The Treasury Market’s Hidden Plumbing
This chapter follows a Treasury repo before a central counterparty changes its risk connections. The specific focus is how a cash borrower, a cash lender, a dealer, a custodian or triparty agent, and an interdealer broker fit together—and where counterparty, rollover, collateral, and settlement exposures remain. That operating model lets us judge central clearing by the problems it actually solves, rather than by the reassuring sound of the word “central.”
One Repo, Three Roles, Several Risks
A repo is easy to mistake for an ordinary bond trade because its legal form is a sale followed by an agreement to repurchase. Economically, however, it is a short-term loan secured by a Treasury security. The party selling the security and promising to buy it back is raising cash. The party buying it and agreeing to return it later is investing cash against collateral. The difference between the collateral’s market value and the cash advanced is the haircut, a buffer intended to protect the cash lender if the borrower defaults while the security’s value falls.
The basic transaction has three roles. A cash borrower may be a hedge fund financing a leveraged position, a dealer funding inventory, or another institution that wants temporary liquidity. A cash lender may be a money-market fund or other institutional investor seeking a short-term return with high-quality collateral. The dealer connects them. The dealer may receive cash from the lender and pass it to the borrower, while moving Treasury securities in the opposite direction. The ultimate borrower and lender do not need to negotiate every operational detail with each other, and they may not even have the same timing, size, or maturity preference.
This intermediation is useful, but it does not make the system risk-free. A dealer’s aim is often to maintain a matched book: repo borrowing from one client is offset by reverse repo lending to another, or the reverse. If the positions match closely, the dealer is not primarily taking a directional view on Treasury prices. Yet the two sides do not disappear. They remain separate legal relationships with separate settlement instructions, collateral terms, credit limits, and failure possibilities. A borrower can fail to return cash, a lender can fail to deliver securities, a settlement agent can encounter an operational problem, or a dealer can be left with one side of a supposedly matched position when the other side breaks.
The timing of the repo matters as much as its collateral. An overnight repo matures on the next business day. That short maturity limits the period of market exposure, but it also turns refinancing into a daily requirement. A borrower must find funding again tomorrow; a lender must decide again whether to reinvest. In normal conditions, that flexibility is valuable. In stress, the same flexibility can become a point of withdrawal. If lenders become cautious or dealers reduce the balance sheet they are willing to use, a borrower may have to sell securities or shrink a position quickly rather than roll it.
A term repo lasts longer—perhaps several days, weeks, or months—while an open repo has no fixed maturity and can be ended under agreed notice terms. Term funding reduces the need to refinance every morning, which can reduce rollover risk and the chance of a forced sale caused by a one-day funding gap. It creates a different exposure in return: the parties are committed for longer while interest rates, collateral values, and the credit condition of the counterparty can change. A longer maturity is therefore not simply a safer overnight repo. It trades daily refinancing risk for longer-lived market and counterparty exposure and may require more collateral.
The haircut is another trade-off rather than a complete shield. A positive haircut means the cash lender advances less than the value of the securities received. That protects the lender against a limited decline in collateral value and some liquidation costs. It does not guarantee that the securities can be sold immediately at the quoted price, that the borrower will perform every settlement obligation, or that the lender can avoid losses in a fast-moving market. Nor does it necessarily protect the repo seller equally. In a bilateral arrangement, the repo buyer may be able to sell or reuse the security, leaving the seller exposed to the buyer’s ability to return it at the end of the trade.
The SEC’s economic analysis cites an Office of Financial Research pilot in which 74% of non-centrally cleared bilateral repo volume was transacted at zero haircut. The report associated much of that result with naturally offsetting repo and reverse-repo packages, often involving the same counterparty and tenor. That observation does not mean every zero-haircut trade is imprudent, or that the pilot describes all repo activity. It does show why private incentives can produce a network with less visible protection than a central risk manager might prefer. A dealer and a sophisticated client may understand their offsetting exposures, while other participants and the wider clearing system cannot see the same net position.
The diagram’s alternatives matter because “repo” does not identify one uniform risk system. Bilateral, triparty, and centrally cleared transactions can all move cash against Treasury collateral while allocating responsibility differently. The same economic purpose—temporary financing—can therefore create different exposures for the borrower, lender, dealer, and infrastructure provider.
Where the Network Becomes Fragile
In bilateral repo, the counterparties agree the terms and settle the trade between themselves, often with custodians or clearing banks supporting the movement of cash and securities. Each relationship can use its own credit limits, haircut, collateral schedule, margin practice, and close-out process. This flexibility serves borrowers and lenders whose needs are not standardized. It also makes risk harder to compare across the market. A haircut negotiated down to zero, a maturity chosen to fit one portfolio, and a settlement procedure accepted by one dealer may not be visible to another dealer or to a central risk manager.
Triparty repo adds a third-party operating platform, but not necessarily a third-party guarantor. The triparty agent can hold the securities, value and allocate collateral, confirm that agreed conditions are met, and move cash and securities among accounts. Keeping collateral within the platform can reduce some settlement failures and limit how the buyer reuses the asset. These are meaningful operational protections. They do not make the agent the buyer to every seller and seller to every buyer. If the cash borrower cannot repay, the lender still faces the borrower’s performance and the consequences of liquidating or managing the collateral. The platform administers the trade; it does not absorb the counterparty risk through novation.
Central clearing changes that relationship. Once a qualifying CCP receives and matches the trade, it becomes the counterparty to both sides, collects margin under its risk methodology, and can manage positions and defaults across its members. That is a different allocation of responsibility from triparty administration. But it does not follow that a centrally cleared trade becomes independent of every other institution. A client may still depend on a sponsoring member to access the CCP, and a settlement may still depend on banks or custodians to deliver cash and securities. Central clearing changes the risk-management connection; it does not erase the execution and settlement connections around it.
The cash Treasury market illustrates a related problem through interdealer brokers, or IDBs. An IDB can provide an electronic facility where buyers and sellers submit orders anonymously. The IDB then stands as principal to each side, booking two transactions rather than exposing the identity of the ultimate participants to one another. This arrangement is valuable to dealers and principal trading firms, or PTFs, that want access to liquidity without revealing their positions or strategies. It also means that the IDB itself sits at a central point in the transaction network.
Suppose a dealer sells a Treasury security to a PTF through an IDB. From an economic perspective, the IDB may be close to flat: it buys from one side and sells to the other. But its two legal transactions can be treated differently for clearing. The IDB’s trade with the dealer may be submitted to the CCP because the dealer is a direct clearing participant. The IDB’s trade with the PTF may remain bilateral because the PTF is not a member and has not entered a client-clearing arrangement. The CCP sees the first leg but not the second. To the CCP, the IDB can appear to have a directional position, even though the IDB believes it has an offsetting trade elsewhere.
That is hybrid clearing. Its danger is not simply that one half of a transaction is outside the central system. The danger is that the two halves are connected economically but governed by different risk controls. If the PTF fails to deliver securities or cash, the IDB may be unable to perform on the dealer-facing leg that has already been cleared. The IDB’s own resources may then be insufficient, and a problem that began with a non-member counterparty can reach the CCP through the IDB. The CCP may have margin and default procedures for the visible leg, but it lacks the information needed to manage the unseen exposure in the same way.
This pattern also explains why the market’s electronic appearance can be misleading. An anonymous order book may broaden the people who can submit interest to a platform, but the resulting trade still has to be cleared, margined, and settled through particular relationships. A PTF may see prices and transact electronically while remaining outside the central clearing system. A customer may receive a quote from several dealers while still depending on one dealer to clear and settle the trade. A triparty platform may process collateral efficiently while leaving the underlying counterparties exposed to one another. Execution access, operational access, and risk access are not the same thing.
Nor are bilateral and triparty arrangements uniformly defective. Bilateral trading can accommodate bespoke sizes, maturities, packages, and collateral that a standardized market cannot easily absorb. A triparty agent can reduce operational friction and settlement risk even without becoming a CCP. A dealer can add value by using relationships and balance-sheet capacity to match a borrower and lender whose needs do not line up. The relevant comparison is not between a perfect central system and an irrational bilateral one. It is between different ways of allocating flexibility, information, collateral, and failure responsibility.
The structural weakness appears when the market depends on many such relationships at once. Overnight funding creates repeated rollover points. Low haircuts leave less buffer against a rapid price move. A dealer that is matched in aggregate may still face a settlement failure on one leg. An IDB that is flat in economic terms may look directional to the CCP because only one leg is visible. A triparty agent may keep collateral organized without guaranteeing repayment. These are separate mechanisms, but they can reinforce one another when liquidity disappears.
The case for central clearing begins with this specific problem: not that dealers are unnecessary, and not that every bilateral trade is unsafe, but that the risk system should be able to see and manage linked obligations more completely. Novation, multilateral netting, uniform margin, and centralized default management can address parts of that problem. They cannot decide who may trade, which venue shows the best price, or whether a repo contract is standardized enough for a broad market. The next question is therefore not whether the network should have a hub, but what the hub can actually see, net, and guarantee once more of the market is connected to it.