Paycheck State Calc

Changelog

What changed, when, and why. The data-layer entries come from the tax year's own changelog file; the page entries come from the repository's CHANGES file. Newest first.

  1. – Trust pages added: about, methodology, changelog, corrections and a worked example

    Added /about, /methodology, /changelog, /corrections and /state-tax-changes-2026, a fully worked $60,000 Texas biweekly example on /how-it-works computed by the engine at build time, a verification line on every content page, and the six trust links in the footer. tests/test_eeat_pages.py added. Georgia's 2026 Table E is verified in data/tax_year/2026/georgia.json; its page is built and not yet published, so the 2026-10-30 note below that says it remains blocked is superseded on that point.

    Why: A calculator that prints a number without saying where the number came from cannot be checked by the person reading it, and a worked example is the only way to show the arithmetic rather than describe it.

  2. – Two information pages added: a coverage ledger and a state comparison

    Added /coverage, a coverage ledger with three status groups generated from PUBLISHED, LIVE_STATES and the engine publish gate, and /compare-states, which runs one $75,000 semimonthly salary through all eight live state calculators. Home page links to both; tests/test_pages.py added to hold the pages to the same rules as the calculators.

    Why: A coverage claim that blurs built and live hides the one state that is built and unpublished, and a cross-state comparison shows that states differ in kind, not only in rate.

  3. – Michigan withholding implemented: annualise, subtract $5,900 per exemption, x 4.25%, divide by pay dates (method per the 2026 federal payroll restatement, figures per Michigan Treasury)

    Georgia standard deduction ($15,000/$30,000) and glide-path target (3.99%) recorded; Georgia remains blocked on the 2026 Table E allowances.

    Why: Withholding is not rate x gross. Without the annualise, subtract the exemption, multiply, divide sequence, the Michigan line was a rate-only estimate that no pay stub would match.

  4. – Washington PFML employee/employer split (71.43% / 28.57%) verified against the ESD news release for the 2026 rate

    WA is now fully publishable; TX and WA are both unblocked.

    Why: The employee/employer split was supplied before a primary source was found. An unverified split would have put a guessed deduction on every Washington page, so it was confirmed against the ESD release before the page could publish.

  5. – Added FUTA (Pub 15), Georgia and Michigan state income tax rates

    GA/MI are rates-only: their withholding formulas are marked pending_release because withholding is not rate x gross. Added engine/employer.py with employer FICA + FUTA and SUTA as a caller-supplied rate.

    Why: An employer-cost figure that hard-codes a statewide SUTA rate is wrong for every employer with an experience rating, and rate-only state data would have forced rate x gross on states that do not work that way.

  6. – Initial 2026 data layer

    Federal withholding schedule transcribed from Pub 15-T (2026) by tools/extract_p15t.py and passed continuity validation. FICA and federal parameters read from Pub 15 / 15-T. TX and WA added.

    Why: Starting from the arithmetic rather than from the pages is the only way the later claims about sourcing can be true.

Data-layer dates come from data/tax_year/2026/metadata.json. Page dates come from CHANGES.md. Read the methodology for how a change gets verified, or about the site for who makes them. Start from the state calculator to check the current figures.