How TaxOptimus is built

The product is a versioned fact store. A rate is never overwritten. A new observation closes the previous one.

Structured official series come from the OECD Tax Database (SDMX), the European Commission’s Taxes in Europe database, World Bank WGI, Transparency International’s CPI, and government-bond yields (OECD / ECB / FRED). National ministry pages are the audit trail you click.

Scrapers and AI extractors write proposals. Nothing they produce is published until a reviewer accepts it. Every published row has a source URL and an as-of date.

Forecasts default to no change: we copy the last published value into the next tax year and mark the method status_quo. A sourced finance bill or government-change note is the only reason to replace that.

“Slow, bureaucratic, unreliable, opaque” are labels mapped from WGI government effectiveness / regulatory quality and the CPI. We do not maintain a secret house score. Rule-change risk is the only curated overlay, and it stays low unless we have a source.

Credit ratings are public letter grades with a link to the agency list. We do not republish rating reports.

Every jurisdiction has a published personal-income-tax bracket schedule for tax year 2026 — the statutory step function, not a fraction of the top rate. Where municipal tax is a large, flat extra (Finland, Sweden, Iceland, Liechtenstein) the schedule uses the official national average or the year's statutory communal multiplier, and says so. Subnational PIT (US states, Canadian provinces, Spanish autonomous communities, Swiss cantons beyond the Zurich illustration) stays flagged and omitted.

Spending (COFOG) and demographics are reserved tables. They are empty in v1.