Pandemic Darlings The pandemic economy, in original documents
Home Data downloads PPP lender processing fees by lender type, modeled

Derived table · Paycheck Protection Program

PPP lender processing fees by lender type, modeled

Fintech lenders earned a modeled $2,272 in SBA processing fees per PPP loan, community banks $4,005 and the 51 largest banks $3,918. The population is the 11,468,206 disbursed loans in SBA's loan-level file of September 30, 2024. The fee schedule paid by loan size, and the fintech type made the smallest loans: 93.5% of its 3,049,574 loans were $50,000 or less, against 67.9% at the big banks and 69.3% at community banks. The six types together earned $38.24 billion, the same total as the per-lender table this page groups.

Rows6
Columns23
Files1 · 2 KB
FormatCSV
Source vintageSBA PPP FOIA 2024-09-30, grouped from the per-lender modeled fee table
Published byPandemic Darlings
Acquired / built2026-10-16
Built fromPPP lender fees by lender and loan size, from SBA PPP FOIA loan-level data
SHA-25605e34e7cf01b3bf569f18831ed5e75ec30083c4f0fc2726592f704d87db70056
RightsDerived from U.S. government public data and published fee rules; attribution to Pandemic Darlings requested.
Dictionarydictionary.csv

Download CSV 2 KBData dictionary CSVAt a glancePreviewColumnsHow to useBefore you quote a row

At a glance

Measured on the file on 2026-10-16; the checksum above is the file these numbers describe.

Rows6
Columns23
Keylender_type · unique
Cells left blank0 of 138 · 0.0%
Lender types6 · 4,689 lenders
Modeled fees$38.24B
Fee per loanfintech $2,272 · big bank $3,918 · community bank $4,005
Fintech type18.1% of fees · 26.6% of loans
Loans of $50,000 or lessfintech 93.5% · CDFI 98.7% · big bank 67.9%
Fintech-type lenders with a bank charter word in the name14 of 31 · $2.52 billion of fees

What it is

One row for each of six lender types, grouped from the 4,689 originating lenders in the per-lender fee table. The types are community bank (3,512 lenders), big bank (51), fintech (31), CDFI (20), credit union (946) and other (129, including the one row with a blank lender name). Each row sums the type's disbursed loans, current approved dollars and modeled processing fees, and splits the fees by draw and by four loan-size bands. Fee per loan divides fees by loans.

Community banks earned the most, $17.06 billion, 44.6% of the total; the big banks $10.75 billion, 28.1%; the fintech type $6.93 billion, 18.1%; CDFIs $2.27 billion, 5.9%; credit unions $1.03 billion, 2.7%; other lenders $200.79 million, 0.5%. The fee-per-loan order is close to the reverse: community bank $4,005, big bank $3,918, other $2,824, credit union $2,588, CDFI $2,396, fintech $2,272. CDFIs made 98.7% of their loans at $50,000 or less and took 95.7% of their fees from that band. The same two shares for the other types: fintech type 93.5% and 79.6%; big banks 67.9% and 25.3%; community banks 69.3% and 28.5%; credit unions 81.7% and 46.7%; other lenders 82.6% and 51.6%.

The type is a channel label, not a charter. The fintech list holds 14 lenders whose names carry a bank charter word, Cross River Bank, Customers Bank, Celtic Bank Corporation and WebBank among them. They account for 1,139,827 loans and $2.52 billion of the type's $6.93 billion, at $2,211 a loan. The 17 fintech lenders without one account for 1,909,747 loans and $4.41 billion, at $2,308 a loan. The companion file flags the charter word on every lender, so the table can be regrouped by charter.

How it was built

Each lender's loans, dollars and fees come from the per-lender fee table, which prices every disbursed loan in SBA's September 30, 2024 release on its current approved amount under the schedule for its draw and approval date. The 2020 CARES Act schedule paid 5% up to $350,000, 3% above that and below $2 million and 1% at $2 million and above. The 2021 Economic Aid Act schedule paid the lesser of $2,500 or 50% on loans of $50,000 or less, then the same 5% and 3%. The 4,689 rows grouped here match that table to the cent on all 14 shared columns.

The lender type is a keyword rule applied to the lender's name as the SBA file prints it, 104 keywords in all. They break down as 39 naming fintech lenders and the chartered banks the rule groups with them and 36 naming the largest banks. The rest: 19 naming CDFIs and mission lenders, 5 naming Farm Credit lenders as community banks and 5 for credit unions. A name none of them matches is a community bank if it contains bank, banc, trust, savings or national association, and other if not. Re-applying the rule to every name reproduces the type on all 4,689 rows. The companion file's charter flag tests the name for Bank, National Association or SSB.

Preview

lender_typecategory_codelendersloansamount_usdmodeled_fees_usdfee_per_loan_usdshare_of_fees_pctshare_of_loans_pctfees_2020_first_draw_usdfees_2021_first_draw_usdfees_second_draw_usdloans_50k_or_lessloans_50k_to_350kloans_350k_to_2mloans_2m_and_abovefees_50k_or_less_usdfees_50k_to_350k_usdfees_350k_to_2m_usdfees_2m_and_above_usdshare_loans_50k_or_less_pctshare_fees_50k_or_less_pctlenders_with_charter_word_in_name
Community bankCOMMUNITY_BANK35124258532401414240549.0717055395726.334004.9944.6137.139567536027.412086437849.465401421849.4529525741078346209202184104860495914.486900724522.804563117666.65731057622.4169.3328.503380
Big bankBIG_BANK512744317275216797853.2410753259672.113918.3728.1223.936699252339.14640098763.983413908568.981862174737542129195154062720745704.214525045795.362876714808.47630753364.0767.8625.3047
FintechFINTECH31304957476482901318.766928802167.722272.0618.1226.591256187733.153445196702.042227417732.532851697180261168317855517922543.441028246500.80348145353.1034487770.3893.5179.6414
CDFICDFI2094563016661068616.592265307612.452395.555.928.2556684397.901932784318.89275838895.669337321055112271202168427070.6362982915.3628739070.385158556.0798.7495.720
Credit unionCREDIT_UNION94639904619425858703.481032711014.992587.952.703.48513174952.93179291733.27340244328.79326101653257066554482657525.64380294035.62147064466.2722694987.4681.7246.740
OtherOTHER129711073400202002.41200787611.022823.740.530.6274462584.3760614044.3365710982.315871111123117994103634921.7668033617.5524929194.484189877.2382.5751.610

First 6 of 6 rows · cells longer than 60 characters are cut with an ellipsis · shown as text, unformatted.

Files

FileRowsSizeSHA-256
ppp-lender-fees-by-lender-type.csv
CSV, UTF-8, header row
61,691 B05e34e7cf01b3bf569f18831ed5e75ec30083c4f0fc2726592f704d87db70056Download ↓
dictionary.csv
data dictionary, 23 rows
23—built with this pageDownload ↓

Columns

ColumnType, measuredBlankDistinctRange / valuesMeaning
lender_typetext06Community bank (1), Big bank (1), Fintech (1)Community bank, Big bank, Fintech, CDFI, Credit union or Other.
category_codetext06COMMUNITY_BANK (1), BIG_BANK (1), FINTECH (1)The same type as the keyword rule names it: COMMUNITY_BANK, BIG_BANK, FINTECH, CDFI, CREDIT_UNION, OTHER.
lendersinteger0620 – 3.5K · median 90Originating lenders of this type; the six rows sum to 4,689.
loansinteger0671.1K – 4.3M · median 1.8MDisbursed loans originated by lenders of this type.
amount_usdnumber063.4B – 401.4B · median 48.0BCurrent approved amount of those loans, dollars.
modeled_fees_usdnumber06200.8M – 17.1B · median 4.6BModeled processing fees on those loans, dollars (computed from loan amounts, not fees SBA paid).
fee_per_loan_usdnumber062.3K – 4.0K · median 2.7Kmodeled_fees_usd divided by loans.
share_of_fees_pctnumber060.53 – 44.61 · median 12.02modeled_fees_usd as a percentage of the $38,236,263,804.62 total.
share_of_loans_pctnumber060.62 – 37.13 · median 16.09loans as a percentage of 11,468,206.
fees_2020_first_draw_usdnumber0656.7M – 9.6B · median 884.7MModeled fees on first-draw loans approved in 2020.
fees_2021_first_draw_usdnumber0660.6M – 3.4B · median 1.3BModeled fees on first-draw loans approved in 2021.
fees_second_draw_usdnumber0665.7M – 5.4B · median 1.3BModeled fees on second-draw loans (SBA processing method PPS).
loans_50k_or_lessinteger0658.7K – 3M · median 1.4MLoans of $50,000 or less.
loans_50k_to_350kinteger0610.6K – 1.1M · median 122.8KLoans over $50,000 up to $350,000.
loans_350k_to_2minteger061.2K – 209.2K · median 11.9KLoans over $350,000 and under $2 million.
loans_2m_and_aboveinteger0694 – 18.4K · median 669Loans of $2 million and above.
fees_50k_or_less_usdnumber06103.6M – 5.5B · median 2.4BModeled fees on loans of $50,000 or less.
fees_50k_to_350k_usdnumber0663.0M – 6.9B · median 704.3MModeled fees on loans over $50,000 up to $350,000.
fees_350k_to_2m_usdnumber0624.9M – 4.6B · median 247.6MModeled fees on loans over $350,000 and under $2 million.
fees_2m_and_above_usdnumber064.2M – 731.1M · median 28.6MModeled fees on loans of $2 million and above.
share_loans_50k_or_less_pctnumber0667.86 – 98.74 · median 82.14loans_50k_or_less as a percentage of loans.
share_fees_50k_or_less_pctnumber0625.30 – 95.72 · median 49.17fees_50k_or_less_usd as a percentage of modeled_fees_usd.
lenders_with_charter_word_in_nameinteger040 – 3.4K · median 7Lenders of this type whose SBA name contains Bank, National Association or SSB.

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

How to use it

Specific to this file, measured on it: keys, joins, encodings, units, and the values that will trip a naive count.

  • Six rows, sorted by modeled_fees_usd, largest first. The key is the type label; category_code is the same type in the form the keyword rule writes.
  • The splits add up. The four loans_* band columns sum to loans, and the four band fee columns and the three draw fee columns each sum to modeled_fees_usd within a cent, on every row.
  • Every lender's type is in the companion file. ppp-lender-fees-by-lender-type-lenders.csv has 4,689 rows: lender name, type, charter-word flag, loans, dollars, fees, fee per loan and the $50,000-or-less band. Group it by lender_type to rebuild this table, or by charter_word_in_name to regroup by charter.
  • Money columns are dollars with two decimals. Divide by 1e9 for billions; round before quoting. LF line endings, UTF-8, 1,691 bytes.

Before you quote a row

  • Modeled, not observed. Fees are computed from loan sizes under the published schedules; fees on loans later canceled or found ineligible are not netted out, and the SBA's actual payments are not in the file.
  • Names, not charters. The type follows the lender's name against a keyword list. A chartered bank the keyword list names sits in the fintech type, and a lender the list does not name falls to community bank or other by the words in its name. The companion file's charter flag is a name test too.
  • Originating lender only. Fees are credited to the lender that originated the loan. Platforms, agents and service providers paid out of those fees do not appear, and a loan sold or serviced elsewhere stays with its originator.
  • Population. 11,468,206 disbursed loans, the 11,468,210 rows of the September 30, 2024 release less 4 rows the disbursement filter removes, the same population as the per-lender table and the fee-schedule table.
Cite this table. Name the source vintage; the numbers depend on it.PPP lender processing fees by lender type, modeled on SBA PPP FOIA loan-level data (2024-09-30 vintage). Pandemic Darlings, https://pandemicdarlings.com/data/ppp-lender-fees-by-lender-type/, SHA-256 05e34e7cf01b3bf569f18831ed5e75ec30083c4f0fc2726592f704d87db70056
Back to top