The Market Is a Web, Not a Ledger
The sovereign-bond market is not held or operated by one kind of investor. This chapter focuses on the relationships among non-bank financial institutions, banks and dealers that determine whether government debt absorbs a shock or passes it through the financial system. The useful judgment is not “Are non-banks risky?” but “Which institution is carrying which obligation, through which interface, and who must raise cash when the bond market moves?”
That question matters because the market’s ownership structure has changed. The Bank for International Settlements reports that non-bank financial institutions, or NBFIs, became the largest holders of sovereign debt in advanced economies: their share rose from 44 percent in 2021 to 53 percent in 2025. This shift reflects both more government debt to absorb and changes in bank balance sheets after the global financial crisis. But the headline share hides different kinds of behavior. Some non-banks provide patient demand. Others depend on leverage, short-term funding or the ability to meet redemptions without selling into a thin market.
A market-wide risk assessment therefore needs a map of functions and interfaces, not a ledger showing who owns the most bonds.
One market, different reasons to hold
An insurer or pension fund may hold government bonds because the assets help match long-term liabilities or manage interest-rate duration. A money-market fund may hold government securities or repos because investors expect access to cash. A hedge fund may hold a large gross position in bonds, futures and swaps because a relative-value trade appears to offer a small spread. An open-ended bond fund may hold government bonds as part of a portfolio while promising investors frequent redemption. These positions can involve the same sovereign security but create very different demands on liquidity.
This distinction is important when interpreting the rise of NBFIs. The BIS explicitly separates real-money investors, such as many insurers and pension funds, from leveraged hedge funds and redemption-sensitive funds. Real-money demand can deepen a market and diversify risk, including across borders. It may also provide a source of stability because a long-term investor is not forced to refinance a position every day. That does not make such an investor immune to losses or margin calls, but the liability structure gives it more time to respond.
A hedge fund has a different relationship with the same bond. Its position may be financed through repo and supported by prime-brokerage or derivatives services. The return can depend on the convergence of two prices, while the ability to remain in the trade depends on collateral values and financing terms. A fund with daily or periodic redemption obligations has another constraint: it may need to sell liquid assets to meet withdrawals even if the underlying credit thesis remains intact.
The term NBFI is consequently a category of legal and economic institutions, not a conclusion about risk. Treating every non-bank as a hidden source of instability would miss the stabilizing role of long-term holders. Treating every non-bank position as patient capital would miss the leverage and liquidity mismatch that make a sovereign market more sensitive to stress.
The structure has also been shaped by the supply of government debt and by the changing economics of bank market-making. Large and persistent fiscal deficits have expanded sovereign issuance faster than credit to the private sector in some advanced economies. At the same time, post-crisis regulation and bank business-model changes made balance-sheet-intensive market-making more costly. Banks remain central to intermediation, but their capacity to warehouse growing amounts of sovereign risk has not expanded at the same pace as the debt stock. Deeper derivative and repo markets helped other intermediaries take a larger role.
That is not a simple substitution of banks by non-banks. It is a redistribution of tasks. The financial system still needs dealers to connect buyers and sellers, banks to provide funding and credit lines, and infrastructure to value collateral and collect margin. A bond can move from a bank balance sheet to a hedge-fund strategy without leaving the banking network that finances, hedges or clears the position.
Interfaces decide who absorbs a shock
The most important connections are contractual. A dealer may provide a hedge fund with repo financing, prime brokerage, securities lending or derivatives intermediation. The same dealer may make a market in Treasury securities and serve other investors that use those securities as collateral. A bank may extend credit lines to private-credit vehicles, insurers or asset managers, while money-market funds and other NBFIs provide the bank with deposits, purchases of bank paper or secured funding.
Each interface carries a different obligation. Repo creates a claim that depends on collateral value and a haircut. A derivative creates margin requirements that can change with prices and volatility. A credit line creates a potential draw on the bank precisely when the borrower’s other funding sources are under pressure. A redemption promise gives the investor a claim on cash even if the fund’s assets are harder to sell. These are not interchangeable forms of exposure, and they do not become safe merely because the common collateral is a government bond.
The connections can also run in both directions. Banks and dealers support the market’s liquidity, but their ability to do so depends on capital, risk limits, inventory capacity and access to funding. NBFIs rely on those services, yet they can also supply funding back to banks. When conditions tighten, a bank may reduce financing to a leveraged client, while the client’s selling makes the market less liquid and increases the risk of further calls on the bank. The network’s behavior is therefore different from a one-way chain in which a loss simply moves from one owner to the next.
Each interface also has its own clock. A long-term liability may be measured in years, but a repo can require renewal much sooner and a derivatives position can require cash as soon as its market value changes. A fund can therefore have an asset that is valuable over a long horizon and still face a short-horizon funding problem. The prime broker monitors the client’s positions and sets financing terms; the repo lender decides how much cash to advance against collateral; a dealer’s risk function determines how much inventory the firm can carry; and the fund decides whether to add liquidity or sell. No single participant controls the whole chain.
That division of control explains why ownership data are insufficient. Imagine two investors holding the same quantity of a government bond. One pays with its own capital and can wait through a temporary price decline. The other has borrowed against the bond and must renew that borrowing. The market sees the same asset in both portfolios, but the second investor brings an obligation that can turn a price move into a sale. If both are served by the same dealer, the dealer may face higher inventory and counterparty risk precisely when the market needs more liquidity. The relevant unit of analysis is consequently the asset plus the liability attached to it.
The reverse connection matters as well. A bank may treat a hedge fund as a client whose trades generate fees and market activity, while the fund treats the bank as a source of essential financing. In quiet markets those incentives can reinforce one another: plentiful funding supports trading, and trading supports dealer revenues. In a stressed market, each side may protect its own balance sheet. The dealer tightens terms to limit exposure; the fund sells to preserve cash; the sale worsens prices used by other counterparties. The behavior is individually understandable but collectively procyclical.
The relationships can be represented without assuming that every connection is active at the same intensity:
The purpose of this map is to show shared interfaces, not to claim that the entities have identical balance sheets. A long-term pension fund and a leveraged hedge fund can both be linked to the sovereign-bond market while responding to a yield rise in opposite ways. One may tolerate a valuation loss; the other may face a margin call. A dealer may provide liquidity to both, but its own capacity can be limited by the volatility and inventory created by their combined actions.
This is also why the same sovereign bond can appear in several risk channels at once. It can be an asset held for long-term duration, collateral for secured borrowing, a component of a derivative hedge and a liquid holding used to meet redemptions. The security’s identity has not changed. The obligations attached to it have.
The bank-sovereign link has moved into the network
Banks’ direct holdings still matter. The BIS reports that banks’ share of sovereign-debt holdings has been broadly stable at about 20 percent over the past five years. In 2025, median bank sovereign exposure relative to Tier 1 capital was around 180 percent in emerging-market economies and just under 100 percent in advanced economies. A rise in sovereign yields can reduce the market value of bonds held by banks and weaken their capacity to lend or make markets.
Yet the direct channel appears less dominant in some advanced economies than it was a decade ago. Banks now hold more and better-quality capital, maintain larger liquidity buffers and rely on more stable funding. European evidence summarized by the BIS suggests that bank-sovereign credit-default-swap co-movement has become less dependent on banks’ direct sovereign holdings. These reforms matter: a stronger bank can absorb a valuation loss that might previously have threatened its solvency or lending capacity.
The indirect channel has become more important. Banks provide repo financing, prime brokerage, derivatives, collateral and margin services to hedge funds, asset managers, insurers and pension funds. If one of those clients is forced to sell government bonds, the bank may face counterparty losses, a draw on a credit line or a sudden change in the collateral supporting a loan. Banks can also be affected when NBFIs withdraw funding, reduce purchases of bank paper or change their demand for secured finance.
The Federal Reserve’s US assessment captures this mixed picture. Bank regulatory-capital ratios were near historically high levels, and broker-dealer leverage remained subdued. Dealers continued to support Treasury intermediation, with activity increasing in part through secured lending. But the same report says hedge-fund leverage remained near all-time highs and concentrated among large funds; life-insurer leverage stayed in the upper quartile of its historical distribution; and fair-value losses on bank fixed-rate portfolios, though reduced from their 2022 peaks, remained sizable.
The conclusion is not that bank resilience is illusory. It is that institutional resilience and network resilience answer different questions. Strong capital can reduce the chance that a bank fails after a bond-price decline. It cannot, by itself, ensure that every leveraged client can refinance, that every dealer will continue to make markets or that a crowded sale will not move yields sharply. Conversely, a market with vulnerable links may still function normally when dealers have capacity and funding terms remain stable.
What the map reveals—and what it cannot
A useful map does not turn every arrow into a forecast. The BIS’s advanced-economy holding figures show a structural shift, but they do not identify each investor’s net position. The Federal Reserve’s data on bank commitments to NBFIs are US estimates, and the report revised the classification of some funds after receiving new information. The measure of hedge-fund gross notional leverage is not the same as balance-sheet leverage, and neither one tells a reader exactly who would sell first.
The map also shows why aggregate calm can coexist with concentrated vulnerability. Government money-market funds are described by the Federal Reserve as historically less susceptible to large-scale redemptions than other cash-management vehicles. Certain private-credit vehicles faced higher redemption requests but generally kept those requests manageable by limiting redemptions. Dealers reported adequate market-making capacity in normal conditions. These are meaningful buffers, not proof that the network has no weak point.
The correct question when a sovereign-bond market moves is therefore relational. Which participants are holding the bond for the long term? Which have borrowed against it? Which contracts require cash after a price move? Which dealer balance sheets connect those positions? And which institutions can absorb a sale without becoming sellers themselves?
Once those questions are asked, the market can be seen as a system of linked obligations rather than a static list of owners. The next useful test is to run one yield shock through that system: from collateral and margin pressure, to sales and market depth, and then to the wider consequences for sovereign financing.