151 | Jordan Ellenberg on the Mathematics of Political Boundaries

Episode Summary

Executive Summary: Sean Carroll and Jordan Ellenberg discuss gerrymandering as a mathematical and geometric problem: how to judge whether political maps are unusually biased by comparing them to an ensemble of reasonably fair maps generated via random walks/Markov chains. The episode also connects this to random walk history, Markov chains, PageRank, and voting reforms like ranked-choice voting.

Main Topics: Gerrymandering as a coarse-graining problem (Priority: 5/5): Carroll frames representative democracy as a kind of coarse-graining, where many citizen preferences are compressed into a small number of elected officials, creating room for distortion when districts are drawn strategically. How gerrymandering works in practice (Priority: 5/5): Ellenberg explains how map drawers can pack opposing voters into a few districts and crack them across others, producing outsized seat advantages without needing bizarre-looking district shapes. Measuring unfairness with math (Priority: 5/5): The conversation focuses on how to assess whether a map is an outlier relative to plausible alternatives, including the efficiency gap and more robust ensemble-based approaches. Random walks and Markov chains (Priority: 5/5): Because the space of all district maps is too large to enumerate, the episode explains using random walks/Markov chains to sample representative maps and approximate what a neutral process would produce. History and intuition of random walks (Priority: 4/5): Ellenberg traces the development of random walk ideas through malaria research, finance, Brownian motion, and Markov’s work on language and probability to show how broad the concept is. Voting reform and alternative systems (Priority: 4/5): The discussion turns to ranked-choice voting and the possibility of redesigning electoral systems to reduce distortions, while acknowledging legal, political, and social constraints. Mathematics, law, and politics as a joint enterprise (Priority: 4/5): The episode emphasizes that no single discipline can solve gerrymandering alone; effective reform requires math, political science, law, and civic legitimacy together.

Key Arguments: Gerrymandering should be judged by how atypical a map is relative to a reasonable ensemble of maps, not by whether it merely has odd-looking shapes. The key metric is not a perfect definition of fairness but identifying extreme unfairness; one can often detect the worst abuses even if fairness itself is philosophically contested. The efficiency gap captures wasted votes, but it can misclassify some non-gerrymandered outcomes, so it is useful but not definitive. A random-walk approach makes the otherwise intractable space of district maps computationally manageable by exploring local changes repeatedly. The current U.S. districting system is especially vulnerable because the people drawing the maps are often the same people whose electoral fortunes are affected. A map that remains an outlier even after many local perturbations is strong evidence of intentional partisan manipulation. Alternative voting systems like ranked-choice voting may change both the incentives for gerrymandering and the tools used to detect it. Mathematical simulation can help social science and legal reasoning by estimating what outcomes would look like under neutral rules when experiments are impossible.

Data Points: Wisconsin assembly districts: 99 - Carroll cites Wisconsin as an example of a state divided into 99 assembly districts, with Senate districts defined as multiples of three assembly districts. Wisconsin wards: about 7,000 - Ellenberg notes that Wisconsin has roughly 7,000 wards, which are the units used in district construction. Republican share in Massachusetts: about 33% to 35% - Used to illustrate how geographic districting can amplify a statewide minority into zero congressional seats under first-past-the-post rules. Republicans from Massachusetts in Congress: none since 1996 - Example of representation distortion not necessarily caused by gerrymandering but by geographic clustering and winner-take-all districts. Libertarian congressional support: about 1% of voters - Ellenberg uses this to show how proportional representation would likely yield some libertarian seats, unlike the current system. Seven shuffles theorem: 7 shuffles - A theorem by Percy Diaconis and Dave Bayer mentioned as an analogy for how a small number of random moves can thoroughly randomize a deck. Markov chain walk duration: 10,000 local moves - Ellenberg describes using repeated local boundary changes to explore the space of district maps. Map sample size: 20,000 maps to 100,000 maps - He says it is easy to generate many thousands of plausible maps for ensemble analysis. Efficiency gap threshold example: 50.001% vs 49.99% - Illustrates how a party could engineer many narrow wins and one lopsided loss to maximize seats efficiently. Historical period of random walk development: 1900-1905 - Ellenberg situates key developments in random walks, Markov chains, Brownian motion, and early financial modeling around this period.

Pivotal Quotes: "the legislators are choosing their voters instead of the other way around" — Sean Carroll: A concise summary of the logic of gerrymandering. "The opposite of gerrymandering is not hewing to some standard that you think of in advance about what the number of seats should be. The opposite of gerrymandering is not gerrymandering." — Jordan Ellenberg: Explains why fairness should be judged against neutral map-drawing, not a pre-imposed seat target. "The geometry is the referee here, not the player." — Jordan Ellenberg: Describes the role of math in evaluating political maps rather than drawing them directly.

Implications: Listeners should see gerrymandering as a measurable statistical anomaly, not just a visual one. The episode suggests reforms and detection tools will increasingly rely on randomization, simulation, and cross-disciplinary methods.