The Risk Taker of Last Resort

Author: Richard Barwell

Published: 5 April 2022

Introduction

Large scale asset purchase programmes by central banks have become a familiar part of the landscape in financial markets. There is a lot of attention on what assets central banks plan to buy and how much money they plan to invest, but rather less on the specific reasons why.

Central banks can buy assets to pursue financial stability as well as monetary stability goals, but it would appear that many, if not most, market participants and commentators view all asset purchase programmes as essentially monetary policy operations.

The events of the opening phase of the pandemic are a case in point: almost everyone seems to treat the torrent of asset purchases that the global central banking system unleased in March 2020 as just another dose of QE, albeit a very large one. That is not the right way to think about what happened. The purchases were no doubt a game-changer. But they should probably be thought of as monetary policymakers engaging in asset purchases to pursue essential financial stability goals.

This article argues that it would be wise to acknowledge that reality and formally establish this function – which could be thought of as a risk taker of last resort – in the central bank crisis management toolkit (as opposed to the monetary toolkit). After all, it may well have a critical role to play in future crises.

Market maker of last resort

A consensus did emerge after the financial crisis – at least within the narrow confines of the small set of academics and central bankers who work on macroprudential policy – that there was a justification for a central bank to purchase assets in the pursuit of financial stability objectives. Central bankers could act as a market maker of last resort (MMLR).

The MMLR function is essentially the capital markets analogue of the lender of last resort (LOLR) function that supports illiquid but solvent systemically important banks. The MMLR intervenes in systemically important markets to restore liquidity by standing ready to buy and sell securities at a tighter spread than the prevailing bid-ask in a dysfunctional market. Early in the financial crisis, the then Deputy Governor of the Bank of England, Paul Tucker, explored the principles that should guide MMLR operations (Tucker (2009)):

1. Do not interfere with the stance of monetary policy
2. Charge a penalty – that is, buy at a discount – relative to the fundamental value of an asset
3. Operations should reveal information about the state of the markets
4. Do not over-reach capital resources
5. Aim to catalyse rather than replace private markets
6. Do not proper up markets that are no longer viable.

The similarity with the principles that should guide LOLR operations was intentional. For example, the Bagehot principle is transposed into the MMLR setting such that the central banks should stand ready to provide unlimited support but at ‘high rates’, which Tucker argues should be understood as not necessarily penal, but in this context, as a wider bid-ask spread than that which would prevail in normal (i.e., pre-crisis) market conditions. In short, the MMLR involves “stepping into the market with a narrower bid-offer spread than private-sector intermediaries would otherwise offer, with a view to those very same dealers bringing their bid-offer spreads inside ours and to investors reducing the liquidity premium they require” (Tucker (2009)).

Central banks had felt compelled to innovate MMLR operations on the fly during the financial crisis. Tucker wanted to stimulate a debate on the rules of the game for MMLR and other financial stability operations so central banks would be better prepared in the next crisis. Unfortunately, that debate soon went cold as the current Executive Director for Markets at the Bank explains (Hauser (2021)): “the Bank – in common with other central banks – chose to say relatively little in public. That reflected a number of factors, including the practical challenges of determining in advance the markets in which central banks might operate, the terms on which they would do so, and the consequences for public money.”

Introducing the risk taker of last resort

A successful MMLR operation would likely move asset prices – in a direct sense by squeezing bloated liquidity risk premia that become embedded in the price of many assets in dysfunctional markets, and then in an indirect sense by squeezing compensation for other sources of risk if the operation is catalytic and helps to shift the system onto the path to recovery. However, an MMLR operation is not designed to squeeze the compensation that investors demand, for say, the risk of default or the delay or deferral of the flow of income from a security.

Barwell (2013, 2017) defined an alternative operation – the risk taker of last resort (RTLR) function – that would have the explicit objective of leaning against significant and unwarranted declines in asset prices that are judged incompatible with fundamentals. An RTLR operation would set out to squeeze an unwarranted increase in the credit risk premium implicit in the price of corporate bonds or the equity risk premium implicit in the price of equities. Put crudely, the RTLR involves a more emphatic attempt to put a floor under asset prices than the MMLR because it is comfortable leaning against a much broader set of factors contributing to a collapse in valuations.

The objective of an RTLR operation must be to lean against what are considered to be excessive movements (relative to fundamentals) in asset prices that are considered systemically important where the collapse in valuations has the potential to exacerbate stress within financial markets and the wider real economy. That amplification might arise because of the implied hit to the net worth of both financial intermediaries and households and companies in the real economy with the associated increase in credit risk premia, or because beliefs about fundamentals are themselves sensitive to observed prices driving privately rational but socially destructive defensive actions, or because one or other mechanisms leads to the collapse of systemically important financial institutions and a default cascade.

The boundary between the MMLR and RTLR functions might look blurred in the real world. The distinction between them boils down to why the central bank is intervening in markets: which movements in asset prices she is trying to lean against. Of course, the distinction between liquidity risk premia and credit risk premia is a theoretical one. In practice, asset prices will be collapsing and central banks will be under pressure to act, regardless of how central bank officials decompose the movement in asset prices. Moreover, in both cases liquidity is likely to be draining out of financial markets, so liquidity risk premia will appear to widen. Nonetheless, it would be wise to have a conceptual framework to guide interventions, or more precisely to establish at what price the central bank is willing to purchase securities. Will you buy at prices consistent with the market’s dire estimates of default and recovery rates in the midst of a panic, or your own more considered estimates of what the future holds?

An RTLR operation will almost surely involve taking more risk onto the central bank balance sheet than a MMLR operation. One could argue that because central banks have the luxury of not facing liquidity constraints they can afford to be patient and “await mean reversion of values” (Coeure (2012)) so a buy low, sell high operation should tend to make, not lose, money for the taxpayer (Farmer (2014)). However, this will not always be the case. A RTLR intervention to stabilise equity prices in Japan in the early 1990s might have proved very expensive.

There is a debate to be had about under what circumstances – if ever – a RTLR operation would be appropriate. It is unclear whether unelected officials have the remit to conduct these operations in any capacity other than as an agent of the state under direction from the government, given the risk to the taxpayer in the event that the operation fails. It is also unclear whether these interventions are wise in the long run, on the basis that putting a floor under asset prices today may encourage excessive risk-taking tomorrow. The discussion should not begin and end with the traditional macroeconomic focus of what is required to stabilise aggregate demand and inflation. Of course, you need to have a serious debate about what exactly the macroprudential regime is for – and in particular, write down a fit for policy purpose model of the system and a loss function – if you want to debate those issues seriously (Barwell (2021)).

Asset purchase programmes of interest

The concept of a market maker of last resort did enter the vocabulary of central bankers after the financial crisis but the terminology was not widely used in the policy debate. Several asset purchase programmes were developed in the period between the financial crisis and the pandemic that have a clear financial stability dimension. There is a temptation to refer to all of them as MMLR operations. But if we adopt the distinction above – that an MMLR operation should focus on squeezing liquidity risk premium whilst the RTLR function has broader ambitions – then these programmes bear more than a passing resemblance to the RTLR function (Barwell (2017)). For example, consider the following two schemes developed by the European Central Bank:

• Flow credit easing schemes: purchases of specific securities, such as covered bonds or asset-backed securities (ABS), in response to concerns that the elevated risk premia implicit in the funding costs of financial intermediaries were constraining the supply of credit to the real economy. As Praet (2014) explained in the context of the ABS programme, “purchases push down the market spreads”, encouraging banks to originate more ABS and thus create more loans “to ensure an unimpeded transmission of our actions to the ultimate borrowing costs of the private sector”.
• Sovereign bond safety nets: the ECB has developed a couple of asset purchase programmes that were designed to stabilise the peripheral sovereign bond market and prevent financial fragmentation. In particular, the Outright Monetary Transactions (OMT) scheme was justified on the basis that “risk premia that are related to fears of the reversibility of the euro are unacceptable, and they need to be addressed in a fundamental manner” (ECB (2012)) or what Draghi (2013) referred to as the “extraordinary risk premia that markets require when self-fulfilling expectations of catastrophic events prevail”.

One could no doubt make the claim that a lack of market liquidity was a cause for concern in both cases, whether it was the market for ABS or peripheral sovereign bonds. However, the objective of both programmes appeared to go beyond a desire to just squeeze the liquidity risk premia implicit in the securities concerned. Policymakers wanted to lean against other drivers of movements in asset prices – compensation for other sources of risk – that were judged unwarranted and a threat to major public policy goals.

However, if we want to discuss the most significant application of the RTLR principle in the real world then we need to discuss Quantitative Easing (QE), and in particular the QE programmes that were launched during the pandemic.

“Go big and go fast” asset purchases to address market disorder

As Cecchetti and Tucker (2021) observe in their excellent survey of central bank purchase programmes: “Whether something described as ‘QE’ is always, in fact, QE as we define it – with the purpose of directly stimulating aggregate spending – is another matter”. The asset purchases that were implemented during the period of market disorder at the start of the pandemic are typically referred to as QE but whether they should be is definitely another matter.

At the special meeting on March 19 the MPC decided to increase the stock of asset purchases by a staggering £200 billion and to implement the decision “as soon as was operationally possible” (BoE (2020a)). The associated Market Notice underlined that the Bank intended to vary the pace of gilt purchases “in response to market conditions and to support market functioning” and “given prevailing conditions, the Bank intends – at least initially – to make purchases at a materially higher pace than in the recent past” (BoE (2020b)). In short, the Bank of England decided to “go Big” and “go Fast” (Bailey (2020)) in an attempt to calm dysfunctional financial markets. The Bank did not act alone. Other major central banks took similar steps, with the Federal Reserve’s March 23 decisions to both continue buying government bonds and mortgage-backed securities “in the amounts needed to support smooth market functioning and effective transmission of monetary policy to broader financial conditions” and to establish new facilities to support the flow of credit (including purchases of corporate bonds in the primary and secondary markets) likely the most decisive (FRS (2020a, 2020b)).

Returning to the policy debate in the UK, the Minutes of the March 19 meeting noted that “there was little evidence as yet to assess the magnitude of the economic shock from Covid-19” and therefore little evidence as yet to calibrate the appropriate increase in the stock of asset purchases to deliver price stability through the normal channels. Instead, central banks were responding to what Governor Bailey called “an incipient financial stability shock” – the so-called dash for cash. Put another way, central banks had to act as a risk taker of last resort and buy whatever it took to restore functioning markets. Indeed, that seems to be, more or less, how the Bank’s Executive Director for Markets sees it (Hauser (2021)): “Central banks were ‘buyers of last resort’ more than ‘market makers of last resort.”  Critically, in the context of Tucker’s rules of the game for MMLR, Hauser (2021) goes on to note that the classic Bagehot principles were not respected in this period: “Purchases typically took place at prevailing market prices: ie the Bagehot principle was not applied … [although] Arguably this was not true at the very height of the dysfunction, when central banks were buying at prices well below pre-stress levels. But this behaviour was not expected by the market ex ante, and prices of government securities in most markets bounced back rapidly following the initial central bank interventions.”

Members of the MPC have pushed back against the idea that the March 19 decision to go big and fast was outside the scope of monetary policy. For example, Ramsden (2021) challenges the claim that the MPC “somehow broadened its objectives and used a monetary policy tool for financial stability purposes”, arguing that if the Committee had not taken action to calm dysfunctional markets then the macro outlook would have deteriorated and hence the asset purchases can be considered a monetary policy operation. But following this chain of logic one could in theory claim that an unsterilised lender of last resort operation was also monetary policy. After all, an unsterilised LOLR operation would involve the creation of reserves and the macro outlook would also deteriorate if the authorities failed to prevent the collapse of a systemically important financial institution.

The intermediate objectives of any policy intervention surely have some bearing on how those interventions are classified.   Indeed, Cunliffe (2022) draws a clear distinction between the rationale for the 50 basis point rate cut and the £200 billion increase in asset purchases that both took place in March 2020 – the former “to keep demand and supply in line” and the latter “to restore market order”. The March 19 decision looks a lot like an unsterlised RTLR operation that helped to preserve price stability. There was a major injection of reserves into system so perhaps this was monetary policy but it is certainly not conventional monetary policy.

You can’t always count on a coincidence

There are two interesting coincidences regarding the events of March 2020. First, there was a case for monetary easing alongside the need to address the market dysfunction and with rates at the lower bound that implied more QE. Second, the epicentre of the dysfunction was to be found in the government bond market which is where conventional QE operations take place. Everything pointed to central banks printing money and buying government bonds. The question is whether those coincidences will always apply. If they do not then we cannot rely on monetary policy committees always coming to the rescue and acting as a risk taker of last resort when the risks to financial stability demand action.

It does seem likely that central banks will often see a case for monetary easing in those moments where there might be a case for an RTLR intervention. Asset prices will be falling fast, financial conditions will be tightening and that tends to drag on the macro outlook. However, it does not follow that there will always be a monetary policy justification for a large RTLR operation in these circumstances. The central bank might not be close to the lower bound in which case QE would probably not be an active tool unless the news on the inflation outlook is sufficiently large to bring the central bank back to the lower bound. Even if QE is the active tool at the moment of market disorder, it is not obvious that an intervention on the scale required to calm financial markets would be justified by the perceived deterioration in the inflation outlook. Indeed, it is possible that an episode of market dysfunction could even arise when inflation is too high – perhaps because the central bank is forced to tighten aggressively in response, potentially triggering stress in the shadow banking system (Aikman and Barwell (2022)). It seems very hard to believe that the MPC would be able to justify a massive QE programme under that circumstance.

Moreover, it is not at all clear that dysfunction will always emerge in markets where the monetary policy committee is comfortable carrying out asset purchases for price stability purposes. For example, a crisis along the lines of the 1987 stock market crash would present an interesting intellectual challenge to the central banking community: there might be a case for central bank purchases of equities under extreme circumstances, particularly if there was a threat to core market institutions, as there arguably was in 1987, but again it seems more plausible that such a purchase programme would be motivated on financial stability rather than price stability grounds.

In short, it is not clear that the central bank will always find itself in a position where the monetary policy committee can be relied upon to act as a RTLR in the name of preserving price stability. The Bank is alive to this possibility: Ramsden (2021) warns of circumstances “where the MPC would not act to quell market disorder, if doing so would run counter to rather than in support of monetary stability”. It would be better to prepare for that circumstance in advance.

Writing down the rules of the game

There will always be pressure on central banks to act in a crisis. Action today then establishes or reinforces a precedent that central bankers act a certain way (buy assets) in a financial crisis that in turn creates additional pressure on future central bankers to do the same in future crises. The RTLR function appears to have been normalised as a result of the innovations of the past decade. The idea of the ECB buying assets in distressed sovereign bond markets is now mainstream. The idea of central banks going big and fast to combat dysfunction in the government bond market is now mainstream.

To be precise, the scale of the interventions by the global central banking community in March 2020 and the role they played in putting a floor under asset prices will not have been lost on the market. It would not be so unreasonable if the market concluded that this is the normal state of affairs: if problems arise in the shadow banking sector then a torrent of purchases will be forthcoming and we will all chalk it up to another round of QE.

Now that the RTLR function is out in the open it would make sense to put a framework around it to anchor expectations on reality. Of course, credibility is key: there is no point writing down rules today which the politicians and the central banks know that their successors will not be able to respect under pressure in a crisis. The first thing that must happen is a period of reflection. The end point of this discussion should be potential revisions to the remit of the Bank’s policy committees, recognising the RTLR function in much the same way that the MPC’s remit currently includes a section on the use of ‘Unconventional Policy Instruments’.

Fortunately, this debate about asset purchases for financial stability purposes does appear to be taking place within the Bank (see Hauser (2021)). However, there also appears to be a debate about whether the MPC acted as a buyer of last resort – or in the terminology of this article an RTLR – or carried out business as usual in March 2020. This question needs to be resolved because, as we have seen, the authorities cannot count on coincidences. This brings us to two key points that need to be addressed in the process of writing down the rules of the game.

First, the authorities need to establish a robust and inclusive decision-making process around RTLR interventions. Of course, that process must involve the Chancellor, since these operations inevitably put taxpayer funds at risk. And under certain circumstances, the MPC may have an active interest in acting once more as a “buyer of last resort”. But the FPC should also be involved. The circumstances that motivate a RTLR intervention and the considerations that influence the calibration of that intervention belong primarily to macroprudential and not the monetary policy sphere. The intermediate objective is to preserve financial stability, now and in the future. That realisation then guides the discussion over which assets to buy, at what price, with input from whom, and so on. It would make sense if the RTLR function was recognised in the remits of both the FPC and MPC.

Second, the authorities have to create a safe space for the RTLR function that is distinct from monetary policy. There is a good reason why Tucker’s first rule of the game for MMLR back in 2009 was: “central banks cannot engage in MMLR operations if to do so would interfere with monetary policy. Securities are purchased in exchange for cash, entailing an injection of reserves into the banking system. That has to be consistent with the target for the central bank’s net supply of reserves”. The same point applies to the RTLR but one can also argue the reverse: wherever possible, the target for reserves should not interfere with macroprudential policy and the execution of RTLR. Careful thought must be given to how to execute this principle in practice, and the feasibility of the authorities simultaneously issuing treasury bills and purchasing assets in distressed markets on a potentially large scale.

Conclusion

The risk taker of last resort function is now acknowledged implicitly but has seemingly been subsumed into the monetary policy toolkit. That could be a mistake because we cannot count on monetary policymakers always being ready, willing and able to execute the RTLR function in future crises. It would be better if the RTLR function was recognised explicitly so that a framework can be built around this critical element of the crisis management toolkit.