Private credit data: separating debt revisions from new borrowing

The Fed’s October 16, 2026 note describes private credit’s integration into September’s Financial Accounts, combining reclassification of existing claims with the inclusion of previously missing instruments. Fed note.

The analytical task is to decide what each number can actually establish. A change in the map may reveal an exposure that deserved attention all along. It cannot, by itself, tell us when a borrower received cash. The following framework is OMN’s analysis; all numerical examples are hypothetical accounting illustrations, not estimates of the market.

Two comparisons that answer different questions

Start with a company that owes 100 units. An old statistical table places 70 under loans and 30 under other liabilities. A revised table moves 20 from the second category into the first. Recorded loans become 90; total liabilities remain 100. The borrower has not received another 20, signed a new contract or acquired more repayment capacity merely because the labels changed.

Now change the example. The company actually draws a new loan of 10 and keeps the proceeds as cash. Assuming nothing else changes, its liabilities rise to 110 and its cash rises by 10. This is a financing event with an identifiable transaction. Both examples produce a larger loan number, but their implications for demand, debt service and liquidity differ.

There is a third possibility: a statistical system previously omitted a claim altogether. Adding it can increase the reported total even without a new transaction on the revision date. It would therefore be equally misleading to insist that every statistical revision must leave every aggregate unchanged.

The essential distinction is between comparing two publications for the same quarter and comparing two quarters within a consistent publication. The first can reveal a revision. The second can help investigate economic change. Splicing an old-release starting value to a new-release ending value mixes these questions and may create a growth rate with no coherent interpretation.

A balance is not a flow

The Financial Accounts distinguish transactions, revaluations and other volume changes. A change in the amount outstanding therefore need not measure new lending. Z.1 concepts.

Consider another invented loan book. It opens at 100, advances 15 and receives 12 of principal repayments, with no other changes. Its closing balance is 103. Gross lending was 15; the net increase was three. Calling all three quantities “credit growth” hides the question being answered.

Next suppose a separate fair-value portfolio rises from 100 to 103 solely because its valuation changes. The same endpoints now describe no new lending at all. A useful chart needs more than a line: it needs the unit, valuation basis, population and treatment of repayments. Otherwise, visual precision can conceal conceptual ambiguity.

For an editor or analyst, the practical test is simple. Before writing that borrowing accelerated, locate the transaction measure or construct a documented reconciliation. If only balances are available, say what changed in those balances and leave the unobserved financing story open.

Fund assets and business loans need different denominators

Imagine a fund with 100 of assets: 70 in loans, 20 in equity and 10 in cash. Suppose it finances those assets with 25 of borrowing and 75 of investor equity. “Assets of 100,” “loans of 70” and “net assets of 75” are all valid descriptions of the same example. They are not interchangeable measures of credit supplied to businesses.

An additional definition issue appears in regulatory reporting: SEC Form ADV includes uncalled private-fund commitments when calculating regulatory assets under management. SEC instructions.

That definition should not be transferred automatically to a differently constructed AUM series. The first question is always what the particular publisher counts. An investor’s promise to contribute, a fund’s cash balance, an outstanding loan and the fund’s equity value occupy different places in an accounting story.

In our example, adding a hypothetical uncalled commitment of 10 does not cause the borrower to receive 10. Nor does a rise in the fund’s equity holdings establish a larger loan book. A comparison between two industry-size estimates should therefore begin with a scope reconciliation, before attempting to explain the gap as growth or disagreement.

Geography follows the question

Suppose Fund A is domiciled domestically and lends 60 to domestic companies and 40 abroad. Fund B is domiciled abroad and lends 30 to domestic companies. Counting the domestic fund’s entire book gives 100. Counting the two funds’ claims on domestic companies gives 90. Neither answer is defective; each describes a different perimeter.

This small example shows why a fund’s domicile cannot substitute for the location of its borrowers. A question about domestic business debt calls for a borrower-based boundary. A question about domestically organized intermediaries calls for an institution-based boundary. A global manager’s marketing footprint answers neither question by itself.

The same discipline applies over time. An acquisition that brings an existing portfolio into a surveyed group could enlarge that group’s reported holdings without increasing the combined borrowers’ debt. Any claim about expansion should identify which boundary stayed fixed.

Follow the chain without counting the borrower twice

The IMF’s April 2024 briefing highlighted leverage across borrowers, funds and investors, alongside the need for better data. IMF briefing.

To explore the accounting implication, imagine a second fund lending 100 to companies, financed by investor equity of 75 and a bank loan of 25. Adding the bank’s 25 claim to the fund’s 100 loans produces 125 of claims at different stages. It does not establish 125 of borrowing by those companies. The intermediate claim still matters, but it answers a question about financing relationships rather than the companies’ loan total.

Under a deliberately simplified assumption that the bank debt stays at 25, a 10 loss in fund assets reduces investor equity from 75 to 65. That arithmetic alone cannot tell us whether a covenant is breached, a lender demands repayment or an investor faces a cash call. Those outcomes require contract terms, cash resources and timing, not just a larger aggregate.

This is where measurement can improve an argument without completing it. A creditor map identifies where to ask the next question. A claim about stress transmission needs evidence about the connection itself.

A better standard for the next headline

A defensible reading should preserve the publication date and observation date, identify the measured population, separate assets from loans and explain any change of definition. It should also distinguish a newly observed relationship from a newly created one.

The goal is not to make a large number look smaller. It is to attach the right verb to it. A liability can be reclassified, discovered, revalued, repaid or newly incurred. Each verb leads to a different economic argument. Choosing the verb carefully is the first step toward understanding the credit cycle rather than merely describing a changing spreadsheet.