The Heist Nobody Noticed
In 2003, Texas Congressman Tom DeLay did something that hadn't been done in modern American history: he redrew the state's congressional districts in the middle of a decade. The census hadn't changed. The voters hadn't moved. But when the dust settled, what had been a 17–15 Democratic edge in the state's congressional delegation became a 21–11 Republican landslide.1
To put that in perspective: roughly the same six-and-a-half million Texans who voted in 2002 voted again in 2004. Their preferences hadn't radically shifted. What shifted were the lines on the map — the invisible borders separating Congressional District 23 from Congressional District 25, deciding which votes "counted" together. Change those lines, and you change who wins. Same voters, different outcome. It's not a bug in democracy. It's a theorem.
In Chapter 7, we met Arrow's Impossibility Theorem — the devastating result that no ranked voting system can satisfy a modest handful of fairness axioms simultaneously. We learned that democracy isn't just hard to get right; it's mathematically impossible to get right, at least if "right" means satisfying every reasonable person's idea of fairness at once. But Arrow's theorem was about the voting rule — the algorithm that converts ballots into outcomes. The Texas redistricting fiasco shows us something even more unsettling: even if you fix the voting rule, you can still rig the game by manipulating which voters get grouped together.
This chapter is about the deeper math of democratic manipulation — the tricks that come after Arrow, the ones that operate on the geography, the incentives, and the strategic behavior of the voters themselves.
The Geometry of Stealing Elections
Gerrymandering — named after Massachusetts Governor Elbridge Gerry, whose 1812 redistricting plan produced a district shaped like a salamander — comes in two flavors, and both are elegant in their awfulness.2
Packing: Cram your opponent's voters into a few districts where they win by absurd margins — 85%, 90%. Every vote beyond 50%+1 is "wasted." You sacrifice those districts but drain their voters from everywhere else.
Cracking: Spread your opponent's remaining voters thinly across many districts so they never reach a majority anywhere. You win each of those districts by a comfortable-but-not-wasteful margin.
The combination is devastating. Pack a few, crack the rest, and you can turn a 55–45 popular vote into a 70–30 seat advantage. The voters haven't changed their minds. The map-drawer has simply sorted them.
The same population, three different maps, three different outcomes. Geometry is power.
The Efficiency Gap
Political scientists Nicholas Stephanopoulos and Eric McGhee proposed an elegant metric in 2014 for measuring gerrymandering: the efficiency gap.3 The idea is to count "wasted votes" — votes that don't contribute to electing anyone. For the winning party in a district, every vote beyond the 50%+1 needed to win is wasted. For the losing party, every single vote is wasted.
A large efficiency gap means one party is systematically wasting more votes — the signature of gerrymandering.
In a fair map, both parties waste roughly the same number of votes, and the efficiency gap hovers near zero. In a gerrymandered map, one side's wasted votes far exceed the other's. The metric was cited in Gill v. Whitford, a Wisconsin gerrymandering case that reached the Supreme Court in 2018.4
Try It Yourself: The Gerrymander Lab
Below is a 6×6 grid of voters — blue and red. Your job: paint district boundaries (five districts of roughly equal size) and watch how the seat allocation changes. Same voters, different outcomes. That's the whole trick.
🗺️ Gerrymander Lab
Click a district color, then click cells to assign them. Try to make Blue win all 4 districts, then try to make Red win — from the same voters.
Detecting the Crime: Random Maps
How do courts tell a gerrymandered map from one that's merely quirky? The breakthrough idea, pioneered by mathematicians like Moon Duchin at Tufts University, is beautifully simple: generate thousands of alternative maps and see where the real one falls.5
The technique uses Markov chain Monte Carlo (MCMC) sampling. Start with any valid redistricting plan. Make a small random change — swap a precinct from one district to an adjacent one, as long as both districts remain contiguous. Repeat millions of times. Each step is a random walk through the space of all possible maps, and after enough steps, you've sampled a representative collection of "neutral" maps that respect basic constraints (contiguity, equal population, compactness) without political intent.
The Ensemble Method
Generate 10,000 random redistricting plans. Compute the partisan outcome of each. If the actual, enacted map produces an outcome more extreme than 99.5% of the random maps, it's strong evidence of intentional manipulation. The map isn't just unusual — it's astronomically unlikely to have arisen by chance.
This approach was used decisively in Common Cause v. Rucho and in Pennsylvania's 2018 redistricting litigation. Jonathan Mattingly at Duke generated ensembles showing that North Carolina's congressional map was a massive outlier — giving Republicans 10 of 13 seats from what was essentially a 50-50 state.6
Most random maps give 6–8 Republican seats. The enacted map (dashed red) gave 10 — a clear outlier, evidence of gerrymandering.
The Liars' Theorem
Gerrymandering manipulates who votes together. But there's an even more fundamental problem: voters themselves can manipulate the outcome by lying.
In Chapter 7, Arrow told us there's no perfect way to aggregate honest preferences. In 1973, Allan Gibbard and Mark Satterthwaite independently proved something worse: there's no way to make voters want to be honest in the first place.7
This is the Gibbard-Satterthwaite Theorem, and it's the evil twin of Arrow's result. Arrow says you can't design a perfect aggregation rule. Gibbard-Satterthwaite says you can't even trust the inputs, because rational voters have incentive to game them.
Strategic Voting in the Wild
You've probably done this yourself. It's 2000, and you're a Nader voter in Florida. You prefer Nader, but you know he can't win. If you vote your true preference, you might help elect Bush — your last choice. So you vote for Gore, your second choice, to block your worst outcome. You've just voted strategically: lying about your preference to improve the outcome.
This isn't irrational. It's mathematically optimal. And it's inevitable under plurality voting (first-past-the-post), where you cast a single vote for one candidate. The French political scientist Maurice Duverger noticed the downstream consequence: plurality voting systems inevitably collapse into two-party systems.8 Third parties get squeezed out not because nobody likes them, but because voting for them feels "wasted." Duverger's Law isn't just a sociological observation — it's a game-theoretic equilibrium.
Try It: The Strategic Voting Simulator
Below, set the preferences of 100 voters across three candidates. See who wins under honest plurality voting, then watch what happens when voters act strategically — and compare it to approval voting, where you can vote for as many candidates as you like.
🗳️ Strategic Voting Simulator
Adjust how many voters have each preference ranking. Then see the results under three systems.
Better Ballots?
If plurality voting is so broken, can we do better? The Gibbard-Satterthwaite theorem says no system is perfectly strategy-proof — but some systems make strategic voting much harder, riskier, or less rewarding.
Approval voting is the simplest reform: instead of voting for one candidate, you vote for every candidate you approve of. This eliminates the "spoiler" problem entirely. A Nader voter can approve both Nader and Gore without penalty. Approval voting has been adopted in St. Louis, Fargo, and several other cities.
STAR voting (Score Then Automatic Runoff) asks you to score each candidate from 0 to 5. The two highest-scoring candidates advance to an automatic runoff, decided by who was rated higher on more ballots. It's a clever hybrid that resists strategic manipulation because even if you "bullet vote" (give your favorite a 5 and everyone else a 0), the runoff stage still reflects broader preferences.
Neither system is perfect — Gibbard-Satterthwaite guarantees that — but both shift the strategic calculus dramatically. Under approval voting, your best strategy is usually just to vote honestly: approve the candidates you genuinely like. The gap between "honest" and "optimal" behavior shrinks, and that's the best we can hope for.
Gibbard-Satterthwaite's bitter punchline: perfect strategy-proofness requires dictatorship. Everything else is a compromise.
The Map Is Not the Territory (But It Decides the Election)
Here's the thread connecting everything in this chapter: democracy is not a single mechanism but a stack of mechanisms, and each layer can be manipulated. Arrow showed that the voting rule itself is imperfect. Gibbard-Satterthwaite showed that voters will exploit those imperfections. Gerrymandering shows that even before a single ballot is cast, the grouping of voters into districts can predetermine the outcome.
But the math doesn't just diagnose the disease — it offers treatments. The efficiency gap and ensemble methods give courts rigorous tools to detect gerrymandered maps. Approval voting and STAR voting make strategic manipulation harder and less rewarding. Independent redistricting commissions, guided by mathematical fairness criteria, can draw maps that resist partisan abuse.
None of these solutions are perfect. The theorems guarantee that. But "no perfect solution exists" is not the same as "all solutions are equally bad." Some voting systems are much less manipulable than others. Some maps are much fairer than others. The math tells us we can't reach the summit, but it also tells us exactly how high we can climb.
And if a politician ever tells you that the lines on the map don't matter — that it's just administrative convenience — remember Texas, 2003. Same voters. Different lines. Different government. The map is the election.