Skip to content

Credible Commitment

Prime #
760
Origin domain
Economics & Finance
Subdomain
game theory → Economics & Finance
Aliases
Precommitment Device

Core Idea

A promise or threat is credible only if it would still be carried out when execution arrives and the committing party would prefer not to. The structural answer to the time-inconsistency problem is deliberate constraint of one's own future choice set — the fix lives in the world, not the will.

How would you explain it like I'm…

Hide-The-Cookies Promise

Imagine you promise not to eat the cookies, but everyone knows you love cookies, so nobody believes you. Now imagine you give the cookie jar to your friend to hide far away. Even when you really want a cookie later, you can't get one. Now your promise is believable, because you made it impossible for future-you to break it.

Burn-The-Bridge Promise

Sometimes you really mean a promise when you make it, but later, when it's time to do it, you'd rather not. Other people can guess that, so they don't trust the promise. To make people believe you, you have to change things so that breaking the promise actually hurts you more than keeping it. Like burning the bridge behind your army so they can't retreat, or signing a contract with a penalty. A Credible Commitment lives in the world as a real cost, not just in your good intentions.

Tying Your Own Hands

A promise or threat is only believable if you'd still carry it out even at the moment you'd most prefer to back down. The problem is time-inconsistency: right now you sincerely plan to do something later, but once the other person has acted on your plan, your incentives flip and you'd rather not. They can foresee this, so they ignore your plan. The fix isn't being more sincere or trying harder. It's deliberately shrinking your own future choices, observably, so that doing the promised thing stays your best option even later. You don't fix it in your willpower; you fix it in the world.

 

Credible Commitment is the structural answer to the time-inconsistency problem in strategic interaction. At time t an actor genuinely intends to do X at t+1; but at t+1, having already induced a counterparty to act on that intention, the actor now prefers Y. A rational counterparty anticipates this and discounts the original intention, so the cooperative outcome collapses. Exhortation and sincerity cannot repair this, because the issue is the future self's incentives, not the present self's honesty. The mechanism is to deliberately constrain your own future choice set so that X remains best even from the standpoint of t+1, making your own non-compliance more costly than compliance, observably and verifiably, before the counterparty must act. The tactics vary enormously: burning bridges, posting bonds, ratifying constitutions, delegating to an independent agency, building irreversible specific assets, automating a response. Yet all share one shape, which is why this recurs across international relations, monetary policy, contract design, constitutional engineering, behavioral economics, and AI safety. The frame it carries with it is game-theoretic and human-bound: preferences, promises, intentional agents, strategic reasoning.

Broad Use

  • International relations: a retaliation threat is credible only if retaliation remains optimal after the provoking move; deterrence rests on surviving second-strike capability.
  • Monetary policy: an independent central bank is more credible at low-inflation promises than an elected government that can re-optimize opportunistically.
  • Contract law: enforceable contracts, collateral, escrow, and liquidated damages make promises mechanically expensive to break.
  • Constitutional design: separation of powers and entrenched rights constrain future majorities — "ambition counteracting ambition."
  • Personal life: marriage vows, automatic savings, and public bets bind the present self against the future self's predicted reversal.
  • AI safety: hard-coded refusals and immutable audit trails are technical commitments — the system cannot misbehave even if instructed to.

Clarity

It distinguishes intention from commitment, relocating believability from the psychology of the promiser to the structure of their future incentives, which is observable where sincerity is not. It also makes sense of deliberately reducing one's own options.

Manages Complexity

A heterogeneous landscape — independent central banks, hard-to-amend constitutions, liquidated-damages clauses, automated weapons command — collapses into one diagnostic: what makes the announced future behaviour incentive-compatible at execution?

Abstract Reasoning

Credibility is the substantive content of subgame-perfect equilibrium: a non-credible threat is one that would not actually be played in the subgame where it must be executed, and the minimal requirement is a cost-of-defection wedge large enough to flip the future self's best response.

Knowledge Transfer

  • Monetary policy to platform governance: independence, explicit targets, and transparency port to a platform promising not to favour its own services.
  • Constitutional design to AI alignment: binding a future legislature and binding a future AI system have the same structural form.
  • Backward induction to negotiation: "is this threat one they will still want to carry out if I call it?" is a transferable bluff-detector.

Example

An incumbent's threat to flood the market against an entrant is cheap talk — fighting loses money post-entry, so a rational entrant enters anyway; building excess capacity in advance (a sunk, observable asset) lowers the cost of fighting until it becomes the best response, and the threat is finally credible.

Not to Be Confused With

  • Credible Commitment is not a Commitment Device because the device is the mechanism (a bond, a burned bridge), whereas credible commitment is the strategic property of being believed because non-compliance was made costly.
  • Credible Commitment is not Sunk Cost because a sunk cost is a backward-looking irrelevance to ignore, whereas credible commitment deliberately sinks costs forward to alter future incentives.
  • Credible Commitment is not Optionality because optionality is the value of preserving future choices, whereas credible commitment is the value of destroying them — the two are exact strategic opposites.