Central Bank Balance Sheet Normalisation: How Repo Facilities Replace Quantitative Easing

The Underlying Problem: Shrinking a Balance Sheet Without Breaking Money Markets

For more than a decade, central banks bought government bonds at scale — quantitative easing — to push down long-term borrowing costs when interest rate cuts alone weren't enough. That left them holding enormous bond portfolios, funded by reserves created specifically to pay for them.

Reversing that process, known as quantitative tightening (QT), raises a much harder operational question than the buying phase ever did: how do you drain hundreds of billions of pounds of liquidity back out of the banking system without triggering a cash scarcity shock in interbank lending markets?

The core challenge in one sentence: shrink the balance sheet and drain reserves, without losing control of the short-term interest rate the central bank actually targets.

Mechanism Part I: How Reserve Creation and Drainage Actually Work

A central bank's balance sheet works like any other — assets have to equal liabilities. On the asset side sit the government bonds and loans it holds; on the liability side sit the reserves it owes to commercial banks, physical currency in circulation, and its own capital.

When a central bank buys a bond under QE, it doesn't spend money it already has sitting somewhere. It creates a new reserve balance — an electronic credit entry — in the selling bank's reserve account at the central bank, and receives the bond in exchange.

Reserves aren't taxpayer cash, and they don't circulate in the wider economy the way banknotes do. They're a form of money that exists exclusively on the balance sheets of the central bank and the commercial banks that hold accounts there, used to settle payments between banks.

QT runs this process in reverse. When a bond the central bank holds matures and isn't replaced, or when the central bank actively sells a bond into the market, the corresponding reserve balance is extinguished — reserves drain out of the banking system in exactly the mechanical mirror image of how QE created them.

Mechanism Part II: The Floor System vs a Scarce-Reserve Regime

When reserves are abundant — the situation QE deliberately created — commercial banks have far more liquidity than they need for day-to-day settlement. In that environment, short-term money market rates like SONIA naturally trade right at the rate the central bank pays on reserves, because no bank needs to bid rates higher to attract scarce cash. This is the 'floor system': the interest rate on reserves acts as a floor that money market rates settle on.

Contrast that with a scarce-reserve regime, where the total pool of reserves in the system is small relative to demand. In that world, a modest, everyday shift in cash needs — a large payment settling, a temporary spike in demand for safe collateral — can send interbank lending rates spiking well above the central bank's target, because banks with a shortfall have to compete for a genuinely limited pool of available cash.

The complication QT introduces is that these two regimes aren't separated by a single, knowable line. Aggregate reserves can look comfortably sufficient in the system as a whole while individual banks — constrained by regulatory buffers like the Liquidity Coverage Ratio (LCR), or simply holding reserves in the wrong place within the system — experience real, localised shortages regardless of the aggregate total.

Mechanism Part III: How Repo Facilities Fill the Gap

A repurchase agreement, or repo, is a short-term collateralised loan structured as a sale-and-repurchase: one party sells a security today under a contract to buy it back tomorrow (or at an agreed future date) at a fixed price, with the difference in price functioning as interest.

Standing repo facilities let commercial banks initiate that trade on demand, pledging high-quality government bond collateral to the central bank in exchange for reserves, at a rate the central bank sets — rather than needing to find a private counterparty willing to lend at an acceptable rate in a moment of stress.

Instrument names matter here. The Bank of England's facility is the Short-Term Repo (STR); the US Federal Reserve's equivalent is the Standing Repo Facility (SRF). They serve the same structural purpose but are separate, differently-designed facilities — don't use the names interchangeably.

The Bank of England has been explicit that it is deliberately transitioning to what it calls a repo-led, demand-driven system for supplying reserves, rather than trying to pre-calculate exactly how many reserves the banking system needs. The STR is priced at Bank Rate with no spread, runs as a twice-weekly auction, and — per the Bank's own reporting — now sees regular borrowing of around £100 billion, with more than 30 firms typically bidding.

A complementary facility, the Indexed Long-Term Repo (ILTR), currently supplies roughly £70 billion of reserves against a wider range of collateral, with around 80 firms participating and an average of 17 firms bidding at any one auction. Together, the Bank has said, repo operations now supply around a quarter of the reserves circulating in the system — directly offsetting the drainage caused by ongoing gilt maturities and sales under QT.

This system has already been tested in practice. A gilt maturity in January 2026 alone drained around £20 billion of reserves from the banking system in a single event — and, per the Bank's own account, that drain passed smoothly, with banks replenishing the shortfall through STR and ILTR borrowing rather than short-term rates spiking. Banks also currently hold more than £450 billion of collateral pre-positioned with the Bank, ready to be drawn on for further borrowing if needed.

On precision: there is no single published figure for the 'minimum comfortable level of reserves' the banking system needs — and treating any specific number as an official threshold would be misleading. It's a dynamic target shaped by individual banks' regulatory buffer choices and demand, which is precisely why a demand-driven repo system, rather than a fixed reserve target set in advance, is the mechanism central banks have moved to.

The Durable Takeaway

The structural shift underneath all of this is a change in who decides how many reserves the banking system holds. Under the old approach, the central bank effectively set that number by deciding how many bonds to buy or hold. Under a repo-led framework, commercial banks decide, each auction, how many reserves they actually need — and the central bank supplies them on demand, at a known price, against good collateral.

That's the mechanism that lets a central bank keep shrinking its balance sheet — as the Bank of England is doing through its ongoing gilt sales programme — without needing to correctly forecast, in advance, the exact point at which reserves become scarce enough to risk a money-market disruption.

For the mechanics of how those reserves were created in the first place, how quantitative easing works covers the buying side of this cycle, and the Bank of England's own current gilt sales programme — the specific QT operation draining the reserves this repo system is built to replace — is covered in its £20bn gilt sales target announcement.