Splitting the bill evenly taxes the light eater

Is splitting the bill evenly fair?

The mathematics of fair division — procedures for splitting a shared cost that people provably can’t object to — has been worked out over decades, and we cover it in is splitting the bill evenly fair. This piece asks a different question: not whether the even split is unfair, but how big the unfairness is, and who pays it. That’s a question about incidence — the same question a tax economist asks. The receipts can’t see who ordered what, but they supply the raw material: how wide the gaps an even split ignores actually run.

The answer is that the burden lands on the person who consumed the least. When you split evenly, your payment is fixed at the average no matter what you ordered, so the ratio of what-you-pay to what-you-got is largest for the smallest order. Say the table averages $30 a head: the person who ordered a $6 salad pays five times their own consumption, while the person who had $80 of food and two cocktails pays a fraction of theirs. The money flows uphill, from light to heavy, every single time.

Equal and fair are not the same value

Psychologists have a precise name for the thing an even split gets wrong. When people divide a shared cost, they can reach for different principles of fairness. One is equality — everyone receives the same outcome “regardless of possible differences in the size of their contributions.” Another is equity, also called the proportionality principle, where “the outcome for an individual should be proportional to his or her contribution.” An even split is the equality rule wearing the costume of fairness. Charging by what each person ordered is the equity rule.

Which one people actually judge to be fair isn’t fixed — it depends on what the group is for. Morton Deutsch’s classic account of distributive justice argues that in relationships built around getting something done efficiently, equity dominates; in relationships built around keeping everyone comfortable, equality does. A 2017 experiment by Kazemi, Eek, and Gärling put that to the test: when the group’s goal was productivity, people’s allocations landed “closest to equity”; when the goal was harmony or social concern, they shifted to equal final outcomes — everyone walking away the same. That was a public-goods allocation task, not a dinner table, but the parallel is hard to miss: split-it-evenly belongs to the same equality family — the smooth-the-moment, don’t-itemize-a-friendship choice — applied to a bill where the orders were never equal to begin with.

The broader research on how people actually judge fairness says the same thing: there is no single rule. James Konow’s survey of justice theories concludes that people’s fairness preferences run on several forces at once — equality, need, and equity among them — and that which one wins is a matter of context. An even split isn’t “the” fair answer; it’s the equality setting, applied by default to a situation — consumption that was never equal — where the proportionality principle has the more obvious claim.

Sources: Morton Deutsch, “Equity, Equality, and Need,” Journal of Social Issues (1975); Ali Kazemi, Daniel Eek & Tommy Gärling, “Equity, Equal Shares or Equal Final Outcomes?” Frontiers in Psychology (2017); James Konow, “Which Is the Fairest One of All?” Journal of Economic Literature (2003).

The more items on the bill, the wider the gap

The even-split tax isn’t a fixed toll — it scales with how spread out the items are. Grouping splitty’s receipts by how many items are on the bill (a rough stand-in for party size — a receipt doesn’t record how many people were there) shows the gap between the priciest and cheapest item widening at every step:

Some of that is mechanical — more items means more chances for an extreme high or low. But that’s exactly the point: a bigger bill is a bigger spread to divide, and an even split ignores the spread entirely. A quieter way to see it: divide a bill’s subtotal equally across its own line items, and the average item still lands about 44% away from that equal slice. “Even” is nowhere near what any one item cost — and the wider the bill, the more the cheap items subsidize the expensive ones.

Item-count bands and dispersion: splitty’s scanned restaurant receipts (US-leaning, itemized), medians, rounded. Item count is a rough proxy for party size, not a count of diners; receipts do not record who ordered what.

Where does the even-split tax bite hardest?

The drink line. On splitty’s receipts, 61% of itemized restaurant bills mix at least one drink with at least one food item — the mixed check is the norm, not the exception. On those mixed bills, drinks are a median 16% of the subtotal, climbing past 29% on the top quarter of them. And the drink itself isn’t cheap on its own terms: the priciest drink on a mixed bill runs a median $15.50, climbing to $30 on the heaviest-drinking quarter of mixed bills — often enough to rival an entrée outright. That’s not a rounding error tucked at the bottom of the check. Split evenly, anyone at that table who skipped the round still absorbs an equal slice of the $15.50-to-$30 line — the arithmetic doesn’t ask who drank it.

This is the cleanest case of the transfer because the divide is so stark: a cocktail can cost as much as an entrée, and the person who skipped it gets none of the enjoyment and a full share of the bill. It’s a big enough problem to have its own playbook — see sober dining and splitting the bill for the non-drinker’s case, and splitting a bill with cocktails for what happens when the drinks outrun the food, and the hybrid bill split for the three-step method that pulls the drinks off the top and splits the rest evenly.

Drink share and price: splitty’s scanned restaurant receipts (US-leaning, itemized bills), medians and shares, rounded. Drink line items are identified by matching OCR item text against a lexicon of drink terms (beer, wine, cocktail names, and similar) — a lower bound, since only matched labels are counted as drinks.

Why nobody notices they’re paying it

A regressive tax you can see gets voted down. This one is invisible for two reasons. First, it arrives as a single number — the check gets divided once, at the end, and no line on it says “you just covered $11 of someone else’s steak.” Second, any one instance feels too small to fight. A few dollars over, on one dinner, isn’t worth the awkward conversation — which is precisely why it goes uncontested meal after meal. The person who always orders light and always pays the average is running a standing subsidy they never agreed to. (The math of why that small overcharge is worth caring about is in “it’s only five dollars”.)

It also lands on the side of the table primed to mind it. Loewenstein, Thompson, and Bazerman found that people weigh disadvantageous inequality — getting less than the person next to you — far more heavily than the advantageous kind. If that asymmetry holds at the table, the split’s two sides are felt very differently: coming out ahead barely registers, while coming out behind is exactly the kind of gap people weigh most heavily.

There’s a second cost the receipts can’t see: splitting equally doesn’t just misdivide a fixed bill, it can make the bill bigger. When no one pays for their own order, the incentive tilts toward ordering up — the pattern the classic diner’s-dilemma experiment documented among unacquainted diners splitting evenly — more on that in why fair splits matter. The even split quietly enlarges the pie and then hands the light eater an equal slice of the bill for it.

The even split is now a button in your banking app

The even split has graduated from social habit to product default. Banks have started shipping “split the bill” as a native feature inside Zelle, and each of these flows divides a total — none of them knows what anyone ordered. U.S. Bank’s version lets you split a bill with up to 10 people, and “to start, amounts are automatically calculated to split the total evenly between you and the others” — changing that means typing a new dollar figure next to each name yourself. Chase’s version starts from a settled card charge: swipe left, tap “Split,” tap “Include my portion” and, in Chase’s words, “we’ll do the math” — no step in the flow asks who ordered what. SoFi’s Zelle FAQ says it plainly: “It will default to even split.”

Why the pre-filled option matters is one of the best-documented results in behavioral science. In Johnson and Goldstein’s Science study of organ-donation choices, an online experiment varied only the default: 42% agreed to donate when agreeing took an opt-in, 82% when donating was preset — and a neutral version with no default at all landed close to the opt-out rate. The one condition that cratered was the one requiring action to say yes. The same paper points to “natural experiments” well beyond medicine — insurance choices, internet privacy policies, pension savings — all showing “large effects” from defaults, “often with substantial financial consequences.” Defaults stick because they read as recommendations, because undoing them takes effort, and because they become the reference point any change is measured against. If the same psychology holds on a bank’s review screen, a pre-filled equal split stops being something the table decides — it becomes what happens when nobody decides. The regressive tax this article measures arrives pre-checked.

Sources: Eric J. Johnson & Daniel Goldstein, “Do Defaults Save Lives?” Science (2003); U.S. Bank knowledge base, “How do I split the bill with Zelle®?”; Chase, “How to split and request money with Zelle®”; SoFi Zelle FAQ — bank pages accessed July 2026.

Apple went the other way: itemized from the receipt

The other big platform split flow of 2026 points in the opposite direction. Apple’s iOS 27 bill splitting — announced for this fall — starts from the receipt, not the total: point the iPhone camera at the check and, per Apple’s own announcement, it can “identify the items on the receipt,” and “as users select their items, their total payment is calculated, including their share of tax and tip.” That is the equity rule — pay for what you ordered, with your share of tax and tip calculated on top — built into the operating system’s receipt-scan flow. The split this article has been arguing for is no longer a spreadsheet enthusiast’s position; it’s what the receipt-scan button does.

The convergence is the tell. When one set of platforms hard-codes the equality rule and the other builds the equity rule into the camera, the question this article opened with — equal or proportional? — has stopped being table etiquette and become a product decision made for you. Apple’s version has its own limits: it’s US-only and settles on a single rail, Apple Cash — the one-wallet problem we cover in why one company’s wallet can’t fix the group bill. But on the fairness mechanic itself, the receipt-first design gets the starting point right.

Sources: Apple Newsroom, “Apple unveils innovative features and intelligence experiences across services” (June 2026); MacRumors, “iOS 27 Adds Six New Features to Apple Wallet” (June 10, 2026).