Pain of paying

Evidence

Prelec and Loewenstein (1998) developed the theory of coupling between payment and consumption. Their own data are ratings of hypothetical payment schedules by 86 to 91 respondents. Knutson et al. (2007) found in a brain-imaging study with 26 participants that insula activation during price display predicted the decision not to buy. It did not correlate significantly with the price differential itself (Z = −1.36); the authors explicitly call the specificity of this response unresolved.

Since 2024 there has been a large meta-analysis — and it answers the main question more clearly than the secondary ones. Schomburgk, Belli and Hoffmann (2024) pooled forty years of research on cashless payment: 392 effect sizes from 71 papers, 11,257 respondents, 338,513 transactions, 17 countries, 1978 to 2022. That people spend more with cashless methods is confirmed — but weakly: Hedges’ g = .135 (95 percent interval .068 to .203). The g measures how large an effect is: 0.2 already counts as small, and .135 sits below that. The pain of paying is the dominant explanation of that finding in this literature.

Against the file-drawer problem the collection is carefully built. Twenty-three percent of the papers were not published in peer-reviewed journals at all — markedly more than in comparable meta-analyses — and three standard checks find no sign that unpublished null results are distorting the picture.

The spread, however, is enormous. Individual findings run from g = −1.51 to 1.85, and 86 percent of the differences between them are not down to chance but to real differences between the studies. A mean of .135 does not describe any reliable single case here.

The two predictions the mechanism rests on were not supported on their own. Prelec and Loewenstein derive the effect from the decoupling of payment and consumption; Soman added the transparency of the payment. Neither works by itself: decoupling b = −.024 (p = .57), transparency b = .073 (p = .12).

Only together do they produce something — and what they produce is a reversal.

Among decoupled methods the difference between high and low transparency is significant (p < .001); among coupled methods it is not (p = .33). Which means: the credit card, the textbook case, sits at the top — buy-now-pay-later at the bottom, with the sign reversed.

Three further findings bear directly on pricing pages. First, in pro-social situations — tipping, donating — the effect is switched off entirely (ḡ = .023, the confidence interval includes zero). Second, in conspicuous consumption meant to display status it is strongest at ḡ = .55 — though from only 13 effect sizes across seven papers. Third, whether a purchase is hedonic or utilitarian makes no difference (p = .32), although the literature long assumed otherwise.

The effect is shrinking. Over the years it declines linearly (b = −.006, p = .025). The authors attribute this to habituation: someone who has always paid cashlessly no longer experiences a difference. Conversely the effect is stronger in periods of growth (b = .033, p = .009) — contrary to the authors’ own expectation. The inflation rate does nothing at all (p = .997).

One methodological finding partly exonerates hypothetical data. Whether a study used real money or invented scenarios made no difference (b = −.021, p = .80); the same holds for student samples and online surveys. The effect came out stronger only where participants were told the price (b = .158, p = .012).

Limits

Brain activity does not unambiguously indicate “pain”. The idea that currency symbols trigger the pain is not even supported by that one restaurant study: in Yang et al. (2009) tables with a dollar sign and with the currency spelled out spent the same (p > .99); only the bare numeral differed. A related money-priming effect failed outright in the Many Labs project: instead of the original effect (d = 0.80), −0.02 remained (99% CI −0.08 to 0.04) across 36 samples and 6,344 participants, with a single sample reaching significance (Klein et al., 2014). That was a different paradigm — images of money and system justification, not the pain of paying. But the null is entirely uniform (I² = 0) and so cannot be explained away by differences between samples.

The term does not come from the authors named here. “Pain of paying” was coined by Zellermayer in a 1996 dissertation. Prelec and Loewenstein contributed the coupling theory. The neurological basis the meta-analysis invokes is a working paper by Mazar and colleagues (2017), which its own authors call preliminary.

The effect is small and the field is not uniform. On the usual reading g = .135 is a small effect, and 86 percent of the spread is not due to chance. Anyone looking for a number to apply to their own shop will not find one in this paper, but a list of conditions.

The authors explicitly advise against exploiting the effect. They recommend that retailers forgo credit cards and buy-now-pay-later as revenue levers, because these draw on money customers may not have, and use debit cards instead. Charities they advise to keep the cash box — for donations, cashless is no better.

The evidence comes almost entirely from wealthy countries. China contributes 5 effect sizes, India 13, Poland and Turkey one each. Personality differences the authors could not test at all; a side calculation shows only that the effect weakens as the proportion of women in the sample rises.

The evidence comes out in two halves. The main finding — that people spend more without cash — is confirmed across 392 effect sizes, small though it is. The mechanisms usually invoked to explain it are not: decoupling and transparency do nothing on their own, and in one of the four combinations the effect reverses. What is supported, then, is that more is spent, not why.