Why equal splitting costs you money
Revolut’s help page gives you both options in one line — “You can adjust the individual amounts owed or just tap Split bill to split the bill evenly” — and only one of them is a single tap. Tap, divide, done. But equal splitting has a well-documented cost.
In 2004, economists Uri Gneezy, Ernan Haruvy, and Hadas Yafe ran a field experiment in a restaurant in Israel, seating diners in groups of six. One group paid individually, one split the bill evenly between the six members, and a third ate free. The samples were small — 24 diners in the individual-pay treatment, 24 in the even-split treatment, 12 in the free-meal treatment — and the authors are precise about which measure carried the result. Counting items ordered did not clear the bar: between individual pay and even split it came in at p = 0.0818 on the t-test and p = 0.0948 on the Mann-Whitney, and the authors warn the item count is unreliable anyway “due to the large number of ties.” What did clear the bar was the money. The cost of the meals, they write, “provided clear-cut evidence to conclude that the samples are significantly different under all three hypotheses” — for individual pay against even split, p = 0.0014 on the Mann-Whitney and below 0.0001 on the t-test.
“When splitting the bill, diners consume such that the marginal social cost they impose is larger than their own marginal utility and, as a result, they over-consume relative the social optimum.”
Gneezy, Haruvy & Yafe, The Economic Journal (2004)
The researchers called this the Unscrupulous Diner’s Dilemma — a game-theory scenario where diffusing the cost makes overspending the individually rational move. Read that at the level the study measured it: the even-split groups spent significantly more on average. The paper does not report that every diner in them ordered more. The telling detail is what diners wanted: asked before ordering whether to split the bill or pay individually, 19 of the 24 subjects asked — 80% — chose to pay individually. Given the choice, people opt out of the mechanism.
The effect is not unlimited, and the paper says so. A fourth treatment had each diner pay one-sixth of their own meal, and comparing that against the even split produced no significant difference on either cost or items ordered. The authors add their own caveat — “the lack of significance could be due to the smaller sample size of the ‘Pay 1/6’ group” — so read it as a hint, not a result. Taken at face value, it suggests the mechanism has a floor: what matters is that you are not paying your own marginal cost, not how finely the bill is carved up afterwards.
Be careful about pointing this study at Revolut, though, because the timing does not line up. Gneezy’s diners were told the rule before they ordered — that is the whole mechanism, since it changes what a starter costs you at the moment you choose it. Revolut’s Split Bill runs at the opposite end of the meal: it opens a transaction you have already paid. Nothing you tap in the app can reach back and change what the table ordered. So the study cannot show that Revolut causes overspending, and this page is not claiming it does.
What survives the timing problem is the expectation, not the app. If everyone sits down already assuming the bill gets divided by five, the Gneezy mechanism has done its work before the first drink arrives — and Revolut’s contribution is that it makes that even division the fastest, most frictionless way to finish the night. The app does not create that norm, and nobody has tested whether it strengthens it. What can be said without a study is narrower and duller: the even split is the one-tap path, and the itemised one is not. For more on why fair splits matter, the direction of the evidence is consistent: the overspending is a response to the payment rule, so changing the rule is what removes the incentive.
The authors also draw a line the article should not blur. Splitting the bill, they note, is “generally prevalent among friends or colleagues and not among strangers,” and that difference is “critical since with friends and colleagues the game is repeated, so punishment strategies in response to excessive waste are feasible.” They go further: given those repeated-game strategies, “it may in fact be individually and socially optimal to split the bill among friends.” A table of six people who eat together every month is not the table Gneezy measured. The waste is worst where the social discipline is weakest — dining with distant acquaintances, or a group that will never assemble again.
The other thing worth taking from the study is how uneven a single table is. The authors flag “the variability in subjects’ costs for any given treatment” as striking enough to worry about — within one treatment, diners eating the same menu under the same rule ran up wildly different bills. That is the shape of a real table too: not a set of equivalent orders, but a salad and a round of cocktails on one receipt, waiting to be divided by five.
Source: Gneezy, Haruvy & Yafe, “The Inefficiency of Splitting the Bill,” The Economic Journal (2004), 114 (April), 265–280. P-values are from Tables 2 and 4 of the published paper.
