Pain of paying

The construct comes out of mental accounting. Prelec and Loewenstein (1998) modelled a purchase as two linked accounts, payment and consumption each partly carrying the other: people prefer money coupled to a benefit they can feel. In the MIT OpenCourseWare notes that teach the paper, 63 percent of respondents would rather pay for a Caribbean holiday in six instalments before the trip than in six identical instalments after it, while only 24 percent want to prepay a washer-dryer on the same schedule. So the decision an interface makes is not how few taps checkout takes but what the tap is attached to: whether the amount is visible, when the money leaves, and whether the user can see which purchase it belongs to.

The famous demonstration is smaller than its reputation. Prelec and Simester (2001) auctioned Celtics tickets to MIT Sloan students, half told they would pay in cash and half by card: mean bids were $29 under cash against $61 under credit card. Their second auction, for a dinner for four, is the one worth reading. Where bidders identified themselves with any four characters, cash bids ran 13 percent above card bids — a reversal, and not significant. Where they had to write their actual card digits, card bids ran 36 percent above cash at p < .05. How committing the card felt flipped the result, in cells of thirty-odd people. Liu and Dewitte (2021) varied cash, credit card and mobile payment across four studies, N = 692, and reproduced the credit card effect on neither willingness to pay nor basket value; mobile payment was rated less painful and more convenient than cash, and that rating did not turn into spending. Treat the card premium as something to measure in your own funnel, not a constant you inherit.

The pooled record is real and small. Schomburgk, Belli and Hoffmann (2024) meta-analysed 71 papers and found a grand mean of g = 0.135, 95% CI [0.068, 0.203]: people do spend more cashless than in cash, by about a seventh of a standard deviation. Heterogeneity was extreme, with individual effects running from g = −1.508 to g = 1.852, and publication bias looked unlikely. The uncomfortable part is the moderator test. Neither feature the pain-of-paying account nominates survived on its own — payment transparency (p = 0.122) and decoupling of payment from consumption (p = 0.568) were both null — and only their interaction held, with transparency mattering inside decoupled methods (mean g = 0.259 high versus −0.350 low, p < 0.001) and not inside coupled ones. The effect is also shrinking: b = −0.006 per year, p = 0.025. Size any payment-visibility change against that seventh of a standard deviation, which the largest study cannot attribute to visibility.

The coupling half holds up at field scale. Quispe-Torreblanca, Stewart, Gathergood and Loewenstein (2019) used 1.8 million credit card accounts and found debt incurred for non-durable consumption about 10 percentage points more likely to be repaid than debt for durables, comparable to a 15 percentage point increase in APR: people clear the hotel before the sofa, because the sofa is still paying for itself. Thomas, Desai and Seenivasan (2011) tracked the shopping of 1,000 households over six months and found a higher proportion of impulsive and unhealthy items in card baskets than in cash baskets, mediated by self-reported pain of payment. Which purchase a charge belongs to is worth naming on the receipt and the statement line, not only in the cart.

The neural story is suggestive and thinner than it is usually sold. Knutson, Rick, Wimmer, Prelec and Loewenstein (2007) found excessive prices activating the insula and deactivating medial prefrontal cortex, each region predicting the purchase beyond self-report. Robitaille, Plassmann, Mazar and Lindner (2026) found anterior insula activity tracking price parametrically, and paying with money engaging affective but not somatosensory pain networks where paying in electric shocks engaged both; their reverse-placebo causal test left those expecting an affective pain enhancer willing to pay more than those expecting a reliever, with neither differing from an untreated control. The feeling is real enough to design around and too blunt to dose.

Pain is also doing work you may want to keep. Shah, Eisenkraft, Bettman and Chartrand (2015) show across four studies that paying by a more painful form, cash or cheque rather than card, leaves people more attached to what they bought, less committed to the alternatives they rejected, more willing to signal that commitment publicly and more likely to transact again, at identical monetary cost and even when the money belongs to someone else. Rick, Cryder and Loewenstein (2008) add the Tightwad–Spendthrift scale: the gap between the two types is largest where the pain is amplified, so stripping it out compresses your users toward one behaviour rather than lifting everyone. Gourville and Soman (1998) add the clock: the felt cost of a prepayment depreciates, so an annual plan stops pulling the user back within months of the charge. A frictionless checkout buys the transaction and spends down the attachment that would have brought the next one.

Where friction has been removed at scale, the cost shows up outside the funnel. In banking records for 10.6 million US consumers, deHaan, Kim, Lourie and Zhu (2024) found new buy-now-pay-later users running up overdraft charges and credit card interest and fees relative to non-users, consistent with overborrowing, though they note a coinciding shock such as a layoff could drive both; Di Maggio, Katz and Williams (2022) find that BNPL rollout raises total spending and the retail share of spending by more than substitution explains. So treat the record as mixed: converging support from self-report, mediation, neuroimaging and a validated scale, against a small pooled effect that is heterogeneous, declining year on year, absent from a four-study replication and unexplained by the two salience moderators the theory nominates. And this literature measures willingness to pay, basket composition and spending, not checkout completion: the widely quoted conversion lifts for one-tap payment and wallets come from vendors, not from this evidence base. The spending side of a payment change is a known small effect; the conversion side is unmeasured in your product.