QE at the Bank of England: a perspective on its functioning and effectiveness

Quarterly Bulletin 2022 Q1
Published on 18 May 2022

By Filippo Busetto, Matthieu Chavaz, Maren Froemel, Michael Joyce, Iryna Kaminska (Monetary Analysis Directorate) and Jack Worlidge (Markets Directorate).

Foreword

In the face of the global financial crisis in 2009, the Bank of England embarked on a programme of large-scale asset purchases. This reflected a need to further ease monetary conditions with Bank Rate at its effective lower bound. The Bank’s Monetary Policy Committee (MPC) decided to purchase mostly UK long-term government bonds, with those purchases financed by the issuance of newly created, interest-paying reserves. These asset purchases were labelled quantitative easing (QE).

In 2021 the Bank of England’s Independent Evaluation Office (IEO) published a report on the Bank’s approach to QE. One of the report’s recommendations was that the Bank should periodically update technical audiences about its latest thinking on QE.

This article is part of the Bank’s response to that recommendation. It has been authored by Bank staff at the request of the MPC, and summarises a large body of research and analysis on QE produced both inside and outside the Bank. The article therefore gives insight into the body of technical work that has underpinned and framed the MPC’s discussion of QE in recent years. It should be read against the background of the academic articles, staff working papers and reports published by Bank staff on this topic, some of which are cited in the article itself.

That said, the article does not seek to give a single MPC view about how QE works. Individual MPC members, in line with the wider academic and economic community, naturally have their own views about which transmission channels of QE are likely to prove most important, about how that assessment may vary across different economic circumstances, and about the effectiveness of QE in easing financial conditions and stimulating the economy. These individual opinions are recorded in MPC meeting minutes and reported in speeches and testimony to Parliament. It is one of the strengths of the UK monetary policy framework that MPC members are individually accountable for their decisions and votes. The resulting expression of differing views and encouragement of debate among them helps to improve the robustness of the final policy decision.

This article also addresses another recommendation in the IEO report, namely that the Bank should raise awareness of the cash transfer arrangements between the Bank’s Asset Purchase Facility (APF) and HM Treasury (HMT). The APF is a financial vehicle that holds the portfolio of bonds accumulated as a result of the MPC’s asset purchases. It is indemnified by HMT. This article provides an updated analysis of the cash flow transfers between HMT and the APF arising from that arrangement.

The stock of assets held in the APF peaked at £895 billion towards the end of 2021, following the implementation of further QE in response to the Covid-19 crisis and its aftermath. But, at its February 2022 meeting, the MPC voted to begin unwinding the APF portfolio by ceasing to reinvest maturing bonds, as well as conducting a programme of corporate bond sales (to be completed no earlier than towards the end of 2023). Furthermore, in May 2022 the MPC asked Bank staff to work on a strategy for UK government bond sales, which would allow the Committee to make a decision at a subsequent meeting on whether to commence such sales.

Just as it is important to flag what this article is about, it is also important to highlight what it is not about. With QE unwind now underway, it is important to clarify that this article does not evaluate the economic and market impact of such quantitative tightening (QT), nor is it intended to offer any signal about the prospects for QT in the UK.

The MPC set out its initial thinking on QT in the August 2021 Monetary Policy Report. It noted there that “the impact on monetary conditions of a reduction in the stock of purchased assets … is likely to be smaller than that of asset purchases on average over the past” provided that such a reduction takes place “in a gradual and predictable manner and when markets are functioning normally”. On this basis, the impact of QE in the past cannot be mechanically applied to anticipate the impact of QT in the future, simply with the sign reversed. The research results presented in this article certainly should not be read in that way.

Now that the gradual reduction of the APF portfolio has started, it is a good time to reflect on the lessons to be drawn from the past decade or so’s experience with QE in the UK. These lessons should help frame any discussion of its potential future use.

At the outset, it is important to recognise the difficulties faced in extracting strong conclusions from the available experience. In the UK, there have been only five episodes of QE since 2009. The evidence base for QE is therefore narrow, certainly narrower than that available for assessing conventional interest rate policy (a ‘small sample’ problem). Moreover, QE has been implemented in the face of economic and market disturbances. It is difficult to distinguish the impact of the QE implemented in response to those disturbances from the impact of the disturbances themselves (an ‘identification’ problem). Researchers at the Bank and elsewhere have made valiant efforts to address these challenges, reflected in the studies summarised in this article. But drawing robust and universal conclusions is difficult. When interpreting the results for policy purposes, an appropriate degree of circumspection is required.

Nevertheless, we can draw out some lessons. From experience in the UK and elsewhere, we have learnt that QE mainly works through lowering interest rates at various maturities, thereby lowering borrowing costs. Indeed, it is by lowering yields at longer maturities that QE can ease monetary conditions further once Bank Rate has reached its effective lower bound.

In lowering interest rates, QE operates via a number of channels, which are explained in the article. These include: a portfolio balance channel, a signalling channel, a liquidity channel, a market functioning channel, and an uncertainty channel. The available evidence suggests that both the absolute and the relative importance of each of these channels has varied through time, as economic circumstances and market conditions have evolved. The impact of QE on interest rates and yields therefore depends on the context in which it is implemented (‘state contingency’).

If the effect of QE on yields is uncertain and time-varying, then the impact of QE on the real economy and ultimately on CPI inflation (which the MPC targets) is, unsurprisingly, also both uncertain and likely to vary over time. This notwithstanding, on balance the evidence surveyed in this article suggests that, in the face of the multiple large shocks to the UK economy since the onset of the global financial crisis, QE has helped the MPC meet its 2% inflation target while also sustaining growth and employment in line with the MPC’s remit.

The Bank seeks to encourage an active and informed debate about its monetary policy framework and choices. Publishing and discussing the analytical work undertaken to support those choices is central to this ambition. By summarising the research that has informed decisions about QE, this article seeks to contribute to that debate.

Huw Pill
Chief Economist
Executive Director for Monetary Analysis and Research

Executive summary

Over the course of the past two decades, and especially since the onset of the global financial crisis (GFC), quantitative easing (QE) has become an important instrument of monetary policy for central banks, including the Bank of England.footnote [1] This article reviews the available evidence on the impact of QE and draws out lessons from that experience for the design and conduct of monetary policy.

The article was written in response to a recommendation of the Bank’s Independent Evaluation Office (IEO) contained in its report on QE that was published last year.footnote [2] The report recommended that the Bank should periodically update technical audiences on its latest collective thinking on QE. The article also addresses their recommendation that the Bank should raise awareness of the cash transfer arrangements between the Asset Purchase Facility (APF)footnote [3] and HM Treasury (HMT), by providing an updated analysis of the cash flow transfers between HMT and the APF.

QE is an additional tool employed by the MPC in pursuit of its price stability remit.

QE was one of several novel policy instruments deployed in the aftermath of the GFC. It has helped the Bank to respond flexibly and efficiently to a number of large economic shocks: not only the GFC itself, but also the euro area sovereign debt crisis, the initial reaction to the EU referendum vote, and the onset of Coronavirus pandemic.

QE involves the central bank purchasing financial assets – in the UK case, mostly longer-term government bonds (‘gilts’) – in exchange for newly created interest-paying reserves. QE is intended to ease financial conditions, boost aggregate demand and thereby prevent inflation from being below the 2% target on a sustained basis. Similar to a cut in Bank Rate, QE seeks to lower interest rates, thereby stimulating demand by reducing borrowing costs for households and corporates, supporting asset prices, and putting downward pressure on the exchange rate.

This article summarises research evidence showing that the announcement of asset purchases under the Monetary Policy Committee’s QE programmes has indeed lowered interest rates at various maturities, consistent with the summary view of QE’s transmission mechanism.

Both the magnitude of this effect and the strength of the evidence is greatest for the first QE programme, announced in 2009 during the most intense phase of the GFC. Nonetheless, on balance, the available evidence suggests QE announcements also helped to lower longer-term interest rates, albeit to a lesser extent, during subsequent rounds, even when market stress and dysfunction were not as pronounced.

QE lowers interest rates via several channels that are complementary and vary in intensity over time.

In the first step of its transmission mechanism, QE influences interest rates and yields through a number of channels: (1) reducing expected future policy rates (‘signalling’); (2) increasing demand for longer maturity or riskier assets (‘portfolio balance’); (3) lowering liquidity premia on eligible assets (‘liquidity’); (4) improving ‘market functioning’ more broadly; and (5) reducing policy and economic uncertainty (‘uncertainty’).

These channels are not mutually exclusive, and the intensity of each of them is likely to vary over time as circumstances change. This ‘state-contingency’ means that the observed impact of asset purchases in one set of circumstances cannot easily be translated into a view about the potential impact of asset purchases - or sales - in other circumstances. At the very least, a cautious reading of the evidence is required when drawing policy conclusions.

A growing number of studies, including some conducted by Bank staff, explore the contribution of these various channels to the observed overall impact of QE in different contexts. While there are few episodes to draw on and it is difficult to generalise from any individual study, overall the evidence reported in this article suggests that the signalling, portfolio balance and uncertainty channels have operated in concert, albeit with varying potency – both in absolute terms, and relative to one another – across the various QE programmes implemented since 2009.

On balance, the evidence suggests that QE supports the objective of monetary policy, namely price stability.

While, on balance, the evidence suggests QE has influenced the yield curve and resulted in lower borrowing costs, quantifying its impact on GDP and inflation is much more challenging and the evidence base is smaller. However, the available literature suggests that QE programmes provided positive stimulus to activity and supported inflation in the face of disinflationary pressure.

In March 2020, QE was one of several policies that sought to improve market functioning and mitigate an unwarranted tightening of financial conditions.

In March 2020, there was an extreme spike in the demand for liquidity in financial markets and the real economy, at a time where dealers’ intermediation capacity was constrained. This was associated with an abrupt deterioration in market functioning and a surge in gilt yields, which could have tightened financial conditions and exacerbated the adverse economic impact of Covid-19. Along with other monetary authorities, the MPC responded with a QE programme of unprecedented scale and pace.

The evidence reported in this article suggests this QE programme had a substantial impact on interest rates, with gilt yields retracing a significant portion of their surge shortly after the 19 March 2020 asset purchase announcement. The evidence also suggests gilt market liquidity improved after the Bank’s purchases began, suggesting there was a greater role for the market functioning channel compared to other programmes. However, other schemes introduced by the Bank at that time to improve market functioning and actions by other central banks are also likely to have contributed.

Beyond gilt purchases undertaken within QE programmes, purchases of UK corporate bonds can be effective in lowering corporate borrowing costs and stimulating corporate debt issuance.

As part of its efforts to stimulate demand following the 2016 EU referendum and the 2020 Covid-19 pandemic, the MPC also purchased bonds issued by corporates that make a significant contribution to the UK economy. The evidence suggests that the 2016 Corporate Bond Purchase Scheme succeeded in lowering interest rates on UK corporate bonds and stimulating sterling corporate bond issuance. Corporate bond yields also stopped increasing and eventually declined after the announcement of corporate bond purchases in 2020.

1: Introduction

Quantitative Easing (QE) involves large-scale asset purchases by a central bank, financed by the creation of central bank reserves.footnote [4] These purchases seek to boost aggregate demand and hence inflation. Since the global financial crisis (GFC), QE has become part of the Bank of England’s monetary policy toolkit. As set out in the response to the 2021 Independent Evaluation Office (IEO) report on QE, the Bank is committed to ensuring it uses QE as effectively as possible, in order to achieve its inflation target. To do so, the Bank contributes to, and draws on, economic research into the impact of QE in the United Kingdom (UK) and overseas.

The objective of this article is to review evidence on the impact of QE in the United Kingdom, and draw out implications for monetary policy.

While little evidence was available to guide the first QE programmes introduced in 2009, over time a growing number of studies have analysed the impact of QE, with a particular focus on advancing understanding of the channels through which QE transmits to financial markets and the real economy. Bank staff have contributed to this literature, with earlier reviews of the UK evidence provided by Haldane et al. (2016) and Bailey et al. (2020). In addition, the 2020 asset purchase programmes, which were unprecedented in terms of their scale and pace, provide further insights into how QE can work in situations of market dysfunction and uncertainty. This article, written in response to the recommendation of the Bank’s Independent Evaluation Office that the Bank should periodically update the technical audience on its latest collective thinking on QE, takes stock of the key insights from this material and discusses how it advances technical understanding of QE, drawing out the implications for monetary policy. It also addresses the IEO’s recommendation that the Bank should raise awareness of the cash transfer arrangements between the Asset Purchase Facility (APF) and HM Treasury (HMT), by providing an updated analysis of the cash flow transfers between HMT and the APF.

2: Quantitative easing so far

In the aftermath of the GFC, QE was one of several policy tools that allowed the Bank of England and other central banks to respond flexibly to shocks to the inflation outlook.

In response to the GFC, the MPC cut Bank Rate to 0.5% in March 2009 in order to stimulate activity and inflation. However, the MPC judged that further monetary stimulus was needed in order to meet its inflation target, but believed that the effectiveness of any additional Bank Rate cut could be ‘significantly impaired’. Instead of using its hitherto ‘conventional’ Bank Rate tool, the MPC then announced that it would pursue its remit by purchasing assets in secondary markets in exchange for interest-bearing central bank reserves.footnote [5]

For more than a decade since, QE played a major role in setting monetary policy, with the MPC reviewing and communicating its target for the stock of purchased assets at each meeting. Over time, purchases in response to shocks such as the Eurozone sovereign debt crisis, the referendum vote to leave the EU, and Covid-19 caused the target stock of purchased assets to rise to £895bn (Chart 1). The large majority of purchases have been of conventional UK government bonds (‘gilts’), although some sterling-denominated corporate bonds were purchased during the first QE programmefootnote [6] and further purchases were made with the start of the Corporate Bond Purchase Scheme (CBPS) in 2016. The MPC has at times used QE alongside Bank Rate cuts, forward guidance and schemes to improve the monetary transmission mechanism, i.e. the Term Funding Scheme (TFS) and Term Funding Scheme with additional incentives for SMEs (TFSME), reflecting complementarities between these tools.footnote [7]

Chart 1: MPC asset purchase target (£bn, left scale) and Bank Rate (%, right scale)

Footnotes

  • Source: Bank of England calculation.

3: Quantitative easing in theory

The main objective of QE is to boost aggregate demand in the economy and hence support inflation, thereby helping the MPC to meet its inflation target.

QE typically involves purchasing longer-dated gilts from financial market participants in the secondary market in return for newly created central bank reserves. At a high level, similarly to a Bank Rate cut, these asset purchases should lower longer-term gilt yields and other interest rates, thereby reducing borrowing costs for households and corporates and stimulating spending and investment.

Research has identified a number of channels through which QE can affect economic and financial conditions.

The primary transmission channels are highlighted in orange on Chart 2 and discussed in detail in Box A. Most of these channels initially affect gilt yields, which can be decomposed into two components: the expected policy rate over the bond’s term to maturity, and a term premium that captures compensation for the investor from interest rate and illiquidity risk, and also potentially from the impact from demand/supply imbalances. QE can lower the expectations component by conveying information about the future path for the policy rate (signalling channel). QE can also lower term premia via several channels. By shrinking the supply of longer-term bonds, QE can push some investors to rebalance their portfolios towards other bonds and thereby lower their yields, as well as increasing the price of other assets (portfolio balance channel). QE can also reduce any liquidity premia on purchased assets (liquidity channel). In stressed circumstances, QE can improve market liquidity more broadly by lowering risks to dealers who are intermediating trades (market functioning channel). In addition, QE can mitigate uncertainty in financial markets and the real economy (uncertainty channel), which may also reduce term premia, as well as having wider effects. And under some conditions QE can stimulate credit supply by increasing banks’ holdings of reserves (bank lending channel).

Chart 2: Stylised transmission mechanism

Whereas three of these channels are also part of the transmission mechanism of changes in Bank Rate, the liquidity, market functioning, and portfolio balance channels are specific to QE. Given the mix of channels, and to the extent that term premia tend to increase in importance with bond maturity, QE should have stronger effects on longer-term rates compared to changes in Bank Rate. In line with this, 10-year gilt yields have become more responsive to MPC announcements since the introduction of QE in 2009 (Chart 3).

Chart 3: Change in short-term interest rates (blue) and 10-year gilt yields (orange) following MPC announcements (a) (b)

Footnotes

  • Sources: Refinitiv Tick History; Kaminska and Mumtaz (2022).
  • (a) In addition to asset purchases, changes in 10-year yields may also reflect other tools used to set monetary policy during the period, such as forward guidance, although in theory other tools should have less impact on long yields, see Greenwood et al. (2016).
  • (b) Difference in yields within a window spanning up to 50 minutes around the MPC announcements. The short rate is a mix of the first four short sterling future contracts.

Box A: The transmission channels of QE

This box summarises the main transmission channels of QE and the theoretical conditions that underpin them. One common thread is that the extent to which these conditions are met is likely to depend on the state of the economy and financial markets. Therefore the mix of channels, and the total impact of QE, may vary across time according to circumstances, implying the different QE programmes may have different effects on asset prices and the real economy.footnote [8]

1: Signalling channel

When the central bank engages in QE, markets may infer that the central bank is credibly committing to a period of protracted low policy rates. One reason for this belief could be related to the expectation by market participants that the central bank will not raise policy rates before the end of the QE programme.footnote [9] This would lower longer-term interest rates by lowering the expectation component of government bond yields.

2: Portfolio balance channel

In aggregate, QE reduces the ‘free-float’ of bonds available to financial market participants. This may increase the price of bonds and lower hence their yields, via two main effects. First, reducing the supply of certain bonds can lower the yield on these bonds and on other assets that investors might consider close substitutes (‘local supply’ channel). Second, by shifting investors’ portfolios towards shorter-dated assets, QE reduces the overall amount of duration risk borne by market participants and may therefore lower term premia across the yield curve (‘duration channel’).

According to some theories, one necessary condition for changes in the supply of bonds to affect yields is that some investors do not view the bank deposits they receive in exchange for bonds as a perfect substitute.footnote [10] This means they would look to rebalance their portfolio towards other assets, pushing up their price. This can be the case, for instance, if some investors have preferences for bonds with a specific maturity. An additional requirement is that other investor groups should not be able to undo this effect by exploiting arbitrage opportunities across bonds; for instance, because they are risk averse or subject to financial constraints.footnote [11]

3: Liquidity channel

Bond yields can also incorporate a liquidity premium to compensate investors for the risk that they might not be able to sell the bond immediately, particularly in times of stress. In this case, purchases of assets by the central bank can improve the liquidity of assets purchased or eligible for purchases by reducing this risk and stimulating trading.

4: Market functioning channel

In exceptionally stressed circumstances, when dealers’ capacity to intermediate trades is limited, large-scale asset purchases can improve wider market liquidity and mitigate the risk of a broader tightening in financial conditions that might disrupt the monetary transmission mechanism. The strength of this channel therefore depends on the degree of market dysfunction and the amount of gilts held by dealers.

5: Uncertainty channel

QE might also help to lower yields by reducing uncertainty about the economic outlook and the path for the policy rate. In particular QE could be seen to lower the probability of the most negative economic outcomes when Bank Rate is close to zero, relative to a situation in which Bank Rate is the only available policy instrument. This channel might be particularly strong when the central bank engages in QE for the first time.

6: Bank lending channel

When selling gilts to the Bank of England, a market participant will see an increase in its deposit account with a commercial bank.footnote [12] At the same time, the commercial bank’s reserves with the central bank will increase by the same amount. Under some conditions this can increase the bank’s ability and willingness to lend, particularly if greater holdings of liquid assets relax the bank’s financial constraints. But whether this is the case in practice depends on regulatory requirements and the bank’s financial position and risk appetite. It will also depend on how stable this new source of funding is.

4: The impact of QE on the yield curve

QE is transmitted to the economy by reducing interest rates across the yield curve.

Since financial market participants are forward-looking, the expected impact of asset purchases should be reflected in asset prices as soon as the central bank announces a change in its purchase target.footnote [13] In line with this idea, a large number of studies have estimated the impact of QE by exploring how asset prices change within a short time interval (window) after each QE announcement, with the length of the window chosen in order to help exclude the influence of other news. Among others, the evidence on the yield impact of QE announcements has been reviewed by Gagnon (2016), Haldane et al. (2016), and Hartley and Rebucci (2020).

There is consistent evidence that MPC QE announcements have lowered interest rates across the maturity spectrum.

On balance, studies focused on the UK find that gilt yields have fallen after QE announcements. In particular, medium to long-term gilt yields fell by about 100bps in response to the first £200bn QE programme (Joyce et al. (2011), Christensen and Rudebusch (2012)).footnote [14] Studies for the US and euro area have found broadly comparable effects when adjusting for differences in programme size.footnote [15]

These studies focus on the first QE rounds, but evidence across all the major QE episodes to date also suggests that gilt yields fell across the maturity spectrum after the MPC announced an increase in the target stock of purchased assets (Chart 4).footnote [16]

Chart 4: Change in gilt yields by maturity after Bank purchase announcements: one-day (blue) and two-day window (orange)

Left: QE1 March 2009 (£75bn). Right: QE1 August 2009 (£50bn).

Left: QE2 October 2011 (£75bn). Right: QE3 July 2012 (£50bn).

Left: QE4 August 2016 (£70bn). Right: QE5 March 2020 (£200bn).

Footnotes

The evidence points to significant variation across rounds, with the largest moves being associated with the first QE programme.

Overall the economic research suggests that the impact of QE has changed materially over time. The largest impact was associated with the first programme in March 2009, with yields falling by relatively less in response to subsequent MPC announcements (Chart 4). Variation between the first and subsequent programmes was also observed in the US, leading to a debate about how effective QE was outside of crises.footnote [17] One argument is that in times of stress, the conditions necessary for QE to lower yields via liquidity, market functioning, portfolio balance, and uncertainty effects are likely to be met to a larger degree (Box A). Applying this argument to the UK, Vlieghe (2021) argues that QE has had little impact outside situations of market stress like March 2009 and March 2020.

Another explanation is that subsequent QE rounds were anticipated by the market to a greater extent (Gagnon (2018); Bernanke (2020). In line with this, survey data for the UK suggest that a majority of MPC QE announcements were expected to a significant degree by financial market participants. In some cases QE announcements were also smaller than expected. However, the fact that QE programmes were anticipated does not mean they were less effective. Instead, it implies that the expected impact of future programmes may have already been incorporated into asset prices ahead of QE announcements, and that yields would therefore be less responsive to these announcements themselves.

The evidence suggests that QE announcements can also lower yields when markets are not stressed.

The impact of QE appears to vary across programmes, even if one allows for what had been expected beforehand. The response to the surprise component of QE – the announced target less what markets had anticipated – was much greater in March 2009 than at other times (Chart 5). But there was still a measurable impact of QE surprises on other occasions, even when markets were functioning well (Chart 6).

Chart 5: Change in 10-year gilt yields after QE announcement and gilt purchase surprise (a)

Footnotes

  • Sources: Bloomberg Finance L.P, Refinitiv Eikon, Reuters and Bank of England calculations.
  • (a) February and March 2009 are shown in orange, while March 2020 is shown in purple. The chart shows two-day windows around announcements except when there are confounding events within this window. The changes over two days after the announcement have been suggested as a plausible time frame for markets to absorb news in the context of QE1 (Joyce et al (2011). October 2011, February 2012, and March 2020 use a narrower window, as MPC announcements coincided with other central bank announcements or major political news. The picture is similar if one uses an average across gilt maturities. Purchase surprise is the change in the expected target stock of QE purchases among market participants surveyed by Reuters (QE1–QE4) and market intelligence (QE5). The purchase surprise for QE1 is the difference in the terminal expectation for asset purchases between April and February 2009, while the surprise in QE5 is estimated using market intelligence obtained shortly before the March 2020 MPC meeting.

Chart 6: Empirical relationship between 10-year gilt yield changes and gilt purchase surprise (a)

Footnotes

  • Sources: Bloomberg Finance L.P, Refinitiv Eikon, Reuters and Bank of England calculations.
  • (a) The average fall in yields is derived by fitting a linear trend line to three different samples from the points displayed in Chart 5 (eg, the fitted line included in Chart 5 is for all announcements). R2 is the fraction of the variation across announcements that is explained by the linear trend.

One critique of this body of research is that the identified impact of QE announcements on gilt yields might be a short-term effect that would reverse before it can have any impact on real outcomes. Bond yields will fluctuate with a range of domestic and global factors, and this can offset any impact from a QE announcement – particularly in an open economy like the UK. But insofar as these factors would exist absent QE, any reversal in yields does not necessarily mean that the effect of QE is transient.footnote [18] In addition, while the full effects would be expected to manifest in gilt yields over a few days, there is evidence that the spillovers to other assets may take place over several weeks.footnote [19] However if QE succeeds in strengthening the economy, there will ultimately be upward pressure on yields. Available studies point to announcement impacts that persist as long as several months and up to two years.footnote [20] Overall, QE has created substantial policy ‘space’ by increasing the MPC’s ability to lower longer-term interest rates.footnote [21]

5: Evidence on the transmission channels of QE

This section discusses the main channels that could explain the reaction of yields to the announcement of QE programmes before 2020. As discussed above, the impact on the yield curve could reflect a combination of signalling, portfolio balance, liquidity, and uncertainty effects. Separating these channels is challenging given the relatively small number of QE programmes that have been undertaken. With the passage of time, the body of evidence on the contribution of different channels is growing. Although generalising from individual studies is difficult, looking across a range of studies suggests a number of common themes.

The evidence suggests that different channels have operated together, but have varied in strength over time.

One key distinction is between channels that work by lowering term premia – including portfolio balance and uncertainty – and those operating by lowering expectations of future policy rates (signalling). Decomposing these effects is difficult because these components are unobserved and must therefore be inferred from models.footnote [22]

Focusing on the UK, Kaminska and Mumtaz (2022) decompose high-frequency changes in the yield curve using a dynamic term structure model over a sample period which includes the first four UK QE programmes. Their decomposition suggests that QE announcements lowered interest rates via both term premia and expectations, often simultaneously. When monetary policy was implemented using QE, announcements had a stronger impact on term premia, compared to cases in which monetary policy was implemented using changes in Bank Rate.footnote [23]

Signalling has been found to have played a significant role in some programmes.

The decomposition by Kaminska and Mumtaz (2022) also suggests that signalling has been one of the key QE channels, although its importance was relatively less pronounced during the first QE programmes. Similarly, other papers have found that initially the signalling channel played a statistically significant but quantitatively modest role. Using different methods to decompose short-term changes in gilt yields, Christensen and Rudebusch (2012) and Joyce et al. (2011) both find that expectations of future Bank Rate fell in response to the announcement of the first Bank QE programme – in line with a signalling channel. But according to their decompositions, signalling effects across the entire programme explain at most up to a third of the announcement impact of QE on gilt yields. By comparison, some studies (see e.g. Christensen and Rudebusch (2012), Bauer and Rudebusch (2014)) find that signalling explains a larger share of the impact of the announcement of the Fed’s first QE programme - perhaps because the Fed also gave explicit guidance about the future path for policy rates.

The finding that signalling explains only a portion of the impact of QE is consistent with a substantial role for portfolio balance in helping to lower yields by compressing term premia.

There is empirical support for ‘preferred habitat’ behaviour.

As discussed in Box A, some theories suggest that for QE to lower bond yields by shrinking the supply of longer-term bonds (‘portfolio balance channel’) some investors must have a preference for bonds of specific maturities. Insurance companies and pension funds (ICPFs) are often thought to have such ‘preferred habitats’ because they tend to hold long-term assets in order to match their liabilities.

In recent research, Giese et al. (2021) show that preferred habitat behaviour in the gilt market exists across the term structure and that foreign central banks (at shorter maturities) and ICPFs (at longer maturities) make up some of these preferred habitat investor groups (Chart 7). These investors’ gilt holdings are less sensitive to changes in the price of gilts, consistent with the idea that they value these assets for non-pecuniary reasons. In addition, one of these groups of investors – including insurance companies – reduced their gilt holdings more than proportionately during the 2016 QE4 programme, in line with these investors playing an important role in the transmission of QE. Similarly, Joyce et al. (2017) find that ICPFs played a significant role in selling gilts to the Bank as part of the QE1 and QE2 programmes. Worlidge (2022) finds a similar result for QE5, in that ICPFs reduced their net acquisition of gilts and increased their net investment in corporate bonds.footnote [24]

Chart 7: Sectoral mapping of gilt investor groups (a)

Footnotes

  • Sources: Euroclear and Bank calculations; Giese et al. (2021).
  • (a) X-axis: weighted average duration of investor’s conventional gilt holdings. Y-axis: 90th–10th percentile range of investor portfolio’s weighted average duration. Bubble sizes reflect the portfolio size.

Evidence suggests that local supply effects (part of the portfolio balance channel) were important in the first QE programmes.

One way to shed light on the role of portfolio balance is to study how the impact of QE announcements varies across individual gilts. If QE lowers gilt yields by shrinking the supply of gilts, those gilts more likely to be purchased by the central bank should see larger drops in yields after announcements. Exploiting changes in the purchase ranges used by the Bank to buy gilts across different maturity buckets, McLaren et al. (2014) find that this was the case for the first two QE programmes. From their estimates, they calculate that this ‘local supply’ effect explains around half of the total impact of the QE1 and QE2 announcements on medium and long-term gilt yields.footnote [25] Comparable estimates suggest that local supply explains around two-thirds of the response of US Treasury yields to the Fed’s first QE programme.footnote [26] McLaren et al. (2014) also find that longer-term bonds react more strongly to QE announcements, in line with a duration effect (Box A).

More recent evidence suggests that the importance of portfolio balance may have diminished over time.

Recent research by Froemel et al. (2022) sheds light on the importance of the local supply channel across all Bank programmes to date. They find that QE announcements lead to larger falls in the yields of gilts that are more likely to be purchased by the Bank, or that could be seen as substitutes, consistent with the operation of this channel. Over time, however, the yield reductions associated with these two effects seem to have become smaller, and to account for a smaller fraction of the overall yield response to announcements (Chart 8).footnote [27]

Chart 8: Average impact on yields explained by relative scarcity, 2-day window around announcements

Footnotes

QE has also been found to reduce uncertainty.

QE can also lower term premia by reducing uncertainty about the economy and the future path of the policy rate, in particular when the policy rate is close to its effective lower bound. In line with this, the decomposition by Kaminska and Mumtaz (2022) suggests that term premia associated with uncertainty about the future path for Bank Rate fell after QE was introduced in 2009. Feeding this decomposition into a macroeconometric model, they estimate that lower uncertainty translates into lower corporate bond spreads and improved economic outcomes. Using a vector autoregression approach, Weale and Wieladek (2016) also find that proxies for financial market and household uncertainty fell in response to earlier MPC QE announcements.

QE might have particularly helped to lower uncertainty when used for the first time at the peak of the GFC. In this situation, QE might have reduced the perceived risk of tail outcomes. Conceptually, this channel could affect both the quantity and pricing of risk in a persistent way.

There is limited evidence that QE purchases lowered yields by reducing liquidity premia in the gilt market.

While asset purchases can improve the liquidity of purchased gilts around auction times, these benefits should be short-lived and limited to periods of poor gilt market liquidity.footnote [28] For example, for the UK, Joyce and Tong (2012) find that during the early stages of the QE1 auctions yields fell most for less liquid gilts, but these effects were small and temporary. But beyond a certain point, by reducing the free float of bonds, central bank purchases may also potentially reduce bond liquidity – although the Bank’s purchase caps and gilt lending scheme are intended to limit this effect (see Box B). The international evidence suggests that asset purchases have a range of impacts on bond liquidity. In Sweden, for example, Blix Grimaldi et al. (2021) found that the Riksbank’s purchases of government bonds improved liquidity as long as the Riksbank held less than 40% of these bonds, but impaired liquidity past that level. These findings are unlikely to be representative of the impact of asset purchases when there is a stress in government bonds and pressure on dealers’ ability to intermediate, as was the case in March 2020. As discussed in Section 7, in these circumstances there is more scope for asset purchases to improve market functioning and liquidity. The potential for purchases of UK corporate bonds to improve liquidity is also greater because they are less liquid than gilts (Section 8).

There is no evidence of a material bank lending channel in the UK.

By lowering gilt rates, QE will tend to lower rates on bank loans to households and corporates as well as on banks’ wholesale funding. All else equal, this should translate into more bank lending. However there is less direct evidence that QE operates through a more specific bank-lending channel.

As discussed in Box A, the influx of reserves and deposits associated with QE could allow commercial banks to increase their holdings of riskier or less liquid assets – including loans to households and businesses – or to grant cheaper loans. But this might not be the case if banks are not constrained by regulations or other constraints in the first place, and if the deposits are ‘flighty’ – possibly because depositors subsequently rebalance their portfolios towards risker securities.footnote [29] In line with this, the evidence suggests that UK banks experiencing a larger influx of reserves as a result of QE did not increase lending relative to other banks during QE1 and QE2.footnote [30]

6: The impact of QE on other asset prices and the real economy

As discussed in the previous sections, lower interest rates across the yield curve should stimulate investment and consumption and therefore support demand in the economy by easing financial conditions and increasing wealth.

Evidence suggests that the impact of QE on the yield curve spills over into other financial market prices.

For a fall in interest rates across the yield curve to stimulate demand, it should lower borrowing costs for households and corporates in a sufficiently persistent way. These costs reflect default-free rates plus a credit risk premium. A fall in the yield curve from QE will lower the first component (by lowering expected risk-free rates via the signalling channel and by increasing the demand for corporate bonds via portfolio balance effects – although this effect might materialise more gradually). QE can also help to lower the second component by improving the outlook for the economy and therefore decreasing default risk for corporates.

In line with this, event studies for the UK suggest that corporate bond yields fell significantly in response to the announcements of the QE programmes (Joyce et al. (2011); Joyce, Tong and Woods (2011); McLaren et al. (2014); Haldane et al. (2016); D’Amico and Kaminska (2019). High-yield corporate bonds seem to react particularly strongly, suggesting that QE lowered downside tail risks in the corporate sector.footnote [31] In addition, D’Amico and Kaminska (2019) find that after QE announcements, there was a larger drop in the yield on investment-grade corporate bonds with maturities similar to the purchased gilts. This impact peaks after a few months, consistent with portfolio rebalancing into riskier assets occurring gradually and having a persistent impact on financial conditions.

Furthermore, lower interest rates could also put some downward pressure on the exchange rate. The empirical evidence suggests that, consistently with this mechanism, the pound has tended to depreciate in response to QE announcements by the MPC.footnote [32]

While estimating the impact on real activity is more challenging, the literature suggests that QE programmes provided positive stimulus.

Averaging across 16 studies by academic and central bank researchers, the Bank‘s QE1 was estimated to have increased UK inflation and annual GDP growth by up to 1.8pp and 1.6pp at its peak, respectively. To put this into context, UK GDP dropped by 4.2% between the start and the trough of the GFC. In comparison, the Fed’s first Large-Scale Asset Purchase (LSAP1) programme was estimated to have had a larger impact on the economy, especially in terms of GDP (Chart 9).footnote [33]

Chart 9: Empirical estimates of peak impact of QE programmes: average (blue dots), minimum and maximum (orange squares) (a)

Left: Inflation. Right: GDP.

Footnotes

  • Sources: Fabo et al. (2021) and Bank calculations.
  • (a) APP: ECB’s Asset Purchase Programme. LSAP: Federal Reserve’s Large Scale Asset Purchase programme. Estimates are re-scaled to reflect a purchase programme worth 1% of domestic GDP.

The range of estimates across studies and QE programmes is very wide – particularly for inflation.

Researchers typically use models calibrated from historical relationships between bond yields and GDP and inflation to estimate the macroeconomic impact of QE announcements. However these models often embody a limited range of potential transmission channels and their estimates depend on the chosen identification scheme.footnote [34]

As Chart 9 shows, impact estimates across different studies and programmes vary considerably, especially for inflation.footnote [35] More broadly, there is substantial uncertainty around each individual estimate (Williams (2013), leading Borio and Zabai (2016) to argue that, even though the positive effect of QE is clearly present, its size and stability are quite uncertain. Therefore, and unsurprisingly given the identification challenges, there remains room for disagreement about the economic significance of individual QE episodes. This is also consistent with these programmes having operated through a mix of channels with different impacts on the real economy.

Understanding the strength of different channels is important to quantify the macroeconomic impact of QE.

While a majority of existing studies focus on the first QE programmes, Kaminska and Mumtaz (2022) study all of the MPC’s programmes before 2020 together and allow for a range of transmission channels (signalling, QE-specific supply, uncertainty). Their results suggest that insofar as it lowers gilt yields on announcement, QE has been associated with an improvement in real activity. They also find that the impact of QE on economic activity varies in strength with the mix of channels. In particular, the signalling channel can have a persistent impact on interest rates at the short end of the yield curve and so affect the pricing of a broad range of loans and financial assets. In contrast, local supply and duration effects have their main impact at the longer end of the curve. In line with these ideas, Kaminska and Mumtaz (2022) estimate that a given reduction in long gilt yields is associated with relatively larger changes in the exchange rate, corporate bond yields, and inflation and activity if it is driven by signalling rather than channels operating via term premia.

7: The role of asset purchases at the onset of the Covid-19 pandemic

At the start of 2020, the spread of Covid-19 and the measures taken to contain the virus triggered a sharp deterioration in the economic outlook and an increase in economic uncertainty. In response, investors suddenly demanded more liquid and safe assets. This ‘flight-to-safety’ was initially evident in an increase in the price of gilts and therefore a decline in their yields. But around mid-March the sell-off also extended to advanced economy government bonds, despite typically being considered the most liquid and safe asset class, reflecting a ‘dash-for-cash’. As a result, between 10 and 18 March, 10-year gilt yields increased by around 60bps (Chart 10), with similar spikes in US Treasuries and German Bunds.

Chart 10: Gilt yields between February and April 2020 (a)

Footnotes

  • Sources: Bloomberg Finance L.P, TradeWeb and Bank calculations.
  • (a) Zero-coupon spot rates derived from gilt prices.

The ‘dash-for-cash’ reflected an extreme spike in the demand for liquidity, at a time where dealers’ capacity to intermediate in the gilt market was constrained.

In the UK, the demand to liquidate gilts was driven in part by a need for investors, such as domestic insurance and pension funds, to meet significant margin calls.footnote [36] In normal circumstances, dealers provide liquidity by holding gilt inventories on their balance sheets. In March 2020, the growing imbalance between the demand from clients wishing to sell gilts relative to those wishing to buy required dealers to ‘warehouse’ unusually high volumes of gilts on their balance sheets (Chart 11), and liquidity provision was insufficient to fully accommodate demand.footnote [37] The combination of these factors resulted in a sharp deterioration of market liquidity. One way this manifested was in an abrupt widening of bid-ask spreads, which is one of the ways dealers are compensated for the risks of holding gilt inventories (Chart 12).footnote [38]

Chart 11: Cumulative net purchases of gilts by dealers (a) (b) (c)

Footnotes

  • Sources: MiFID II transaction data and Bank calculations.
  • (a) Includes all conventional UK government bonds. Measured in nominal value.
  • (b) Estimated from transaction reports.
  • (c) Smoothed using a centred five day moving average of cumulative purchases net of sales.

Chart 12: Gilt bid-ask spreads

Footnotes

  • Sources: Eikon Refinitiv and Bank calculations.

Central banks responded with asset purchases of unprecedented scale and speed, along with other measures aimed at improving market functioning and mitigating an unwarranted tightening in financial conditions.

Other things equal, the deterioration in gilt market conditions and the associated tightening in financial conditions would have further depressed the UK economy. Therefore, on 19 March the MPC announced that it had voted unanimously to increase the stock of asset purchases by £200bn, and that it would complete the programme ‘as soon as operationally possible, consistent with improved market functioning’.

At the same time, the Fed and ECB also announced unusually large and fast-paced asset purchase programmes (Table A). In addition, major central banks reinstated and extended swap line programmes. This allowed the Bank to lend US dollars to its counterparties, potentially mitigating their liquidity needs.footnote [39] On 24 March the Bank also launched the Contingent Term Repo Facility (CTRF), with the objective of further supporting market functioning.

Table A: Timeline of key policy announcements in March 2020

Date

Announcement

11.03

UK: Bank Rate cut to 0.25%; TFSME

15.03

US: Fed Funds cut to between 0% and 0.25%

US: >$500bn Treasury purchase programme

Swap lines costs reduced

17.03

US: Primary Dealer Credit Facility

18.03

EA: Pandemic Emergency Purchase Programme

19.03

UK: £200bn QE programme

UK: Bank Rate cut to 0.1%

US: Treasury purchases surge to over $60bn/day

20.03

Swap lines frequency switched to daily

23.03

US: Treasury purchase ‘in the amount needed’

US: Corporate bond purchase facilities;

24.03

UK: Contingent Term Repo Facility (CTRF).

Gilt yields retraced a significant portion of their surge after the MPC’s announcement.

Summing together the high frequency reactions to the 19 March MPC announcement and the subsequent Market Notice, 10-year gilt yields fell by 24 basis points (Chart 13) – around 40% of the surge in yields over the ten preceding days.footnote [40] Since there was no other significant news over this window, this effect can be attributed to the Bank’s announcements. Since it might have taken time for markets to fully absorb the news however, these moves could provide a lower bound for the full impact of the announcements.footnote [41] On average, medium and long-term gilt yields fell by around 40 bps in the two days after the announcements. This might be an upper bound because the parallel extension of Fed and ECB asset purchases and swap lines could also have helped to lower gilt yields by improving market functioning.

Chart 13: Gilt yields on 19 and 20 March 2020