Commitment devices, and what they are for

There is a word for the thing that separates a plan from a resolution, and economists have been studying it for about forty years.

The definition worth keeping

A commitment device is a voluntary, self-imposed constraint that makes a future choice either impossible or costly, adopted specifically because you expect your future self to want something different from what you want now.

The three parts each matter. Voluntary, or it is just a rule someone imposed on you. Self-imposed in advance, at a moment when you can think clearly. Costly to break, or it is a note on the fridge.

Gharad Bryan, Dean Karlan and Scott Nelson set out the framework in the Annual Review of Economics in 2010. The economic motivation is time-inconsistent preferences: your ranking of options changes depending on when you are asked. At 9pm you would prefer to be asleep by eleven. At 1am, sleep ranks below one more video. Neither preference is fake. They simply belong to different moments, and only one of them gets to hold the phone.

A commitment device is how the 9pm version reaches forward and takes an option away from the 1am version.

What the trials found

Two randomised studies do most of the work in this literature, and they are worth quoting accurately, because the popular versions of both have drifted.

Savings, Philippines, 2006. Ashraf, Karlan and Yin offered bank customers a savings account that restricted withdrawals until a self-chosen date or amount. The account paid no extra interest — its only feature was the restriction. Reported in the Quarterly Journal of Economics: of 710 clients offered the account, 202 took it — 28.4% — and after twelve months average savings balances in the treatment group were 81 percentage points higher than in the control group. People paid, in liquidity, for the privilege of not being able to get at their own money.

Smoking, Philippines, 2010. Giné, Karlan and Zinman tested CARES: smokers deposited their own money for six months, then took a urine test for nicotine. Pass and the money came back; fail and it went to charity. Published in the American Economic Journal: Applied Economics, the result was that 11% of those offered it took it up, and being offered the contract made someone three percentage points more likely to pass at six months. The effect was still visible in surprise tests at twelve months.

Three percentage points is a modest number, and you will find blog posts quoting figures ten times larger for this study. They are wrong. It is worth knowing the real one, because the real one is still remarkable: a voluntary contract, offered once, with the participant's own money, produced a detectable and durable change in a behaviour that resists nearly everything.

Note also the take-up rates. 28.4% and 11%. Most people offered a commitment device decline it — which is itself the most honest thing anyone can tell you about this category. It is not for everyone, and the people it helps are the ones who already suspect they will need it.

The four rules

From the research and from building one, the properties that separate a working commitment device from theatre.

1. The stake must be real and it must actually leave. Money that returns to you on failure is not a stake, it is a deposit. Points, streaks and badges are not stakes at all — losing a 43-day streak costs you a number. The literature's designs forfeit real money to a third party, and that is not incidental.

2. The terms must be fixed before the moment. A rule you can renegotiate under pressure is not a constraint, because renegotiating under pressure is precisely the behaviour you were trying to prevent. Every price in our own product is printed before it is charged, including the next one, for this reason and no other.

3. The exit must cost something that does not get cheaper. This is the rule most consumer software breaks. Friction made of seconds and taps is cheap at 1am, when your willingness to spend seconds is at its peak. A fixed cost has the property you actually need: it is the same amount at 1am as it was at noon.

4. Paying must not dissolve the constraint. If paying the penalty also resets the terms, you have built a subscription to breaking your own rule. In Last Ferry the price ladder never resets on payment — top up while three rungs up and you resume three rungs up. If money could reset the ladder, the ladder would be theatre.

Applying this to a phone

The phone case has one feature that makes it harder than smoking or saving: the decision recurs dozens of times a day, and each individual instance is trivially defensible. Nobody's evening is ruined by one more video. It is ruined by ninety of them, and no single one is the culprit.

That rules out the classic design. A six-month contract with a lump-sum forfeit does not fit a behaviour that happens ninety times before bed. What fits is a small, immediate, escalating price attached to each instance — enough to be noticed, not enough to be punitive, and rising so that the second instance is a fresh decision at a new number rather than a repeat of a decision already made.

That is the shape of what we built, and the reasoning for the escalation is written up in why the price rises. The first unlock of the day costs $0.50; the second $1.00; the third and every one after $2.50. At midnight the ladder resets, because tomorrow is a different day and there is always another boat.

Two details keep it from becoming a shop. Credit bought now becomes spendable thirty minutes later, so a purchase can never turn into scrolling in the instant it was made — the reasoning is in you cannot buy a ticket for a boat that has already left. And the honest metric: the percentage of users who buy nothing in a month is tracked as a success measure, on the same dashboard as revenue. Most days the right outcome is that we make no money.

The uncomfortable part

Stated coldly, a business that charges for unlocks makes more money the worse your day is going. We wrote that sentence about ourselves before anyone else could, and writing it does not stop it being true.

The defence is not that we are nice. It is structural: blocking is free forever, because a commitment product that puts its commitment behind a subscription has built a paywall-shaped exit and pointed at it. Cancelling would be the bypass. So the constraint costs nothing, and only going past it costs anything — which means the version of this product that makes the most money is also the version that works least well, and we would rather you could see that in the design than take our word for it.

Sources

• Bryan, G., Karlan, D., & Nelson, S. (2010). Commitment Devices. Annual Review of Economics, 2, 671–698.
• Ashraf, N., Karlan, D., & Yin, W. (2006). Tying Odysseus to the Mast: Evidence From a Commitment Savings Product in the Philippines. The Quarterly Journal of Economics, 121(2), 635–672.
• Giné, X., Karlan, D., & Zinman, J. (2010). Put Your Money Where Your Butt Is: A Commitment Contract for Smoking Cessation. American Economic Journal: Applied Economics, 2(4), 213–235.