Do Commitment Devices Actually Work? What the Evidence Shows

You already know what to do. The problem is that your future self keeps voting against it.

A commitment device is a voluntary constraint you impose now to bind that future self — a bet with a referee, a locked savings account, a deadline with real teeth. Do they work? Randomized field trials say yes, with one sharp caveat. A Philippine commitment savings product raised balances 81 percentage points, and a smoking-cessation contract raised the odds of quitting by 3 percentage points. The catch is adoption: only a minority ever opt in — 11% to 28% in these trials — and a device that is too rigid can punish you when life legitimately intervenes. The lever is real. The bottleneck is using it.

What is a commitment device?

It is a way to change your own choice set so the tempting option costs you something or disappears entirely. The mechanism it fights is present bias — why we choose now over later: you overweight immediate rewards and steeply discount future ones, so the important-but-not-urgent reliably loses to whatever pays off now.

Willpower attacks that bias head-on and usually loses. A commitment device sidesteps it. Like Ulysses lashing himself to the mast before the sirens, you make the decision once, while you are clear-headed, and remove your later self’s ability to renegotiate.

What did the savings experiment actually find?

The cleanest evidence comes from Ashraf, Karlan and Yin (2006, Quarterly Journal of Economics). Working with a bank in the Philippines, they built a product called SEED: a savings account that restricted withdrawals until the client hit a self-set goal — a target date or a target amount. Crucially, it paid the same interest as a normal account. The only new feature was the lock.

They randomly offered it to 710 clients. 202 — 28.4% — opened one. After twelve months, average savings balances rose 81 percentage points for the treatment group relative to control. And it held: the authors conclude the effect “represents a lasting change in savings, and not merely a short-term response to a new product.”

Sit with the information most people miss: these clients gave up nothing in returns. They paid in liquidity — the freedom to raid their own money — because they knew that freedom was exactly what kept sinking them.

Can a commitment device help you quit something?

Giné, Karlan and Zinman (2010, American Economic Journal: Applied Economics) tested this with a study memorably titled “Put Your Money Where Your Butt Is.” Smokers deposited their own money into an account called CARES. After six months, they took a urine test for nicotine. Pass, and the money came back. Fail, and it was forfeited to charity.

Of smokers offered CARES, 11% took it up. Those offered the account were 3 percentage points more likely to pass the 6-month test than control — and the effect persisted at a surprise test twelve months later, after the money was long gone.

Three points sounds modest, but read it correctly: that is the effect across everyone offered, including the 89% who declined. Concentrated among the people who actually signed a contract with their own money on the line, the bite is far larger. The design is what a founder can copy: real stakes, a clear pass/fail test, and a loss that stings.

So why don’t they work more often?

Two honest limits.

First, adoption is the ceiling. Uptake was 11% and 28.4%. A commitment device only works if you use it, and most people will not voluntarily hand over their own escape hatch — the same present bias that makes the device necessary also makes you reluctant to install it.

Second, rigidity cuts both ways. A device strict enough to bind you when you are weak will also bind you when circumstances genuinely change. A forfeiture triggered by a real emergency is not discipline — it is a penalty for being human. The failure mode of commitment devices is not that they are too soft. It is that they are too brittle.

How do you build one that actually works?

Put real stakes on the line. CARES worked because losing money hurts more than gaining it feels good. A public bet with a referee beats a private resolution every time.
Restrict the option — don’t just intend to. SEED physically locked withdrawals; that is why it outperformed good intentions. Pair the lock with implementation intentions — if-then plans so the constraint fires on a specific trigger, not on willpower.
Size the stake to bite, not to ruin. Enough that failing is painful; not so much that a legitimate emergency becomes a catastrophe.
Add a carrot, not just a stick. Commitment devices are the penalty side of present bias. Temptation bundling is the reward side — pairing a want with a should. Run them together and you attack the bias from both directions.

The takeaway

Commitment devices are among the few self-control tools with randomized field evidence behind them: real, durable behavior change in savings and in quitting. But the data is blunt about the limits. They move the minority who adopt them, and they backfire when they are too rigid to flex. The skill was never willpower. It is designing a constraint your future self cannot easily overrule — then actually opting in. For where this fits in a full behavior-change system, see our AI hypnotherapy and behavioral change pillar.