research collaborators across the country.
As before, I’ll give a brief blurb for each talk, along with a link to my notes and the video of the talk (if available).
1. Talk 1: Guy-Uriel Charles on A Legal and Conceptual Primer for Redistricting
Guy is a law professor, and approached the problem of redistricting from a more judicial standpoint, in contrast to the civil rights-focus of Kristen’s talk on the first day and the historical lens of Ellen’s talk on the second. Guy layed out the constitutional framework under which redistricting cases are justiciable, and picked apart the judicial rationale for the opinions in Bandemer and Vieth, two important redistricting cases the Supreme Court heard. He also described some of the differences between the judicial approach to political and racial gerrymanders. Weighing in on a perceived political gerrymander could be seen as an infringement on a state’s right to structure the legislature and a clear case of judical overreach, while there are constitutionally-based arguments that racial gerrymanders violate guaranteed rights and federal law, like the VRA. Link to Zach’s Notes | Link to Video of the Talk
2. Talk 2: Robert Cheetham on DistrictBuilder
Robert is the CEO of Azavea, a GIS software engineering B-corporation in Philadelphia. His talk was a presentation of a tool called DistrictBuilder they built a few years ago. DistrictBuilder is free and open source (and available on Github somewhere…) and allows users to draw their own legislative maps while being able to see live statistics about their plan. The software was used for a few redistricting contests, notably one in Virginia where the contestants’ maps were used as part of a suit against the state for gerrymandering, and here in Philly where active citizens using the tool were able to get the legislature reengaged in drawing more representative districts. Link to Zach’s Notes | Link to Video of the Talk
3. Talk 3: Justin Solomon on Computational Geometry
Justin is a computer scientist at MIT EECS and he does things in the realm of computational geometry. Since redistricting has an inherent geometric component, the expertise of geometers may be useful in the problem. To begin with, most reasonable framings of the redistricting problem are at least NP-Complete. Additionally, some formulations of the problem look like problems for which there does not exist a good polynomial time approximation (unless $P=NP$). Therefore, we need to be a little more creative with our approach. Additionally, since we are working with highly numerical information in practice, we also have to worry about issues like measurement precision of the data and arithmetic underflow. There are a number of open problems in this area which I am excited to try to tackle. Link to Zach’s Notes | Link to Video of the Talk
3. Talk 4: Nestor Guillen and Justin Solomon on Optimal Transport
This talk was very mathematical and it went very quickly, so I’m not sure I’m in an excellent position to give a good summary, but here we go. Optimal transport is a topic in mathematics/statistics/physics/computer science which addresses problems of the flavor “If I have X cannonballs scattered on the ground and they need to be brought to Y different cannons, what is the most efficient way of moving them?” There are discrete and continuous versions of this problem, but for redistricting, we are interested in a model where we allocate parcels of land (a continuous thing) into a fixed number of districts (a finite and discrete number of things). It turns out this problem looks a lot like linear programming, which means that there is a nice correspondence between this approach and other optimization techniques. At this point, it is unclear how or if optimal transport will be useful in the redistricting problem, but according to Nestor and Justin (experts), it is certainly an area worth exploring. Link to Zach’s Notes
Share this post on social media:
Zach is a Postdoc at MIT IDSS. His research interests include game theory and mechanism design, machine learning, algorithmic fairness, and computing for the social sciences.