Jonathan Acosta-Smith, Benjamin Guin, Mauricio Salgado-Moreno and Quynh-Anh Vo

Over the past years, a growing consensus has acknowledged the need to construct a ‘system [wherein] every financial decision takes climate change into account‘. While such a system is still far from reality, market participants already produce and demand an increasing amount of climate-related information. Equally, many authorities around the world are considering mandatory climate-related reporting. These developments raise myriad unanswered questions. We focus on the following in a recent working paper:

1. How have voluntary, climate-related disclosures of UK financial institutions changed over time?
2. Can prudential regulators influence current climate-reporting levels just by announcing a future shift to mandatory reporting?

This post summarises the main insights from this paper.

A novel data set on disclosures of financial firms in the UK

In order to answer the above questions, we build a novel data set of voluntary, firm-level climate-related disclosures that are in line with the recommendations of the Task Force on Climate-Related Financial Disclosures (TCFD). These recommendations, published in 2017, are organised into four themes, beneath which are 11 more granular recommendations on the information to be disclosed.

Our approach consists of four consecutive steps summarised in Figure 1.

Figure 1: Machine learning pipeline

First, we manually collect corporate reports of the biggest banks, building societies and insurance companies between 2016 and 2020 in the UK. Second, we extract the information from each page of the reports, identifying potential climate-related pages using natural language processing (NLP) techniques. Third, we reduce the size of our truncated sample by identifying pages that are truly climate-related using a machine learning classifier and a supervised learning approach. Finally, we train 11 independent classifiers to identify the information disclosed on each corporate report’s climate-related page that is closely in line with each of the individual TCFD recommendations.

Our machine-learning approach allows us to measure the extent of voluntary disclosures in several ways. We can simply count the number of corporate report pages that contain disclosed information in line with the 11 TCFD recommendations. We can also construct binary variables indicating if a UK financial firm discloses specific climate-related information in a given year. Lastly, we can also compute the average number of disclosed recommendations within each TCFD specific theme (ie governance, strategy, risk management as well as metrics and targets), or count the total number of recommendations that firms provided information on.

Consequently, our novel data set allows us to obtain a detailed picture of the voluntary, climate-related disclosures in the UK, and its evolution over time.

Evolution and determinants of climate-related disclosures in the UK

Looking at the evolution of climate-related disclosures by UK banking and insurance companies in our data set, we can see an encouraging trend. Not only have these sectors all increased the average number of climate-related pages in their corporate reports since 2016, but they have also published more pages providing greater information in line with the recommendations issued by the TCFD.

Figure 2: Climate-related pages and TCFD recommendations

Figure 3: TCFD recommendations disclosed

Across both banking and insurance sectors, climate reporting in line with TCFD recommendations has been quite similar, as can be seen in Figure 4. While in 2016, only about 30% of institutions in each sector disclosed climate-related information, by 2020 the shares of reporting institutions reached over 70% (at the group level) across the four TCFD themes.

Figure 4: TCFD disclosure themes at the group level

We consider how different characteristics of financial institutions are related to disclosure levels. We observe that institutional size (measured via total assets) seems to matter the most. Figure 5 shows the large positive correlation between institutional size and the number of TCFD recommendations disclosed.

Figure 5: Firm’s size and TCFD disclosures by sector

Background on UK regulatory policy announcements

A general pattern we observe in the figures above is that there is a significant increase in climate-related information published after 2018. There are several potential explanations for this. A first set of potential reasons relate to ‘global‘ factors, while a second set relate to UK ‘internal‘ factors.

Global factors behind this increase in climate reporting could arise both from unusually large and devastating natural disasters (eg, the extreme wildfires in Australia in 2019), and/or from pressure by international climate groups (eg FridaysForFuture and Extinction Rebellion).

However, this increase in climate disclosures could also be driven by UK internal regulatory changes discussed and announced between 2018 and 2020. Over this period, in the UK, the Prudential Regulation Authority (PRA) issued a set of announcements and publications to encourage the management of climate-related financial risks (Table 1).

Table 1: Climate-related policy publication by the PRA

For example, in 2018 the PRA published a consultation paper (CP23/18) proposing that financial institutions under its supervision develop and maintain an appropriate approach to disclosing climate-related financial risks. In 2019, a supervisory statement (SS3/19) mentioned the TCFD framework as an example for financial institutions to consider when developing their approach to climate reporting. And in July 2020, the PRA issued a so-called ‘Dear CEO letter‘ with the explicit purpose of managing expectations by announcing a concrete timeline for financial firms to incorporate climate reporting in line with TCFD recommendations. Concretely, this letter made clear that ‘firms should have fully embedded their approaches to managing climate-related financial risks by the end of 2021’.

The role of policy announcements for climate disclosures

We examine whether any of the aforementioned policy publications affected financial institutions’ climate-related disclosures. Our empirical strategy allows us to analyse how institutions affected by the policy announcements change their climate-related disclosures relative to those not directly affected by the announcements.

In particular, we employ a difference-in-differences approach in the spirit of Bolton and Kacperczyk (2021). We construct the treatment group as the subset of financial institutions whose level of disclosure was below the average level of disclosure when the TCFD recommendations were published in 2017. The intuition behind this approach is as follows: institutions already disclosing sufficient climate information by 2017 – ie our control group – would not have been affected by these policy announcements, because they had already decided to disclose prior to any of the policy publications.

Our regression specification allows us to control for the global factors mentioned before, as we include institution and sector-time fixed effects. Additionally, we further control for other possible characteristics, such as institutional size (total assets), profitability (ROE), and leverage. In other words, our empirical setting allows us to identify the contribution of the internal UK policy announcements on UK financial institutions’ climate-related disclosures.

We find evidence of a statistically significant effect on treated institutions’ decisions to disclose climate-related information across all four TCFD recommendation themes, but only after the 2020 Dear CEO letter. Specifically, we find that those institutions that previously disclosed less catch up in terms of their disclosures after the Dear CEO letter (as can be seen in Figure 6 where the three vertical lines represent publication dates of the three policy communications we consider).

Figure 6: Evolution of two disclosure measures between control and treated firms

Conclusion

In this blog post, we have studied the levels of climate-related disclosures in the UK financial sector and their evolution over time. We use NLP and machine-learning techniques to generate a novel data set that collates the reporting information directly from corporate reports.

We find an increasing trend in climate disclosures across all TCFD themes, and across both banking and insurance sectors. We also find evidence of a significant effect of policy announcements on institutions’ decisions to disclose climate-related information.

Our findings provide some interesting insights for policymakers who are considering mandatory climate-related disclosure. Our results suggest that prior to regulatory interventions, only a fraction of firms disclosed climate-related information in line with TCFD, and these were mostly larger institutions. This gap in voluntary disclosures creates a case for regulatory intervention to encourage smaller institutions to disclose too. Indeed, our results suggest that regulators setting clear timelines for mandatory disclosures can help accelerate the trend, which leads to convergence across institutions.

Jonathan Acosta-Smith works at the OECD, Benjamin Guin works in the Bank’s Strategy and Policy Approach Division, Mauricio Salgado-Moreno works in the Bank’s Monetary and Financial Conditions Division and Quynh-Anh Vo works in the Bank’s Banks Resilience Division.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Sangyup Choi, Tim Willems and Seung Yong Yoo

How does monetary policy really affect the real economy? What kinds of firms or industries are more sensitive to changes in the stance of monetary policy, and through which exact channels? Despite advances in our understanding of the monetary transmission mechanism, existing studies have not reached a consensus regarding the exact mechanics of transmission. In a recently published Staff Working Paper, we aim to contribute to this understanding by analysing the impact of monetary policy on industry-level outcomes across a broad international industry-panel data set, exploiting the notion that different transmission channels are of varying degrees of importance to different industries.

Covering 105 countries and 22 industries from 1973 to 2019, our study combines estimates of monetary policy surprises with industry-level data to identify the industries which are particularly sensitive to changes in monetary policy. Industry-level data are especially informative on the monetary transmission mechanism since factors determining the sensitivity to monetary policy typically vary more across industries within a country than across countries.

New measures of international monetary policy shocks

We compile a comprehensive international data set on monetary policy shocks, which covers not only advanced economies but also numerous emerging market and developing economies. In order to identify the causal impact of monetary policy, it is necessary to disentangle unexpected changes in the stance of monetary policy (also known as monetary policy ‘shocks’ or ‘surprises’) from policy rate movements that occur systematically in response to changes in variables like inflation or growth. Otherwise, it is not clear whether the observed subsequent movement in (say) inflation caused monetary policy to respond, or whether the causality actually runs in the opposite direction (from monetary policy to inflation, which is the direction of causality monetary policy makers are most interested in). Understanding the direction of causality is crucial when it comes to conducting policy counterfactuals (‘what would happen if the central bank increased the interest rate by 50 instead of 25 basis points?’), which is why we are interested in creating a broad database of such shocks.

In many emerging/developing economies, it is difficult to identify unexpected monetary surprises due to data limitations. Consequently, we utilised a hierarchical approach that prioritised surprise measures generated by methods that are deemed superior, according to the following hierarchy: (i) shocks identified by others via high-frequency methods (such as Cesa-Bianchi et al (2020) for the UK) which is seen as the gold standard, (ii) changes in the short-term yield around dates involving monetary policy decisions (the idea being that these changes capture the ‘surprise’ component associated with each monetary policy decision), (iii) the surprise-component implied by interest rate forecasts from Bloomberg’s survey of financial market participants, (iv) deviations from an estimated Taylor rule (a way for the central bank to set its policy rate as a function of inflation and growth), and (v) in case of countries that peg their exchange rate: the estimated monetary policy shock in the anchor country (often the US, for which we conveniently have high-quality shock estimates generated through high-frequency methods).

Figure 1 shows that using the resulting monetary policy shocks in a panel Vector Auto Regression model produces conventional ‘contractionary’ responses in the cyclical components of real GDP and the GDP deflator, giving credence to the underlying shock series that sits at the core of our analysis. These impulse-responses were estimated on data from the 105 countries included in our analysis and can be seen as cross-country averages.

Figure 1: Impulse responses following a positive monetary policy shock

Note: Dashed lines represent the 95% confidence interval.

Test of various theoretical channels of monetary policy transmission

To investigate the transmission of monetary policy, we use our newly constructed monetary surprise data and adopt a ‘difference-in-differences’ approach that interacts monetary surprises with industry-level characteristics – essentially asking whether industries that score higher along a particular dimension are more sensitive to monetary surprises. In particular, we implement this strategy by estimating regressions of the following type (which follows the approach underlying Rajan and Zingales (1998) who used it to estimate the impact of financial development on growth):

Y_i,c,t+1 = α_i,c + α_i,t + α_c,t + β (X_i × MPS_c,t) + ε_i,c,t+1 (1)

In equation (1) the subscript i denotes industries, c countries, and t years. Y_i,c,t is a measure of output growth in industry i, in country c in year t. The variable X_i characterises industry i along eight dimensions (such as external financial dependence, asset tangibility, and durability of output; see Table B); MPS_c,t is our measure of the monetary policy shock for each country c during year t (with positive values indicating monetary contractions). Regression (1) also contains industry-country, industry-time, and country-time fixed effects (α_i,c, α_i,t, and α_c,t, respectively). This constitutes a powerful set of controls (with α_c,t for example controlling for the aggregate state of the economy), reducing any lingering concerns about omitted variables, model misspecification, or reverse causality; the fact that we use monetary policy shocks further helps on this front.

The main object of interest in equation (1) is β, the coefficient on the interaction term (X_i × MPS_c,t). The interpretation of β is akin to a difference in differences approach, which measures the differential impact of monetary contractions in industries with characteristics as proxied by X_i. This coefficient is informative about what type of industries are particularly affected by the monetary policy shock, which is, in turn, informative about the importance of the various transmission channels. When the estimate of β < 0, this means that a monetary contraction (MPS > 0) ends up having a larger negative effect on output growth in industries that score higher along characteristic X.    

This approach enables us to examine four prominent transmission channels that have been identified in the literature, namely: the interest rate channel, the credit channel, the exchange rate channel, and the cost channel (see Table A for a brief description of each channel). The dimensions we investigate are summarised in Table B, along with their predicted effect according to the various transmission channels.

Table A: Description of transmission mechanisms included in our analysis

Table B: Industry-level characteristics and associated theoretical channels

Our results reveal that industries with assets that are more difficult to collateralise (ie, industries with lower asset tangibility, lower investment intensity, greater labour intensity, and higher depreciation) experience a more substantial decline in output in response to an unanticipated monetary contraction, followed by industries that produce durable goods. The latter finding lends support to the interest rate channel (predicting that consumption of durables falls after a monetary tightening), while the former finding highlights the crucial role of financial frictions and the associated credit channel. In particular, our results point to the importance of unsecured financing in the monetary transmission mechanism, with a ‘flight to quality’ (ie, towards secured financing) in downturns. This poses a challenge for models in the spirit of Kiyotaki and Moore (1997), where the liquidation value of secured debt drives/amplifies the business cycle, but provides support for models featuring both secured and unsecured lending (see, eg, Luk and Zheng (2022)).

Conversely, we do not find consistent evidence to support the hypothesis that exporting industries are more vulnerable to monetary tightening. Instead, our findings are more in line with the theory of ‘dominant currency pricing’ where widespread US dollar-invoicing (meaning that trades between countries tend to be denominated in US dollars, even when neither country that is party to the trade uses the dollar as legal tender) implies that exports are relatively insensitive to fluctuations in the exchange rate (with most of the action happening on the side of imports). Additionally, we fail to find evidence to support the cost channel (predicting that prices go up after a rate increase, due to borrowing costs being part of the production process): if anything, we find that relative prices of products produced by industries that are more likely to borrow to satisfy their working capital needs tend to decrease following monetary contractions, thus going against the prediction of the cost channel. This supports the conventional view among policymakers that rate hikes work to reduce inflation, not fuel it (as critics occasionally claim).

Conclusions and policy implications

Our results suggest that the effects of changes in the stance of monetary policy are likely to be heterogeneous, with output in industries producing durables and industries that have lower access to collateral being more responsive. This points to transmission taking place via the interest rate channel and the credit channel. Estimates in our paper also indicate that the credit channel becomes less important as a country’s level of financial development increases. That suggests that, of the channels considered by our paper, it is the interest rate channel (running via durable purchases) that may be most important to a financially developed country like the UK. 

With respect to our results on prices (in particular, the absence of evidence for the cost channel of monetary policy), our results provide support to the conventional view that interest rate hikes work to lower inflation.

Finally, we hope that our database (containing monetary policy shock estimates in over 170 countries, available for downloading) will prove useful to other researchers in answering related or different questions.

Sangyup Choi works at Yonsei University, Tim Willems works in the Bank’s Structural Economics Division and Seung Yong Yoo works at Yale University.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Aidan Saggers, Lukas Alemu and Irina Mnohoghitnei

Decentralised Finance (DeFi) may seem a tempting option for those seeking financial gain, autonomy, and self-governance… But how safe is a world in which ‘code is law’? Closer inspection reveals an ecosystem experiencing several hacks, attacks, and fraud. Estimates show at least US$6.5 billion has been stolen since DeFi’s inception, and one particular DeFi feature is often at the centre of this theft – flash loans. Unlimited, ungoverned, and uncollateralised, flash loans give hackers the toolkit to highly leverage their potential attacks. The only cost is the gas fees required to send the transaction. In this blog post we consider the world of flash loans and their criminal counterpart – flash attacks.

What exactly is a ‘flash loan’?

Flash loans are unlimited uncollateralised loans, in which a user both receives and returns borrowed funds in the same blockchain transaction. Currently they exist exclusively within the DeFi ecosystem. DeFi aims to be an alternative to traditional financial (TradFi), with centralised intermediaries replaced by so-called decentralised code-based protocols. These protocols, based on distributed ledger technology, eliminate, in theory, the need for trust in counterparties and for financial institutions as we know them.

Flash loans are most commonly used for arbitrage opportunities, for example if traders look to quickly profit from a mismatch in cryptoassets’ pricing across markets. Flash loans can also be used for collateral swaps – a technique where a user closes their loan with borrowed funds to immediately open a new loan with a different asset as collateral – or debt-refinancing through ‘interest rate swaps’ from different protocols.

In TradFi, borrowers generally need to go through a due diligence process and, depending on the loan amount, provide a number of documents, including proof of identity, proof of income and, most importantly, collateral. None of this is necessary in the case of a DeFi flash loan.

It is important to understand that the lender is exposed to almost no credit risk when participating in a flash loan, hence collateral is not required. Flash loans leverage smart contracts (code which ensures that funds do not change hands until a specific set of rules are met) and the atomicity of blockchains (either all or none of the transaction occurs) to enable a form of lending that has no traditional equivalents.

Flash loans are therefore only available to the borrower for the short duration of the transaction. Within this brief period, the borrower must request the funds, call on other smart contracts to perform near-instantaneous trades with the loaned capital, and return the funds before the transaction ends. If the funds are returned and all the sub-tasks execute smoothly, the transaction is validated.

In TradFi, collateral is key because it reduces or eliminates the lender’s exposure in a default. However, if the borrower does not repay the flash loan as part of the same transaction in which it was taken out, then the entire transaction gets reverted, including the initial amount borrowed and any other actions that follow. In other words, if the borrower does not repay the flash loan, they never receive the loan in the first place.

A non-refundable fee that covers the operational costs of running the smart contracts must be paid up-front, known as the ‘gas fee’ for the transaction – this is true for any Distributed Ledger Technology transaction and not specific to flash loans. Further commission fees are charged only once the transaction executes successfully, making the whole endeavour nearly ‘risk free’ to both the borrower and lender.

Flash loan features

To better understand flash loans, we analysed the Ethereum blockchain (using Alchemy’s archive node) and gathered every transaction which has utilised the ‘FlashLoan’ smart contract provided by DeFi protocol Aave V1 and V2. The Aave protocol, one of the largest DeFi liquidity providers, popularised flash loans and is often credited with their design. Using this data we were able to gather 60,000 unique transactions from Aave’s flash loan inception through to 2023, letting us take a closer look at this new financial primitive.

In general, the properties of flash loans differ from other DeFi transactions. This is not only because they are near-instantaneous, uncollateralised, and unlimited, but because they tend to be complex, as measured through the number of events or logs emitted during a transaction. This higher complexity contributes to the second distinguishing feature, which is that flash loans typically incur much higher gas fees than standard DeFi transactions, see Figure 2. The more events included in a transaction, the more space it takes on the Ethereum Virtual Machine. Given the uncertain execution of these loans, some users are also willing to pay additional prioritisation fees for their transaction to be included in the most immediate block added.

Keeping those attributes in mind, we used the Aave data set to answer the following questions: Which assets are these flash loans borrowing and why? How complex are these transactions? And how expensive are these transactions compared to the average transaction?

Figure 1: Top five assets borrowed on Aave V1 and V2^[1]

Given flash loans require both price stability and deep liquidity to execute successfully, which assets are most commonly borrowed are not surprising. Figure 1 shows that three stablecoins and the two largest cryptocurrencies, Bitcoin and Ether, make up the top five most borrowed assets.

Figure 2: Distribution of the ratio between the gas fee paid by a flash loan transaction and the average gas fee paid on the same day, for all transactions on the Ethereum blockchain

Source: Etherscan Average Transaction Cost.

What is surprising though, is the outsized cost of flash loan transactions. Figure 2 shows that, on average, flash loans cost roughly 15 times as much as a standard DeFi transaction. As previously mentioned, cost is proportional to the complexity of a transaction, and on this count, flash loans also stand out from typical transactions. Flash loans typically contain between 35–70 logs (Figure 3) per transaction compared to roughly 5–10 logs for the average Aave transaction.  

Figure 3: Count of logs per flash loan transaction 

Flash attacks

Figure 4: Cumulative total exploited vs total value locked in DeFi

Source: DefiLlama.

While giving benefits to some users, the DeFi ecosystem has been exposed to significant attacks, hacks, and fraud, with flash loans a particular vulnerability.

In general, hacks, exploits, or price manipulations implemented using flash loans are dubbed ‘flash attacks’. Flash attacks take advantage of the unregulated, uncollateralised, and near-unlimited capital that flash loans enable to, for example, manipulate crypto markets or exploit platform vulnerabilities and generate profits. To this date over US$6.5 billion dollars’ worth of cryptocurrency has been stolen in attacks directly attributable to flash loans.

Flash attacks are unlike anything we have seen in TradFi because flash loans, and therefore flash attacks, are a function of the underlying DeFi technology. A typical flash attack involves taking out a flash loan to borrow a large amount of crypto from a DeFi platform. Next, these funds might be used to manipulate the price of a particular cryptoasset, or to exploit a vulnerability in the DeFi platform. If the flash attack is successful, then the final step involves repaying the borrowed funds along with any fees due, while keeping the profits. However, should the attack not materialise, then the entire transaction is reversed as if it never happened (bar gas fees). In accordance with the unofficial DeFi ethos that ‘code is law’, some argue that select forms of flash attacks are legitimate, describing them as ‘complex arbitrage’.

Flash attacks can be implemented in a multitude of ways, for example by utilising smart contract code in unintended manners, or to generate and exploit price slippage through oracle manipulation. DefiLlama’s list of known hacks^[2] records the largest DeFi hacks, ranging from rug pulls and re-entrancy attacks to flash attacks. Out of roughly 150 attacks, 45 were supported using flash loans. Additionally, Table A shows that out of the top five largest amounts borrowed via flash loans, four of these were used to attack protocols.

Table A: Top five flash loans by amount borrowed on the Aave protocol

Are flash attacks preventable?

By enabling a whole host of low-risk avenues for attack, flash loans increase the cost to DeFi protocols of securing themselves from cyber threats. Despite that, there are steps which DeFi systems are already starting to take to protect themselves.

One of the simplest attack vectors, price manipulation, could be reduced, to some extent, by employing decentralised pricing oracles. While they are not without faults, these services provide live-pricing data by using a host of independent off-chain sources to validate an exchange rate.

A common approach to minimising code errors or unexpected behaviours is to use audits, which are thorough code reviews undertaken by independent third-party entities. It is important to note that even well-audited protocols have been exploited in the past. Similarly, separate ‘test networks’ called testnets, which replicate the ‘live’ blockchain environment, allow developers to simulate common attack methods and test their protocol’s resilience.

More similar to TradFi, ‘circuit breakers’ can be implemented when suspicious activity is detected. These are similar to TradFi’s trading halts, and have encountered great scepticism in the crypto ecosystem. Further, time-locks could be used to delay the execution of certain transactions, allowing the platform time to respond to potential flash attacks.

Conclusion

From the perspective of those involved in TradFi, flash loans might seem somewhat reality-bending, despite being entirely possible using technology developed within the DeFi ecosystem. Although flash loans and DeFi are in their relative infancy, what is apparent is that while they may service valid uses, they have also enabled some of the biggest thefts in the DeFi space. Whether they will be widely adopted and how they might look in the future remains to be seen.

What are your thoughts? Do flash loans have a place in DeFi? Let us know in the comment section below.

[1] The term ‘wrapped’ describes an interoperable token that mirrors the entire value of the underlying cryptoasset referred to.

[2] This is almost certainly a lower bound for the actual number of attacks.

Aidan Saggersworks works in the Bank’s Foreign Exchange Division, Lukas Alemu works in the Bank’s Current Economic Conditions Division and Irina Mnohoghitnei works in the Bank’s Fintech Hub.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Ambrogio Cesa-Bianchi, Federico Di Pace, Aydan Dogan and Alex Haberis

The recent steep rise in energy prices led to a rise in the price of energy-intensive tradable goods, with inflationary pressures subsequently broadening into services in many economies. Because services are less traded and have little energy input some have suggested this broadening might indicate inflationary pressures becoming more persistent. In this post, we explore the issue through the lens of a stylised two-country model with a tradable and a non-tradable sector. It suggests that following an energy price shock: i) the broadening of inflation from goods to services need not imply more persistent inflationary pressure or changed longer-run expectations, but may reflect one-off adjustments via domestic labour markets; and ii) Inflationary pressures in non-tradable sectors can still have sizable international spillovers.

Our stylised framework

To analyse the issue, we use a dynamic stochastic general equilibrium model with two countries that are connected through trade and financial linkages. We refer to the countries as ‘home’ and ‘foreign’, as in the economics literature.

Four key features of the model are important for our discussion. First, the countries are asymmetric in size, with the foreign economy much larger and relatively more closed than the home economy. Second, both economies are energy importers, where energy is modelled as an exogenous input into production. Third, households in both economies consume domestically produced non-tradable goods (eg theatre) and tradable goods (eg theatre snacks), which may be produced domestically or imported. Fourth, workers can move freely between the tradable and non-tradable sectors and have a degree of market power in their wage setting.

We model the energy shock in a simplified way by considering a global input cost shock that affects both home and foreign tradable sectors. We also assume that the shock is more severe at home than abroad. The shock therefore materialises as an increase in the input cost of snacks, which affects the domestic economy more.

How can an input shock in the tradable sector led to inflation in the non-tradable sector?

To answer this question, we can focus on the impact of the shock on the foreign economy. Because it is large and relatively closed, sectoral spillovers within the foreign economy are largely unaffected by international spillovers and developments in the home economy, so we can abstract from the latter.

The immediate consequences of the shock are to increase tradable goods prices in the foreign economy. This is a result of firms in the traded goods sector attempting to protect their profit margins, which have been squeezed by the increase in input costs.

In turn, households in the foreign economy cut their consumption. Demand for traded goods falls in response of their now higher prices. Demand for non-traded goods falls because households prefer to consume them alongside traded goods: when the price of snacks increases and its demand falls, demand for theatre also falls.

The input cost shock is therefore recessionary abroad.

Turning to the labour market, this plays a key role in generating inflation pressure in the non-traded sector. To understand why, it is important to note that the rise in tradable goods prices reduces the real wages of workers in both sectors. In an attempt to maintain their real incomes, workers use their market power to restrict their labour supply, pushing up on nominal wages. This process may be viewed as a kind of ‘real wage resistance’. Importantly, because wages are common across sectors, non-tradable firms now face higher labour costs. This is what generates the higher inflation in the non-tradable sector.

Monetary policy in this framework is assumed to be credible and brings inflation to target through an increase in nominal rates.

In sum, we can observe rising inflation in sectors not directly affected by the energy shock as a result of a joint labour market and a form of ‘real wage resistance’. Mutually reinforcing price and wage inflation need not, therefore, be a sign of de-anchoring of inflation expectations, which is ruled out by our assumption of rational expectations and credible monetary policy.

Why might inflation for foreign non-tradable goods matter for the domestic economy?

The global input cost shock generates domestic inflationary pressures in the tradable and non-tradable sectors in the home economy, through similar mechanisms as in the foreign economy.

But, in contrast to the foreign economy, open economy considerations play a key role in shaping domestic outcomes. The open economy dimension can be summarised by the bilateral real exchange rate (RER), which is determined by two separate components:

where P^X and P^M are the prices of domestic exports and imports to and from abroad, respectively; and and and and denote the price of tradables and non-tradables in the home and foreign economies.

It is helpful to unpack these components and their effects on the domestic economy in turn.

Starting with the domestic bilateral terms-of-trade (ToT). In response to the shock, this improves (an increase). Note that if we were to explicitly model the third block of commodities exporters (where the global input cost shock for snacks originates), P^M would now include energy prices and hence rise significantly, causing a deterioration in the home aggregate ToT.

The bilateral ToT improvement vis-à-vis the foreign economy reflects our assumption that the global input cost shock for tradable goods hits the home economy more severely: prices for domestically produced snacks increase by more than those produced abroad. All else equal, the ToT improvement is associated with an appreciation of the domestic RER and a deterioration in the domestic trade balance: home consumers switch to the now cheaper imported foreign snacks.

Chart 1a shows in a stylised way the relative supply and demand for foreign traded goods relative to home traded goods. The larger input cost shock for domestic snacks shows up here as a fall in the relative supply of domestic snacks, represented by the inward shift in the relative supply schedule (from the black line to the green dashed line).

Chart 1a: Bilateral terms of trade

Chart 1b: Internal relative prices

Turning to the ratio of internal relative prices. As discussed by Broadbent (2017), alongside the ToT, two further relative prices determine relative demand (and hence the allocation of resources) across countries and between the different types of goods within each country. These are the relative prices between non-tradable (theatre) and tradable goods (snacks) at home and abroad, respectively.

Our assumption that the global input cost shock for tradable goods hits the domestic economy more severely implies that the price of theatre relative to snacks falls by more at home than abroad (although, in absolute terms, all prices are rising). This can be seen in a stylised way in Chart 1b, which shows the demand and supply curves for non-tradables relative to tradables within a particular economy. The shock shows up as an increase in the relative supply of theatre tickets, represented by the outward shift in the relative supply schedules (from the black line to the blue dashed line for the UK; and to the red dashed line for the rest of the world).

All else equal, this movement in relative prices is associated with a depreciation of the RER, which helps to compensate for the loss of competitiveness due to the higher tradable input costs. Domestic consumers shift away from tradables (both domestically produced and imported) to non-tradables, and do so more than foreign consumers (point C versus B in Chart 1b). In our example, the domestic imports of snacks from abroad fall by more than foreign imports of snacks from home (ie domestic exports). As a result, the trade balance improves.

In sum, the overall response of the RER is the result of two opposing mechanisms: (i) a ToT mechanism, which appreciates the RER and leads to a worsening in the trade balance and (ii) an internal relative price effect, which depreciates the RER and leads to an improvement in the trade balance. In our model-based simulations, the ToT effect dominates on impact. The resulting appreciation helps to contain the inflationary pressures of the input cost shock (via lower imported inflation). However, the worsening of the trade balance contributes to a bigger fall in domestic output. 

Conclusions

The global input shock leads to a global recession, widespread global inflationary pressures, and a pickup in nominal wage inflation. Monetary policy authorities, therefore, face a trade-off: a tighter monetary policy stance to stabilise inflation at target needs to be balanced against a shortfall in output.

We show that price dynamics in the non-tradable sector can have important implications for the RER. In the absence of the non-tradable goods sector, the RER would move one-to-one with the ToT, and hence would appreciate by more than in our baseline simulations. A stronger appreciation would imply lower imported inflation but, at the same time, a larger fall in economic activity.

Ambrogio Cesa-Bianchi works in the Bank’s Global Analysis Division, Federico Di Pace works in the Bank’s Research and Structural Policy Team, Aydan Dogan and Alex Haberis work in the Bank’s Global Analysis Division.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Mahmoud Fatouh

Small banks tend to have more specialised business models, likely focusing on commercial and retail banking activities, and show limited interconnectedness to other financial institutions. Hence, they are likely to show less intense cyclical patterns compared to large banks. This post investigates whether large and small banks in the UK and US differ in the cyclical patterns of capital positions and credit provision.

Introduction

Following the Global Financial Crisis, the Basel III reforms introduced stricter capital requirements and reinforced them with cyclical components (the capital conservation buffer and the countercyclical capital buffer). The reforms aimed to ensure that banks have sufficient capital resources to absorb losses and reduce the cyclical effects of bank capital (and regulation) on the supply of bank credit in stress. The stricter and more cycle-sensitive capital requirements should reduce the pro-cyclicality of bank capital positions as they discourage unsustainable credit growth in credit booms, and so limit the need for deleveraging in stress.

Under Basel III reforms, systematically important banks face stricter requirements. Small banks mostly face regulatory requirements like those of larger banks but benefit from some exemptions that differ in scope between jurisdictions. Additionally, regulators in various jurisdictions have been trying to create simpler (but not weaker) regulatory frameworks for small banks. Ideally, a framework for smaller banks would take into consideration their simpler and specialised business models, the limited financial stability implications of their failure, and the disproportional compliance costs they face compared to larger banks. For example, the Bank of England is designing a ‘strong and simple’ regulatory regime for small banks with simpler business models.

Owing to simpler business model and weaker interconnectedness, small banks should be less sensitive to cyclical fluctuations insofar as their credit supply and capital positions should be less pro-cyclical than larger banks. In this post, I investigate this hypothesis and study the implications of stricter cycle-sensitive capital requirements under Basel III for these cyclical patterns.

The data

I use bank-level data and GDP growth for the UK and the US. UK bank-level data comes from a confidential data set at the Bank of England and includes financial data on UK banks between 1990 and 2021. The data for US banks runs from 1979 to 2021 and was collected from multiple sources including Refinitiv® Eikon, Capital IQ®, and published financial statements. GDP data was collected from the Office for National Statistics and FRED® for the UK and US respectively. Table A includes summary statistics of the bank-level and GDP growth data. The aim of having a long time series is to have a sample that covers at least a big portion of a credit cycle before the introduction of the much stricter capital requirements by Basel III standards in 2010. Stricter standards would likely affect the cyclical patterns of bank credit and capital positions, potentially reducing the validity of results. Although capital standards existed before 2010, they were significantly weaker. Such time series can be used to answer two questions. First, analysing years up to 2009 can be used to measure the cyclical patterns in a less regulated environment, providing evidence on whether Basel III needed cyclical components. Second, expanding the analysis beyond 2009 provides evidence on whether Basel III addressed the pro-cyclicality of bank capital and lending.

Table A: Summary statistics

1: UK data (£ millions)

Source: Bank of England internal database.

2: US data (US$ millions)

Sources: Refinitiv Eikon, S&P Capital IQ and published financial statements.

Empirical strategy

In order to measure the cyclicality patterns of capital ratios and total lending of banks and investigate whether they differ between small and large banks, I first categorise banks by size. For the UK, I use internal Bank of England classification of small and large banks. Meanwhile, for US banks, I define small and large banks as those in the lowest 80% and the highest 5% of assets distribution, respectively.

Following Fatouh and Giansante (2023), I measure cyclicality of a variable by the correlation between that variable and GDP growth. I estimate this correlation using the following panel regression:

(1)

where, β_i: bank fixed effect; Y_i,t, capital ratio (equity to total assets) or log of total lending of bank i at time t; X_i,t, a set of bank-level controls, including total assets, capitalisation, and ratios reflecting business model (eg, loans to asset and deposits to liabilities); GDP_t, GDP growth rate at time t.

Analysis

I apply the model in Equation 1 to small and large banks separately at the bank-level to detect differences in the cyclical behaviour of capital ratios and total lending. The results of the regressions are presented in Table B.

As the table shows, capital ratios of large banks were positively correlated with GDP growth in the UK and US before the introduction of Basel III in 2010. On average, a 1 percentage point fall in GDP growth was associated with an 80 basis points and 61 basis points drop in capital ratios of large banks in the UK and US, respectively. Meanwhile, capital ratios of small banks were either not correlated (UK) or negatively correlated (US) with GDP growth.

Total lending of large banks was pro-cyclical pre-Basel III, especially in the UK. On average, a 1 percentage point fall in GDP growth was associated with a 302 basis points and 71 basis points fall in total lending of large banks in the UK and US, respectively. The total lending of small UK banks did not show cyclical patterns. However, the total lending of small US banks was pro-cyclical, but significantly less than that of large banks (11 basis points compared to 71 basis points for each 1 percentage point change in GDP growth).

Table B: Regression results for bank capital ratios and total lending

1: UK banks (1990–2009)

Notes: Coefficient estimates of quarterly capital ratios and total lending of UK banks between 1990 and 2009. Capital ratio is equal to equity to total assets, and total lending is the log of net lending. Small and large banks are defined based on internal Bank of England classifications. Standard errors reported between parentheses, * p<0.10 ** p<0.05 *** p<0.01.

2: US banks (1979–2009)

Notes: Coefficient estimates of quarterly capital ratios and total lending of US banks between 1979 and 2009. Capital ratio is equal to equity to total assets, and total lending is the log of net lending. Small and large banks are those in the lowest 80% and the highest 5% of assets distribution, respectively. To ensure robustness of the results, I also run regressions based on different thresholds. Results of the additional regressions are consistent with the baseline results. Standard errors reported between parentheses, * p<0.10 ** p<0.05 *** p<0.01.

In other words, the capital positions of large banks were more sensitive to economic fluctuations than small banks in both the UK and US prior to Basel III. These trends in capital positions affect the credit supply of banks, depending on their size. Large banks become relatively capital-constrained in downturns, and hence tend to ration lending. The lower cyclicality of small banks’ capital positions allows them to keep their supply of credit steadier over the cycle. Nevertheless, as large banks provide most of bank credit, aggregate credit crunches are expected, especially in deep downturns.

Stricter capital requirements and cyclical components (the capital conservation buffer and the countercyclical buffer) introduced by Basel III should reduce the pro-cyclicality of large banks’ capital positions and supply credit. To investigate this, I re-run the regressions above using data sets that extend beyond 2009. As Table C shows, the coefficient on GDP growth for large banks falls from 80 basis points to 43 basis points (at a lower significance level) for UK banks, and from 61 basis points to 29 basis points for US banks. Results for small banks’ capital ratios using the extended sample are consistent with the baseline in Table B.

Post Basel III, the pro-cyclicality of total lending of large banks fell from 301 basis points to 165 basis points for large UK banks and 71 basis points to 49 basis points for large US banks. The pro-cyclicality of total lending of small US banks fell further (11 basis points to 5 basis points) and remained well below that of large banks.

Table C: Regression results for bank capital ratios and total lending; Basel III impact

1: UK banks (1990–2021)

Notes: Coefficient estimates of quarterly capital ratios and total lending of UK banks between 1990 and 2009. Capital ratio is equal to equity to total assets, and total lending is the log of net lending. Small and large banks are defined based on internal Bank of England classifications. Standard errors reported between parentheses, * p<0.10 ** p<0.05 *** p<0.01.

2: US banks (1979–2020)

Notes: Coefficient estimates of quarterly capital ratios and total lending of US banks between 1979 and 2009. Capital ratio is equal to equity to total assets, and total lending is the log of net lending. Small and large banks are those in the lowest 80% and the highest 5% of assets distribution, respectively. To ensure robustness of the results, I also run regressions based on different thresholds. Results of the additional regressions are consistent with the baseline results. Standard errors reported between parentheses, * p<0.10 ** p<0.05 *** p<0.01.

In summary, the capital positions and credit supply were clearly more pro-cyclical for large banks than small banks. The introduction of more cycle-sensitive capital requirements under Basel III reduced differences between the two groups of banks. As such, it can be argued that the cycle-sensitive components of capital requirements are more effective in reducing the pro-cyclicality of credit supply of large banks (than small banks), as well the aggregate supply of bank credit, reducing the severity of credit crunches in deep downturns.

Summary

This post assesses whether small banks’ total lending and capital ratios show different cyclical patterns from larger banks, and whether the introduction of stricter cycle-sensitive capital requirements under Basel III affects these cyclical patterns. The analysis uses data for small and large banks in the UK and US. The empirical results suggest that prior to Basel III reforms, capital positions and credit supply of large banks were much more pro-cyclical than small banks. The introduction of more cycle-sensitive capital requirements under Basel III reduced capital and credit supply pro-cyclicality for large banks, while having smaller effects for small banks. This suggests that the cycle-sensitive capital requirements are more effective in reducing the pro-cyclicality of credit supply of large banks and reducing severity of credit crunch in deep downturns.

Mahmoud Fatouh works in the Bank’s Prudential Framework Division.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Luke Heath Milsom, Vladimír Pažitka, Isabelle Roland and Dariusz Wójcik

Exports of financial services decline with geographical distance at a rate comparable to that for international trade in goods (eg, Portes and Rey (2005)). This is surprising since there are no transportation costs involved. The consensus is that distance is a proxy for information frictions. We show how cross-border syndication can help overcome information barriers to trade in financial services. We zoom in on the equity underwriting industry where international syndicates reduce information asymmetry between issuers and investors located in different countries.

International linkages between financial institutions are a key feature of global financial market integration. Recent contributions have shed light on their positive effects on international trade in goods (Caballero et al (2018), Claessens and Van Horen (2021)). Surprisingly, little is known about the empirical impact of international linkages on international trade in financial services. We set out to fill this knowledge gap.

Geographical patterns of international equity transactions are heavily influenced by information frictions, which represent barriers that investors face in accessing and interpreting price-relevant information, particularly soft information, about foreign issuers (Sarkissian and Schill (2004), Portes and Rey (2005)). Information frictions can, however, be mitigated by financial and reputational intermediaries, including venture capitalists, equity research analysts, accountants, auditors, law firms and securities underwriters among others (Dunbar (2000), Pollock et al (2004), Ljungqvist et al (2009), Jeon and Ligon (2011)). We focus on the role of equity securities underwriters’ networks, which play a crucial role in enabling firms to access global capital markets.

Cross-border syndication plays an important role in global equity issuance activity. When firms issue equity securities domestically or abroad, the offering can be underwritten by a single underwriter or a so-called underwriting syndicate, a group of financial institutions formed temporarily to sell the new securities to investors. The purpose of underwriting syndicates goes far beyond risk sharing among underwriters. In fact, information production is one of their key purposes. Underwriters involved in a syndicate collectively have a higher potential for information collection and dissemination than any individual syndicate member. Ties formed through international underwriting syndicates therefore facilitate cross-border information flows and mitigate information frictions between issuers and investors located in different countries.

In a recent Bank of England Staff Working Paper (Milsom et al (2023)) we argue that, through this channel, syndication ties promote exports of equity underwriting services, and provide supporting evidence using theory-consistent gravity equations. We construct a comprehensive country-pair panel data set of trade flows by aggregating transaction-level data on revenue flows associated with the underwriting of new issues of equity securities from the Dealogic Equity Capital Market database. Our data set covers 122 countries of origin and 145 countries of destination for the period 2000–15.

Exploiting syndicate structure to measure information flows generated by syndication ties

We build a measure of newly formed ‘core syndication ties’ that proxies for information flows between the trading partners. To do so, we exploit the structure of underwriting syndicates. The lead underwriter is primarily responsible for due diligence, while the other syndicate members mainly sell securities to investors within their networks. In other words, information acquisition is mainly conducted by lead underwriters. Therefore, information flows between the trading partners are strongest when the underwriter in the importing (issuing) or exporting country is the lead underwriter. We denote such syndication ties as ‘core syndication ties’. While information is still expected to flow between the trading partners when the underwriters in both countries are non-lead underwriters, the information content of new syndication ties should be lower. We denote such syndication ties as ‘peripheral syndication ties’. The question we ask is whether newly formed ties increase the flow of underwriting services between trading partners? Therefore, our explanatory variables measure the number of newly connected bank dyads for each pair of countries in each year.

Identifying the impact of newly formed syndication ties on exports

We estimate gravity equations and find that a 1% increase in new core syndication ties increases exports by 0.243%. In other words, doubling the change in the intensity of core ties is associated with a 24.3% increase in exports. In accordance with their lower information content, new peripheral syndication ties have a smaller effect on exports. The effect of a 1% increase in new core syndication ties on exports is roughly seven times larger than the effect of new peripheral ties.

Our aim is to isolate the causal effect of syndicates on exports through information production (supply-side effects). The identification of this causal effect is endangered by reverse causality, whereby underwriters from export markets establish new linkages in import markets in anticipation of an increase in exports due to demand shocks in the importing country (demand-side effects). To rule out reverse causality, we develop an instrumental variable approach which focuses on plausibly exogenous supply-side shocks. Specifically, we measure shifts in the interest of the exporting country’s underwriters’ in each importing country as an export destination unrelated to changes in demand-side prospects in the importing country itself. When instrumented, the effect of new core ties is only slightly smaller: doubling the change in the intensity of core ties is associated with a 22.1% increase in exports.

Syndication mitigates information frictions between issuers and investors

We provide further evidence that syndicates promote exports by mitigating information frictions. First, we show that new core syndication ties matter more when the importing (issuing) country is riskier. The effect of new core syndication linkages on exports is larger when the destination country scores worse on the International Country Risk Guide index, has a worse sovereign credit rating as measured by Moody’s, and performs worse in terms of resolving insolvencies according to the World Bank Doing Business surveys. Second, we show that new core syndication ties also matter more for more information-sensitive transactions, namely initial public offerings as opposed to follow-on offerings and convertible debt.

In conclusion, cross-border syndication is a crucial mode of supply in financial services trade, especially when information frictions between the trading partners are severe. The ability of a country’s underwriters to form international syndication ties is a crucial, but little researched, determinant of a country’s export potential.

Luke Heath Milsom works in the Department of Economics, University of Oxford, Vladimír Pažitka works in the Leeds University Business School, University of Leeds, Isabelle Roland works in the Bank’s Macro-financial Risks Division and Dariusz Wójcik works in the School of Geography and the Environment, University of Oxford.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

John Hillier, Tom Perkins, Ryan Li, Hannah Bloomfield, Josie Lau, Stefan Claus, Paul Harrington, Shane Latchman and David Humphry

In 2022 a sequence of storms (Dudley, Eunice and Franklin) inflicted a variety of hazards on the UK and across Northwest Europe, resulting in £2.5–4.2 billion in insured losses. They dramatically illustrate the potential risk of a ‘perfect storm’ involving correlated hazards that co-occur and combine to exacerbate the total impact. Recent scientific research reinforces the evidence that extreme winds and inland flooding systematically co-occur. By better modelling how this relationship might raise insurers’ capital risk we can more firmly argue that insurers’ model assumptions should account for key dependencies between perils. This will ensure that insurers continue to accurately assess and manage risks in line with their risk appetite, and that capital for solvency purposes remains appropriate.

Introduction

UK insurers use simulated extreme weather events to inform their pricing, manage their accumulation of risk, and decide how much capital they need to operate from both an economic and regulatory viewpoint. Historically, for simplicity, major modes of natural threat were often modelled separately. Yet, different types of adverse scenarios can correlate and occur together. If the correlations are too weak in an insurer’s model, it could lead to under-capitalisation, thereby weakening financial protection for policyholders.

Our work here builds on exploratory work in 2021. It picks on some of the UK’s most spectacular and destructive winter weather, strengthening the evidence that it is important not to neglect the co-occurrence of severely wet and windy conditions. Critically and globally, however, this is but one of numerous correlations that might be under-represented in many insurers’ models.

Wintertime windstorms tend to co-occur with inland flooding on many timeframes

The UK’s two most impactful hazards are extreme wind (including storm surge) and inland flooding. Over nine days, storms Dudley, Eunice and Franklin brought a mixture of damaging winds and inland flooding, snowfall and rain-triggered landslips. Was this an exception, or something to be expected? To shed new light on this question, Bloomfield et al (2023) measured flooding-wind dependency using consistent methods on a range of data sets, which included 240 modelled years of UK Met Office climate projections and historical loss data. They used a spectrum of time frames for correlation (days to seasons) and they modelled river flows rather than just rainfall. The key result is that a correlation of ~70% exists between the hazards of extreme wind and inland flooding (Chart 1).

Chart 1: Plots of wintertime correlation between flooding and extreme wind in Great Britain (GB) and western Europe

Notes: Adapted from our recent scientific study.

(a) The level of correlation in Great Britain between wind hazard and rain (purple), and between wind and river flow (yellow) in the October–March season. Error bands are 95% confidence.

(b) To illustrate a broader context, a map of correlation at a seasonal time frame across Europe, between wind and historical river flows; explore this further in an online tool.

In reading this chart, it is important to recognise that heavy rain does not necessarily result in a dangerously high flow in a river, which in turn does not always convert into flooding. In panel (a) the historically observed losses (2006–18) on Great Britain’s rail network are used as a sense-check on the climate projection results. It is reassuring that the historical loss correlations (black line) are similar to those for river flow and wind (yellow lines). 

Impact on insurers’ solvency requirements more robustly established

Identifying that windstorm events tend to co-occur with inland flooding is one thing. Quantifying a selected potential financial impact to an insurer is another. Taking whole years, we investigated how the level of capital required to remain solvent is affected. Our baseline is a typical commercial situation wherein the perils are assumed to be independent. We used totals of hazard and loss for the UK from two Verisk catastrophe models, one for inland flooding and one for wind and storm surge. The take-home messages are listed after the next two paragraphs, which are for more technical readers.

First, looking at the whole UK market, the choice of method used to join the independent flooding and wind damage events sets was examined. To link total annual hazard severities, copulas (two t-copulas, Gaussian, Gumbel) and a rank-swapping algorithm common in (re)insurance were implemented. Chart 2 shows their effect on joint losses, quantified at a 1-in-200 year return period using the Aggregate Exceedance Probability (AEP) measure. 70% correlation is likely most appropriate (Chart 1), which induces a 10%–12% uplift net of reinsurance. The uplift is enhanced by 1%–2% using a Gumbel copula, which more strongly associates extremes. Alternatively, it is reduced to 7%–10% by a lower correlation (40%), or equivalently to 8%–10% if the ratio of wind to flooding losses exceeds 3:1 (typically c. 2:1). 

In a second analysis, impact on capital was assessed for four selected firms. Outputs are shown in Table A. A Gaussian copula is taken as a ‘best estimate’ because it is in the middle of the range (Chart 2) and best fits the joint distribution of hazard proxies – Site W in Hillier and Dixon (2020). The firms are a representative sample of significant firms with exposure to natural catastrophes. AEP uplift cases a Solvency Capital Requirement (SCR) impact of 2%–4%, depending upon factors such as how well diversified a firm is (eg with man-made catastrophe), and can be raised plausibly to 6%–10% in a stress test that increases the relative influence of natural catastrophes in order to more fully account for the range of firms in the market.

Chart 2: Indicative impact of a correlation between flooding and wind hazards on annual losses for the whole UK market at a 1-in-200 year return period

Notes: Box plots display the distribution created by five types of correlation (eg copula). Pragmatically, reinsurance is applied to events with 1 reinstatement, attaching at 1.5x annual expected loss, exhausting at a 1-in-100 year return period event loss. These are defined on and applied to the joint set of events, but before correlation is considered, and prior to annual aggregation of losses. Gaussian is ‘best’ as it best fits the data of Site W in Hillier and Dixon (2020), displayed in Chart 1b of our previous article.

In summary, two main statements can be drawn from this work, which involves c. 20 million years of statistical simulation:

1. The effect on 1-in-200 year joint net aggregate (AEP) losses is estimated at 10%–12% (Chart 2).
2. This net AEP uplift causes an impact of 2%–4% on firms’ SCR, plausibly up to 6%–10% depending on a firm’s diversification and reinsurance (Table A).

Table A: Indicative impact on firms’ risk capital (top) and appetite (bottom)

Notes: For capital, rows 1–3 show the AEP uplift from wind-flooding correlation propagating into impact on an internal model’s SCR. Four large retail insurers (A–D) illustrate a range of SCR impacts that might arise, with row 4 a stress test to account for less diversified firms. The bottom two rows relate to risk appetite.

We extended our initial analysis by including a greater variety of firms, longer simulation runs, and better constrained scientific inputs. Yet, the headline AEP uplift (~10%) is similar. As such, with results robust to various choices and details of implementation, we believe that a basis for cautiously and carefully incorporating flooding-wind dependency into regulatory tools (eg GIST and CBES) and policy is more solidly established.

Wider implications for risk management and premiums

In addition to solvency considerations, failure to recognise correlations might be detrimental to firms’ risk management. Illustratively, consider a firm writing UK wind and flooding with a risk appetite defined such that surplus capital should be able to withstand a 1-in-10 year for catastrophes. Joint losses assumed to occur every 10 years in a view with no correlation in fact occur every nine years, with the 5%–8% uplift in joint AEP (Table A). Since, for a typical flooding to wind ratio (c. 2:1), the maximum AEP uplift is 13%–17% at a 1-in-50 return period, the effect might actually drive up the frequency of a 1-in-10 year risk threshold defined for all natural catastrophes. Certainly, the 1-in-10 year aggregate AEP will be bigger, so management could think they still have enough headroom to expand their book when they do not. At least, a light touch exercise to scope this possibility might be wise.

Looking more widely, we signpost a recent interesting paper. This also considers inter-peril correlation, but by modifying a scenario used in the Climate Biennial Exploratory Scenarios (CBES), to give insights into the wider implications (eg on necessary future premiums). In other words, the ramifications of hazard co-occurrence are not limited to the thin slices of interest we selected in this blog.

Conclusions and future work

Our main insight from this work is that we can now more firmly argue that insurers’ and reinsurers’ model assumptions should account for key dependencies to allow firms to hold sufficient capital for solvency requirements, price premiums, and to accurately reflect their risk appetite.

A second conclusion is that neither uncertainty (eg in science) nor variability (eg between firms) are sufficient reasons to ignore this message. Thus, in line with climate and weather-related risk more widely, we argue for capability building in both regulators and the wider industry. The market should be responsive to emerging information about risk correlations, whilst not over-reacting. Additionally, there is a potential systemic risk if many firms rely on third-party risk models that omit correlations (ie model uncertainty). So, we specifically highlight a CBES finding, namely that it is good practice for insurers to identify limitations of any third-party models used. Are key correlations captured? If not, what adjustments can address the limitation? Or, what methods need to be developed for insurers to do this? This said, note that overall risk might be reduced by perils in anti-phase (Hillier et al (2020)), which may present the opportunity to actively diversify risk. What constitutes a proportionate response, to provide internal and external comfort, will differ by firm.

Looking into the future, Bloomfield et al (2023) tentatively identify a threefold increase in days where very UK severe flooding and wind co-occur by 2060–80. Results like this justify efforts to understand and jointly model such perils in future climates. A significant benefit of funding scientific hazard research is the possibility of more effectively using of private and public funds in future physical risk mitigation initiatives.

John Hillier works at the University of Loughborough, Tom Perkins, Ryan Li, Stefan Claus and Paul Harrington work in the Bank’s Insurance Division, Hannah Bloomfield works at Newcastle University, Josie Lau and David Humphry work in the Bank’s Insurance Policy Division and Shane Latchman works at Verisk.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Natalie Burr

The challenge of measuring financial conditions

Imagine you were tasked with thinking about how financial conditions have changed over a policy tightening cycle. Different economists would come to very different conclusions, and none would necessarily be wrong. Why? Because measuring financial conditions is challenging – for a variety of reasons. A financial conditions index (FCI) is a common solution, and its advantage lies in the disadvantage of the alternative: it is simpler than making a judgement across a range of individual variables. In this post, I propose one method to create a UK FCI. I find that financial conditions have tightened significantly over the past two years, coming from a period of accommodative conditions following Covid. 

What are financial conditions?

Generally financial conditions are defined as the ease with which households and firms can gain access to financing. But FCIs can also be used as an (imperfect) measure of the effectiveness of the first stage of the monetary transmission mechanism. Imperfect because factors other than monetary policy can also influence financial conditions, but useful nevertheless to assess how effectively changes in policy rates are feeding through to financial markets, such as money market interest rates, credit spreads, asset prices, risk premia and the exchange rate.

How do you measure financial conditions?                                                                            

When it comes to the choice of methodology, there are generally two ‘strands’ of FCIs. Some indices have economic interpretation, where the weights of individual components depend on their historical comovement with macroeconomic variables like GDP, or inflation. Other indices use a data-driven approach, deriving the weights statistically. My new FCI falls within the latter group. The advantage over the first group of indicators is that it does not impose a static relationship between financial conditions and GDP. I argue that while making this connection is helpful, these indices risk trying to achieve too much in just one index. The assumption that the relationship between financial and credit variables, and GDP are constant over time is a strong one. In reality, these relationships can be time-varying, and state-dependent.

Methodology

I offer an FCI for the UK constructed using principal component analysis (PCA), inspired by Angelopoulou et al (2013). Why PCA? The definition of financial conditions can incorporate a broad range of variables. PCA is a dimensionality reduction technique which decomposes the covariance structure of selected series into factors that are common to all, and idiosyncratic noise. PCA is simple and intuitive, allowing to combine a set of explanatory variables that are closely related, optimising the information embodied in the input variables. The choice of methodology was motivated by insights from Arrigoni et al (2020), who find that simpler financial conditions measures perform better than sophisticated statistical methods. 

The index uses monthly data on a range of short and long-term nominal government bond yields, term spreads (the difference between some short and long-term yields), credit spreads (mortgage and unsecured lending), the exchange rate and risky asset prices. The data selection is motivated by the asset prices and credit variables that likely matter for economic activity. Risky asset prices matter due to wealth effects which affect spending decisions, and represent the cost of market-based financing to firms. A wide range of interest rates (including mortgage rates) matter because they represent the cost of borrowing firms and households face. In much the same way, the exchange rate matters as it represents the cost, in £, of other currencies. The question is, what information matters most for aggregate financial conditions? That is what PCA can help with.

First, I prepare the data for analysis. The index contributes to the literature of existing FCIs by addressing the issue of non-stationarity. Stationarity of inputs is important particularly for the FCI to have a reasonable interpretation over a longer period of time. I purge these trends by subtracting an estimate of the long-run equilibrium real interest rate (r*) from interest rates, and applying log differences to equity prices. Long run, structural factors that are not directly relevant for financial conditions have driven these trends, and these would otherwise imply a continuous loosening of financial conditions over time. By detrending, I ensure a reasonable interpretation of the index, which I discuss in the next few paragraphs.

Second, the variables entering the PCA have very different magnitudes and units. Therefore, all variables are standardised by subtracting the mean and dividing by their standard deviation. Lastly, variables are also normalised to account for the way in which the series affect financial conditions. In order for an increase in the FCI to denote a tightening in financial conditions, variables where an increase reflects a loosening in financial conditions enter the model with an inverted sign.

Finally, the index is compiled using weights implied by the first three principal components (which explain almost 80% of the overall variation in the data) on the normalised variables. Chart 1 shows the index, and a decomposition into the contribution from various groups of variables. Financial conditions have tightened sharply over the latest tightening cycle, but moves have retraced somewhat since a spike in September 2022, on account of falling yields and Sterling appreciation.

Chart 1: A new UK financial conditions index

Sources: Bloomberg Finance L.P., Moneyfacts, Refinitiv Eikon from LSEG, Tradeweb and Bank calculations. Latest observation: January 2023.

By scaling the index to be mean-zero, the index should be interpreted as a relative, not an absolute measure of financial conditions. Although an increase (decrease) in the index denotes a tightening (loosening) in financial conditions, it is unclear to what extent they are ‘tight’ or ‘loose’ in absolute terms. In other words, financial conditions above 0 are tight conditions only relative to the historical average.

Similar to existing FCIs, it is not a perfect, or holistic measure of financial conditions. This index is very much a reduced form measure and does not tell us anything about ‘why’, for example, interest rates rise and fall. Individual variables have not been purged of their responses to macroeconomic variables (due to the difficultly of doing this accurately), and therefore the index does not capture pure exogenous shifts in financial conditions. Moves may be endogenous to changes in the macroeconomic environment or monetary policy, and movements in asset prices attributed solely to macroeconomic shocks are not necessarily meaningful changes in financial conditions.

A thought experiment

This index was built on the premise that financial conditions are very hard to measure. Staying true to the belief that one index cannot provide the final word, I test three alternative specifications of the index.

First, I do not allow the weights to be time-varying, so they are not generally robust to changes in the sample. I therefore re-estimate the index only over a post-financial crisis (GFC) sample period, shown in Chart 2. Conditions appear tighter in the aftermath of the GFC, where spreads and a steepening yield curve contribute the most. But since the Brexit referendum in 2016, when the Bank cut interest rates and undertook QE, the index loosens relative to historical experience.

Chart 2: UK FCI estimated over a post-financial crisis sample period

Notes: Solid line represents the post-GFC, dotted line shows Chart 1 version.

Sources: Bloomberg Finance L.P., Moneyfacts, Refinitiv Eikon from LSEG, Tradeweb and Bank calculations. Latest observation: January 2023.

Second, I explore a ‘real’ version of the index, which uses real interest rate variables, exchange rate and equity prices. Chart 3 shows that in real terms, financial conditions have tightened less, reflecting the fact that real interest rates in the latest tightening cycles are still largely in negative territory.

Chart 3: A ‘real’ UK financial conditions index

Sources: Bloomberg Finance L.P., Moneyfacts, Refinitiv Eikon from LSEG, Tradeweb and Bank calculations. Latest observation: December 2022.

Finally, I explore a version of the index that strips out the effects, on both interest rates and equity prices, of international spillovers. I use model-based estimates that are identified through the heteroscedasticity of asset prices, which pin down the geographic origin of the underlying shocks. Comparing the post-GFC FCI (dotted) and the FCI excluding spillovers (solid line) in Chart 4, international spillovers (mainly US and EA) have slowed the relative tightening of UK financial conditions over 2021, but softened the spike in financial conditions around the mini-budget turmoil of September 2022.

Chart 4: A UK financial conditions index, stripping out international spillovers

Notes: Solid line represents the FCI excluding spillovers, dotted line shows the post-GFC version.

Sources: Bloomberg Finance L.P., Moneyfacts, Refinitiv Eikon from LSEG, Tradeweb and Bank calculations. Latest observation: January 2023.

Having looked at a variety of specifications, Chart 5 brings back the link to policymaking, by portraying the relationship between financial conditions and Bank Rate, focusing on tightening cycles in the UK since the Bank’s operational independence in 1997. For each episode, I fit a linear trend to illustrate the relationship. There are reasonable arguments as to why this relationship may not be linear, not least the different pace, speed and size of hikes. But broadly speaking, tightening Bank Rate has been associated with tightening financial conditions (with the exception of the early period of operational independence). And in the latest tightening cycle, each unit of Bank Rate increase had brought a considerable amount of tightening, more so even than in previous cycles. 

Chart 5: Scatter plot of UK financial conditions index against Bank Rate over past tightening cycles

Notes: For the 2021–22 hiking cycle, two observations (denoted by the grey diamonds) were excluded from the estimation of the linear trend. These are observations for September and October 2022, which are likely to be influenced by the mini-budget turmoil in September 2022. The FCI used for this chart is one presented in Chart 1.

Sources: Bloomberg Finance L.P., Moneyfacts, Refinitiv Eikon from LSEG, Tradeweb and Bank calculations. Latest observation: January 2023.

To sum up, financial conditions is a tough concept to capture in just one index. I have argued that FCIs are nevertheless useful, to assess how changes in policy rates transmit to aggregate financial conditions. I find that UK financial conditions have tightened significantly over the most recent tightening cycle, but the degree of tightening is subject to much uncertainty. Robustness checks undertaken by looking at different variations of the FCI demonstrate this. Therefore, it is important to focus on a variety of indices to make a robust and holistic assessment of financial conditions.

Natalie Burr works in the Bank’s External MPC Unit.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Jelle Barkema

How concerned should policymakers be as UK business insolvencies have soared to 60-year highs? This phenomenon has been extensively covered in the media; with media outlets attributing the record-breaking numbers to a ‘perfect storm’ of energy prices, supply-chain disruptions and the cost of living squeeze. Insolvencies are a popular measure of economic distress because they have implications for both the financial system and the real economy. For the financial system, an insolvency generally means creditors will incur losses. Insolvent firms will have to cease trading and lay off workers, which affects the real economy. In this blog post, I assess the evolution of corporate insolvencies over time, including the post-Covid surge to understand what these record numbers mean for the UK economy. 

What is an insolvency?

Let us start with the basics – what is an insolvency? An insolvency occurs when a company can no longer meet its debt obligations. These obligations can be bank loans, but can also include outstanding electricity bills or tax liabilities. A director of a company is obliged to file for insolvency once they realise that their company cannot pay its debts. Hence, most insolvencies are voluntary and instigated by the company itself. These insolvencies are called creditors’ voluntary liquidations (CVLs). In most other cases, the company in question has failed to abide by this obligation and creditors are compelled to go to court and issue a so-called winding-up petition. A judge will then consider the petition, and, if deemed valid, will issue a winding-up order. Following either CVL or a winding-up order, a liquidator will take control of the company and attempt to liquidate its assets – the proceeds of which will be used to repay (some of) the debts. In the remainder of the blog, I will refer to winding-up orders and CVLs as liquidations. Insolvencies, in contrast, will include all insolvency procedures, even those that do not result in liquidation (like administrations).  

Insolvencies over time

In the UK, the liquidation rate, which measures the number of liquidations per 10,000 firms, is cyclical and has followed a clear downward trend. Chart 1a below shows increases in the liquidation rate (orange line) after the early 1990s and 2008 recessions. Overlaying this trend with a line depicting Bank Rate (blue line) shows that the long-term decline in the liquidation rate coincides with a loosening in financing conditions. This is consistent with the likelihood of a firm going insolvent being a function both of the economic environment and the cost of their debt. The literature corroborates this: Liu (2006) finds that interest rates are strong predictors of the liquidation rate in the UK, both in the short and long term. In contrast, a measure of corporate dissolutions since the mid-1980s (Chart 1b, green line), which tracks all company exits (whether they had debt or not), seems more stationary and follows real economy developments – as measured by real GDP growth – more closely. It is important to add that structural changes to the insolvency regime and/or company register also play an important role in determining insolvency and dissolution trends. For example, Liu finds that the 1986 Insolvency Act, which introduced the administration process as an alternative to liquidation, caused a structural downward shift in UK liquidations.

Chart 1a: Corporate liquidation rate and Bank Rate over time

Chart 1b: Inverse real GDP growth and corporate dissolution rate

Sources: Bank of England, Companies House and Insolvency Service.

Note: Liquidation rate equals the number of liquidations per 10,000 firms. Dissolution rate equals the total number of dissolutions divided by the total number of incorporations.

Setting the record straight

So given that Bank Rate was at an all-time low until 2021, how did insolvencies reach an all-time high? Some necessary nuance to this record is that it only pertains to voluntary insolvencies and, importantly, does not account for the growth of the company register over time. The liquidation rate mentioned in the previous paragraph does factor this in and shows the 2021 numbers are nowhere near their all-time maximum. Moreover, insolvencies are only a fraction of all firm exits (4% in 2022) so by themselves are not a reliable gauge of real economy risk.

That is not to say that all is well. UK corporates are facing a unique series of shocks with Covid followed by a sharp increase in energy prices. In addition, financial conditions are tightening faster than they have in decades, making refinancing more challenging and thus insolvency more likely. Business insolvencies can trigger defaults and significant write-offs, which, in theory, could threaten financial stability if occurring in large numbers or in particular sectors of the economy.

Analysing insolvencies at a company-level

To better understand the steep increase in insolvencies and potential financial stability risk, it is helpful to move away from aggregate numbers and to look at insolvencies at a micro-level. I do this by web scraping individual insolvency notices from the Gazette and matching them to company balance sheets obtained through Bureau van Dijk. Having this matched, firm-level data allows us to analyse patterns across insolvency types, sectors, age and size bands.

A first look at the data reveals insolvencies are partially making up for lost ground during the pandemic. Targeted legislation meant that Covid-related insolvencies were temporarily suspended. The suspension of lawful trading rules (targeting CVLs) was in effect from March 2020 until June 2021 while restrictions on winding-up petitions (targeting compulsory insolvencies) remained in place until March 2022. After these measures had been lifted, insolvencies increased rapidly. Chart 2a below demonstrates this clearly: monthly voluntary insolvencies (blue line) fell significantly in 2020, but have since moved past their pre-Covid average, reaching all-time highs. Meanwhile compulsory liquidations (yellow line) were slower to recover but are now surpassing 2019 levels. As of 2022 Q4, the difference between cumulative insolvencies in the 11 quarters before Covid and the 11 quarters since Covid (the ‘insolvency gap’) has almost disappeared.

Chart 2a: Business insolvencies by category (number of insolvencies)

Chart 2b: Business insolvencies by company size (number of insolvencies)

Sources: Insolvency Service, Gazette and Bureau van Dijk.

Note: Micro firms have <£316,000 in total assets, small firms between £316,000 and £5 million, medium firms between £5 million and £18 million, and large firms over £18 million.

Micro firms drive the recent surge in insolvencies

Analysing the post-Covid insolvency surge across company size bands shows that it is largely driven by micro firms – those with less than £316,000 in assets (Chart 2b). In 2022, 81% of insolvencies comprised micro firms, compared to 73% in 2019. This uptick can in part be attributed to timing. The insolvency process tends to be more drawn out for large firms, so it will take longer for the impact of Covid and the energy price rises to be reflected in the statistics. But that is only part of the story. Data from responses to the ONS Business Insights and Conditions Survey (BICS) shows that smaller firms (fewer than 50 employees) consider themselves at a substantially higher risk of insolvency compared to their larger peers (Chart 3a). At the latest wave (ending December 2022), small firms perceived the risk of insolvency to be twice as high. This corresponds with the disproportionate impact of rising energy prices on small businesses (Chart 3b).

Chart 3a: BICS – Business at moderate/severe risk of insolvency (share; by number of employees)

Chart 3b: BICS – Energy prices as main concern (share; by number of employees)

Source: ONS BICS.

Note: Different BICS waves will not necessarily contain the same questions, hence the difference in x-axes between the two charts.

The prevalence of small firms in the insolvency numbers is reassuring from a financial stability perspective; the UK banking sector is well capitalised and exposure to these companies is simply not large enough to present a material risk. Moreover, because of the unprecedented financial support provided during the pandemic in the form of loan schemes, some of this debt will be guaranteed by the government. Indeed, close to 60% of all insolvencies between May 2020 and March 2022 were incurred by firms who had also taken out a Bounce Back Loan. This is also reflected in the company-level data with small firms boasting higher debt levels prior to insolvency compared to pre-Covid (Chart 4). The debt to assets ratio of young firms going insolvent is two times higher in 2022 than it was in 2019.

Chart 4: Indebtedness prior to insolvency by size (total debt/total assets)

Sources: Gazette and Bureau van Dijk.

Sectoral and age distributions remained unchanged

Financial risk could also arise if insolvencies are concentrated in particular parts of the economy. There is no evidence of this so far: the sectoral distribution of insolvencies, for example, looks very similar to 2019 despite the heterogenous impact of the pandemic. One explanation for this is that industries particularly hard hit by the pandemic, like accomodation and food, are also significant beneficiaries of government support schemes. The same goes for the age profile for insolvent firms, which has largely remained the same compared to before the pandemic despite widespread dissolutions among newly incorporated firms.

A succession of macroeconomic shocks has pushed UK business insolvencies to all-time highs. Insolvencies only constitute a small share of all firm dissolutions so it is not an accurate representation of real economy risk. Furthermore, the majority of firms going insolvent are small while exposures are in part government-guaranteed, so I cannot conclude they constitute an imminent financial stability issue either. However, this can change as macroeconomic challenges continue to accumulate, government loan payments become due, financial conditions tighten, and larger, more complex insolvencies start to crystallise. This is definitely a space worth watching.

Jelle Barkema works in the Bank’s Financial Stability Strategy and Risk Division.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Image source: Shutterstock.

Vania Esady

In macroeconomic models, economic agents are often assumed to perfectly observe the current state but in reality they have to infer current conditions (nowcast). Because of information costs, this is not always easy. Information costs are not observable in the data but it can be proxied. A good proxy is disagreement on a near-term forecast because significant disagreement indicates a period when it is difficult to observe the current economic conditions – ie higher information frictions. Suppose people’s ability to nowcast varies over time, this may affect their ability to respond to various shocks, including monetary policy shocks. My recent paper shows that when disagreement is higher, to bring down inflation, a contractionary monetary policy will lead to a greater fall in economic activity.

What does disagreement look like in the data?

Disagreement is time varying. This is a stylised fact that has been captured in many different surveys, ranging from households, firms and professional forecasters, as well as for a variety of variables and a range of different forecast horizons, from nowcasts to 10 year ahead. The intuition behind this stylised fact is that people are not fully informed all the time and this naturally creates heterogeneity in beliefs. People use the information they have to take decisions and they take decisions not only once but repeatedly over time. When making decisions, at each period, people choose whether it is beneficial to reallocate their attention and by how much – making the degree of disagreement change over time.

As a starting point, it is useful to first be familiar with what disagreement looks like in the data. In my paper, disagreement is captured by the interquartile range of real GDP nowcast from the US Survey of Professional Forecasters (SPF). SPF is one of the longest standing macroeconomic surveys, covering a variety of episodes in US macroeconomic history, including important economic events in the 1970s. Professional forecasters are some of the most informed group in the economy, so SPF serves as a conservative benchmark for measuring information costs (information frictions) in their forecasts’ cross sectional variation. If there was an increase of information frictions that reduces a professional forecaster’s ability to predict macroeconomic aggregates – despite all publicly available information and forecasting techniques – then, one could expect there would be even higher information frictions among firms and households.

How do varying degrees of information frictions affect the transmission mechanism of monetary policy?

To answer this question, I estimate state-dependent local projections on US data over the period 1970–2013. Local projections method has recently been applied to study time-varying effects as it can be easily adapted for estimating state-dependent model. This method allows the response of output and inflation to a monetary policy shock to vary depending on how much disagreement there is. Monetary policy shock is identified with a narrative approach à la Romer and Romer (2004) and estimate the shocks in high versus low disagreement periods.

The results show that when disagreement is higher, prices respond more sluggishly in response to monetary shocks. These stickier prices yield a flatter Phillips curve, leading to the empirical result that monetary policy has stronger real (output) effects. During high disagreement periods, output responds fairly quickly to the narrative monetary policy shocks. Conversely, the response of output is muted for a longer period when disagreement is lower. This result arises from the higher stickiness of prices in the high disagreement periods. At its trough, during heightened disagreement prices fall by 0.8% and output by 1%. These results are robust to using forecast (as opposed to nowcast) and inflation (rather than output) disagreement.

Interpreting the empirical results according to rational inattention

To understand why disagreement could be crucial for monetary policy, I design a tractable rational inattention model where nowcasting is costly and firms decide how to optimally allocate their attention.

There is a fast growing literature on designing rational inattention models to understand monetary policy transmission. However, these mechanisms have not been utilised much to explain the empirical evidence of state-dependent monetary transmission. In my paper, rational inattention model gives an interpretation of the empirical results as it demonstrates how price-setting changes with varying information frictions, as well as how it impacts output and inflation. When there are a lot of information frictions, price-setting firms pay less attention to demand conditions. This implies that their prices will respond sluggishly to aggregate monetary shocks. The slower prices respond, the more ‘sticky’ prices appear. These stickier prices lead to smaller price adjustments, but also because of the higher nominal rigidities, it causes a flatter Philips curve, leading to larger output effects.

Dissecting disagreement and uncertainty

While there is a large literature on uncertainty, disagreement has received relatively less attention while possibly being more relevant in studying information effects. A novel insight from model is the ability to distinguish uncertainty versus disagreement, and shows how the two have quite different effects on the monetary transmission mechanism – sometimes they work in same direction, other times quite differently. Uncertainty about demand condition comoves with disagreement when attention on aggregate demand condition is already relatively high, such that paying additional attention may result in a lower marginal benefit. Hence firms do not reallocate more attention to demand conditions, resulting in a rise in disagreement. On the other hand, when the allocated attention on demand is still relatively low, an increase in demand uncertainty raises the benefit to monitoring demand conditions. Firms could then optimally reallocate much more attention to monitoring demand and thus decreases demand disagreement.

Policy implications

For the purpose of monetary policy, my paper shows the important role of central bank communication. During periods of low disagreement, contractionary monetary policy is able to reduce inflation significantly with relatively little output loss. That raises the potentially important role of central bank in communicating aggregate conditions to people, enabling them to internalise contractionary monetary policy, which effectively makes prices more flexible. This leads to a lower sacrifice ratio and enables an inflation-targeting central bank to better achieve its objective.

Vania Esady works in the Bank’s Current Economic Conditions Division.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Nicola Garbarino, Benjamin Guin and Jonathan Lee

5.2 million properties in England are at risk of flooding. To ensure the availability and affordability of flood insurance to households in flood-prone areas, the UK Government introduced an innovative reinsurance scheme, Flood Re, in April 2016. It provides insurers with an option to pass the flood-risk element of their policies on to the reinsurer at a lower fixed price according to property council tax band. In a recently published Staff Working Paper, we employ a granular data set of all property transactions and flood events in England. We estimate the effect of Flood Re on property values. We also explore if Flood Re effects are heterogeneous across different subpopulations in England.

Flood Re reduces insurance premiums for households in flood-risk areas by over half for four out of five properties with previous flood claims. This should increase the values of flood-prone properties. However, in theory, the magnitude of this property price effect depends on the expected reduction in future insurance premiums caused by Flood Re as well as the rate in discounting future insurance premiums. Moreover, the Flood Re effect might vary by area and demographic groups due to differences in flood risk and home ownership. 

Empirical approach to identify the house-price effect of Flood Re

In our research, we examine properties in England. We employ a ‘repeat sales approach’ examining only those 1.5 million properties built before 2002 and sold multiple times. Specifically, we regress the change in property price on their exposure to flooding. In our regression, we interact this flood exposure with a binary variable capturing the time period when Flood Re was operating. The nature of this research design allows us to account for unobservable and time-invariant property characteristics, by examining the effect of floods on property prices both before and after Flood Re came in. To gauge heterogeneous effects of the policy, we then estimate our regression in different subsamples. To that end, we slice our data by property values and different regional characteristics, such as their income level, obtained from the Office of National Statistics.

Average impact of Flood Re on property prices

Chart 1 illustrates the effect of Flood Re on prices of flooded properties. We first run our regression on the full sample of properties in England. We find that flood events reduce property values by more than 1.5% relative to their first transition before the introduction of Flood Re. This negative effect is completely mitigated after the introduction of Flood Re.

Chart 1: Effect of Flood Re on property prices

For an average property, the introduction of Flood Re increases the relative price of properties as a result of flooding by £4,083. We then conduct a back-of-the-envelope calculation among the 5.2 million properties that are at risk of flooding in England. It suggests that the subsidisation of Flood Re increases the total value of flooded properties by £212.3 million per year, assuming only 1% of the at-risk properties are flooded annually. The total effect of Flood Re on property values would double to £424.6 million if flood-risk probability further increases to 2%.

Our results suggest that Flood Re also increases the transaction volumes of properties in at-risk areas. Chart 2 outlines the effects of Flood Re on the transaction volumes of flooded properties. Our results suggest that a flooded property has a 3.6% reduction in the annual probability of transacting before Flood Re came into place. Flood Re more than offsets this negative effect on the transaction probability and increases the transaction of flooded properties.

Chart 2: Effect of Flood Re on market liquidity

Regional heterogeneity in the response of property prices

We also explore heterogeneous effects of Flood Re in different areas across England. To that end, we create subsamples by employing median values of key variables to split the original sample in half. The effect of Flood Re is stronger in urban areas with a higher income and an older population as well as areas with more rental properties. While it would be interesting to identify the channels leading to the heterogeneous effects, our study leaves this question to future research due to the lack of more granular data.

Conclusion and discussion

Our results illustrate two key implications. First, we document heterogeneous Flood Re effects. Interestingly, our results suggest that Flood Re has a weak impact in lower-income areas but a stronger impact in higher-income areas. These results can be seen as evidence hinting at distributional consequences of this reinsurance scheme in terms of wealth among homeowners in England.

Second, the results are relevant for the transition risk of public policy interventions. Flood Re is expected to be phased out in 2039. The flood-risk component of property insurance may therefore be fully priced into premiums by that time. Consequently, the value of properties at flood risk may experience a sudden adjustment, reflecting the increase in current and future premiums. However, the magnitude of the effect that we estimate does not appear large enough to disrupt property and financial markets, but this may change if flood-risk perceptions change in the next decades.

Nicola Garbarino works at the Ludwig Maximilian University of Munich and the Ifo Institute, Benjamin Guin works in the Bank’s Strategy and Policy Approach Division and Jonathan Lee works at the University of Glasgow.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Qun Harris, Ieva Sakalauskaite and Misa Tanaka

After the 2007–08 Global Financial Crisis (GFC), several jurisdictions introduced remuneration regulations for banks with the aim of discouraging excessive risk-taking and short-termism. One such regulation is the bonus cap rule which was first introduced in the European Union (EU) and the United Kingdom (UK) in 2014. This post examines whether the bonus cap mitigates excessive risk-taking and short-termism, both in theory and in practice. It also discusses unintended consequences highlighted by the literature.

Does the bonus cap work – in theory?

So what is the economic case for regulating bankers’ pay? In general, regulation is justified if two conditions are met: first, a market failure is identified, and second, the regulation improves on the market outcome. In the absence of any market failure, there is no case for regulating pay, as firms would offer a compensation package that incentivises their employees to take appropriate levels of risk. A high bonus itself is not an evidence of a market failure. Indeed, some studies (eg Rosen (1981); Gabaix and Landier (2008); Edmans and Gabaix (2016)) explained how both the rise in the level of executive remuneration and the very large levels of compensation for the most senior employees could reflect the efficient outcome of a competitive market for talent against the backdrop of growth, globalisation and technological advances. 

The case for post-GFC remuneration rules was based on the argument that the market-determined pay of bankers incentivised excessive risk-taking and short-termism. This could happen when banks are ‘too big to fail’ (TBTF), or when the deposit insurance premium is mispriced. In order to maximise the implicit subsidy for risk-taking arising from these, banks would incentivise excessive risk-taking by rewarding their employees with a high bonus when their risky bet succeeds, without penalising them when it fails.

The aim of the post-GFC remuneration rules was to rectify this asymmetry in bankers’ reward structure. Some of the UK remuneration rules aim to reduce short-termism and excessive risk-taking in banks by exposing the so-called material risk-takers’ (MRTs’) compensation to losses which may materialise over a longer time horizon. This includes requirements to delay the payment of a part of the bonus (‘deferral’) and pay a proportion of it in bank shares, where deferred bonuses can be withdrawn if adverse circumstances materialise before the deferred bonus is paid out (‘malus’) or even after it is paid out (‘clawback’). By contrast, the bonus cap is supposed to mitigate excessive risk-taking by limiting the reward from risky bets. The bonus cap rule in the EU and the UK restricts the variable pay of MRTs at banks to be no more than 100% of their fixed pay, or 200% with shareholders’ approval. Crucially, the existing bonus cap rule limits the ratio of variable-to-fixed pay, but it does not limit the total pay or total bonus. Thus, the existing bonus cap rule can be justified only if capping the ratio of variable-to-fixed pay can improve on the market outcome.

The theoretical literature on the effectiveness of the bonus cap in preventing excessive risk-taking is mixed. For example, Hakenes and Schnabel (2014) argue that the case for a bonus cap arises when banks have a strong incentive to encourage excessive risk-taking by offering a large bonus, in order to exploit the implicit taxpayer subsidy arising from TBTF. Their analysis, however, assumes that bankers are rewarded in bonus only and so a bonus cap also puts a limit on total reward from risk-taking. It also does not consider the possibility that banks may adjust the pay structure in response to the regulation.

Thanassoulis and Tanaka (2018) also consider the impact of regulating bankers’ pay when banks’ incentives are distorted by TBTF, but they explicitly analyse the possibility that banks adjust the sensitivity of bonus to equity returns in response to regulation. They show that banks can restore excessive risk-taking even in the presence of a clawback rule by offering a bonus which rises more than proportionally with (ie convex in) the equity returns, and that a bonus cap does not prevent this.

Thanassoulis (2012) highlights the unintended consequences of a bonus cap, arguing that it would shift pay from bonuses to fixed salaries, and thereby increase banks’ fixed costs and their probability of failure. This is because in a competitive market for bankers, total pay will be determined by the banker’s ability and the bank’s size.

Does the bonus cap work – in practice?

There is only a handful of empirical studies on the impact of the bonus cap rule. Colonnello et al (2018) examine the impact of the EU bonus cap and find that the risk-adjusted performance of EU banks deteriorated following the introduction of the bonus cap in 2014, possibly because the bonus cap reduced incentive to perform. The paper also looks at how the bonus cap affected bank executives’ turnover, as restrictions on their bonus could lead them to move to non-banks (eg hedge funds) which are not subject to the bonus cap rule. They find that the cap did not impair European banks’ ability to retain their best executives, and that CEO turnover increased only in under-performing banks, possibly due to increased shareholder monitoring.

Colonnello et al (2018) also show that, for those top executives whose variable-to-fixed pay ratio exceeded the bonus cap before its introduction in 2014, fixed pay increased after 2014 so as to keep their total compensation unaffected. These findings have been confirmed by Sakalauskaite and Harris (2022). Using data on a larger number of MRTs in major UK banks between 2014 and 2019, the authors find that the 100% variable-to-fixed pay limit is not binding in practice for most MRTs. Around one third of MRTs in the sample have bonuses exceeding this limit, and there is no clear evidence that getting close to the 100% threshold affects the developments in individuals’ remuneration. However, when an MRT’s bonus ratio gets close to 200%, their fixed pay grows faster while their bonus grows more slowly relative to other MRTs in the subsequent year. Their total remuneration growth does not differ significantly from that of their colleagues whose bonus is not constrained by the bonus cap. These findings are consistent with banks increasing fixed pay to maintain a desired level of total pay for each individual when the bonus cap starts binding. The proportion of MRTs close to the regulatory limits (variable-to-fixed pay ratio of 175%–200%) is however low, at around 4% of MRTs receiving bonuses in a given year.

There is currently no empirical paper which has clearly identified how the bonus cap affects risk-taking of individual MRTs, due to data limitations. In this context, Harris et al (2020) conducted a lab experiment in which participants were asked to undertake investment decisions on behalf of a hypothetical bank, in order to examine how constraints akin to bonus regulations, such as a bonus cap and malus, affect individuals’ risk choice. The bonus cap in this experiment capped the total pay, the total bonus, as well as the bonus-to-fixed pay ratio. When bonus depended on their own investment performance only, participants who were subject to bonus cap and malus took less risks than those who were paid a bonus which was proportional to their investment returns. But when bonus was paid only when their investments outperformed those of their peers, all participants took greater risks and the risk-mitigating effects of bonus cap and malus were significantly weaker.

Conclusions

There is limited support from the existing literature that the bonus cap rule, as it is currently designed, is effective in curbing excessive risk-taking. The theoretical literature suggests that a bonus cap could curb incentives for excessive risk-taking if it caps the total reward from risk-taking, and banks do not adjust other pay parameters in response. However, this is not how the actual bonus cap rule is implemented, as the cap applies to variable pay only.

The theoretical literature also suggests that a bonus cap could be ineffective in mitigating risk-taking given that banks can adjust various pay parameters, and that it can have an unintended effect of driving up fixed pay, thereby increasing banks’ fixed cost and their probability of failure. The evidence based on UK data suggests that banks are prone to increasing fixed pay when the variable pay of an MRT is close to the bonus cap, consistent with the predictions from the theoretical literature. Finally, there is no clear empirical evidence that the bonus cap rule has curbed excessive risk-taking, though data limitations mean that such effects are difficult to identify.

Qun Harris works in the Bank’s Strategy and Policy Approach Division, Ieva Sakalauskaite works in the Bank’s Prudential Framework Division and Misa Tanaka works in the Bank’s Research Hub.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Martina Fazio and Gary Harper

During recessions, and indeed pandemics, housing prices usually fall. Yet between March 2020 and December 2021 (‘the pandemic’), housing prices grew in the UK, reaching at the time their highest growth rate in a decade. During this pandemic, many more people could work from home, which potentially influenced their housing choices. In a recent Financial Stability paper, we analyse how changes in peoples’ preferences might have played into house price growth. We find that about half the growth in housing prices was linked to shifts in preferences. This was mostly due to an increased premium paid for houses over flats, with changes in location preferences only contributing marginally. But other interventions and macroeconomic factors also affected housing price growth.

Changes in the UK housing market trends since the start of the pandemic

To analyse trends in the housing markets during the pandemic, we examine differences in growth rates of housing prices across different regions of the UK, as well as across different types of housing (such as flats versus detached, semi-detached or terraced houses).

First, housing prices continued to grow across all UK regions through the pandemic. But prices grew faster in regions outside London, especially in the North of England, as well as Wales and Scotland (Figure 1).

Figure 1: From the start of the pandemic to the end of 2021, housing prices grew more rapidly in the North and West than in the South and London
Average 12-month housing price growth from March 2020 to December 2021

Sources: ONS and authors’ calculations.

Prices of all housing types grew rapidly after the pandemic started, but house prices grew more quickly than flat prices.

These trends are not completely new: higher growth rates for the prices of houses rather than flats, and in regions other than London has been seen since around 2017 (Chart 1a and 1b). What’s more, they are independent: the slower growth in London is not simply the result of a higher proportion of flats there.

Chart 1a and 1b: The higher housing price growth outside of London and for houses rather than flats pre-dated the pandemic
Year-on-year growth rates of property prices across the UK, by regions and housing types

Sources: ONS and authors’ calculations.

Did household preferences for housing shift during the pandemic?

During the pandemic, the media reported on a ‘race for space’ narrative, suggesting increased demand for bigger houses away from city centres. To check if this theory holds in the data, we recreate a version of the Office for National Statistic’s House Price Index and compute what the growth rate would have been if preferences remained the same as pre-pandemic.

To do this, we combine three data sets giving us information on all housing transactions in England and Wales between 2010 and 2021, including prices paid, detailed property characteristics, as well as features of the Local Authority District where each property is located.

Following the ONS’s methodology, we run ‘hedonic regressions’ on each month of data. These can be used to decompose the price of a property into the value households place on its observable characteristics, such as number of rooms, floor size, property type (flat, terraced, semi or detached house), and whether the property is a new build.

For example, if we compare a flat and a detached house on the same street with very similar characteristics (number of rooms, floor size etc), the detached house usually sells for more as households typically value houses more highly than flats. Similarly, if faced with two identical properties, but one happens to be in the South of England and the other in the North, the property in the South typically sells for more. The ‘race for space’ story suggests that the value that is typically attributed to certain characteristics, such as property type, size and location, may have changed during the pandemic.

We use the coefficients from each monthly regression, together with information on the relevant mix of properties transacted to compute the price of a ‘typical’ property sold in each month. From this we can derive an index and associated growth rate, both of which largely track the ONS’s series. But they are more volatile than the ONS’s, so we smooth our series (Chart 2).

Chart 2: The HPI and growth rates produced by our model broadly tracked those produced by the ONS
HPI and growth rates produced by our analysis compared to the ONS series

Sources: Department for Levelling Up, Housing and Communities, HM Land Registry, ONS and authors’ calculations.

Were changes in housing preference during the pandemic associated with rapid price growth?

According to the ‘race for space’ narrative, households may have increased their willingness to pay for houses outside of busy city centres instead of flats in densely populated areas such as London. Also, some of the growth in housing prices in 2021 could have been mechanically driven by changes in the types of properties households were buying in 2020. This is because the ONS mix adjusts its calculation to take into account household buying habits, but it recalculates the mix every year.

From our calculations, just under 50% of the rapid housing price growth during the pandemic is associated with changes in buying habits, an increased premium paid on houses over flats, and a reduction in the discount on properties outside of London. We have not found a clear pattern for any of the other housing characteristics in our model, such as floor size or number of rooms. And while the premium paid on houses outside of London increased, we do not find a corresponding reduction in the willingness to pay for flats in London, which remained approximately constant throughout the period.

Additionally, the importance of the factors we control for varied over the course of the pandemic (Chart 3). The increased premium on houses relative to flats is the most important and consistent contributor, while the reduced discount on properties outside of London played a smaller role and was only positively correlated with growth rates between January and August 2021. Outside of this period, the discount increased slightly on average across regions compared to pre-pandemic. Alongside these factors, changes in buying habits play a small but consistent role throughout the whole of 2021. 

Chart 3: Our analysis suggests an increased premium for property types other than flats played an important role
Effect of different factors on housing price growth since the pandemic started

Sources: Department for Levelling Up, Housing and Communities, HM Land Registry, ONS and authors’ calculations.

Other factors are likely to have influenced housing price growth during the pandemic

During the pandemic, technology made remote working feasible in many sectors, so more people could work from home. This may have altered both the type of housing that people look for and the total amount they were willing to spend on housing compared to other goods and services. Our analysis is specifically focused on the former and how changes in households’ preferences may affect growth in housing prices. But other factors, some standard and some pandemic-related, also affected housing price growth at the same time.

On the demand side, both the monetary policy environment and government support to household income may have sustained housing demand. On the supply side, construction came to a halt at the start of the pandemic, which may have contributed to price pressures, but this was short-lived. While relevant, all these forces tend to put uniform pressure on housing prices, differently from the relative shifts in pricing across different types of property that we saw in the pandemic.

Some pandemic-specific factors also affected the housing market, in ways that could more easily induce or mimic a shift in preferences. First, due to limited possibilities to spend on recreation, wealthier households in particular accumulated ‘forced’ savings, which they may have channelled into the housing market. Second, government interventions to reduce the rate of Stamp Duty Land Tax directly supported housing demand, with the largest savings accrued for transactions around £500,000. Both these factors are likely to have supported the change in preferences we observed.

We do not think the change in preferences will continue to boost housing price growth forever

This work offers a new framework for analysing the role of changes in preferences for explaining housing price growth. The particular shift we looked at could be transitory, if for example the trend for working from home reverses, causing housing price growth to unwind. But the shift might be structural, for example hybrid forms of work may persist, in which case we would not expect price growth to completely unwind because of a reversal in preferences. That said, we do not expect continued upwards pressure on housing prices from the shift in preferences either. The pressure we saw probably subsided once households adjusted their housing choices by moving. In either case, other, more traditional macroeconomic and business cycle related factors, alongside long-term supply side considerations will continue to drive housing prices well beyond the effects of the pandemic.

Martina Fazio and Gary Harper work in the Bank’s Financial Strategy and Risk Division.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Gerardo Martinez

In 1936, John Maynard Keynes coined the famous term ‘Animal Spirits’ to illustrate how people take decisions based on urges, overlooking the benefits and drawbacks of their actions. To what extent are prices of Environmental, Social and Governance (ESG) assets driven by the sentiment of market participants, as opposed to economic fundamentals? To answer this question, I make use of Natural Language Processing (NLP) tools and an original corpus of tweets to capture market sentiment around climate change. Estimating a factor model, I find that sentiment is associated with immediate returns of climate change related stock indices. These results are stronger for days with the most extreme returns. Market sentiment might be particularly useful in explaining large movements in ESG asset prices.

Up and coming: ESG assets

ESG assets are portfolios of equities and bonds whose underlying companies satisfy environmental, social and governance factors. They represent a fast-growing share of asset management portfolios: according to Bloomberg Intelligence, ESG exchange-traded funds (ETFs) cumulative assets reached over $360 billion in 2021, and that figure is expected to reach $1.3 trillion in 2025.

The growing importance of these assets makes ESG returns and volatility an important object of study. First, we would like to measure to what extent market sentiment around ESG can drive asset prices. And if the effect is significant, ESG assets could act as a trigger or amplifier of stress in financial markets if there was a significant adverse turn in sentiment.

To the best of my knowledge, this post is the first to use a sentiment indicator on climate change, constructed using NLP tools and an original sample of tweets, as an input into models that explain asset returns. I look at three stock market indices designed to measure the performance of companies in global and UK clean energy-related businesses:

• The S&P Global Clean Energy Index (GCEI).
• The FTSE Environmental Opportunities Renewable and Alternative Energy Index (EORE).
• The FTSE Environmental Opportunities UK Index (EOUK).

Chart 1 plots the performance of the indices, which move closely with political events related to climate change policy.

Chart 1: Climate change related stock indices and overall benchmarks (01/01/2016 = 100)

Sources: Bloomberg and author’s calculations.

All about attitude: measuring market sentiment

To construct a measure of market sentiment around climate change, I extract from the Twitter API an original sample of over 700,000 tweets filtered by keywords closely associated with climate change. I restrict my search to English-language tweets posted in the US and UK. I follow a standard pipeline to remove duplicates, clean and pre-process the text of each tweet.

I apply two existing, pre-trained Natural Language Processing tools (FLAIR and VADER) to the resulting data set. Chart 2 shows the resulting counts of tweets, divided into positive and negative sentiment according to VADER. It also shows the average FLAIR score for every day in the sample. The three metrics are normalised using the Z-score.

There is a strong correlation between the three indicators. Spikes in the count of negative and positive tweets track extreme values of the average FLAIR score closely. These extremes are often linked to political developments around climate change.

Chart 2: Measures of market sentiment around climate risk

Source: Author’s calculations.

Linking market sentiment and one-day ahead returns

To assess to what extent market sentiment influences ESG asset returns, I estimate a factor regression which links the return on the ESG indices to the VADER and FLAIR scores, controlling for additional factors. These factors include the price-to-earnings ratio of each index, the difference between 20-year and 30-day government bonds (time-horizon risk), investment-grade corporate bond spreads (confidence risk), and the returns of a benchmark index (the S&P 500 in the case of the S&P GCEI, and the FTSE 100 for the FTSE EORE and FTSE EOUK indices).

Table A shows that the effect of market sentiment on returns is statistically significant, but modest. The effect is especially clear for the FTSE EORE index. A 1 standard deviation increase in the count of positive tweets is associated with an increase in daily EORE returns of 10 basis points. Reversely, a 1 standard deviation increase in the count of negative tweets is associated with a decrease in daily returns of 14 basis points. For comparison, the unconditional standard deviation of EORE returns in the sample is of 76 basis points.

Note that the effect of positive and negative tweet counts is similar, but of opposite signs. This is encouraging, as it is natural to interpret market sentiment as the difference between positive and negative individual sentiment.

The estimated effects on the S&P GCEI are of similar magnitude and direction, although the coefficient on the count of positive tweets is not significant at the 10% significance level. However, I find no significant effect of positive and negative tweet counts on FTSE EOUK returns. One possible explanation is that the FTSE EOUK index captures UK companies. In contrast, the majority of tweets in the sample were located in the US, and thus might not capture sentiment around climate change specific to local UK factors.

Focusing our analysis on the 10% most extreme (highest and lowest) returns yields larger coefficients on the VADER sentiment metrics. For example, on the day of the 2016 US election, I estimate that market sentiment lowered returns for the FTSE EORE and S&P GCEI by around 30 basis points, based on the difference between the negative and positive tweet counts. The regression on the more extreme sample estimates that effect to be of 300 basis points, which would explain 60% and 85% of the observed negative returns respectively.

While FLAIR and VADER scores react to important events, they are likely to contain a significant amount of noise on a day-to-day basis. Adding periods with smaller returns to the sample is likely to add noisy FLAIR and VADER observations, which drives down the regression estimates towards zero.

The opposite happens to FLAIR sentiment scores. Taking the regression results at face value, days with negative market sentiment are associated with higher returns. But on days with extreme returns, the effect of sentiment as measured by FLAIR scores disappears. Given the strong correlation between FLAIR and VADER scores, it is likely that sentiment is captured through the VADER scores, with FLAIR estimates driven mostly by noise.

Table A: Effect of market sentiment on ESG returns

***p < 0.01: coefficient significant at the 1% level **p < 0.05 *** p<0.10

Table A also shows the same set of coefficients, estimated on five-period-ahead returns. No coefficient is statistically significant. This is encouraging: we would expect changes in market sentiment to be quickly included in the information set of investors and for market prices to adjust accordingly.

Market sentiment across time

In order to shed light on the dynamic relationship of ESG returns and market sentiment (as well as the other factors), I run a Vector Autoregression (VAR). I am particularly interested in the pass-through of shocks in the market sentiment indicators to ESG returns. To that effect, Chart 3 plots the variance decomposition of the estimated model for each of the three ESG indices. The variance decomposition is computed over forecast errors over a 20-day horizon, and then averaged for ease of exposition.

The three market sentiment indicators together explain a very small fraction of the forecast error variance. Combined with the results of the regressions for the one-day and five-day returns, these findings suggest that shocks to market sentiment do not explain returns beyond a one-day time horizon. One interpretation is that shocks to market sentiment typically happen around important political events (see Chart 3), and that market participants are able to quickly price in their effects, hence having little effect on returns over a longer horizon.

Chart 3: Variance decomposition, average over 20-day horizon forecast

Source: Author’s calculations.

Conclusions

The results of this analysis suggest that market sentiment on climate change is associated with one-day returns of ESG stock indices. The estimated effect is of modest magnitude, but is especially clear and strong when the analysis is limited to the periods with the most extreme returns. However, it is not universal across all indices and sentiment indicators. And a dynamic analysis shows that exogenous shocks to market sentiment do not explain returns beyond a one-day horizon.

Nevertheless, these results have several implications for financial markets regulators. Firstly, they open the door to enriching models for forecasting asset prices, by including additional inputs such as fundamentals or market sentiment and new tools such as machine learning models. Secondly, regulators will be able to leverage on the novel data set on market sentiment and asset prices to study patters of market reaction to changes in sentiment, such as procyclical asset purchases or asset reallocations.

Gerardo Martinez works in the Bank’s Capital Markets Division.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Kristina Bluwstein, Sudipto Karmakar and David Aikman

Introduction

Inflation reached almost 9% in July 2022, its highest reading since the early 1990s. A large proportion of the working age population will never have experienced such price increases, or the prospect of higher interest rates to bring inflation back under control. In recent years, many commentators have been concerned about risks to financial stability from the prolonged period of low rates, including the possibility of financial institutions searching for yield by taking on riskier debt structures. But what about the opposite case? What financial stability risks do high inflation and increasing interest rates pose?

Maintaining financial stability means looking out for low probability high-impact events like financial crises and devising policies to prevent and mitigate these ‘tail’ risks from materialising. There is no simple method for measuring tail risks – but in recent years researchers have begun exploring an approach called ‘GDP-at-Risk’ as a financial stability metric. The idea in a nutshell is to model the relationship between indicators for the health of the financial system, including the strength of household and corporate balance sheets, and the probability of experiencing a very severe recession. A typical finding is that when the risk appetite in the financial system increases, the risks of a severe recession over the next three years or so also increase.

Our recent research paper present a novel model of GDP-at-Risk. We apply it to answer the question of how a fairly persistent rise in inflation would affect financial stability. Just to emphasise, this is a ‘what if’ scenario rather than the most likely outcome for the economy. 

We find that higher inflation and interest rates increase financial stability risks in the near term, as higher rates put pressure on debt-servicing costs. This in turn means greater risk of ‘debt deleveraging’ by heavily indebted households and firms, who may be forced to reduce their spending in order to meet their debt obligations, potentially amplifying any recessionary effects. There is also a risk of higher loan defaults eroding banks’ equity capital, which could lead banks to tighten lending conditions. However, this effect is small in our model given the size of banks’ capital buffers. Interestingly, financial stability risks actually fall in the medium term, as the increase in Bank Rate allows for greater scope to cut interest rates in any future stress.

A model of GDP-at-Risk

We build a novel macroeconomic model with financial frictions to study the drivers of GDP-at-Risk. The model is grounded in the New Keynesian tradition: inflation dynamics are driven by the output gap and cost push shocks via a Phillips curve; monetary policy works by altering the real interest rate via an IS curve. 

We augment the model to include nonlinearities associated with three occasionally binding constraints: (a) an effective lower bound on interest rates, which reduces the capacity of the central bank to cushion shocks; (b) a bank capital constraint, which creates the potential that banks may restrict lending sharply (ie a credit crunch) when their capital position becomes impaired; and (c) a debt-service constraint, where households and companies deleverage sharply when their debt-service burdens become too large. The model is calibrated to match salient features of the UK economy. 

To characterise tail risk, we focus on the 5th percentile of the GDP distribution. To measure this, we simulate the model a large number of times, sort the predicted GDP outcomes according to their severity, and find the drop in GDP that is only exceeded in 5% of the simulations. This is akin to the concept of ‘value-at-risk’ used in financial risk management. We do this for different forecast horizons and focus especially on GDP-at-Risk at the 3–5 years horizon, as this provides policymakers with sufficient time to recognise risks and apply macroprudential tools to head off any build-ups in vulnerabilities found. 

Non-linearities lead to a fat-tailed GDP distribution

Chart 1 plots the distribution of GDP (relative to trend) from this model. The distribution is asymmetric and has a pronounced left tail. The point (a) is the GDP-at-Risk in our baseline model, while (b) represents the GDP-at-Risk in the linear model. The fat tail reflects the possibility of one or more of the three occasionally binding constraints amplifying the effects of negative shocks, triggering a deep recession. This fragility of the model is absent in standard, linear New Keynesian and Real Business Cycle models, meaning that standard models underestimate the risk of a large recession.

Chart 1: Model implied GDP distribution

In some situations, the constraints in the model interact with one another to make recessions particularly severe – these are the circumstances in the far left-hand tail of the GDP distribution in the chart. For instance, when interest rates are very low, banks are less profitable and find it harder to replenish their equity capital making the financial system prone to bank credit crunch episodes. Similarly, when indebtedness is very high, debt deleveraging episodes will be more common and the deflationary consequences of these episodes makes it more likely that monetary policy will be trapped at the lower bound. 

Inflation and tail risks: a thought experiment

To understand how inflation affects GDP-at-Risk in our model, we perform a thought experiment: we feed in a persistent inflation shock into the model, which results in inflation of 8% at the end of 2022, 5%–6% in 2023–24 and remaining at target through mid-2026. The monetary policy response is modelled very stylistically via a simple Taylor Rule, which responds to inflation by increasing the policy rate substantially in 2023. We then draw other shocks randomly and use these to simulate the model. Given the simplicity of the model and the purely hypothetical assumptions about the path of inflation, this should be viewed as a ‘what if’, illustrative scenario rather than the most likely outcome for the economy. 

The predicted impact of this scenario on GDP-at-Risk is shown in Chart 2, which plots the 5th percentile of GDP in the scenario compared to a baseline where the economy is growing at trend. Overall, high inflation is unambiguously bad news for financial stability risk over the next 2–3 years. The model predicts a significant decline in the 5th percentile of GDP, compared to prevailing conditions, in the next 4–8 quarters. While around half of this would be captured by standard macroeconomic models (dark blue bars), the rest is amplification from the risk of higher interest rates pushing some borrowers’ debt burdens into unsustainable territory leading to abrupt ‘belt tightening’ (green bars). Banks do little to amplify this shock because their capital buffers can absorb the increase in defaults without triggering concerns about their solvency (yellow bars, barely visible). Eventually, by 2025 GDP-at-Risk is back to baseline – and even improved – as these recessionary forces are offset by the benefit of having more monetary policy headroom to cushion other adverse shocks in the future (light blue bars). 

Chart 2: GDP-at-Risk forecast decomposition following a persistent inflation shock

Policy implications

Our model is highly stylised and its quantitative predictions should be treated with caution. There are, however, some insights from this exercise that will be of potential interest to policymakers concerned with addressing financial stability risks in the period ahead. 

First, the banking sector does little to amplify the effects of an inflation shock in our model. This reflects the build-up in capital ratios over the past decade via Basel 3, stress tests and other measures, which means that banks appear resilient to inflationary shocks. Given this, there would be little additional benefit to raising bank capital requirements further in our setting. This channel would matter more, however, if banks’ ‘usable’ capital buffers were smaller than we assume.

Second, our model highlights that the main downside risks from a persistent inflation scenario stem from debt deleveraging by borrowers facing increased debt-servicing costs alongside a broader cost of living squeeze. This is a particular issue given the large outstanding stock of private sector debt. These risks will need to be monitored closely in the period ahead.

Kristina Bluwstein works in the Bank’s Monetary and Financial Conditions Division, Sudipto Karmakar works in the Bank’s Financial Stability Strategy and Projects Division, and David Aikman works at King’s College London.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Gerardo Ferrara, Gerardo Martinez, Pelagia Neocleous, Pierre Ortlieb and Manesh Powar

The Russian invasion of Ukraine in February 2022 and subsequent sanctions led to unprecedented increases in key commodity prices. While prices briefly abated in late spring and early summer, these surged again over late July and August, with EU and UK gas prices reaching new peaks on 26 August. These moves created a sudden and significant demand for liquidity from market participants with derivatives positions. This post examines how non-financial firms (henceforth ‘commodity traders’) reacted to this liquidity pressure, and how their reactions impacted the functioning of commodity derivatives markets. Commodity derivative markets are important for the real economy and the recent events underscored the need to better understand the interdependencies between margin and counterparty risk management practices.

Margin requirements and liquidity pressures

Price moves on some key futures contracts following the invasion were extremely sharp but were surpassed by moves in late August. Dutch TTF natural gas, for example, peaked at eight times the pre-invasion price in March 2022, only to later reach 11 times the pre-invasion price in late August. As of early November 2022, natural gas prices in Europe have significantly retraced and are more in line with pre-invasion levels, but remain volatile, and are still far above levels typically seen over the 2010s.

As a result of this elevated volatility and sharp changes in prices, central counterparties (CCPs) called for more initial and variation margin to cover short derivatives positions in commodities. Variation margin calls must be met with cash, while initial margin calls can be met with cash or a range of eligible securities (usually government bonds). Notably, initial margin rates on natural gas in ICE Clear Europe increased sixfold from January to April 2022; they have remained elevated since. Chart 1 shows the margin rate – which is a base level of required initial margin for a given contract – for the front-end (shortest available maturity) futures contract for key commodities. On top of this, some CCP clearing members applied margin add-ons or multipliers when sending the margin calls to their clients to reflect their credit risks.

Chart 1: Prices and margin scanning range of front-end futures: TTF gas and Brent oil

Notes: Scanning ranges are representative of initial margin requirements on a single contract of the product. Levels are reported in monetary units per unit of underlying. The depicted Dutch TTF futures are traded in EDX-ICE Endex and Brent futures in ICE Futures Europe Commodities. Vertical dashed line marks the start of the Russian invasion of Ukraine on 24 February 2022.

Sources: Bloomberg, ICE Clear Europe and Bank calculations.

When prices are rising, variation margin is paid by those who have short positions; in commodity derivatives, non-financial institutions such as commodity traders, producers, and end-suppliers typically hold structural short positions as they seek to protect the portfolio against price drops during future sales. These short positions are offset by commodity users (eg airlines) and financial players, like dealers and asset managers. As such, it is predominantly non-financial corporates, such as energy firms or commodity traders, who would have been required to pay margin during the commodities stresses of 2022. In principle they could sell assets to meet margin demands. However, sale of the corresponding physical product can take anywhere between 20 and 90 days, depending on the commodity in question. This means there is a timing mismatch between the margin calls they are required to pay on their derivative positions, and the time required to sell their actual physical products. The various cash flow needs of non-financial commodity traders are illustrated in Figure 1.

Figure 1: Cash flow needs of non-financial commodity trader

This timing mismatch between realised cash flows has been at the crux of recent liquidity pressure on commodity firms, but it is not a new problem. For example, in 1993, German industrial conglomerate Metallgesellschaft required a liquidity injection from a group of banks after sudden changes in the shape of the oil curve left it facing large hedge accounting losses and margin claims.

Liquidity management and market incentives

A spike in volatility like that observed in March–April 2022 interacted with these inherent features of traders’ business models to produce changes in market behaviours. While some of these were expected, some were not, and raise important questions for policymakers.

First, differences in liquidity management between non-financial and financial corporations strained the functioning of physical commodities markets. Comparing non-bank financial institutions and commodity traders is instructive. During the March 2020 ‘dash for cash’, non-bank financial institutions sold government bonds or used them as collateral for borrowing in repo markets, in order to access immediate liquidity. The liquidation of safe assets, coupled with limits in dealer intermediation capacity, led to exacerbated volatility and dysfunction in core financial markets.

In contrast, non-financials such as commodity traders hold little to no investments in securities (such as government bonds) on which they might draw during stress. For this reason, in normal times, commodity traders rely on credit lines (such as revolving credit facilities) with banks to meet their liquidity needs. Following the steep and sudden rise in margin calls, commodity traders have primarily sought to increase borrowing capacity of their credit lines. This worked to a certain extent; however, as banks’ risk appetite in commodities markets tightened, some traders sought credit elsewhere (such as from private equity) or reduced their hedging activities altogether. This may leave them unable to capture the benefits of hedging forward future production and vulnerable to sharp swings in the prices. Or lead to a reduction of physical supply should an inability to hedge lead participants to exit the market. Both enhance pass-through of price shocks and amplify impacts to the real economy.

Second, market liquidity and trading volumes changed in unexpected ways as the shock evolved. In theory, differences in the reactivity of margins models to volatility increases should incentivise a shift from centrally to non-centrally cleared over-the-counter (OTC) derivative markets in times of stress. Initial margin requirements on non-centrally cleared transactions (such as those calculated via eg the ISDA Standard Initial Margin Model) are generally less reactive to increases in market volatility and therefore less costly for investors. For centrally cleared markets, a reduction in activity was visible in available data: Chart 2 shows open interest in TTF one-month futures falling more than 40% from pre-invasion levels. Significant drops are also seen in other markets such as Brent oil and aluminium.

Chart 2: Open interest of generic first futures: TTF gas, Brent oil, aluminium and wheat

Notes: Open interest is defined as the number of outstanding agreements for the selected contracts. Open interest belongs to contracts traded at the Shanghai Futures Exchange for aluminium, Chicago Board of Trade for wheat, ICE Futures Europe for Brent crude oil and EDX-ICE Endex for TTF gas. Vertical dashed line marks the start of the Russian invasion of Ukraine on the 24 February 2022.

Sources: Bloomberg and Bank calculations.

However, the non-centrally cleared OTC activity in UK markets also declined sharply. Chart 3 shows a 50% decline in OTC trading activity between June 2021 and June 2022. It should be noted that this is only a partial cut of data relative to the European Gas (Netherlands TTF), and that other jurisdictions may have observed different patterns. One reason for the decline is the tightening of risk appetite: at the first peak of the crisis in March, there was less willingness to enter into OTC trades with commodity traders due to counterparty credit risk concerns, particularly less well-rated ones. Further complicating the incentives at play is the fact that multiple factors other than margin width, influence the choice of which the derivative markets have to use for their trades (eg portfolio netting, temporary amendment of collateral requirements, etc). This is consistent too with the observation that, for a given market structure, reactions of market participants and banks can amplify volatility shock.

Chart 3: Total volume of European Gas (Netherlands TTF) in MWh of centrally and non-centrally cleared OTC GAS derivative contracts (millions)

Sources: London Energy Broker’s Association data and Bank calculations.

Policy implications

In addition to implications for markets and liquidity management, this episode has underscored three important lessons for future policy work.

First, given that initial margin is designed to cover potential future losses in the event of a counterparty default, it is natural for levels to increase as volatility increases. However, these sudden changes can cause liquidity stress whereby parties posting margin have to find additional liquid resources, often at just the times when it is most difficult to do so.

Second, changes in investor behaviour and trading activity have further underscored structural issues in commodity derivatives markets and margining practices. In theory, differences in the reactivity of margins mean that liquidity could shift from centrally to non-centrally cleared OTC derivative markets in times of stress. However, other aspects of margin models, and differences in margin practices, including waivers and multipliers, or eligible collateral may have played a bigger role in incentivising behavioural shifts. A review of margining practices by the Basel Committee on Banking Supervision, the Committee on Payments and Market Infrastructures and the International Organization of Securities Commissions looked at margin practices during the Covid ‘dash for cash’ in March and April 2020. The report recommended increasing transparency of margin practices in centrally cleared markets and evaluating the responsiveness of CCPs’ initial margin models to different market scenarios.

Third, the episode has raised questions about the different liquidity needs and strategies of financial and non-financial corporations. The latter do not hold liquidity buffers in the same way the former typically do yet are subject to similar liquidity demands during times of stress via margin calls on derivatives used for hedging. This episode should prompt more careful thought about non-financial firms’ use of derivatives, their liquidity resilience, and their interconnection with the real economy.

In response to the extreme liquidity pressures energy companies face as a result of steep margin calls, the Bank and HM Treasury launched the Energy Markets Financing Scheme on 17 October 2022. It seeks to quell one of the main dynamics outlined above by enabling the provision of short-term financial support to energy firms of good credit quality for the purpose of meeting collateral requirements that arise due to hedging activity.

Gerardo Ferrara, Gerardo Martinez, Pelagia Neocleous, Pierre Ortlieb and Manesh Powar work in the Bank’s Capital Markets Division.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

As is our annual tradition, we take a look back and present you once more with our most popular posts of the year. In case you missed any of them the first time round, the five most viewed posts for 2022 were:

• Predicting exchange rates
• Cryptoassets, the metaverse and systemic risk
• Capitalising climate risks: what are we weighting for?
• The impact of shipping costs and inflation
• Old problems with new assets: some of crypto’s challenges look strangely familiar

We hope you enjoyed the blog in 2022. Happy New Year and we look forward to you reading our posts in 2023!

Philip Schnattinger, Managing Editor

Ieva Sakalauskaite and Qun Harris

Following the Global Financial Crisis of 2007–08, some regulators introduced rules on bankers’ bonuses with an aim to mitigate incentives to take excessive risks, and in turn promote financial stability. In a recent paper we use detailed data on remuneration of staff in six large UK banks to look at how two of those rules – the bonus cap and deferral – affected bankers’ pay. We find that the bonus cap did not reduce bankers’ total remuneration but rather shifted it from the variable to the fixed part of the package. And while requirements to defer bonus pay can be expected to affect bankers’ risk-taking incentives, we find some evidence that they increased their total compensation.

There is broad consensus that bankers’ remuneration packages contributed to the Global Financial Crisis because they created a reward structure which encouraged excessive risk-taking: giving bankers a large share in upside rewards, but smaller and more limited exposure to the downside. Following the crisis, regulators introduced remuneration requirements aiming to remedy this. First, they aimed to better align bankers’ incentives with longer-term bank performance through requirements such as deferral, payment in shares, malus or clawback. Furthermore, in some countries (EU and UK) regulators attempted to reduce excessive risk-taking by imposing a bonus cap.

Ten years since their implementation, evidence on the effects of remuneration rules on pay or behaviours in banks is still limited. Some evidence suggests that when they were introduced, some remuneration requirements were associated with a reduction in banks’ risk. But other researchers found that the bonus cap decreased affected banks’ performance at its introduction in 2014 – potentially by reducing bankers’ effort. Furthermore, for top executives most affected, pay shifted from bonuses to salaries.

In this paper we use detailed data on remuneration in UK banks to study two specific regulations. First, the bonus cap, introduced in the UK in 2014, which limits the ratio of bankers’ variable remuneration (comprising mostly of bonuses) to fixed pay (comprising of salaries, role-based allowances and benefits) to a maximum of 100%, or 200% with shareholders’ approval. At the time of its introduction, it was expected that the cap would limit the gains bankers could achieve through excessive risk-taking, in turn reducing such incentives. In our paper, we study whether and how the bonus cap affected bankers’ pay sizes and structures.

Second, we study deferral rules which require that key risk-takers in banks (material risk-takers (MRTs)) receive a proportion (40% or 60% depending on seniority) of their bonuses with a delay. The purpose of this regulation is to increase bankers’ accountability by allowing part of their bonuses to be received only once the longer-term effects of their decisions and banks’ performance have emerged. Deferral periods are set to be sufficiently long to reflect the timescale over which problems come home to roost in banking. Because of discounting effects, deferring bonuses for several years reduces their net present value. Therefore, economic theory predicts that if banks wished to defer a proportion of bankers’ pay, they would need to ‘compensate’ by paying them more. As deferred bonuses are subject to downside risks in the form of a malus and a proportion of them being paid in shares, the risk to bankers, and the need to compensate them, could be even higher. We test this theoretical prediction by looking at whether longer deferral requirements introduced by the Prudential Regulation Authority in 2016 were associated with increases in pay.

The data

To explore the effects of remuneration rules, we use regulatory data on the sizes and structures of MRT remuneration in six major UK banks during 2014–19. MRTs are individuals whom banks identify as having scope to take decisions that can materially affect the risk profile and soundness of their banks due to seniority, ability to create large exposures, and other criteria. Our data covers information on the sizes of their variable and fixed remuneration, the proportion of bonuses deferred as well as deferral periods (in years). Based on data available, we focus on MRTs whose remuneration we can follow for at least three consecutive years; this gives us around 60% coverage across the banks observed.

What does the data tell us about the bonus cap?

To identify the effects of the bonus cap on MRTs’ pay, we look at how individuals’ bonuses, fixed pay, and total pay grow in the year after they reach a bonus-to-fixed pay ratio close to the bonus cap threshold (ie 200%) as compared to colleagues who are further away from it during 2014–19. In particular, we regress individual MRTs’ year-on-year remuneration component growth (in %) on a dummy variable equal to one if in the previous year, that MRT’s bonus to fixed pay ratio was between 175%–200% of fixed pay, ie close to the bonus cap limit.

As individuals with higher bonus ratios are arguably different from their colleagues and experience different year-on-year developments in pay sizes and structures, we account for MRTs’ previous period bonus amounts and the bonus ratio itself. This means that our analysis attempts to capture the additional effect of one’s remuneration being close to the regulatory limit. Furthermore, we factor in shocks that hit each of our sample banks every year.

We find that when an MRT’s bonus ratio got close to 200% (being in the 175%–200% bonus/fixed-pay range), the following year their fixed pay grew much faster than that of other MRTs. We find such effects both when we use data on all MRTs, or only the closer comparison group with ratios already exceeding 100%. We do not find statistically significant evidence that affected MRTs’ total remuneration decreased, consistent with bonuses being replaced by higher fixed pay. These effects are seen on a relatively narrow cohort of employees: on average, only around 4% of MRTs had their bonuses between 175%–200% of fixed pay throughout the sample period.

These effects are illustrated in Chart 1 which plots average remuneration growth figures for MRTs depending on their bonus/fixed pay ratios. It shows that although bankers with higher bonus ratios have overall tended to experience higher fixed-pay growth and lower bonus growth the year after, there is a discontinuity in this trend for individuals closest to the 200% limit.

Chart 1: Average year-on-year (%) growth in MRT remuneration components depending on their previous year bonus/fixed-pay ratios

Notes: This chart plots average year-on-year growth figures for material risk-takers’ fixed pay (salaries, role-based allowances, and other), bonuses (including both bonuses and other components of variable pay), and total pay according to their initial bonus to fixed-pay ratio bucket (from 0% to 200% variable to fixed-pay ratios, at 25% increments). Analysis is based on data in six major UK banks through 2014–19 for material risk-takers which we could track for at least three consecutive years.

What have we learned about deferral requirements?

To measure the effects that bonus deferral has on MRTs’ pay sizes, in particular whether they are compensated for delayed earning, we study the consequences of a change in UK remuneration requirements implemented in 2016. Until then, all largest UK bank MRTs faced the same requirement where at least 40% (or 60%) of bonuses needed to be deferred for at least ‘three to five years’, with banks generally setting it at the minimum of three years for most individuals. In 2016, regulators increased minimum bonus deferral periods to five or seven years for some senior MRTs, and kept the minimum unchanged for the rest. As a result, this policy change affected only some individuals in each sample bank, which allows us to comparing developments in affected and unaffected MRTs’ pay around the time of the rule change.

Specifically we implement difference-in-difference analysis, regressing MRTs’ pay sizes during 2014–17 on (i) a dummy variable equal to one for all MRTs in the years after the rule change (2016–17), and (ii) its interaction with a dummy variable equal to one for MRTs who were affected by the rule change. Whereas the first dummy variable measures how all MRTs’ pay changed in 2016/2017, the interaction term captures how this differed for MRTs affected by longer deferral requirements.

Our results show that the total remuneration of MRTs affected by the rule change indeed increased more than that of the unaffected MRTs around 2016. This is consistent with affected MRTs being compensated for the longer periods over which their bonuses were deferred. These findings are subject to several caveats which do not allow us to definitely conclude the changes were the sole result of the change in deferral regulation – for example, as MRTs affected by the rule change tended to be more senior than those unaffected. Nevertheless, they provide some support for the theoretical point that deferring bankers’ pay could lead to them being compensated via increases in total remuneration. We also find that around 2016, the proportion of affected MRTs’ bonuses deferred voluntarily beyond regulatory minima diminished more than that of unaffected MRTs, consistent with banks trying to reduce the impact of the rule by decreasing the share of bankers’ pay exposed to it.

Conclusions

We have sought to shed some light on how two specific post-crisis rules affected remuneration in UK banks.

We do not find evidence that the bonus cap significantly constrained MRTs’ total pay growth, but rather led to slower bonus and faster fixed pay growth. Our findings also give some support to theoretical predictions that deferring individuals’ pay might mean they need to be compensated for postponed consumption.

But this analysis does not establish whether these costs outweigh the benefits of those rules, which is beyond the scope of our work.

Ieva Sakalauskaite and Qun Harris work in the Bank’s Prudential Policy Division.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.

Before Bank Underground takes a festive break, it’s time for the Annual Bank Underground Christmas Quiz. Test yourself against our 10 festively themed questions on economics, finance and all things central banking…

Daniel Christen and Nicola Shadbolt

Geoeconomic fragmentation is one of the greatest risks to the international monetary and financial system at present, particularly since Russia’s war of aggression against Ukraine. Fragmentation is likely to have wide-ranging implications for the global economy, including increasing the volatility of capital flows and exposing gaps in the global financial safety net (GFSN). In this post, we argue that increased take up of the IMF’s ‘precautionary facilities’ would reinforce the GFSN and help prepare it for these challenges. The IMF’s upcoming review of precautionary facilities is an opportune moment to find ways to reduce stigma and increase uptake.

What is the GFSN and is it big enough? 

The GFSN consists of a set of financial instruments and institutions that act as insurance for countries facing sudden stops in capital flows. It includes countries’ foreign reserves, Regional Financing Arrangements (RFAs), central bank swap lines, and IMF lending.

It proved resilient to the Covid shock, albeit with the help of a large injection of liquidity from the Federal Reserve, and a fresh allocation of IMF Special Drawing Rights, which boosted global liquidity further.

However, there are three key emerging risks that may result in it proving too small relative to demand in future. First, it may be more challenging for major central banks to extend liquidity on a large scale to mitigate an unwarranted tightening in global financial conditions while domestic monetary policy is in a tightening cycle. Second, continued growth in external liabilities, especially those intermediated by the non-bank sector, could increase the likelihood and size of sudden stop episodes.

Third, geoeconomic fragmentation, caused by tensions over security, trade and investment, could increase economic volatility, thereby increasing demands on the GFSN. It could also weaken the safety net itself, if, for example, geopolitical splits jeopardise existing RFAs and swap lines.

Fragmentation could play out in different ways. Since the global financial crisis, cross-border investment and trade have been shrinking as a share of world GDP, a phenomenon referred to as ‘slowbalisation’, partly reflecting geopolitical tensions. Meanwhile, Covid led to calls to reduce economic dependence on geopolitical rivals. Most recently, increased emphasis on national security and political values in economic relations – particularly following the start of Russia’s war of aggression against Ukraine – risks fragmenting the global economy into blocs.

To see how this might impact the global economy, we explore an illustrative scenario in which the global economy splits into blocs based on the votes cast on the UN General Assembly Resolution condemning Russia’s invasion of Ukraine. The emergence of these blocs could entail restrictions on trade and financial flows across blocs. We identify around US$12 trillion of ‘cross-bloc’ financial liabilities – roughly 13% of global external liabilities (excluding intra euro-area liabilities) – which might be at risk of disruptive unwinds (Chart A).    

Chart A: ‘Cross-bloc‘ financial liabilities

Sources: BIS International Banking Statistics, IMF Coordinated Portfolio Investment Survey and IMF Coordinated Direct Investment Survey.

Notes: Bloc 1 assets are those held by countries who condemned Russia’s invasion, and are located in Bloc 2, which includes all countries who abstained or voted against the UN resolution on Ukraine. The opposite is true for Bloc 2 assets.

Taken together, these emerging risks point to the need for countries to seek greater insurance.

How can we strengthen the GFSN efficiently?

The most efficient way for countries to insure themselves is through pooling resources globally. The IMF, with its near-universal membership and mandate to promote and protect global macroeconomic stability, is the obvious vehicle for this. Accumulating large reserve positions individually is costly for the holder and, collectively, for the global economy. It has been identified as one factor that has contributed to global current account imbalances, which can have significant negative spillover effects, as well as encouraging risky ‘search for yield’ behaviour, by compressing global yields.

IMF resources are limited, however, and so it is important they are used efficiently. An increase in their use on a precautionary basis is one way to do this.

Having access to a precautionary line means countries that face external risks, but have sound policy frameworks, are ‘pre-approved’ to draw on IMF resources, should a qualifying balance of payments need crystallise and provided they continue to maintain sound policies.

The country’s commitment to good policies, the Fund’s endorsement, and the access to IMF resources that accompany this, send a reassuring message to markets. The overall package incentivises sound policymaking, promoting economic and financial stability.

This, ultimately, helps prevent financial bumps in the road ending in crises – which create even greater financing needs (and might require conventional IMF lending, which comes with higher costs and more severe adjustment requirements), as well as damaging spillovers to the rest of the world.

Precautionary lending, a preventative approach rather than reactive one, is therefore a more efficient use of IMF resources. ‘A stitch in time saves nine’, as the proverb says.

Why haven’t more countries taken precautionary facilities in the past?   

Uptake of precautionary facilities has been fairly limited. Since the Flexible Credit Line (FCL) and Precautionary and Liquidity Line (PLL) were created in 2009, only eight countries have used them. The Short-term Liquidity Line, created during the pandemic, has only had one user.   

Our analysis shows that the limited use of precautionary facilities isn’t because there hasn’t been a need. Due to data constraints, we focus on the FCL, which is designed to provide additional tail-risk insurance to countries with very strong fundamentals and policy track records when faced with potential balance of payments (BoP) pressures.

Using a capital flows-at-risk framework, we use financial market data to identify episodes where countries face heightened risk of BoP pressures in the year ahead. We find that there have been a significant number of cases where countries have faced heightened risks, but did not take up an FCL programme, despite likely being ‘eligible’ (Chart B).

Chart B: Episodes where FCL ‘eligible’ countries faced heightened risk of future balance of payments pressures

Sources: Bloomberg Finance L.P., Eikon from Refinitiv, IMF International Financial Statistics, IMF World Economic Outlook and Bank calculations.

Notes: Periods of heightened BoP risk are defined as episodes where ‘capital-flows-at-risk’, defined as the 5th percentile of the distribution of non-resident capital flows (portfolio and other), is more negative than the average historical level of ‘capital-flows-at-risk’ when counties first took-up or augmented their FCL programme. The full distribution of capital flows is estimated for a group of 19 emerging markets using financial conditions indices as in Eguren-Martin et al. FCL ‘eligibility’ is determined using the methodology similar to that set out in Birdsall et al.

This might be because countries preferred to build up large reserves positions instead. Indeed, countries identified as having heightened risks of future BoP pressures that did not take up an FCL, had much higher average levels of reserves than those that had an FCL in place (Chart C). Excessive reserve accumulation comes with a cost though, both for the holder and the global economy.

Chart C: FCL ‘eligible’ countries’ foreign exchange reserves following a period where countries faced a heightened risk of future BoP pressures

Sources: Bloomberg Finance L.P., Eikon from Refinitiv, IMF International Financial Statistics, IMF World Economic Outlook and Bank calculations.

Notes: Level of reserves during periods of heightened risk as defined as in chart B. Eligibility also determined as in Chart B. Episodes included in the ‘FCL in place’ panel are where ‘eligible’ countries had an FCL in place or took one up within a year of facing heightened risk. Episodes included in the ‘No FCL in place’ panel are where ‘eligible’ countries did not have or take up an FCL.

Countries have also increased their access to other parts of the GFSN since the FCL was created in 2009. The total size of RFAs has expanded more than ten-fold and the estimated value of bilateral swap lines has almost doubled. Neither of these provide perfect substitutes for reserves or IMF support. RFAs are vulnerable to regional systemic shocks and remain largely untested. Access to swap lines is not universal or guaranteed.

Despite having access to other elements of the GFSN, our analysis suggests having an FCL in place helps protect against the realisation of capital outflows following a period where risks of future BoP pressures are heightened. On average, countries with FCLs tended to experience more positive capital flows following the period of heightened risk (Chart D), and were much less likely to experience significant outflows.

Chart D: Cumulative non-resident capital flows following a period where FCL ‘eligible’ countries faced heightened risk of BoP pressures

Sources: Bloomberg Finance L.P., Eikon from Refinitiv, IMF International Financial Statistics, IMF World Economic Outlook and Bank calculations.

Notes: Realised non-resident capital flows (portfolio plus other) following periods of heightened risk as defined as in Chart B. Eligibility also determined as in Chart B.

Given the realised benefits of FCL programs, and the downsides of other elements of the GFSN, why haven’t more countries come forward? Geographical patterns suggest historical experience with the IMF may be a factor. In our sample, South American countries tended to get an FCL following a period of heightened risk, while in Asia, no country has come forward for a programme (Chart E). This might relate to historical experience: some Asian countries felt they were treated unfairly by the IMF during the 1997–98 Asia crisis. 

Chart E: Regional differences in FCL take-up following periods where countries faced heightened risk of future BoP pressures

Sources: Bloomberg Finance L.P., Eikon from Refinitiv, IMF International Financial Statistics, IMF World Economic Outlook and Bank calculations.

Note: Periods are defined as number of quarters where ‘FCL’ eligible countries faced a heightened risk of capital outflows (as defined in Chart A).

What could an increase in the uptake of precautionary facilities achieve now?    

We estimate that around a fifth of countries that would likely qualify for an FCL or PLL wouldn’t currently have enough insurance – via reserves, swaps and RFAs – to absorb a moderate sudden stop shock. With a precautionary arrangement in place, nearly all would be able to absorb the shock.         

Moreover, in a fragmented world, it is likely that a significantly larger number of qualifying countries would require access to precautionary facilities to withstand external shocks. This is because only a minority could rely on reserves alone, while smooth and timely access to RFAs and swaps, which make up, on average, around 10% of their insurance, might be jeopardised by fragmentation. Moreover, if these 30 countries faced a more ‘severe’ shock, with some larger economies facing large capital outflows, aggregate financing needs could rise by around 50%.         

What does this mean for policy?

The IMF should look at ways to increase the uptake of its precautionary facilities, while ensuring they continue to incentivise sound macroeconomic policies. Not least because this commitment to sound policies combined with up-front access to additional resources when faced with elevated external risks, should ultimately reduce the actual use of IMF financing.

To increase uptake, it is important to consider ways to increase demand, including by addressing stigma. But it is also important to address concerns around their supply. This might include ensuring they are effectively resourced, as well as adopting a ‘risk-weighted’ accounting treatment on the IMF’s balance sheet, reflecting the fact that many precautionary facilities are often approved but not drawn on.

The upcoming IMF review of its precautionary facilities is a timely opportunity to do it. Risks from fragmentation mean the need for the IMF to provide robust and efficient insurance at the heart of the GFSN has never been more important.

Daniel Christen and Nicola Shadbolt work in the Bank’s International Directorate.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here