The Game
Game theory is the operating system of politics. Every tariff threat, every filibuster, every "bipartisan" handshake is a move in a game — players, strategies, payoffs, everyone anticipating everyone else. Learn the games and the news starts reading like a playbook. Taught from zero. No math degree required.
The core idea, in one paragraph
A game is any situation where your best move depends on what everyone else does. Politics is nothing but that: a candidate's message depends on the opponent's message, a country's tariff depends on the other country's tariff, a senator's vote depends on how the vote will be used against them in the next primary. Game theory is the study of these situations — it asks: given the payoffs (who gains what from each outcome), what will rational players actually do? And the uncomfortable answer, over and over, is: not what's best for everyone. The structure of the game decides the outcome more than the character of the players. Change the payoffs, change the behavior. That's why this page exists — because policy design is game design, and most bad policy is just a badly designed game.
1. The Prisoner's Dilemma — why cooperation keeps dying
The setup: two suspects, held separately. If both stay silent, both get a light sentence. If one rats and the other stays silent, the rat walks free and the silent one gets hammered. If both rat, both get a medium sentence. Each prisoner's best move — no matter what the other does — is to rat. So both rat, and both end up worse than if they'd cooperated. That's the dilemma: individual rationality produces collective stupidity.
In politics: tariff wars. Two countries would both be richer with open trade — but each one's best move, assuming the other might cheat, is to slap on tariffs first. Arms races work the same way: both sides would be safer spending the money on hospitals, but neither can afford to disarm first. Every trade war in history is two prisoners ratting on each other and calling it strategy. See the tariff autopsy for a live specimen.
The way out is the whole story of civilization: binding agreements, enforcement, and repeated play (more on that in № 9). Treaties, trade deals, arms control — they're all attempts to rewrite a prisoner's dilemma into a game where cooperation survives.
2. Chicken — brinkmanship as a governing strategy
Two drivers speed toward each other. Swerve and you're the coward; hold your line and the other driver swerves, you're the hero. Both hold their line and everybody dies. The winning move is convincing the other driver you're crazy enough — or committed enough — never to swerve.
In politics: the debt ceiling. Every few years, one party drives the full faith and credit of the United States straight at a cliff and dares the other party to swerve. Government shutdowns are chicken with federal workers as the hood ornament. The game rewards whoever is most believable about being willing to crash — which is why the most extreme voices get the most leverage. Chicken structurally rewards extremism. Remember that the next time someone asks why Congress keeps electing people who won't swerve.
3. Credible threats — why bluffs have an expiration date
A threat only works if the other side believes you'll carry it out. "Do what I want or else" is worthless if the "or else" would hurt you more than them — everyone can see that, so nobody flinches. Game theorists call a believable threat credible, and credibility has to be built: by tying your own hands, by making the threat automatic, or by paying a visible cost.
In politics: this is the entire art of deterrence. Sanctions with automatic triggers are credible; sanctions the president can waive with a phone call are suggestions. A politician who stakes their reputation on a red line — publicly, on camera — has created what theorists call audience costs: backing down now means humiliation, so the threat gets teeth from the politician's own ego. That's why leaders make threats on television instead of in private. The audience isn't the public. The audience is the other leader, watching you trap yourself on purpose.
The tell: when someone threatens something they'd obviously never do, it's not a strategy — it's theater for the home crowd. Ask of every threat in the news: would carrying this out hurt the threatener more than the target? If yes, it's a bluff. Bluffs work until the day they don't, and the day they don't is always expensive.
4. Cheap talk vs. costly signals — words are free, actions aren't
Cheap talk is communication that costs the speaker nothing — and rational players should ignore it, because anyone would say it. Costly signals are messages you can only afford to send if you mean them: burning money, burning bridges, burning political capital.
In politics: "thoughts and prayers" after a tragedy is cheap talk in its purest form — zero cost, zero information. A campaign promise is cheap talk. But a politician who spends their own limited political capital — endorsing a risky candidate, breaking with their party on a recorded vote — is sending a costly signal, and you should update your beliefs accordingly. Military exercises near a disputed border are costly signals. So are resignations on principle: quitting a comfortable job to make a point is a signal you can't fake.
The rule: ignore what cost them nothing. Watch what cost them something. Most political "news" is cheap talk about cheap talk.
5. The Stag Hunt — trust is a technology
Two hunters can cooperate to take down a stag (feeds both for a week) or defect to chase a hare (feeds one for a day). If you trust the other hunter to hold position, you hunt the stag. If you doubt them, you grab the hare before they do. Unlike the prisoner's dilemma, both outcomes are stable — cooperation works if there's trust, and collapses without it. The game doesn't pick the outcome. Trust does.
In politics: climate agreements. Every nation would be better off if all cut emissions (the stag), but each is tempted to free-ride on others' cuts (the hare). The Paris Agreement isn't really about carbon targets — it's a trust-building machine trying to move the world from the hare equilibrium to the stag equilibrium. Same with pandemic preparedness, ocean fishing quotas, and every commons problem ever.
The lesson: some of the most important political work isn't policy at all — it's manufacturing the trust that makes the good equilibrium reachable. Summits, verification regimes, inspectors, transparency: that's all trust technology. Mock the pageantry if you want, but the pageantry is load-bearing.
6. Principal–agent — why your representatives don't represent you
A principal hires an agent to act for them, but the agent has their own interests and the principal can't watch them every minute. The agent shirks, self-deals, or pursues their own goals — and the principal can't fully stop it. It's the equilibrium.
In politics: voters are the principals, elected officials are the agents. Your representative's interests (re-election, fundraising, a future lobbying job) overlap with yours only partially. Monitoring them is expensive — that's why economists talk about rational ignorance: it genuinely doesn't pay any individual voter to master the details of every bill, so nobody does, and the agents operate in the dark. Lobbyists, meanwhile, monitor full-time — which is why they get the meetings and you get the form letter.
The fix is structural, not moral: transparency rules, recorded votes, short leashes (frequent elections), and bans on the revolving door all shrink the gap between what the agent wants and what the principal can verify. "Throw the bums out" feels good; redesigning the contract works better.
7. The Median Voter — why both parties sound the same, until they don't
In a two-candidate race with voters spread along a spectrum, both candidates converge on the voter in the exact middle — because whoever wins the middle wins. That's the median voter theorem: competition drives both sides to the center.
In politics: it's why general-election candidates suddenly sound moderate after sounding radical in the primary. But here's the catch the textbooks bury: the theorem assumes one election. In reality there are two games — the primary (where the median is the party's base, far from center) and the general (where the median is the whole electorate). Primaries reward extremism; general elections punish it. The candidate who wins both is the one who can play two different games six months apart and hope nobody kept the receipts. This is also why safe districts produce radicals and swing districts produce moderates — the relevant median moves, and the politician follows it like a compass.
8. Nash equilibrium — gridlock is nobody's fault and everybody's choice
A Nash equilibrium is a standoff where nobody wants to move first, because moving alone makes you worse off. Nobody's happy. Nobody moves. It's stable anyway.
In politics: congressional gridlock. Every member would prefer some deal to nothing — but the first one to compromise gets primaried (challenged in their own party's primary for not being pure enough), so nobody compromises. The filibuster, mutually assured obstruction, the endless debt-ceiling dance: they're the equilibrium of the system as currently designed. Yelling at individual politicians for gridlock is like yelling at water for being wet. If you want different behavior, you don't need better people — you need a different game: different rules, different payoffs, different consequences for the first mover.
9. Repeated games — the shadow of the future
Everything above assumes one round. But politics is played again and again — and that changes everything. In a repeated game, today's defection gets punished tomorrow, so cooperation becomes rational. The classic winning strategy in tournaments was tit-for-tat: cooperate first, then copy whatever the other player just did. Nice, retaliatory, forgiving, clear — and brutally effective.
In politics: this is why long-serving legislators could cut deals across the aisle — they'd be seeing each other for decades, and betrayal had a price. It's also why lame ducks are dangerous: a politician who'll never face voters again is playing their last round, and game theory says the last round always defects. Term limits, retirement announcements, lost primaries — every one of them dissolves the shadow of the future. And notice what that implies about our era: as politics gets more short-term, more primary-driven, more performative, the games get shorter — and shorter games defect more.
10. Moral hazard — privatize the gains, socialize the losses
Moral hazard is what happens when someone is protected from the consequences of their risks, so they take bigger ones. Insurance against your own recklessness doesn't make you careful — it makes you reckless with confidence.
In politics: the 2008 bailouts. Banks kept the profits from reckless bets and handed taxpayers the losses — and the game-theoretic prediction is exactly what happened: the lesson Wall Street learned wasn't "don't gamble," it was "gamble bigger, you're covered." "Too big to fail" is moral hazard as official policy. It shows up everywhere: disaster relief that subsidizes building in flood zones, farm bailouts that dull the signal of a bad market, and every corporate rescue that teaches executives the downside isn't theirs.
The test: whenever a policy shields someone from consequences, ask what behavior it subsidizes. The answer is always: more of the behavior that needed shielding.
The meta-lesson: policy design is game design
Here's the whole page in one sentence: if a policy keeps producing the behavior it claims to hate, the payoffs are wrong. Don't ask why people are selfish — ask what the game rewards. Corruption is what happens when the payoff to cheating exceeds the payoff to honesty and nobody's watching (principal–agent, № 6). Partisan extremism is the equilibrium of primary elections (median voter, № 7). Endless trade wars are prisoners ratting (№ 1).
This is why "follow the incentives" is the whole masthead. Incentives are the payoff matrix. And it's why reform that ignores game design always fails: you can't shame people out of an equilibrium. You have to rebuild the game — change who moves first, change what defection costs, change what cooperation pays. Every autopsy on this site is really a game report: what were the payoffs, who played optimally, and who got played.
Where the models break
Honesty requires the limits. Game theory assumes players are rational, self-interested, and well-informed — and people are routinely none of the three. Culture, identity, spite, loyalty, and sheer confusion move real politics as much as payoff matrices do. A model that predicts everyone defects can't explain the volunteer fire department, the whistleblower, or the voter who shows up in the rain for a school board race. Use game theory the way a carpenter uses a level: it tells you which way the structure leans, not what the house looks like. The pull is real. The person is more complicated than the pull. Anyone selling you game theory as a complete theory of politics is selling you № 4 — cheap talk.