# What the held studies say about PPP outcomes

Findings about the outcomes of the Paycheck Protection Program, each read off a document this archive holds as a file: 82 findings from 40 documents, each with the estimate exactly as printed, the study's own words for its data, period, method and caveat, and the page the sentence sits on.

Published at https://pandemicdarlings.com/data/ppp-outcomes-literature/ . Compiled table; not part of release pandemic-darlings-data-v1-2026-07-10. Built 2026-09-24.

## Files

| file | rows | bytes | sha256 |
|---|---|---|---|
| `ppp-outcomes-literature.csv` | 82 | 83835 | `3551707b0df06d5cc58dbbea6e07d220e9ca32a1de648cc9c63b54ee4c590bf7` |
| `ppp-outcomes-literature-comparability.csv` | 68 | 85171 | `b2283eba13b1ade9326f79a344710e21cb9f8d77889c4f6f6ac92788b3471f11` |

`dictionary.csv` beside this file describes every column of the served table: measured type, blanks, distinct values, range, example and meaning. `SHA256SUMS` carries the SHA-256 of every file on the page, including the page itself.

## What this is

One row is one finding printed in one held document. A row is not a study: the 82 rows come from 40 documents, and nothing is weighted, pooled, rescaled or averaged. A study with four rows is not counted four times in any statement made from this file.

Each row carries the estimate exactly as printed, the study's own description of its data, period, method and caveat where our copy prints them beside the estimate, the path and SHA-256 of our copy, the page the sentence sits on, and the sentence itself.

ppp-outcomes-literature-comparability.csv holds every pair of findings this archive reads as estimating the same quantity, and prints the two rows' periods, data, methods and publication types side by side. It does not say which estimate is right.

## How it was built

Every row was re-read from the held PDF with a text extractor other than the one used to build the table. For all 82 rows the file was present, its SHA-256 matched the value in the table, and the quoted sentence and every fragment of the printed estimate were found on the pinned page.

Re-verified again when this page was built: 82 of 82 files present, 82 of 82 hashes matching the row, and the quoted span re-found on the pinned page for 82 of 82 rows. Four of them needed a third reading: one reader breaks a hyphenated word across the line ending and renders an accented letter as two characters, so on those four pages it finds the sentence only once the line-ending hyphens are closed up. The printed sentence is on the pinned page in every case; none of the four estimates changed.

The thirteen outcome classes are derived from the held rows, not imposed on them. question_studied and the comparability file's quantity_label are the only free-text fields written by this archive rather than copied from a document.

## Read this before you use it

This is not the literature. The documents here are the ones the archive happens to hold as files. Counting rows or documents measures the archive's holdings, not the field. A study missing from this table may be missing only because the archive does not hold it.

It is not a meta-analysis. Two rows in the same comparability group can rest on different samples, periods and methods. The group prints those differences and stops there. No row is described as right, best or settled.

It does not grade studies. publication_type records what a document is, not how good it is. A working paper here has not been peer reviewed, and the table draws no conclusion from that.

Blank is not zero. Where the data, period, method or caveat column is empty, our copy does not print that thing beside the estimate. An empty public URL means no URL is recorded in a repository manifest, which says nothing about whether one exists.

Quotation is not endorsement. The caveat column quotes the document's own warning. This archive adds none of its own and corrects none of the numbers it quotes.

## What it does not measure

Working papers are revised. Each row describes the copy held here, and the version column prints what that copy says about itself. Where two versions of one study are held, both appear and the comparability file says so rather than choosing.

What is missing is listed. Our missing-documents list names 91 documents that the archive's acquisition manifest or a page on this site cites; we do not have a copy of any of them.

What the documents say about fraud is a separate table. Pandemic fraud estimates records what each published fraud figure measures; this table is about the program's outcomes.

## Citation

What the held studies say about PPP outcomes, compiled from documents held by Pandemic Darlings and read on 2026-09-20.

## Rights

Quotations remain the works' own; attribution to Pandemic Darlings requested for the compiled table.