Commitment Devices: Make Backing Out Harder
Commitment devices make backing out of a goal costly or impossible. What the research shows, the catches most guides skip, and how to set one up.
Goal4
Habit Tracker & Streaks
Track the habit as you build it
Goal4 logs a habit with one tap and keeps your streak.
Free · No sign-up needed
Disclaimer: This article is for informational purposes only and is not medical, financial or psychological advice. Individual results vary; consult a healthcare professional for personalized guidance.
On Sunday night, you mean it. Three runs this week, no snacks after dinner, the phone stays out of the bedroom. By Wednesday morning a different person is in charge, and that person has excellent reasons to start next week instead.
A commitment device is how Sunday-night you gets a vote on Wednesday. You make a choice now, while you’re calm, that makes backing out later harder, costlier or impossible. Odysseus had his crew tie him to the mast so he could hear the Sirens without steering onto the rocks.
The modern versions are less nautical: a savings account you can’t touch, a contract with money on the line, a friend who checks. The research behind them is better than most habit advice can claim. It’s also more mixed than most articles let on. This guide covers what they are, the three ways to build one, what the studies found and how to set one up for a single habit.
Key Takeaways
- A commitment device is a choice you make in advance to bind your future self. Hard ones carry actual costs (money, a removed option). Soft ones carry only psychological costs (a public promise).
- There are three basic moves: remove the option, attach a cost or add a witness.
- The wins are real but modest. Smokers offered a forfeit-to-charity savings contract were 3 percentage points more likely to pass a nicotine test at 6 months, and the effect held at 12 months.
- Most people don’t sign up, and many who do still fail. Take-up was 11% in that smoking study. In a separate savings study, 55% of clients defaulted and lost money.
- A tracker is the record, not the device. It makes a commitment checkable. The stake, the lock or the referee has to live somewhere else.
What Is a Commitment Device?
Economists Bryan, Karlan and Nelson define a commitment device as “any arrangement, entered into by an individual, with the aim of making it easier to fulfill his or her own future plans.” You decide now, and that decision makes the later choice harder to reverse.
The idea is sometimes called a Ulysses pact, after Odysseus’s Latin name. It also has a game-theory side: Thomas Schelling, who shared the 2005 Nobel Prize in economics, wrote a 1984 paper on what he called “self-command.”
Why we need them: present bias
Planner-you, on Sunday, weighs the long-term payoff of a run. Doer-you, standing in a cold kitchen at 6 a.m., weighs the next ten minutes.
Behavioral economists call this present bias, or time inconsistency. A reward that looks worth waiting for from a distance loses to whatever is on offer right now. A commitment device lets the planner set terms the doer has to live with.
Hard vs soft commitments
Bryan, Karlan and Nelson split commitment devices into two kinds:
- Hard commitments carry actual costs: money at stake, or an option that’s gone.
- Soft commitments carry only psychological costs: a public promise, or a rule that clashes with how you see yourself.
Soft doesn’t mean useless. Their review argues that soft commitments in particular can help, both in explaining how people behave and in designing commitment devices that work.
One more condition: a commitment device is voluntary, and you set it up in advance. A rule someone else imposes on you may change what you do, but it isn’t one.
Three Ways to Bind Your Future Self
Most commitment devices do one of three things. Some do two at once.
1. Remove the option (hard)
The strongest move is to make the bad choice unavailable.
- A locked savings account. The SEED account at a bank in the Philippines let clients set a goal date or amount, then blocked withdrawals until they reached it.
- Put it out of reach. Leave the credit card at home, don’t buy the snack, hand someone your car keys, or put your phone in a lockbox.
- Shut yourself away. Atomic Habits tells of Victor Hugo isolating himself to finish The Hunchback of Notre-Dame.
Breaking a habit rather than building one? This is often the simplest place to start. See how to break bad habits for more ways to take the option off the table.
2. Attach a cost (hard)
When you can’t remove the option, make taking it expensive.
- A stickK contract. stickK calls itself “a free goal-setting platform created by behavioral economists at Yale University,” and Dean Karlan, a co-author of the savings and smoking studies below, is a co-founder. Its FAQ has set the minimum stake at $5 per reporting period, with lost money going to a friend or foe, a charity, or an “anti-charity” you’d hate to fund. Check the current terms before you sign up.
- A Beeminder goal. You log data against a “bright red line,” and if you end a day on the wrong side, you “derail” and pay. The pledge starts at $0 (on a freebee) or $5 and can climb with each derail: $5, $10, $30, $90, $270, $810. Its “akrasia horizon” means you can’t make a goal easier any sooner than 7 days out.
- A deposit you forfeit. In the CARES study, smokers in the Philippines deposited money for six months, then took a urine test for nicotine and cotinine. Fail, and the money was “forfeited to charity.”
- A bet with a friend. Cheap, and only as strong as the friend’s willingness to collect.
- A contract that only punishes. Kaur, Kremer and Mullainathan found that full-time data entry workers “increase output by voluntarily choosing dominated contracts (which penalize low output but give no additional rewards for high output).”
3. Add a witness (soft, or hard if they enforce a stake)
Sometimes the cost is just someone else knowing.
- A public pledge. Tell your team, post it, stick it on the fridge.
- A signed habit contract. The habit contract generator builds a paper contract with a partner, ready to print and sign.
- A referee. On stickK, a referee confirms each report and “may choose to take your word for it, or ask you for proof.” With no referee, the contract runs “On Your Honor.”
Witnesses have some evidence behind them. In a review of non-financial commitment devices for weight loss, Coupe and colleagues found that commitments “appeared most successful when made publicly.”
Bonus: decide now for later
Save More Tomorrow, from Richard Thaler and Shlomo Benartzi, has employees commit in advance to save part of their future raises. In their 2004 study, 78% of those offered the plan joined, and joiners’ average saving rates rose from 3.5% to 13.6% over 40 months.
It’s a soft device. Participants can opt out, and in Benartzi and Thaler’s own words, the design “uses inertia.”
A gym membership is a weak device on its own
In DellaVigna and Malmendier’s study of 7,752 gym members, people on a flat monthly fee “of over $70 attend on average 4.3 times per month.” The authors calculate that they paid more than $17 per expected visit, when a 10-visit pass would have cost $10 a visit.
Prepaying doesn’t remove the option to skip. The money is gone either way, so skipping costs you nothing extra.
Commitment Devices vs Temptation Bundling, Accountability Partners and Environment Design
These tactics get lumped together, and they do overlap. What separates them is whether backing out costs you anything.
| Tactic | What it changes | When it counts as a commitment device | Example |
|---|---|---|---|
| Temptation bundling | Adds an immediate reward to the should-do | When the want is locked to the should, so you can’t get it any other way | Audiobooks you can only play at the gym |
| Accountability partner | Adds a witness and support | When they act as the referee who enforces a stake you agreed on | A friend who collects $10 for each missed run |
| Environment design | Adds friction to the bad choice | When undoing it is costly or impossible, and you decided it in advance | A timed phone lockbox, not just a phone in another room |
| If-then plans | Decides what you’ll do when a cue hits | Not on its own: skipping costs nothing | “If it’s 6 p.m., I put on my running shoes” |
Temptation bundling becomes a commitment device when the want is locked to the should. In Katy Milkman’s gym study, audiobooks available only at the gym initially led to 51% more gym visits, though the effect faded over time. Afterwards, “61% of participants opted to pay to have gym-only access to iPods containing tempting audiobooks, suggesting demand for this commitment device.”
An accountability partner is a witness and a source of support. They become part of a commitment device once they’re the referee who enforces a stake.
Environment design adds friction you can still undo: the phone in the other room is a short walk away. Friction is often enough, and it’s cheaper. A lock is for the habits where friction keeps losing.
They stack. A temptation bundle plus a $10 stake that your partner referees covers reward, witness and cost in one setup.
Do Commitment Devices Actually Work?
Yes, for some people in some areas, with modest average effects and real catches. The wins come first, then the catches.
The smoking and weight-loss studies below are research results, not treatment advice. Talk to a doctor before using a commitment device to quit smoking or lose weight.
The wins
- Savings. 710 bank clients in the Philippines were offered the SEED account, and 202 (28.4%) opened one. “After twelve months, average savings balances increased by 81 percentage points” for clients assigned to the treatment group, relative to the control group. That figure counts everyone assigned to get the offer, not only those who opened an account (Ashraf, Karlan and Yin, 2006).
- Smoking. 11% of smokers offered CARES took it up. Those randomly offered it were 3 percentage points more likely to pass the 6-month test than the control group, and the effect “persisted in surprise tests at 12 months” (Giné, Karlan and Zinman, 2010).
- Exercise. At a Fortune 500 company, a four-week gym incentive alone “generated only small lasting effects.” Offering a self-funded commitment contract at its end produced “significant long-run changes, detectable even several years after the incentive ended” (Royer, Stehr and Sydnor, 2015). A gym study of 1,248 people by Carrera and colleagues also found high take-up of contracts for more gym visits, and significant increases in exercise.
- Weight loss, without money. Coupe and colleagues’ review left out financial devices and included 10 of 2,675 studies. Across three trials with 409 participants, adding a commitment device to a weight-loss program increased short-term weight loss by a mean 1.5 kg (95% CI 0.7 to 2.4). Two trials with 302 patients suggested the benefit held at 12 months (mean difference 1.7 kg; 95% CI 0.0 to 3.4). The authors call the evidence “limited and of variable quality.”
Catch 1: most people don’t sign up
Take-up was 11% for CARES and 28.4% for SEED. In a smoking-cessation trial with 2,538 participants, Halpern and colleagues found that 90.0% accepted reward-based programs, but only 13.7% accepted deposit-based ones (which paired a refundable $150 deposit with $650 in rewards).
Counting everyone assigned, rewards did better: 15.7% sustained abstinence at six months, against 10.2% for deposits. But among people who would have accepted either, deposit-based programs came out 13.2 percentage points higher.
A stake can work for people willing to put one up. Getting them to is the hard part.
Catch 2: many who commit still fail
J-PAL’s summary of CARES reports that “a surprisingly large proportion (66 percent) of smokers who voluntarily committed to CARES ended up failing to quit.”
In John’s (2020) savings study, low-income clients in the Philippines designed their own commitment plan and chose their own default penalty. “While the average effect on bank savings is large, 55% of clients default and incur monetary losses.” The author suggests the penalties they chose were too low.
Catch 3: choosing commitment is noisy
Carrera and colleagues found that take-up was “influenced both by noisy valuation and incorrect beliefs about one’s time inconsistency.” About half of the people who took a contract for more gym attendance also took one for less.
Their estimates imply that offering the contracts “lowers consumer surplus.” In plain terms, people aren’t great at judging which commitments they need.
How to Set Up a Commitment Device for One Habit
Start with one habit, not your whole life. Six steps:
- Name the exact moment you fail. “I don’t exercise” is too vague to act on. “I turn off the 6 a.m. alarm” is a moment, and that’s where the device has to bite. Habit triggers and cues will help you pin it down.
- Match the mechanism to the failure. A temptation in the house: remove it. A choice you can always make: attach a cost. A habit nobody notices you skipping: add a witness.
- Write a bright-line rule with a number and a period. “Exercise more” can’t be checked. “Run 3 times a week, Sunday to Saturday” can.
- Size the stake. It should sting, but it should also be money you can afford and would actually hand over. Never pick one that harms your health, your finances or another person. Too low, and John (2020) suggests you’ll default. Too high, and you’ll be tempted to hide misses. The bad habit cost calculator shows what a bad habit already costs you.
- Pick a referee who checks, or decide honestly to go “on your honor.” Accountability partners for habits covers choosing and briefing one.
- Decide in advance what one miss means, and set a review date. Does one missed week cost the stake, or only two in a row? The never miss twice rule is a sensible default.
For money held by a third party, stickK and Beeminder handle the stakes.
Keep a record your commitment can be checked against
A stake with no record is a bet nobody can settle. A habit tracker doesn’t bind you to anything, but it can make the commitment visible and checkable.
In Goal4, mirror the contract: “Run 3 times a week” becomes a Weekly habit with a goal of 3, and Goal4 weeks run Sunday to Saturday. The streak counts complete weeks, shows “(ending)” when this week isn’t done yet but last week was, and resets to 0 after a whole missed week. The contribution grid shows recent weeks at a glance, and the monthly calendar lets you page back through each month of the contract.
On check-in day, share the habit’s card as an image through your phone’s share sheet and send it to your referee yourself. On a Sunday the card’s progress has already moved on to the new week, so your referee should judge last week from the grid and the streak. Log any payment in that day’s “Add a note…” field (“paid Sam $10”).
Build in slack on purpose. Goal4 has no skip days or streak freezes. If you want room for bad days, write it into the goal: a weekly goal of 5 rather than a daily habit. Habit streaks covers when a streak helps and when it backfires.
Know the limits, and tell your referee. Any past day can be edited, so the card is a self-reported snapshot, not verification. Goal4 holds no money, has no referee or partner features, doesn’t block apps and doesn’t connect to stickK or Beeminder.
Reminders with quiet days and a home screen widget keep the habit in view, but they add no cost to skipping. There’s no account, and habit data stays on your device.
Troubleshooting and Tips
- You keep paying and nothing changes. The stake is too small or the rule is too big. Change one of them, not both at once.
- You’re tempted to fudge the record. The stake is probably too high. Lower it until paying feels better than lying to your referee.
- You can’t bring yourself to sign up. You’re in good company: CARES take-up was 11%. Start soft, with a public pledge or a paper contract that has no money in it.
- Your referee lets everything slide. Pick someone else, or let a service hold the money.
- The contract is ending. Renew it, lower the stake, or drop it and keep tracking to see whether the habit holds on its own.
Related Reading
- Temptation Bundling: The Science-Backed Habit Hack: Milkman’s gym study in depth, and how to make a bundle strict enough to work.
- Accountability Partners for Habits: Making It Work: How to pick, brief and check in with the person who keeps you honest.
- Environment Design for Habits: A Room-by-Room Guide: Friction you can add today, before you reach for a lock.
References
- Bryan, Karlan & Nelson (2009). “Commitment Contracts.” Yale Economic Growth Center Working Paper No. 980.: The source of the definition quoted here and of the split between soft commitments (psychological costs only) and hard commitments (actual costs).
- Bryan, Karlan & Nelson (2010). “Commitment Devices.” Annual Review of Economics 2(1): 671-698.: The published review, which distinguishes hard and soft commitments and shows how soft commitments can help.
- Ashraf, Karlan & Yin (2006). “Tying Odysseus to the Mast.” Quarterly Journal of Economics 121(2): 635-672.: The SEED commitment savings study, with an 81 percentage point rise in average savings balances after twelve months for the group assigned to be offered the account.
- J-PAL. Commitment Savings Products in the Philippines.: Product details for the SEED account, including the goal date or amount clients chose before they could withdraw.
- Giné, Karlan & Zinman (2010). “Put Your Money Where Your Butt Is.” American Economic Journal: Applied Economics 2(4): 213-235.: The CARES smoking contract: 11% take-up, a 3 percentage point rise in passing the 6-month test, and an effect that persisted in surprise tests at 12 months.
- J-PAL. CARES: Commitment Savings for Smoking Cessation in the Philippines.: J-PAL’s summary, including the finding that 66 percent of smokers who committed to CARES failed to quit.
- Royer, Stehr & Sydnor (2015). American Economic Journal: Applied Economics 7(3): 51-84.: A workplace field experiment in which a self-funded commitment contract produced exercise changes detectable several years later.
- Carrera, Royer, Stehr, Sydnor & Taubinsky. “Who Chooses Commitment? Evidence and Welfare Implications.” NBER Working Paper 26161.: A gym experiment with 1,248 participants on who takes up commitment contracts and why, later published in the Review of Economic Studies.
- Halpern et al. (2015). New England Journal of Medicine 372(22): 2108-17.: A trial of four financial-incentive programs for smoking cessation in 2,538 participants, with 90.0% acceptance for reward-based programs and 13.7% for deposit-based ones.
- Coupe, Peters, Rhodes & Cotterill (2019). BMC Public Health 19: 816.: A systematic review of non-financial commitment devices in weight-loss interventions, finding a modest added short-term weight loss and stronger results for public commitments.
- John (2020). “When Commitment Fails: Evidence from a Field Experiment.” Management Science 66(2): 503-529.: A savings experiment in which 55% of clients defaulted on self-designed commitment plans and lost money.
- Milkman, Minson & Volpp (2014). “Holding the Hunger Games Hostage at the Gym.” Management Science 60(2): 283-299.: The temptation bundling study, with an initial 51% rise in gym visits that faded over time, and 61% of participants later paying for gym-only access.
- Kaur, Kremer & Mullainathan (2015). “Self-Control at Work.” Journal of Political Economy 123(6): 1227-1277.: A year-long field experiment in which data entry workers raised output by choosing contracts that only penalized low output.
- Thaler & Benartzi (2004). “Save More Tomorrow: Using Behavioral Economics to Increase Employee Saving.” Journal of Political Economy 112(S1): S164-S187.: The program in which employees commit in advance to save part of future raises, with joiners’ saving rates rising from 3.5% to 13.6% over 40 months.
- Benartzi & Thaler (2013). “Behavioral Economics and the Retirement Savings Crisis.” Science 339: 1152-1153.: The description of Save More Tomorrow as a plan that uses inertia, with participants free to opt out.
- DellaVigna & Malmendier (2006). “Paying Not to Go to the Gym.” American Economic Review 96(3): 694-719.: A study of 7,752 gym members finding that flat-fee members paid more than $17 per expected visit when $10 per visit was available.
- stickK. FAQ: Stakes.: stickK’s own description of stakes, recipients and its minimum stake per reporting period. See also its referee FAQ.
- Beeminder. Glossary.: Definitions of the bright red line, derailing and the pledge schedule. See also the akrasia horizon.
- Schelling (1984). “Self-Command in Practice, in Policy, and in a Theory of Rational Choice.” American Economic Review 74(2): 1-11.: Schelling’s paper on self-command. His 2005 Nobel Prize is summarized on the Nobel Prize site.
Goal4
Habit Tracker & Streaks
Start with one habit
Name a habit, pick a daily, weekly or monthly goal, and tap once to log it. Goal4 keeps the streak and the grid.
Free · No sign-up needed
Frequently Asked Questions
What is a commitment device?
A commitment device is a choice you make now that makes backing out of a goal later harder, costlier or impossible. Economists Bryan, Karlan and Nelson define it as 'any arrangement, entered into by an individual, with the aim of making it easier to fulfill his or her own future plans.' The classic image is Odysseus, tied to the mast so he could hear the Sirens without steering toward them.
What are examples of commitment devices?
A savings account that blocks withdrawals until a goal date or amount, a stickK contract with money at stake and a referee, a Beeminder pledge that can rise each time you derail, a bet with a friend, leaving your credit card at home, a phone lockbox, a public pledge, or a signed habit contract with a partner.
Do commitment devices actually work?
For some people and some goals, yes, with modest average effects. Smokers offered a forfeit-to-charity savings contract were 3 percentage points more likely to pass a nicotine test at 6 months than a control group, and the effect held at 12 months. But only 11% of those offered signed up, and J-PAL's summary reports that 66% of those who committed still failed to quit. This is not medical advice.
What is the difference between hard and soft commitment devices?
Hard commitments carry actual costs, such as money you lose or an option you can no longer take. Soft commitments carry only psychological costs, such as breaking a public promise or acting against your self-image. The split comes from economists Bryan, Karlan and Nelson, who argue soft commitments can be useful too.
Is temptation bundling a commitment device?
Only when the want is locked to the should, so you can't get the reward any other way. In Katy Milkman's gym study, audiobooks available only at the gym initially led to 51% more gym visits, though the effect faded over time, and afterwards 61% of participants opted to pay for gym-only access. The paper calls that 'this commitment device.' A bundle you can break on the couch is just a pairing.
Is an accountability partner a commitment device?
On their own, a partner is a witness and a source of support. They become part of a commitment device when they act as the referee who enforces a stake you agreed on in advance. That is how stickK's referees work: they confirm each report and can ask for proof.
Is a gym membership a commitment device?
A weak one at best. In a study of 7,752 gym members, people on flat monthly fees of over $70 went 4.3 times a month on average, which the authors calculate as more than $17 per expected visit, when a $10 per-visit option was available. The money is gone whether you go or not, so skipping costs you nothing extra.
Can commitment devices backfire?
Yes. In one field experiment, 55% of low-income savers defaulted on a commitment plan they designed themselves and lost money, and the author suggests the penalties they chose were too low. Another study found that about half the people who took a contract for more gym visits also took one for fewer. Size the stake so you would actually pay it, and never risk your health, your finances or another person.
Can a habit tracker be a commitment device?
Not on its own. A tracker records whether you kept a commitment, but it doesn't make backing out cost anything. Pair it with a device set up elsewhere, such as a stickK contract or a paper contract with a referee, and use the tracker as the record you both check against. If the tracker lets you edit past days, what you share is your own report, not verification.