For a decade, financial-stability officials described two problems as if they lived in separate buildings. One was sovereign debt: too much of it, growing too fast. The other was the migration of finance out of banks and into funds, insurers, and vehicles that supervisors can see only dimly. The Bank for International Settlements’ 2026 Annual Economic Report, released on 28 June, put them in the same room. Its verdict, in effect: high public debt and the rise of non-banks are no longer two risks. They are one loop.
The problem
Start with the numbers, because they are unusually clean.
Global public debt reached just under 94% of GDP in 2025 and is on track to cross 100% by 2029 — a year earlier than the IMF projected only twelve months ago (IMF Fiscal Monitor, April 2026). The accumulation is concentrated in the largest economies, and it is being driven by spending that does not reverse easily: defense, aging, industrial policy, and a rising interest bill on top of all of it.
Now the other side. Non-bank financial intermediaries — investment funds, pension funds, insurers, money-market funds, and the vehicles in between — held 51% of global financial assets in 2024, or $256.8 trillion, according to the Financial Stability Board’s latest monitoring report (FSB, December 2025). That sector grew 9.4% over the year — double the 4.7% pace of banks. The “other financial intermediaries” bucket, which includes money-market and hedge funds, grew 11% to $169.4 trillion. The narrow measure the FSB flags for bank-like risk rose 12% to $76.3 trillion.
Put the two facts together and the nexus appears. Someone has to buy the record supply of government bonds. Banks, constrained by capital rules and their own balance sheets, are not the marginal buyer they once were. Increasingly, the marginal buyer is a non-bank.
The analysis
This would be unremarkable if non-banks held sovereign debt the way a pension fund is supposed to — patiently, unlevered, indifferent to a bad week. Much of it is held exactly that way. The risk lives at the margin, in the corner of the sector that funds long positions with short money.
The mechanism is not theoretical; it has misfired twice in living memory. In March 2020, the “dash for cash” saw funds sell Treasuries into a market that could not absorb them, forcing the Federal Reserve to become buyer of last resort and stand up emergency facilities to funnel cash back into the system (Federal Reserve Bank of New York). In September 2022, UK liability-driven investment funds — pension vehicles using leverage to match their liabilities — hit margin calls as gilt yields spiked, and sold the very bonds whose fall had triggered the calls. The Bank of England intervened in the gilt market to halt what it called a “self-reinforcing spiral” (Bank of England).
Both episodes share a structure. A non-bank holds government debt. Stress forces it to sell. The selling worsens the stress. The central bank, unwilling to watch the sovereign market seize, steps in and buys — backstopping an institution it does not supervise, in a market it did not choose to make.
That is the loop the BIS is warning about, now with more debt on one side and a bigger non-bank sector on the other. The sovereign needs the non-bank to fund it. The non-bank, under stress, needs the central bank to fund it. And the central bank, having promised to hold inflation down, finds itself buying the bonds of a government whose borrowing it is trying not to accommodate. Each actor is behaving rationally. The system is not.
There is a second-order twist. When a central bank intervenes to calm a bond market, it blurs the line between financial-stability policy and monetary policy — between “we are repairing market functioning” and “we are financing the state.” Every intervention makes the next one more expected, and expected backstops are subsidies. They reward exactly the leveraged, liquidity-mismatched positioning that made the backstop necessary in the first place. The insurance changes the behavior of the insured.
The implications
The uncomfortable conclusion is that macroprudential regulation has a blind spot the size of half the financial system. The post-2008 rulebook was built for banks: capital ratios, liquidity coverage, resolution regimes. Non-banks are supervised more lightly, disclose less, and — critically — sit outside the perimeter where a central bank has both the tools and the mandate to act before a crisis rather than during one. Yet they are now the dominant holders of the asset that sits at the center of the system.
Three things follow.
First, the tools are converging on the problem slowly. Regulators are pushing on non-bank leverage, money-market fund reform, and margin preparedness for exactly the LDI-style vulnerability that surfaced in 2022. This is sensible and overdue, but it is regulating the plumbing, not the pressure. The pressure is the debt.
Second, the sovereign side has less room than the debt-to-GDP number suggests. A government whose bonds are held by patient banks can absorb a bad auction. A government whose marginal buyer is a leveraged fund that can be forced to sell into weakness has a more fragile bid than the headline holdings imply. Debt sustainability is not only about the level; it is about who holds the paper and how quickly they can be made to sell it.
Third, central-bank independence is quietly on the line. Each time a central bank backstops a bond market to preserve stability, it takes another step toward being the market-maker of last resort for sovereign debt — a role that looks, from a distance, hard to distinguish from monetary financing. The BIS’s framing is, at bottom, a plea to defend that distinction while it still exists.
The reassuring version of this story is that none of it is new: non-banks have always intermediated, governments have always borrowed, and central banks have always cleaned up afterward. The BIS’s point is that the three have grown entangled enough that the cleanup is no longer occasional. Watch the next liquidity event — the first forced-seller episode in a sovereign market since debt crossed these levels — for whether the backstop arrives faster and larger than the last. If it does, the nexus will have stopped being a chapter in an annual report and started being the operating model.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.