A cash lender accepting negative interest on a Treasury repo may be paying for something valuable: temporary use of the exact bond it needs to deliver. Reading that rate as a simple price of money misses the collateral shortage embedded in the bargain.[2]

The New York Fed's September 28, 2026 explainer makes the distinction timely. Repo serves both institutions seeking funding and traders seeking particular securities. Those motives can produce very different prices inside the same market.[1] The useful question is how far a bond's repo rate sits below comparable general funding rates—and what happens when more of that bond becomes available.

The lender wants the bond

A repurchase agreement begins with a security changing hands for cash, alongside an agreement to reverse the exchange later at a specified price. Economically, one party borrows money against securities; the other lends money and temporarily receives them.[1]

In a general collateral, or GC, transaction, the cash provider accepts securities meeting agreed eligibility criteria. Its main concern is the return on its money. In a transaction arranged around a particular security, the cash provider may care intensely about which bond arrives.[1]

Imagine a dealer that has sold a Treasury note and must deliver it to a customer. A similar note cannot discharge that particular obligation. The dealer can obtain the required security through repo, passing cash to its owner and accepting a lower return on that cash as the price of access. Competition for the same scarce issue can push the repo rate below zero.[2]

This is what traders mean when a security trades special. Merely naming a specific bond in a contract is insufficient: the defining evidence is a repo rate below the general collateral rate. The gap measures an implicit securities-borrowing fee.[2]

A negative rate has a positive price

Consider a deliberately simplified example, using hypothetical rates rather than current market quotes. A dealer can lend $1 million overnight at a 3.6% annualized GC rate. Obtaining the particular Treasury it needs instead requires lending the same amount at −0.9%. Assume the transactions have comparable counterparty and settlement terms, run for one calendar day, and have no additional fees. Use the standard dollar money-market convention of a 360-day year.[4]

The economic cost of obtaining the bond is the difference between those cash returns:

$1,000,000 × (0.036 − (−0.009)) ÷ 360 = $125.

That cost includes both interest the dealer gives up and the reduction in cash it receives when the special repo unwinds. A statement showing only the negative interest charge would understate the opportunity cost. For the security's owner, the other side of the bargain is unusually cheap funding.

These are two prices packed into one quoted rate: a return for providing cash and a charge for obtaining a useful security. Our calculation isolates their difference. It does not establish that the dealer's overall trade is profitable; that depends on the customer transaction, the hedge and the dealer's own funding costs.

It also explains why crossing zero is an incomplete signal. A specialness spread can widen while both quoted rates remain positive. Conversely, the special rate can turn negative because general funding rates have fallen, even without a larger scarcity premium.

Why SOFR can look untroubled

The Office of Financial Research documented this combination in its study of negative bilateral repo rates during 2021. Broad conditions pushed general collateral rates down, while demand for particular securities pulled some bilateral rates further below them. The study supports separating these effects; it does not establish that a comparable squeeze is occurring today.[3]

SOFR, the Secured Overnight Financing Rate, is designed to measure the general cost of financing Treasury securities overnight. Its inputs include GC transactions and trades from the Fixed Income Clearing Corporation's delivery-versus-payment, or DVP, repo service, where counterparties can specify securities.[4]

Under the current methodology, the New York Fed removes the lowest-rate 20% of DVP transaction volume, after relevant affiliated-trade exclusions, to reduce the influence of specials. It then calculates SOFR as a volume-weighted median across the included transactions. The filter removes some special trades, not all of them.[4]

The implication is practical: a calm SOFR reading cannot establish that every Treasury issue is easy to borrow. The benchmark deliberately dampens the very prices that can reveal a local collateral shortage. Equally, an isolated negative repo quote cannot establish that cash has become cheap throughout the system.

What would overturn the scarcity reading?

The strongest counterweight is the ordinary price of money. The OFR's historical evidence shows that general funding conditions and security-specific demand can move together.[3] Comparing unmatched maturities, counterparties or contractual terms introduces further ambiguity; the New York Fed identifies all of these as influences on repo pricing.[1]

For a particular episode, the scarcity explanation would fail if the supposedly special rate simply followed a comparable GC rate downward, with no unusual or widening discount. That observation would support a broad funding explanation. A negative sign alone cannot choose between them.

Supply offers another test. Treasury reopenings add more of a previously issued security.[6] If newly settled bonds become available for lending, a smaller specialness spread would support the shortage interpretation. Persistence would call for a closer look at demand and lendable inventory: additional bonds outstanding need not mean additional bonds offered in repo. This is a diagnostic inference, not a promise that every reopening will normalize financing.

Three events make the explanation testable:

• The next daily repo-rate publication: compare SOFR and the Broad General Collateral Rate for the funding backdrop, then compare the issue-specific quote with suitably matched GC terms. Public benchmarks alone cannot identify the scarce bond.[4]
• The next reopening settlement for the issue being monitored: check whether its discount narrows as supply arrives. The settlement date is when the extra securities become usable; an auction announcement alone does not deliver them.[6]
• The next Thursday primary-dealer statistics release: inspect Treasury settlement fails for signs that borrowing friction is disrupting delivery. These are weekly aggregates and cannot, by themselves, attribute trouble to a particular issue.[5]

Sources

1. Gara Afonso, Jun-Davinci Choi, Gonzalo Cisternas and Will Riordan, “Who's Borrowing and Lending in Repo Markets?” Federal Reserve Bank of New York, September 28, 2026 — repo mechanics, participant motives and pricing differences.
2. International Capital Market Association, “What is a ‘special’ in the repo market?” — specialness, negative rates, delivery demand and the implicit borrowing fee.
3. Samuel J. Hempel and R. Jay Kahn, “Negative Rates in Bilateral Repo Markets.” Office of Financial Research, September 27, 2021 — historical evidence separating broad funding conditions from collateral demand.
4. Federal Reserve Bank of New York, “Additional Information about Reference Rates Administered by the New York Fed” — current SOFR inputs, DVP filtering, median calculation, publication and dollar money-market day-count convention; accessed October 9, 2026.
5. Federal Reserve Bank of New York, “Primary Dealer Statistics” — Thursday release schedule and the definition and aggregation of settlement fails.
6. TreasuryDirect, “Treasury Reopenings” — issuance of additional amounts of an existing security.
7. Ajay Suresh, “Federal Reserve Bank of New York Building — Entrance,” June 22, 2019. Wikimedia Commons — photograph, CC BY 2.0; reduced-resolution copy used.