Data and methods

The article relies on a set of 51 interviews with central bank personnel and financial market participants primarily in the UK and Germany (see Table 1). Twenty-one of these interviews were carried out with central bankers, primarily Bank of England personnel, including former members of the Monetary Policy Committee and staff in the Markets and Monetary Policy divisions of the Bank of England. Thirty interviews with financial market participants in the UK and Germany were also highly informative. Of note are a significant number of interviews with interviewees whose role at the time of the interview was that of a market participant but who also held a central banking experience and the inverse for other interviewees. Importantly, these interviews provided rich detail spanning the fluid ‘boundaries’ of central banks and financial markets (Coombs and Thiemann, Reference Coombs and Thiemann2022), with interviewees discussing in detail the oft-concealed relations between central banks and markets.

Table 1.

Interviews broken down by professional and organisational category.

This article draws more directly on interviews with central bankers, primarily those on the Bank of England’s Monetary Policy Committee and those central bankers in the engine rooms of central banks who design and run the models, especially DSGE models. In many cases, the latter are experts on DSGE modelling and have spent years working on, tweaking, and in some cases even revamping these models given that their experiences span the pre-crisis and post-crisis periods. Equally important are those interviews with traders in hedge funds and in investment banks who experienced the market disruptions of the financial crisis and the March 2020 debacle. The rest of the interviews informed the article indirectly.

The interviews were conducted between April 2019 and January 2021. Those interviews conducted between April 2019 and March 2020 were done in person, while those interviews conducted following the onset of the pandemic were carried out via audio/video call. All interviewees, irrespective of seniority and position, were guaranteed full anonymity. Interviews lasted between 28 and 180 minutes, with the average interview taking 69 minutes. During the interviews, I asked the interviewees for views about the relationship between central banks and financial markets and the role of the socio-technical infrastructures that constituted this relationship. In this light, I also asked how central bankers and market participants made sense of the relatively new policies known as unconventional monetary policies, including QE and forward guidance, among others, and how these shaped their practices.

I triangulate my interview data with a large set of documents from digital archives and the Bank of England website, ranging from 2009 to 2023, primarily involving Bank of England documents. These include Monetary Policy Committee meeting minutes, speeches given by MPC members, Inflation Reports and Monetary Policy Reports, press conference transcripts, public statements and letters, research papers, Asset Purchase Facility Reports, Quarterly Bulletins, oral evidence to Treasury Committee, and other relevant documents.

I use the case of the Bank of England because it was one of the first central banks in which the programme of expectation governance was erected alongside the development (and making) of deep and liquid financial markets during the 1980s and 1990s (Wansleben, Reference Wansleben2018), and it was also one of the first central banks to turn to QE in response to the financial crisis of 2007/08, which led to internal conflicts on the form QE would take and to undertake significant relational and political work around it (Cassar, Reference Cassar2024). The process of scientisation of the Bank of England since the 1990s has also meant alignment work between its socio-technical devices and financial market developments, making it an appropriate case study for an investigation of post-crisis monetary policy. Its generative mould, as a technocratic central bank implementing procedures and policies in line with other central banks, also allows a degree of generalisability to other central banks across the world, though contextual and institutional conditions will necessarily differ, and in spite of comparative assessments not being within the scope of this article.

The model of efficient markets in pre-crisis monetary policy

This section explores how conventional monetary policy – revolving around the performativity of expectations – was conditioned on a model of complete and efficient financial markets and on this model being an acceptable real-world approximation for the central banker. While a lot of work has been done on how central banks govern market expectations (Holmes, Reference Holmes2013; Braun, Reference Braun2015; Reference Braun2014; Wansleben, Reference Wansleben2018), the role of the model of efficient markets as the backbone of this policy arrangement has not been given much attention. I show how, in treating their model as an acceptable approximation of reality, central banks came to rely on ‘the arbitrageur’ from a distance. That is to say, central bankers assigned much of the work in the transmission mechanism of monetary policy to the arbitrageur (i.e., fixed income relative value traders such as those in leveraged relative value hedge funds). This analysis serves as pre-amble to understand precisely how post-crisis QE implementation implied a performative intervention in financial markets.

The arbitrageur is a direct descendent of modern financial theory developed in the second half of the 20th century, and the no-arbitrage ‘epistemic culture’ to which this gave rise in financial markets (MacKenzie, Reference MacKenzie2006; MacKenzie and Spears, Reference MacKenzie and Spears2014). As MacKenzie (Reference MacKenzie2006) shows, these arbitrageurs enacted market efficiency as they exploited deviations from their theoretical models embodying efficient markets. In exploiting deviations, therefore, arbitrageurs pushed market prices towards the theoretical prices as exemplified by models. The socio-materiality of arbitrage, particularly of the kind this article deals with, has been the subject of study in the social studies of finance, in what has been called ‘a material sociology of arbitrage’ (Beunza and Stark, Reference Beunza and Stark2004; Beunza et al., Reference Beunza, Hardie and MacKenzie2006; Beunza, Reference Beunza2019). Because it is not within the scope of this article to take up a discussion on the murky definitions of what constitutes arbitrage (see Hardie, Reference Hardie2004, for such a discussion), for the purposes of this article, I will restrict the term to the way market participants and central bankers themselves refer to it: any form of market-neutral trading that may involve (non-directional) risk and that seeks to profit from price discrepancies between two (or more) securities that in principle should trade close to each other.

So how do arbitrageurs enact complete and efficient markets, thus supporting the central bank’s governance of expectations? The answer lies in a sociomaterial device that is a key tool on the desk of the arbitrageur – the yield curve (Zaloom, Reference Zaloom2009; Cassar, Reference Cassar2022), a tool around which both central banks and markets are cognitively and materially entangled. The yield curve is a technical illustration of yield levels and a representation of market pricing in bonds and their derivatives. The graph is a comparative representation of yields on similar-quality bonds with different maturities. As a representation of market pricing, most of the analysis and activity happening in the bond market revolves around it (Zaloom, Reference Zaloom2009; Choudhry, Reference Choudhry2019).

The yield curve is the arbitrageur’s world, existing as a ‘material system’ in the sense of mechanical engineering or physics – a mathematical space holding a set of relationships along the curve, that should, in theory, be mean-reverting (Interview 32). For the arbitrageur, what matters is not the future direction in which yields move (e.g. via a central bank announcement). Indeed, the latter would be the remit of directional traders who place trades on the future yield (or price) of a security, such as a bond, which action is entirely reliant on the formation of expectations which central banks seek to govern. (It is to these directional traders, or central bank watchers, that the social studies of central banking refer to when they speak about market participants and their expectations.) On the directional trader’s desk, the yield curve functions as an agglomeration of market expectations, along or against which a view (or a bet) is expressed. However, on the arbitrageur’s desk, the yield curve is ontologically an entirely different object, one that represents statistical and mathematical relationships between yields. For the arbitrageur, then, what matters is that yields move in some predictable fashion relative to similar securities, hence relative value (Interview 25).

While there are multiple yield curves (swaps curve, money market curves, government bond market curves, etc.), it is often the bond market yield curve that concerns central bankers. But this restricted focus is only possible because of the existence of arbitrageurs who connect yield curves together (Woodford, Reference Woodford2003) via the analysis and trading-based exploitation of spreads (Beunza and Muniesa, Reference Beunza, Muniesa, Latour and Weibel2005) across curves. Hence the transmission of monetary policy can move relatively seamlessly across financial markets. They do so by exploiting (and therefore bringing about) converging relationships, or associations (Beunza and Stark, Reference Beunza and Stark2004), between on-the-run and off-the-run bonds; basis trades between government bonds and futures; and spreads between government bonds and swap curves. The result of such trading is that different markets become connected, not unlike the socio-materially constituted ‘world market’ that Beunza et al. (Reference Beunza, Hardie and MacKenzie2006) speak of. Central bank’s action on short-term interest rates will, in theory, reverberate to other markets (see also Gabor, Reference Gabor, Goodhart, Gabor, Vestergaard and Erturk2014) because of the work of the arbitrageur. Indeed, this is an aspect of the perlocutionary dimension of monetary policy which Walter (Reference Walter2024) emphasises and which has often been neglected in the literature.Footnote ²

Central bankers could then adopt this conception of markets as an acceptable ‘real-world approximation’ underlying their policy arrangement. This model is inscribed in the material devices through which central banks make policy. One interviewee who has worked for years in modelling at the Bank claimed that it is precisely for this reason that central banks’ key macroeconomic models of DSGE did not, at least not until after the crisis, incorporate the financial and banking systems within them (Interview 17). In this model world, financial markets are absent because they are treated as largely efficient and frictionless, and hence there is no need for them – in other words, the transmission of monetary policy would seem to happen efficiently and instantaneously, and hence one could abstract away financial markets in the model (Woodford, Reference Woodford2003; Dale, Reference Dale2010; Braun, Reference Braun2014; Bailey et al., Reference Bailey, Bridges, Harrison, Jones and Mankodi2020). DSGE models involve a range of models, and many participants cautioned against lumping all kinds of DSGE models into one basket. There are some, for instance, that incorporate different kinds of financial frictions in them. Others, like those dominant in central banks prior to the crisis, however, do not. These come in the form of an Arrow-Debreu style of models, in which markets clear instantly and in efficient ways. They include an intertemporal real interest rate channel that is powerful and which can be split into inflation and policy expectations (Interview 4; see also Braun, Reference Braun2014).

The very fact that DSGE models did not incorporate frictions also suggests that central banks largely assumed away those frictions. In practice, central bankers are well-aware that markets are never fully complete and efficient and always entail some degree of frictions (Shleifer and Vishny, Reference Shleifer and Vishny1997; Nyborg and Östberg, Reference Nyborg and Östberg2014). The very existence of arbitrageurs would suggest that markets are not completely arbitraged, since arbitrageurs depend on the identifying of frictions in order to waive them away via trading. And the transmission mechanism of central bankers, therefore, is never instantaneous. Even so, central bankers often accept these marginal frictions if they are able to deem markets as efficient and complete to an approximation.

For central bankers, the degree to which markets approximate perfectly complete and efficient markets is largely down to their assessment of the leverage abilities of arbitrageurs. Because price deviations are typically small, the arbitrageur needs to deploy several magnitudes of leverage to extract reasonable profit. Leverage is crucial for arbitrage trading; leverage-constrained market participants would not exploit the arbitrage opportunities because the profits generated from such trades are so low that it is only leverage, which makes these trades worth exploiting. In fact, central bankers tie the notion of complete and efficient markets directly to leveraged arbitrage. As one central banker told me:

Their assessment of leveraged arbitrage, and the model of complete and efficient markets, allows central bankers to restrict their focus to market expectations. The yield curve itself, taken as a representation of expectations, has been enrolled into the expectations management policy arrangement of pre-crisis central banking. In other words, the yield curve became the world of the Bank’s monetary policy, supported by an interpretive space that redefined the yield curve in terms of expectations. Assessing that arbitrage relations are functioning allows central banks to engage in what Jessop (Reference Jessop2010) calls complexity reduction, by carving and opening up an epistemic space around the yield curve of the government bond market and the expectations hypothesis. Rather than being overly concerned with multiple markets, with a multitude of expectations and how to tame them, and with how that is transmitted to the ‘real economy’, they could therefore assume away the transmission mechanism of monetary policy ‘outside’ of the government bond yield curve. Concealed beneath the Bank’s expectations management policy arrangement, therefore, was an understanding that a muted process was labouring in the background, led by arbitrageurs whose trading made markets more complete and efficient, and inscribed in the very models that render the economy visible and the governance of expectations calculable.

Modelling incomplete and inefficient markets in post-crisis monetary policy

The central banks’ policy arrangement of expectation governance met its limits at the height of the 2007/08 financial crisis. As central bankers pushed their policy rates close to zero, they hit the zero-lower bound. The model by which central bankers operated, as argued in the previous section, was a New Keynesian DSGE model which incorporated an idea of efficient and complete markets (by omission), and importantly, one in which only interest rates would have some bite on the real economy (Woodford, Reference Woodford2003). Given that the policy rate was at zero, central bankers needed to turn to other policy tools for monetary policy purposes.

This required central bankers to design a new policy toolkit, which itself depended on the model of financial markets underlying the transmission mechanism. And this is precisely where the question of the model of financial markets becomes important. David Miles, speaking in 2009, had argued that ‘[t]here are economic models which would imply that the impact [of QE] is small – though I believe they rely upon unrealistic assumptions, specifically about the efficiency and completeness of financial markets’ (Miles, Reference Miles2009: 13). Indeed, at the height of the crisis, disruptions in financial markets were glaringly evident for most economists. One MPC member (Interview 12), referring to 2008, claimed that it was obvious that markets were not perfectly arbitraged and that there would hence be supply effects.

The model which came to the fore was one going all the way back to Culbertson (Reference Culbertson1957) and Modigliani and Sutch (Reference Modigliani and Sutch1966) and recently reworked by Vayanos and Vila (Reference Vayanos and Vila2009), the latter of which became the key model for central bankers. A DSGE model specialist argued: ‘So then for us central bankers, this work by Vayanos and Vila is important because for the first time it established a link between quantities and supply effects’ (Interview 24). In this model, agents are heterogeneous, assets are not perfectly substitutable, and therefore in which there are limits-to-arbitrage because of preferred habitats or market segmentation. If market participants prefer to inhabit local parts of the curve, and arbitrageurs fail to undo those preferred habitat effects (e.g. because they have capital/liquidity/collateral constraints) (Tenreyro, Reference Tenreyro2023: 9), then there are limits-to-arbitrage because assets are no longer perfectly substitutable, and, for instance, rates at the shorter end bear little relation to rates at the longer end (Interview 17; Interview 24). If these preferred habitats are not undone (see Bernanke, Reference Bernanke2020), QE can exploit local supply effects by influencing the risk component (risk premia) of the yield curve, rather than expectations.

And yet this was a world away from the model of efficient markets which central bankers had worked with previously. For decades Bank economists had slowly and incrementally built tools and devices that shaped the doing of monetary policy, such as DSGE models, which implied the existence of complete and efficient financial markets. Replacing the model of complete and efficient market was no straightforward mission, even materially. The model of efficient markets was so dominant that model specialists across central banks (and at the Bank of England) had never needed to formalise consensus around how to model financial frictions, market inefficiencies, and incompleteness. And the edifice of the post-crisis New Keynesian DSGE model known as COMPASS (which replaced BEQM in 2011) at the Bank of England was still built on an inscription of ‘perfectly functioning financial markets’ which retained an inability to allow for an impact of QE. One model specialist at the Bank argued:

The DSGE model itself, therefore, retained the assumption of efficient markets, hardcoded into the material model. And forecasters at the Bank of England resorted to introducing some of these frictions in their macroeconomic modelling practices:

Bank staff would superimpose QE effects as if they were expectational effects. They:

Qualitative and non-model-based judgements about the impacts of QE were superimposed on the model via expectational effects, supported via the satellite suite of models around the DSGE (Burgess et al., Reference Burgess, Fernandez-Corugedo, Groth, Harrison, Monti, Theodoridis and Waldron2013; Interview 17).

But while materially the device failed to directly include frictions, superimposing these frictions in the policy arrangements implied that the cognitive model of complete and efficient markets had been replaced by incomplete and frictional markets. Indeed, this is the conclusion reached by some scholars and observers of the central banking world (Borio, Reference Borio2011; Gabor, Reference Gabor, Goodhart, Gabor, Vestergaard and Erturk2014: 158). The question this raises is: have central banks permanently replaced the underlying model of the transmission mechanism from one of complete and efficient markets to one of frictional and imperfect markets?

Performing ‘states of the world’ via quantitative easing

The above empirical sections have elucidated how the state of the world in which central banks operate changed following the crisis and how the Bank scrambled to realign its operative procedures with the shifting state of the world (i.e., financial market conditions). This section will first lay out the state contingency of monetary policy from the perspective of the Bank of England, detailing how those very same interventions may shape the state of the world, thus giving rise to a performative element to monetary policy as world-making ontological shaping.

The notion that monetary policy interventions are state-contingent is reflected in the documents published by the Bank of England – see Miles (Reference Miles2009), Haldane et al. (Reference Haldane, Roberts-Sklar, Wieladek and Young2016), Broadbent (Reference Broadbent2018), Vlieghe (Reference Vlieghe2018), Bailey et al. (Reference Bailey, Bridges, Harrison, Jones and Mankodi2020), Vlieghe (Reference Vlieghe2020), Ramsden (Reference Ramsden2021), and Tenreyro (Reference Tenreyro2023). Interventions and their transmission channels mirror the state of financial markets (see Table 2 for a schematic illustration). The governance of expectations is most relevant in the benign state of the world when financial markets are more efficient and complete. In the less benign state of the world (when arbitrage relations break down), QE takes stronger hold via risk premia as markets are less arbitraged (Buiter and Sibert, Reference Buiter and Sibert2007; Mehrling, Reference Mehrling2011; Bailey et al., Reference Bailey, Bridges, Harrison, Jones and Mankodi2020).

Table 2.

The state contingency of monetary policy.

Source: Author’s own.

The Bank’s underlying model of financial markets is therefore not static, but changes to reflect their assessment of the ‘states of the world’. And this will, in turn, shape the contingent interventions. For instance, from this viewpoint, QE will be deployed more actively when financial markets are disrupted, but not when markets are seen to be well-arbitraged and hence approximating the efficiency and completeness of markets. This is not to say that the channels are mutually exclusive but that QE’s effectiveness and channels rely much on the states of financial markets. Bailey et al. (Reference Bailey, Bridges, Harrison, Jones and Mankodi2020: 18) argue:

Similarly, Vlieghe (Reference Vlieghe2018: 17) argues: ‘large quantity effects are mostly temporary, and larger when intermediaries are constrained, as in the financial crisis. But they have much smaller or no long-run effects once intermediaries are unconstrained’. And on the same lines, Tenreyro (Reference Tenreyro2023: 14) claims that ‘QE effects on yields depend on the risk capacity of arbitrageurs. If that risk capacity is state-contingent, QE effects will be state-contingent and largest when market liquidity dries up’. Indeed, it is perfectly reasonable to think that in a liquidity crisis, arbitrageurs become increasingly constrained by limited capital and liquidity, resulting in those preferred habitat/market segmentation effects to remain in place for longer.

Another central banker interviewee argued:

But while monetary policy is contingent on the state of the world, there is also a performative element underneath the process of monetary policy. To the extent that QE is state-contingent, the very act of purchasing assets in financial markets will have an effect on financial markets conditions, that is, the states of the world themselves. In line with a MacKenzian and Callonian conception of performativity, I mean an act or intervention (QE) that brings about the (state of the) world of which it speaks. The model of efficient markets that was assumed away in the benign state of the world (i.e., as one which explains the conditions under which central bankers govern) is transformed into a performative object to be achieved in the less benign state of the world.

While in the previous section we have seen how the Bank sought to replace its own model to realign it with the state of financial markets, the model of complete and efficient markets remained operative in the background – the model drives contingent interventions to shape financial markets in its image. The intervention of QE is hence more than mere monetary policy – it is an intervention in the very texture of financial markets that alters them and their conditions. And the intervention is intended to ontologically shape markets into the image of the model of complete and efficient markets. This is particularly visible in those particular instances when central bankers judge the model to have detached from reality and when there is a threat that the model remains detached from reality in a state of permanent performative failure. Instances of severely impaired market functioning are one key example of this. The latter implies arbitrageurs who, despite having trades that seek to bring about some degree of efficiency and completeness in the market, are constrained from doing so. Even worse, they are pushed to wind up the trades they already have on, as assets/securities that normally trade close to each other break away from each other.

The March 2020 ‘dash for cash’ is perhaps the most clear-cut example to illustrate the shift from one state of the world to the other, a shift brought about performatively by central banks. As the World Health Organisation declared Covid-19 a pandemic, markets responded quickly and intensely: a modern-day ‘bank run’ had set off. Global asset prices moved several basis points in the space of a few minutes and margin calls grew.Footnote ³ The arbitrage relationships described earlier stood at the heart of the event. One of these in particular, known as the ‘cash-futures basis trade’, represented by far the most popular trade for relative value hedge funds in the Treasury market. This trade involves hedge funds taking large positions in government bonds funded on repo (and thus leveraged multiple times over) as the first leg of the trade, and shorting the corresponding futures contract as the second leg. In making the spread between the two, hedge funds normally enforce ‘the law of one price’ – that is to say, the exploiting of the opportunity moves the two rates (government bond and future rates) closer until there are no more opportunities left.

In the case of the March dash for cash, however, the same actors that are in other circumstances crucial in maintaining efficiency in financial markets, this time they turned into a source of systemic risk.Footnote ⁴ Rising volatility as investors turned cash-hungry triggered margin calls as collateral prices fell, requiring them to either deposit additional funds or sell assets to maintain the required margin level. Non-banking institutions in particular, mostly leveraged hedge funds, saw the normally mean-reverting arb spreads moving against them, resulting in mark-to-market losses. For instance, the cash-futures basis that would, in efficient markets, trade at close to zero shot up to hundreds of basis points. This led to higher margin calls by dealers – that is, a demand by the broker for the trader to deposit additional securities as collateral. However, with not enough capital to meet those margins and to keep their positions on, traders were forced to unwind their positions by dealers. The result was further volatility and pressure on dealers’ balance sheets, leading to a lower intermediation capacity by dealers and hence a further contraction of liquidity. This was a looping effect, a ‘doom loop’ (Hauser, Reference Hauser2020), or margin spiral (Brunnermeier and Pederson, Reference Brunnermeier and Pederson2009) on those same spreads and the market-makers’ bid-ask spreads, in which a severe shortage of liquidity, price volatility, and rising margin calls fed on each other, thus threatening the very functioning of financial markets (Duffie, Reference Duffie2020; Schrimpf et al., Reference Schrimpf, Shin and Sushko2020).

Severe market disruption, then, and perfect conditions for QE to make an impact. A decision to ‘go big and fast’ by major central banks, including the Bank of England, helped restore some stability to financial markets (FSB, 2020; Hauser, Reference Hauser2021; Reference Hauser2020). The central bank’s performative intervention worked insidiously through a (temporary) reformatting of the network that constituted the market (and the very liquidity that this made possible). The central bank enters the market, armed with a balance sheet, and makes bids for assets, mostly in the form of government bonds. Exchanging liquid cash for bonds, it first offers a reprieve for the conduits and intermediation between arbitrageurs. Dealers offload their inventories onto the central bank and hence also offload risk from their own balance sheet to the central bank’s, lowering risk premia. This frees up their capacity and appetite to intermediate between arbitrageurs. Lower margin calls remove some of the pressure off arbitrageurs to wind down their trades, thus allowing them to maintain them until they are successful.

The injection of liquidity was only possible by a reconfiguration of the network in which the central bank intervened, and the behaviour of the network was reformatted as market participants regained confidence and risk appetite to make markets, at which point the central bank could slowly exit the market. An intervention such as QE targets confidence and risk, minimising those constraints on arbitrage, which is an attempt to bring about the performative object of complete and efficient markets.

From Callon’s (Reference Callon, MacKenzie, Muniesa and Siu2007) point of view, the formula/theory/model is not brought into being merely by the mere existence of the formula/theory/model itself. There are other forces and interests that cause the socio-technical network, or agencement, to be enacted. It is not enough, therefore, that both central bankers and market participants held a shared world of efficient markets in their activities. At certain points (such as the market impairment of March 2020), reality could stray away from the model, and it required efforts and an alignment of interests of the various actors to bring reality back to an approximation of efficient markets. Firstly, it required the effort of a central bank to exercise its balance sheet and to enter the network as it announced its intervention. And second, it required a condition in which it was in the interest of market participants – namely, arbitrageurs, dealers, and the central bank itself – to restore market efficiency. Indeed, the arbitrageurs and dealers willingly participated and cooperated with central banks in a reconstitution of the socio-technical agencement via which liquidity and confidence flowed once again. As risk abated, the yield curve could be once again interpreted primarily in terms of the expectation hypothesis (given the drop in risk premia), while the DSGE model became aligned more closely to the conditions of financial markets post-intervention.

QE, therefore, had a performative effect on the state of the world in which it was intervening. One member of the MPC explained that QE is primarily a monetary policy instrument, but given that the transmission mechanism relies on the state of the world, the way it operates has an element of world-making performativity to it, particularly in periods such as March 2020. One of my interviewees argued so:

The same process has been manifestly evident more recently in the Bank of England’s conduct in 2022. As inflation in the UK skyrocketed, the MPC announced several rate hikes and later what it dubbed as ‘quantitative tightening’ (QT).Footnote ⁵ The planned commencement of the programme for September 2022 came at an unfortunate moment in which the new British Chancellor, Kwasi Kwarteng, announced a ‘mini-budget’ involving planned freezes of household energy bills to be funded by the issuance of gilts and coupled with tax cuts for wealthier households. The announcement triggered a sell-off in the gilt market and a downward spiralling of the pound. The Bank, all set to start QT, reversed course on 28 September and announced a temporary programme of QE in an attempt to ‘restore market functioning and reduce any risks from contagion to credit conditions for UK households and businesses’ (BoE, 2022). In effect, the Bank was banking on the performative powers of QE to alter the state of the world in which it is effective.

But QE’s performative powers are also evident when there is no severe shortage of liquidity in markets. For instance, the Bank provides empirical evidence that the portfolio rebalancing channel was stronger in the first few years after QE was implemented (Joyce and Tong, Reference Joyce and Tong2012). As we have seen, prior to the crisis, central bankers were mostly concerned with governing expectations, and consequently, they interpreted interest rates as incorporating expectations about future rates. And while they were aware of the ‘risk premia’ that certain bonds entailed beyond the expectational component (especially at the longer end of the curve), they never turned them into an instrument of governance.Footnote ⁶ Although not modelled in DSGE models, risk premia are primary targets of QE, which intervention is intended to minimise or eliminate risk premia (Lane, Reference Lane2019). Central bankers seek to nudge market participants to rebalance their portfolios towards riskier assets, thus taking on more risk on their balance sheet. Rather than targeting expectations, the liquidity channel and portfolio rebalancing channel target risk.

Both these channels may have a performative effect on the state of the world, as they diminish risk premia through a shift in the arrangements and behaviour within the market. This means that QE would have had a differential impact according to the changing state of the world. In fact, Busetto et al. (Reference Busetto, Chavaz, Froemel, Joyce, Kaminska and Worlidge2022) claim ‘Therefore there were good reasons to believe that the impact of the March 2020 programme would be closer to the one from QE1 [in 2009], whereas the impact of subsequent programmes was likely to be lower, as market dysfunction and risk aversion had abated’.Footnote ⁷

I have argued that the model of complete and efficient markets was temporarily abandoned as an epistemic tool on which monetary policy hinged. In other words, the pre-crisis model with which they operated no longer approximated the world that central bankers were observing. And yet, despite the temporary abandonment of the model in the sense just outlined, it was still labouring in the background of monetary policy. Not only did it remain the underlying model of the workhorse DSGE model, but it was turned into a performative outcome to be achieved. A clue can be found when central bankers spoke of ‘returning to normality’ or a variation thereof (Bailey et al., Reference Bailey, Bridges, Harrison, Jones and Mankodi2020; Lagarde, Reference Lagarde2022; Schnabel, Reference Schnabel2023). While this can be easily interpreted as a return to pre-crisis economic growth, stable inflation rates and the like, it also refers to the pre-crisis monetary policy of expectation management. As QE brings about the ‘performative outcome’ of efficient markets, it ends up enacting a world in which it itself is less effective. In other words, as it moulds financial markets into efficient ones, QE is seen to become less effective due to the neutrality result in efficient financial markets (Wallace, Reference Wallace1981).Footnote ⁸ And this is perfectly in line with the Bank’s intention to return to the pre-crisis policy. This is how MPC members anticipated the return to a QE-less world, a process that the Bank of England started in February 2022. The wind down itself, it was argued, would not have significant effects because the state of the world under which it will be wound down would be different from that in which QE was first implemented. In this ‘state of the world’, balance sheets could be wound down in a context of efficient markets, as had been predicted by David Miles as early as 2014:

By the time that central banks start the process of returning to the management of expectations, which was always an end goal for central banks (Bailey et al., Reference Bailey, Bridges, Harrison, Jones and Mankodi2020: 27–28), the state of the world would have altered in such a way as to approximate complete and efficient markets. Central bankers, then, have always aimed to – and worked towards a – return to a form of monetary policy that relies on the governing mechanism of expectations management in a world of efficient markets.