Government debt: the price of a dealer’s balance sheet

Four pages into the charter governing France’s primary dealers, the commitment becomes tangible. The banks supporting French government borrowing must sustain a market for its debt, quoting prices at which they will buy and sell and helping to finance the securities. Their job continues long after an auction closes. A government borrowing for decades relies on intermediaries willing to serve investors who may need their money back tomorrow. ^[02]

The first part of this series followed a bond from auction to settlement and separated the initial subscriber from a subsequent holder. One question remained: who carries the security between buyers? The gap may last seconds or stretch into days. It requires funding, creates exposure to price movements and uses part of a bank’s balance sheet. Bridging that gap is one of the services an intermediary provides. ^[03][04]

That service is now the subject of a regulatory contest. On October 1, 2026, Federal Reserve Vice Chair for Supervision Michelle Bowman presented early evidence in support of easing leverage constraints on large US banks. The aim is to make dealers more available to the Treasury market. A Fed survey of eight banking groups and a British stress exercise point to another condition worth testing: whether a bank is willing to use the capacity when its clients ask for it. ^[10][11][13]

This second episode follows that availability, from a quote on French debt to a banking group’s funding decisions. The public documents distinguish securities held, turnover and regulatory headroom. Each measures something different. Treating them as interchangeable can make the same banks appear peripheral in one account and overwhelmingly dominant in another.

The leaders change once the bonds start trading

France’s debt management agency, Agence France Trésor, published its 2025 dealer rankings on February 27, 2026. BNP Paribas led the primary-market category, covering auctions and buybacks. J.P. Morgan came first in the secondary market, where existing securities change hands. Morgan Stanley moved from fourth in primary-market activity to second in secondary-market activity, while BNP Paribas ranked fifth in the latter. The same five institutions occupied the leading places in both lists, in a different order. ^[01]

The changes reflect two sets of functions. Participating in an issue involves preparing bids and reaching investors. Keeping the securities trading involves matching a continuing flow of different needs: an insurer buying, a fund selling, an investor moving between maturities. AFT’s overall score allocates 45 points to primary-market activity, 30 to secondary-market activity and 25 to qualitative criteria. ^[01][04]

The published ranks reveal neither the size of each bank’s bond portfolio nor the nationality of the savers behind its clients. They describe intermediation. A security can pass through several firms before reaching a longer-term holder, accumulating turnover without generating an equivalent amount of new government debt. The illustration therefore shows ranks, with no invented market shares. ^[01][03]

The French charter also requires a presence in repo, the market through which securities can be exchanged temporarily for funding. It explicitly joins the circulation of bonds to the funding that supports them. US primary-dealer eligibility likewise requires market-making in both Treasury cash and repo markets. This comparison concerns functions; the two countries’ dealer arrangements and legal obligations remain distinct. ^[02][03]

Buying time between two investors

Consider a wholly illustrative transaction. A bank buys government bonds at a price of 100 per 100 of face value, requiring a €100 million payment before accrued interest. Clients immediately take €60 million. The bank must hold the remaining €40 million for two days before finding another buyer. We assume the first €60 million of purchases and sales settle together and set intraday liquidity requirements aside. None of these prices, amounts or holding periods describes an observed trade.

The market-maker has agreed to buy before finding every subsequent buyer. Its balance sheet covers the interval. This temporary inventory is often described as warehousing risk. Another trading arrangement might bring two investors together directly, leaving no inventory with the intermediary. Warehousing becomes valuable when the seller wants to transact immediately and the eventual buyer is not yet ready. ^[04][07]

The bank can fund its €40 million inventory through a repurchase agreement, or repo. It transfers securities in exchange for cash and agrees to repurchase them on specified terms. The economics resemble a collateralised loan; the legal structure transfers title to the securities. The bank retains the economic exposure associated with changes in the value of the bonds it is financing. ^[05]

Suppose the cash lender advances 98% of the collateral’s market value. The difference, known as a haircut, provides a buffer against a fall in that value. The bank borrows €39.2 million against its €40 million inventory and finds the other €800,000 from its remaining treasury resources. At an assumed annual rate of 4%, over two days using a 360-day convention, the repo interest amounts to €8,711. ^[05][06] (illustrative calculation)

The €800,000 is the portion of the purchase price this lender leaves unfunded. It does not measure the regulatory capital required for the transaction. Capital has a separate purpose: absorbing losses and supporting the institution’s commitments. The repo interest rate also differs from the bond’s coupon. One pays for very short-term funding; the other belongs to the government’s contract, which may run for decades. ^[05][07]

A dealer earns revenue in part through the difference between its buying and selling prices, known as the bid-ask spread. Buying the entire illustrative lot at 100 and selling it at 100.05 would produce a gross price difference of €50,000 on €100 million of face value. Funding interest, hedging and execution costs, overheads and any adverse price movements still need to be considered. The difference alone cannot reconstruct a bank’s profit. ^[04] (illustrative calculation)

A delay can absorb much of that amount. A 0.1% decline in the value of the remaining €40 million inventory would create a €40,000 gross loss before hedging. Prices can move favourably too. Either way, the dealer has accepted an exposure over time, beyond processing an instruction. A futures hedge can reduce interest-rate risk, while introducing its own cash-flow requirements and exposure to differences between the instruments’ behaviour. ^[07][08] (illustrative calculation)^[06]

This explains how a bond considered safe to hold to maturity can be expensive to carry through a few turbulent days. The government’s creditworthiness determines neither tomorrow’s resale price nor the terms on which the intermediary can obtain funding. ^[06][08]

The bond leaves; the financing remains

Selling the security ends the dealer’s inventory position. It may also begin a different relationship. An investor buys €100 million of bonds, contributes €2 million and borrows the remaining €98 million from a dealer against the securities. The dealer can seek the €98 million from another cash lender, such as a money-market fund, passing on the collateral it receives where its agreements permit reuse. The amounts again are hypothetical. ^[05][06]

Two repo contracts now form a financing chain. The outside lender advances cash to the dealer, which finances the investor. Securities move in the opposite direction as collateral. With identical cash amounts and aligned maturities, assuming the opening settlements also coincide, the dealer’s initial cash flows offset. Its exposure to the client, obligations to its own lender and operational responsibilities remain. It must recover the money and return the securities on the agreed terms. ^[05][06]

The investor retains the economic exposure to the purchased bonds. The title transfers involved in the repo agreements create no additional government bonds or coupons. They do create a set of private financial contracts around the asset. Analysis therefore needs to distinguish government securities outstanding from the financing commitments supported by those securities. ^[05][06]

A bank can reduce the bonds in its inventory while remaining essential to their holders’ financing. Looking only at its securities portfolio misses that role. An October 2024 Fed research note explicitly separates dealer positions from the financing dealers provide to clients. Its historical data describe two channels with different magnitudes and patterns. ^[07]

The risk distinction matters as well. Borrowing and lending the same amount can limit the dealer’s net funding need without eliminating client default risk, collateral liquidation risk or settlement mismatches. If its borrowing expires before its loan, the dealer needs replacement funding for the gap. The amount, maturity and counterparty determine how much of a balance-sheet commitment the business creates. ^[06]

The name appearing in a holdings statistic answers a question about ownership within that statistic’s accounting scope. Assessing market resilience also requires asking who advanced the money, how long it is available and what must happen for it to be renewed. Part three follows investors who borrow to buy.

Several limits, several decision-makers

“Balance-sheet capacity” can sound like a single pool of money available for government bonds. Within a financial group, usable capacity is the product of several decisions. The group has capital; a market-facing legal entity receives resources; a trading desk operates within risk limits; and treasury arranges funding. Room at the consolidated group level can coexist with a desk that is already close to its own ceiling. ^[07]

Prudential rules help shape those decisions. Risk-based capital requirements attempt to align loss-absorbing resources with the nature of exposures. The leverage ratio adds a broader backstop: it compares Tier 1 capital with an exposure measure that includes certain off-balance-sheet commitments, without weighting those exposures by credit risk. That requirement can matter for a low-margin, high-volume business even when its securities carry relatively little credit risk. ^[07][09]

Capital, supplied in part by shareholders and retained earnings, absorbs losses. It is not a separate heap of idle cash. An institution can have adequate capital while lacking cash in the right place at the right moment. It can also possess cash and decline an additional exposure. This distinction is particularly important when a reform is said to release trillions of dollars of capacity. ^[07][13]

Internal limits react to market conditions as well. If prices become more volatile, the same unchanged position can generate a larger modelled potential loss at a trading desk. The desk approaches its risk ceiling without having bought another bond. Reducing its position becomes one possible response. The Fed’s 2024 analysis identifies this channel alongside regulatory constraints, with particular relevance during market stress. ^[07]

Research by Darrell Duffie and co-authors at the New York Fed, published in 2023, adds another piece of evidence. When dealer balance-sheet utilisation is especially high, observed deterioration in liquidity can exceed what volatility alone would explain. March 2020 provides a prominent example. The paper supports the importance of intermediation capacity without establishing one regulatory ratio as the sufficient cause of every disruption. ^[08]

Clients encounter these decisions through execution terms. To take more securities, a bank may require greater compensation, quote for smaller sizes or prioritise particular relationships. A price may remain on a screen even as a large sale becomes harder to execute on those terms. Market availability depends on executable quantity and the price impact of an order as well as on the presence of a quote. ^[04][08]

Washington creates regulatory room

On November 25, 2025, US banking agencies adopted changes to the enhanced supplementary leverage ratio, or eSLR, for systemically important banking groups and certain subsidiaries. The rule took effect on April 1, 2026, with optional early adoption from January. It recalibrated the backstop to make it less constraining for some low-risk activities, including Treasury intermediation. The leverage requirement remained in place. ^[09]

The official record preserves an important disagreement. The agencies describe the aggregate reduction in Tier 1 capital requirements at the affected holding companies as limited, at less than 2%. Governor Michael Barr, who dissented, emphasises an estimated $219 billion reduction at bank subsidiaries. These figures describe different organisational levels. Barr’s number is an estimated reduction in requirements, rather than a sum already distributed to shareholders. ^[09][12]

The argument for the reform is straightforward. At a given level of capital, a lower leverage constraint permits more exposure. A group may consequently find holding Treasuries or extending low-margin client financing more attractive. The counterargument concerns how it uses that latitude and how resilient it remains: the room can support other activities, while less capital at a bank subsidiary changes the protection available if difficulties arise. Barr challenges the presumed progression from regulatory relief to market resilience. ^[09][12]

Bowman’s October 1 speech offers early observations. She reports almost $5 trillion of additional eSLR headroom in the first quarter of 2026 at the parent holding companies of six dealers. She also reports Treasury positions in supervisory data rising from roughly $600 billion to more than $700 billion, between the beginning of the modification period and the end of April. The increase was concentrated among firms that had previously been more constrained. ^[10]

The distinction between these measures is crucial. The $5 trillion is capacity calculated under a regulatory standard at holding-company level. It is neither cash received nor an amount of Treasury purchases. The roughly $600 billion and more than $700 billion refer to reported positions in a separate aggregate. The October 1 speech refers to a research note that was then forthcoming. The figures here are attributed to that speech; this investigation does not reconstruct the underlying institution-level supervisory data. ^[10]

Eight banks, two positive responses

A separate piece of evidence is already public. In a survey conducted from March 20 to March 30, 2026, with results released on May 14, the Fed specifically asked the eight US global systemically important banks about the reform’s effect on their own or their affiliated dealer’s repo activity. Two attributed an actual or expected increase in repo activity to the reform; six reported no effect. The question encompassed both realised and anticipated effects during the year. ^[11]

Each institution counts once in this result. It does not mean 75% of repo volume was unchanged. Large expansions at two firms could dominate an aggregate even if the other six reported no effect. The observations in the October speech also extend to late April, beyond the March survey, and include securities positions. The two documents illuminate uneven effects rather than providing a contradictory causal test. ^[10][11]

Their juxtaposition supports a more useful question than declaring the reform an outright success or failure. Which firms converted regulatory room into additional services, for which clients and which transactions? Relief may improve intermediation at some institutions while barely changing the decision at others. Evaluating the collective benefit also requires evidence about execution prices, market depth and financing terms during difficult episodes.

The public documents show differing responses and developments, with Michelle Bowman making the favourable case in remarks she identifies as her own views. Marginal availability during a crisis remains an open question. Capacity deployed in ordinary conditions is already occupied when fresh demands arrive. The starting position matters alongside the regulatory ceiling. ^[10][12]

The client expecting another loan

The Bank of England examined availability using a different method. Its system-wide exploratory exercise, published on November 29, 2024, compared the reactions of banks, funds, insurers and other participants to a hypothetical market shock over ten business days. The exercise used positions as of October 31, 2023. It models interacting behaviour; it describes neither an actual November 2024 crisis nor participants’ current balance sheets. ^[13]

One finding is especially relevant. In the scenario, banks generally had the funding and balance-sheet resources but little appetite to extend additional repo. Counterparty risk was the main restraint. They were more willing to renew existing trades than to expand financing lines. Meanwhile, some asset managers overestimated the extra funding they could obtain. ^[13]

The mechanism is intelligible at the level of an individual credit decision. A lender may accept the quality of a government bond and still doubt that its client will survive the week. It must assess how quickly it could recover and sell the collateral, what the securities might be worth during that interval, and its other exposures to the same client. The credit relationship extends beyond the security delivered as collateral. ^[06][13]

Such caution may protect the bank individually. Across the market, it requires the investor to seek an alternative: use its cash, find another lender or reduce a position. A sale becomes necessary when alternatives are inadequate or judged too expensive. The British exercise also found that many participants still had fallback resources. Equating every declined loan with a forced liquidation would misrepresent the findings. ^[13]

The collective problem emerges when multiple firms rely on the same availability and it falls short. An investor hoping to buy newly cheap bonds may also need financing. If lenders’ caution prevents that demand from materialising, prices may have to fall further to attract a buyer able to pay without the loan. This provides a plausible transmission channel from a bank’s decision to government-bond market stress. Its strength depends on the shock and on alternative sources of funds.

More circulation from the same balance sheets

Dealers need not warehouse every transaction. Investors can trade directly with one another where market arrangements make the connection possible. A New York Fed study published in February 2025 examines all-to-all trading: expanding the set of accessible counterparties so that certain trades no longer routinely depend on a dealer’s inventory. Access, credit relationships and settlement remain among the challenges. ^[04]

The benefit depends on compatible orders existing. A willing insurer can meet a fund that wants to sell, with the platform facilitating their trade. When sellers arrive before buyers, matching technology alone cannot absorb the time between them. Someone must take the exposure, or the price must move enough to attract a buyer who can transact.

Another route reduces obligations that offset. A central counterparty interposes itself between participants, becomes their counterparty and calculates net obligations under its rules. This can economise on securities and cash transfers and, where the accounting and prudential conditions allow, reduce balance-sheet usage. In exchange, it requires collateral and robust management of default risk. ^[06][14][15]

Consider three simultaneous trades in the same security, at the same price and for the same settlement date. A buys 100 from B; B buys 100 from C; C buys 80 from A. Gross payments total 280 units. With complete multilateral netting, A must pay 20 and receive the corresponding securities, C must receive 20 and deliver them, and B has a zero net obligation. The CCP receives 20 from A and pays 20 to C; both legs settle the same net amount. Original trading volume remains 280, and the government’s outstanding debt is unchanged. ^[14][15] (illustrative example)

The example demonstrates a settlement saving. The precise balance-sheet benefit depends on additional conditions, including the legal enforceability of netting and the accounting treatment of the exposures. Margin posted to the clearing house is itself a liquidity requirement. Clearing also concentrates a critical function within an infrastructure that must withstand a member’s default. ^[06][14]

In the United States, the compliance dates confirmed on the SEC’s implementation portal, updated on September 23, 2026, and repeated in industry documentation are December 31, 2026 for eligible cash transactions and June 30, 2027 for eligible repos. Both remained in the future at this investigation’s cut-off. Scope depends on the categories and counterparties covered by the rules; the same obligation does not apply indiscriminately to every Treasury transaction. ^[15][16]

This route changes how trading is organised. In a July 17, 2026 article based on a speech delivered in May, the Bank of England’s Sarah Breeden highlights the potential of clearing and standardisation in UK repo. She also notes that gilt markets had recently weathered volatility without major dysfunction. Their ability to adapt belongs in an investigation of vulnerabilities as much as their possible weaknesses. ^[14]

The cost reaches the state through the price

After issuance has settled, a resale between investors brings no additional proceeds to the government. Yet the terms of that resale matter for subsequent issues. An investor anticipating a difficult exit may demand extra compensation at the outset. An intermediary expecting to fund inventory for longer may adjust its bids too. Secondary-market quality therefore helps determine the price at which the government’s promise finds a buyer. ^[04][17]

A larger debt stock does not mechanically imply deteriorating conditions. Research published by the New York Fed in March 2026 and revised in July examines price movements around US Treasury auctions over 1991–2024. It finds no recent increase in those pressures despite debt growth and highlights a greater role for nondealer investors in absorbing issuance. The sample ends before the 2026 US regulatory change and cannot measure its effects. ^[17]

Taken together, the documents support a bounded conclusion. Government-bond markets rely on intermediaries able to warehouse securities, finance purchasers and organise settlement. Regulatory requirements can constrain those functions. So can internal risk limits, available funding and confidence in a client. Improving any single component does not measure the availability of the entire chain. ^{[06][07][08][13]}

Public sources identify approved counterparties, rankings, some aggregate positions and responses to supervisory surveys. They do not provide an exhaustive chain linking every bond to its ultimate buyer, lender and financing terms. This investigation therefore attributes neither the inventory nor the profit on a specific bond lot to a bank merely because of its place in a ranking. ^{[01][03][10][11]}

The next episode moves to the investor borrowing to buy. On October 6, 2026, the IMF again highlighted hedge funds’ dual role: supporting price formation while also being capable of amplifying financial shocks. Understanding which role dominates requires following their funding, collateral and maturities. Having examined the balance sheet that helps sell the debt, the series turns to the one that makes the purchase possible. ^[18]

Method and scope

Documentary investigation with a research cut-off of October 11, 2026. Main sources are AFT, the Federal Reserve, the New York Fed, the Bank of England, the FSB, the SEC and the IMF. ICMA and SIFMA are used for professional-practice documentation and identified as market organisations in the references. Illustrations with assumed figures are reproducible teaching examples, separated from observed data. No interviews, private portfolios or bank-specific profit calculations are represented as obtained. The references below record publication dates, observation periods and collection limitations.