Pandemic Darlings The pandemic economy, in original documents
Home Key findings

Original research

Key findings

Three results computed from the public records on this site. Each gives the number, who it covers, how it was measured, what it does not show, the file behind it and the code that rebuilds it.

1 · PPP lenders

Which lenders' PPP loans went bad

Among the 67 lenders that made at least 20,000 PPP loans, the share SBA lists as charged off runs from 0.14% at Pinnacle Bank to 24.3% at Harvest Small Business Finance. Ten lenders made 76.7% of all the PPP loans charged off, and 27.2% of all PPP loans.

0.14%–24.3%charge-off rate across the 67 largest lenders
2.2%the median lender of the 67
76.7%of charge-offs made by ten lenders
Population
The 67 originating lenders with 20,000 or more loans: 7,049,210 of the 11,468,206 loans in SBA's file (61.5%) and 608,450 of its 643,039 charge-offs (94.6%). The 220 lenders with 5,000 or more loans run from 0.00% to 24.3%. The ten with the most charged-off loans, in order: Harvest Small Business Finance, LLC; Cross River Bank; Benworth Capital; Capital Plus Financial, LLC; Prestamos CDFI, LLC; BSD Capital, LLC dba Lendistry; Fountainhead SBF LLC; Kabbage, Inc.; Customers Bank; Itria Ventures LLC.
Method
Charged off is a status SBA gives some loans in its loan-level release of September 30, 2024. The lender is the one SBA records as originating the loan; servicing later changed hands for some loans. Loan mix explains part of the gap. Small loans, sole proprietors' loans and loans approved late in each round charged off far more often, so each lender was also compared with other lenders' loans of the same draw, size band, borrower type, approval month and state. The 67 still run from 0.08 times the comparable rate (Pinnacle Bank) to 2.6 times (Kabbage); 20 are at half or less and 2 at double or more. Within loans of $20,833.33 or less the range is 0.15% to 24.9%. Counted by servicing lender instead, the 62 servicers with 20,000 or more loans run from 0.14% to 23.8%.
By draw and month
First-draw loans charged off at 6.2%, second-draw loans at 4.0%. Timing mattered more: 0.7% of loans approved in April 2020, 13.6% of August 2020's, 1.4% of January 2021's and 13.5% of April 2021's.
Caveat
A charge-off is not a fraud finding. The status covers loans that defaulted and loans that were never forgiven, for whatever reason, and SBA's file does not say which. 169,086 loans (1.5%) carry a status that names a Freedom of Information Act exemption instead of an outcome; counting all of them as charged off would move the range to 0.30%–31.1%.
Data
PPP charge-offs by originating lender: one row per lender, all 4,689, with draw, size, borrower-type and comparison columns (CSV). Source: SBA PPP loan-level release, September 30, 2024.
As of
SBA's release dated September 30, 2024. Computed September 23, 2026.
Reproduce
reproduce_chargeoffs.py rebuilds every number in this block from SBA's 13 files. check_chargeoffs.py, written separately with pandas, recomputes them: 54 of 54 agree. Both are published with the table, on its data page.
The 67 largest PPP lenders by charge-off rate — lenders with 20,000 or more loans, counted by the share of their loans SBA lists as charged off in its September 30, 2024 release.
Lenders by charge-off rateHorizontal bars, one per band of charge-off rate, length equal to the number of the 67 largest PPP lenders in that band.ratelendersunder 1%191–2%102–3%113–5%85–10%410–15%815–20%420–25%3
The numbers behind this chart
Charge-off rateLenders
under 1%19
1–2%10
2–3%11
3–5%8
5–10%4
10–15%8
15–20%4
20–25%3

2 · Justice Department

What the Justice Department announces: fewer new charges, more settlements

The Justice Department put out half as many announcements of new pandemic relief fraud charges in 2025 as in 2021, 115 against 235, while its announcements of civil settlements and judgments rose from 7 to 86. So far in 2026, settlement announcements (71) have passed new-charge announcements (69).

235 → 115new-charge announcements, 2021 to 2025
7 → 86civil-resolution announcements
$1.5 million → $306.8 millionin settlements and judgments announced
Population
The site's collection of Justice Department press releases dated October 25, 2019 to September 22, 2026, 135,785 in all. 4,996 name a pandemic relief program; 4,734 once repeats of one event are merged. The department's output as a whole did not shrink: 19,247 releases in 2021, 19,466 in 2025.
Method
A release counts if it names a relief program: PPP, COVID EIDL, pandemic unemployment benefits, the Employee Retention Credit, the restaurant and venue grants, the Provider Relief Fund and others. What it announces is read from the headline. Round-ups, takedown summaries and office news count as neither charges nor settlements. The rules were checked against 415 releases read by hand: 98 of the 102 releases the rules called charging announcements were, and all 100 called civil resolutions were. A release posted twice for one event counts once. The charging announcements name about 539 people in 2021 and 304 in 2025, 2.3 and 2.6 per release.
Caveat
These are announcements, not caseloads, and the department announces a changing share of its work. Its COVID-19 Fraud Enforcement Task Force counted about 3,500 people charged through 2023, under a wider definition of pandemic fraud; the charging announcements name about 2,100 over the same years. Its year-end False Claims Act figures report about 270 civil resolutions over PPP loans in fiscal 2023 and more than 250 pandemic-related ones in fiscal 2024; it announced 20 and 42. The counts do not show that the department charged fewer people or settled more cases. 2026 runs to September 22.
Data
Justice Department pandemic-relief announcements by year and stage (CSV), with the 4,996 releases behind it. Sources: DOJ press releases, 2020–2026; COVID-19 Fraud Enforcement Task Force 2024 Report; the department's False Claims Act statistics releases of February 22, 2024 and January 15, 2025.
As of
Releases through September 22, 2026: the site's copy through September 15, plus the department's public press-release list read on September 23, 2026.
Reproduce
reproduce_enforcement.py rebuilds the table. check_enforcement.py recounts it with separately written code: 238 of 238 cells agree. The 415 hand-read releases and the script that scores the rules against them are published with the table, on its data page.
New charges against civil settlements, as announced — Justice Department press releases on pandemic relief fraud, repeats merged, by year. *2026 runs to September 22.
Justice Department pandemic-relief announcements by yearTwo bars per year, 2020 to 2026: announcements of new criminal charges and announcements of civil settlements or judgments. 2026 runs to September 22.new chargescivil settlements20201670202123572022196112023189262024181452025115862026*6971
The numbers behind this chart
YearNew-charge announcementsCivil-resolution announcements
20201670
20212357
202219611
202318926
202418145
202511586
2026 (to Sept. 22)6971

3 · SBA programs

Businesses found in more than one SBA pandemic program

Between 77,155 and 91,448 businesses appear in two or more of PPP, COVID EIDL, the Restaurant Revitalization Fund and the Shuttered Venue Operators Grant. The low end needs a matching ZIP code on every link; the high end takes a name-and-state match alone, generic names included.

77,155matches confirmed by ZIP code
91,448all matches
$51.7B–$60.9Bmatched program dollars, low and high end
Population
Recipients in SBA's award files for the four programs, matched across programs on a normalized business name plus state.
Method
A link between two programs is confirmed when name, state and a five-digit ZIP code agree, and probable on name and state alone. 77,155 businesses are confirmed on every link, 9,013 more are probable, and 5,280 more have names too generic to trust. The counts were taken from the site's matched-business table and recounted from its second file of the same matches.
Caveat
A name match is not proof that two awards went to the same business, and drawing on more than one program is not in itself a problem. The one pairing the program rules did not allow, an RRF grant plus an SVOG grant, appears once in the file.
Data
Cross-program recipient matches; the range as a table: program-overlap-range.csv.
As of
The matched-business table as served on September 23, 2026.
Reproduce
reproduce_overlap.py rebuilds the range; check_overlap.py recounts it from the second file: 7 of 7 agree.
Businesses in two or more SBA pandemic programs, by strength of match — PPP, COVID EIDL, RRF and SVOG recipients matched on normalized name and state.
Matched businesses by strength of matchOne bar split into three parts: businesses matched with a ZIP code on every link, with name and state only, and with a generic name.77,155 ZIP code confirmed9,013 name and state only5,280 generic name
The numbers behind this chart
MatchBusinesses
ZIP code confirmed77,155
name and state only9,013
generic name5,280
Back to top