So far you’ve seen the paradox: throwing away options can make you stronger, and a two-move game resolves cleanly by backward induction once one side truly locks itself in. But “truly” is doing a lot of work in that sentence. A commitment that exists only in your head, or only in your words, changes nothing. This lesson is about the one property that makes the whole machine run — and the toolkit for engineering it.
Before you read — take a guess
A small airline tells the dominant carrier on its route: 'Cut your fares to match ours and we'll cut ours even further — we'll bleed you dry.' The dominant carrier has ten times the cash reserves and could outlast a price war easily. What should the dominant carrier expect?
Talk is cheap
Here is a threat: “Raise your price and I’ll start a ruinous price war that destroys us both.” Here is why nobody flinches: if the war destroys you both, then when the moment comes to actually pull the trigger, you won’t want to. You’d be setting fire to your own house to smoke out the neighbor. A rival who can think one step ahead knows this, so the threat is just noise.
Economists call this cheap talk — a message that costs nothing to send and commits the sender to nothing. Cheap talk is the default state of most threats, promises, and ultimatums. “I’ll never accept less than $50,000.” “Do that again and I’m walking out.” “Match our price and it’s war.” All free to say, all free to abandon, all therefore ignorable by anyone paying attention.
The opposite of cheap talk is a real commitment, and the property that separates the two is credibility — a commitment is credible when the other side genuinely believes you’ll follow through with it. Not hopes you might. Believes you will. Credibility is not a feeling you have about your own resolve; it’s a belief the other side holds about your future behavior. If they don’t believe it, you don’t have a commitment — you have a sentence.
The whole course in one line
A commitment device only works to the exact degree the other side believes you’re stuck with it. Everything in this lesson is about manufacturing that belief.
The credibility test
You already know how to test a commitment — you just have to point the tool at yourself. In the previous lesson you used backward induction (reasoning from the end of the game back to the present) to figure out what a rival would do. The credibility test is the same move, aimed inward: when the decisive moment actually arrives, will carrying out your threat or promise be in your own interest?
- If yes — following through is what you’d genuinely choose when the time comes — the commitment is credible. The other side, running the same calculation, sees this and takes you seriously.
- If no — you’d rather quietly not do it — the commitment is an empty threat: a threat you wouldn’t actually execute, which a clear-eyed rival will simply call.
The classic empty threat lives at your own kitchen table. A parent snaps, “Behave, or we’re turning this car around and going straight home — no vacation.” But everyone in the back seat has run the numbers. The parent has paid for the hotel, taken the week off, and wants the vacation as much as the kids do. When the decisive moment comes, turning around punishes the parent more than the child. So the child, doing backward induction from the booster seat, keeps kicking the seat. The threat failed the credibility test before it left the parent’s mouth.
Contrast a threat that passes: “Behave, or you lose screen time tonight.” Confiscating a tablet costs the parent nothing and is easy to carry out. Follow-through is in the parent’s interest, so the threat is credible — and, not coincidentally, it works. The lesson isn’t “threaten harder.” It’s “threaten something you’d actually do.”
| Empty threat | Credible threat | |
|---|---|---|
| Turn the car around | Hurts the parent as much as the child → won’t happen | — |
| Lose tonight’s screen time | — | Costless to the parent, easy to execute → will happen |
| Fare war vs. a giant | Bankrupts you first → bluff | — |
| Signed penalty clause | — | Backing down now costs you more than following through |
The pattern is always the same: a threat is credible when the person making it has arranged things so that following through is their best remaining move. Which raises the obvious question — how do you arrange that on purpose?
How to make a commitment credible
You can’t just decide to be credible; the other side won’t take your word for it, because your word is cheap talk. Instead you change the game so that following through becomes your genuine best move — and, crucially, you make that change visible so the other side can see you’re stuck. Here is the toolkit.
Contracts and penalties
The most direct move: make backing down expensive. Sign something that inflicts a penalty on your future self if you fail to follow through, so that when the moment comes, executing the commitment is cheaper than eating the penalty.
A supplier promises delivery by March 1. Cheap talk. Now attach a penalty clause — $10,000 for every week late, enforceable in court — and the promise has teeth. The supplier’s future self, weighing “ship on time” against “pay $10,000 a week,” genuinely prefers to ship. You didn’t ask them to be trustworthy; you rearranged their incentives so that being trustworthy is now their cheapest option. Posting a performance bond (money held in escrow, forfeited if you renege) does the same thing.
Why a penalty on yourself helps you
It feels backwards to want a clause that can hurt you. But the penalty is a gift to your negotiating position: it’s proof, visible to the other side, that you can’t cheaply wriggle out. The cost you might pay is exactly what makes the promise worth believing.
Burning the option
Sometimes you don’t add a penalty — you delete the tempting alternative entirely, so there’s nothing left to be tempted by. When Hernán Cortés scuttled his own ships on the Mexican coast in 1519, he wasn’t being dramatic. He was removing “retreat by sea” from his army’s option set. With no ships, “sail home” is off the table, so every soldier’s backward induction ends at “advance.” The commitment to fight became credible because the alternative literally no longer existed.
You see the civilian version everywhere. A shop’s “no refunds, no exceptions” sign eliminates the shopkeeper’s discretion to cave to a persistent complainer — and because customers know the option is gone, they stop trying to argue it back. A writer who deletes the streaming apps and hands their phone to a friend has burned the option to procrastinate. The trick is that burning must be real: a “no refunds” sign that the manager quietly overrides is just cheap talk with better typography.
Public commitment and reputation
Stake your name. If you announce your commitment loudly enough, backing down becomes publicly humiliating and — worse — it teaches everyone watching that your future words are cheap talk. That reputational cost, paid across every game you’ll ever play again, can be large enough to make following through your best move even when the immediate stakes are small.
A central bank that publicly targets 2% inflation has bound itself in exactly this way: reversing course would shred the credibility it needs for every future policy statement, so it fights inflation harder than a bank that made the same promise privately. A negotiator who announces a red line on live television has done the same thing on purpose — retreating now means eating the humiliation in front of the whole audience, which is precisely why the other side believes the line will hold.
Reputation is the version of this that compounds. In a repeated game (the same players interacting again and again), a track record of always following through is your commitment device — people believe today’s threat because you executed yesterday’s. The pitfall: this only works if enough people are watching and remember. Which is exactly why the next section matters.
A CEO tells employees privately, one-on-one, 'If quarterly numbers slip, there will be layoffs' — but never says it in any meeting, memo, or public forum. As a commitment device, what's the core weakness here?
Delegation to an agent who must follow through
Here’s a subtle one: hand the decision to someone who can’t or won’t concede, and their inflexibility becomes your commitment. You gain credibility precisely by giving up control.
A union representative at the bargaining table can say, truthfully, “I am not authorized to accept anything below the members’ vote.” That’s not a bluff — the rep genuinely lacks the power to cave, so management can’t squeeze them, and everyone knows it. “My hands are tied, my boss won’t sign off on that” is the everyday move: you’ve delegated the yes/no to an absent decider who won’t budge. Hiring a famously ruthless lawyer does it too — you’ve delegated your response to someone whose incentives and temperament won’t let them settle cheap.
This is the principal–agent relationship (a principal hands authority to an agent who acts on their behalf) used as a commitment device. Normally a principal worries about an agent who won’t do what they want; here you deliberately pick an agent who can’t do the soft thing, and their constraint is your strength. The pitfall: it only works if the delegation is real. If the other side suspects you can pick up the phone and overrule your “agent,” you’re back to cheap talk — now with an extra layer of theater.
Automaticity — the doomsday machine
The strongest move is to remove your own discretion entirely, so the response is automatic and you literally cannot chicken out. In Dr. Strangelove, the Soviets build a doomsday machine — a device that automatically triggers global retaliation if the USSR is attacked, with no human able to stop it. Its whole value is that no one can lose their nerve at the last second: the retaliation isn’t a choice anymore, so the deterrent is total.
You don’t need the apocalypse to use this. A stop-loss order on a stock (“sell automatically if it drops to $40”) commits your future, panicking self to a decision your calm self made in advance — you can’t talk yourself out of it in the moment because you’re not consulted. An automatic tripwire — a tax that kicks in the instant a condition is met, or a defensive garrison stationed where any attack must hit it — makes the response mechanical.
The twist that trips people up: automaticity is credible because it takes the decision out of your hands. A threat you control is a threat you might soften; a threat that fires by itself can’t be softened, so the other side has to believe it. But — as the movie’s grim joke insists — a doomsday machine only deters if the enemy knows it exists. A secret automatic response is the ultimate cheap talk: it changes nothing, because nobody adjusts to a machine they can’t see.
Which of the following genuinely make a threat MORE credible? (Select all that apply.)
Sort the credible from the empty
Time to calibrate your ear. The test never changes: a commitment is credible when the actor has arranged things so following through is now their best (or only) move — a signed penalty, a burned option, an automatic device, a delegate who can’t concede, or real reputation on the line. It’s empty when the actor would quietly rather not follow through and everyone can see that — a threat that hurts the threatener more, an unenforced promise, a private commitment nobody witnesses, or a deadline with no teeth.
Drop each statement into the bucket that fits. Ask the credibility test of each: when the moment comes, is following through this actor's best move — and can the other side see that?
- A startup burns its bridges by publicly deleting its old product line, so there's no falling back to it.
- A supplier signs a contract with a steep per-week late penalty enforceable in court.
- A tiny airline threatens a price war that would bankrupt itself long before the giant it's aimed at.
- A parent threatens to cancel the paid-for family vacation they want as much as the kids do.
- A CEO warns of layoffs only in private one-on-ones, never where it costs them to back down.
- A boss announces a 'hard deadline' that has quietly slipped four times with zero consequences.
- A union rep who literally lacks the authority to accept less than the members' vote.
- An investor sets an automatic stop-loss that sells the moment the price hits its floor.
Where credibility bites back
The toolkit is powerful, which means it fails in instructive ways. Three traps to watch:
Invisible commitments change nothing. A commitment device only works through the other side’s beliefs, so if they can’t see it, it does no work. The doomsday machine kept secret; the penalty clause the rival never learns about; the red line drawn in a diary. You have to pay the cost of the commitment and broadcast that you’ve paid it. A brilliant commitment nobody witnesses is just an expensive way to constrain yourself for no strategic gain.
Over-committing and over-signaling. Burn too many options and you lose the flexibility to adapt when the world shifts — the general who scuttled the ships can’t retreat even when retreat is genuinely the smart move. Signal too aggressively — red lines on every issue, penalty clauses on every promise — and you either box yourself into a corner you can’t get out of, or you train the other side to discount your signals as noise. Commitment is a scalpel, not a hammer; each one you make is an option you can never use again.
Threats too costly to execute read as bluffs. This is the cheap-talk problem in disguise. If following through would hurt you enormously — mutually assured destruction over a minor squabble, a lawsuit that costs you more than you could ever win — a thinking rival runs your backward induction for you, sees you’d never actually do it, and calls the bluff. The most credible threats are often the modest ones you’d genuinely carry out, not the apocalyptic ones you obviously wouldn’t. Bigger is not more believable; executable is more believable.
A country draws a dramatic public 'red line' — vowing overwhelming retaliation over a fairly minor provocation — but everyone knows that retaliating would trigger economic self-harm far worse than the provocation itself. Why does the threat likely fail?
Key takeaways
- Cheap talk is the default. Threats, promises, and ultimatums cost nothing to say and nothing to abandon, so a thinking rival ignores them until proven otherwise.
- Credibility is a belief the other side holds about your future behavior — not a feeling you have about your own resolve.
- The credibility test: aim backward induction at yourself. When the decisive moment comes, is following through your best move? If not, it’s an empty threat and a smart rival will call it.
- The toolkit all works by making follow-through your best move: contracts & penalties (make backing down expensive), burning the option (delete the tempting alternative), public commitment & reputation (make retreat humiliating and costly across repeated games), delegation (hand the decision to an agent who can’t concede), and automaticity (remove your own discretion so you can’t chicken out).
- Visibility is non-negotiable. A commitment the other side can’t see changes nothing — you must pay the cost and broadcast it.
- Executable beats apocalyptic. A modest threat you’d genuinely carry out is more credible than a devastating one everyone knows you’d never pull the trigger on.