Where the Numbers Come From
6 min read · Published by EasyPayCalc, easypaycalculator.com.au
Sources behind this guide last checked 15 August 2026.
A pay calculator is a stack of claims. Every figure it shows you rests on a tax rate, a threshold or a rule that somebody chose to put into it, and you have no way of knowing from the outside whether that somebody was careful. This page is our answer to that problem. It explains where our values come from, what order of authority we hold them to, and what stands between a typing mistake and your screen.
A different page covers what the calculator does and does not calculate. This one is about something narrower and, for money, more important: whether the numbers underneath the calculation can be trusted, and how you could check.
You do not need any of this to use the calculator. It exists for the moment you want to rely on a result, when you are about to sign a lease against it or challenge a payslip with it, and for anyone whose job is to judge whether a site like this deserves to be shown to people at all.
The law outranks the guidance
Tax values are set by legislation. The tax office publishes pages explaining that legislation, and those pages are what most people, and most calculators, actually read. Nearly all the time the two agree. When they do not, because a page lags behind an amendment or summarises it loosely, the law wins, so our values are sourced to acts and legislated amendments first and to official pages second.
This is not a theoretical preference. In checking values against the legislation we have found an official page describing a change as if it were already in force when the amendment had not passed, and a competitor site quoting a family threshold from a superseded year. Reading one layer deeper is slower. It is also the entire point.
The superseded threshold is worth dwelling on, because it shows how these errors spread. Most sites do not read the law. They read other sites, or last year's page, and a stale number can travel a long way looking perfectly confident. The only defence is boring: go to the source, quote it, and write down the day you looked.
Announced is not enacted
Governments announce tax changes months before parliament passes them, and announcements make headlines while enactments do not. A calculator that adopts announced values early will match the news and mismatch your pay, because employers withhold under the law as it stands.
Our rule is that a value enters the calculator when it is law, on the day it takes effect, and not before. Where an announced change is widely expected and users may wonder about it, the calculator says so in plain text next to the affected figure rather than quietly adopting it. You will have seen this on the Medicare line if you have used the calculator while a proposed threshold sat before parliament.
Every value carries its receipt
Inside this site, next to each stored value, sits a note recording where it came from, the words of the source it was read from, and the date we last checked it. None of that is visible in the calculator, which stays deliberately uncluttered, but it means any value can be traced back to its origin without guesswork, by us or by anyone reading the code.
The table below shows when each group of values was last checked against its official source. Dates move forward as checks happen, and a monthly reminder run compares the stored values against their sources so a silent change upstream does not stay silent here.
| Values | Last checked |
|---|---|
| Income tax brackets | 15 July 2026 |
| Low income tax offset | 15 August 2026 |
| Medicare levy settings | 19 August 2026 |
| Medicare surcharge tiers | 7 July 2026 |
| Study loan repayment thresholds | 3 July 2026 |
| Super rate and contribution caps | 16 July 2026 |
| Super maximum contribution base | 18 August 2026 |
| High income super surcharge | 16 July 2026 |
| Working holiday maker brackets | 3 July 2026 |
| Minimum wage and casual loading | 12 August 2026 |
| Luxury car tax thresholds | 16 July 2026 |
| Car limit for depreciation and GST credits | 15 August 2026 |
| Fringe benefits tax rates | 16 July 2026 |
| Not-for-profit FBT exemption caps | 21 August 2026 |
| Electric car FBT exemption | 21 August 2026 |
The checks also differ in depth, and the honest table admits it. Some dates record a full comparison of every value in the group against its source. Others record a narrower look, a single rate or rule confirmed while the rest of the group stood unchanged, and the notes inside the site record which kind each date was.
Read the table as a freshness stamp, not a warranty. Most of these values change once a year at the start of July, so a check date from months ago usually means nothing has moved since, not that nobody has looked. What the dates rule out is the quiet failure mode, a value that has been wrong for a long time with nobody able to say when it was last known to be right.
Tests stand guard
Recording a value correctly is not the same as using it correctly. Between the stored values and the results sits arithmetic, and arithmetic is where mistakes hide. So the calculations are pinned down by a suite of automatic tests that runs before any change ships, comparing the engine against answers worked out independently.
The word independently is doing real work in that sentence. A test whose expected answer was copied from the calculator itself proves nothing, and worse than nothing, because it will defend a mistake as loyally as a correct answer. Our rule is that every expected answer must state where it came from: a worked example published by the tax office, a calculation done by hand that a reader could repeat, or a separate recalculation done outside the engine. When we add a test, we also break the code on purpose to watch the test fail, because a test that has never failed has never been tested.
We have been wrong before
A page like this earns nothing if it only describes intentions, so here is the failure that shaped the rules above. For a time, our casual pay tab applied the casual loading to an hourly rate that already contained it, quietly inflating every result on that tab. Worse, the automatic tests approved: their expected answers had been derived from the faulty output, so the suite defended the bug and failed the fix.
We corrected the calculation, rebuilt those tests from answers a person can verify by hand, and turned the lesson into the standing rule about independent expected answers you read a moment ago. The mistake is also told, in user terms, inside the hourly rate guide, because the same double counting trap catches people doing their own sums.
None of this makes the site infallible, and this page is not a promise that nothing is wrong. It is a description of the nets we have strung, and an invitation. If a figure here disagrees with an official source, we want to know, and the contact address in the footer reaches us.