Savings commitment devices

Why do people lock money into an account that pays nothing to give it back?

A Christmas club takes a few pounds from you every week from January, pays no interest at all, and hands the total back in November. A goal-locked savings pot refuses withdrawals until a date you set. A rotating savings club takes your contribution on pay day and gives you nothing until your turn comes round.

Line any of these up against an ordinary instant-access account and they lose on every measurable dimension. Same money in, less money out, and you cannot reach it. In standard economic terms the flexible account dominates: an option you are free not to exercise can only help you. So the demand for these products is a small scandal. What exactly is being bought?

Start by stripping the product down to what is genuinely different about it. Not the return — there usually isn't one. Not the safety — a bank account is safer. The single feature a commitment account has that a flexible one lacks is a restriction. Which means the restriction is not a side effect of the product. It is the product.

So ask when a restriction could be worth paying for. Only in one circumstance: if you expect a future version of yourself to make a choice that the present version would rather it did not make. If your preferences were stable, this could never arise — the you of June would want exactly what the you of January wanted, and the lock would sit there doing nothing but costing interest. The willingness to pay for a lock is therefore evidence about the person, not about the account: it says the saver expects to disagree with themselves.

Now the step that does the real work here, because it separates two people who are both bad at saving. Suppose someone is impatient in the way that shifts when the money gets close — the reversal that delay discounting describes. If they are naive about it, they will not buy a commitment device, because they sincerely believe they will save next month. Demand for a lock comes only from the sophisticated case: the saver who has watched themselves fail and now treats their own future choices as a hazard to be designed around. The device does not improve willpower. It removes the moment where willpower would be needed.

Is there evidence, or is this a tidy story? There is a fairly clean test. Ashraf, Karlan and Yin ran a field experiment with a rural bank in the Philippines in 2003 offering a product called SEED — an account paying the ordinary interest rate, with withdrawals blocked until a date or an amount the client chose themselves. No extra reward, purely a self-imposed restriction. Around a quarter of those offered it took it, and after a year their balances were roughly 80 per cent higher than the control group's. Notably, take-up was concentrated among clients whose survey answers showed exactly the reversing-preference pattern — which is the sophistication prediction, tested rather than assumed.

There is a second thing a lock can bind, and it is easy to miss because it is not you. Money that is visibly reachable is also reachable by everyone with a claim on you: a spouse, a relative in difficulty, a neighbour who knows you were paid on Friday. A rotating savings club — susu, tanda, chit fund, ROSCA — makes the contribution a social obligation and puts the fund beyond anyone's individual reach. Work by Anderson and Baland on Kenyan ROSCAs found participation concentrated among married women with income but limited say over household spending, which fits the reading that the device binds other people's claims as much as one's own. That interpretation is argued rather than settled, but the participation pattern itself is well documented.

Then the cost, which the framing so far has hidden. Illiquidity is not free, and its price is charged at the worst possible moment: the boiler fails in week nine of a fifty-week club. The saver either breaks the lock and pays the penalty, or borrows at a rate far above the interest they gave up. So the sensible amount of commitment is a dose, not a maximum — enough to protect a savings habit, not so much that an ordinary emergency has to be financed by credit. And there is a counterparty question too. When the lock works by handing your money to a company rather than by restricting your own account, you have acquired a new risk: the Farepak Christmas hamper scheme collapsed in 2006 owing tens of millions to savers who turned out to be unsecured creditors.

Which points at where the design has gone since. Save More Tomorrow, the scheme Thaler and Benartzi described in 2004, commits future pay rises to pension contributions, so the saver never experiences a fall in take-home pay. Automatic enrolment does the same job through a default rather than a lock. Both answer the same question — how do you bind the future self without punishing the present one?

Three things are worth pinning down.

First, this is not irrationality wearing a clever hat. The saver is making a perfectly coherent choice at the moment of signing up; the incoherence lies between that moment and the later one, and the lock is what stops the two from meeting.

Second, the penalty on early withdrawal is not meant to be collected. It exists to alter a decision that will now mostly not be taken — which is why a device that earns its provider a lot of penalty revenue is probably mis-sold rather than working well.

Third, the missing interest is a real cost, not a rhetorical one. A commitment product has to beat a flexible account plus the yield forgone, and it can only do that through the deposits it causes to exist at all. If someone would have saved the money anyway, the lock is pure loss.

• Whether commitment devices build a lasting habit or only work while the lock is on.
• How automatic enrolment achieves similar results through defaults, with no restriction at all.

Every term the collection defines is gathered in the glossary.