---
name: Mental accounting
slug: mental-accounting
category: Decision making
categories: [Pricing, Onboarding, Retention]
evidence: moderate
page: /principles/mental-accounting/
updated: 2026-10-26
---

# Mental accounting

> People keep money in separate mental pots, so the same price feels affordable or extravagant depending on which pot it lands in.

## What it is

Money is treated as earmarked rather than interchangeable: a household budgets for fuel, groceries and entertainment separately, and a cost charged against a budget with room in it is accepted while the identical cost charged against an exhausted one is refused.

## Evidence

Thaler (1985) described consumers as running an internal accounting system that codes each purchase against a category and a reference point, and Thaler (1999) collected the demonstrations into seventeen numbered problems. So a pricing surface decides two things beyond the amount: which of the user's existing budgets the charge lands in, and whether the total stays legible while you do it. Earmarking holds up in the field, where the money leaves a paper trail. Hastings and Shapiro (2018) estimate a marginal propensity to consume SNAP-eligible food of 0.5 to 0.6 out of SNAP benefits, against a much smaller propensity out of cash. Beatty, Blow, Crossley and O'Dea (2014) found households spending 47 percent of the UK Winter Fuel Payment on fuel, where treating it as ordinary cash predicts 3 percent. In Abeler and Marklein's restaurant experiment an €8 voucher labelled for beverages raised per-person beverage spending by €3.90 over an €8 voucher for the whole bill, and in their lab version subjects put 57 percent of a labelled subsidy on the targeted good against 28 percent of a cash one, driven by subjects with weaker numeracy.

Hastings and Shapiro (2013) give the magnitude and the boundary. As gasoline went from $2.98 to $4.10 a gallon in the 2008 spike, the share of transactions going to regular grade rose 1.4 percentage points, from 80.2 to 81.6 percent. The spike cost a typical household about $1,313 a year, which the cross-sectional income relationship puts at 0.02 points of regular share; reading the shift as an income effect needs income to fall by almost $100,000. Households respond almost 20 times more to income lost at the pump than to the same income lost elsewhere. The placebo is the part to design from: in the same households' orange juice and milk purchases, gasoline money and other money were fungible. Fungibility breaks inside the category being spent from, so a frame moves nothing unless your price sits in the category you named.

The lab paradigms replicate less cleanly. Li and Feldman (2025) replicated all seventeen problems in Thaler (1999) as a registered report, N = 1,007: eleven supported, three mixed, three unsupported. The casualty is the one people cite most — the calculator-and-jacket problem, drive twenty minutes to save $5 on a $15 item but not on a $125 item, did not hold (V = 0.07, p = 0.066, against 0.38 originally) — and both house-money results reversed sign (h = −0.46 and −0.55). Survivors shrank: the theatre-ticket sunk-cost problem to V = 0.17 from 0.44, the waiting-time problem to d = 0.23 from 0.48. The budgeting problem split: a prior dinner reduced willingness to buy a theatre ticket as predicted (d ≈ 0.26 to 0.28), a prior flu vaccination did nothing (d ≈ 0.04 to 0.05). Whether your charge counts as the same category as the last one does the work, and you cannot tell from outside which pairs will count.

The label cuts both ways. Heath and Soll (1996) found that consumers do set budgets and that budgeting can produce underconsumption, more strongly for purchases highly typical of the category: name a category the user has exhausted and the label argues against the purchase. Milkman and Beshears (2009) followed 4,435 grocery customers redeeming a $10-off coupon and saw spending rise by just $1.59, concentrated in groceries the customer does not usually buy — a windfall goes to a loose account, not the main one. Sussman and Alter (2012) show where the categories leak: forecasting a week ahead, participants predicted $39 of exceptional spending and spent $70, while predicting ordinary spending almost exactly ($83 against $81). Each unusual purchase is filed as a one-off, so the series never gets budgeted for.

The per-day framing in every pricing deck is the weakest-evidenced piece. Gourville (1998) named it temporal reframing and put the mechanism in retrieval: a pennies-a-day frame brings small ongoing expenses to mind as the comparison, an aggregate frame brings large infrequent ones. Atlas and Bartels (2018) added the benefit side across eight lab experiments and one field test, arguing a periodic price makes people imagine many small benefits rather than one lump. Their field test is the only field result for this half: of 15,127 visitors to a prepared-meal delivery site shown a landing banner quoting either $16 per day or $99 per week, first-time visitors converted at 1.3 percent against 0.7 percent — 81 transactions against 47, a 77 percent lift — with the daily frame priced 13 percent higher. Across the whole sample including returning visitors the lift was 19 percent at p < .10, over five weeks on one site before the partner stopped it. Their own moderator is the caution: periodic pricing raised donation intentions only under high affective involvement, and in the low-involvement cell the effect was exactly zero.

Hiding the total is the version that reliably makes money, which is why it is increasingly illegal. Abraham and Hamilton (2018) meta-analysed partitioned pricing across 149 observations from 43 studies and found consumers preferring a split price principally when the total price is absent, with perceived benefit of the surcharge and how typical partitioning is in that category as the robust moderators. Blake, Moshary, Sweeney and Tadelis (2021) ran the field version on StubHub: shrouding a 15 percent fee until checkout raised revenue about 21 percent, transactions 14.1 percent and average price per seat 5.73 percent, with upgrades accounting for about 28 percent of the gain. Across 99 Major League Soccer matches buyers were 13 percent less likely to purchase when fees were shown upfront (p = 0.076), and the effect persisted among users with ten or more visits. Under 16 CFR Part 464, effective 12 May 2025, live-event ticketing and short-term lodging must disclose the total price clearly, conspicuously and more prominently than other pricing information; in the EU, Article 6(1)(e) of Directive 2011/83/EU requires the total price inclusive of taxes before the consumer is bound, and for subscriptions “the total price shall include the total costs per billing period,” plus total monthly costs at a fixed rate.

Grade the two halves separately. That people treat money as earmarked is supported by independent large-scale field work on benefits, transfers, fuel and vouchers, and mostly survived a preregistered replication of the founding paradigms. That you can move your own conversion by relabelling a price into a different budget rests on one five-week field test with a known moderator, a meta-analysis whose main lever is the absence of the total, and a lot of hypothetical choice. Treat the category as real and your framing as an experiment.

## When it holds

- The spending decision sits inside the category you labelled. Hastings and Shapiro (2013) reject fungibility for gasoline grade choice but cannot reject it for the same households’ orange juice and milk, so a frame naming a category the price does not belong to changes nothing.
- The comparison expense is close enough to count as the same category. In Li and Feldman’s (2025) replication a prior dinner suppressed theatre-ticket purchase as predicted (d ≈ 0.26 to 0.28) while a prior flu vaccination did not (d ≈ 0.04 to 0.05).
- The user is seeing you for the first time. Atlas and Bartels’ (2018) banner test lifted first-time visitors 77 percent and the whole sample only 19 percent at p < .10.
- The benefit accumulates in units the user can picture and feel rather than calculate. Periodic pricing raised donation intentions only in the high-involvement cell; with low involvement the effect was exactly zero.
- The surcharge reads as buying something, and split pricing is normal in that category. Those were Abraham and Hamilton’s (2018) two robust moderators, alongside absence of the total price — the condition you should not be using.
- Weak numeracy amplifies it. In Abeler and Marklein’s lab experiment the label effect came from subjects with weaker mathematical ability, so a frame that works on consumers may do nothing to a procurement team.
- It reverses when the budget is already spent. Heath and Soll (1996) found budgeting producing underconsumption, strongest for purchases most typical of the category, so naming an exhausted category argues against your own price.
- Three textbook demonstrations did not survive replication: the calculator-and-jacket problem (V = 0.07) and both house-money problems, which reversed sign (h = −0.46 and −0.55). Do not build a flow on those.

## How to apply

- Give the total at least the visual weight of the per-day, per-seat or per-unit figure. This is a legal floor rather than a style preference: 16 CFR § 464.2 requires the total price to be disclosed more prominently than other pricing information in covered sectors, and Article 6(1)(e) of Directive 2011/83/EU requires the total costs per billing period for subscriptions.
- Choose the denominator from a budget the user already keeps, and name it. “Less than your weekly coffee” and “$12 per seat per month” land in different accounts than “$3,600 a year”, which is what a finance approver will compute anyway.
- Put the periodic figure on acquisition surfaces and the aggregate figure on account and renewal surfaces. Returning users already have a category for you, and re-anchoring them on a daily number hides what they are renewing.
- Pair the per-day price with the per-day thing the user gets, not with an annual feature list, so the benefits are as countable as the price.
- State the renewal as the amount that will leave the account, on the date it leaves, in the billing period’s own units. A plan sold per day and renewed per year is where disputes come from.
- Split seat count from price per seat in the cart and recompute the total live as the count changes, instead of revealing it at checkout.
- Aim a trial credit or coupon at a feature the user has not tried, not at the core plan — extra spending from a windfall goes to things the customer does not usually buy.
- In a budgeting or expense tool, total irregular spending as a category across the year rather than itemising each surprise, because users forecast the instance and not the series.

## What to avoid

- Quoting a per-day price with no total anywhere on the surface. That absence is the one condition under which the partitioned-pricing meta-analysis finds a reliable advantage, which makes it a statement about what the user cannot see rather than about value.
- Drip-feeding fees to checkout. Shrouding a 15 percent fee raised StubHub revenue about 21 percent, and under 16 CFR Part 464, effective 12 May 2025, doing it in live-event ticketing or short-term lodging is an unfair and deceptive practice carrying civil penalties.
- Pricing the daily frame higher than the aggregate one and letting the frame carry the difference. Atlas and Bartels’ winning banner was 13 percent more expensive, so copying it commercially means charging more for being harder to total.
- Inventing the budget you compare against. A “cheaper than your daily commute” claim needs a real figure for that user, or it is a fabricated reference price.
- Splitting one charge across several currencies, credits or passes until the category is lost. A price the user cannot assign to a budget is a price they cannot refuse on the merits.
- Relabelling a renewal to keep it out of the user’s subscriptions budget — renaming it a “membership”, moving it to a different billing day, or splitting one annual charge into twelve unannounced ones.
- Shipping a lab result as your result. The classic topical-account demonstration failed to replicate at V = 0.07, and the per-day field evidence is one five-week test on one site.

## Measuring it

- Primary: conversion and revenue per visitor against the aggregate frame, split by new and returning users. The only field estimate splits exactly there — 77 percent for first-time visitors against 19 percent overall at p < .10 — so a pooled read will understate or invent the effect.
- Secondary: revenue at 90 days net of refunds, plus average contract value or seat count, which is where the upgrade half shows up. On StubHub price per seat moved 5.73 percent and accounted for about 28 percent of the revenue gain.
- Comprehension check before launch: ask a sample of users what the plan costs per year and record the error rate per arm. If the periodic arm cannot state the total, you built the shrouded condition rather than the framing one.
- Backfire metrics for the awareness the frame removed: refund rate, chargeback and dispute rate, “what is this charge?” contacts, and cancellations inside the first billing period.
- Backfire metrics for the budget you named: downgrade and non-renewal rate in that category, and win rate and sales-cycle length on per-seat deals where the buyer totals the year anyway. Heath and Soll’s budgeting effect runs toward underconsumption, strongest for the purchases most typical of the category you invoked.

## Live examples

- Instacart leads the annual plan with its $99 total and prints the derived ($8.25 / month) underneath it, so the yearly commitment never appears as a monthly-sized number alone.
- Rocket Money sets each post-trial price as a monthly figure with the per-day equivalent in smaller type under it — $10/mo against ($0.33/day) — so the charge can be weighed against a daily budget without the monthly amount leaving the screen.
- WHOOP splits the number of tests from the price per test — four frequencies priced from US$199/test down to US$150/test — and puts the US$349 the chosen plan costs for the year in the largest type on the card.
- Deezer states the renewal as a date and an amount, 02/05/2024 and SGD 21.90, on the same screen as the cancel control, so the figure under review is the one that will leave the account.
- DoorDash lists the $3.00 service fee, $0.83 estimated tax and US$1,00 tip under the $8.38 subtotal and totals them at US$13,21 on the order screen, so the parts and the sum arrive before the order is placed rather than after.

## References

- Thaler (1985). Mental Accounting and Consumer Choice. Marketing Science. — https://doi.org/10.1287/mksc.4.3.199
- Thaler (1999). Mental Accounting Matters. Journal of Behavioral Decision Making. — https://doi.org/10.1002/(SICI)1099-0771(199909)12:3<183::AID-BDM318>3.0.CO;2-F
- Li & Feldman (2025). Revisiting Mental Accounting Classic Paradigms: Replication Registered Report of the Problems Reviewed in Thaler (1999). Royal Society Open Science. — https://doi.org/10.1098/rsos.250979
- Hastings & Shapiro (2018). How Are SNAP Benefits Spent? Evidence from a Retail Panel. American Economic Review. — https://doi.org/10.1257/aer.20170866
- Hastings & Shapiro (2013). Fungibility and Consumer Choice: Evidence from Commodity Price Shocks. Quarterly Journal of Economics. — https://doi.org/10.1093/qje/qjt018
- Hastings & Shapiro (2012). Mental Accounting and Consumer Choice: Evidence from Commodity Price Shocks. NBER Working Paper 18248. — https://www.nber.org/papers/w18248
- Beatty, Blow, Crossley & O'Dea (2014). Cash by Any Other Name? Evidence on Labeling from the UK Winter Fuel Payment. Journal of Public Economics. — https://doi.org/10.1016/j.jpubeco.2014.06.007
- Abeler & Marklein (2017). Fungibility, Labels, and Consumption. Journal of the European Economic Association. — https://doi.org/10.1093/jeea/jvw007
- Abeler & Marklein (2008). Fungibility, Labels, and Consumption. IZA Discussion Paper 3500. — https://docs.iza.org/dp3500.pdf
- Heath & Soll (1996). Mental Budgeting and Consumer Decisions. Journal of Consumer Research. — https://doi.org/10.1086/209465
- Milkman & Beshears (2009). Mental Accounting and Small Windfalls: Evidence from an Online Grocer. Journal of Economic Behavior & Organization. — https://doi.org/10.1016/j.jebo.2009.04.007
- Sussman & Alter (2012). The Exception Is the Rule: Underestimating and Overspending on Exceptional Expenses. Journal of Consumer Research. — https://doi.org/10.1086/665833
- Gourville (1998). Pennies-a-Day: The Effect of Temporal Reframing on Transaction Evaluation. Journal of Consumer Research. — https://doi.org/10.1086/209517
- Atlas & Bartels (2018). Periodic Pricing and Perceived Contract Benefits. Journal of Consumer Research. — https://doi.org/10.1093/jcr/ucx127
- Abraham & Hamilton (2018). When Does Partitioned Pricing Lead to More Favorable Consumer Preferences? Meta-Analytic Evidence. Journal of Marketing Research. — https://doi.org/10.1177/0022243718800724
- Abraham (2015). A Multi-Method Examination of Partitioned Pricing. PhD dissertation, University of Maryland. — https://drum.lib.umd.edu/handle/1903/16560
- Blake, Moshary, Sweeney & Tadelis (2021). Price Salience and Product Choice. Marketing Science. — https://doi.org/10.1287/mksc.2020.1261
- Blake, Moshary, Sweeney & Tadelis (2018). Price Salience and Product Choice. NBER Working Paper 25186. — https://www.nber.org/papers/w25186
- Federal Trade Commission (2025). Rule on Unfair or Deceptive Fees, 16 CFR Part 464. 90 Federal Register 2066, effective 12 May 2025. — https://www.govinfo.gov/content/pkg/FR-2025-01-10/html/2024-30293.htm
- Directive 2011/83/EU (Consumer Rights Directive), Article 6 — information requirements, including total price and total costs per billing period. — https://www.legislation.gov.uk/eudr/2011/83/article/6

## Similar principles

- Framing effect — /agents/framing-effect.md
- Pain of paying — /agents/pain-of-paying.md
- Anchoring — /agents/anchoring.md
- Left-digit effect — /agents/left-digit-effect.md
- Sunk cost fallacy — /agents/sunk-cost-fallacy.md
- Hyperbolic discounting — /agents/hyperbolic-discounting.md