Present bias & under-saving
What these dates mean
Published is the original release. Updated records a material content change. Reviewed records the latest documented factual or editorial review; it does not mean personalized professional advice.
Why 'I'll save more next month' never happens — and why the fix is structural, not a bigger supply of willpower.
"I'll save more next month" is one of the most expensive sentences in personal finance. It rarely happens, and the reason is a well-documented quirk called present bias. This is a core spoke of the Money psychology hub.
What present bias is
Present bias (a form of hyperbolic discounting) is our tendency to weigh a reward now far more heavily than a bigger reward later. Future-you — the retiree who needs the money — is abstract and easy to short-change. Present-you, who wants the purchase today, is vivid and loud. So saving keeps losing to spending, one reasonable-feeling decision at a time.
That much is not seriously disputed. What is disputed is the story people usually tell about how to beat it.
What the replication crisis did to the willpower story
The standard advice says willpower is a limited resource that gets used up over the day — the "ego depletion" model. It is a memorable idea, and this page used to repeat it.
It did not survive testing. A pre-registered replication across 23 labs with roughly 2,100 participants found essentially nothing (Hagger et al., 2016). A second, larger multi-lab attempt with 3,531 participants across 36 labs found a pre-registered effect of d = 0.06 — not statistically distinguishable from zero, with the Bayesian analysis favouring the null about four to one (Vohs et al., 2021).
Take the honest lesson, not the cynical one. Self-control is real; people obviously differ in it and obviously find some periods harder than others. What collapsed is the specific mechanism — the tank that drains and must be refilled — and with it any plan whose logic is "budget your willpower carefully". You should not build a thirty-year savings plan on a laboratory effect that disappeared when people looked properly.
Here is where each claim in the usual advice actually stands:
| Claim you will hear | Evidence status | What it can support |
|---|---|---|
| Willpower is a fuel tank that depletes (ego depletion) | Failed two large pre-registered replications | Nothing. Do not build on it. |
| People discount the future steeply and inconsistently (present bias) | Well documented in lab and field | Deferring the sacrifice beats demanding it today |
| People will commit future raises to saving (Save More Tomorrow) | One well-known field study; the mechanism is now standard in auto-escalation features | Auto-escalation |
| Automatic enrolment raises participation | Robustly reproduced since Madrian & Shea (2001) | Use the default |
| Auto-enrolment raises long-run wealth as much as it raises participation | Contested — Choukhmane (2025) finds a far smaller long-run effect | Enrol, but do not stop at the default rate |
The finding that did hold up: defaults
Madrian and Shea (2001) studied a firm that switched to automatic 401(k) enrolment and found participation jumped from 37% to 86%. That result has been reproduced repeatedly, across firms and countries, and it is now the basis of US retirement policy. Inertia is powerful. The only question is which direction it points.
The honest footnote: participation is not the same as wealth. Choukhmane (2025) finds the long-run wealth effect of auto-enrolment is much smaller than the participation jump implies — many auto-enrolled workers would have joined a few years later anyway, and some offset the forced saving by saving less elsewhere. So enrol, yes. But do not mistake "I am enrolled at the default 3%" for "I am on track."
Structural fixes that work
The fix is not more discipline — it is removing the decision.
- Automate on payday. Auto-transfer to savings and investments the day you are paid, before the money feels spendable.
- Escalate with raises. Thaler and Benartzi's "Save More Tomorrow" worked because the sacrifice was deferred to a raise that had not arrived. That is the one trade present bias is happy to make. Set the auto-escalation and forget it.
- Use the default, then beat it. Enrolment gets you in the door. The contribution rate is what actually funds the retirement.
- Check the number, not the feeling. Run your rate through Am I on track? or the retirement checkup once a year, and change the standing instruction rather than the monthly intention.
Key takeaways
- Present bias is real and well documented: future-you loses to present-you by default.
- The "willpower is a depletable tank" story failed two large pre-registered replications. Do not build a plan on it.
- Defaults did replicate. Automate the transfer, auto-escalate with raises.
- Participation is not the same as being on track — the default contribution rate is usually too low.
Educational only — not financial advice.
FAQ
Why does 'I will save more next month' never happen?
Because present bias weighs a reward now far more heavily than a bigger reward later. Future-you is abstract; present-you is vivid and in the room. Next month arrives with exactly the same trade-off, and loses it exactly the same way. Nothing about the calendar changes the maths.
What actually fixes present bias?
Commit future money rather than current money. Escalating your contribution with your next raise works because the sacrifice is deferred, which is the one thing present bias does not object to. Automating the transfer on payday works because it removes the decision altogether.
Is willpower really a limited resource that runs out?
Probably not, or at least not in the way the popular story says. The 'ego depletion' finding failed two large pre-registered multi-lab replications (Hagger 2016; Vohs 2021). We are not saying willpower does not exist — we are saying you should not build a savings plan on a lab effect that vanished under scrutiny. The case for automation does not need it.
If ego depletion failed to replicate, why still automate?
Because automation's case rests on a finding that did replicate: defaults are powerful. Automatic enrolment reliably raises participation. And the structural argument stands on its own — a plan that needs one decision beats a plan that needs 480 of them, whatever the neuroscience turns out to say.
Does automatic enrolment actually make people richer?
It reliably raises participation. Whether it raises long-run wealth by as much is contested: Choukhmane (2025) finds the long-run effect is far smaller than the participation jump suggests, because some auto-enrolled workers would have joined anyway and later save less. Enrol — but do not treat the default contribution rate as sufficient.
What contribution rate should I automate?
More than the default. Plan defaults are often set at 3%, which is a starting point chosen for opt-out rates, not for your retirement. Pick a rate from your own numbers and set an annual auto-escalation, so the rate rises without you having to decide again.
Does present bias go away as you get older?
The trade-off gets easier because retirement stops being abstract, but the mechanism does not disappear. What changes is that a late saver has less time for compounding to cover the gap, which makes the structural fixes more urgent, not less.
Sources and notes
- Save More Tomorrow: Using Behavioral Economics to Increase Employee SavingThaler & Benartzi, Journal of Political Economy 112(S1), 2004 · Accessed 2026-07-11The SMarT program: employees who would not cut spending today committed future raises to saving. Used for the auto-escalation section, and for the caveat that this is one firm's field study.
- The Power of Suggestion: Inertia in 401(k) Participation and Savings BehaviorMadrian & Shea, Quarterly Journal of Economics 116(4), 2001 · Accessed 2026-07-11Automatic enrolment raised 401(k) participation from 37% to 86% at the firm studied. Used for the default-effect claim, which has been reproduced many times since.
- A Multilab Preregistered Replication of the Ego-Depletion EffectHagger et al., Perspectives on Psychological Science 11(4), 2016 · Accessed 2026-07-1123 labs, ~2,100 participants, pre-registered. Found essentially no ego-depletion effect. Used to retire the 'willpower is a depletable resource' claim.
- A Multisite Preregistered Paradigmatic Test of the Ego-Depletion EffectVohs et al., Psychological Science 32(10), 2021 · Accessed 2026-07-1136 labs, 3,531 participants. Pre-registered confirmatory result d = 0.06, not significant; Bayesian analysis favoured the null roughly 4 to 1.
- Default Options and Retirement Saving DynamicsChoukhmane, American Economic Review 115(11), 2025 · Accessed 2026-07-11Auto-enrolment's long-run effect on wealth is much smaller than the participation jump implies, because many auto-enrolled workers would have joined later anyway and offset by saving less afterwards. Used for the honest caveat on defaults.