How the number is built

No black box. This page states the assumptions, the formula, the source for every jurisdiction, and what we deliberately do not model.

Source of record

The calculation order, and why it is written down

The order operations apply in is what decides the answer, and it is not the same in every jurisdiction. A social contribution that reduces the income tax base produces a different figure from one that does not. A surtax on the tax due is a different number from a surtax on taxable income at the same rate. A non-refundable credit applied before a sub-central ladder is not the credit the statute describes.

So the order is stated in the engine file rather than left implicit in the sequence of function calls, and it is this: gross, less any contribution the jurisdiction allows against the base, gives taxable income; the statutory ladder runs on taxable income; any surtax applies to whichever base its own rule names; any credit comes off the tax rather than the income and cannot take it below zero; any sub-central ladder runs on the base its rule names. Employee social security is computed on its own thresholds against gross. Employer contributions are excluded everywhere, because they never appear in your gross pay and including them would flatter or damn a country depending on where its wedge sits.

One consequence of stating it is worth naming. A scalar in an international dataset does not carry its own meaning: the same field is a percentage of the tax due in one country, a contribution on gross in another and a levy on taxable income in a third. Rather than assert a reading, the ingest builds every reading the data allows, computes each with the real engine, and keeps only a reading that reproduces the published outcome at all three incomes. A reading that survives that is a finding. A reading asserted would be a guess wearing a citation.

The nine gates

These run in order before anything is published and the build refuses on any failure. Each names the file that implements it, so the description can be checked against the code rather than believed.

What we agree with, and by how much

Gate F computes tax for a single person with no children at 67%, 100% and 167% of each country's own average wage, from our statutory rules, and compares the result with the average rates the OECD publishes for exactly that person in period 2025. The tolerance is 0.5 of a percentage point. The United Kingdom regime, built entirely from HMRC documents and owing the OECD nothing, reproduces the published British row to 0.001 of a percentage point, and that is what made the series usable as an acceptance test rather than a coincidence.

What we refuse, and why

31 jurisdictions are refused. That is not a coverage gap we are working up to hiding: it is the same gate, applied to the countries where it fails. 26 of them fail by a measured amount, and the amount is the useful part. It says exactly how wrong a page of ours would have been, which is a thing no site publishing all of them can tell you about any of them.

The remaining 5 are refused for reasons no number describes. Germany: 3 rates but 4 thresholds; a ladder needs exactly one fewer threshold than rates. United Kingdom: already ingested from its national authority; OECD is its cross-check, not its source. Greece: rate steps down from 90.00% to 22.00% at band 2; a progressive ladder does not fall as income rises, so one of these rates is wrong at source. Portugal: rate steps down from 21.50% to 0.24% at band 4; a progressive ladder does not fall as income rises, so one of these rates is wrong at source. United States: already ingested from its national authority; OECD is its cross-check, not its source.

Two of those are worth dwelling on because the refusal is about the source rather than about us. Where a published ladder steps down as income rises, a progressive schedule does not do that, so one of the rates is wrong where it was published. We do not repair a source's arithmetic and publish the result as a calculator. We say the series fails a sanity check and leave it.

What we deliberately do not model

This list is the honest part of any methodology page and it is the part most sites leave out. Everything named here is a known gap rather than an oversight, and each is named on the page it affects as well as here.

• Scotland. Scottish rates apply to non-savings non-dividend income on their own band structure. Until those are sourced separately the United Kingdom pages cover England, Wales and Northern Ireland, and say so on every one of them.
• Personal circumstances of every kind. Reliefs, other income, marriage allowances, student loans, salary sacrifice, pension arrangements, residence history and the timing of a move. Any one of these can move the answer more than the headline rate does.
• Employer social security. Excluded everywhere, deliberately. It never appears in your gross pay, and including it would change how a country ranks for a reason that has nothing to do with what lands in your account.
• Employee social security in some jurisdictions. Where it is real and not modelled, the page does not call its figure take-home pay. The label on the number is the claim; a disclaimer underneath does not repair it.
• Cost of living. No index is used, because none has been sourced to the standard everything else here is held to. It is directional at best in any case: the city moves it more than the tax does.
• Currency conversion, except where it is stated. Comparisons are made at each country's own average wage wherever possible, which needs no rate at all. Where a common currency is genuinely required, the rate is the European Central Bank's daily reference rate and the page prints the date it was published beside the converted figure.

The store behind all of it is 959 rules from 5 authorities, published in full at /data/ with every quote, locator and hash intact, under CC BY 4.0 over our compilation. If you want to check this page rather than read it, that file is the place to start.

Written and maintained by Nkosi Ndwandwe, who is not a qualified tax adviser and does not claim to be. The most recent figure behind this site was read from its source on 06 September 2026, which is the date this page reports as its last modification. Everything here can be checked against the document it came from without asking us anything.