WhatPays.org

Data sources

Every dataset behind the numbers

All of it is public, free and federal. None of it is proprietary to us. The advantage is not access — it is that a site funded by course sales cannot afford to publish what this data says.

Sources currently in use. Vintage is the most recent complete year available.
Source What it gives us Geography Access
Census Nonemployer Statistics
US Census Bureau, annual
Counts and receipts for businesses with no employees, broken out by receipts-size class. That size-class breakdown is what makes a failure rate calculable at all. National, state, metro, county Public API
IRS Statistics of Income
Sole proprietorships, annual from 2000
Receipts, deductions and net income by industry from Schedule C filings. The long series is what lets us detect a method decaying rather than guess at it. National, some state and county Published spreadsheets
Platform disclosures
Varies by platform
Official payout schedules, published rate cards and transparency reports. Used for tier-2 evidence where federal codes are too coarse. Varies Public

Planned, not yet in use

Coverage expands only where a country publishes administrative records good enough to support a real failure rate. Four do. Listing them here is a commitment, not a claim that they are live.

  • France — INSEE base Non-salariés, built from URSSAF social-security records. The strongest non-US source we have found.
  • Australia — ATO Taxation Statistics, individuals in business by industry code and broken out by income range.
  • United Kingdom — HMRC Survey of Personal Incomes, covering self-employment profit by industry.
  • Canada — Statistics Canada, which uses the same industry classification as the United States.

What we deliberately do not use

Screenshots of earnings dashboards, income reports published by people selling a course on the method, and community survey results presented without a response rate. Each of these describes a self-selected group of people who chose to talk about it, which is the specific bias this site exists to remove.

Known limits

Industry codes are coarser than methods, so several methods may share one denominator; where that happens the figure is labelled as shared. The smallest receipts are excluded from nonemployer data, meaning the true share earning almost nothing is likely higher than we report. Gig income paid on a W-2 does not appear in sole-proprietor statistics. And federal data lags by roughly two years, so a method that collapsed recently can still look healthy in the numbers. Every record carries a confidence marker reflecting these.

How these sources become a published figure