DOJ pandemic-fraud press-release index

Download CSV 3.0 MBData dictionary CSVAt a glancePreviewColumnsHow to useBefore you quote a row

At a glance

Preview

First 10 of 3,093 rows · cells longer than 60 characters are cut with an ellipsis · shown as text, unformatted.

Files

Columns

Type, blanks, distinct count, and range are measured on the file. The same table as CSV: dictionary.csv.

How to use it

• uuid is the key (0 duplicates) and is the join to the identity-theft classification (all 1,977 of its rows are here) and to record_id values prefixed doj: in the sentencing table. 159 rows share a headline with another row; join on the uuid, never the title.
• programs has two vocabularies. On the 778 rows whose fact_source is canonical it is the sentencing table's program_category (PPP, PPP/EIDL, UI, …); on the 2,315 parsed rows it is a comma-separated list of the phrases matched, so one program can appear twice as its full name and its abbreviation (the most common value is PAYCHECK PROTECTION PROGRAM, PPP). Map both forms before counting.
• dollar_amount depends on fact_source. On canonical rows it is the case amount the sentencing table records (its amount_basis says what the figure is); on parsed rows it is the largest dollar figure parsed from the release. Data release v17 checked the 46 parsed figures of $1 billion or more against the release text: 45 were program totals, nationwide tallies or other context figures. Each now carries the case figure from the same release or is blank. The one left is the more than $2.9 billion three defendants were charged with falsely seeking from the IRS. It also corrected three smaller figures marked reliable in the identity-theft classification. They were a $903 million District of Oregon total (the release's own case is $178 million) and two nationwide totals, $836 million in a Maryland roundup and $569 million in a Tulsa plea release. Data release v18 corrected 177 more parsed figures below $1 billion that were not the case figure of their release, such as a statutory maximum fine or income claimed on an application. Smaller context figures remain on other parsed rows: the largest, $832 million, is the capital a PPP lender borrowed through the PPP Liquidity Facility, in a 2023 release about its chief executive's guilty plea.
• 2 parsed rows have a sentence while case_type is charged or indicted: both read pending. On canonical rows the sentence is the one imposed. Filter on case_type = sentenced before using parsed sentence text.
• Blanks: date is empty on 6 rows, district on 271, defendants on 689, restitution on 2,789. Blank means not found in the text, not zero.
• LF line endings, UTF-8 with non-ASCII characters (curly quotes, accented names), no byte-order mark. The longest cell is 1,636 characters (summary).

Before you quote a row

• Names are headline fragments. defendants is whatever the headline pattern captured; it can be a description (Cheshire Man) rather than a name, and multi-defendant releases list several or none.
• Sentences are recorded only where the release reports a term imposed on the person it names. Charge and plea releases carry no sentence, and releases about several defendants give the range.
• Roundups, reports and notices have no case amount. Where such a release gave a figure of $1 billion or more, dollar_amount is blank (data release v17).
• This is an index, not a docket. Go to the charging document or judgment before quoting a fact about a case; the court filings section holds many of them.
• Headline names and dollar figures require case-level verification.