How this data is built
Everything on this page traces back to eight government releases. The Small Business Administration publishes every Paycheck Protection Program loan under the Freedom of Information Act; the copy used here is the 2024-09-30 vintage, 11.5 million loans in thirteen files. USASpending.gov publishes each COVID Economic Injury Disaster Loan as a prime award; 3.68 million of them, April 2020 through December 2022, were exported in 38 date-window files. The SBA also publishes its Restaurant Revitalization Fund award file and a Shuttered Venue Operators Grant award list. And the Department of Justice exposes its press releases through a public API, harvested on July 4, 2026 back to a January 1, 2020 cutoff: 131,293 releases, carrying DOJ dates from October 25, 2019 through July 3, 2026.
Each source file was downloaded, hashed with SHA-256, and frozen as part of a numbered release (v1, 2026-07-10). The eight derived tables in that release were built from those frozen copies by scripts, then published as UTF-8 CSV with a column schema and their own checksum. Internal file paths were stripped from the tables before publication, and cells that begin with a formula character are quoted so a spreadsheet will not execute them. The source vintage is stated on every table because the numbers depend on it.
The tables fall into four groups. Three describe PPP lenders: who originated how many loans and dollars, with the fintech aggregators Womply and Blueacorn consolidated into their own rows. One models PPP processing fees under the 2020 and 2021 schedules. Three are built from the DOJ corpus and cover enforcement: a parsed index of every pandemic-fraud press release, a sentencing database, and a classification of PPP and EIDL cases by whether someone else's identity was used. The last matches businesses across programs.