Moloch Comes to Virginia

Today Virginians vote on a constitutional amendment I would normally oppose on its merits. The mechanics are ugly: a mid-decade reapportionment of the state’s congressional districts, drawn for transparent partisan advantage, pushed through by the party that happened to control both chambers of the General Assembly and the governorship when the opportunity appeared. Under ordinary conditions the answer is an easy no. These are not ordinary conditions.

Here is the board. Virginia currently sends a 6-5 Democratic congressional delegation to Washington, which tracks the state’s political geography well enough: strong blue clusters in the north, strong red territory to the south and west, a slight statewide tilt toward blue. The bipartisan redistricting commission that voters approved in 2020 produced a map that, for all its imperfections, comes close to honest. The amendment on today’s ballot replaces that map with one projected to yield ten Democrats and a single Republican.

Ten to one. In a state whose statewide partisan divide sits near 53-47. That is not a correction; that is a mugging dressed as a correction.

The game-theoretic context makes the vote difficult. The opening move came from Texas in the summer of 2025, when Republicans there pushed through a mid-decade redraw that would net the GOP roughly five seats. California answered with Proposition 50 in November, a constitutional amendment that passed with 64 percent of the vote and set up a roughly five-seat flip the other way. Missouri moved. Other states moved. Virginia is moving now.

This is Moloch in perfect form. A textbook multipolar trap, so clean I could teach from it.

Scott Alexander’s original Moloch essay

Here is the structure. Every participant is making the locally correct move. A player who dislikes the norm-violation but refuses to match it gets strictly worse outcomes than a player who matches the norm making move. Unilateral restraint in a defection game does not produce a better world; it produces the same defected world with one fewer voice at the table. The prisoner who refuses to defect after the other prisoner already has is not nobler, just convicted. Each state legislature can correctly reason: “I would prefer a world where no one gerrymandered between censuses; given that someone already did, my choice is not between gerrymandering and clean maps but between gerrymandering and letting the other side gerrymander alone.”

That reasoning is right. It is also how the worst equilibrium arrives. Every locally rational move adds up to a House of Representatives drawn by partisan operatives in state capitals rather than by anything resembling the population.

Virginia’s proposed 10-1 map illustrates the second ugly property of these dynamics: the response need not be proportional, and typically isn’t. California’s gerrymander at least had the rough shape of a counterweight to Texas. Virginia’s map has no such character. Once the norm against mid-decade partisan redistricting cracks, nothing constrains the scale of each response except the size of the ruling majority and the nerve of its members. This is the ratchet property of multipolar traps: each round resets the baseline, the baseline only moves in one direction, and the moves get bigger as participants learn what the game now permits.

How do multipolar traps actually get resolved? Exactly one way, historically. By agreement among the players. The form of the agreement varies: an informal norm maintained by reputational cost, a treaty, a domestic statute, a constitutional provision. The mechanism is always the same. The players bind themselves together to a rule that none of them can afford to follow alone. Nuclear non-proliferation, the Montreal Protocol, the medieval Peace and Truce of God, fisheries quotas that actually hold: the shape is recognizable across scales and centuries.

For congressional redistricting, the shape of a workable agreement is not mysterious. Federal legislation establishing that congressional maps may be redrawn only after a decennial census, absent court order, would dissolve the trap in a single stroke. So would a constitutional amendment to the same effect. Either requires cross-partisan cooperation of a kind not currently on offer. The trap will therefore continue, and the equilibrium will worsen, until the cost of the game exceeds the cost of coordinating. That crossover point is not close.

So today’s vote is not a referendum on gerrymandering. The referendum on gerrymandering was lost in Austin last summer, and every subsequent state has been voting inside the wreckage. What remains on the ballot is whether to accept further decline in return for maintaining one’s strategic position, or to absorb the cost of unilateral restraint and watch the other side collect the winnings. That is the choice a multipolar trap presents by design. It is the choice it will keep presenting until somebody changes the rules.

You could not ask for a cleaner demonstration. It is also not a demonstration I wanted to live inside.

Today I voted yes on the amendment, even though it is part of a bad pattern. That is the power of a multipolar trap: game theory forces us to do the wrong thing. Outlawing the move is the only way to release all of us from the trap.