Private credit is a community bank problem

Private credit gets discussed at community banks, when it gets discussed at all, as somebody else's story. The funds are raising money to lend to mid-market companies at scale, in deals larger than most community institutions would write, in structures they would not use. It reads as an event in a different part of the market.

That reading misses two things, and the second one is the part that will eventually show up in an examination.

What the bank balance sheets already show

Start with what has happened to commercial lending as a share of what banks hold.

Commercial and industrial lending, as a share of all bank credit

Percent of total loans and leases, all commercial banks, annual average

Total bank credit grew about a third across that period. Commercial and industrial lending grew roughly a seventh, and has been falling outright since 2023. As a share of what banks hold, it has gone from very nearly a quarter to about a fifth.

Some of that is the pandemic distortion washing through, and some is rate-driven demand. But the direction is persistent, and it has a straightforward reading: the business-lending franchise is a shrinking part of what a bank does, and something else is doing it instead.

The adverse selection problem

Competition for borrowers is the obvious concern and the less serious one. A fund that can move faster, hold more, and write a covenant package a bank cannot is going to win deals. Losing a deal on price or terms is an ordinary commercial outcome.

The problem is which deals get lost. Private credit is at its most competitive where the borrower is largest, best documented, most sponsor-backed and easiest to underwrite: precisely the credits a bank would most like to keep. It is least interested in the smaller, messier, relationship-dependent borrower whose file takes local knowledge to read.

That is a sorting mechanism, and it runs in one direction. Over several years it does not simply shrink the commercial book; it changes what is in it. The bank retains the credits that a fund with a spreadsheet and no local presence could not get comfortable with, which is sometimes a genuine informational edge and sometimes exactly what it sounds like.

The distinction matters enormously and it is answerable. Run the loss experience on originations you won against originations you lost, over a full cycle, and you find out which one you have. Very few institutions do this, because losing a deal generates no file.

Losing a deal on price is an ordinary outcome. Losing every deal of a particular quality, for years, is a change in the risk profile of what you kept.

The exposure that comes back, and who it comes back to

The next move in this argument is usually that the funds need financing, banks provide it, and the exposure returns through the back door as a loan to a nondepository financial institution. Subscription lines, NAV facilities, warehouse lines, leverage on the vehicles themselves.

All of that is real. It is also, at community bank scale, almost entirely somebody else's business, and it is worth being precise about that rather than borrowing an alarm that belongs to a different tier of the industry.

Lending to nondepository financial institutions, by size of bank

Loans to NDFIs as a share of total loans and leases

The gradient is about as clean as anything in bank data gets. At the largest institutions, lending to nonbank financial firms is close to a seventh of the loan book and 97% of them do some. In the $500m–$1bn band it is roughly a third of one percent, and 86% of those banks report none at all.

So the fund-financing channel is a large-bank story. A community bank that loses a commercial borrower to a private credit fund does not typically end up financing that fund. It simply loses the borrower.

That is a different problem, and in one respect a worse one. The large bank at least retains an economic interest in the credit and gets paid for it. The community bank gets the sorting effect with none of the compensation: the book it keeps is reshaped by which credits it lost, and there is no offsetting position anywhere on its balance sheet.

The exposure that does come back arrives indirectly, and it is easy to miss because nothing on the balance sheet is labelled for it:

• Through your borrower's competitors. A mid-market firm financed by a fund at terms and speed you cannot match is competing with the borrower you kept. That is a credit-quality question about your own book with no trade attached to it.
• Through the local economy. In markets where fund financing has displaced bank lending, the credit cycle is now partly driven by an investor base with redemption windows and mark-to-model valuations rather than by lenders with deposits and an examiner.
• Through the minority that do have it. Fourteen per cent of banks in that band report some NDFI exposure. For those institutions the earlier concerns are live and specific: one counterparty, correlated legs, somebody else's underwriting, collateral carried at a mark. They should be measured as a concentration, not filed as a financial-institution line.

What the sorting leaves behind

There is a question at the centre of this that a community bank cannot currently answer about itself: whether the commercial book it retains reflects an informational edge or an adverse selection.

Both stories fit the same facts. A bank that lends to borrowers a fund cannot underwrite from a spreadsheet may be exercising exactly the local knowledge that justifies its existence. A bank that keeps only what nobody else wanted looks identical from the inside, and for several years performs identically too.

The evidence that would separate them is the loss experience of the credits it lost against the credits it won. That evidence does not exist anywhere, because a lost deal generates no file and there is no field on any core system to put it in. Which means the question is answerable in three years only by institutions that decided today it was worth recording, and the ones with the most to learn from the answer are, by construction, the least likely to have started.