How to split tax and tip: in proportion, not per head

How should tax and tip be split when you itemize the bill?

Proportionally to each person’s share of the subtotal. Tax, tip, and any percentage-based fee should ride along with the food they were computed from. The cleanest way to do it isn’t to allocate each overhead line separately — it’s to collapse the whole bill into a single multiplier and apply it to each person’s items once:

That’s the entire method. Take the invented example we’ll walk through below: a bill that lands at $154.20 on a $120 subtotal makes the multiplier 1.285, so the person who ordered $20 of food owes $25.70, all lines included. No separate tax arithmetic, no debating the tip line — the multiplier distributes everything at once, in proportion, and the shares reconcile — up to a rounding cent — because proportional shares of a total always sum back to that total.

What the gap looks like on a real-shaped bill

Here’s an illustrative table of four — the numbers are invented to be arithmetic-friendly, but the shape (one heavy order, one light one) is what real bills look like. Subtotal $120, tax at 8.5% ($10.20 — inside the 8–9% band where the largest share of splitty’s US receipts land), a 20% pre-tax tip ($24.00). Overhead: $34.20, which is 28.5% of the subtotal. Bill total: $154.20.

Look at Jordan’s line. Their food was $10; their fair, all-in total is $12.85. Under the half-itemized split they pay $18.55 — 44% more than their fair total — because the per-head overhead slice charges them the tax and tip on Sam’s average order, not on their own. Dev’s $8.55 discount is funded, dollar for dollar, by the two lightest orderers. And notice Sam: for whoever ordered exactly the table average, the two methods agree to the penny. The per-head split isn’t wrong everywhere — it’s wrong in proportion to how far each order sits from the average, which is what makes it so easy to miss on your own line and so costly on the lightest one.

If that pattern sounds familiar, it should: it’s the same regressive transfer that a fully even split inflicts on the whole bill, shrunk down to the overhead lines. We measured that larger version against real receipts in the even-split hidden tax; this piece is about the last place it hides after a group thinks they’ve fixed it.

Tax and tip were never per-head charges to begin with

The strongest argument for the proportional rule isn’t a fairness theory — it’s how the two lines got onto the bill in the first place. Neither one was generated per head. Both were generated as a percentage of what was ordered. That is also why offering to cover the tip instead of itemizing the drinks doesn’t settle up: the tip is indexed to the whole table’s order, so the amount it hands back has nothing to do with the gap it’s meant to close.

The tax line first. Sales tax on a restaurant meal is levied on the transaction: a percentage rate applied to the taxable items on the check. California’s tax authority illustrates its own guidance with a guest check — two items totaling $13.70, then “8.25% sales tax: $1.13” — the tax computed directly on the item subtotal. That California example is just one illustration of what a percentage tax does by definition: it scales with the taxable amount, so a pricier taxable item carries proportionally more of it than a cheaper one on the same check. Which items are taxable, at what rate, and how mandatory charges are treated all vary by jurisdiction — but wherever a percentage sales tax applies, that scaling holds. The tax line, in other words, already tracks what was ordered — it’s the sum of proportional charges on each taxable item. Dividing it per head takes a charge the state assessed against the items and silently reassigns it by headcount.

The tip line is the same shape by convention rather than by law. The Emily Post Institute’s guideline for sit-down wait service is 15–20%, pre-tax — a percentage of the food, not a flat amount per diner. (Even the suggested-tip math printed on that California guest check — “15%=$2.06” on a $13.70 subtotal — is computed on the pre-tax items.) A tip that was created as a proportion of the table’s consumption carries that proportionality with it; splitting it evenly discards the very logic that produced the number.

Sources: California Department of Tax and Fee Administration, Tips, Gratuities, and Service Charges (Publication 115); The Emily Post Institute, general tipping guide. These establish how each line is calculated (tax on the taxable items; the tip convention on the pre-tax bill), not how a table ought to divide it — the case for splitting them the same way they were computed is the structural argument above.

When is splitting tax and tip evenly actually fine?

Three cases, and they’re worth naming because they’re why the per-head habit survives.

What’s not fine is the unexamined middle: a table with a $60 order and a $10 order that itemizes the food because the gap is large and then splits the overhead evenly out of habit. If the spread was big enough to justify itemizing, it’s big enough that the overhead lines carry the same spread — scaled down to 23–29 cents on every subtotal dollar, but the same spread.

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).

What about service charges, auto-gratuity, and delivery fees?

Sort every extra line by how it was generated, and the rule assigns itself:

Percentage-generated lines inherit the proportionality of the items underneath them — that includes the “18% added to parties of 8 or more” auto-gratuity, which is computed on the bill exactly like a tip. (In California, a mandatory charge like that is also part of the restaurant’s taxable receipts — percentage lines can even compound into the tax line. What these charges are and how they differ from a tip is covered in service charge vs. tip and automatic gratuity for large parties.) The exception is a service charge billed as a flat dollar amount rather than a percentage — that one follows the flat-fee rule below, not the tip’s. And a true flat fee is the honest per-head case: a genuinely fixed fee, like a flat delivery charge, doesn’t scale with anyone’s order, so a per-head split of that line has a real argument behind it. (A charge that scales with a particular person’s order — a per-bag fee on the one who took the bags — is theirs, not the table’s.)

These lines are worth getting right because they’re real: a small share of splitty’s US receipts — about 2%, and that’s a lower bound, since it only counts charges the scanner parsed as their own line item — carry a separate service-charge or auto-gratuity line on top of tax and tip.

What total do you split now that cash rounds to the nearest nickel?

Split the total the check prints. If the table settles in cash, the amount actually handed over can differ from that printed total by a few cents — and that gap belongs to the payment, not to anyone’s food. The reason is new: the US Mint struck its final circulating penny on November 12, 2025, and as the remaining ones thin out, registers round cash totals to the nearest five cents. Nineteen states passed rounding laws in 2026, though several — New Mexico’s, Minnesota’s, Oklahoma’s — cover state or tax payments rather than a restaurant checkout.

That range is wider than it looks, for a reason most coverage skips: outside Arizona — the one state to mandate a method, the symmetrical “Swedish” ladder — the direction is generally left to the merchant, and the POS vendors ship it as a setting. Toast offers three: always round down in the guest’s favor, always round up in the location’s favor, or round to the nearest nickel. Only the last is symmetrical, and only under it is a total never moved more than two cents. Pick a direction instead and the move reaches four — Toast’s own example takes $3.97 to $3.95 one way and $4.00 the other. None of this touches the proportional rule: the rounding lands on the final total after tax is computed, and Arizona is explicit that it “does not change the amount on which tax is calculated.” Toast books it as a non-taxable adjustment outside net sales, since “rounding is not considered a sale.” Nobody ate it, so the cleanest place to leave it is with whoever settled the check. Worth not confusing this with dual pricing, where a menu posts a separate cash and card price up front — that one changes the item prices themselves, and it’s covered in the line-by-line guide to a restaurant receipt. Rounding happens at the register, after everything is totalled.

Why the number of checks matters more than the size of the bill

The rounding applies once per cash transaction, to that transaction’s own final total — a consequence the merchant documentation never spells out, since Toast’s guide is silent on split and multi-payment checks, but one that follows directly from the rule. Four separate cash checks create four rounding events instead of one. Under a symmetrical ladder they may well cancel rather than pile up; what more checks reliably add is more chances to diverge, and so a wider gap between what the table hands over and what the same meal on one check would have cost. That exposure tracks the number of checks, not the size of the bill: a $600 dinner on one cash check rounds once, and a $60 dinner split four ways rounds four times.

Keep it in proportion, though — this is a reconciliation problem, not a money one. The Richmond Fed put the nationwide “rounding tax” at roughly $6 million a year using the Federal Reserve’s 2023 Diary of Consumer Payment Choice, and that is a national total falling only on cash payers, modelled on symmetrical rounding. Cash was 14% of US transactions in 2024. The literature doesn’t even agree on the direction: as the St. Louis Fed summarizes it, Canadian grocery evidence found a small transfer to stores while a 2007 convenience-store study found a small benefit to consumers, and Chande and Fisher’s simulations found nearest-nickel rounding was not inflationary. So don’t chase it. Just agree on which number you’re dividing before you divide it.

Sources: Federal Reserve Bank of St. Louis, “When a Penny Costs More Than a Penny” (Page One Economics, May 2026) — final strike date, cash share, and the mixed rounding literature; Zhu Wang & Russell Wong, “Rounding Up: The Impact of Phasing Out the Penny”, Federal Reserve Bank of Richmond Economic Brief No. 25-27 (July 2025); Arizona Department of Revenue, “Arizona Penny-rounding Law” (A.R.S. § 44-7952); Toast, “Configure Cash Rounding During U.S. Penny Phase Out”; Arizona Capitol Times, “19 states pass rounding bills since last penny was minted” (July 2026); Dinu Chande & Timothy C. G. Fisher, “Have a Penny? Need a Penny?”, Canadian Public Policy 29(4), 2003.

Other split calculators build the proportional option in

The proportional rule isn’t a splitty invention — other bill-splitting tools offer it too. ItemSplit, an itemized-bill calculator, states the rule on its landing page in exactly the terms above: “Tax and tip are distributed based on what each person ordered. If you had 30% of the subtotal, you pay 30% of the tax and tip” (accessed July 2026). Easy Check Splitter’s January 2026 update made the choice explicit per line — its v1.7 changelog “enabled tax, tip, flat fees, and discount to be split evenly or relative to items ordered,” with a toggle next to each field. Notably, Easy Check Splitter offers both modes rather than forcing one: proportional overhead is a first-class option in these tools, not a fringe setting.

splitty now has a browser version of the same rule: a free proportional bill split calculator that divides shared items among their sharers and allocates tax and tip by each person’s subtotal, so the per-person totals sum back to the bill.