Authors

Matteo Falagiarda^[1]European Central Bank. and Petra Köhler-Ulbrich^[2]European Central Bank.

 

1. Introduction

The coronavirus (COVID-19) pandemic had a strong impact on firms’ business plans and financing needs. In view of the importance of bank borrowing as a source of financing for euro area non-financial firms,^[3]For more details on bank lending to euro area firms in recent years, see Adalid et al. (2020). the banking sector has played a key role in facilitating the flow of credit to the corporate sector during the COVID-19 pandemic. This role has been crucially supported by the sizeable support measures by monetary, fiscal and supervisory authorities, which have so far acted as a backstop against the risk of an adverse feedback loop between the real and financial sectors. This article discusses the main drivers of bank lending to euro area firms during the pandemic. Understanding the relative role of credit supply and demand forces as well as the impact of the various policy measures is crucial for policy makers in order to draw appropriate conclusions with respect to the effectiveness of the implemented measures and the possible need for further action. Against this background, the article first focuses on the early stages of the pandemic, when acute emergency liquidity needs arising from the lockdown measures were satisfied by bank borrowing at very favourable conditions. Then, it examines bank lending dynamics in the second phase of the pandemic, which was characterised by abating liquidity needs, a continuation of the policy support measures, but also by the emergence of pressures on bank intermediation due to intensifying concerns about the deterioration of borrowers’ creditworthiness. The article concludes by highlighting some of the risks to banks’ credit intermediation capacity in the near future.

 

2. The first phase of the COVID-19 crisis: emergency liquidity needs met by favourable bank lending conditions amidst ample policy support

In the first months of the pandemic, the unprecedented nature of the shock led to a marked increase in bank lending to euro area firms. Demand from firms for bank loans soared to record levels in most euro area countries from March to May 2020 as firms scrambled to bridge liquidity gaps originating from the COVID-19 shock (Chart 1, left panel, and Chart 1A in the Appendix). This increase in demand was driven by a decline in the capacity of firms to finance their ongoing costs via operating cash flows, owing to a sharp fall in their revenues during the lockdown period in the first half of 2020. This situation resulted in acute liquidity needs to finance working capital, as also indicated in the euro area bank lending survey (BLS) (Chart 2).^[4]The euro area bank lending survey (BLS) provides information on bank lending conditions in the euro area. It supplements existing statistics with information on the supply of and demand for loans to … Continue reading Moreover, in a context of high uncertainty, firms drew their credit lines and applied for new loans, often with government guarantees, with a view to building up precautionary liquidity buffers, as suggested by the same survey (Chart 2A, left panel). This is also visible in the exceptionally large accumulation of bank deposits by firms in the first months of the pandemic (Chart 2A, right panel). The aggregated balance sheet of the corporate sector reveals that firms overcompensated the large decline in revenues with an even larger recourse to bank loans and market-based financing.

In March 2020, acute emergency liquidity needs were mainly satiated by the recourse to short-term loans by drawing down previously agreed credit lines. In later months, the substantial lending flows largely reflected the use of medium- and long-term loans (Chart 1, right panel), maturities which were typically backed by the public guarantee schemes implemented since April 2020 in most euro area countries. The flat yield curve, the perceived longer duration of the pandemic and the ensuing high degree of uncertainty have also contributed to the increase in firms’ demand for long-term borrowing. This maturity shift mitigated firms’ rollover and liquidity risks that would have intensified had the new loans been granted in the form of short-term commitments. The increase in the demand for long-term loans contrasts with historical regularities, as acute liquidity needs for working capital are typically associated with higher demand for short-term loans, while long-term loans are used to finance fixed investment projects (Chart 3A).^[5]For more details on the drivers of firms’ loan demand in the euro area during the pandemic, see Falagiarda et al. (2020a). In fact, as indicated by the BLS, in contrast with firms’ financing needs for working capital, those for fixed investment declined sharply in the first half of 2020 (Chart 2),^[6]The fact that firms have used external financing mainly for inventories and working capital and less for fixed investment is confirmed by the Survey on the Access to Finance of Enterprises (SAFE) in … Continue reading mirroring the steep fall in business investment, which reflected either a reduction or a postponement of capital expenditure by firms, driven by the need to compensate revenue losses in a context of elevated uncertainty.

 

Chart 1. Bank loans to firms (flows in EUR bn) Source: ECB (BSI) and authors’ calculations. Notes: (lhs panel) Bank loans to non-financial corporations adjusted for sales, securitisation and cash pooling activities. The term “Other countries” includes flows to other euro area countries as well as seasonal adjustment residuals to preserve the additivity to the total euro area flows. The term “avg.19” refers to the quarterly average flow recorded in 2019. (rhs panel) Bank loans to non-financial corporations non-adjusted for sales, securitisation and cash pooling activities. The term “avg.19” refers to the quarterly average flow recorded in 2019.

Chart 2. Changes in demand for loans to firms and contributing factors (net percentages of banks) Source: ECB (BLS). Notes: Net percentages are defined as the difference between the percentages of banks reporting an increase (contribution to an increase) and the percentages of banks reporting a decrease (contribution to a decrease). “Other financing needs” are an unweighted average of “M&A and corporate restructuring” and “debt refinancing/restructuring and renegotiation”; “use of alternative finance” is an unweighted average of “internal financing”, “loans from other banks”, “loans from non-banks”, “issuance/redemption of debt securities” and “issuance/redemption of equity”. “General level of interest rates” was introduced in 2015 Q1.

COVID-19-related concerns about physical contact and the concomitant lockdown policies caused a large loss of value added in trade, transport, accommodation and food service activities. Strict lockdowns, a lack of demand, interruptions to supply chains and high uncertainty also spilled over into large segments of the manufacturing sector. A comparison of financing needs across sectors shows that the increase in corporate lending in the first half of the year had been highest in these sectors, the hardest hit by the COVID-19 pandemic (Chart 4A), pointing to acute liquidity needs for firms in these segments. Firms in less affected sectors have also increased their borrowing in the first half of 2020, with a view to building up precautionary liquidity buffers in an environment of high uncertainty. Developments in sectoral activity are broadly in line with the evidence from the BLS, according to which, in the first half of the year, loan demand increased considerably in the manufacturing sector, services sector (excluding financial services and real estate) and wholesale and retail trade sector (Chart 5A). Loan demand increased less in the construction sector, and more particularly in the real estate sector, where firms were less affected by the crisis. This can be attributed to the lower labour intensity and fixed costs of real estate activities, which resulted in smaller liquidity needs during the lockdown period in the first half of 2020.

A comparison across firm sizes shows that the surge in bank borrowing recorded in the first half of 2020 was more pronounced for small and medium-sized enterprises (SMEs) than for large firms. SMEs have benefited substantially from policy support measures for bank lending, such as the TLTRO III operations, as well as from public loan guarantees, which are typically targeted to this specific group of firms (see below for a more detailed discussion of the policy measures). In particular, the take-up of guaranteed loans has been significantly higher for SMEs and the self-employed than for large firms, reflecting their relatively larger emergency liquidity needs, smaller liquidity buffers, their greater dependence on banks for financing compared with large firms and overall easier and fast-track procedures in the provision of guaranteed loans for smaller amounts. The BLS confirms that the increase in the demand for loans (and, in particular, guaranteed loans) in the first half of 2020 was higher for SMEs than for large firms (Chart 3, left panel).

 

Chart 3. Changes in firms’ demand and credit standards for loans with and without government guarantees   (net percentages of banks)  Source: ECB (BLS). Notes: (lhs panel) Net percentages are defined as the difference between the percentages of banks indicating an increase and the percentages of banks indicating a decrease; (rhs panel) net percentages are defined as the difference between the percentages of banks indicating a tightening and the percentages of banks indicating an easing.

Chart 4. Bank lending rates on new loans to firms (lhs panel: percentages per annum; rhs panel: basis point changes since February 2020)  Source: ECB (MIR). Notes: (rhs panel) Very small loans are loans up to Eur 0.25 million, small loans are loans of more than Eur 0.25 million and up to Eur 1 million and large loans are loans of more than Eur 1 million.

The surge in the demand for loans by euro area firms in the initial period of the pandemic was met by historically low bank lending rates and favourable bank lending conditions.^[7]For more details on bank lending conditions for euro area firms in recent years, see Burlon et al. (2019). This is especially important in a strongly bank-based financial system like the euro area. Bank lending rates charged on loans to euro area firms have declined significantly in the first half of 2020, reaching new historical lows in many euro area countries (Chart 4).^[8]Among the large euro area countries, very low interest rates have been recorded in France, reflecting the very favourable pricing conditions of guaranteed loans, the take-up of which was very large … Continue reading The decline in rates was concentrated on those charged on very small loans, suggesting that SMEs benefitted the most from the favourable financing conditions over this period. At the same time, credit standards (i.e. banks’ internal guidelines for their lending policies or loan approval criteria) for loans to firms, both to large firms and to SMEs, tightened slightly in the first quarter of 2020, when support measures were still on their way and uncertainty was exceptionally high, but remained broadly unchanged at the euro area level in the second quarter when such measures were implemented (Chart 5). While some sectors were more affected than others (Chart 5A), banks’ credit standards remained overall beneficial across sectors in the first half of 2020. Given the size of the pandemic shock, the continuation of favourable bank lending conditions was remarkable and very much in contrast with developments during the global financial and sovereign debt crises.

This notwithstanding, banks already indicated in the first half of 2020 increased concerns about economic developments, industry-specific risks and borrowers’ creditworthiness for their lending policy, as reflected in the tightening impact of risk perceptions on their credit standards as well as in the tightening impact of banks’ risk tolerance (Chart 5). At the same time, banks’ balance sheet situation did not have a tightening impact, reflecting the persistent positive impact of pre-crisis improvements in the resilience of bank balance sheets as well as the effective policy support. In particular, following the global financial and sovereign debt crises, banks stepped up their capitalisation, partly related to stricter supervisory and regulatory requirements. In addition, banks, in particular in some jurisdictions, have cleaned their balance sheets and reduced their share of non-performing loans. This is a noticeable difference to the financial and sovereign debt crises, during which banks’ balance sheets constraints were a relevant factor in the tightening of bank lending conditions.

 

Chart 5: Changes in credit standards on loans to firms and contributing factors (net percentages of banks)  Source: ECB (BLS). Notes: Net percentages are defined as the difference between the percentages of banks reporting a tightening and the percentages of banks reporting an easing. “Cost of funds and balance sheet constraints” are an unweighted average of “cost related to capital position”, “access to market financing” and “liquidity position”; “risk perceptions” are an unweighted average of “general economic situation and outlook”, “industry or firm-specific situation and outlook/borrower’s creditworthiness” and “risk on collateral demanded”; “competition” is an unweighted average of “bank competition”, “non-bank competition” and “competition from market financing”. “Risk tolerance” was introduced in 2015 Q1.

Chart 6: Terms and conditions on loans to firms (left panel: cumulated net percentages of banks and cumulated basis points; 2014Q2=0; right panel: net percentages of banks) Source: ECB (BLS and MIR). Notes: Net percentages are defined as the difference between the percentages of banks indicating a tightening and the percentages of banks indicating an easing. “Margins” are defined in the BLS as the spread of lending rates over a relevant market reference rate; a widening of margins is defined as a tightening. The cumulated lending rate spread refers to the composite lending rate for firms minus 3-month OIS.

The benign developments in banks’ credit standards in the first half of 2020 were consistent with banks’ actual credit terms and conditions, as agreed between banks and borrowers in the loan negotiation process. In line with historically low bank lending rates and squeezed spreads of bank lending rates over market reference rates, margins on average loans remained narrow in the first half of 2020, while they tightened for riskier loans (Chart 6). Over the same period, non-price terms and conditions tightened slightly. A comparison across firm sizes indicates that terms and conditions applied by banks on loans to SMEs were reported to have developed more favourably than for large firms in the first half of the year (Chart 6A), in line with actual lending rate developments. This evidence confirms that SMEs, which generally tend to be more at risk of becoming credit constrained during crisis periods, benefitted the most from the supportive lending conditions engendered by the strong policy response to the COVID-19 crisis.

In the first half of 2020, bank lending conditions for euro area firms remained beneficial in spite of the unprecedentedly large demand for loans and the deteriorating creditworthiness of many borrowers. This evidence points to the effectiveness of the response by monetary policy, fiscal policy and supervisory authorities to the COVID-19 crisis. Besides their direct impact on lending, these policies also provided assurance to the private sector on forceful counter-measures, thereby reducing overall macroeconomic uncertainty.

 

Chart 7. Take-up of loans covered by COVID-19-related public guarantees (EUR bn) 

Sources: National authorities and authors’ calculations.

Notes: The take-up data refer to approved amounts of guaranteed loans. As guaranteed loans can also be granted in the form of revolving credit facilities, the approved amount is higher than the amount actually disbursed. In the absence of a breakdown by firm size for Italy, it is assumed that guaranteed loans to SMEs are those granted via the Fondo di Garanzia, while guaranteed loans to large firms are those granted via SACE (the Italian export credit agency).

 

The monetary policy accommodation introduced by the Eurosystem in response to the crisis supported considerably euro area firms’ financing conditions. First, under the Pandemic Emergency Purchase Programme (PEPP) announced in March 2020, the ECB’s asset purchases were expanded and made more flexible by allowing fluctuations in the distribution of purchases over time, across asset classes and among jurisdictions. The PEPP, by impacting yields across the maturity spectrum, exerted significant downward pressures on lending rates. Second, the ECB’s targeted longer-term refinancing operations (TLTRO III) have offered attractive bank funding conditions, which banks passed on to their customers, thereby facilitating bank lending to euro area firms during the pandemic.^[9]The TLTRO III recalibrations in March and April 2020 increased further the attractiveness of the TLTRO III. Altavilla et al. (2020) show that banks’ ability to supply credit would have been … Continue reading Third, the ECB introduced temporary collateral easing measures in April 2020. These measures eased the conditions at which credit claims are accepted as collateral in the liquidity providing operations of the Eurosystem and facilitated the availability of eligible collateral to support the provision of credit via the Eurosystem’s  refinancing operations.^[10]Among other things, the eligible collateral was expanded to include very small loans and loans covered by COVID-19-related public guarantees. Fourth, the ECB’s negative interest rate policy (NIRP) contributed to historically low lending rates, thereby supporting bank lending.^[11]The NIRP has proven to be effective in easing financing conditions for euro area firms. For more details, including a discussion on the channels through which the NIRP may impact bank loan provision, … Continue reading Overall, according to banks’ assessment, the ECB’s monetary policy measures contributed positively to an increase in lending volumes and an easing of bank lending conditions during the COVID-19 period (Chart 7A).^[12]Other measures implemented by the ECB as a response to the COVID-19 crisis included a recalibration of the Asset Purchase Programme (APP) and the pandemic emergency longer-term refinancing operations … Continue reading

Besides monetary policies, also other policy domains provided critical support to the credit provision to euro area firms. The microprudential policy response to the crisis has provided important capital relief for banks, which created further space for bank balance sheet expansion. National fiscal policies have also been instrumental in providing liquidity support, thereby averting so far a potential wave of corporate bankruptcies. Schemes of public guarantees on bank loans were implemented by most euro area governments in April 2020 in order to help banks accommodate the surge in loan demand at favourable conditions. These programmes transferred some of the credit risk (in some cases the entire credit risk) and potential credit losses from banks to governments, thereby mitigating the costs for banks.^[13]The features of the loan guarantee schemes vary across countries but they must all comply with the guidelines adopted by the European Commission (see Section 3.2 of the Communication from the … Continue reading The window for applying for loans covered by guarantee schemes was initially set to close at the end of 2020. In addition, public and private moratoria were introduced in most euro area countries to provide short-term relief through the suspension of principal and/or interest payments on loans.^[14]These schemes were compliant with the guidelines of the European Banking Association (EBA). These schemes avoided that loans to solvent corporates became non-performing due to temporary liquidity needs to bridge the pandemic.

Public loan guarantee schemes have played a key role in supporting corporate lending dynamics in the second quarter of 2020, thereby contributing to the surge in loan demand provided at favourable lending conditions to firms, as described above. The substantial lending flows recorded over this period largely reflected the take-up of loans covered by public guarantees, most of which were granted to SMEs (Chart 7). Gross flows of guaranteed loans were higher than overall net lending flows in all large euro area countries, implying a shift from non-guaranteed loans into guaranteed loans. Moreover, lending dynamics were proportionally stronger in countries with a higher take-up of guaranteed loans, such as Spain and France. In these two countries, where fiscal support for firms was delivered mainly via guarantee schemes, more than 60% of new business volumes in the second quarter of 2020 consisted of guaranteed loans. The impact of loan guarantee schemes was also reflected in the favourable developments of bank lending conditions. First, guarantees crucially contributed to the drop of lending rates to historically low levels, especially for small loans, the ones typically backed by these programmes. Moreover, they exerted considerable easing pressures on credit standards (Chart 3, right panel) and credit terms and conditions, particularly in the countries where the use of this type of loans was the largest.

 

3. The second phase of the COVID-19 crisis: liquidity needs abated, while incipient signs of tighter credit supply counteracted by continued policy support

As the spread of COVID-19 temporarily decelerated and lockdown restrictions were relaxed in mid-2020, activity started to rebound and firms’ sales recovered. Firms’ demand for credit started to abate correspondingly (Chart 1, left panel, and Chart 1A), also dampened by the significant precautionary liquidity buffers built-up over the period from March to May 2020. The marked moderation in bank borrowing by firms over the summer of 2020 was also reflected in the diminished demand for loans benefitting from a public guarantee (Chart 7, left panel). In the last quarter of 2020 and at the beginning of 2021, bank lending to firms stabilised at modest levels, in spite of the resurgence of the pandemic and the associated tightening of containment measures. The absence of a large surge in emergency borrowing reflected available liquidity buffers and direct government support measures, which shielded firms in affected sectors in an environment of renewed revenue shortfalls. In addition, at the aggregate level, the second and third waves of COVID-19 infections and the resultant containment measures have not been as disruptive to firms’ sales and operating cash flows as during the first wave. This is because some economic sectors were less affected, partly as they profited from the recovery of the global economy. In addition, firms and customers seem to have adapted better to the new environment.^[15]For more details, see Battistini and Stoevsky (2021). At the same time, loan demand continued to be dampened by the high uncertainty, especially for financing fixed investment in the sectors more affected by the pandemic. The use of other sources of financing by large firms in less affected sectors also weighed on loan demand over this period. Finally, some firms may have also become more reluctant to take on more bank debt because they might have had accumulated already significant amounts of debt.

In the second half of 2020, longer-term loans continued to support lending dynamics (Chart 1, right panel), reflecting their coverage under guarantee schemes and the flat yield curve. While the increase in corporate lending in the first half of the year had been highest in the sectors hardest hit by COVID-19, the deceleration in the second half of the year has been broad-based across activities (Chart 4A). In addition, bank lending dynamics displayed increasing heterogeneity across countries. Positive lending flows were recorded in France, Italy and to a lesser extent Germany, while net redemptions were recorded in Spain. In Italy, lending to firms continued to reflect the take-up of loans benefitting from a public guarantee. In this country, net lending was also supported by moratoria on loan repayments (i.e. implying temporarily less loan redemptions), as the usage of these schemes continued to be larger than in other countries.

The moderation in bank lending dynamics observed since the summer of 2020 is confirmed in the BLS by a net decline in firms’ loan demand for the second half of 2020 and the first quarter of 2021, following the highest net balance ever recorded in the second quarter of 2020 (Chart 2). The net decline in loan demand was somewhat stronger for SMEs than for large firms, especially for non-guaranteed loans. In line with lower emergency liquidity needs and existing liquidity buffers, banks reported during this period overall lower financing needs for inventories and working capital than in the first half of 2020. Still, liquidity needs and precautionary buffers continued to be relevant factors for firms’ demand for loans with public guarantee (Chart 2A, left panel). Importantly, firms’ financing needs for fixed investment continued to dampen loan demand, suggesting that firms’ long-term business plans have been put on hold due to the high uncertainty, especially in sectors more affected by the pandemic, which may postpone a sustained recovery.

After declining significantly since the outbreak of the COVID-19 pandemic, bank lending rates on loans to euro area firms have rebounded but remained around record lows in the second half of 2020 and in the first months of 2021 (Chart 4). Lending rates on very small loans displayed a marked U-shaped pattern, characterised by an increase since May 2020 that has been almost equally steep as the prior decline. This mirrors the developments in the use of guaranteed loans, the majority of which were granted to SMEs at very attractive conditions. Overall, the developments in lending rates support the view that the deceleration in credit dynamics observed in the second half of 2020 was largely driven by the reversal of the extraordinarily high demand for loans seen in the early stages of the crisis.

Despite bank lending rates remaining around historically low levels, banks became overall less forthcoming in their attitude towards credit expansion in the second half of 2020. Credit standards on loans to firms tightened both in the third and fourth quarters of 2020 (Chart 5). This was the first significant tightening in the last eight years and was above the historical average since 2003, while remaining considerably below the peak during the great financial crisis. It also remained below the euro area peak during the sovereign debt crisis, where only some countries were affected. The tightening of credit standards was driven by heightened concerns of banks about intensifying risks to borrowers’ creditworthiness and possible loan losses in the future, in particular in the sectors most affected by the pandemic. The tightening was stronger in the commercial real estate and in the trade sectors, while the services sector (that covers both those businesses which suffered and those which profited from the pandemic) and manufacturing were somewhat less affected (Chart 5A). Construction and residential real estate sectors were the least affected from tightening credit standards in the second half of 2020, reflecting the resilience of residential real estate markets to the COVID-19 shock. In the first quarter of 2021, a less pronounced tightening of credit standards was reported by banks in net terms, on account of a smaller contribution of risk, both in terms of perceived risk and risk tolerance of banks. This likely reflected the prolongation of fiscal support measures, the continued support from monetary policy and supervisory measures and the broader improvement in risk sentiment in the first quarter of 2021. Still, risk perceptions related to the economic and firm-specific situation and outlook continued to be the main factor contributing to the tightening of credit standards on loans to firms.

In line with the reduced use of guaranteed loans, the easing impact of these loans on credit standards was more limited during this period than in the first half of 2020 (Chart 3, right panel), while credit standards continued to tighten for non-guaranteed loans. In addition, in the fourth quarter of 2020, the tightening of credit standards became somewhat stronger for SMEs than for large firms, on account of a stronger net tightening of credit standards for non-guaranteed loans to SMEs, despite the continued presence of ample policy support, often tailored specifically towards SMEs. While this development became less acute in the first quarter of 2021, it may nonetheless signal that banks consider credit risks for SMEs being larger, in line with typical patterns of higher risks for SMEs being credit constrained given their more opaque balance sheets due to lower and later reporting requirements.

During past episodes of stress, the BLS indicator of credit standards has proved to be a reliable harbinger of future weakness in bank credit. Historical regularities suggest that credit standards tend to lead lending to corporates by around five quarters (Chart 8A). At the same time, the predictive information content of credit standards tends to be state-contingent, as it emerges more prominently in periods of stress. This is because, over these periods, a significant tightening in credit standards is typically associated with binding supply constraints. However, unlike previous crisis episodes, the net tightening of credit standards for loans to firms in the second half of 2020 and in the first quarter of 2021 was not accompanied by a tightening contribution of banks’ cost of funds and balance sheet constraints (which in fact had overall an easing impact in this period), a factor historically associated with worsening credit supply conditions. This reflected the more resilient state of the banking system, compared with the great financial and sovereign debt crises, as well as the policy response to the pandemic, which has been much more proactive than in prior crisis episodes. Both factors have been key to mitigating the adverse supply pressures originating from deteriorating risk perceptions.

At the same time, the favourable lending rate developments observed since the summer of 2020 might have concealed compositional effects, arising from a shift of new loans to lenders with a better credit risk profile as well as changes in the non-price terms and conditions of loans. Reflecting banks’ increased concerns about the riskier loan segments, margins on riskier loans widened further in the second half of 2020 and in the first quarter of 2021 (Chart 6, left panel). Banks also intensified their tightening of non-price terms and conditions, in particular their collateral requirements (Chart 6, right panel), which reached in the last quarter of 2020 a level unseen since 2011 (although remaining well below the peak during the global financial crisis). This indicates that banks aimed to protect themselves against higher credit risks by demanding the pledging of assets as security. On more general grounds, banks often tend to adjust their non-price terms and conditions when they perceive higher credit risk, as this provides an opportunity, compared with changing the pricing of the loans, to reduce potential adverse selection issues in lending. Consistent with this, a comparison across firm sizes shows that bank lending policies, in particular as regards collateral requirements and margins on riskier loans, tended to become stricter especially for SMEs in the last quarter of the year (Chart 6A). This evidence confirms that banks’ attitude towards SMEs may have become more cautious.

Banks’ concerns about firms’ debt servicing and repayment capacity and possible loan losses were also reflected in their indications on the impact of non-performing loans (NPL) on their bank lending conditions. Following a modest impact of NPL ratios on banks’ credit standards in 2018 and 2019, the impact has increased in the course of 2020 (Chart 9A). At the same time, euro area banks’ actual NPL ratios remained broadly stable. Nevertheless, actual NPL developments should not be interpreted as a reassuring sign of unchanged credit risk on banks’ balance sheets. First, according to the accounting procedure of NPLs, loans are considered as non-performing only if borrowers do not meet their agreed repayment arrangements for 90 days or more. Second, support measures such as moratoria on loan repayments have contributed to delays in NPL recognition, although credit risk was already materialising. The phasing out of these schemes could lead to an increase in NPLs. In fact, euro area banks have built up their provisions for loan losses, dampening bank profitability in the second half of 2020. The surfacing of new NPLs may constitute an important headwind to banks’ intermediation capacity in 2021.

In order to prevent the emergence of bottlenecks in the provision of bank financing resulting from the economic fallout from the resurgence of the pandemic, policy support measures were prolonged and in part recalibrated in the second half of 2020 and in the first half of 2021.

On the monetary policy side, the ECB announced in December 2020 various measures, including: (i) a recalibration and prolongation of TLTRO III with the aim of preserving favourable funding conditions for banks and further incentivise their lending to the real economy, (ii) an increase of the envelope of its asset purchases under the PEPP, and (iii) an extension of the duration of the set of collateral easing measures adopted at the onset of the crisis.

Similarly, on the fiscal side, support measures were extended into 2021. After being phased out in September 2020, EBA’s guidelines on moratoria were reactivated in December 2020. Besides setting the new deadline for application at the end of March 2021, a cap of nine months to the length of payment extension was introduced in order to mitigate the risk faced by banks. In addition, following the prolongation of the Temporary Framework for state aid measures by the European Commission in October 2020, the window for applying for loans covered by guarantee schemes has been extended by an additional six months until the end of June 2021 in most euro area countries. Some governments also loosened conditions on the original guarantee schemes, e.g. in the form of longer maturity and grace periods for repayments, or proposed programmes of participative loans. These loans will be still granted by banks and guaranteed by the state, but will be treated as equity, thereby improving firms’ debt position. Finally, in view of the persistence of the pandemic, in January 2021 the European Commission extended for additional six months until the end 2021 the Temporary Framework for state aid measures.^[16]The European Commission will also allow governments to convert guarantees granted under the Temporary Framework into other forms of aid, such as direct grants.

4. Concluding remarks

The vigorous and prompt policy response to the COVID-19 shock has been key to keeping bank lending conditions favourable in the euro area, thereby supporting the financing of firms. While the anatomy of the moderation in bank lending dynamics since the summer of 2020 points to a preponderance of demand-side factors, incipient signs of tighter credit supply conditions have emerged. Moreover, the uncertainty surrounding the evolution of the pandemic and related containment measures continued to weigh on firms’ demand for financing fixed investment, especially in the sectors more affected by the pandemic. In this environment, the expected further deterioration of the balance sheet health of borrowers and lenders may pose risks of adverse financial amplification effects. The continuation of a supportive policy environment will thus be crucial for staving off the risk of a deterioration in credit supply conditions. This would also improve the confidence that firms need in order to engage in long-term investment projects, on which a sustained recovery in economic activity depends.

 

Notes

The views expressed in this paper are those of the authors and do not necessarily reflect the views of the European Central Bank or the Eurosystem.

 

References

Adalid, R., Falagiarda, M., and Musso, A. (2020). Assessing bank lending to corporates in the euro area since 2014. Economic Bulletin, Issue 1, European Central Bank.

Albertazzi, U., Bijsterbosch, M., Grodzicki, M., Metzler, J., and Ponte Marques, A. (2020). Potential impact of government loan guarantee schemes on bank losses. Financial Stability Review, May 2020, European Central Bank.

Altavilla, C., Barbiero, F., Boucinha, M., and Burlon, L. (2020). The great lockdown: pandemic response policies and bank lending conditions. Working Paper Series, No 2465, European Central Bank.

Anderson, J., Papadia, F., and Véron, N. (2021). COVID-19 credit support programmes in Europe’s five largest economies. Working Paper 03/2021, Bruegel.
Bańkowska, K., Ferrando, A., and García, J. A. (2020). The COVID-19 pandemic and access to finance for small and medium-sized enterprises: evidence from survey data. Economic Bulletin, Issue 4, European Central Bank.

Battistini, N., and Stoevsky, G. (2021). The impact of containment measures across sectors and countries during the COVID-19 pandemic. Economic Bulletin, Issue 2, European Central Bank

Boucinha, M., and Burlon, L. (2020). Negative rates and the transmission of monetary policy. Economic Bulletin, Issue 3, European Central Bank.
Burlon, L., Dimou, M., Drahonsky, A.-C., and Köhler-Ulbrich, P. (2019). What does the bank lending survey tell us about credit conditions for euro area firms. Economic Bulletin, Issue 8, European Central Bank.

ECB (2020). Survey on the Access to Finance of Enterprises in the euro area – April-September 2020. European Central Bank.

ECB (2021). The euro area bank lending survey – Fourth quarter of 2020. European Central Bank.

Falagiarda, M., Köhler-Ulbrich, P., and Maqui, E. (2020a). Drivers of firms’ loan demand in the euro area – what has changed during the COVID-19 pandemic? Economic Bulletin, Issue 5, European Central Bank.

Falagiarda, M., Prapiestis, A., and Rancoita, E. (2020b). Public loan guarantees and bank lending in the COVID-19 period. Economic Bulletin, Issue 6, European Central Bank.

Ferrando, A., and Ganoulis, I. (2020). Firms’ expectations on access to finance at the early stages of the Covid-19 pandemic. Working Paper Series, No 2446, European Central Bank.

Köhler-Ulbrich, P., Hempell, H. S., and Scopel, S. (2016). The euro area bank lending survey. Occasional Paper Series, No 179, European Central Bank.

 

Appendix

Bank Lending to Euro Area Firms – What Have Been the Main Drivers During the COVID-19 Pandemic?

 

Chart 1A. Bank loans to firms (lhs panel: monthly flows in EUR bn; rhs panel: monthly flows in EUR bn, index) Source: ECB (BSI), University of Oxford and authors’ calculations. Notes: Bank loans adjusted for sales, securitisation and cash pooling activities. The stringency index is a composite index produced by the University of Oxford that captures the strength of government restrictions on social and businesses in response to COVID-19. The index for the euro area is the GDP-weighted average of the indexes for individual euro area countries. A level of 100 denotes the maximum level of restrictions.

Chart 2A. Purpose of loans and firms’ deposit inflows (lhs panel: net percentages of banks; rhs panel: flows in EUR bn) Sources: ECB (BLS, BSI) and authors’ calculations. Notes: (lhs panel) Factors affecting the demand for loans or credit lines with COVID-19-related government guarantees. The net percentage refers to the difference between the sum of the percentages for “increased considerably” and “increased somewhat” and the sum of the percentages for “decreased somewhat” and “decreased considerably”. Banks can select more than one factor that affects loan demand. Therefore, the sum of the net percentages can exceed 100 in this chart. (rhs panel) Borrowing of firms include bank loans and debt security issuance. The term “avg.19” refers to the quarterly average flow recorded in 2019.

Chart 3A. Firms’ financing needs for fixed investment and demand for long-term loans (lhs panel: four-quarter moving average of net percentages of banks, annual percentage changes; rhs panel: net percentages of banks, annual percentage changes)  Sources: ECB (BLS), Eurostat and authors’ calculations. Notes: “GFCF” stands for gross fixed capital formation. Demand for long-term loans and financing needs for fixed investment are net percentages of banks indicating an increase or a positive impact on firms’ loan demand.

Chart 4A.Bank loans to firms and gross value added by sector (p.p. contributions to percentage changes 2020Q4 vs 2019Q4, percentage changes 2020Q3 vs 2019Q4)  Sources: ECB (BSI), Eurostat and authors’ calculations. Notes: Based on outstanding amounts of non-adjusted loans to non-financial corporations. Services include trade, transportation, accommodation, food service activities and ICT.

Chart 5A. Changes in credit standards and loan demand across economic sectors (net percentages of banks) Source: ECB (BLS). Notes: Sectors are defined based on the NACE Rev. 2 classification. Construction (excluding real estate), services (excluding financial services and real estate). Net percentages for credit standards are defined as the difference between the percentages of banks reporting a tightening and the percentages of banks reporting an easing. Net percentages for loan demand standards are defined as the difference between the percentages of banks reporting an increase and the percentages of banks reporting a decrease.

Chart 6A. Credit standards and terms and conditions by firm sizes (net percentages of banks)  Source: ECB (BLS). Notes: Net percentages are defined as the difference between the percentages of banks indicating a tightening and the percentages of banks indicating an easing. “Margins” are defined as the spread over a relevant market reference rate.

Chart 7A. Impact of the ECB’s unconventional monetary policy on bank lending (net percentages of banks) Source: ECB (BLS). Notes: Net percentages are defined as the difference between the sum of the percentages of banks indicating a tightening or an increase and the sum of the percentages of banks indicating an easing or a decrease. “Net tightening of credit standards” is not available for the negative deposit facility rate.

Chart 8A. Correlation at different leads/lags between loans to firms and BLS indicators (correlation coefficient by quarter, where 0 denotes contemporaneous correlation)  Source: ECB (BSI, BLS) and authors’ calculations. Notes: Correlation between 4-quarter moving averages of BLS indicators and annual growth rate of loans to non-financial corporations.

Chart 9A. Impact of banks’ non-performing loan ratios on their lending conditions and actual NPL ratios for loans to euro area firms

(net percentages of banks and percentages) 

Sources: ECB (BLS and Supervisory banking statistics). Notes: In the BLS, the NPL ratio is defined as the stock of gross non-performing loans on banks’ balance sheets as a percentage of the gross carrying amount of loans. The actual NPL ratios refer to euro area significant institutions and are defined as the gross carrying amount of non-performing loans (and advances), as a percentage of total loans (and advances). They are calculated as an average over the respective periods. The first period for the actual NPL ratio refers to 2015 Q2 – 2017 Q4.

 

Authors

Ignazio Angeloni

 

One year has passed since the Covid-19 pandemic was discovered and recognized as such. The world economy plunged into a major recession; some areas have recovered, some are in the process of doing so while others are still deep into it. Policymakers have responded promptly with measures to protect the economy; in particular, massive support has been provided to the banking sector in the form of credit moratoria and guarantees. These measures have helped spared people, firms and banks the brunt of the crisis but have also suspended the normal functioning of the market mechanisms. As a result, the full consequences of the crisis are not visible yet. As one ECB supervisor put it to me recently, referring to eurozone banks: “we stopped the car; when we will have to start it again, we don’t k now what we will find under the hood”.

Virus and lockdowns impact the banks through multiple channels. The first to manifests itself is an increase in the demand for credit, as households and firms experiencing revenue shortfalls draw on their credit lines, often with the support of public guarantees. The increase in the amount of guaranteed credit is revenue-positive for the banks; this explains, for example, why 2020 was a surprisingly good year for small banks in the US^[1]Wall Street Journal: “The best year ever: 2020 was surprisingly good to small banks”, 14 December … Continue reading. This positive effect is dampened, and may even be reversed, by the reduction of lending margins which follows from a more accommodative monetary policy. Over time, however, both of these impacts are likely to be dwarfed by the deterioration of credit quality resulting from the recession. This effect becomes evident with a considerable time lag, after the public support measures are lifted.

In the eurozone, an increase in the demand of credit was observed in early 2020. The growth rate of bank loans to non-financial firms, close to 3 percent in the pre-Covid period, rose to 5 percent in the first quarter and reached a plateau around 7 percent in the summer months^[2] See ECB Economic Bulletin, various issues.
https://www.ecb.europa.eu/pub/pdf/ecbu/ecb~b6a4a59998.eb_annex202101.pdf.. Intermediation margins shrunk somewhat, due to the decline of lending rates on certain components of the loan portfolio, mainly overdrafts. By contrast, no deterioration of credit quality has been observed so far in the supervisory statistical reports. The (gross) NPL ratio for the euro area as a whole, slightly over 3 percent at the end of 2019, continued to decline, reaching 2.8 percent in September 2020^[3] ECB supervisory statistics,
https://www.bankingsupervision.europa.eu/banking/statistics/html/index.en.html.. However, recent surveys by the ECB suggest that this benign phase may be ending and the post-Covid “wave” of NPLs may now start^[4]A. Enria, “European banks in the post-Covid world”, speech given at the Morgan Stanley European Financials Conference, 16 march … Continue reading.

Eventually, NPLs are expected to rise sharply in the eurozone. An estimate based on an adverse scenario, published by the ECB, puts the peak at 1.4 trillion euros^[5]A. Enria, “An evolving supervisory response to the pandemic”, Speech given at the European Banking Federation, October 2020; … Continue reading, which would imply a CET1 ratio depletion of up to 5.7 percent. It is interesting to compare this estimate with the NPL increase observed after the great financial crisis (GFC). Between 2007, the last pre-crisis year, and 2013, the peak year, the NPL ratio in the euro area rose by roughly 6 percentage points, while NPLs in nominal terms increased by over 600 bn. euros. If one makes the milder assumption that NPL may rise up to 1 tn. euros, the increase relative to today’s level would be comparable in magnitude to that occurred after the GFC. Under the aforementioned adverse scenario, it would be significantly greater.

While magnitudes may be comparable, the context in which the NPLs increase occurs this time is completely different. In the GFC, the epicenter of the crisis were the banks themselves – their excessive risk taking in the earlier period and later the delays in recognizing the problem and dealing with it. Now, the banks are “victims” of an exogenous and unpredicted shock, which they are in fact contributing to mitigate. As Augustin Carstens, general manager of the BIS, put it at an early stage, banks this time are part of the solution, not of the problem^[6]A. Carstens, “Bold steps to pump coronavirus rescue funds down the last mile”, Financial Times, 29
March 2020.. And they have in fact already started doing so, by keeping credit channels open. Supervisory and regulatory measures to deal with the problem should accordingly be different.

Broadly speaking, four were the main areas of response of eurozone supervisors and regulators after the GFC, in dealing with NPLs:

1. Supervisory action by the ECB. ECB action was organized in a specific action plan, which included guidelines, regular and ad-hoc reviews and inspections, as well as guidelines and Pillar II requirements applied to capital and provisions;
2. Pillar I provisioning requirements. These requirements, embodied in EU law in 2019, are often referred to as “calendar provisioning”;
3. Accounting rules. They relate to the way in which NPLs are quantified for accounting purposes, and were introduced in the EU as part of the new IFRS9 framework;
4. Asset management companies (AMCs). Various proposals were made to establish AMCs either at national or at area-wide level, to help banks remove NPLs from their balance sheets. These proposals were extensively discussed but never implemented.

In the following sections, these areas are examined from the viewpoint of whether they can help in the new situation. The conclusion is that the two main new regulatory elements which were introduced, points 2 and 3, are no longer suited or at least would require significant adaptation. Asset management companies, at national or at area-wide level, are an interesting avenue to consider but for several reasons are not likely to become part of a realistic policy package in the foreseeable future. Traditional micro-supervisory tools will therefore continue to occupy center stage. The final section expands on this conclusion with some comments on how the ECB can overcome the challenge.  

 

1. Supervisory action

ECB supervision started dealing with NPLs immediately after its inception, in 2014. It did so by launching a dedicated “action plan”, which was started in 2015 and virtually completed, except for routine follow-ups, before Covid struck at the beginning of 2020. Details on the ECB NPL action plan are available from several sources^[7]ECB Guidance on Non-Performing Loans, 2017; see https://www.bankingsupervision.europa.eu/ecb/pub/pdf/guidance_on_npl.en.pdf; and I. Angeloni, Beyond the pandemic: reviving Europe’s banking union; … Continue reading. For our purpose here, three aspects need highlighting.  

First, the plan put major emphasis on the need for banks to maintain efficient structures to measure and monitor the state of their exposures and the debtors’ ability to pay. These structures would include ad-hoc internal units able to collect all relevant information, with direct access to top management and decision-making boards. Before the ECB action plan, this was not regarded as a priority by many banks. Often, information on credit quality was not available in a systematic way and therefore boards and management were not always properly informed. As part of the action plan, the ECB requested banks to set up dedicated units in charge of monitoring loan performance, with direct reporting lines to the board, responsible also for proposing solutions for NPL disposal if needed.  

This aspect remains crucial today; in fact, good internal information and governance are going to be particularly important in the post-Covid scenario. While bank exposures are provisionally protected by moratoria and guarantees, banks need to continue to maintain an updated picture of the clients’ ability to pay. This is an aspect the ECB supervision has repeatedly insisted on in 2020. Using the earlier metaphor, maintaining good internal information will lower the probability of bad surprises when the “hood of the car” will be opened. 

Secondly, the ECB action plan was based on the idea that the NPL strategies should be tailored to the specific conditions of each bank. For this purpose, emphasis was placed on a constant dialogue between the teams of examiners and the bank. Their interaction would exploit the best information available on the situation of the bank’s loan portfolio, in order to propose to the bank’s decision makers and to the supervisory authority itself, the strategy most appropriate in each case.  

Third, while tailored to the bank’s specific condition, the NPL strategies should also satisfy criteria common across all supervised banks. Consistent criteria fulfil the banking union’s principle of a single supervisory concept applied to all participating banks. Criteria should be not only consistent, but also transparent. Transparency, a universal principle of good governance, is also a contributor to effectiveness because policies which are well understood tend to be more easily accepted and followed. 

The ECB meant to fulfil the twin requirement of consistency and transparency by announcing “supervisory expectations” regarding NPL provisioning. Banks with a significant NPL problem were asked to set-up provisioning plans within specific time frames, different across loan types. “Supervisory expectation” were not rigid rules but rather starting points of supervisory dialogues, during which specific elements could be taken on board and modifications in the provisioning calendar could be made. NPL strategies would eventually become an input in the annual supervisory reviews (Supervisory Review and Evaluation Process, or SREP^[8]See https://www.bankingsupervision.europa.eu/about/ssmexplained/html/srep.en.html ), thereby contributing to the determination of Pillar II requirements. 

This combination of general criteria and bespoke elements helped exert the right amount of supervisory pressure while not losing sight of individual considerations. This approach was successful: the (gross) NPL ratio for the euro area declined between 2013 and 2019 from close to 7 percent to close to 3 percent, with a marked convergence across countries. The plan and the recapitalization processes that followed did not prevent, in that period, a restart of the bank lending process in the eurozone and a recovery of its economy. 

After the pandemic, the SREP was essentially suspended. Pillar II requirements have been kept constant except for a few specific cases. This means that the underlying conditions of the banks’ exposures are no longer reflected in supervisory policies. However, the underlying approach with its blend of rule-based and ad-hoc elements remains valid; in fact, it will be particularly useful during the exit from the pandemic. At that time, bank specific conditions will be particularly important because each bank is impacted differently by the virus and the lockdowns depending on the sectoral and geographical mix of its exposures. The quantum of discretionary decisions by the supervisor is likely to increase. This raises the bar for the ECB, which will need to apply in each case the proper mix of flexibility and determination. Common principles regarding NPL disposal and provisioning plans will remain useful but will require adaptation to individual circumstances. Excessively rigid instruments (like the legal provisioning calendars discussed in the next section) are not going to be helpful. 

 

2. Calendar provisioning

The concept of “supervisory expectation” mentioned in the previous section was initially not universally well understood. While parts of the banking community and some member countries were resisting the ECB’s pressure towards cleaning balance sheets, the European Parliament objected on the legal side, arguing that supervisory expectations invaded the prerogative of legislators by being akin to general rules rather than specific risk-based requirements applied on a case-by-case basis^[9]See letter sent to the ECB by the President of the EP on .. 2017 ( See https://www.politico.eu/wpcontent/uploads/2017/10/Letter-to-President-Draghi.pdf). .  

Misunderstandings and criticisms converged in putting in motion a process leading to a legislative package dealing with NPLs, which after a long gestation entered into force in 2019^[10]See a Council summary here https://www.consilium.europa.eu/en/press/pressreleases/2019/04/09/council-adopts-reform-of-capital-requirements-for-banks-non-performing-loans/. The full text is here … Continue reading . The law prescribed minimum legal coverage levels for loans (so-called “prudential backstops”), with percentages increasing with the time of non-performance (between 1 and 10 years), distinguishing among different loan categories: secured by immovable collateral, secured by movable collateral, and unsecured. The legal (Pillar I) requirement was intended to coexist with possible additional requirements set by the supervisor as part of Pillar II. 

Unlike the “expectations”, however, the legal requirement lacked any flexibility in responding to bank specific conditions. This may have been unfit to individual banks in some cases. More seriously, it could become inapplicable to the system as a whole in case of system-wide adverse shocks outside the banks’ control – for example: a pandemic like Covid-19.  Not surprisingly, the prudential backstops were de-facto suspended as a result of the entry into force of moratoria and government guarantees^[11]The EU “banking package” introduced in 2020 is available here
https://ec.europa.eu/commission/presscorner/detail/en/qanda_20_757. 

Even beyond the short term, the prospect of restoring the “prudential backstop” in its present form after the pandemic is questionable. Provisioning calendars enshrined in law may at times become an alibis discouraging supervisors from proactively applying Pillar II powers for the same purpose. Parameters set by law across the board, as already mentioned, may not fit individual circumstances. More seriously, in presence of certain shocks they become impossible to apply. Rules whose application is impeded by circumstances difficult to foreseen in advance lose credibility, especially when such circumstances occur.  

 

 3. Accounting treatmentof NPLs 

As part of the reforms undertaken globally after the GFC, accounting rules for financial institutions were changed in several respects, with the aim of making financial statements responsive to changing economic conditions. Part of the amendments regarded NPLs. The underlying logic there was to make NPL recognition and provisioning no longer based on incurred (past) losses, but rather corresponding more closely to the moment in which the corresponding risks were undertaken.  

Fig. 1 provides a graphical representation. During normal demand-driven business cycles, risks are perceived to be low in the upswing. In this phase banks tend to undertake more risky lending (left-hand part of the curve), which normally results in NPLs later in time. If provisions are based on incurred losses, they end-up being made when the economy declines (right-hand side of the curve), hence strengthening the recession. It may then be appropriate to anticipate the provisions to match the time when risks originate. Early provisioning dampens growth in booms and stimulates it downswings. Basing provisions on the expected level of NPLs therefore exerts a desirable counter-cyclical effect. 

Figure 1: Demand cycle 

Following this type considerations, and consistent with the general move towards mark-to-market accounting after the crisis, new IRFS9 rules were introduced in EU law in 2016^[12] Commission Regulation 2016/2067; see https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A32016R2067, effective in 2018 but with a gradual transition which foresaw a full phasing in only in 2023.  

The new approach has two problems. First, it requires banks to formulate accurate expectations of their future losses. This may not be easy, not only because of the inherent uncertainty but because, as already noted, expectations tend to be optimistic in booms and pessimistic in busts^[13]See for example J. Abad and J. Suarez, “IFRS 9 and COVID-19: Delay and freeze the transitional arrangements clock“; VoxEU 2 April 2020, see … Continue reading. The second, more serious problem is that the logic just described applies only under the specific demand-driven cycle depicted in fig. 1.  

Fig 2 represents a different economic cycle, more similar to that occurred under Covis-19. The left side of the curve represents the time when the pandemic hits the economy with the related initial lockdowns; say, the first half of 2020. The wave of NPL is not yet manifest in that phase; it will occur later. If provisions are based on expected losses, they tend to worsen the economic cycle when it is already declining due to the pandemic shock. It is better, in this case, to delay the provisioning to a later time when the economy recovers (right side of the curve). Under this type of cyclical pattern, unlike in the previous one, traditional, backward looking NPL provisioning based on incurred losses is counter-cyclical, while that stemming from the new accounting rules is pro-cyclical.  

Figure 2: COVID-19 Cycle 

In 2020, the transitional regime of IFRS9 was further prolonged to take this into account. De-facto, its implementation was suspended. Once again, unexpected circumstances required suspending application of an element of the post-GFC reform program right after it was adopted. 

As for the case of calendar provisioning, whether the IFRS9 rules for NPLs can be revived as such after the pandemic is questionable. The new rules are inherently fragile because of the uncertainty of loss expectations. Even abstracting from that, undesired effects arise in a variety of circumstances, as soon as one departs from the textbook case of demand-driven cycles. Well-functioning accounting rules for NPLs need to be designed in a way to respond appropriately in all circumstances, so as to be robust from a macro-prudential perspective. This is a complex question, requiring further analyses which go well beyond the limited scope of this paper.  

 

4. Asset management companies

The idea of removing NPLs from eurozone banks and relegating them in an area-wide AMC was suggested while the ECB was still in the early phases of its NPL action plan. Though an AMC does not in itself necessarily involve mutualization of bank risks (this depends on how the scheme is designed), the proposal immediately faced opposition from some eurozone members, fearing that the proposal would allow countries with large amounts of legacy assets, preceding the launch of the single supervision, to offload part of the burden onto others.  

In 2018 the Commission, fulfilling a mandate given by the Council, issued a “blueprint” with criteria for member countries willing to set up their own, national AMCs^[14] See https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=SWD:2018:72:FIN. The document spelled out conditions for creating such schemes, making suggestions on various aspects including accounting, risk management, transfer pricing, impact on public finances and so on. The blueprint raised interest but as such was not applied, for several reasons. First, no explicit relaxation of state-aid criteria was included in the scheme, thereby limiting its feasibility for countries facing public finance constraints (countries with public finance problems often have also high NPL levels). Second, in the meantime the ECB supervision had advanced in its NPL action plan, and a more active secondary market for NPLs had developed. This allowed several banks in high-NPL countries to conclude important offload operations, alleviating the problem in the countries concerned. In the background, there was also a perception of stigma annexed to national AMCs, whose creation may in itself signal a systemic fragility in the banking sector of the country in question.  

The eurozone-wide AMC proposal was revived in 2020 by the ECB^[15]6 A. Enria, “The EU needs its own ‘bad bank’”; Financial Times, 27 October 2020 and echoed by the European Commission as part of its Covid strategy^[16]Coronavirus response: Tackling non-performing loans (NPLs) to enable banks to support EU households and businesses”; 16 December 2020. See … Continue reading. The Commission proposal, however, dropped the idea of an area-wide scheme arguing instead in favor of a “network” of cooperation, of unspecified content, among national AMCs.  

These new proposals, while still rather general, are of interest and should be carefully considered. An element in favor of them is that in the situation created by the pandemic the AMC solution is less prone to the criticisms that had plagued the proposal previously. NPLs derived from Covid cannot be regarded as a “legacy” of past errors by bankers or attributed to national supervisors, as had been the case in the past. These NPLs are the result of a common shock which hit all countries and was outside of their control. The underlying logic of the proposal is therefore stronger. 

Yet, there are hurdles in this new proposal as well. First and foremost, the entity of the problem is not known. The wave of Covid-related NPLs has not been observed yet; we do not know when it will develop, how large it will be, how it will be distributed across countries and banks. It seems unlikely that such scheme can be agreed upon, let alone implemented, before this information is available.  

Second, certain obstacles faced by the earlier proposals persist, to some extent. Even before Covid, an NPL problem still existed in certain countries and banks. Distinguishing between new, Covid-related losses and the preceding ones may not be easy in all cases. As a result, the objections raised in the past with reference to “legacy” problems may resurface. In addition, the “stigma” effect may still discourage certain countries from setting-up national schemes. The set-up of national “bad banks” could be regarded as a sign of underperformance in a broader sense, not only in dealing with banks including but also in the way the health situation has been handled or the supports to the economy have been provided.  

 

5. Conclusions

The wave of NPLs expected to develop in the eurozone as a consequence of Covis-19, while perhaps not too different in size from to the one observed after the financial crisis, is different in nature and will therefore require different remedies. Predictions are premature, because the phenomenon has not been observed yet. But it is already possible to make some reasoned conjectures on whether the regulatory tools put in place after the earlier crisis are going to be helpful in the new situation. 

The two main regulatory instruments introduced before the pandemic in the eurozone’s Pillar I structure for tackling the NPL problem, namely, the so-called “calendar provisioning” and the new accounting principles based on expected losses, are not suitable to deal with the new situation. Even prospectively, after the pandemic will be overcome, their usefulness in their present form is questionable, because either they are excessively rigid, or excessively sensitive to uncertainty, or both. Conversely, the proposals to create AMCs, at national or supranational level, are valid but cannot be seriously considered before the dimension of the post-Covid NPL problem is known.  

Absent these, traditional micro-supervisory instruments will continue to play a key role. One more time, the responsibility of cleaning eurozone banks from their NPLs will be predominantly fall on ECB supervision. Pillar II powers will have to be applied flexibly, depending on the conditions of individual banks. But when the moment comes, supervisory pressure should be exerted with determination, using all the independent power that the law and the statutes accord to the single supervision. Not an easy task; but the ECB has the instruments and the expertise necessary to carry it out. 

 

Notes

This draft is based on an intervention made on 11 February 2021 at the Global Annual Conference organized by the European Banking Institute in Frankfurt.

Authors

Howard Davies ^[1]Chair of the Natwest Group. Professor at Sciences Po, Paris.

 

“This time the banks are not part of the problem, as was the case in the financial crisis of 2008, but part of the solution” ^[2]Dividend payouts and share buybacks of global banks. ECB Financial Stability Review. May 2020. www.ecb.europa.eu . Such was the verdict of Felix Hufeld, then the President of the Bafin, the German financial regulator, in June 2020.

Hufeld himself has since moved on, a casualty of the Wirecard scandal, but his sentiment has been echoed by many regulators, commentators, and even some politicians who have been sparing in their praise of the banking sector in the past. It has even become something of a cliché, beloved of bankers themselves, who have enjoyed basking in the warmth of unaccustomed praise.

Bankers are human too (at least they like to think they are), so congratulations are always welcome, but some have been uncomfortably aware that these golden opinions may have come at a hefty price. Banks have been strongly encouraged, even required, to keep their branches open through the Covid lockdowns even when the footfall has been very light. They have given extended mortgage holidays to personal borrowers on demand. And they have extended loans to distressed companies, to help them through dips in demand, or even enforced closures. Some of those loans have been fully or partly guaranteed by governments, but it would be unrealistic to assume that the banks will not incur major losses on that and other lending. Some have been pushed into loss for 2020. And these losses come at a time when bank profitability is under serious threat from very low, or even negative interest rates. With yields on low risk assets almost flat as far as the analyst’s eye can see, the usual attractive and rewarding banking business of maturity transformation on an upwardly sloping curve has not been available.

That banks have survived this very difficult period can largely be attributed to the strong capital ratios they displayed when the crisis hit. Regulators have therefore taken some credit for the banks’ success. The aggressive re-regulation of the sector since 2008, led by the Basel Committee, has paid off. In spite of the sharpest recession for a century in the largest European economies, no significant bank has fallen over, or needed to be rescued by the state. And banks demonstrated remarkable operational resilience. The ECB acknowledges that there was no noticeable rise in operational losses due to business disruption or system failures. That is as remarkable as the strong capital position.

But in spite of this robust performance at a very challenging time the market has drawn the unsurprising conclusion that future profitability is uncertain and that bank stocks are to be treated with great care. Most large European banks have continued to trade at a significant discount to book value, well below 100% and systematically below their US counterparts in most cases, even though there was something of a rally in early 2021 (1).

Some might be tempted to think that if this is what it means to be a solution, maybe being a problem was not such a bad thing after all.

As we emerge from what we must hope to have been the worst of the pandemic, it is time to ask whether, from the banks’ perspective, anything has changed. Will politicians and regulators conclude that large banks, which many saw as dinosaurs, ready to be wiped out by agile digital fintech newcomers, or by the BigTech monoliths, have their social uses after all, and should not be allowed to vanish into the primeval swamp along with diesel engine plants and high street fashion retailers? Or will the Covid crisis be seen merely as a temporary respite in a process of secular decline?

To attempt an answer to that question we need to parse it a little, and address four sub-questions:

1. Can we expect the regulatory environment to change as a result, in ways that might benefit traditional banks?
2. Might the experience of the crisis, and the solidity the banks displayed, affect customer behaviour, and create a kind of ‘flight to safety’?
3. Has the crisis weakened some new competitors and demonstrated weaknesses in their business models?
4. Are banks therefore now in a stronger competitive position, or is their predicament fundamentally unchanged?

 

1. Regulation

The European regulators’ initial response to the Covid crisis was not encouraging from a bank perspective. In March 2020 both the European Central Bank and the Bank of England imposed restrictions on bank dividends, indeed they effectively banned any capital distributions during the whole of the year, to retain as much capital as possible within the banking system. The Federal Reserve did not take the same line, allowing normal dividends, typically accrued quarterly in the US, to continue, but did impose a moratorium on share buybacks, which in recent years have dominated US bank distributions.

The banks reacted negatively, arguing that their capital positions were strong enough to sustain normal dividends, and that preventing them from rewarding their shareholders would adversely affect investors’ views of the investability of bank stocks, thereby raising their cost of capital in the longer term. They pointed out that the ban was also inconsistent with the capital framework put in place since the crisis, with its higher ratios, buffers and rigorous stress tests.

By early 2021 there were signs that the regulators were beginning to soften their position, and allowing modest distributions to go ahead. The Bank of England revised its guidelines. The ECB allowed stronger banks to resume dividends within strict limits, noting that the average tier 1 capital ratio for the banks it supervised had risen from 14.4% at the start of 2020 to 15.2% at the end ^[3]Interview with Felix Hufeld. 3 June 2020. www.bafin.de . The revised rule was that dividends in 2021 should not exceed 15% of 2019-20 profits, or 20 bps of CET1 capital. Though the secretary of the Basel Committee, Carolyn Rogers, alarmed bankers (and some regulators alike) in November 2020 by arguing that the dividend ban should continue until the full extent of the covid hit to the economy was clear ^[4]Supervisory banking statistics. October 2020. www.bankingsupervision.europa.eu . That may take some time, as the pandemic rumbles on for longer than expected.

In other respects, however, the regulators were somewhat more helpful to the banks. The ECB implemented a series of relief measures, which were broadly paralleled by the Bank of England and others. They allowed, indeed encouraged banks to dip into their capital conservation buffers, and allowed some capital instruments which would not normally be counted towards pillar 2 requirements to be incorporated. The ECB revealed in January 2021 that nine banks, which would otherwise have fallen below its CET1 guidance, had taken advantage of that flexibility, though most have not needed to do so. The regulators also allowed the use of transitional IFRS 9 provisions, which somewhat reduced the procyclicality of the expected loss calculations. Banks could operate below the 100% liquidity coverage ratio until the end of 2021, and that may be extended. Furthermore, a series of other supervisory interventions were deferred or abandoned, notably the deadline for meeting the 2019 qualitative guidance.

But these transitional relief measures are specifically related to the crisis period, and there has been no suggestion from the ECB, or the Bank of England, that capital requirements will be relaxed in the longer term. Indeed the full implementation of Basel 3, to which the regulators are committed, would increase minimum capital for a number of institutions, putting further pressure on profitability, which is already challenged. As the ECB itself concludes: “Banks profitability and business model sustainability remain under pressure from the economic environment, low interest rates, excess capacity, low cost efficiency, and competition from banks and non-banks” ^[5]Quoted in the Financial Times. 17 November 2020. www.ft.com . They do not include high capital and liquidity requirements in that list of obstacles. While in the US there have been some signs of willingness to lighten capital requirements on small institutions in particular, there is no sign yet of a similar move in Europe.

The banks, while not requesting a major relaxation of the rules, have asked the ECB to rethink the remainder of the Basel 3 reforms, and invited the Commission to use its discretion to reduce the scale of the levy paid to the Single Resolution Fund. Both requests have so far been declined.

A recent report by the independent banking analyst at Autonomous has argued the capital rules for banks in the UK, and the same could certainly be said of banks in the Eurozone, are now arcane and in some respects dysfunctional. “The UK capital framework is creaking under the weight of its own complexity”, the author Christopher Cant maintains, and “the level of complexity is a deterrent for investors” ^[6]ECB supervisory priorities 2021. 28 January 2021. www.bankingsupervision.europa.eu . The stress testing arrangements are opaque, and there is still no clarity on the transitional arrangements for IFRS 9. There is uncertainty over the MREL and liquidity requirements. Overall, they conclude, “the scenario doesn’t exactly bode well for a rapid normalisation of dividends”.

There is another dimension of regulation, however, where change might be in prospect. For some time the banks have maintained that new digital competitors, whether small fintech start-ups or Bigtech giants, have benefited from lighter regulation in areas such as data usage and anti-money laundering, where banks seem to be held to higher standards. And there has been a bias towards promoting new competition, through forcing the opening up of banking relationships (open banking) and regulatory sandboxes, in which the regulators help new entrants to develop compliant business systems.

The response from regulators to date has been that the same activity is subject to the same regulation, and that most of these new entrants have chosen not to be banks, which brings obligations as well as rights.

There are signs that this line may be in the process of being rethought. A February 2021 paper ^[7]UK banks: creaky capital. Autonomous. 27 January 2021. www.autonomous.com by Fernando Restoy, of the Financial Stability Institute, a think tank linked to the Bank for International Settlements in Basel, questioned the current approach. Restoy notes that the ‘same activity, same regulation’ mantra is not accurate, and that incumbent banks have specific entity-based prudential and other obligations which do not facilitate a level playing field. He argues that ‘the growth potential of fintech and big tech companies could be, in part, the consequence of lighter regulatory requirements’. He goes on ‘regulation specific to banks entails higher compliance costs and can therefore put them at a competitive disadvantage’.

The policy implications of his analysis are intriguing. His main point is that while banks have argued that regulation should be activity-based to promote a level playing field, that may well not be the consequence, and that fintechs may ‘generate concrete threats to relevant policy objectives such as market integrity or stability or fair competition’. Those threats may create a case for entity-based regulation of these new entrants, which would achieve a better balance of policy objectives, and would in practice level what is now a very bumpy playing field.

It is too early to say whether this argument will influence key decision-makers in the European Commission, or elsewhere in the Tower of Basel for example, but the implications could be far-reaching.

It is possible, too, that payments initiatives led by central banks themselves will alter the competitive landscape. The most recent survey by the BIS shows that 86% of the central banks surveyed are working on their own digital currencies ^[8]Fintech regulation: how to achieve a level playing field. Fernando Restoy. February 2021. Financial Stability Institute Occasional Paper No. 17. www.bis.org . The gauntlet thrown down by Facebook’s Libra initiative, now dubbed Diem, has stung the central banks into a response. Depending on the nature of the response CBDCs could disintermediate commercial banks or strengthen them. The ECB has ^[9]Digital Currencies and the future of the monetary system. Remarks by Agustin Carstens, General Manager, Bank for International Settlements. Hoover Institution policy seminar. Basel 27 January 2021. … Continue reading suggested in a consultation paper that individuals should hold digital euros through their accounts at private sector banks. If they maintain that view commercial banks could find their position in the payments landscape reinforced.

So the incumbent banks robustness and resilience in the Covid crisis has pleased regulators, and there are signs that the nature of desirable competition may be under review. But in the long run customer preferences will be decisive. Has their performance paid dividends with customers?

 

2. Flight to Safety

The key lending support schemes for businesses affected by the covid crisis were backed by governments in various ways. But while that was true, lenders still needed the balance sheet strength to participate in the schemes. For the most part they took the view that, at least in the early stages, they would lend only to existing clients. Performing new ‘know your customer’ checks was almost impossible in the timescales involved. So businesses which had moved their business to challenger banks or peer to peer lenders faced a problem if those lenders could not extend their facilities rapidly.

Some of the new lenders – Tide is an example in the UK – were able to participate fully in the government schemes, but others had less balance sheet flexibility. There are no reliable data on how many companies were affected by the inability of their principal bank to extend further credit, but there is some anecdotal evidence. Alan McIntyre, head of Accenture’s global banking practice, commented, “Part of the fintech challenge is that in times of uncertainty and stress, traditional banks are seen as a safe haven. This partly reflects a flight to safety, as people hew closer to institutions with long track records that they judge more likely to survive an economic downturn” ^[10]Report on a digital euro. European Central Bank. October 2020. www.ecb.europa.eu.

How significant has this factor become? Have new competitors in the banking sector in fact lost share to the larger incumbents. The answer is not clearcut. A research note by Jeffries in July 2020 entitled “Will Corona kill the Digital-Only Challenger? ^[11]Alan McIntyre. Quoted in Tearsheet. 14 August 2020. www.tearsheet.com ”, focussing on the UK market, argued that “digital engagement has inflected back into the hands of large incumbents in the era of coronavirus”. Their evidence to back this claim showed that the rates at which customers were installing apps from large and small banks had begun to change in 2020. For some time the app share of challenger banks had been rising, but the trend changed in early 2020. The significance of this change of trend is disputed. Starling, a strong digital challenger, said “we simply do not recognise the picture outlined in this report”. It may also simply reflect an improvement in the digital offerings of the larger banks, rather than a lack of confidence in the stability of new entrants.

 

3. Competition

There are signs, however, that the competitive environment for the big banks may have become a little less intense. Some fintechs have struggled in the new landscape. While finance has remained available to fund the growth of the most promising and competitive, the implied equity valuations have fallen when new money has been raised. Some have withdrawn from markets in which they are marginal players. N 26 pulled out of the UK, for example, but the cost advantages of the new entrants which focus on payment services, with up to date technology and without the cost drag of large branch networks, remain strong. Both Monzo and Revolut have continued to grow their customer base, though profitability remains elusive.

And the societal and behavioural changes driven by lockdown restrictions may work to their advantage. Deloitte point out that “as social distancing has taken hold worldwide, there has been tremendous growth in the use of digital services and e-commerce ^[12]Will Corona kill the digital-only challenger? Jeffries equity research. July 2020. www.jeffries.com ”. The footfall in traditional bank branches has necessarily fallen, which may have the effect of reducing brand loyalty in the medium term. The number of bank branches in the EU fell by over 6% in 2019: the fall is likely to have been sharper in 2020. Deloitte’s conclusion, which is plausible, is that “an important outcome of COVID-19 for fintechs may well be the continued acceleration of partnerships with financial institutions, which can offer the benefits of capital, distribution access, and compliance infrastructure, but often lack highly sought-after digital solutions”.

Different considerations apply to the Bigtech companies, Apple, Google, Amazon and Facebook in particular. They can hardly be described as financially challenged. Their balance sheets are stronger than those of any major bank, and their market valuations are of a different order. Amazon’s market capitalisation in early February 2021 was around $1.7 trillion, compared to JP Morgan’s $420 billion.

The challengers and peer to peer lenders who offer credit face a different challenge. They will almost certainly experience a credit environment which will be far more hostile than they have encountered hitherto. I suspect some may be crushed under the wheels of an unforgiving credit cycle. There will be an element of chance in who survives and who does not. Those which had completed a funding round shortly before the crisis hit may well have the resources to ride out the storm. Others, who need more capital to grow (and many are still loss-making) will find new money harder to raise except on terms which may constrain their growth ambitions. Investors in peer to peer lenders have found it difficult to access their cash, with waits of several months at some providers ^[13]Beyond COVID-19. New opportunities for fintech companies. Deloitte Center for Financial Services. January 2021. www.deloitte.com . That is likely to constrain growth in the future as investors will be far more reluctant to fund them if they fear their money is locked up. Some have sought wholesale funding to replace the retail funds, which may guarantee short-term survival but will put pressure on margins in the longer run.

A continued shake-out in the challenger bank and peer to peer sectors seems very likely. But will that be enough to alter the competitive dynamics of the European banking sector, and return it to acceptable levels of profitability, with share prices at or above book value?

 

Are banks now stronger?

Generalisations about the prospects for European banks are hazardous. Some large banks, especially those in Scandinavia, have remained acceptably profitable throughout the last difficult decade. They have achieved low cost-income ratios, maintained strong market positions and innovated successfully and repeatedly. Their reputations have remained strong, too, though in some cases tarnished through money-laundering problems. But, on average, large European banks have found it difficult to earn their cost of capital.

Looking forward, the most decisive influence will be the level and shape of the yield curve. That in turn will be influenced ultimately by the supply of and demand for investment funds. The central banks will not raise rates to rescue the profitability of the banking sector. Negative interest rates will make the problem more severe for banks, as it is both technically and presentationally difficult to charge negative rates to retail customers who have the opportunity to switch money holdings into cash. The ECB has tried to mitigate the impact of very low rates on the banks, with mixed success. They may continue to do so, as may the Bank of England if it also imposes negative rates. In February 2021 they asked the banks to prepare for that eventuality.

When challenged about the viability of the banking sector the ECB typically points to a lack of concentration, and high costs, suggesting that many of the remedies lie in the hands of the banks themselves. In 2016, for example, Mario Draghi said: “Overcapacity in some national banking sectors, and the ensuing intensity of competition, exacerbates this squeeze on margins ^[14]Peer to peer lending. Martin Lewis and Amy Roberts. 4 February 2021. www.moneysavingexpert.com ”. How valid is this argument, and what scope is there for further bank consolidation in Europe?

On a conventional definition, concentration in EU banking seems quite high. On average the top 5 banks per country have 65% of the market as defined by balance sheet size, with the range running from 28 to 97% ^[15]Speech by Mario Draghi . Frankfurt 22 September 2016. www.reuters.com . But the ECB have attempted a more sophisticated analysis to try to determine what we mean by overcapacity in the banking sector, and where it is present.The research ^[16]EU structural financial indicators: end of 2019. 8 June 2020. www.ecb.europa.eu identifies three overlapping definitions of overcapacity. The first is size, measured by bank assets as a percentage of GDP, and as a percentage of the whole financial sector. The second is the intensity of competition. As proxies they use the number of banks per 100,000 inhabitants, the concentration ratio and also measures of Net Interest Margin and Return on Assets. The third dimension they call “Infrastructure/efficiency” which includes a basket of measures such as the number of people per bank branch, customer deposits per branch and total assets per bank employee. From these three components they construct a composite indicator of overcapacity.

The methodology may be open to criticism, and the composite measure involves a degree of subjective judgement on the weights to be attached to individual factors. But the results are intuitively reasonable. They show that those Scandinavian countries where returns on equity, and price to book ratios, are healthy, show low volumes of overcapacity. At the other end of the European scale Germany, Austria, France and Italy have relatively more overcapacity. As the authors point out, ‘the banking systems of these countries are often characterised by the traditionally strong role of savings and cooperative banks, and, thus, a high number of banks, lower degree of concentration and an extensive physical infrastructure”.

Where that is the principal reason for overcapacity it is not easy for private sector banks to solve the problem Draghi identified. There are countries where consolidation is possible, and there has been some recent activity in Spain and Italy, but the analysis suggests that different approaches are needed in different countries. In some cases progress can be made through conventional efficiency improvements, such as branch closures. In others exit of some players may be needed. These are controversial and time-consuming changes.

Pre-crisis, the ECB’s solution was threefold: reductions in Non-Performing Loans, for those still with high stocks of such loans, in-market consolidation by weak-performing small banks and a combination of bank-level restructuring and cross-border M&A activity for poor performers among the large banks ^[17]Overcapacities in banking: measurements, trends and determinants. Sandor Gardo and Benjamin Klaus. Occasional Paper No. 236. November 2019, www.ecb.europa.eu . The first option now looks harder to achieve. In-market consolidation is difficult but not impossible and the crisis may give those efforts a boost, as we have seen in some cases. But significant cross-border consolidation looks as far off as ever, for cultural, political and regulatory reasons. In 2018 bank M&A activity in Europe was lower than at any time this century ^[18]Euro area bank profitability. Where can consolidation help? Deislava Andreeva, Maciej Grodzicki, Csaba More and Alessio Reghezza. ECB Financial Stability Review November 2019. www.ecb.europa.eu . Andrea Enria, the Chairman of the ECB’s Supervisory Board, has acknowledged that countries are still ringfencing liquidity and capital at the national level, which means that limited benefits emerge from operating across borders.

 

Conclusions

One conclusion from this review might be that nothing fundamental has changed.

Banks with high costs and weak positions in slow-growing markets remain as challenged as before. Indeed the likely resurgence of NPLs, which had been declining for several years, will make their dilemma sharper.

The interest rate prospect, from a bank’s perspective at least, has become even more pessimistic. The prospect of strongly positive real interest rates has retreated further into the future.
The attractiveness of new digital competitors in the payments arena, unburdened by the legacy costs of unwieldy technology stacks, remains strong.

But that conclusion does require some qualification. Politicians and regulators have seen that the financial re-regulation they oversaw since 2008 has indeed delivered a banking sector which is robust, even in a sudden and unparalleled economic crisis delivered by the pandemic. Over time, that will reduce the pressure for ever higher capital ratios, which were in prospect before the crisis hit. They have seen that strong bank balance sheets are a highly valuable asset at times when the private sector needs credit and liquidity support on a massive scale, and that bank systems can deliver sharply higher volumes of activity very quickly. As a result, the reputation of banks, and trust in bankers, have risen, after a long period in which the latter were languishing near the bottom of the trust league, along with politicians and journalists. That reputational benefit does not translate into an enhanced return on equity in the short term but it will have a value over time.
We have also seen that non-bank credit provision can have fragile foundations, causing some business customers to appreciate the value of a solid banking relationship more. That may also translate into business opportunities in the longer run.

But the pressures on banks to reduce cost income ratios, to focus on business areas where they have a defensible market position, to control NPLs and to upgrade their technology to compete effectively with new competitors will remain intense. Covid is not going to offer the banks a ‘get out of gaol card’ but some of the more fanciful predictions of the death of banking may need to be revised. In 1997 Bill Gates said “We need banking. We don’t need banks any more” ^[19]Quoted in “Bye Bye banks” www.jessleitch.co . It is fortunate for the global economy that this is one of his predictions which did not come true.

Authors

David Bholat^[1]Bank of England, Oliver Thew^[2]Bank of England, and Mohammed Gharbawi ^[3]Bank of England

 

Abstract

The Covid-19 crisis continues to have a profound effect on the financial sector, with firms reassessing and adapting strategies, business models, and investment plans. Technological transformation is likely to be a significant part of this adjustment and early evidence from a survey conducted by the Bank of England suggests that the use of machine learning (ML) and data science (DS) could have an increasingly important role to play in these shifts. The technological, financial, and social changes wrought by the pandemic have also compelled businesses across the economy to look for opportunities in using different processes, developing different products, and exploring different markets. This paper looks at some of the early signs of what those changes are likely to be and how banks are responding. 

 

The use of ML and DS in UK banking before Covid

 

Recent trends in ML and DS

Over the past two decades, digitalisation of society and the economy has generated vast amounts of data (WEF, 2019). DS has therefore become an increasingly important tool for businesses looking to capitalise on data-driven insights (McKinsey & Company, 2019a). This has also led to the increased use of ML across a range of businesses and sectors (McKinsey & Company, 2019b), including finance (Centre for Data Ethics and Innovation, 2020), which has seen widespread adoption of ML and DS in recent years. In 2019, The Bank of England conducted a joint survey (Bank of England, 2019) with the Financial Conduct Authority (FCA) to understand how ML was being used in UK financial services. The survey showed that ML was already being used by a majority of firms across a range of financial sub-sectors and business lines (Chart 1).

 

Chart 1: Two thirds of respondents have ML applications in use

Source: Bank-FCA (2019), Machine learning in UK financial services.

 

Banking was the sub-sector with the highest number of ML applications and second highest share of ML applications relative to the number of survey respondents. The two most prominent uses were customer engagement and risk management. For a majority of the banks surveyed in 2019, the use of ML had matured to the point where it was being deployed in the regular run of operations. Moreover, the majority of banks expected the number of ML applications to triple by 2021(Chart 2).

 

Chart 2: Banks expect significant growth in use of ML

Source: Bank-FCA (2019), Machine learning in UK financial services.

Other regulatory authorities have reported similar findings. For example, in 2019, Canada (Bank of Canada, 2019) and Hong Kong (HKMA, 2019) reported similar increases in the importance of ML and its adoption by banks in their respective jurisdictions.

These pre-Covid patterns in banks’ use of ML and DS were backed-up by a survey conducted by the Economist Intelligence Unit in February and March 2020 (The Economist, 2020b), as well as a study published in January 2020 by the Cambridge Centre for Alternative Finance (CCAF) and the World Economic Forum (WEF) (Ryll, et al., 2020). The CCAF and WEF surveyed 151 fintech start-ups and incumbent firms across 33 countries. They found that 85% of respondents already used some form of AI, most commonly in risk management, 65% expected to use AI in three or more business areas within two years, and 77% anticipated that AI would have high or very high overall importance to their business by 2022 (Chart 3).

 

Chart 3: Before Covid financial firms expected AI to become strategically more important by 2022 
Current and expected strategic importance of AI to firms (surveyed pre-Covid)

Source: CCAF and WEF (2020), Transforming Paradigms: A Global AI in Financial Services Survey.

 

Benefits for households, banks and the economy

ML and DS have wide-ranging applications in financial services, which can bring benefits to consumers, businesses and the economy. For example, many banks use ML and DS for anti-money laundering (AML) processes (Delle-Case et al., 2018). In many instances, this has reduced the rate of false positives in money laundering detection,^[4]False positives are notifications of potential suspicious payments or financial activity that do not end up resulting
in the filing of a suspicious activity or suspicious transaction report. with one large UK bank lowering its false positives by 70% (IBM, 2019). For consumers, this helps reduce the number of erroneously blocked or delayed payments. For banks, this frees up scarce resources and speeds up internal processes. For the economy as a whole, this can help banks and authorities more precisely identify illicit financial activity.

ML and DS also have the potential to provide more inclusive and tailored products to consumers. For example, ML is already being used by banks and fintech companies around the world to analyse newer data sources (such as social media data) to provide risk assessments of individuals with limited credit histories, which might help underserved or unbanked customers access financial services (Ryll, et al., 2020). Some UK fintechs (Holmes, 2020) and banks (McKinsey & Company, 2020) are using new data sources for consumer and business risk assessments. This trend looks set to continue with one credit rating agency announcing plans to offer UK banks access to a broader range of transactional data for consumer credit scores, including money earned and spent, council tax payments, savings and investments, and subscription payments (Business Insider, 2020).

 

Risks and challenges

At the same time, existing risks may increase and new risks may emerge from the use of ML and DS in financial services. Respondents to the Bank-FCA survey and a similar report from the UK’s Centre for Data Ethics and Innovation note that risks may increase due to ML’s lack of explainability (the so-called ‘black box’ problem), meaning the outputs cannot always be easily understood (Bundy et al., 2019; CDEI, 2020). In addition, ML models may perform poorly when applied to a situation they have not encountered before in the training data. This is particularly relevant in the context of the Covid pandemic when the underlying data may have changed (data drift) or the statistical properties of the data may have changed (concept drift) (Robotham, 2020; Ma and Jarrett, 2020, respectively).

These risks could materialise at an individual bank or system-wide level. Systemic risks are particularly concerning as they can create financial instability, which can in turn adversely affect the real economy and the prosperity of households and businesses. Therefore, regulators and central banks have an interest in understanding how ML and DS are being deployed and managed

 

The impact of Covid on ML and DS in UK banking

To better understand the impact of Covid on ML and DS in the UK banking sector, the Bank of England conducted a survey of Prudential Regulation Authority (PRA) regulated banks in August 2020.^[5]The survey consists of 32 submissions in total, with 17 from UK banks, nine from foreign banks with operations in the UK, and six from insurers. The sample of insurers was too small to be judged … Continue reading The survey focused on banks’ perception of ML and DS, as well as the resourcing for current and planned ML and DS projects.

Around 40% of respondents reported an increase in the importance of ML and DS for future operations, and a further 10% of banks reported a large increase. None of the banks reported a decrease in the importance of ML and DS. This is an unexpected finding given the suggestion from some commentators that a new ‘AI winter’^[6]An AI winter is shorthand for a time when interest and investment in AI wanes, for example, as occurred in the
early 1970s (Frankenfield, 2020). might unfold in as a result of reduced investment budgets due to the economic impact of Covid or because pre-pandemic ML systems may not have performed well (The Economist, 2020).

 

Chart 4: Half of banks view ML and DS as more important for future operations since Covid 
Impact of Covid on banks’ plans for, and current use of, ML and DS

Source: BOE (2020), ML, DS and Covid survey.

 

Around a third of banks said there was an increase in the number of ongoing ML and DS applications. Yet only 16% of banks reported an increase in funding and/or resourcing for existing applications and a similar number reported a decrease. Similarly, around 35% of banks reported an increase in the number of planned applications. But only 23% of banks reported an increase in funding and/or resourcing for planned applications and 12% of banks reported a decrease.

Banks may be looking to use ML and DS to increase efficiency and improve digital customer channels as they manage the cost and revenue impact of Covid. The crisis has accelerated use of ML-powered tools to manage an unprecedented uptick in customer enquiries (Motsi-Omoijiade, 2020). Half of the banks in the survey reported a ‘positive’ impact on plans for customer engagement applications. Around a third of banks also reported a ‘positive’ impact on planned investment in internal operations and financial crime applications. As the 2019 Bank-FCA survey found, ML models have already been used in all three of these areas.

 

Chart 5: Banks plan to invest more in ML and DS across a range of business areas due to Covid
Impact of Covid on planned investment by use case

Source: BOE (2020), ML, DS and Covid survey.

 

The overall planned investment picture is largely similar for all banks in the survey, with UK- headquartered banks having slightly more positive expectations. More specifically, nearly 60% of banks headquartered in the UK reported that Covid has had a ‘positive’ impact on planned investment in customer engagement applications. Similarly, almost half of these banks noted the ‘positive’ impact on planned investment in DS and ML applications in credit (including origination and pricing), with 29% of the banks reporting a large positive impact. One reason could be the use of ML to deal with the high volume of customer enquiries (Motsi-Omoijiade, 2020) and government guarantee loan applications (Hinchliffe, 2020).^[7]There were more than 1.6 million applications for the Bounce Back Loan Scheme, 159,277 applications for the Coronavirus Business Interruption Loan Scheme and 1,034 applications for the Coronavirus … Continue reading These banks may also use ML and DS as they look to refine expected credit loss calculations in line with the IFRS 9 accounting regulation.^[8]Expected credit loss calculation under IFRS 9 involves the definition of forward-looking scenarios to derive provisioning. The extreme nature of the Covid shock has meant that these forecasts have … Continue reading

The survey shows that around 35% of banks reported that ML and DS had a ‘positive’ impact on technologies that support remote working among employees. The same percentage also reported a positive impact on their overall risk appetite for ML projects, meaning these banks are more willing to use these techniques. At the same time, around 35% of banks reported a negative impact on ML model performance with just 8% reporting a positive impact. This is likely because the pandemic has created major movements in macroeconomic variables, such as rising unemployment and mortgage forbearance, which required ML (as well as traditional) models to be recalibrated. Other areas where banks noted a negative impact were in ‘resourcing’ and in ‘hiring/retention of skilled staff’.

 

Chart 6: Covid had a net negative impact on model performance
Issues (opportunities and risks) encountered by existing applications as a result of Covid

Source: BOE (2020), ML, DS and Covid survey.

 

It is important to note that while Chart 6 indicates where net ‘positive’ or negative effects are felt, the numbers do not tell us the extent of these effects, beyond small or large, nor indeed how they may impact banks’ business models or financial performance. More research is needed to gauge how material the affected ML/DS models are to banks’ overall performance, operations and risk profile, and hence the overall impact of the crisis.

Finally, there were marked differences between small and large UK^[9]The PRA divides all deposit-takers it supervises into five ‘categories’ of impact. ‘Large banks’ here refers to the Category 1 banks, namely, the most significant deposit-takers whose size, … Continue reading banks with respect to their use of third-party vendor products and services. Chart 7 shows that smaller banks reported a ‘positive’ impact (eg in terms of performance, impact, use) of Covid on all categories of DS and ML, with data collection, and model testing and validation being the areas with the largest ‘positive’ impact. Large UK banks reported a ‘positive’ impact on use of outsourced platforms and infrastructure. These findings are in line with market intelligence that smaller banks are looking to increase their use of off-the-shelf ML products. This stands to reason given the generally more substantial in-house data and analytical capabilities of large banks.

Alongside the usual risks associated with outsourcing, the use of ML and DS can pose additional risks and challenges (PRA, 2021a). For example, outsourced ML models may be more difficult to interpret because detailed knowledge in terms of how they were developed resides outside the bank. This can make it more difficult for banks to understand how the model works and to monitor performance, which could result in unexpected or unexplained performance, and risks materialising. If multiple banks use the same third-party provider and ML model, this could also potentially lead to an increase in herding, concentration and even the possibility of systemic risks where methodologies are common.

 

Chart 7: Covid had a positive impact on outsourcing and the use of third-party providers by large banks

Source: BOE (2020), ML, DS and Covid survey.

 

Explaining the survey findings

Prior to the survey, we expected that UK banks’ investment in ML and DS in response to Covid might follow the same historical pattern as other business investments during an economic downturn. Most businesses tend to respond to negative macroeconomic shocks by reducing expenditure, including spending on investment and innovation (Archibugi, 2013). In this way, business investment is typically pro-cyclical (Younes et al., 2020), rising in upswings and falling in downturns (Barlevy, 2007).

A major reason businesses reduce investments in innovation during economic downturns is the need to prioritise near-term cash flow rather than long-term technology projects. Businesses may become increasingly hesitant to invest in long-term capabilities when revenues are declining, and when there is higher uncertainty around future profits. There is plenty of evidence to suggest that uncertainty has increased during the pandemic (Altig et al., 2020). Our Decision Maker Panel survey (DMP, 2020), designed to be representative of the population of UK businesses, found that 70% of firms viewed overall economic uncertainty as high or very high in August 2020 when we conducted our survey.

Yet the Covid survey shows that banks’ investment and interest in ML and DS has held up. The strategic imperative to drive efficiency through automation and has perhaps been reinforced by the low interest rate environment. Furthermore, the nature of this shock means that demand for banking and other financial services may not have suffered to the same degree as other industries like hospitality (Office for National Statistics, 2020), given the extent and impact of lockdown measures.

The pandemic has also catalysed more extensive use of computers and smartphones for commerce, remote working and socialising (Kemp, 2020). This has likely increased the amount of data businesses have available to them. This in turn is likely to increase demand for data scientists, data engineers and other IT professionals (CDEI, 2020). Ultimately, if necessity is the mother of all invention, then Covid has arguably accelerated demand for data and technical innovation (Taalbi, 2017).

 

Banks have benefited from ML during the pandemic

In March 2020, the Bank of England put in place a package of measures to help mitigate the economic shock resulting from Covid (BoE, 2021). The UK Government also provided a range of financial support for businesses, including government-guarantee loan schemes. As noted earlier, some UK banks used ML (Temenos, 2020) to process the high volume of government guaranteed loan applications (Curtis, 2020), resulting in increased operational efficiency.

As the emphasis was on providing finance to businesses quickly during the early stages of the pandemic, lenders were given a 100% government guarantee on Bounce Back Loans, and borrowers could apply in a streamlined process with no assessment of their creditworthiness. Market intelligence suggests that banks are now using ML to enhance their credit risk management and to help identify and manage higher risk loans within certain portfolios, some of which may be expected to have higher default rates compared to other loan portfolios (NAO, 2020).

 

Covid may amplify certain risks associated with ML

As the survey highlights, Covid has had a negative impact on the performance of some ML models. This is linked to the fact that ML models’ performance can change or deteriorate under conditions different to those displayed in the data on which they were originally trained. This can occur either when the underlying data changes (data drift) or the statistical properties of the data change (concept drift). The Covid crisis has led to both data and concept drift, which has challenged models in unusual and unexpected ways. Therefore, monitoring for data drift and concept drift is one of the key challenges for firms to ensure appropriate risk management.

Our survey also showed that small banks have increased their use of third-party providers of data, infrastructure, and off-the-shelf or bespoke ML models as a result of Covid. As previously mentioned, while there are many advantages to outsourcing and third-party provider models, they can carry additional operational risks that may be amplified as banks seek to integrate new ML applications into existing legacy IT systems PRA (2021b).

What next for ML and DS?

The repercussions of the Covid crisis, including its impact on ML and DS trends in financial services, will likely be with us for many years to come and firms across the sector, including banks, are reappraising the use of data-driven technologies to augment revenue streams, refine cost reduction programmes, and enhance risk management processes. Against a potential post-pandemic background of persistently low interest rates, increasing competition, and subdued economic conditions, the opportunities to grow revenues in retail and commercial banking may be limited. The focus would therefore remain on cost containment or cost reduction as the main drivers of profitability. This implied increase in efficiency and productivity is likely to be achieved primarily through technological transformation, including the adoption and use of ML and DS.

Technological transformation carries with it significant execution risk and the operational risks run by firms with ambitious automation programmes may well increase in the short term. Firms will also need to keep a keen eye on the skills base necessary to ensure that the technical and cultural aspects of change are managed effectively and appropriately.

The directions in which ML and DS will move over the next few years are dictated by an evolving set of factors propelled and accelerated by the pandemic. These could be at the technological level, with ever increasing use of alternative data, cloud, or off-the-shelf ML solutions; the firm level, with more automation given additional impetus by virtual and agile working patterns; and the societal level, where regional and demographic variations may push firms to use ML and DS in devising more localised and narrowly targeted products.

The pandemic has also altered, perhaps permanently, strategic priorities, objectives, and plans. It has made it much easier to argue for more decentralised business models and organisational structures. ML and DS are key components in managing an effective distributed and networked business.

 

References

Altig, D., Baker, S., Barrero, J.M., Bloom, N., Bunn, P., Chen, S., Davis, J.D., Leather, J.,  Meyer, B., Mihaylov, E., Mizen, P., Parker, N., Renault, T., Smietanka, P., and Thwaites, G. (2020). Economic uncertainty before and during the Covid-19 pandemic. Staff Working Paper No. 876. Available at: Bank of England Staff Working Paper No. 876 (Accessed on April 21, 2021).

Archibugi, D., Filippetti, A., and Frenz, M. (2013). The impact of the economic crisis on innovation: Evidence from Europe. Technological Forecasting and Social Change, 80 (7), 1247-1260.

Bank of Canada (2019). Financial System Survey Highlights. May 2019. Available at: Financial System Survey Highlights—May 2019 – Bank of Canada (Accessed on April 21, 2021).

Bank of England (BoE) (2019). Machine learning in UK financial services. Financial Conduct Authority. October, 2019. Available at: Machine learning in UK financial services (bankofengland.co.uk) (Accessed on April 21, 2021).

Bank of England (BoE) (2020). Our response to coronavirus (Covid). Available at: Our response to coronavirus (Covid) | Bank of England (Accessed on April 21, 2021).

Barlevy, G. (2007). On the cyclicality of research and development. American Economic Review, 97 (4), 1131-1164.

Bundy, A., Crowcroft, J., Ghahramani, Z., Reid, N., Weller, A., McCarthy, N., and Montgomery, J. (2019). Explainable AI: the basics. The Royal Society. Available at: Explainable AI: the basics (royalsociety.org) (Accessed on April 21, 2021).

Business Insider (2020). Experian brings its open banking-powered credit score tool to the UK. Available at: Experian Debuts Credit Score Improving Tool in the UK (businessinsider.com) (Accessed on April 21, 2021).

Centre for Data Ethics and Innovation (CDEI) (2020). CDEI AI Barometer. Available at: CDEI AI Barometer – GOV.UK (www.gov.uk) (Accessed on April 21, 2021).

Curtis, J. (2020). £2bn of ‘bounce back’ loans for SMEs approved in first 24 hours. CityAM. Available at: £2bn of bounce back loans for SMEs approved in first 24 hours : CityAM (Accessed on April 21, 2021).

Decision Maker Panel (DMP) (2020). Monthly Decision Maker Panel data. Bank of England. August 2020. Available at: Monthly Decision Maker Panel data – August 2020 | Bank of England (Accessed on April 21, 2021).

Delle-Case, A., Bailey, N., Carr, B., and Ekberg, M. (2018). Machine Learning in Anti-Money Laundering. Institute of International Finance. Available at: https://www.iif.com/portals/0/Files/private/32370132_iif_machine_learning_in_aml_-_public_summary_report.pdf (Accessed on April 21, 2021).

Hinchliffe, R. (2020). Temenos helps UK banks speed up BBLS distribution “in days”. Fintech Futures. Available at: Temenos helps UK banks speed up BBLS distribution “in days” – FinTech Futures (Accessed on April 21, 2021).

HM Treasury (2021). HM Treasury coronavirus (COVID-19) business loan scheme statistics. Available at: HM Treasury coronavirus (COVID-19) business loan scheme statistics – GOV.UK (www.gov.uk) (Accessed on April 21, 2021).

Holmes, C. (2020). Credit scoring: Going alternative. Finextra. Available at: Credit scoring: Going alternative (finextra.com) (Accessed on April 21, 2021).

Hong Kong Monetary Authority (2019). Report on Artificial Intelligence (AI) Application in Banking. Available at: https://www.hkma.gov.hk/eng/news-and-media/press-releases/2019/12/20191223-4/ (Accessed on April 21, 2021).

IBM (2019). Fighting financial crime with AI. How cognitive solutions are changing the way institutions manage AML compliance, fraud and conduct surveillance. Available at: WKLQKD3W (ibm.com) (Accessed on April 21, 2021).

IFRS (2021). IFRS 9 Financial Instruments. Available at: IFRS – IFRS 9 Financial Instruments (Accessed on April 21, 2021).

Kemp, S. (2020). Digital 2020: July global statshot. Datareportal. Available at: Digital 2020: July Global Statshot — DataReportal – Global Digital Insights (Accessed on April 21, 2021).

Ma, V., and Jarrett, D. (2020). The COVID-19 Concept drift, using Sydney ferry activity data. Available at: The COVID-19 Concept drift, using Sydney ferry activity data | by Drew Jarrett | Towards Data Science (Accessed on April 21, 2021).

McKinsey & Company (2019a). Catch them if you can: How leaders in data and analytics have pulled ahead. Available at: https://www.mckinsey.com/business-functions/mckinsey-analytics/our-insights/catch-them-if-you-can-how-leaders-in-data-and-analytics-have-pulled-ahead (Accessed on April 21, 2021).

McKinsey & Company (2019b). Global AI Survey: AI proves its worth, but few scale impact. Available at: Survey: AI adoption proves its worth, but few scale impact | McKinsey (Accessed on April 21, 2021).

McKinsey & Company (2020). Managing and monitoring credit risk after the COVID-19 pandemic. Available at: Credit risk after COVID-19 | McKinsey (Accessed on April 21, 2021).

Motsi-Omoijiade, I. (2020). How AI is helping banks meet the challenges of COVID-19. Birmingham Business School Blog. Available at: How AI is helping banks meet the challenges of COVID-19 – Birmingham Business School Blog (bham.ac.uk) (Accessed on April 21, 2021).

National Audit Office (NAO) (2020). Investigation into the Bounce Back Loan Scheme. Available at: Investigation into the Bounce Back Loan Scheme – National Audit Office (NAO) Report (Accessed on April 21, 2021).

Office for National Statistics (2020). Coronavirus and the impact on output in the UK economy: September 2020. Available at: Coronavirus and the impact on output in the UK economy – Office for National Statistics (ons.gov.uk) (Accessed on April 21, 2021).

Prudential Regulation Authority (PRA) (2021a). PS7/21 | CP30/19 Outsourcing and third-party risk management. Policy Statement 7/21 | Consultation Paper 30/19. Available at: PS7/21 | CP30/19 Outsourcing and third party risk management | Bank of England (Accessed on April 21, 2021).

Prudential Regulation Authority (PRA) (2021b). PS6/21 | CP29/19 | DP1/18 Operational Resilience: Impact tolerances for important business services. Policy Statement 6/21 | Consultation Paper 29/19 | Discussion Paper 1/18. Available at: PS6/21 | CP29/19 | DP1/18 Operational Resilience: Impact tolerances for important business services | Bank of England (Accessed on April 21, 2021).

Robotham, P. (2020). Why your models might not work after Covid-19. Available at: Why your models might not work after Covid-19 | by Patrick Robotham | Eliiza-AI | Medium

Ryll, L., Barton, M.E., Zhang, B.Z., McWaters, J., Schizas, E., Hao, R., Bear, K., Preziuso, M., Sege, E., Wardrop, R., Rau, R., Debata, P., Rowan, P., Adams, N., Gray, M., Yerolemou, N. (2020). A Global AI in Financial Services Survey. University of Cambridge and World Economic Forum. Available at: Transforming Paradigms – CCAF publications – Cambridge Centre for Alternative Finance – Centres – Faculty & research – CJBS (Accessed on April 21, 2021).

Taalbi, J. (2017). What drives innovation? Evidence from economic history. Research Policy, 46 (8), 1437-1453.

Temenos (2020). Temenos Helps UK Banks Respond Rapidly to Surging Demand for Bounce Back Loans During Covid-19. Available at: Temenos Helps UK Banks Respond Rapidly to Surging Demand for Bounce Back Loans During Covid-19 – Temenos (Accessed on April 21, 2021).

The Economist (2020a). An understanding of AI’s limitations is starting to sink in. Available at: An understanding of AI’s limitations is starting to sink in | The Economist (Accessed on April 21, 2021).

The Economist (2020b). Forging new frontiers: advanced technologies will revolutionise banking. Available at: Forging new frontiers: advanced technologies will revolutionise banking – The Economist Intelligence Unit (EIU) (Accessed on April 21, 2021).

World Economic Forum (2019). How much data is generated each day? Available at: How much data is generated each day? | World Economic Forum (weforum.org) (Accessed on April 21, 2021).

Younes, G.A., Ayoubi, C., Ballester, O., Cristelli, G., van den Heuvel, M., Zhou, L., Pellegrino, G., de Rassenfosse, G., Foray, D., Gaule, P., Webster, E.M. (2020) COVID-19: Insights from Innovation Economists SSRN Working Paper No. 3575824. Available at SSRN: https://ssrn.com/abstract=3575824 or http://dx.doi.org/10.2139/ssrn.3575824 (Accessed on April 21, 2021).

Authors

Nicole Jonker^[1]De Nederlandsche Bank (DNB), the Netherlands., Carin van der Cruijsen^[2]De Nederlandsche Bank (DNB), the Netherlands., Michiel Bijlsma^[3]SEO Amsterdam Economics, the Netherlands. ^[4]Tilburg University, the Netherlands., Wilko Bolt^[5]De Nederlandsche Bank (DNB), the Netherlands. ^[6]Vrije Universiteit, Amsterdam, the Netherlands.

 

Abstract

COVID-19 has affected almost every aspect of our daily lives including the way we pay. The pandemic caused a large drop in the use of cash triggering more contactless payments at the point of sale. The question comes up whether this change in payment usage is temporary or permanent. We use payment diary survey data to study the shift in payment behaviour and payment preferences. Since the start of the lockdown in the Netherlands the usage of debit card versus cash has increased by 10 percentage points. The initial drop in cash usage by 17 percentage points has only partly been reversed. The reversal occurred especially with people aged 65 and above as well as with people with a low-income. Thus the shift appears to be long-lived. Moreover, the pandemic has also resulted in a shift in payment preferences. People who used to prefer electronic payment have shifted to contactless payment. The percentage of people preferring cash only slightly decreased from 21% to 20%.

 

1. Introduction

The COVID-19 pandemic is affecting almost every aspect of our daily lives. The pandemic has not only made the way we live more contactless but also the way we pay. Although lockdowns have led to a sharp fall in consumer spending, electronic payment instruments at the point of sale (POS) became more attractive relative to cash by avoiding close physical contact with the cashiers. Retailers promoted the usage of contactless payments at the expense of cash, and banks made it easier and more convenient for consumers to pay contactless. As a result, electronic payment instruments gained further ground.

In this article we illustrate the impact of the COVID-19 pandemic on consumer payment behaviour and payment preferences using unique payment diary data collected among a representative panel of Dutch consumers. This payment diary data includes information both on cash payments and electronic POS payments. Moreover, in addition to payment usage information, our diary data also provides useful information on payment preferences. The daily data used in this paper covers the Netherlands and ranges from January 1 2019 until December 31 2020. The nature of the data allows us to examine whether the effects of the outbreak of COVID-19 and its associated measures led to a shift in payment behaviour and payment preferences. If the shift increases adoption by forcing consumers to incur learning cost and breaking cash habits, or if the shock leads to a change in preferences, the ensuing change in payment behaviour is expected to persist. This would then create extra incentives for commercial banks and other third parties to speed up the provision of contactless, mobile and online payment services.

The Netherlands offers a particularly good setting to analyse the impact of the COVID-19 pandemic and its associated containment measures on payment behaviour and preferences. First, cash and debit cards are – de facto – the only two payment instruments that matter at the point of sale. Payment choice in the Netherlands is effectively a binary choice; credit cards, store value cards or checks hardly play any role for shopping at the point of sale. In 2019, 32% of POS payments were in cash, 24% by debit card with PIN and 43% contactless (i.e. a total of 99%; DNB 2020a). Moreover, paying contactless by debit card happens much more often than paying contactless by smartphone, as 90% of all debit cards are contactless-enabled (DNB 2020a).

Second, since the pandemic started half-March 2020, Dutch government, commercial banks and merchants have been taking measures to limit and contain the spread of the virus. Among others, this facilitated easier and more convenient contactless payments. Pre-COVID-19, consumers were required to enter their PIN code when they made a payment of more than EUR 25 and to insert their payment card into the payment terminal. If payments of EUR 25 and below reached a cumulative limit of EUR 50 the PIN code was also required. In 2020 the cumulative limit was (temporarily) increased to EUR 100 on March 19, while the transaction limit was raised to EUR 50 on March 24.

Third, merchants stimulated people to pay contactless as it lowers the likelihood of hand contact between customers and cashiers. For example, they applied doorplates and notices next to the cash counter asking and steering people to pay contactless. Moreover, the Centraal Bureau Levensmiddelenhandel (CBL) – the Dutch organisation that looks after the interests of supermarkets – appealed to consumers to pay contactless.^[7]See the press release https://www.cbl.nl/pinnen-als-voorzorgsmaatregelen-tegen-coronavirus/ (in Dutch). Also the WHO (2020) has been advising to not use paper tender and to use as many cashless … Continue reading

Fourth, during the first lockdown in the Netherlands, which started on March 16 2020, people were still allowed to leave their home and visit a POS as often as they wanted, except for POS in particular sectors such as restaurants and bars, recreation and culture and the services sector. Furthermore, kindergartens, schools and universities were closed and people were encouraged to work from home and to avoid public transport as much as possible. From mid-May 2020 onwards, the imposed containment measures have gradually been relaxed, and by the beginning of July 2020 the pandemic appeared largely under control. As a result, from July 1 onwards, many of the COVID-19 measures were further relaxed by the government: the maximum of people that could visit a pub, restaurant or recreational/cultural venue was increased to 100 (but subject to 1.5m distance), people were allowed to participate in sport competitions, and those working from home, were – although not recommended – (partially) allowed to go to the office. However, on October 14 2020 the Dutch government tightened the COVID-19 containment measures again to combat a second wave of infections. Cafes and restaurants were closed again, as well as cinemas, museums, theatres, sports clubs, conference centers and sport events were cancelled. On December 15 2020 the Dutch government took a third set of measures in order to mitigate the impact of new more contagious COVID-19 variants. For instance, it closed all schools and non-essential service providers as well as non-essential shops. The interplay between containment measures, as described by the so-called Oxford OSI index, and the number of daily COVID-19 related hospitalizations is shown in Figure 1.

 

Figure 1: Containment measures and COVID-19 hospitalizations in the Netherlands 

The remainder of this paper is structured as follows, section 2 reviews the related literature on the main drivers of payment patterns as well as the potential impact of COVID-19 on payment behaviour. Section 3 briefly describes the payment data and diary setup, while section 4 presents our main stylized facts and results. We end with a discussion and conclusion in Section 5.

 

2. Related literature

In the past decades numerous studies were conducted on the drivers of payment patterns and how to influence them. A wide range of factors emerges. Various studies find that cash usage increases with age, decreases with education and income, and negatively correlates with transaction amount (e.g. Jonker 2007; Arango-Arango et al. 2018, Wang and Wolman 2016). Moreover, it is shown that payment choice depends on the ability to monitor liquidity (von Kalckreuth et al. 2014), keep control of one’s budget (Hernandez, Jonker and Kosse 2017) and the perceived speed of payment, its user-friendliness, and safety (Jonker 2007; Schuh and Stavins 2010; van der Cruijsen and Plooij 2018). Financial incentives matter too (Arango-Arango et al. 2018; Bolt et al. 2010; Stavins 2018; Simon et al. 2010). In addition, payment behaviour depends on how well a payment instrument is accepted at the POS (Bagnall et al. 2016; DNB 2020a).^[8]There is a limited number of studies showing the importance of socio-psychological factors – e.g. social norms, attitudes, perceptions and feelings – for payment behaviour (van der Horst and … Continue reading

In spite of this large literature on the drivers of payment behaviour, relatively little is known about the effect of external shocks on consumers’ payment behaviour. Using 2005-2008 data on the Netherlands, Kosse (2013) shows that newspaper articles on skimming fraud has a limited but significant negative impact on debit card usage on the same day. However, these small effects of informational shocks do not sustain or accumulate in the long run. In a recent paper, Choi and Loh (2019) find empirical evidence that downsizing the network of ATMs in Singapore – a densely populated city – has increased customers’ travel distances to ATMs and increased their usage of the bank’s digital platform.

The COVID-19 pandemic offers a unique opportunity to study to what extent external shocks and associated containment measures by the government, banks and retailers can result in a change in payment behaviour and payment preferences. There are a few first studies. According to Chen et al. (2020) there is some early survey evidence from Spring 2020 that cash usage at the POS by Canadian citizens has decreased at the expense of debit and credit card payments, but that the role of cash as a store of value has somewhat increased. In particular, a third of the survey respondents reported that they had decreased their use of cash in response to the pandemic. Similarly, based on a yearly payment diary carried out in the U.S., Coyle et al. (2021) find that, in general, respondents hold more cash in their wallet and as a store of value in their homes, compared to trends reported in the 2019 diary. Moreover, focusing on changing payment behaviour, the results show that approximately 20% of the respondents have switched to paying online or over the phone.

Other recent reports show that the pandemic has accelerated the use of electronic payment instruments in Europe. Four out of ten respondents in an ECB study carried out in July 2020 say they use less cash since the start of the COVID-19 pandemic and a majority of these people expect to stick to this behaviour after the pandemic has faded away (ECB 2020). The fact that electronic payment instruments have been made more convenient is the most often mentioned reason for the change in behaviour. In a recent report by the Danish central bank (Danmarks Nationalbank 2020), the analysis shows that contactless and online payments quickly gained ground while cash payments fell during the lockdown in the spring of 2020. More specifically, 30% of the Danish respondents reported increased payment card use relative to before the lockdown, and 41% reported less cash usage. The Danish study also indicates that the use of cash gradually increased during the reopening of the economy by the end of August 2020. In addition, online payments have returned to pre-lockdown levels in Denmark.

Other recent reports show that the pandemic has accelerated the use of electronic payment instruments in Europe. Four out of ten respondents in an ECB study carried out in July 2020 say they use less cash since the start of the COVID-19 pandemic and a majority of these people expect to stick to this behaviour after the pandemic has faded away (ECB 2020). The fact that electronic payment instruments have been made more convenient is the most often mentioned reason for the change in behaviour. In a recent report by the Danish central bank (Danmarks Nationalbank 2020), the analysis shows that contactless and online payments quickly gained ground while cash payments fell during the lockdown in the spring of 2020. More specifically, 30% of the Danish respondents reported increased payment card use relative to before the lockdown, and 41% reported less cash usage. The Danish study also indicates that the use of cash gradually increased during the reopening of the economy by the end of August 2020. In addition, online payments have returned to pre-lockdown levels in Denmark.

 

3. Payment data

Given the COVID-19 outbreak and its associated measures by the government, banks and retailers, we expect a reduction of the share of POS transactions paid in cash and an increase in the share of electronic payments since the lockdown. Moreover, there may be reasons to believe that this change in payment usage is escorted by a shift in payment preferences causing a long-lived shift in payment behaviour. It is likely that at least part of the group of prior non-users who made the step towards paying contactless and experienced the convenience and ease of paying contactless, became enthusiastic about this payment method and changed their payment preferences accordingly. Second, payment preferences depend on perceived payment instrument characteristics. COVID-19 has changed the relative cost and benefits of different payment methods in terms of health risk and safety, ease of use and likelihood of acceptance. Third, social norms may have changed. People may infer that the social norm has moved toward paying contactless and not to using cash if more and more people do so. Prior research has shown that people copy the payment behaviour of others (van der Cruijsen and Knoben 2020).

To assess the impact of COVID-19 on consumers’ payment behaviour and preferences we use unique payment diary data collected from Dutch consumers. De Nederlandsche Bank (DNB) and the Dutch Payments Association (DPA) commissioned the data collection. The main goal of the DNB/DPA Survey on Consumers’ Payments (SCP) is to monitor consumers’ payment behaviour at the POS (Jonker et al. 2018). Members of the GfK market research-panel, aged 12 years and over fill out the questionnaire. The results give a representative picture of cash and debit card usage at the POS by the Dutch. Survey participants register their payment behaviour on the registration day. They give detailed information about the transactions they made during the day such as the payment instrument used, how much they spent at each POS, and what sector the POS belongs to. In addition, participants answer an additional questionnaire. We use this part to get insight in payment preferences.

For our analysis of payment behaviour and payment references we use data from January 1 2019 until December 31 2020. For our analysis of payment preferences we used information from all participants. This results in information from around 48,000 different people. On average, we have close to 70 diaries per day. For our analysis of payment behaviour we selected the payment dairies where the respondent made at least one payment at a POS on the registration day. We exclude payments that were not made with cash or the debit card, leaving us with more than 63,000 POS payments. We focus on cash and debit card usage, as these are by far the most frequently used means of payment at the POS in the Netherlands.

 

4. Main stylized facts and results

Figures 2 and 3 show the change in payment behaviour and payment preferences over time. Figure 2a shows 14-days moving averages of the share of POS payments made by cash or by debit card between January 2019 until December 2020 by age class and Figure 2b by income class. We highlight three key moments in time (vertical lines): the start of the lockdown on March 16 2020, the start of the second lockdown on October 14 and the implementation of a set of additional containment measures on December 15. At the start of 2020, the proportion of cash in the total number of cash and debit card POS payments still stood at 31%. Bottoming out at 13% on 12 April 2020, cash transactions rebounded to 23% at the end of June. Overall, this suggests that a large part of the shift in payment behaviour appears to be persistent because from the second half of July onwards no major changes have been observed in consumers’ payment behaviour. While the decrease in cash use is seen across all age groups, it is most pronounced among consumers older than 65 (see Figure 2a). However, in this age group there seems to be a trend back to normal as well, which was only temporarily reversed in December 2020 after the second set of measures in the second lockdown. Thus, this age group appears to be very responsive to the containment measures.

With respect to people from different income groups, we also observe a drop in cash usage at the start of the first lockdown (see Figure 2b). This decline was largest for people in the lowest income group. Many consumers who reduced their cash payments since the COVID-19 pandemic first broke out, still made fewer cash payments at the end of the year. However, those in the lowest income group did not. They have reverted almost entirely to their previous payment behaviour. In February 2020, just before the pandemic started, they used cash for 41% of their purchases, and at the end of the year this percentage stood at 39%.

Consumers in the Netherlands have given various reasons for the change in their payment behaviour at the POS (DNB 2020b). The most frequently cited reason was that merchants strongly encouraged them to pay contactless or they did not accept cash (37%). Consumers also responded that they paid electronically more often because the government advised them to do so (36%), because paying electronically has been made more convenient, or because the virus can be transmitted via banknotes or through hand-to-hand contact with the cashier (29% for all three).

 

Figure 2: Payment usage (2019-2020) – by age and income group  

Payment preferences have also changed considerably since the start of the lockdown; contactless payments clearly gained ground. Figure 3 shows 14-days moving averages of the share of people preferring different payment methods during January 2019 until December 2020. Since March 16 2020 substantially more people developed a preference for paying contactless, whereas the share of people preferring to pay with their debit card in the traditional way (so by inserting the card into the payment terminal and providing a PIN code) decreased. Both paying contactless by debit card and mobile phone became increasingly popular. The share of consumers preferring these payment instruments increased from 39% to almost 49%, respectively, from 7% to 10%, so a combined total increase of 13 percentage points. This occurred mainly at the expense of the share of people preferring to use the PIN debit card, which dropped from 29% to 19%. Surprisingly, the share of people preferring to pay with cash only declined by 1 percentage point, from 21% to 20%.

 

Figure 3: Payment preferences (2019-2020) 

Relatedly, Figure 4 depicts the percentage of people unable to pay in their preferred way. We observe a peak after the first lockdown, which seems to subside relatively quickly as people get used to the new status quo. After the end of the lockdown another peak is observed, which may indicate that people were able to go out and shop again, but cash payments were still not possible in many places. Then, after the second lockdown comes into force, there is no indication of a substantial effect.

 

Figure 4:  POS payments not with preferred instrument (2019–2020) 

Note: 14-days moving averages

A more detailed regression analysis (using data until October 13 2020) shows that the COVID-19 pandemic has led to an increase in the likelihood of a debit card transaction at the POS in the Netherlands at expense of cash (Jonker et al. 2020). Since the start of the March lockdown the likelihood that consumers use their debit card instead of cash increased by 13 percentage point compared to debit card usage before the lockdown. Moreover, empirical evidence suggests that part of the shift in payment behaviour may be long-lived. That is, about 60 percent of this shift has persisted seven months after the start of the pandemic in the Netherlands.

Furthermore, the likelihood that someone prefers to pay contactless increased by debit card by 8 percentage points compared to before the lockdown, whereas the likelihood of preferring to use the debit card with a PIN code decreased with 6 percentage points. There appears to be no significant effect of the pandemic and associated containment measures on the likelihood that people prefer to use cash or their mobile phone. More importantly, there is no reversal of payment preferences after the end of the first lockdown. These results suggest that the changes in payment behaviour and payment preferences stem for a large part from the measures taken by government, banks and retailers and are not caused directly by the fear of getting infected. The likelihood of using the debit card for payments and the likelihood of preferring contactless do not significantly relate to the severity of the pandemic, which is proxied by the number of new infections by province (Jonker et al. 2020).

 

5. Conclusions

In our analysis, we find that COVID-19 and its associated containment measures initially led to a 13 percentage point increase of the likelihood of debit card versus cash usage. The impact of the pandemic on people’s payment behaviour appears to be mainly triggered by the containment measures to control the pandemic. In addition, for many people payment behaviour has not returned to pre-COVID-19 levels. The share of cash payments at the POS has only partially reversed since its lowest point in April 2020. Thus, at least part of the effect appears to be persistent, although it seems too early to tell what part of the shock is temporary and what part of the shock is permanent. Also, we conclude that the lockdown did not have a homogeneous impact on people’s payment behaviour; the effect differs across age classes and income classes. Elderly people older than 65 and people with low income seem to return to their pre-COVID payment behaviour to a larger extent relative to younger age and higher income groups. Their payment behaviour seems to fluctuate more implying that the availability of cash remains important for these groups.

Overall, our results suggest that payment behaviour will not return to its pre-pandemic level in the future as payment preferences of many people have changed. Substantially more people now prefer to pay contactless. The share of people preferring to use their debit card with a PIN code has seen a large decline, whereas the share of people fond of cash usage only slightly declined. There are several possible interpretations for the persisting lower share of cash usage. People who preferred to use cash may have continued to pay electronically because of the fear of getting the virus. It could also be that they still perceive that retailers and other people do not want them to pay by cash. Paying electronically may now be perceived as the new social norm. COVID-19, together with its associated containment measures, may have helped to break old social norms. Another plausible explanation is that COVID-19 and the subsequent measures have induced people to break their cash habits.

Our findings provide a better understanding of how an external health shock and its associated measures by the government, banks and retailers can shift payment behaviour and payment preferences. Only within a few months’ time, a persistent change in payment behaviour appears to have taken place that, if we extrapolate pre-pandemic trends, normally would have taken several years. Compared to other external shocks, the impact of the pandemic on payment behaviour has been relatively large in magnitude and long-lasting in duration.

The COVID-19 pandemic is reshaping consumer shopping and retail payments across the globe. Although consumer spending is likely to recover over time, the way people shop and pay for goods and services may persistently change. At the same time, this creates new opportunities for banks and other payment providers to speed up the provision of contactless and online payment services. However, there is a group of elderly and low-income people that prefers to stick to cash. These people may on average be less digitally able or use cash as a way to monitor and manage their spending. Also, they may prefer some specific properties of cash, such as the anonymity of a cash payment, or perceive cash as less risky than digital money. For them, cash is still king.

 

Notes

Wilko Bolt is the corresponding author. E-mail: w.bolt@dnb.nl. For econometric approach and results this article draws on Jonker et al. (2020). The views expressed in this article do not necessarily reflect the views of DNB or those of the Eurosystem.

 

References

Arango-Arango, C., Bouhdaoui, Y., Bounie, D., Eschelbach, M. and Hernandez, L. (2018). Cash remains top-of-wallet! International evidence from payment diaries. Economic Modelling, 62, 38–48.

Bagnall, J. Bounie, D., Huynh, K., Kosse, A., Schmidt, T., Schuh, S. and Stix, H. (2016), Consumer cash usage: a cross-country comparison with payment diary survey data. International Journal of Central Banking, 12 (4), 1–61.

Bolt, W., Jonker, N. and van Renselaar, C. (2010). Incentives at the counter: An empirical analysis of surcharging card payments and payment behaviour in the Netherlands. Journal of Banking and Finance, 34, 1738–1744.

Chen, H., Engert, W. Huynh, K., Nicholls, G., Nicholson, M. and Zhu, J. (2020). Cash and COVID-19: The impact of the pandemic on the demand for and use of cash. Bank of Canada Staff Discussion Paper 2020-6.

Choi, H. and Loh, R. (2019). Physical frictions and digital banking adoption. Available at SSRN: https://ssrn.com/abstract=3333636 or http://dx.doi.org/10.2139/ssrn.3333636.

Coyle, K., Kim, L. and O’Brien, S. (2021). Consumer payments and the COVID-19 pandemic: The Second Supplement to the Findings from the 2020 Diary of Consumer Payment Choice. FED Notes, February, Federal Reserve Bank of San Francisco.

van der Cruijsen, C. and Knoben, J. (2020). Ctrl+C Ctrl+pay: Do people mirror electronic payment behavior of their peers? Journal of Financial Services Research, 59, 69-96. https://doi.org/10.1007/s10693-020-00345-6.

van der Cruijsen, C., and Plooij, M. (2018). Drivers of payment patterns at the point of sale: stable or not? Contemporary Economic Policy, 36 (2), 363–380.

van der Cruijsen, C. and van der Horst, F. (2019). Cash or card? Unravelling the role of socio-psychological factors. De Economist, 167 (2), 145–175.

Danmarks Nationalbank (2020). Payments before, during and after the corona lockdown. 16 September 2020-no 16.

DNB (2020a). DNBulletin: Shift of cash to debit card continues. Available at:  https://www.dnb.nl/en/actueel/dnb/dnbulletin-2020/shift-of-cash-to-debit-card-continues/

DNB (2020b). DNB Factsheet “Point of sale payments during the COVID-19 pandemic”. Available at: https://www.dnb.nl/media/tehjvslf/betalen-aan-kassa-uk.pdf

ECB (2020). Study on the payment attitudes of consumers in the euro area (SPACE). December.

Golec, P., Kapetanios, G., Neuteboom, N., Ritsema, F. and Ventouri, A. (2020). Consumption during the covid-19 pandemic: Lockdown or fear? Evidence from transaction data for the Netherlands, King’s College London, Working Paper no. 2020/4, September.

Hernandez, L., Jonker, N. and Kosse, A. (2017). Cash versus debit card: the role of budget control. Journal of Consumer Affairs, 51 (1), 91-112.

van der Horst, F. and Matthijsen, E. (2013). The irrationality of payment behaviour. DNB Occasional Studies No. 11(4).

Jonker, N. (2007). Payment instruments as perceived by consumers: results from a household survey. De Economist 155(3), 271–303.

Jonker, N., van der Cruijsen, C., Bijlsma, M. and Bolt, W. (2020). Pandemic payment patterns. DNB Working Paper No. 701.

Jonker, N., Hernandez, L., de Vree, R. and Zwaan, P. (2018). From cash to cards. How debit card payments overtook cash in the Netherlands. DNB Occasional Studies No. 16(1).

von Kalckreuth, U., Schmidt, T. and Stix, H. (2014). Using cash to monitor liquidity: implications for payments, currency demand, and withdrawal behavior. Journal of Money, Credit and Banking, 46 (8), 1753–1785.

Khan, J., Belk, R. and Craig-Lees, M. (2015). Measuring consumer perceptions of payment mode. Journal of Economic Psychology, 47, 34–49.

Kosse, A. (2013). Do newspaper articles on card fraud affect debit card usage? Journal of Banking and Finance, 37(12), 5382–5391.

Mínguez, J., Urtasun, A. and García de Mirasierra, M. (2020). Consumption in Spain during the state of alert: an analysis based on payment card spending. Economic bulletin 3/2020.

Schuh, S. and Stavins, J. (2010). Why are (some) consumers (finally) writing fewer checks? The role of payment characteristics. Journal of Banking and Finance, 34 (8), 1745–1758.

Simon, J., Smith, K. and West, T. (2010). Price incentives and consumer payment behaviour. Journal of Banking and Finance, 34, 1759–1772.

Stavins, J. (2018). Consumer preferences for payment methods: Role of discounts and surcharges. Journal of Banking and Finance, 94, 35–53.

Wang, Z. and Wolman, A. (2016). Payment choice and currency use: Insights from two billion retail transactions. Journal of Monetary Economics, 84, 94–115.

WHO (2020). Guidance note on the role of Cash and Voucher Assistance to reduce financial barriers in the response to the COVID-19 pandemic. April.

Authors

Olivier Darmouni^[1]Associate Professor, Columbia Business School. and Kerry Y. Siani^[2]Ph.D. Candidate in Finance and Economics, Columbia Business School.

 

Abstract

Corporate bonds and bank loans are the two main sources of credit for large firms. Economic theory and practice have shown that they are quite different, and thus that debt composition has implications for firms, the macroeconomy and economic policy. In this article, we map out some key trends in corporate bond issuance and bank lending in the United States and discuss how the COVID shock in 2020 affected firms and credit markets. We draw some comparisons with Europe as well as some implications for policymakers.

 

1. Bond Issuance vs. Bank Lending

A first important fact is the striking difference in firms’ debt composition between the United States and Europe. Langfield and Pagano (2016) refer to this difference as a European “bank bias.” In general, U.S. firms are much more reliant on market financing and bonds relative to European firms of the same size.  Using micro-data from public firms, Darmouni and Papoutsi (2021) estimate that the bond share of corporate credit is roughly twice as large in the United States. For instance, in 2009, bonds represent 35% of U.S. firm’s total debt, relative to only 13% in the Euro Area. Accordingly, it is appropriate to label the European financial system as ‘bank-based’ and the American as ‘market based’. While the reason for this long-standing gap is complex, differences in institutions are often deemed to play an important role. De Fiore and Uhlig (2011) cite differences in the informational environment. Becker and Josephson (2016) emphasize differences in insolvency resolution; the existence of Chapter 11 bankruptcy tilting the scale in favors of bonds in the United States.

However, this fact should not suggest that this picture is static. Firms rely on both sources of financing, and the relative share of bonds vs. bank loans has changed over time. Berg, Saunders and Steffen (2020) provide evidence that bond financing has grown in the recent decade in the United States, even though it started at a relatively high level relative to Europe. They estimate that bond financing has grown from 17% of GDP in 2008 to 21% of GDP in 2019. Crouzet (2021) finds similar trends using a variety of data sources, as shown in Figure 1. Stricter bank regulation and loose monetary policy likely played a role in this trend. Mota (2020) also highlights the role of a growing demand for safe assets, Grosse-Rueschkamp (2021) of universal banks. Note however that the growth in bond financing has been even larger in Europe, implying a reduction in the loan-bond gap in recent years (Darmouni and Papoutsi, 2021).

 

Figure 1: Aggregate loan share relative to bonds in the United States

Source: N. Crouzet, “Credit disintermediation and monetary policy.” IMF Economic Review (2021): 1-67.

 

What are the implications of corporate debt composition for firms? It is well understood that bank lending and market financing are not perfect substitutes. A central aspect of this difference is that loans are made through banking relationships, while bond financing is done at ‘arm’s length’. Relationships allows for monitoring and screening, while bond investors tend to rely on public information like credit ratings (Holmstrom and Tirole, 1997). In addition, relationship lending allows for the potential renegotiation of the terms of credit, while there is much less flexibility in bond financing (Bolton and Scharfstein, 1996). A key implication of this difference is that firms with more bonds have a larger cost of financial distress in bad times. The reason is that bonds tend to be widely held by a dispersed base of investors, which makes them harder to renegotiate. This coordination (free rider) problem across bond creditors means that market financing is typically seen as less reliable in bad times compared to relationship lending from banks.

The firm’s decision to issue bonds as opposed to getting a bank loan is often viewed as a trade-off between growth and risk. The bond market can offer significantly larger amounts and longer maturities than banks, allowing firms to make big, long-term investments. However, this additional capacity has a potential cost if the borrower faces a negative shock that impairs its ability to service its debt. This is especially true in case of recessions that do not originate from the banking sector, such as the COVID-driven recession of 2020. The growth in bond financing has indeed been associated with a shift towards higher risk. For instance, the BBB-rated segment (one notch above the Investment Grade rating threshold) has been growing the fastest in recent years. In Europe, Darmouni and Papoutsi (2021) shows that new bond issuers tend to be smaller, more levered, and less profitable relative to historical issuers.

Will bank lending eventually be replaced by bond financing for large firms? This should not be case, because bonds cannot replace one key role of banks: the provision of liquidity on demand. Indeed, credit plays a dual role: a firm can borrow to finance a long-term investment that will pay off in the future (term lending); or borrow to withstand temporary cash-flow shocks (liquidity provision). Bank-issued credit lines are the corporate analog to households’ credit cards: firms have an available balance that they can draw when they need to and repay when able to. Banks thus have a special advantage in liquidity provision; there is no market substitute that provides liquidity on demand, even in the U.S.

Why are banks unique in providing liquidity? The main explanation is related to banks’ deposit-taking activities. Gatev and Strahan (2006) argue that funds tend to flow towards safe bank deposits in bad times because of a ‘flight-to-safety’ effect. Thus, banks are flush with liquidity precisely in times when firms need funds the most. Kashyap, Rajan and Stein (2002) relatedly argue that banks have an incentive to hoard liquid assets to meet potential deposit outflows, and that these liquid assets can also be used to meet drawdowns on credit lines. Another line of argument is given by Holmstrom and Tirole (1998), which show that credit lines set up in advance can alleviate financial frictions through a liquidity insurance mechanism. In contrast, the bond market, by its very nature, cannot provide funds in advance.

Bank credit lines account for a significant portion of firms’ access to credit. Large U.S. firms maintain sizeable credit lines with banks even if most of their term funding comes from the bond market (Sufi, 2009; Greenwald et al. 2020). The importance of credit lines has been growing in the recent years following the financial crisis (Berg et al., 2020). Notably, credit lines have grown while bonds have crowded out bank term lending. The common view is that banks are still central to corporate credit markets, but that their role has shifted towards providing relatively more liquidity provision in the form of undrawn credit lines, rather than term lending in the form of term loans.

These pre-2020 facts lead to natural predictions about the effects of a large aggregate shock on corporate credit markets. In the absence of a banking crisis, bank loans should take precedence over bonds. Specifically, bank credit lines should play a very special role in providing liquidity to firms. In contrast, the bond market should be suppressed due the lack of profitable investment opportunities and greater risk aversion in market participants. The next section compares these predictions with patterns of loans and bonds issuance in 2020.

 

2. The COVID Shock: Liquidity-Driven Bond Issuance and the Federal Reserve Response

The spread of COVID led to a large drop in corporate cash-flows in spring 2020.  There was a widespread “dash for cash” across the corporate sector as firms scrambled for liquidity (Acharya and Steffen, 2020a; Li et al., 2020). This episode raises many questions: what is the role of the bond market in providing liquidity in bad times?  What form of debt do firms prefer to raise to meet their emergency liquidity needs? What are the implications for monetary policy and the real economy?

The COVID period is particularly useful to study the firm’s side of the equation, as neither the supply of bond capital nor bank capital was severely constrained. The bond market lent extensively to firms in this period, a surge that was partly due to a spectacular change in the Federal Reserve credit policy that supported the corporate bond market directly for the first time.^[3]See for example Haddad et al. (2020), Boyarchenko et al. (2020), Kargar et al. (2020), O’Hara and Zhou (2020), Gilgricht et al. (2020) or Liang (2020). Both investment-grade (IG) and high-yield (HY) markets reached historical heights in the post-March 2020 period. Figure 2 shows that, as of end of May 2020, investment grade (high yield) issuance by reached $500 billion ($110 billion), compared to $200 billion ($89 billion) over the same period last year.^[4]The sample includes U.S. firms and firms that issue in USD and report financial statements in USD.

 

Figure 2: Bond issuance in 2020

Source: Darmouni and Siani (2020). Data from Mergent FISD, http://bv.mergent.com/view/scripts/MyMOL/index.php, retrieved July 30, 2020.

Note: Red lines correspond to March 23, 2020 (first Fed announcement to buy corporate bonds); April 9, 2020 (first Fed announcement to buy high yield corporate bonds); and May 12, 2020 (start of Fed bond buying program).

 

How did firms choose to use the bond capital that became more available due to policy intervention? How does bond issuance interact with bank financing? To explore these questions, it is necessary to first understand how firms’ balance sheets change around bond issuance. Analyzing balance sheets before and after bond issuance helps inform what firms do with the funds raised from the bond market in bad times vs. normal times. Below, we present a summary of some key facts studied in more detail in Darmouni and Siani (2020).

 

Borrowing Without Investment

During COVID, firms used the bond market differently than in normal times. First, while in normal times, firms follow an issuance pattern and raise bonds when they have lower cash balances and debt coming due, firms issuing during COVID raise bond capital earlier in their bond financing cycle and have less debt coming due. This fact indicates that bond issuance during this time was not simply due to firms rolling-over bonds as they mature. Firms actively sought to increase their reliance on the bond market.

Second, after issuance, COVID-era issuers are more likely to hoard the proceeds from bond issuances rather than invest in real assets. We find that in normal times, 58% of IG issuers increase non-cash assets by the second quarter following issuance; however, in COVID times, only 18% issuers did.  In addition, firms were less likely to payout to equity holders after issuing during COVID. This pattern lends credence to the view that a large share of issuance was “precautionary” and thus unlikely to be immediately reinvested. Chevron, for example, issued $650 million in bonds on March 24th, but cut its 2020 capital spending plan by $4 billion.

The spike in debt issuance in bad times can be explained by recalling the dual role of credit. Liquidity-driven debt issuance spikes because the real recovery is expected to be slow. On the other hand, investment-driven debt issuance is delayed. These bond issuance patterns are drastically different from normal times. The textbook view of bond issuance exclusively financing long-term investment holds only in good times. 2020 has shown that “liquidity-driven” bond issuance can be equally as important as investment-driven issuance.

The Crowding-Out of Bank Loans

One key aspect of the 2020 crisis is that it did not originate in the banking sector. In fact, banks were healthy and entered the year with strong balance sheets, largely because of tighter regulation put in place since the Great Financial Crisis. In fact, according to the Federal Reserve Senior Loan Officers Survey of April 2020, less than 10% of banks cited capital or liquidity positions as a reason for tightening their lending standards. This is important to frame predictions: the common view would suggest that banks provided most of the funding relative to the bond market. Indeed, while firms issued bonds in the GFC, the main interpretation is that loan supply was restricted after a banking crisis (Becker and Ivashina, 2014).

However, even though the shock did not originate in the banking sector, bond issuance crowded out bank loans in 2020, in two ways.

First, many firms left their existing credit lines untouched while issuing bonds instead.  For instance, CVS had $6 billion of its credit line available at the beginning of 2020, yet it still issued $4 billion in BBB-rated bonds. Strikingly, this behavior includes many riskier HY firms: almost 40% of HY issuers received no new net bank funding between January and March. Only 21% had maxed out their credit line by end of March, and the average draw-down rate was below 50%. Many of these riskier firms had available “dry powder” from banks, arranged before the crisis, that they did not use. The pattern is even stronger for IG firms, which represent the bulk of issuance in this period, with over 60% not drawing on their existing credit lines. In aggregate, the amount of undrawn bank credit available at the beginning of 2020 was larger than the total funds raised from bond issuance. HY issuers in our matched sample issued $90 billion in bonds while having $142 billions of undrawn credit available. The gap is even larger for IG issuers.

Second, a large share of issuers that did borrow from their bank early in the crisis repaid by issuing a bond in the following weeks. For example, Kraft Heinz, which was downgraded from IG to junk in February 2020, drew $4 billion from its credit line between February and March. In May, it issued $3.5 billion in bonds (up from a planned $1.5 billion, due to strong investor demand) and used these funds to repay its credit line. In six months, the share of Kraft’s credit coming from banks went from zero to 12% and then back to zero. We find that Kraft is far from an isolated example: among HY issuers repaying bank loans, the median firm paid back 100% of its Q1 borrowing, representing 60% of their bond issuance. In aggregate, a full quarter of HY firms’ bond proceeds went to pay back bank loans. The pattern is similar for IG firms, although a smaller share drew on their credit lines in the first place. We estimate that at least $70 billion was repaid by bond issuers to banks between April and July 2020. Moreover, the majority of the Federal Reserve single-name corporate bond portfolio consists of issuers that had access to bank funds which they did not draw.^[5]Based on Federal Reserve portfolio as of July 31, 2020, as reported on August 10, 2020. https://www.federalreserve.gov/monetarypolicy/smccf.htm

Figure 3: Credit lines draw-downs in 2020 Q1 vs. Q2

 

Source: Darmouni and Siani (2020). Based on Capital IQ Capital Structure Summary table, separately by high-yield and investment grade issuers. For ease of interpretation, the figure also displays the negative 45-degree line (exact repayment in Q2) and horizontal line (no change in credit line in Q2). Excludes large outliers Volkswagen, Ford, and GM.

Why would firms prefer issuing bonds over drawing on credit lines in spite of the prediction of common wisdom? There are at least two reasons why this was the case in the spring of 2020.

First, bond financing is more committed for a long period of time: it typically has a longer maturity and no maintenance covenants that banks can use to renegotiate credit (Sufi, 2009). This is attractive because recessions typically imply cash-flow shocks that last for as long as a few years, and firms that need to cover operational fixed costs thus prefer sources of funds that are committed for a longer period. This implies a more nuanced perspective on the value of bank “flexibility” relative to market financing.

Second, the spectacular reversal of the Federal Reserve credit policy has at least partially eliminated one key aspect of banks’ specialness: the implicit and explicit government support they receive. This support implies that banks are viewed as a safe haven by investors, enhancing their willingness to hold deposits in bad times (Gatev and Strahan, 2006). Historically, the corporate bond market has been outside the scope of government support, but this has changed in dramatic fashion in Spring 2020. Correspondingly, investor demand for bonds was sufficiently strong during the COVID episode to finance record levels of issuance in April and May 2020. Moreover, while Falato et al. (2020) document unprecedented outflows from corporate bond funds in March and early April, the phenomenon was short-lived. Following the Federal Reserve’s announced intent to support corporate bond markets on April 9, there were significant net inflows to both HY and IG bond funds that remained very large through August.

 

Implications for Monetary Policy

Our findings have important implications for the conduct of monetary policy. In particular, direct support for the corporate bond market has received a lot of attention, with many open questions. Our evidence shows that it is important to account for the crowding out of bank loans when evaluating the aggregate effects of these new public programs on the real economy. For the majority of issuers, propping up bond markets does not alleviate a hard credit constraint, since they already have available bank funding. Moreover, firms by and large did not re-inject the record amount of bond issuance into their operations: they instead hoarded most of it in cash on their balance sheet or repaid existing debt. This evidence suggests that the V-shaped recovery of bond markets, propelled by the Federal Reserve, is unlikely to lead to a V-shaped recovery in real activity.

Preventing large bank credit line drawdowns is nevertheless valuable for at least three reasons: (1) it guarantees a longer-term funding source for firms, (2) it helps weaker issuers to “keep their powder dry” to weather any further negative shocks, and (3) it reduces balance sheet constraints on banks (Acharya and Steffen, 2020b). However, as of now, there is little evidence that corporate bond purchases have “trickled down” to smaller borrowers. In fact, it seemed that small firms were largely unable to borrow from banks during the spring of 2020 (Chodorow-Reich et al., 2020, Greenwald et al., 2020). Moreover, the benefits of supporting the bond market directly by extending lender of last resort policies beyond the banking sector must be balanced against potential losses on central bank bond holdings or asset price distortions leading to excessive risk-taking.

 

References

Acharya, V.V., and Steffen, S (2020b). The risk of being a fallen angel and the corporate dash for cash in the midst of covid. CEPR COVID Economics, 10, 2020b.

Acharya, V.V., and Steffen, S. (2020a). ‘Stress tests’ for banks as liquidity insurers in a time of covid. VoxEU.org, (March 22, 2020a).

Becker, B., and Ivashina, V. (2014). Cyclicality of credit supply: Firm level evidence. Journal of Monetary Economics, 62, 76-93.

Becker, B., and Josephson, J. (2016). Insolvency resolution and the missing high-yield bond markets. The Review of Financial Studies, 29 (10), 2814-2849.

Berg, T. Saunders, A., and Steffen, S. (2020). Trends in Corporate Borrowing. Annual Review of Financial Economics.

Bolton, P., and Scharfstein, D.S. (1996). Optimal debt structure and the number of creditors. Journal of Political Economy, 104 (1), 1-25.

Boyarchenko, N., Kovner, A.T., and Shachar, O. (2020). It’s what you say and what you buy: A holistic evaluation of the corporate credit facilities.

Chodorow-Reich, G., Darmouni, O., Luck, S., and Plosser, M. (2020). Bank liquidity provision across the firm size distribution. Working Paper.

Crouzet, N. (2021). Credit disintermediation and monetary policy. IMF Economic Review, 1-67.

Darmouni, O. and Papoutsi, M. (2021). The Rise of Bond Financing in Europe. Working Paper.

Darmouni, O., and Siani, K. (2020). Crowding-Out Bank Loans: Liquidity-Driven Bond Issuance. CEPR COVID Economics, 51, (October 2020).

De Fiore, F., and Uhlig, H. (2011). Bank finance versus bond finance. Journal of Money, Credit and Banking, 43 (7), 1399-1421.

Falato, A., Goldstein, I., and Hortaçsu, A. (2020). Financial fragility in the covid-19 crisis: The case of investment funds in corporate bond markets. Technical report, National Bureau of Economic Research.

Gatev, E., and Strahan, P.E. (2006). Banks’ advantage in hedging liquidity risk: Theory and evidence from the commercial paper market. The Journal of Finance, 61 (2), 867-892.

Greenwald, D.L., Krainer, J., and Paul, P. (2020). The credit line channel. Federal Reserve Bank of San Francisco.

Grosse-Rueschkamp. (2021). Universal banks and firms debt structure, Working Paper.

Haddad, V., Moreira, A., and Muir, T. (2020). When selling becomes viral: Disruptions in debt markets in the covid-19 crisis and the Fed’s response. Technical report, National Bureau of Economic Research.

Holmstrom, B., and Tirole, J. (1997). Financial intermediation, loanable funds, and the real sector. The Quarterly Journal of Economics, 112 (3), 663-691.

Kargar, M., Lester, B.T., Lindsay, D., Liu, S., and Weill, P.O. (2020). Corporate bond liquidity during the covid-19 crisis. Covid Economics, 27, 31-47. Kashyap, A.K., Rajan, R., and Stein, J.C (2002). Banks as liquidity providers: An explanation for the coexistence of lending and deposit‐taking. The Journal of Finance, 57 (1), 33-73.

Langfield, S., and Pagano, M. (2016). Bank bias in Europe: effects on systemic risk and growth. Economic Policy, 31 (85), 51-106.

Li, L., Strahan, P.E., and Zhang, S. (2020). Banks as lenders of first resort: Evidence from the covid-19 crisis. The Review of Corporate Finance Studies.

Liang, N. (2020). Corporate bond market dysfunction during covid-19 and lessons from the Fed’s response.

Mota, L. (2020). The Corporate Supply of (Quasi-) Safe Assets. Working Paper.

O’Hara, M., and Zhou, X.A. (2020) Anatomy of a liquidity crisis: Corporate bonds in the covid-19 crisis. Available at SSRN 3615155.

Sufi, A. (2009). Bank lines of credit in corporate finance: An empirical analysis. The Review of Financial Studies, 22 (3),1057-1088.