Quick answer: a repurchase agreement (repo) is economically a secured loan: one party receives cash and provides securities, agreeing to reverse the transaction later. A repo-pricing algorithm has to determine the financing rate, haircut or margin, term and collateral treatment while distinguishing two different motives. In general collateral (GC) repo, the borrower mainly wants cash funding and can deliver any eligible collateral in a class. In a special repo, the cash lender specifically wants a scarce security, so the repo rate can fall below GC because part of the economics is payment for access to that security. Strong pricing models therefore separate funding value from collateral scarcity, and then layer in haircut, term, counterparty, settlement, balance-sheet and clearing costs.
The repo rate is not always just the price of cash. Sometimes it is also the inverse price of getting hold of a scarce bond.
Page role: this article owns the mathematics of repo pricing. It is distinct from collateral optimisation, which asks which asset should be pledged to which obligation, and from wholesale-funding rollover risk, which asks whether funding can be renewed through time.
1. Repo is a forward-linked cash transaction
Suppose cash principal P is exchanged today and repaid after d days at annualised repo rate r using day-count basis B. A simplified repurchase amount is:
Repurchase amount = P × (1 + r × d/B).
For example, if P = S$100 million, r = 4.00%, d = 7 days and B = 360:
Interest ≈ 100,000,000 × 0.04 × 7/360 = S$77,777.78.
The economic return therefore comes from the difference between the initial cash exchange and the later repurchase amount, while securities collateral secures the obligation.
2. General collateral: pricing the cash side
In a GC repo, the cash provider is broadly indifferent among a set of eligible securities. The collateral mainly protects the loan; the economic question is the secured funding rate.
Useful inputs include:
- overnight secured funding conditions;
- term and expected rate path;
- counterparty and clearing route;
- collateral class;
- haircut/margin;
- balance-sheet and capital cost;
- settlement and operational cost.
DTCC’s GCF Repo service describes GC trading in exactly this generic-collateral way: dealers trade by rate, term and underlying product, with specific securities allocated later after netting. See DTCC GCF Repo.
3. Specialness: when the collateral side dominates
A repo becomes “special” when a particular security is especially valuable to obtain. The cash lender may accept a lower interest rate because access to the collateral is itself valuable.
A simple specialness measure is:
Specialness spread = GC rate − special repo rate.
If GC is 4.20% and a specific Treasury trades at 3.10%, its specialness spread is 1.10 percentage points. The lower repo rate does not mean the cash lender made a pricing mistake. It means part of the lender’s return is the ability to borrow that security.
The New York Fed’s SOFR methodology explicitly mitigates the influence of specials in the centrally cleared DVP segment because SOFR is intended to measure the broad cost of overnight Treasury financing rather than the scarcity value of particular securities. See the November 2024 SOFR methodology modification.
4. SOFR is a funding benchmark, not the repo price of every bond
The New York Fed describes SOFR as a broad measure of the general cost of financing Treasury securities overnight. It incorporates tri-party repo, GCF Repo and qualifying FICC DVP activity under its methodology. See New York Fed Reference Rates.
A bank can use SOFR as one benchmark input, but a transaction-specific repo price can differ because of:
- collateral specialness;
- term;
- haircut;
- counterparty;
- clearing route;
- settlement timing;
- balance-sheet capacity;
- specific collateral optionality.
“SOFR + spread” can therefore be a useful starting representation, not a universal pricing law.
5. Haircuts change the amount of cash raised
If collateral market value is V and haircut h, a simplified cash amount is:
P = V × (1 − h).
If a bond is worth S$105 million and the haircut is 5%, cash advanced is approximately S$99.75 million.
The haircut protects the cash lender against adverse price moves and liquidation uncertainty if the borrower defaults. But it also reduces the borrower’s funding efficiency. The FSB’s haircut framework exists partly because excessively low haircuts can support too much leverage, while abrupt increases in haircuts can amplify stress. See FSB haircut standards.
6. Rate and haircut are different price dimensions
Two repos can have the same interest rate but different economics if one advances 98 cents of cash per dollar of collateral and the other advances 90 cents.
A full economic comparison should consider:
effective funding cost ≈ repo interest cost + cost of overcollateralisation + capital/liquidity/operational costs.
This matters especially when collateral is scarce or expensive to finance elsewhere. A low repo rate with a severe haircut may not be the cheapest all-in funding route.
7. Term repo prices the expected path of secured funding
An overnight repo exposes the borrower to tomorrow’s refinancing rate. A one-month repo locks the funding rate for longer. The term rate therefore reflects expected future overnight rates plus term, balance-sheet, liquidity and collateral effects.
A teaching decomposition is:
Term repo rate ≈ expected compounded overnight secured rates + term premium + balance-sheet/liquidity adjustments ± collateral specialness.
The sign on collateral specialness can be negative because a highly desired security can pull the repo rate below ordinary funding levels.
8. Open repo and evergreen structures add optionality
Not every repo has a fixed maturity. Open or evergreen structures can continue until terminated under agreed notice. That introduces behavioural optionality: one party may choose to end funding when market conditions become less favourable.
A pricing engine should therefore distinguish a fixed term from a funding relationship that can disappear with notice. The contractual notice period and actual behaviour affect liquidity value.
9. Margin maintenance creates a dynamic collateral equation
Collateral prices move after the repo starts. If collateral value falls relative to cash exposure, the lender can require additional collateral or other adjustment under the agreement.
A simplified margin deficit is:
Deficit = required collateral value − current collateral value.
This turns repo into a recurring state-update problem rather than a one-time trade. The same security that looked sufficient on Monday can require additional collateral on Tuesday after a price shock.
10. Fails can distort specialness
When a security is extremely scarce, settlement fails can rise because firms cannot obtain the bond they need to deliver. That creates a feedback loop:
scarcity → lower special repo rate → stronger demand to borrow the security → potential settlement pressure.
The Treasury Market Practices Group uses fails-charge conventions to reduce incentives for persistent fails. The New York Fed’s securities-lending programme also applies penalties when borrowed securities are not returned on time. See the New York Fed securities-lending FAQ.
A repo-pricing model that ignores expected fail cost can misread an extremely low special rate as pure profit opportunity.
11. Collateral substitution changes the option value
Some repo agreements allow collateral substitution subject to eligibility rules. That option is valuable to the collateral provider because it can replace one asset with another if it needs the original security elsewhere.
The cash lender should therefore price not only today’s collateral but also the range of collateral that may legally replace it. A broad substitution right can make a transaction behave more like GC; a specific-security no-substitution trade behaves more like a collateral acquisition.
12. Balance-sheet cost can push repo away from pure market rates
Repo can consume dealer balance sheet even when collateral is high quality. Capital and leverage constraints can therefore affect spreads and capacity, especially near reporting dates or during periods of heavy Treasury intermediation.
This creates a link to balance-sheet optimisation: the repo desk does not price inside an infinite balance sheet. It competes for leverage exposure, liquidity, capital and collateral capacity with other businesses.
13. Central clearing changes netting, margin and counterparty structure
Central clearing can reduce bilateral counterparty exposures through multilateral netting but introduces CCP margin and default-management requirements. That can change the all-in economics of repo even if the headline repo rate is unchanged.
As of August 2026, the SEC’s current Treasury-clearing implementation page states that the compliance date for eligible US Treasury repo transactions is 30 June 2027, following a one-year extension. See SEC Treasury Clearing Implementation. That future date matters for model governance because pricing engines should not assume the post-mandate clearing structure is already universal.
DTCC has also introduced new cleared-repo access and collateral models in preparation for wider Treasury clearing. Market-structure change is therefore an update trigger for pricing, netting and margin assumptions.
14. A compact repo-pricing objective
For transaction j, a simplified bank-side objective can be written:
Net valuej = repo spread income − balance-sheet cost − liquidity cost − expected counterparty loss − margin/collateral cost − settlement/operational cost + collateral option value.
For a special repo, collateral option value can be substantial. For a GC repo, it may be close to zero because the cash lender is indifferent among eligible securities.
15. Inputs and outputs
| Inputs | Outputs |
| SOFR/GC benchmarks, term curve | Base funding rate |
| Security availability and specialness signals | Collateral scarcity adjustment |
| Collateral price, volatility and liquidity | Haircut/margin |
| Counterparty and clearing route | Credit/capital adjustment |
| Term, substitution, call/termination features | Optionality/term adjustment |
| Settlement and fail history | Operational penalty |
16. Failure modes
- GC-everywhere assumption. A scarce security is priced as if collateral were interchangeable.
- SOFR=trade-rate assumption. A broad benchmark substitutes for collateral- and counterparty-specific pricing.
- Rate-only comparison. Haircut and overcollateralisation costs are ignored.
- Static collateral. Margin needs are not updated after price moves.
- Specialness extrapolation. Yesterday’s scarcity premium is projected after supply returns.
- Fail-cost omission. Very special collateral looks profitable despite repeated settlement failures.
- Balance-sheet blindness. Desk pricing ignores leverage and capital consumption.
- Clearing-regime drift. Models continue using bilateral assumptions after central clearing changes netting and margin.
17. Diagnostics, alternatives and falsifiers
- How far is this repo rate from GC and why?
- Does the specialness spread move with observable collateral scarcity?
- What is the all-in funding cost after haircut and balance-sheet charges?
- How much does term pricing change if the expected overnight-rate path shifts?
- What happens if collateral value falls 5% before margin is exchanged?
- Would a simpler SOFR-plus-spread model explain the trade as well as the complex model?
- How much netting benefit arises from the clearing route?
- Do fails rise when specialness becomes extreme?
Falsifier: suppose someone claims, “A lower repo rate always means cheaper funding.” A counterexample is a special repo where the cash borrower accepts a low rate because it is lending a scarce bond that the cash lender values highly. The cash rate fell, but the collateral provider gave up scarcity value.
18. Verification and update triggers
- Reconcile contractual rate, principal, day count and repurchase amount.
- Compare transaction rates with current GC and SOFR references.
- Backtest specialness against lendable supply, fails and settlement demand.
- Revalue collateral and haircut sufficiency under stress.
- Measure actual all-in cost after margin and balance-sheet charges.
- Update after major central-clearing rule changes.
- Recalibrate after persistent shifts in repo market structure or collateral supply.
- Keep a simple benchmark model as an independent challenger.
Research anchors
- New York Fed — Treasury repo reference rates and SOFR.
- New York Fed — SOFR methodology and treatment of specials.
- DTCC — GCF Repo.
- FSB — Securities-financing haircut framework.
- SEC — Treasury clearing implementation and current dates.
The deeper lesson
Repo pricing is a two-sided market in cash and collateral. GC repo mostly prices secured funding. Special repo can price access to a specific scarce security. Haircuts change how much cash is actually raised. Term changes refinancing certainty. Margin makes the collateral state dynamic. Fails reveal when scarcity becomes operationally damaging. Clearing changes netting and capital economics. A strong algorithm therefore asks not only, “What is the repo rate?” but also, “Which side of this transaction is scarce, and which constraint is really determining the price?”
Educational note: This article explains public repo-market mathematics. It is not trading advice, funding advice, a repo quote or an institution-specific collateral strategy.