Methodology
This page describes how every number on this site is obtained, checked and computed. It is deliberately specific, because a tax data site that will not show its working is asking to be trusted for no reason.
1. Where the data comes from
Every rate, threshold and deduction is read from a primary source. We use a strict hierarchy, and a lower tier never overrides a higher one.
| Tier | Source type | Used for |
|---|---|---|
| 1 | State department of revenue or taxation | Rate schedules, instruction booklets, withholding formulas, official tax alerts. |
| 2 | State statute or enacted legislation | Used where a department has not yet republished after a law change, or where a bill sets a future rate. |
| 3 | Internal Revenue Service | All federal figures, and the federal standard deduction for states whose base conforms to it. |
| 4 | Other official government | Census population, Social Security Administration wage bases, municipal finance departments for local rates. |
| 5 | Federation of Tax Administrators | Orientation and cross-checking only. |
| 6 | Research organisations | Cross-check only. Never the source of record for any published figure — this is enforced by our build-time validator, which fails the build if a research organisation is recorded as a primary source. |
Where sources disagree — which happens more than you would expect, particularly when a state cuts rates mid-year and its own withholding tables lag its statute — the higher tier wins, and we record the conflict and the resolution in the state's source notes. Every source we cite was actually retrieved; we record the date. See the full source list.
2. Confidence levels
Not all data is equally solid, and pretending otherwise is how tax content goes wrong. Every jurisdiction record carries one of three states.
| Level | 2026 count | What it means |
|---|---|---|
| Verified | 40 | Read from a primary source that explicitly covers this tax year, and confirmed. Published normally. |
| Provisional | 10 | The rate schedule is confirmed, but one component — usually a cost-of-living-adjusted deduction the state has not yet published — is carried forward from the prior year. Published with a visible notice, and the specific pending item is named on the page. |
| Unverified | 1 | We could not confirm this year’s figures from a primary source. These are never published as current fact: the state page shows the most recent verified year with a prominent banner, and the jurisdiction is excluded from every ranking, comparison and national table. |
3. How the tax is calculated
Every figure on the site comes from one tax engine, so the calculator, the state pages, the rankings and the comparisons can never disagree with each other. The sequence is:
- 1Start from the income you enter, treated as income already adjusted for federal purposes.
- 2Subtract the state’s standard deduction. Where a state computes it as a percentage of income within a floor and cap, we apply that formula rather than a flat figure.
- 3Subtract personal and dependent exemptions, where the state grants them as deductions.
- 4Apply the bracket schedule incrementally. Each rate touches only the income inside its own bracket. Where a state publishes a "base tax plus rate on the excess" table whose base is not the simple cumulative sum, we use the state’s published base.
- 5Add any surtax that applies above its threshold.
- 6Subtract exemption credits, where the state grants exemptions as credits against tax rather than deductions from income. The result is floored at zero — we never show a negative bill.
The engine is covered by an extensive unit test suite that checks bracket boundaries and the dollar either side of them, zero and negative income, deductions larger than income, very high incomes, all four filing statuses, flat states, no-tax states, zero-rate brackets, surtaxes and credit-style exemptions.
4. What we deliberately do not model
5. The model taxpayer
Comparative pages — the national table, the rankings, the income-scenario pages — all use the same hypothetical person so the comparison is like for like: one filer, filing single unless stated otherwise, with the stated income already adjusted for federal purposes, claiming the state standard deduction and personal exemption, no dependents, no itemised deductions, no credits, and no local income tax. Changing any of these changes the answer, which is what the calculator is for.
6. Which pages we refuse to create
There are 1,275 possible pairs of U.S. jurisdictions. We publish 150. A pair gets a page only if both states have verified or provisional data, at least one of them taxes income, and they genuinely differ — by at least $500 a year at $100,000, or in structure, local-tax applicability or capital gains treatment. It then has to be a comparison someone would plausibly make: bordering states where at least one has three million residents, a large no-income-tax state against a large taxing one, or two of the biggest taxing states head to head.
Pages that fail the test are not generated at all — not generated and hidden behind a noindex tag, which would still waste your time if you found one. Wyoming versus Alaska is the canonical example: both levy no income tax, so the page would read "$0 versus $0". The same principle applies everywhere on the site: a URL exists because it answers something, not because two dimensions can be combined.
7. Freshness and corrections
Every data page shows the tax year it describes and the date the figures were last checked against their source. An automated audit flags records that have not been re-verified recently, states missing current-year data, source URLs that have stopped resolving, and year-over-year movements large enough to suggest a transcription error. A separate build-time validator blocks a release outright if any bracket schedule has a gap, an overlap, an implausible rate, a missing filing status, or a jurisdiction without a primary source.
If you find an error, tell us. We would rather be corrected than be wrong quietly.