The Bank of England redraws its exit from QE

Britain’s Treasury could soon borrow to buy back its own debt from the Bank of England and cancel it. Investors would receive newly issued securities to finance the purchase. The funding requirement would remain. The bonds used to meet it could change. S04

This is one part of the framework announced on 17 September 2026. The path for running down the monetary-policy portfolio has been adopted. £120 billion of gilts, British government bonds, measured at their original purchase cost, will be retained to back banknotes. Sales to the government, however, still require a final decision. These are separate decisions at different stages of implementation. S01 S04

The proposal changes the securities offered to investors: the central bank could dispose of its holdings without requiring investors to take back the exact securities it originally bought. Understanding the change requires following the money, then separating the maturity of the bonds from the maturity of the funding behind them.

A portfolio with three destinations

Under quantitative easing, or QE, a central bank buys securities from investors and creates reserves, the balances commercial banks hold with it. In Britain, the purchases were housed in the Asset Purchase Facility (APF), a subsidiary funded by a loan from the Bank of England. Quantitative tightening, or QT, runs that arrangement down through sales and bonds reaching maturity. S06

The allocation published on 17 September divides £488.2 billion at original purchase cost into three groups. £221.7 billion will be held until redemption before 2035. £146.5 billion is earmarked for sale. £120 billion will back banknotes. The sale portfolio matures between 2035 and 2049; the banknote portfolio extends to 2071. The holding of the 1.75% 2049 gilt is split between the latter two groups. S02

Adding the first two categories leaves £368.2 billion to be removed from holdings for monetary-policy purposes. None of these amounts is a current market valuation or a measure of face value, the principal contractually due at maturity. A bond bought at a high price years ago can now trade for considerably less.

The Monetary Policy Committee, or MPC, plans to eliminate this monetary portfolio by 2034, with an average annual reduction of £46 billion, including £20 billion of sales. Redemptions account for the remainder. On interest rates, members remain divided: Bank Rate was held at 3.75% by six votes to three. The QT programme was approved unanimously. S01

The immediate suspension is narrower than a halt to QT. APF sale auctions are paused while the arrangements are prepared; bonds can still mature. Operational details must be published by April 2027, even if the government-purchase proposal is ultimately rejected. S02

One public debt supplier instead of two

With direct APF sales, two public bodies offer gilts to investors. The Treasury issues new debt, while the central bank sells some of the older stock. The proposal would bring that supply together through the Debt Management Office, or DMO, Britain’s debt agency.

The Treasury’s letter sets out the proposed chain. The DMO would buy APF gilts at market prices through the Debt Management Account. It would then sell them to the National Loans Fund, the government account for borrowing and lending, for cancellation. The Treasury would authorise corresponding debt issuance to fund those purchases. S04

The operation can be understood as a substitution. On one side sits an inherited portfolio, assembled at different dates, with coupons and maturities already fixed. On the other is a cash requirement that can be folded into the debt agency’s issuance programme. The amount of cash needed does not, by itself, determine the securities used to raise it.

That is the purpose of debt management: choosing among instruments, maturities and risks rather than automatically taking the lowest interest rate available today. The DMO’s mandate seeks to minimise financing costs over the long term, subject to risk and consistency with monetary policy. S12

The Treasury states that the sales model would not change the overall supply of gilts to the private sector. That comparison concerns alternative ways to implement the same QT path. It does not establish that the whole package, including the £120 billion retained for banknotes, has no effect compared with selling everything. S04

A single public seller does not create an additional source of finance. It changes the intermediary between the inherited portfolio and the ultimate investor.

Cancellation does not fund the loss

The transfer price is crucial. Buying the bonds at their original cost when market prices are lower would shift the loss artificially to the Treasury. The proposal instead specifies market-price transactions and preserves the government indemnity protecting the APF. S03

Consider an entirely fictional example, in units of pounds. The APF bought a bond with a face value of 100 at par, funded by a loan of 100. The bond now trades at 70. Ignore coupons, accrued interest, fees and all other cash balances to isolate the transaction.

A sale raises 70. Repaying the original loan requires another 30. In this example, the Treasury indemnity supplies the difference. Changing the buyer does not fill the gap: whether a private fund or the DMO buys at 70, the sale proceeds are identical. This is a simplified application of the public guarantee described in the Treasury’s accounts. S15

Now assume that the Treasury funds both the purchase and the indemnity with newly issued debt at par. It needs 70 to acquire the bond and another 30 to complete repayment of the APF loan. Total private cash financing is 100. Under a direct sale, investors would have paid 70 for the old bond and supplied the 30 needed for the indemnity: the same cash requirement, met through a different set of transactions.

The unit of comparison matters. After the direct sale, investors hold an old bond with a face value of 100 that they bought for 70. Under the buyback model, that bond is cancelled and new securities take its place. The same cash raised does not imply the same face amount of debt or the same future coupons. A reduction in face value is not an equal-sized economic gain.

This is why APF transfers, the deficit and public debt cannot be used interchangeably. The independent Office for Budget Responsibility, or OBR, distinguishes interest costs, valuation effects and indemnity calls across fiscal aggregates. The example above follows cash. It does not predict the final entries in the national accounts. S13

The maturity question has two baselines

The potential benefit of the DMO route lies in the composition of supply. Investors may be reluctant to buy an old, very long-dated bond without being unwilling to fund a new, shorter issue. They would provide the cash, but accept interest-rate exposure for less time.

Think of a payment promised far in the future. When the interest rate available on new investments rises, that old promise becomes less attractive at its previous price. The more distant the payment, the larger the effect can be. Duration summarises this exposure: it measures a bond price’s approximate sensitivity to a change in yields.

As an illustration, not an announced DMO policy, replacing the return of a thirty-year bond to the market with a new five-year issue would reduce the duration investors must absorb. That could make the financing easier to place. The government’s trade-off would be an earlier refinancing date. The debt-management report identifies this balance between cost and rollover risk. S12

There is, however, another baseline: leaving QE in place. Viewed across the consolidated public sector, much of the long-dated debt bought by the central bank was replaced by bank reserves whose remuneration follows Bank Rate. The OBR has highlighted how that transformation increased the sensitivity of public finances to rising interest rates. S14

In this simplified comparison, five years is short against thirty years, but long against a funding cost that can reset immediately. Describing the proposal as “shortening government debt” without specifying the alternative can produce the wrong conclusion.

Its market impact remains a separate empirical question. A term premium compensates investors for the risks of holding a long-term bond rather than rolling over short-term investments. In its July 2026 report, the Bank estimated that QT had added 20 to 30 basis points to term premia since 2022. A basis point is 0.01 percentage point. That estimate uses the estimated effects of announcements and auctions, with assumptions about persistence. It is neither an estimate of the 17 September decision nor a forecast of an equal decline in yields. S08

A plausible direction is not a measured magnitude. Long-term rates also reflect expected inflation, the prospective path of short rates, fiscal issuance and investor demand. Leverage in the gilt market adds another source of fragility to price moves. Separating this framework’s contribution will require more than watching prices on announcement day.

Banknotes give retained gilts a different job

The £120 billion portfolio addresses a different question: which assets should the central bank keep against the physical money it issues? The weekly statement records £98.962 billion of banknotes on 16 September 2026, or approximately £99 billion. This is the balance-sheet item, not merely cash held in domestic households’ wallets. S11 S16

A banknote pays its holder no interest. A persistent stock of notes can therefore fund assets differently from an interest-bearing bank balance. Any individual note can be returned to the central bank, but the entire stock does not necessarily leave together. The relevant stability is that of the aggregate liability.

That characteristic helps explain the choice to retain very long-dated gilts. A document published in 2008 already describes long-dated bonds backing banknotes, before QE. S17 The September decision avoids selling the entire portfolio only to make separate purchases for this other purpose. The retained securities nevertheless remain in the APF, covered by the Treasury indemnity. S03

The British accounting chain is unusual. Banknotes sit in the Issue Department, which holds an internal deposit with the Banking Department. The latter holds the loan to the APF. APF gilts therefore back banknotes indirectly, through these successive claims. Over time, gilts bought in the secondary market will be held directly by the Issue Department, as the transition proceeds from the retained bonds’ maturities beginning in 2049. S05

This describes the assets behind the currency; it does not turn a banknote into a unit in a bond fund. More importantly, eliminating holdings “for monetary-policy purposes” is not the same as eliminating all central-bank holdings of government debt.

One qualification changes the financial interpretation: the entire £120 billion is not already funded by non-interest-bearing notes. The calibration allows for expected growth in banknote demand and an additional buffer. While demand remains below that amount, part of the portfolio continues to back reserves. The £120 billion is also an original-cost figure, not a market valuation. S05

A new label is not a reset of the losses

Multiplying £120 billion by Bank Rate and calling the result “savings” would be tempting. It would assign notes a funding role they do not yet fulfil in full and treat a partly pre-existing arrangement as a new benefit. It would also overlook income and costs distributed across different accounts.

Reclassification does not cancel the APF loan. The Governor’s letter preserves the accounting arrangements, the indemnity and quarterly cash transfers with the Treasury. Elsewhere in the chain, the net income generated by issuing banknotes, known as seigniorage, reaches the Treasury through a separate channel. A public-sector assessment must bring these flows together without counting them twice. S03 S05

Past transfers show the scale involved. Cumulative net cash paid by the APF to the Treasury peaked at £123.9 billion at the end of September 2022. By the end of June 2026, the cumulative net balance was down to £16.2 billion. The difference, £107.7 billion, measures the erosion of that balance by transfers in the opposite direction. It is neither an annual loss nor the total economic cost of QE. Published on 4 August, these data predate the reform. S07

They cannot price the new plan’s savings. Earlier projections use runoff paths and interest-rate assumptions that do not constitute an evaluation of the 17 September framework.

Holding a bond can avoid a low-price sale today. It does not guarantee a better lifetime outcome. Coupons, funding costs, eventual redemption and their timing still have to be compared. Deferring a loss and avoiding one are different propositions.

The opposite criticism deserves equal care. APF losses alone do not establish that QE made the country poorer. As the OBR notes, a comprehensive assessment would need to include effects on economic activity, tax receipts and financing conditions. Our analysis of who pays QE losses examines the other routes through which these costs reach public finances and banks. S13

Running down QE while lending to banks

A third change concerns reserves. For several years, the Bank has been preparing a framework in which their quantity responds more directly to banks’ demand. It supplies reserves through secured lending rather than relying primarily on bond purchases. This is the repo-led operating framework. Our guide to repo and the creation of liquidity explains the collateral supporting these transactions. S09

A repo is different from QE. A bank obtains reserves for an agreed period and provides assets as collateral. The central bank holds a loan; it does not permanently acquire the borrower’s securities portfolio. The Short-Term Repo supplies funds for one week. The Indexed Long-Term Repo provides six-month funding, with remuneration linked to Bank Rate under the facility’s terms. S10

On 16 September 2026, reported balances were £124.141 billion in STR and £84.919 billion in ILTR, a combined £209.060 billion. These are outstanding loans at that date, not new liquidity injected during a single week. S11

The coexistence matters. QE bond holdings can decline while secured lending rises. The central bank is then removing one type of asset and adding another. Abundant reserves or heavy repo usage alone cannot demonstrate that QE has restarted or that banks are panicking.

An indexed-rate loan brings the yield on the central bank’s asset closer to its variable reserve funding cost. That better match does not eliminate every risk. Counterparty quality, collateral quality, valuation haircuts and renewal conditions still matter. A lending framework must manage those exposures rather than conflate them with the risk of holding long-dated bonds. S09 S10

The boundary will be tested in implementation

The stated division of responsibilities remains intact. The MPC chooses the monetary path; the Bank’s Executive decides on implementation and banknote backing; the Treasury sets the DMO’s financing remit. The committee retains exceptions to the path, including circumstances in which Bank Rate alone cannot deliver the inflation target or markets are severely distressed. S01 S03

Fiscal dominance means monetary policy is subordinated to the government’s financing needs. The announced framework retains several institutional boundaries: market-price transactions, an MPC-determined runoff path and a structural rationale for the banknote portfolio. Their resilience will also depend on future choices between these objectives and public funding constraints.

The useful tests will be concrete: buyback pricing rules, the new issues funding them, their maturities and any changes in pace. Easier financing today could leave a future government with more frequent refinancing. Conversely, better coordination could reduce transaction costs without weakening the monetary path. Both possibilities remain open until the operational arrangements are published.

September’s framework separates three requirements: exiting the monetary portfolio, backing banknotes and supplying reserves to banks. Investors will continue to fund the public sector. They may simply receive different bonds. The testable question is which securities replace the old portfolio, and which risks stay with the public sector.

Sources

1. S01 · Bank Rate maintained at 3.75%: September 2026 Monetary Policy Summary and MinutesBank of England · 2026-09-17. Meeting ending 16 September 2026; summary and paragraphs 31 to 48.
2. S02 · Asset Purchase Facility: Gilt Sales. Market Notice, 17 September 2026Bank of England · 2026-09-17. Opening paragraphs; Table 1, category totals and footnote on the 1.75% 2049 gilt.
3. S03 · Letter from the Governor to the Chancellor: APFBank of England / HM Treasury · 2026-09-17. PDF pages 1 to 2 (runoff), 4 to 5 (banknotes), 6 to 7 (sales and indemnity).
4. S04 · Letter from the Chancellor to the Governor: APFHM Treasury · 2026-09-17. PDF pages 1 to 2 (banknote portfolio) and page 3 (purchase, cancellation and financing).
5. S05 · The promise to pay: what backs banknotes?Bank of England · 2026-09-17. “Current arrangements”, “The new banknote-backing strategy”, “Why £120 billion?”, Boxes B and C and the transition section.
6. S06 · QE at the Bank of England: a perspective on its functioning and effectivenessBank of England · 2022-05-18. Implementation, creation of reserves and portfolio-balance transmission.
7. S07 · Asset Purchase Facility Quarterly Report: 2026 Q2Bank of England · 2026-08-04. 1 April to 30 June 2026; “Cash-flow arrangements between the APF and HMT” and cash-flow projections.
8. S08 · Monetary Policy Report: July 2026Bank of England · 2026-07-30. Box G: QT estimates, cumulative period and identification method.
9. S09 · Transitioning to a repo-led operating framework: discussion paper feedback statementBank of England · 2025-06-11. Demand-driven reserve supply, STR and ILTR design.
10. S10 · Bank of England Market Operations Guide: Our toolsBank of England · Accessed 2026-09-20. “Short-Term Repo” and “Indexed Long-Term Repo”; current guide consulted on 20 September 2026.
11. S11 · Bank of England Weekly Report: 16 September 2026Bank of England · 2026-09-17. £ million table: STR, ILTR, notes and reserve balances. Positions at 16 September 2026.
12. S12 · Debt Management Report 2026-27HM Treasury · 2026-03-03. Sections 2.3, 2.4 and 2.7; Annex B on instruments and risks.
13. S13 · Fiscal accounting for quantitative easing and tighteningOffice for Budget Responsibility · 2023-11. November 2023 Economic and fiscal outlook, Box 4.5, page 121; accounting distinctions and final paragraph.
14. S14 · Fiscal risks and sustainability: July 2025Office for Budget Responsibility · 2025-07. Paragraph 5.24, “Higher interest rate sensitivity”.
15. S15 · Annual report and accounts: year to 31 March 2025HM Treasury · 2025-07-21. Financial instrument notes, “Liquidity risk” and the APF indemnity; year ending 31 March 2025.
16. S16 · Further details about the Bank of England Weekly Report and consolidated balance sheet dataBank of England · Updated 2023-01-31; accessed 2026-09-20. Definitions of notes, reserves and APF loan; last updated 31 January 2023.
17. S17 · Markets and operations: Quarterly Bulletin 2008 Q1Bank of England · 2008-Q1. Quarterly Bulletin 2008 Q1, printed pages 22 to 23, PDF pages 17 to 18: “The injection of reserves via bond-purchase OMOs”.

Limitations

This analysis is current to 20 September 2026. Government purchases of APF gilts remain a proposal; the MPC path and the banknote allocation have been decided. Pricing rules, the detailed timetable, new-issue maturities and fiscal effects have not all been published. No yield benefit or savings estimate is produced here.

The 17 September allocation uses original purchase cost. Loan and banknote balances are stocks at 16 September. APF cash transfers run only to 30 June. These series are not added together to reconstruct a complete balance sheet. The weekly APF loan entry must not be confused with the gilt allocation in the implementation notice; this analysis has not established their reconciliation.

The 100/70/30 transaction and the five-year versus thirty-year comparison are fictional. They isolate cash financing and maturity choice respectively, without modelling national accounts or the entire portfolio. Other central-bank operations can offset a reserve reduction associated with loan repayment.

The British authorities are the authoritative sources for their decisions, but also designed the framework. Their objectives for cost, market functioning and independence are not already demonstrated outcomes. OBR analysis provides a separate institutional check on fiscal mechanisms, not an assessment of the September plan. A complete macroeconomic evaluation of QE lies outside this article’s scope.