Coalition Politics Is an Iterated Game That Ends Every Election, So Everyone Defects

I have written before about Robert Axelrod's The Evolution of Cooperation and what it says about relations between states. That was the foreign-policy half. There is a domestic half, and Israeli coalition politics is about the clearest illustration of it that exists.

The Result, Briefly

Axelrod's tournament pitted strategies for the iterated prisoner's dilemma against each other. The winner was tit-for-tat: cooperate first, then do whatever the other player did last time. It won not because it was clever but because it was nice, retaliatory, forgiving, and clear, and because the game repeated. In a repeated game, defection today is punished tomorrow, so cooperation becomes the profitable strategy.

The condition matters more than the strategy. Axelrod was explicit about it: cooperation emerges when the shadow of the future is long, meaning the players expect to meet again many times and value those future meetings. Shorten the shadow and cooperation collapses. In a game known to end at a fixed round, the rational move in the last round is to defect, and by backward induction, so is every earlier move.

Israeli Coalitions as a Game With a Known End

Now look at how an Israeli government is formed. Several parties, none with a majority, negotiate a coalition. Each promises the others support on their priorities in exchange for support on its own. Ministries are divided. Written agreements are signed.

This is a repeated prisoner's dilemma among the coalition partners: each can honour the deal or defect (vote against a partner's bill, threaten to bring down the government, extract more than was agreed). And it has a structural feature that Axelrod's analysis says is fatal: everyone knows when the game ends. The election is scheduled, or will be forced, and at that point all accounts are settled by the voters, not by the partners.

So as the election approaches, every partner faces the last-round logic. Cooperating with a coalition partner in the final months buys nothing, because there will be no further rounds in which the partner can reciprocate; the next coalition will be a different game with different players. Defecting, on the other hand, pays immediately: a party that publicly breaks with the government over an issue its voters care about takes credit for the break and pays no future cost.

Which is exactly what happens. Coalitions in Israel rarely reach the end of their term. They come apart in the last year, over issues that could have been managed a year earlier, because a year earlier the shadow of the future was still long enough to make managing them worthwhile.

Why Tit-for-Tat Cannot Save It

Tit-for-tat needs to be able to punish. In a coalition, punishment means voting against your partner's bills, and in the final stretch that punishment has no bite: the partner was going to lose those bills to the election anyway. Worse, the punisher looks like the defector to voters who are not following the sequence, so retaliation carries a cost that Axelrod's abstract players never paid.

Forgiveness fails too. Tit-for-tat forgives a defection the moment the other side returns to cooperation. But a party that defected in month 40 and returns in month 44 is returning to a game with four months left; there is nothing to forgive it into.

The strategy that wins iterated games cannot be played in a game whose iterations are counted down in public. This is not a failure of the politicians involved. It is what the structure rewards, and a coalition of saints would face the same incentive.

What the Game Would Have to Look Like

Axelrod's own recommendations for promoting cooperation are, in effect, design specifications: enlarge the shadow of the future, make interactions more frequent and more durable, make defection visible and its punishment certain.

Apply them to a legislature and you get something that looks little like the current arrangement. Decisions would be more frequent and smaller, so that each one is a round in a long game rather than a stake in a single all-or-nothing settlement. The players would persist across elections, so that a defection in one term is remembered in the next. And the reckoning would come continuously, from the public on each question, rather than once every few years on a bundle.

That description is, not coincidentally, a description of deciding issues directly rather than electing coalitions to decide them. A public that votes on questions is a player that never leaves the game. There is no last round to defect in, because there is always another question next month, and the party that betrayed the public on this one will face the same public on that one. The shadow of the future becomes infinite, which is the one condition under which Axelrod's tournament says cooperation can be relied upon.

Coalition politics is a game designed, by accident, to end. Everything Axelrod found about cooperation says that a game designed to end is a game everyone eventually defects in. Israel demonstrates it roughly every three years.