Financial stability buy/sell tools: a gilt market case study

1: Introduction

Background on central bank financial stability tools

Traditionally, central banks have focused on the commercial banking system as the most common source of financial instability because most countries’ financial systems have been dominated by these banking entities for much of their history. Central banks have therefore typically aimed to mitigate financial stability risks by acting as lender of last resort (LOLR), to commercial banks primarily, following Bagehot (1873).

Since the global financial crisis (GFC) of 2008–09 there has been increasing discussion around the role of central banks in responding to liquidity shocks that threaten financial stability, and how it must evolve to reflect the changing structure of financial markets. A key component of that changing structure has been the growing role of non-bank financial institutions (NBFIs), and their influence on vulnerabilities in core markets central to the stability of the financial system. In response, central banks have gradually broadened their focus from providing liquidity insurance to banks to backstopping market liquidity more broadly in situations where severe dysfunction threatens financial stability.^{footnote [2]}

Much work has been done internationally on this topic since the GFC. The Financial Stability Board (FSB) initially undertook significant work to assess and address the risks from NBFIs following the 2008–09 crisis, creating a system-wide monitoring framework to identify the build-up of systemic risks in the sector. Early contributions to thinking about how central banks might need to extend their traditional role to deal with episodes of severe market dysfunction, frequently termed ‘market maker of last resort’ (MMLR), were made by Buiter and Sibert (2007), Tucker (2009 and 2014) and Mehrling (2014), among others.

Dysfunction in core funding markets following the spread of Covid in Spring 2020 triggered a range of central bank interventions to prevent a sudden ‘dash for cash’ from undermining monetary and financial stability^{footnote [3]} and accelerated these debates.^{footnote [4]} While these interventions were successful in restoring market functioning, they highlighted a need to tackle the sources of market dysfunction more comprehensively. The FSB enhanced its NBFI programme, following the March 2020 ‘dash for cash’, with a focus on assessing the adequacy of current policy tools to address the specific issues that contributed to market turmoil in 2020.

Alongside that core work, there was a recognition that central banks need to expand their toolkits to ensure they are better placed to provide a backstop to market liquidity through more targeted interventions in core markets, particularly where future shocks – unlike the Covid shock – did not call for an expansion of the monetary stance. While there are historical precedents for central banks intervening to restore market functioning via asset purchases at critical moments, many of these operations involved one-way asset purchases, without any exit mechanism other than awaiting the maturity of those assets. The need for such tools was identified by, among other bodies, the Bank of England’s Financial Policy Committee (FPC (2020)) and expanded on by a Bank for International Settlements (BIS) working group on tools for market dysfunction (BIS (2022)), drawing together 25 central banks co-led by Andrew Hauser (Bank of England) and Lorie Logan (Federal Reserve). Further contributions on principles for such interventions have followed, including Hauser (2021, 2022, 2023a), Logan (2023), Duffie and Keane (2023), and Duffie (2023).

In 2022, the Bank of England put this theory into practice, to protect UK financial stability following severe dysfunction in the UK government bond (or ‘gilt’) market. The Bank’s temporary and targeted gilt purchase operations were successful in restoring orderly market functioning by breaking adverse feedback loops in the market and buying time for LDI funds^{footnote [5]} – whose vulnerabilities lay at the heart of the crisis – to improve their resilience. By January 2023 the Bank had fully unwound those purchases, delivering on its commitment to ensure the financial stability intervention was temporary, targeted, and had minimal spillovers to the stance of monetary policy.

Context for the Bank’s intervention

In Autumn 2022, the UK government bond market exhibited extreme volatility following the UK Government’s fiscal statement of 23 September. The sequence of events is covered in detail in other Bank of England publications.^{footnote [6]}

In summary, the gilt market became imbalanced, characterised with excessive one-way selling pressures. As a result, gilt market intermediation started to break down and market functioning deteriorated rapidly. The quoted bid-ask spread – a measure of liquidity – on 30-year gilts increased from around 0.5 basis points on 22 September to 2.5 basis points on 11 October, almost twice as high as the previous peak seen during the ‘dash for cash’.

The speed and scale of the moves in gilt yields was unprecedented. By 28 September, the yield on 30-year gilts had risen by 130 basis points in just three days of trading – a move three times larger than any other historical move over a similar period.

The very large and rapid rise in gilt yields meant sharp falls in gilt prices. That exposed vulnerabilities in LDI funds, which are large holders of gilts and that are invested in by many UK pension funds.^{footnote [7]} As gilt prices fell, the net asset value of these levered LDI funds dropped rapidly and their leverage increased further. Challenges were experienced in meeting increased margin calls on their derivative and repo positions. Pooled funds were often the most levered and most exposed to the fall in prices. A number of funds rapidly headed towards a net asset value of zero at which point lenders would seize and liquidate collateral.

Mechanisms to recapitalise LDI funds by transfers in of cash from the underlying pension scheme investors were too slow for the speed of the move in gilt prices. Deleveraging was therefore required to stem the decline in net asset values. That required selling gilts, and an adverse feedback loop took hold. Desired sales volumes of LDI funds rapidly accelerated to more than the volume of typical trading days and the willingness of others to intermediate that spiralling flow also declined rapidly. The market became highly illiquid. The Bank’s FPC deemed conditions to be a material threat to UK financial stability.^{footnote [8]}

Outline of the Bank’s intervention

On 28 September, following a recommendation from the FPC, the Bank announced a gilt market intervention on financial stability grounds, using temporary and targeted asset purchases to restore market functioning in long-maturity conventional gilts. The aim of this intervention was to temporarily ‘support market functioning as a backstop’,^{footnote [9]} in order to buy time for LDI funds to build resilience, thereby reducing risks of contagion to credit conditions to UK households and businesses. On 11 October, inflation-linked gilts (‘linkers’) were added to these operations.

In total, the Bank bought £19.3 billion of gilts between 28 September and 14 October 2022, comprising £12.1 billion of conventional gilts and £7.2 billion of index-linked gilts. These were subsequently sold back to the market in a timely but orderly way, over 12 trading days between 29 November 2022 and 12 January 2023. Chart 1 shows the timeline of key events alongside UK index-linked and conventional 30-year yields, US 30-year yields (for comparison) and the Bank’s purchases and sales.

Chart 1: Financial stability gilt purchases and sales alongside UK and US 30-year yields ^(a)

Footnotes

• Sources: Bloomberg and Bank calculations.
• (a) This chart shows a daily time series for the duration of the financial stability intervention and unwind. While unwind of the financial stability portfolio took place over a longer period, the sales were completed over 12 days of operations compared to 13 days for purchases.
• (b) 22 September: MPC votes at its September meeting to increase Bank Rate to 2.25% and confirms the start of Asset Purchase Facility (APF) active gilt sales (also referred to as ‘quantitative tightening’, or QT).
• (c) 23 September: Fiscal event – The Chancellor announces the Government’s Growth Plan.
• (d) 26 September: Governor statement that Bank is monitoring developments in light of significant repricing in the gilt market.
• (e) 28 September: FPC recommendation for the Bank to intervene in the gilt market on financial stability grounds. 28 September: Bank intervention begins with temporary purchases of long-maturity conventional gilts. The end date of purchases is set for 14 October. Previously planned APF QT gilt sales are postponed.
• (f) 10 October: Bank increases maximum daily operation size from £5 billion to £10 billion for the final week of operations. Bank also announces launch of Temporary Expanded Collateral Repo Facility, which would run until November 2022.
• (g) 11 October: Bank expands temporary gilt purchases to include index-linked gilts for the final four days of the intervention.
• (h) 14 October: Bank’s financial stability gilt purchases end on pre-set deadline.
• (i) 1 November: Previously postponed APF QT gilt sales begin.
• (j) 10 November: Bank announces plans to unwind financial stability purchases using demand-led approach in prompt but orderly fashion.
• (k) 23 November: Bank sets out operational arrangements for sales of financial stability gilt holdings.
• (l) 29 November: Bank’s sales of financial stability gilt holdings begin.
• (m) 30 November: Co-ordinated regulatory action by The Pensions Regulator, Financial Conduct Authority, and other National Competent Authorities on interim resilience of LDI funds.
• (n) 13 December: Publication of FPC Record and Financial Stability Report.
• (o) 12 January: Bank’s financial stability purchases fully unwound.

The remainder of this article describes the design and implementation of financial stability operations in practice. Section 2 sets out the core principles behind extraordinary central bank financial stability interventions to restore orderly functioning in core markets, which guided the Bank’s design of these operations. Sections 3, 4 and 5 detail how these principles were implemented in the Bank’s intervention during the LDI crisis – namely that purchases were temporary, targeted and applied backstop pricing. Section 6 covers the demand-led unwind of the financial stability purchases. Section 7 concludes with some lessons learned.

2: Core principles for central bank financial stability operations to restore orderly market functioning

Principles for central bank financial stability interventions to restore orderly market functioning

This section sets out some key principles that guided the Bank’s thinking on extraordinary financial stability interventions aimed at restoring orderly functioning in core funding markets.^{footnote [10]} These principles build on the BIS (2022) work on central bank tools for market dysfunction.

1: Extraordinary central bank financial stability interventions to restore functioning in core markets are a backstop and should not substitute for private self-insurance by economic agents, including non-banks.

Although central banks are increasingly contemplating new tools to deal with market dysfunction (eg see Hauser (2021) and Duffie and Keane (2023)), it is imperative that firms themselves – including non-bank financial institutions that make up the increasingly important world of market-based finance – have foremost responsibility to build up sufficient ex ante resilience to all but the most severe shocks. In other words, the bar to extraordinary central bank intervention is very high. The key feature of these interventions is that they should act as a backstop, with private self-insurance as the frontstop. This is because excessive public insurance dampens private incentives to manage risks, resulting in moral hazard, and can impose unacceptable costs to the public purse.

Nonetheless, there will be ‘tail’ scenarios (very low probability but very high impact) where an economic shock results in extreme self-reinforcing dynamics in markets (such as an asset fire-sale spiral), beyond what is warranted by economic fundamentals. If such dynamics cause severe dysfunction in core funding markets that are critical to provision of financial services, this may threaten the stability of the financial system as a whole. In such a severe situation, extraordinary central bank intervention may be justified on the ground that the public cost of not intervening exceeds the cost of intervening.

Where the central bank does have to intervene, the design of the backstop intervention should, to the extent possible, limit the central bank’s footprint in private markets; limit the potential for distortions to market pricing; and support private mechanisms for market-making and price discovery.

2: Where they are likely to be effective, central bank financial stability interventions in the form of collateralised lending operations are preferable, as a first port of call, to asset purchase/sale operations.

Central banks have two main tools in responding to severe market dysfunction:

• Lending operations, which aim to prevent asset fire sales by allowing firms to post these assets as collateral at the central bank in return for borrowing liquidity in the form of central bank reserves, for a fixed term.^{footnote [11]}
• Outright asset purchases which the central bank unwinds once risks to market functioning have receded.

Subject to their effectiveness, lending operations are preferable for several reasons:

• They pose lower risks to the stance of monetary policy. Because lending operations have a built-in exit date (determined to the maturity of the loan), there is less uncertainty around how long-lived the injection of reserves into the financial system will be, relative to asset purchase operations.
• They pose fewer moral hazard risks. Unlike asset purchases which remove assets (and their associated risk) entirely from private firms’ balance sheets, in central bank lending operations the counterparty retains the asset in question on its balance sheet following the maturity of the loan. This incentivises firms to prudently manage the risks associated with the assets they hold, and reduce the expectation that the central bank will step in to purchase assets in a stress.
• They pose lower risks to public funds. From the central bank’s perspective, lending operations pose significantly lower financial risks. This is because under collateralised lending operations, the central bank’s exposure to the asset in question is contingent on the counterparty defaulting, and the central bank can manage its contingent exposure to market, credit and counterparty risks via margin calls and haircuts applied to the collateral. In contrast, when buying and selling assets, the central bank is directly exposed to the bond issuer, and to any outright mark-to-market losses in the asset value in the period between purchasing the asset and selling it.
• They may be preferable for the NBFIs themselves. This is because asset fire sales can be very costly for NBFIs, who may prefer to borrow liquidity in a stress rather than sell assets at a discounted price and then purchase them back at a higher price.

Consistent with this, the Bank recently outlined ambitious plans for tackling systemic risks in market-based finance by developing a new lending tool for NBFI, starting with UK insurance companies and pension funds.

However, lending operations may not always be effective. In particular, if the need for de-leveraging by financial market participants is driving market dysfunction, lending operations would not be attractive for participants as they would increase their leverage. And some market participants may be prohibited by regulation from taking on leverage. As set out in Hauser (2022), the Bank considered lending via banks as the primary response to the LDI crisis but deemed that asset purchases were likely to be the most effective tool because the imperative to deleverage was driving LDI funds’ behaviour.

3: Where central banks need to undertake asset purchase operations for financial stability reasons, these must be clearly distinguished from asset purchases for monetary policy reasons, in design, impact and communication.

Asset purchases for financial stability reasons fundamentally differ in motivation from asset purchases for monetary stability reasons. The former aim to narrowly tackle severe market dysfunction while minimising spillovers to the stance of monetary policy. The latter aim to adjust the stance of monetary policy in pursuit of a policy goal – in the Bank’s case, the 2% inflation target. To achieve these differential goals, key design features such as the duration of purchases; the plan for exiting the purchases; the assets under consideration; and the pricing approach are likely to vary across the two types of operations. Communications (and where possible governance arrangements) should also make clear the distinction, to help market participants understand the central bank’s motivation and reaction function.

4: Exceptional central bank financial stability interventions to restore market functioning should be coupled with action by regulators and macroprudential authorities to identify and remedy underlying vulnerabilities that contributed to the dysfunction.

This may include taking steps to improve the resilience of sectors or markets, where appropriate.

Application of these principles to the 2022 LDI intervention

With these principles in mind, to address the severe gilt market dysfunction observed in September 2022, the Bank’s financial stability intervention took the form of asset purchases and had a number of key features:

• Temporary: The Bank defined a strict, time-limited, purchase period ex-ante, lasting 13 working days from 28 September to 14 October 2022. This period was calibrated to allow sufficient time to allow LDI funds and their pension fund investors, working at speed, to de-lever and improve their balance sheet position. The Bank also committed to unwinding those purchases once orderly market functioning had been restored, ensuring that its holding period was also time limited.
• Targeted: Assets eligible for purchase were limited to those segments of the gilt market where dysfunction – and hence risk to financial stability – was greatest. To that end, the Bank began by purchasing long maturity (20-year+) conventional gilts. The Bank subsequently expanded its purchases to include index-linked gilts, once it became clear this was necessary to restore orderly market functioning.
• Backstop pricing: The Bank sought to buy only as much as necessary to restore market functioning. To achieve this, purchase operations were implemented using a ‘backstop pricing’ approach, in which the Bank only bought at relatively distressed prices. This meant that, while the Bank initially announced auctions with capacity for up to £65 billion of purchases, its pricing approach meant market forces ultimately determined how much, in practice, the central bank backstop would be called upon.^{footnote [12]}
• Timely but orderly unwind: Once risks to market functioning had subsided, the Bank promptly began unwinding the portfolio on 29 November 2022. Unwind was implemented via a demand-led approach that allowed market participants to express interest in any of the gilts held in the portfolio via reverse enquiry windows. This demand-led approach ensured unwind commenced promptly but did not rekindle market dysfunction. The portfolio was fully unwound over 12 working days during a four-week period comprising 11 reverse enquiry windows and a final day of bilateral sales, ending on 12 January 2023.
• Market intervention combined with regulatory response to improve resilience: The Bank’s intervention restored orderly functioning in the gilt market, buying time for LDI funds to deleverage and improve their balance sheet position. This was followed by a co-ordinated regulatory and supervisory response, involving The Pensions Regulator, Financial Conduct Authority, and Bank of England, to identify vulnerabilities in LDI funds and pension funds, and take remedial action to increase their resilience to future gilt market volatility (Bank of England (2023a)).

Table A, sets out how the design of the operations was targeted to address financial stability risks, and how this differed from quantitative easing (QE) asset purchases for monetary policy purposes.

Sections 3–6 below focus on key features of the design of the market intervention in more detail, setting out how the Bank’s financial stability operations were designed in practice to i) be temporary, ii) be targeted, iii) utilise backstop pricing, and iv) be unwound in a timely but orderly, demand-led way.

3: Temporary intervention

The importance of a temporary financial stability intervention

A financial stability (FS) intervention is intended to buy time for the underlying vulnerabilities that led to the dysfunction to be addressed. In the case of the Bank’s 2022 gilt market intervention, the aim was to provide the affected LDI funds and their pension fund investors with time to put the positions of the LDI funds on a sustainable footing and increase their resilience to future stress, halting the destabilising dynamic caused by their actions.

In addition to ensuring LDI and pension funds had a strong incentive to address LDI vulnerabilities at pace, the time-limited purchase period and prompt unwind of the intervention was also vital in ensuring that any undesirable spillovers to the stance of monetary policy were minimised.

As set out in Chart 1, the entirety of the Bank’s intervention – from the announcement of the operation, to the completion of purchases, to the sale of all holdings in this financial stability portfolio – took 16 weeks. Combined with the critical element of backstop pricing (see Section 5), which ensured that the Bank only purchased as much as needed to restore market functioning, the strictly time-limited nature of the intervention meant that the extent of the Bank’s additional holdings in the gilt market and the injection of central bank reserves to fund these were not just relatively small but also short-lived. This curtailed their impact on the overall stance of monetary policy. It also curtailed the risks taken on the Bank’s balance sheet.

Chart 2 illustrates how limited the FS asset purchases were relative to the stock of QE assets, both in size and duration.

The end of the temporary purchase period allowed active sales of gilts for monetary policy reasons (part of quantitative tightening (QT)) to commence on 1 November, following a brief postponement with the launch of the temporary financial stability intervention.^{footnote [13]}

Chart 2: Comparing QE/QT and FS purchase portfolios

Temporary purchase period

In its 2022 gilt market intervention, the Bank of England curtailed the purchase phase of its intervention ex ante by noting upon the launch of the intervention on 28 September that the purchases would be ‘strictly time limited’ and setting out upfront an explicit end date for purchases of 14 October.^{footnote [14]} This provided a 13-day window for LDI firms – or any other market participants – to sell their gilt holdings to the Bank as part of their deleveraging process.

This deadline was intended to underline the temporary nature of the operation and was informed by specific information on the challenge at hand. In this case, the 14 October deadline was informed by market intelligence regarding the scale of rebalancing LDIs were required to undertake, as well as an estimate of how long this would take if executed at speed by the funds in question. The firmness of the deadline was reiterated by the Governor on 11 October, to underscore to LDI managers that no extension would be granted if they failed to complete their deleveraging process by that date. The Bank’s gilt purchases picked up after the Governor’s statement, reflecting increased sales by LDI funds.

Any ex ante deadline will need to be credible. There may be cases where setting a credible ex ante deadline is not judged to be possible and an open-ended approach to purchases may be required. In that situation, a central bank would need robust criteria for assessing when purchases could cease as the stress evolves; an approach to communicating that; and a way to guard against the incentives some market participants may have to keep the central bank in the market for an extended period.

Temporary holding period and timely unwind

At the outset of its intervention, the Bank of England committed to unwind the purchases ‘in a smooth and orderly fashion once risks to market functioning were judged to have subsided’. While it was not possible to commit to a detailed exit timeline upfront, following through on this high-level commitment to timely unwind was particularly important for the credibility of this new type of financial stability intervention and its distinction from monetary policy tools, especially QE.

Limiting the holding period of assets purchased entailed two elements: first, commencing unwind of the purchases as promptly as possible; and second, conducting that unwind as quickly as was consistent with orderly market functioning.

To achieve this timely but orderly unwind, the Bank adopted a demand-led approach. Rather than conducting sales at a set pace with a fixed end-date for exiting the portfolio, the Bank allowed eligible counterparties to express interest in purchasing stock via a form of reverse enquiry window. The Bank’s acceptance of those bids was linked to the pattern of demand received. This allowed the Bank to take advantage of demand for particular stocks where it existed, while limiting the impact of sales on market conditions.

Using this approach, the Bank was able to complete unwind of the portfolio within four working weeks of opening the reverse enquiry window, with the portfolio fully sold by 12 January 2023.^{footnote [15]}

The Bank’s approach to unwinding the portfolio is set out in further detail in Section 5.

4: Targeted purchases

Selection of assets to be purchased

Financial stability asset purchases should be targeted at addressing the underlying vulnerability that is amplifying stress. Consistent with the desire to limit spillovers to monetary policy and avoid excessive disintermediation, the central bank should not purchase more than is necessary, and should not have undue impact on the ability of markets to set prices. Operations should be focused narrowly where they are needed.

In this case, that meant the Bank bought only the assets in core markets most affected by forced selling by LDI and pension funds, namely long-dated conventional gilts and – in the latter phases of the operation – index-linked gilts across the curve.^{footnote [16]} As set out in Table A, this contrasts to QE, where the Bank bought conventional gilts across a wider range of maturities to be broadly market neutral.

The decision on which assets to target was informed by a combination of data and market intelligence from LDI and pension funds, dealers and other gilt market participants, which revealed where selling pressure was most acute and how it might evolve. Transaction data available to the Bank showed which gilts LDI funds and other firms in the sector used to back their repo borrowing, and so which assets they would most likely sell if deleveraging. Chart 3 shows a snapshot before the Bank’s intervention. Real-time market monitoring metrics were also used to understand where liquidity was most impaired.

Chart 3: Gilts used by LDI and pension funds to back repo borrowing, by maturity and type

Footnotes

• Sources: Bank of England sterling money markets data and Bank calculations.

These financial stability operations were targeted at stabilising particular parts of core markets where there was selling pressure. As in the Bank’s other market operations, there is a general principle that this is a market backstop, not a firm-specific one. The Bank’s direct counterparties were Gilt-edged Market Makers (GEMMs), also known as ‘dealers’. But any market participant facing potential fire sales – not just LDI funds – could sell to the Bank via those dealers, either directly on an agency basis or indirectly by selling to the dealer before or after the operations.

Size and frequency

The Bank conducted purchase operations daily, making clear that it would buy as much as necessary to resolve dysfunction and initially offering to buy up to £5 billion in each operation.

This initial maximum size of each auction was calibrated to be more than sufficient given expected selling pressure (using market intelligence and data as an input), to ensure the intervention was credible and quickly restored confidence. Moreover, announcing a potentially large size may have meant ultimately needing to buy less due to the confidence-boosting announcement effect.

The daily maximum operation size was raised to up to £10 billion for the final five days of the intervention based on evolving data and market intelligence that suggested an increase in selling pressure ahead of the deadlines for the Bank’s purchases ending.

5: Backstop pricing of purchases

Backstop pricing is a core tenet of financial stability buy/sell tools, as set out by the BIS Markets Committee Working Group (2022). Its aim is to ensure prices accepted in operations do not duly interfere with price discovery or substitute the need for market participants to manage their own risks over the medium term. Calibrated effectively, backstop pricing would also help ensure the central bank purchases only as many assets as required to resolve the specific financial stability issue and recatalyse the market, while minimising undesirable spillovers to the stance of monetary policy.

The ‘outside spread’

Backstop pricing stems from the standard description of MMLR, which has the central bank standing ready to tackle dysfunction in securities markets via asset purchases, by setting a bid-ask spread that is wider than ‘normal’ market conditions but narrower than in stress. Chart 4 illustrates this in yield terms. As described in Tucker (2009) and Tucker (2014), the bid-ask spread in MMLR ‘should be unattractive relative to peacetime conditions in private markets but better than those available in crisis conditions’. Merhling (2014) describes this as an analogue to the high rate in the Bagehot rule for lender of last resort since ‘the purpose of the wide (or outside) spread is to ensure that the intervention is only supporting, not replacing, the market until it recovers and begins to trade again at a narrower (inside) spread’.

Chart 4: Standard MMLR ‘outside spread’ (in yield terms)

In practice, the Bank implemented this style of backstop pricing by defining a ‘reserve spread’ to market mid yields, reflecting where we were willing to buy. This was suitable for conventional gilts, where a market mid-yield could be identified and trusted, given the market size and active trading in these bonds. However, for index-linked gilts market mid-yields were poorly defined, with a wide dispersion in traded prices. So, for linkers we set a further minimum absolute ‘reference yield’, below which we could not make purchases.

Implementing the ‘outside spread’ via a ‘reserve spread’

For conventional gilts, the reserve spread was fixed for all bonds ahead of each auction and set in yield terms relative to market mid-yields. The reserve spread was not disclosed ahead of each auction, with the intention of promoting competition and ensuring purchases acted as a genuine backstop, with auctions used to the greatest possible extent by genuinely distressed forced sellers.

To implement this, offers received in the auctions were ranked in order of attractiveness to the Bank and accepted, subject to the chosen reserve spread and maximum size of the auction. At the end of each auction, all offers to sell were ranked (in yield terms) as a spread to market mid-yields. Only offers at yields higher than the chosen reserve spread were accepted. Chart 5 illustrates a stylised auction with each oval representing an individual offer, plotted as a spread to the market mid-yield.^{footnote [17]}

Chart 5: Stylised auction demand curve illustrating offers from market participants and the reserve spread

Calibrating the reserve spread

The Bank considered several factors when setting the reserve spread for each operation (Table B). These were tailored to the specific nature of the shock and reviewed by Bank staff on a daily basis.

Given the Bank’s desire to operate in a way that was predictable, where possible Bank staff sought to ensure that changes to the chosen reserve spread were made in a gradual and predictable manner, while keeping the actual reserve spread private.

Evolution of the reserve spread over time

In practice, the Bank – and the wider market – learned iteratively from the first few operations.

Chart 6 shows aggregate offers accepted and rejected for conventional gilts, alongside the weighted average accepted spread, for each purchase auction. This provides a good indication of how far above market yields (below market prices) gilts were sold to the Bank over the period.

Chart 6: Financial stability purchases, daily demand and allocation for conventional gilts ^(a)

Footnotes

• Source: Bank of England.
• (a) The white line shows the weighted average spread to market mid-yield of accepted bids for each auction.

While our chosen reserve spreads were non-public ex ante, information was available from the published results after each auction.footnote [18] Over time we chose to supplement this by publishing the basis point spread to market mid-yields on the marginal unallocated offer. This approach was intended to balance the need to minimise gaming with providing sufficient certainty of execution – both for end-users and GEMMs who may be intermediating flows in the secondary market – and was one example of the Bank adapting its approach, learning from experience.

‘Reference yield’ for purchases of index-linked gilts

As referenced in Section 2, an important principle in the Bank’s purchase operations was that gilt market prices should continue to be set primarily by market forces. But in the event of extreme market dysfunction there may be periods where no reliable market mid-yield is available, as observed in Buiter and Sibert (2007).

This was the position in the index-linked gilt market on Monday 10 October. The structure of that market differs markedly from the conventional gilt market, making it more vulnerable to becoming a one-sided market in stress. It has a much smaller investor base – a large proportion of which is concentrated among LDI funds – and is typically less liquid.

During the first part of the Bank’s operations, the market was able to process a large quantity of linker sales by LDI funds and others relatively effectively. Much of this was absorbed by GEMMs intermediating across markets – ie purchasing linkers and simultaneously selling conventional gilts to the Bank – effectively warehousing the inflation risk themselves. Up until 10 October, around 80% of all gilt sales by LDI funds and pension schemes (£13 billion in total) over the crisis period had been in linkers (Chart 7).

This process quickly hit its limits, however. By 10 October, market intelligence revealed that LDI funds still needed to sell a significant quantity of linkers ahead of the Bank’s deadline. The ability to transact in size in the market had dried up due to a near-complete absence of willing buyers. With dealer capacity exhausted, the market for index-linked bonds had become almost entirely one way, impeding its ability to function correctly.

Chart 7: Cumulative net gilt sales by LDI funds and pension schemes with an open gilt repo or interest rate derivatives position, between 22 September and 21 October 2022, and cumulative gilt purchases by the Bank of England

Footnotes

• Sources: Bloomberg L.P., MiFID and Bank calculations.

In such circumstances, live screen prices can be unreliable and very easily moved by small transactions. This posed real challenges for their use as a robust benchmark in the pricing of large-scale asset purchases.

As the only sizable buyer in the market, the Bank had no option but to accept some form of role in price determination for a temporary period, rather than simply taking live price observations as its sole benchmark.

To do this, the Bank opted to establish a ‘reference’ yield level, which functioned as a lower limit on the yields (or an upper bound on prices) that the Bank would accept in its purchase operations for index-linked bonds. For each bond, this was set at the closing yield published on TradeWeb for Monday 10 October, before the Bank’s linker purchases were announced. Setting the reference yield based on publicly available prices ensured the Bank’s pricing was simple and transparent, and supplemented the existing reserve spread pricing approach (used for conventional gilt purchases).

Chart 8 illustrates how the reference yield worked. The ovals reflect a stylised set of offers from market participants (a demand curve) for linkers. Any offers submitted above the specific reference yield and also within the chosen reserve spread were accepted. Offers below the pre-determined reference yield went unallocated.

Chart 8: Stylised auction demand curve for offers from market participants and interaction with the reference yield

Importantly, this reference yield was not a level the Bank was attempting to defend, as with ‘yield curve control’ policies which have been used by some other central banks for monetary policy purposes.^{footnote [19]} Market prices remained freely floating and could fluctuate both above and below the reference level (see Chart 9). Instead, this level functioned as an automatic safeguard which meant the Bank was providing a backstop at particularly high yields but allowed the Bank to avoid making transactions that could instead take place in the market if buyers emerged at lower yields. And, as described above, this remained a very short-term policy, effective for only four days.^{footnote [20]}

To calibrate the absolute reference yield, the Bank chose a point in time at which the price level was known to reflect stressed conditions, but not so stressed as to permit adverse feedback loops to re-emerge. The Bank’s market intelligence was again critical to this, identifying that by 10 October yields were approaching further such trigger levels but had not yet breached them. The closing prices published on TradeWeb for that day were the best available measure of that level, and would be used that evening across the market.

Overall, this innovative reference yield approach proved both effective and necessary. A further damaging spiral of forced selling in index-linked gilts was averted, reducing financial stability risks. Moreover, the approach was strongly supported by market participants due to the clarity it provided over pricing in a highly uncertain environment. Over the period of purchases, private markets remained thin and volatile, with pricing fluctuating both above and below the reference levels. In a number of instances significantly sized sales were made to the Bank at around its reference yields despite market pricing on screens appearing to suggest much higher prices were available, demonstrating the unreliability of these quoted levels, particularly in large size.

Chart 9: UK 30-year index-linked and conventional yields, with reference yield

Footnotes

• Sources: Bloomberg and Bank calculations.

6: Timely but orderly, demand-led unwind

High-level approach

As discussed in Section 3, timely unwind of purchases made for financial stability reasons is critical to ensuring the intervention is temporary, targeted, and minimises undesirable spillovers to the stance of monetary policy. At the same time, the speed of unwind must ensure that the sales process does not trigger renewed market dysfunction, which would be self-defeating. To achieve these twin goals of ‘timely but orderly’ unwind, the Bank adopted a demand-led approach.

The key features of this were:

• A prompt commencement of the unwind process, once risks to financial stability were judged to have subsided.
• A demand-led pricing approach, with the whole portfolio of gilts purchased in the financial stability operations available for sale in each operation. This meant the Bank’s sales would be guided by market appetite for the gilts in the portfolio, rather than proceeding at a fixed pace.
• To ensure unwind did not trigger renewed market dysfunction, discretion for the Bank not to sell individual gilts that had experienced a significant fall in price ahead of each sales window.
• Communications that emphasised that the Bank would be sensitive to demand conditions through the unwind process and therefore would be comfortable with minimal (or no) sales in periods where demand was low.
• At an operational level, careful scheduling taking into account UK Debt Management Office (DMO) supply events to reduce the risk of (actual or perceived) excessive supply on any one day.

The remainder of this section examines each of these key features in turn.

The decision to commence unwind

The Bank announced its approach to unwinding its financial stability portfolio on 10 November 2022, and commenced the unwind process on 29 November 2022, six and a half weeks after purchases ended.

This decision was guided by qualitative market intelligence, including on the resilience of LDI funds which had by now deleveraged, and supported by a dashboard of quantitative indicators of gilt market functioning. Both suggested that, while gilt market conditions had not returned to the levels prevailing prior to the announcement of the Government’s Autumn Statement, risks of severe market dysfunction that could threaten financial stability had subsided.

The implementation of demand-led unwind – pricing approach

The Bank’s demand-led approach was designed to take advantage of the general re-emergence of demand for gilts among financial market participants while limiting the impact of sales on market conditions.

In practice, this was implemented via a form of ‘reverse enquiry windows’. These were held three times per week over the unwind phase. In these 30-minute lots, eligible counterparties were able to express interest in purchasing any of the index-linked and/or long-dated conventional gilts that had recently been purchased by the Bank in its financial stability intervention. The Bank’s acceptance of submitted bids in those windows was based on its assessment of the pattern and strength of demand seen in those windows.

This was encapsulated by the Bank’s pricing approach in conducting these sales, which entailed two key elements:

• Reserve pricing: First, the Bank would only accept bids at yields lower than prevailing market levels (ie only accepted prices higher than the market mid-price). This meant that the Bank set a reserve spread of zero basis points relative to market mid-yields.
• Yield change ‘safety valve’: Second, to ensure unwind did not trigger renewed market dysfunction, the Bank would not typically sell individual gilts that had experienced a significant fall in price ahead of each window.

Reserve pricing

This element mirrored the reserve spread approach used in the purchase phase. As above, the Bank set a consistent reserve spread of zero basis points relative to market mid-yields^{footnote [21]} and the quantity sold in each reverse enquiry window was not fixed. Instead, the amount sold was endogenous to the strength of market demand for gilts held in the portfolio, as reflected in the volume and pricing of bids received in each window.

This decision was arrived at by balancing two sets of considerations:

• On the one hand, for the unwind approach to be demand-led, it was important not to offer to sell gilts at prices below market mids (ie the Bank would not ‘sell at a discount’ to offload a greater quantity of gilts). Forcibly selling bonds into an improved but still febrile gilt market also risked triggering renewed market dysfunction.
• On the other hand, it was important to encourage potential buyers to come to the Bank’s reverse enquiry windows to support timely unwind. This motivated the Bank’s reserve price being as attractive as possible without crossing market mid-levels.

To aid calibration, the Bank conducted a survey of GEMMs’ cost of trading (bid-ask spreads). Chart 10 illustrates how setting a reserve spread of zero basis points would be broadly comparable to prevailing market levels for small trade sizes, but would be increasingly attractive relative to the market as trade size increased. This made the Bank’s sales operations an attractive opportunity for market participants to buy gilts in large size, since dealers typically quote a wider bid-ask spread for transacting in larger size.

Chart 10: Comparing the cost of buying in the market to the Bank’s reserve spread ^(a)

Footnotes

• Source: Bank of England.
• (a) Stylised diagram informed by a survey of GEMMs.

The second element of the Bank’s pricing approach consisted of a yield change ‘safety valve’. This allowed the Bank to exercise its discretion not to sell gilts at a price ‘significantly lower than the previous trading day’s close’. The Bank did not publicly disclose any further information on the thresholds used in making this assessment, nor whether those thresholds varied across conventional and index-linked gilts, or over time.

This safety valve served two purposes. First, it ensured that the Bank would not sell into a significant market sell-off, mitigating the risk that the Bank’s unwind itself would result in disorderly moves in the gilt market. Second, it helped limit excessive market moves driven by investor positioning (intentional or otherwise) in the run-up to, and during, each window. This was supported by value-for-money considerations (by putting an indicative floor on the price level the Bank would be willing to sell at). This second motivation also drove the Bank’s decision not to publish the level of the thresholds.

The quantitative thresholds that defined a ‘significant’ fall in gilt prices since the previous day were calibrated to balance the trade-off between supporting market functioning, minimising uncertainty around allocation, and delivering timely unwind of the overall portfolio. To aid this calibration, Bank staff back-tested various options for thresholds against historical gilt market yield moves. The Bank left open the possibility of adjusting those thresholds, if needed.

In practice, although these thresholds did not bind regularly, they are likely to have helped limit the impact of the Bank’s sales on gilt yields and broader market conditions.

Consideration of ex ante limits on sales

To support the objective of an orderly unwind, the Bank considered imposing ex ante limits on the quantity of gilts it would be willing to sell at each reverse enquiry window. Ultimately, this option was rejected. Bank staff considered that such limits could choke off genuine demand, thereby unduly undermining the objective of timely unwind. At the same time, Bank staff considered that the yield change safety valve component of the pricing approach, and the flexibility to adjust it as necessary, was sufficient to ensure unwind was orderly. The relatively small total size of the portfolio was a factor in this judgement.

Communication approach

Clear communication to financial market participants was critical to the successful unwind of the portfolio. In particular, it was paramount that financial market participants did not perceive there to be a risk that the Bank would sell more gilts than the market was willing or able to absorb, in a price-insensitive manner.

The lack of ex ante quantity limits, beyond the size of holdings in the portfolio, put more onus on the Bank to successfully communicate this.

The Bank therefore emphasised, in announcing its plans for unwinding the portfolio on 10 November 2022, that its approach would be demand-led and responsive to market conditions. In particular, the Bank emphasised that it would be patient if required, stating that it was happy to sell few or no bonds during a reverse enquiry window if demand was insufficient.

This News Release, which set out the key tenets of ‘timely but orderly’ unwind, was followed by a Market Notice on 23 November that explained in more detail how this would be implemented, including the reserve price and yield change safety valve elements of the Bank’s pricing approach.

Operational considerations

Interactions with DMO auctions/syndications

To ensure that there was no actual or perceived risk of oversupply of gilts on any given day arising from the Bank’s unwind, and to avoid interference with the DMO’s issuance programme, the Bank set out the following criteria for the gilts eligible for sale in each reverse enquiry window. The Bank would not make available for sale:

• Any long-dated conventional gilts held in its financial stability portfolio on days when the DMO would be auctioning long-dated conventional gilts. The same would be true for index-linked gilts.
• Any individual gilt held in its financial stability portfolio that the DMO had announced it would reopen, including via a gilt tender, for one week before or after that date.
• Any long-dated conventional gilts held in its financial stability portfolio on days where the DMO confirms it will undertake a syndication of a long-dated conventional gilt. The same would be true for index-linked gilts. In the event of a syndication, the Bank would reserve the right to cancel a reverse enquiry window altogether.

Bilateral sales of small residual holdings

Bank staff were mindful that, as sales progressed and holdings in the financial stability gilt portfolio decreased, reverse enquiry windows would not be an efficient mechanism for selling small residual holdings of stock (defined as <£5 million nominal).

The Bank therefore provided a bilateral sale process for such holdings, to be used at the Bank’s discretion. This involved offering the small residual holding of that gilt to GEMM counterparties who had successfully bid for it in a reverse enquiry window.^{footnote [22]} In practice, less than 2% (£311 million) of the total £19.3 billion in holdings was unwound in this manner over the final few days of operations.

7: Conclusion and lessons learned

The Bank of England’s temporary gilt purchases in 2022 were a successful application of key principles behind financial stability operations. While the precise design of such operations will necessarily depend on the shock in question, the Bank’s experience demonstrates the value of these high-level principles and offers a number of lessons for future interventions.

By design, the operations were temporary, targeted and utilised backstop pricing. This halted the self-fulfilling feedback loop of LDI fund gilt sales, providing time for the financial stability vulnerability to be addressed, and was coupled with action by the regulatory authorities and macroprudential authorities. Consistent with its temporary nature, the financial stability gilt portfolio was unwound in a timely but orderly way. The intervention was clearly distinguished from asset purchases for monetary policy reasons in design, impact and communication, successfully minimising unintended spillovers to the stance of monetary policy.

There are several key lessons to highlight from this intervention.

First, part of the success of these operations came from the combination of preparation in advance and a flexible ‘learning by doing’ approach. While the exact design of the purchase tool was calibrated in real time for the specific stress, the design drew on advanced planning for the eventuality of a generic gilt market stress where QE would not be a suitable response. Work by Bank staff on a financial stability asset purchase tool had been at an advanced stage by Autumn 2022 and was supported by the May 2022 BIS Markets Committee report. This meant design and implementation did not start from a blank sheet of paper – in particular on the reserve pricing approach and operational preparations – which meant purchases were able to begin very quickly when they were needed.

Second, the decisions to intervene, including the timing, were based on a combination of market intelligence, data and judgement. Market intelligence played a critical role in judging when to commence financial stability gilt purchases, how to set the Bank’s reserve pricing, when to cease them, when to begin unwinding them, and how to price that. Such judgements are inherently forward looking. They require supplementing real-time data with market intelligence to understand the drivers of market moves and their expected evolution (Rosen (2022)). This speaks to the importance of a strong, broad-based and dynamic market intelligence function within central banks, with relationships across all key categories of financial market participants. The existence and effectiveness of these relationships is likely to be a key determinant of the quality of the information set available to central banking in making judgements around the design of its operations in conditions of uncertainty. Wuerffel (2023) draws the same lesson from the Federal Reserve’s 2020 market functioning purchases.

Third, clear communication to financial market participants, at all stages of the intervention, was vital to its success. The Bank made it clear from the beginning that this was a targeted and temporary intervention on financial stability grounds, distinguishing it from previous asset purchase programmes which were conducted for monetary policy reasons. The detailed and transparent explanation of operational arrangements was crucial to ensuing the market understood the Bank’s approach, including the novel demand-led approach to sales.

Fourth, there are also lessons from ‘learning by doing’ around backstop pricing, in particular how to set the reserve spread and reference yields, and how much should be communicated publicly on this. The Bank adjusted its reserve spread, particularly in the first week of purchase operations, in response to revealed demand, market conditions and market intelligence. Some of that adjustment could inform how to approach setting the reserve spread in any future financial stability buy/sell tools. While the reserve spread was not revealed publicly, some information on it could be inferred from operational results. And partway through the purchase operations, the Bank did add some information on the pricing of the marginal unsuccessful offer, giving further information about reserve pricing. There remains an open question on how much transparency on reserve pricing is desirable.

Fifth, the reference yield approach used for index-linked gilt purchases proved an effective innovation in the context of extreme illiquidity, providing a temporary anchor when it was hard for the market to find a price and protecting the Bank from adverse pricing outcomes. There are open questions for future research on how to calibrate this and whether it should vary over the purchasing period, especially in the case of a prolonged intervention period, which would be significantly more challenging.

Finally, the ‘demand-led’ sales approach deployed by the Bank did mean its temporary gilt portfolio was unwound in a timely but orderly way. This was a novel approach that resulted in a successful disposal of the portfolio without rekindling market dysfunction. Further work could be undertaken to explore the applicability of such an approach in different situations.

Taken as a whole, the Bank’s intervention stands as an important case study in this field. While the Bank achieved its objectives in this situation, there are important lessons to be learned and areas for further research to inform future policy design.

NBFIs have grown in importance in recent decades and have introduced new sources of systemic risk. As set out in Bank of England (2023b), the FPC continues to further improve risk identification in, and the functioning and resilience of, market-based finance. Alongside that, the Bank is continuing to develop its toolkit, recently outlining ambitious plans for tackling systemic risks in market-based finance by developing a new lending tool for NBFIs, starting with UK insurance companies and pension funds.