Capital Review - Part 5 November 2019: Setting the Countercyclical Capital Buffer

MEMORANDUM FOR FSC FROM Financial Policy (Maisie Prior) MEETING DATE 7 November 2019 SUBJECT Setting the Countercyclical Capital Buffer FOR YOUR Decision It is recommended that the Committee:

1. Note that as part of the Capital Review we have consulted on the calibration of the Countercyclical Capital Buffer (CCyB). In the consultation document it was proposed to be set at 1.5% for all banks, as part of the 15/16% Tier 1 Capital ratio.
2. Note that this is a companion paper to #8544142, which outlines submitters and External Experts’ feedback. Most submitters were in favour of having a CCyB as part of the framework, but some questioned the calibration and some had operational concerns. The two External Experts who commented on the CCyB also expressed concerns around the CCyB’s effective operation, and suggest that it be removed from the framework.
3. Note that in light of the views put forward on the CCyB, we propose three options to FSC:

1. Implement the proposed 1.5% early-set CCyB. This would require further consultation regarding implementation and indicators for releasing the CCyB during times of stress and building up the CCyB.
2. Set the CCyB at 0% (late-set), and leave the overall size of the prudential capital buffer unchanged (i.e. maintain the 16% Tier 1 ratio). This would mean that the prudential capital buffer would become larger when the CCyB is increased.
3. Remove the CCyB from the framework (and maintain the proposed 16% Tier 1 ratio). This reflects concerns raised by some submitters and two of the External Experts around identifying ‘the cycle’ and when to build up or release the CCyB.

4. Decide on a preferred option. The Banking Steering Group’s preferred option is option one (as proposed). While there are legitimate concerns about the effective release of the early-set CCyB, and understanding where in ‘the cycle’ the economy is, Financial Policy believes that there are mitigants that can be put in place to address concerns that the CCyB is released inappropriately, and by doing so assist it to meet the objective of absorbing losses and/or promoting lending. Having more flexibility in our capital rules with respect to the CCyB can provide benefits, especially given the size of the proposed capital buffer. Background
5. This paper follows a previous FSC paper1 that outlined the feedback from submitters and External Experts on the 1.5% early-set countercyclical capital buffer (CCyB), consulted on as part of the Capital Review.
6. The External Experts expressed doubts over the use of CCyBs in general, due to operational concerns. They ultimately suggest that the CCyB be rolled into the rest of the prudential capital buffer. While most submitters were in favour of having a CCyB as part of the capital framework, there were mixed views on whether the early-set calibration was appropriate. On the one hand, some submitters may have been making a case for having a lower Tier 1 capital ratio, rather than having a philosophical opposition to an early-set CCyB. On the other hand, 1 2019.10.01 CCyB background and submitters feedback FSC paper #8544142

2 some submitters questioned whether the CCyB (both early- and late-set) would work in practice.

1. Introduction
2. The key question to revisit in response to submissions and the External Experts’ reports is whether we want the proposed capital buffers to be more flexible or whether we expect the 16% Tier 1 capital ratio to be more strictly maintained.  Having a more flexible buffer (via an early-set CCyB) means that the buffer will be more likely to support lending in - or following - a crisis.  Having more conservative buffers provides more certainty that buffers will be there to absorb losses when required, and will be expected to be maintained at all times (i.e. through having a larger Capital Conservation Buffer).
3. This paper seeks FSC’s decision on the proposed direction for the CCyB. In particular, Financial Policy (FP) considers that there are three options available, taking into account feedback from submitters and the External Experts:

1. Implement the proposed 1.5% early-set CCyB. This would require further consultation regarding implementation and indicators for releasing the CCyB during times of stress and building up the CCyB.
2. Set the CCyB at 0% (late-set), and leave the overall size of the prudential capital buffer unchanged (i.e. maintain the 16% Tier 1 ratio). This would mean that the prudential capital buffer would become larger when the CCyB is increased.
3. Remove the CCyB from the framework (and maintain the proposed 16% Tier 1 ratio). This reflects concerns raised by some submitters and two of the External Experts around identifying ‘the cycle’ and when to build up or release the CCyB.

2. Considering the options 2.1. Removal from the Framework
3. Given that two of the External Experts were cautious about the CCyB, mainly due to concerns that it is difficult for anyone to predict where in ‘the cycle’ the economy is, one option is to completely remove the CCyB from the framework.
4. For example, Professor David Miles is concerned that the CCyB could be used at the wrong time, as it is difficult to tell whether an economic shock to banks is complete, or may become much worse. He cites the experience in the UK where the hit to confidence following the demise of the mortgage lender Northern Rock at the end of 2007 proved not to be the eye of the storm – in fact it was a mild precursor to something that was much more serious and which came in late 2008.
5. However, FP sees that it would be useful to have a tool in place, noting that having a CCyB in the framework could provide optionality to lower capital requirements following (or potentially during) a crisis. Although the CCyB is a relatively untested tool, and most of the theoretical literature around CCyBs is based off evidence during the GFC, there are benefits from having the flexibility of a CCyB. Caution will need to be exercised when considering the use of the tool, however, to mitigate the potential negative impacts that releasing the CCyB at the wrong time may cause.
6. Theoretically, the flexibility achieved through a CCyB could be achieved through the Escalating Supervisory Response (ESR), but this would make operation of the ESR more complex (e.g. by having a ‘normal times’ ESR and ‘crisis times’ ESR). The CCyB is a system-wide tool, whereas the ESR provides clarity to all banks about the regulatory response it can expect as it enters into the buffer.

3 2.2. Keeping the CCyB in the framework Having a CCyB in the framework provides options, over just having a capital conservation buffer. 9. The CCyB has been given two potential objectives in the international literature: building resilience and dampening the financial cycle. Resilience is generally viewed as the primary CCyB objective, while the scope for the CCyB to mitigate the upswing of a cycle is generally seen as a secondary benefit.2 10. For us, the benefit of having a CCyB is to allow for flexibility to reduce capital requirements should there be an economic downturn or crisis. There are other, perhaps better tools that may stem the impact of credit growth rather than a CCyB (e.g. LVR restrictions, sectoral capital requirements). 11. During a significant downturn or where there is a materialisation of systemic risk, banks are likely to suffer losses which can reduce their capital levels or growth in risk weighted assets (RWA). If bank capital ratios reduce to levels closer to their regulatory capital requirements, institutions may restrict the flow of credit during the downturn to reduce their leverage. Such a reaction can further depress economic activity in a pro-cyclical manner. The release of the CCyB, which takes immediate effect, looks to limit the potential that the interaction between higher losses and minimum regulatory capital requirements act as an impediment to the supply of credit to the economy. This could be especially beneficial during the post-crisis ‘recovery’ period. Figure 1 in the Appendix provides an illustration of the impact of capital requirements on bank capital ratios following a crisis under no CCyB, an early-set, and a late￾set CCyB. 2.2.1. Early-set or late-set CCyB? 12. The late-set CCyB strategy raises capital buffers when risks are high, and in most capital frameworks these rates have to be announced one year in advance. 3 The early strategy begins building the CCyB towards a prescribed neutral level, even if risks are not elevated. This ‘neutral’ CCyB setting is effectively part of the through-the-cycle capital calibration that can be removed during a severe downturn.4 While most jurisdictions have a late-set CCyB as part of their framework, some jurisdictions (including the UK, Ireland, and Denmark) have implemented variants of an early strategy, while APRA has indicated interest in an early-set CCyB (for more information, see section 5 in the Appendix). 13. When considering these options, it is also important to consider these in the context of the overall capital ratios proposed, as the early-set CCyB would take up a portion of the proposed prudential capital buffer, while a late-set CCyB would not. 14. There are also trade-offs between having an early- or late-set CCyB. In a paper to MFC in 2017, some of the broad pros and cons of each strategy were outlined, and this table is replicated in the Appendix. 5 Given that the literature on the use of CCyBs has not changed significantly following this paper, and a companion paper on the use of CCyBs and

2 O’Brien, O’Brien, and Velasco (2018) provides an overview of how the CCyB has been developed, and the BCBS (2017) provides an overview of how the CCyB has been implemented as part of Basel III, available here. 3 Raising the early-set CCyB has also been announced one year in advance in the jurisdictions that have adopted an early strategy. 4 Figure 2 in the Appendix provides an illustration of the broad CCyB strategies over the financial cycle. 5 Thornley (2017a). MFC 19 May 2017 - The countercyclical capital buffer strategy #6994986

4 conservation buffers6 , this helps to provide a framework for thinking about some of the relative pros and cons on each of the options. There are clear benefits to having an early-set CCyB, but there is a question on how flexible we want our proposed capital buffers to be. 15. The idea of the early-set CCyB is that it is built up sufficiently early in the cycle to maximise the likelihood that a buffer is available if and when required. This approach accounts for the time lag in implementing a CCyB and can be seen as being prudent, given the inherent uncertainty involved in assessing the level and potential materialisation of cyclical systemic risks.7 In addition, by moving early in the cycle the CCyB could be raised more gradually than the late CCyB, with a view to minimising potential (unwanted) impacts on the real economy. 16. An early-set CCyB can also be raised when risks in the financial system appear elevated, providing the same potential benefit of having a late-set CCyB (while still having the flexibility to lower capital requirements if a shock were to occur unexpectedly). The literature indicates that there is potential for CCyBs to have a mitigating impact on the growth period of the cycle. However, this impact is often less certain and can be dependent on the reaction of individual institutions to the (change in the) CCyB rate. 2.2.2. Operational concerns regarding release of the CCyB There are operational concerns around when a CCyB – early- or late-set – should be released. 17. As the BCBS and central banks overseas note, there are likely to be two kinds of circumstances when the CCyB is released. Where risks decline gradually, the buffer may similarly be reduced or partially released to acknowledge the reduction in cyclical systemic risk. The countercyclicality of the CCyB is of particular relevance however where there is a materialisation of systemic risk (i.e. a period of financial stress or instability), and bank capital is released to absorb losses and encourage ongoing lending. The buffer could be released in a number of alternative scenarios. 18. The Bank of England notes that there are two points at which the early-set CCyB can be lowered: 8  Before losses have crystallised: may reduce banks’ perceived need to hoard capital and restrict lending  If losses have crystallised: reducing CCyB allows banks to recognise losses without having to restrict lending to meet capital requirements. 19. However, if the CCyB is released too early, there is potential for it to accentuate the expansion of the financial cycle and worsen the crisis that may follow. If it is released too late, it may entail a significant cost to the real economy in the form of reduced credit availability. 20. Considering the impact of a ‘too early’ release of the CCyB in the context of our proposed capital requirements, if the early-set CCyB was released, this could lower the proposed Tier 1 ratio to 14.5% for D-SIB banks. If the late-set CCyB was released (presuming it was raised before the crisis hit), the 16% ratio would still be expected to be maintained. However, when

6 Thornley (2017b). What’s the use of the countercyclical capital buffer? #7304976 7 The Central Bank of Ireland notes that inherent uncertainty over the systemic risk outlook may be especially relevant in small open economies where events in the external environment, beyond the control of domestic policy makers, can have significant consequences for the domestic macro-financial environment and real economy. See O’Brien, O’Brien, and Velasco (2018) Measuring and mitigating cyclical systemic risk in Ireland: The application of the countercyclical capital buffer 8 Bank of England (2016). The Financial Policy Committee’s approach to setting the countercyclical capital buffer. This paper is available on the Bank of England’s financial stability webpage.

5 considering the size of the capital buffers, a 14.5% capital buffer could still provide ongoing loss absorbency for banks should a second shock occur. While there are risks around lowering the CCyB at the wrong time, mitigants can be put in place to ensure that the released CCyB is used as intended. 21. While the risk of releasing the CCyB at the wrong time cannot be eliminated (nor not releasing when it may have been beneficial to do so), the scheme could be designed to reduce the downside of releasing a CCyB before a second shock, if one were to occur. These issues should be separately considered in the consultation on the operationalisation of the CCyB, expected next year.  For example, we could impose dividend restrictions to ensure that banks use the released CCyB to absorb the increase in RWAs that occurs during a downturn and/or support additional lending.  If removing the CCyB is logical from a system-wide perspective, but we are concerned about an individual bank’s position (e.g. if a given bank’s lending is highly concentrated in a particular sector or there are prudential concerns about releasing the buffer for that bank but not others in the system), we could remove the CCyB and put a Pillar 2 overlay on that bank to neutralise the capital effect for the bank in question.  While the CCyB’s original intent is for it to be reduced for the entire system, it could potentially be removed for subsets. For example, we could just reduce it for those banks identified as D-SIBs to support lending, if we were concerned about non-D-SIBs capital ratios. 22. Would releasing the CCyB cause all banks to lower their buffers or just healthy banks?  It is likely that unhealthy banks may be under market pressure to retain capital. As outlined in Thornley (2017b), 9 there is evidence that releasing capital (by lowering capital requirements) during a crisis could encourage banks to increase lending but only for banks with relatively high levels of capital. Using data on international banks, Brei et al (2011) finds that higher capital ratios are only associated with higher lending for banks with total capital ratios above 10% during the height of the GFC (2008-09). Jimenez et al (2015) also finds that the reduction of Spanish dynamic provision requirements in Q4 2008 and Q4 2009 (which is equivalent to a CCyB cut) had a larger impact on the lending of well capitalised banks. 23. Would releasing the CCyB encourage risky lending before the second shock?  Again, as outlined in Thornley (2017b), international studies show that during the GFC, euro-area banks reduced risky assets (securities) ahead of less risky assets (loans) (Maurin & Toivanen (2012)), and US banks tightened lending by raising the risk premium on more risky loans (Kwan (2010)). Prudential regulation could also help guard against banks substantially increasing their risk profile after a crisis. The optionality of a CCyB is worthwhile, but caution will need to be exercised when using it. 24. As the CCyB (both early- and late-set) is a relatively untested tool internationally, the RBNZ will need to exercise the precautionary principle when deciding whether to lower, or raise, the CCyB. We will need to develop a set of indicators that will guide the CCyB’s release or build￾up, and exercise judgement on its release depending on the type of shock that is faced. 25. Some examples of potential shocks where we may consider using the CCyB include:  Asset prices dropping and stabilising at a new, lower level.  Banks have funding issues due to offshore funding markets becoming stressed.  A single, bank specific shock that has spill-overs to the rest of the financial sector.

9 Thornley (2017b). What’s the use of the countercyclical capital buffer? #7304976

6 2.2.3. Additional CCyB considerations Relationship with other prudential and macro-prudential instruments, and Phase 2 of the Reserve Bank Act Review 26. The main benefit of an early-set CCyB comes from having the option to lower capital requirements following a downturn to support lending in recognition that there are either reduced solvency concerns and/or maintaining lending to the economy reduces the negative feedback loop which may undermine bank solvency further. LVR restrictions, on the other hand, deal with both the ‘upswing’ and the ‘downswing’ in the economy, and are more targeted. Put another way, LVRs have the potential to reduce the amplification effects of a boom during the upturn by reducing the availability of credit and reducing the negative feedback effects in a downturn in the economy, by decreasing the level of debt that households have taken on (and decreasing the risk of households’ defaulting on their loans or being forced to reduce spending). Policies such as the CCyB or boosting risk weights help to keep banks solvent but do not directly address the ‘externalities’ related to bank lending and household debt levels. 27. In accordance with the Memorandum of Understanding (MoU), we would need to keep the Minister of Finance (MoF) informed, and consult with MoF and Treasury, 10 on raising the CCyB. To this end, there is a question on whether the CCyB may lend itself to political interference. Politicisation of macroprudential policy decisions may also arise as part of the Phase 2 process. MoF and the Phase 2 Review team intend to consult further on providing the RBNZ with the ability to regulate lending standards as part of Phase 2. MoF is supportive of giving the RBNZ this ability (and the potential ability for these to be applied permanently rather than temporarily), however there is still some debate about whether the use of macroprudential tools should be subject to additional consultation requirements and take into account other public policy considerations. The outcome of Phase 2 will take several years to come into force. Voluntary buffering 28. If 1.5% of the prudential capital buffer is assigned to an early-set CCyB, banks may have a lower voluntary buffer above the regulatory requirement in ‘normal times’, as the CCyB is likely less restrictive than a conservation buffer. Reciprocity 29. The BCBS text suggests that member countries should reciprocate another jurisdiction’s CCyB rate for a banking group’s exposures to that jurisdiction. Some submitters were concerned about reciprocity, however given the CCyB only applies to a banking group’s exposures in New Zealand, the increase in capital required at the group level can largely be dealt with by the higher capital that the subsidiary must have. That is, reciprocity will not result in a ‘doubling up’ of capital required at the group level. 30. One potential ancillary benefit from having an early-set CCyB that is reciprocated is that foreign bank branches would be captured. This partially deals with the uneven playing fields between bank branches and locally incorporated banks, as branches would be required to hold the 1.5 percentage point increase in capital on their New Zealand exposures.

10 Under the MoU with MoF we have agreed to keep the Minister and Treasury regularly informed on our thinking on significant policy developments relating to macro-prudential policy, and to consult with the Minister and treasury from the point where macro-prudential intervention is under active consideration, and to inform the Minister and Treasury prior to making any decision on deployment of a macroprudential policy instrument.

7 Coordination with monetary policy 31. Another possible benefit is the ability to coordinate the setting of the CCyB with the setting of monetary policy in situations where coordination may be valuable, and objectives align. However, this should not be the norm, as the objectives and governance of monetary policy and macroprudential policies are separate for good reasons. 3. Recommendation 32. While there was overall support for recommending having an early-set CCyB as part of the prudential capital buffer, members of the Banking Steering Group (BSG) noted that all of the arguments for having a CCyB rely on the RBNZ knowing when a crisis has occurred, and whether it has ended. There are risks of lowering the CCyB at the ‘wrong’ time (e.g. too early, which may delay and/or worsen the crisis that follows). To this end, constrained discretion would need to be exercised. 33. On balance, however, BSG’s preferred option was to have a 1.5% early-set CCyB (as proposed). While there are legitimate concerns about the effective release of the early-set CCyB, and understanding where in ‘the cycle’ the economy is, Financial Policy believes that there are mitigants that can be put in place to address concerns that the CCyB is released inappropriately. By doing so, this can assist the CCyB’s use to meet the objective of absorbing losses and/or promoting lending. Having more flexibility in our capital rules with respect to the CCyB can provide benefits, especially given the size of the proposed capital buffer. 34. Should the Committee decide to have the early-set CCyB as part of the capital stack, the design and framework for deciding to lower the CCyB will require further work, and caution would need to be exercised when using this tool. 4. Future CCyB developments 35. We will need to develop the strategy for releasing and building up the CCyB, including the set of indicators we would use regarding the release and build-up of the buffer. This could be part of future consultation, or consultation as part of the exposure draft. 36. Another consideration is how the CCyB will interact with other elements in the capital framework, including:  Escalating Supervisory Response (ESR).  Dividend restrictions (e.g. we may want dividend restrictions to apply when the CCyB is released, so that it can be used to absorb losses and/or stimulate lending, rather than be fully paid out as dividends).  Capital instruments (e.g. conditions for redemption of redeemable perpetual preference shares). 37. In the future, we could also consider the relationship between CCyB decisions and stress testing. For example, the UK uses stress tests to help detect changes in the structure of bank balance sheets that affects their resilience to unexpected developments, and this has helped to inform their CCyB calibration.

8 Appendix Figure 1: Stylised illustration of bank capital ratios and capital requirements under an early-set, late-set, and no CCyB during a crisis period NB: The illustration assumes that banks hold a management buffer on top of their capital requirements.

10 5. International CCyB developments 38. While much of the literature around the use of the CCyB has not changed since the theoretical and empirical literature was reviewed in 2017, there have been developments in how jurisdictions have been applying the CCyB. In particular, a number of jurisdictions have implemented a positive CCyB rate, and many have implemented this as a variant of an early￾set rather than a late-set strategy. 39. Differences in observed CCyB levels across countries are in part due to the range of macroprudential tools available to stem the build-up in financial vulnerabilities, and the extent to which those tools have been the preferred means for addressing identified vulnerabilities. 40. We review a sample of regulators’ decisions in implementing variants of an early-set CCyB, to provide some background to how this operates in other countries. Should the Committee be interested, we can provide a further survey of how other regulators that have implemented the (late-set) CCyB. The UK 41. The UK is one of the only jurisdictions to have integrated a CCyB with structural capital requirements (with 1% of the prudential capital buffer allocated to the CCyB in standard risk conditions). This policy allows the buffer to be varied - both up and down - in line with the changing risks that the banking system faces over time. 42. The Bank of England (2016) notes that this tool enables the Financial Policy Committee to adjust the resilience of the banking system to the changing scale of risk of losses it faces on its UK exposures over time.11 By increasing the CCyB when risks are judged to be building up, banks have an additional cushion of capital with which to absorb potential losses, enhancing their resilience and helping to ensure the stable provision of financial intermediation services. When threats to stability are judged to have receded, or when credit conditions are weak and banks’ capital buffers are judged to be more than sufficient to absorb future losses, the CCyB can be reduced. This will help to mitigate a contraction in the supply of lending to households and businesses. 43. The UK lowered its CCyB from 0.5% to 0% following the Brexit referendum in June 2016, to encourage banks to use the capital released to support lending to UK households and businesses. As markets settled down, the FPC announced it was raising the CCyB in June 2017 to 0.5% (effective June 2018) and then in November 2017 announced it would raise it to 1% (effective November 2018), as the risk environment was judged to be standard apart from Brexit. Ireland 44. The Central Bank of Ireland recently implemented a 1% neutral CCyB.12 The primary objective to promote banking sector resilience (i.e. the ability of the Irish banking system to withstand potential losses). To meet this objective, it is acknowledged that the buffer should be positive sufficiently early in the cycle to effectively promote resilience, while also accounting for the relative sensitivity of the Irish macro-financial environment to external developments. 45. The high level approach of the Central Bank of Ireland to the utilisation of the CCyB can be summarised as follows:13

11 Bank of England (2016). The Financial Policy Committee’s approach to setting the countercyclical capital buffer. This paper is available on the Bank of England’s financial stability webpage. 12 Information about the Central Bank of Ireland’s CCyB decisions is available here. 13 O’Brien, O’Brien, and Velasco (2018) outline Ireland’s strategy for applying the CCyB.

11  The objective in using the CCyB is to build resilience in the banking system, so as to protect it against potential losses associated with a build-up of cyclical systemic risk, thereby supporting the sustainable provision of credit to the real economy throughout the financial cycle.  The buffer should be positive sufficiently early in the cycle to effectively promote resilience, while also accounting for the relative sensitivity of the Irish macro-financial environment to external developments. Consequently, when there is a sustained trajectory in indicators related to emerging cyclical systemic risk the Central Bank of Ireland expects to maintain a positive CCyB rate.  When that trajectory is persistent or reflects emerging imbalances, the buffer rate is expected to be above 1%. The level of the buffer will be informed by the level of resilience expected to be sufficient to support the sustainable provision of credit to the real economy in a subsequent downturn.  When such a downturn or the materialisation of cyclical systemic risk is identified, the Central Bank of Ireland expects to reduce the buffer rate to a level consistent with mitigating pro-cyclicality, which includes reducing the buffer rate to zero if necessary to limit the impact of the downturn on credit supply. Denmark 46. The Danish Systemic Risk Council has increased their CCyB early, and gradually over time.14 It is currently set at 1.5% (required in 2020). The Council has also indicated that they intend to recommend an increase of the CCyB to 2.5%, as risks are still building in the Danish financial system (for example, due to international market developments related to Brexit, trade conflict and geopolitical tensions). 47. In Denmark, the CCyB is used to help limit the negative effects on the real economy of a future financial crisis. To achieve this, the CCyB is built up before financial imbalances become too large. The aim is for institutions to be more resilient whenever risks materialise, e.g. in the event of a negative shock to the financial system. The buffer can be reduced, releasing capital for use by institutions. Insofar as the institutions do not use the released capital for absorbing losses, it may be used for new lending or as a contribution to their excess capital adequacy. This helps the credit institutions to maintain a suitable level of lending in periods of stress in the financial system. 48. The Danish Systemic Risk Council will recommend the CCyB’s release if the financial sector is under pressure and there is a risk that banks and mortgage credit institutions will tighten their lending to households and firms such that a credit crunch may occur. They do note that the buffer does not necessarily need to be released in an economic slowdown. When, at some point the buffer is to be released, it can be released gradually or all at once. The decision depends on the specific situation. For instance, if there is a risk of a systemic financial crisis, the Council may hold an extraordinary meeting and recommend an immediate release of the buffer. There is also some appetite from Australian and US regulators to consider an early-set CCyB. 49. APRA noted in their January 2019 CCyB decision that they are continuing to analyse whether a non-zero default level of the CCyB should be a component of its changes to capital requirements.15 As part of this, there is appetite within APRA to consider a non-zero baseline CCyB. 14 Information about the Danish Systemic Risk Council’s CCyB decisions, as well as how the purpose of the CCyB in Denmark, is available here. 15 See APRA’s January 2019 release on the CCyB s 6(b)

12 50. Randal Quarles, The Vice Chairman for supervision of the Board of Governors of the Federal Reserve System, noted in a speech earlier this year that a non-zero default level of the CCyB may allow for increased regulatory flexibility.16 He argues that policymakers may be able to react more quickly to economic, financial, or even geopolitical shocks that occur amid otherwise normal conditions, without relying on the slow-moving credit aggregates contemplated in the original Basel proposal. This setting of the CCyB allows for more gradual adjustments in the CCyB, especially in periods with a high degree of uncertainty about the level of financial vulnerabilities. However, he emphasised that the overall capital framework in the US has been designed to ensure high capital levels without having to activate the CCyB, with the implication being that the bar for activation would be a high one; however, as a result, much of the time there would not be any buffer to reduce if conditions were to precipitously deteriorate.

16 Quarles (2019 speech text): Frameworks for the Countercyclical Capital Buffer.

MEMORANDUM FOR FSC FROM Financial Policy (Paula Hontalba and Matthew Brunton) MEETING DATE 7 November 2019 SUBJECT D-SIB buffer calibration FOR YOUR Decision It is recommended that the Committee:

1. Note that a previous paper (#8559445) provided background information on other jurisdictions’ D-SIB buffer approach, submitters’ views, and factors to consider in calibrating the D-SIB buffer. This paper focuses on the key issues in previous discussions and provides further supporting analysis.
2. Note that the key question that this paper tries to address is, “what level of resilience should we require of banks that are not designated as systemically important?”.
3. Note that we recommend an increase in the D-SIB buffer from our initial proposal of 1% to 2%. This is to recognise that the impact of a small bank failure on the financial system is significantly smaller than the impact of a D-SIB failure.
4. Decide on one of the following options: ● No D-SIB buffer (Status quo) ● 1% D-SIB buffer (Proposed in consultation) ● 2% D-SIB buffer (Recommendation) Overview
5. As part of the internal discussions last year in setting proposed capital requirements, Financial Policy focused on calibrating capital requirements to ensure that the financial system is able to withstand severe shocks (1 in 200 years). This was calibrated assuming a single representative and systemic bank. Given that the failure of a small bank does not have as much impact as a failure of a D-SIB, we proposed that non-D￾SIBs be required to meet a 15% Tier 1 ratio requirement, or a 1% D-SIB buffer.
6. Submitters’ views were mixed on the size of the D-SIB buffer. The Big 4 banks, which are designated as D-SIBs, generally agreed with the proposed D-SIB buffer of 1%, while domestic banks argued for a D-SIB buffer of at least 2%. IMF also argued for a D-SIB buffer of 2%, noting that much of the structural systemic financial risks identified in the last NZ FSAP arose from the D-SIBs. For more detail on submitter views see #8559445.
7. The main focus of this paper is to provide a framework doe the question “what level of resilience should we require of banks that are not designated as systemically important?”. This follows from discussions on previous papers on submitter views and subsequent analysis. A summary of the different lenses used in these papers is provided in the appendix.
8. In general, the aim of a D-SIB in the New Zealand context is to reflect the larger cost of failure of a systemic bank relative to that of a non-systemic one. This also needs to be weighed up against counter-acting considerations, such as the lower moral hazard risk from the OBR policy. For the purposes of calibrating the D-SIB buffer we use international benchmarks and a risk-appetite framework for the calibration decision.

6 Funding costs and ‘levelling the playing field’ 19. In some bank submissions and on-going discussions with the domestic banks, the interaction between the D-SIB buffer and ‘levelling the playing field’ was raised. However, we believe the proposed changes to the denominator sufficiently address these points and we see this as a strictly ancillary factor for the D-SIB buffer calibration. Recommendation 20. On balance, we recommend a higher D-SIB buffer and therefore a lower capital requirement for the small banks compared to the proposed D-SIB buffer of 1%, given that the impact of a small bank failure on the financial system is significantly smaller than the impact of a D-SIB failure. While there are some mitigating factors to moral hazards in the New Zealand context, we believe a 2% D-SIB buffer would better reflect the balance of considerations. Although there is a risk that imposing a lower capital requirement lowers the resilience of small banks to survive large shocks, the riskiness should be commensurate with the risk the banks pose to the system. A 2% D-SIB buffer would reflect a failure probability of approximately 1-in-115 years. Next steps 21. The Committee is asked to decide on the following options: i. Status quo (0% D-SIB buffer), ii. Current proposal (1% D-SIB buffer), or iii. Recommended (2% D-SIB buffer)

MEMORANDUM FOR FSC FROM Financial Policy (Richard Downing) MEETING DATE 7 November 2019 SUBJECT Capital Review: Regulatory Impact Assessment and Cost Benefit Assessment FOR YOUR Information It is recommended that the Committee:

1. Note that the Financial Policy team is continuing to develop a draft Regulatory Impact Assessment, which includes a full Cost Benefit Assessment (CBA).
2. Note that the draft will continue to be developed and then revised to incorporate the impacts of final decisions in November.
3. Note that the draft includes alternative scenarios that show: a. Benefits exceed costs at a 16% capital ratio under a range of alternative specifications of the key inputs. b. The net benefits of a 16% capital ratio exceed the net benefits of a 13% ratio (1 in 100 year risk appetite) and an 18% capital ratio. c. The net benefits approach zero as CET1 capital is at 18% and above.
4. Discuss whether there is value in adding further scenarios to the final RIA where costs are substantially higher, in the context that we do not consider these credible or likely. Background
5. The Reserve Bank Act requires an assessment of all regulatory impacts associated with policy decisions. The assessments must be provided to the Minister of Finance. There is also a requirement to publish regulatory assessment reports on the Reserve Bank website.
6. The draft Regulatory Impact Assessment (RIA) for the Capital Review follows the approach set out the Reserve Bank’s Statement of Policy-Making Approach document. The Statement of Policy-making Approach notes that impact analysis is generally conducted in a comparative context with different options for achieving the policymaker’s objective(s) being analysed and compared.
7. The draft RIA consists of two parts:  Part One: an analysis of the options considered at each stage of the review, including comparing the options against the six principles underlying the review and an overview of costs and benefits.  Part Two: a detailed assessment of the quantified and unquantified costs and benefits of the preferred options identified in Part One. The RIA covers each of the numerator, denominator and capital ratio
8. The draft RIA has a detailed assessment of all of the options considered for each of the numerator, denominator and capital ratio. The options in each components are

2 compared against the six principles for the Capital Review. The draft RIA shows that, for the decisions taken so far during the Capital Review, the preferred options are the ones that record the highest ratings against the objectives. Draft RIA shows quantified benefits exceed costs at a 16% capital ratio 5. The interactions between all the various parts of the framework mean that when considering the quantified the costs and benefits of the Capital Review, the impacts should to be considered as a whole, rather than trying to quantify the costs and benefits of each individual component of the Review. For example, this means that the RIA considers the costs and benefits of the package of Capital Review changes, rather than quantifying the costs and benefits of numerator, denominator and capital ratio decisions separately. 6. On 10 September 2019 FSC discussed the draft CBA. The Financial Policy team has continued to develop the draft CBA, including carrying out one round of internal formal QA review. John Yeabsley (NZIER) has also reviewed an earlier draft of CBA. His comments have been incorporated, and he is in the process of reviewing the latest draft. 7. The quantified benefits exceed costs at a 16% capital ratio, and are broadly in line with the draft discussed on 10 September. As in the 10 September draft, the key quantified cost and benefits are:  Expected GDP: the draft CBA incorporates estimates of the change in the probability of a crisis if the proposals are implemented, estimates of the impact of a crisis, and estimates of the impact of higher capital on interest rates. All of these are used to calculate the change in expected GDP.  Wealth transfers: higher interest rates are estimated to increase wealth transfers from New Zealander borrowers to foreign owners of banks, with this impact partially offset by higher tax payments from foreign owners of banks to New Zealand. 8. Table 1 below summarises the quantified costs and benefits of the proposals at a 16% capital ratio. The draft CBA is being tested and refined and should not be regarded as a finalised assessment at this stage. Table 1 shows that the benefits are estimated to exceed the costs. 9. There are also a range of non-quantified costs and benefits that are covered in narrative developed in the draft CBA. These include:  Benefits from avoiding the social costs associated with financial crises.  Welfare benefits for risk averse people from a lower risk environment.  Less fiscal risk associated with possible bailouts.  Benefits from well capitalised banks being able to lend during a crisis.  Costs if access to credit falls.  Social costs from higher interest rates. 10. The conclusion of the CBA is that the unquantified benefits exceed the unquantified costs. Hence, on balance, the unquantified impacts add to the case for the proposals. This means there is additional upside to the proposals that is not captured in the quantified net benefit estimates.

MEMORANDUM FOR Financial Stability Committee FROM Financial Policy (Author: Charles Lilly) MEETING DATE 7 November 2019 SUBJECT Capital Review: Reverse stress tests and going concern capital FOR YOUR Information It is recommended that the Committee:

1. Note that many submitters on the Capital Review cited the outcomes of repeated stress tests of New Zealand banks as evidence that current capital setting are sufficient to meet our soundness objective.
2. Note that a key sensitivity of our stress tests is the role of banks’ underlying earnings (net interest and other income, less operating expenses), which provides a first loss buffer with which to absorb losses prior to running down their capital.
3. Note that another key sensitivity of our stress tests is the scale of credit losses, where by necessity our modelling needs to rely on either extrapolation from recent historical New Zealand data, or using judgemental calibrations of overseas data.
4. Note that previous advice to MFC argued that plausible downside scenarios to these sensitivities, in the context of our existing stress tests, could justify CET1 requirements of around 11-12%.
5. Note that, for the Capital Review, the type of 1-in-200 year event we are trying to calibrate to is significantly further into the tail of the potential loss distribution than what we attempt to model with our regular stress tests. Our regular stress tests could be characterised as modelling perhaps a 1-in-50 or 1-in-75 year event. This limits the degree to which our detailed stress test exercises, though still valuable exercises on their own, can guide our regulatory calibration.
6. Note that we have constructed a small-scale reverse stress test model to assess the level of resilience implied by the December 2018 proposals (16% Tier 1). The model estimates the quantum of losses that would be needed to result in a systemically important bank to breach a minimum capital requirement, or to fully deplete its Tier 1 capital, in a scenario with a highly adverse decline in underlying earnings.
7. Note that, based on the model results, and notwithstanding its assumptions and limitations, capital ratios at the level we are proposing would be sufficient to absorb the scale of losses seen in a number of recent credit risk-driven crises (the Nordics, Italy, Spain and Portugal). Comparison of individual bank episodes suggests our proposed capital levels would be sufficient to absorb some but not all of the losses experienced at the major Irish banking groups over the past decade, and sufficient to absorb an early 1990s BNZ-style loss event.
8. Note that not all crisis events in the sample we have used would be accommodated by our proposed capital ratios. Losses on the scale of the Greek and Cypriot crises, or for some of the riskier Irish banking groups, would fully deplete our proposed Tier 1 capital levels.

2 Purpose and overview

1. This paper focuses on the calibration of the Capital Review’s proposed Tier 1 (going concern) capital requirement for large banks of 16%, and the corresponding notional risk tolerance of a 1-in-200 year systemic banking crisis.
2. The paper seeks to address a common point raised in submissions: that repeated stress tests of New Zealand banks show they are resilient to even severe stress scenarios, and as a result, the proposed increase in capital is unwarranted.
3. The paper explains some of the limitations to using stress tests of specific scenarios to calibrate capital requirements to the type of tail risk event we are seeking to protect against (the 1-in-200 year event). As an alternative, we also illustrate, using a reverse stress test model, the scale of loss event that would be needed for a bank to breach a minimum capital requirement, or use up all of their Tier 1 capital. Feedback received through consultation
4. Many submissions pointed to the robust results that the four major banks achieve in our stress tests as evidence that a 16% Tier 1 requirement is beyond what is necessary to achieve a 1-in-200 risk tolerance. Submitters commented on what they perceive to be our downplaying or dismissal of the role of stress tests in setting our capital requirements, including comparisons to other regulators who use stress tests as a direct input to calibration of banks’ individual capital requirements (e.g. the Federal Reserve’s CCAR and the Bank of England’s Pillar 2 framework).
5. The 2017 RBNZ-APRA industry stress test found that, in a combined severe macroeconomic downturn and operational risk scenario, major banks’ buffer ratios would decline from a starting point of around 5.5% to around 1.25% prior to management actions.1 Similar declines in buffer ratios were reported in previous regulator-led stress tests.
6. The key driver of these outcomes is the underlying earnings of the major banks. Despite applying more severe credit loss assumptions than some other regulators’ stress tests, the strong pre-provision net revenue (PPNR, i.e. net interest and other income, less operating expenses) capacity of the big four banks helps to deliver robust results.
7. PPNR is not specifically modelled in our other Capital Review analytical work (the risk appetite/ASRF approach), as that risk model does not take into account a bank’s income profile when calculating capital required for solvency. By contrast, stress tests model both the income statement and balance sheet/credit portfolio of a bank over a given scenario. This is because the ASRF model assumes that the loss event occurs within a year (and implicitly that no underlying profit is earned during that year), whereas stress tests allow for credit losses to happen over a number of years, which is the typical experience of credit risk-driven events as illustrated later in this paper. Some submitters stated that too much emphasis was placed on the ASRF model in our analytical work, given its limitations. Sensitivities and limitations of stress tests
8. One aspect of the design of our industry-wide stress tests is that they generally assume the shock event affects all participating banks in a similar way. The large

1 Lilly, C., “Outcomes from the 2017 stress test of major banks”, RBNZ Bulletin July 2018.

3 New Zealand banks’ PPNR, even allowing for NIM decline in a stress scenario, allows them to absorb impairment losses of approximately 2% of RWA each year before those losses would begin to draw down on their capital resources. 9. The relatively short repricing profile of the banks’ assets helps to limit their exposure to interest rate risk-related losses, meaning an assumption of a significant loss of NIM (and thereby erosion of PPNR) is difficult to justify in a stress scenario where all banks are affected by the shock in a broadly comparable way. Collectively, banks would seek to pass on a funding cost shock to their borrowers relatively quickly, and thereby preserve their net interest income as a buffer to absorb significant credit losses. 10. Previous work has assessed the sensitivity of our stress testing to variations in the scenario assumptions.2 In particular, a ‘name crisis’, where one bank stands out from its peers in funding markets, can produce much more adverse outcomes than the typical system-wide shock event we have modelled in industry stress tests. This is because, by raising a single banks’ funding costs relative to peers, and not allowing the bank to restore their NIM (due to a need to match peer banks’ lending rates), NIM compression would erode the lone bank’s ability to absorb losses. The modelling exercise suggested CET1 ratios in the region of 12-13.5% (approximately 10.7%- 12% after the proposed recalibration of IRB RWAs) would be necessary to ensure a systemic bank would not breach its minimum requirements, in a scenario also using higher credit loss assumptions than in previous industry stress test exercises. 11. A point that we regularly try to make about stress tests is that there is considerable uncertainty as to the scale of credit losses we should expect to see in a given scenario being tested, in the New Zealand context. For the most part, credit losses in past exercises have been based on extrapolations of banks’ loss data in relatively benign credit stress events (e.g. the GFC in New Zealand), or calibrated, with reliance on judgemental overlays, to loss rates observed in recent credit stress events overseas. For example, our 2014 stress test had mortgage loss rates calibrated to about half of the loss rates observed in the US in the GFC, and around a third of the loss rates observed on Irish banks’ mortgages over the same period.3 Moreover, it is difficult to obtain accurate and comparable historical data from credit risk-driven crisis episodes, limiting the number of benchmarks against which to assess the reasonableness of stress test assumptions.4 12. This is not to say that stress tests have no meaningful role: they are a very useful risk management tool for modelling how banks’ financial position would respond to a severe but plausible event. For example, they can be used to identify credible management actions that could be taken to mitigate a decline in a bank’s capital position when faced with losses. Other regulators use stress tests to calibrate capital buffer requirements such as the CCyB or Pillar 2, although this is not yet a step we have taken given the relative maturity of our stress testing techniques. 13. However, the scale of the 1-in-200 year event that we are calibrating our Tier 1 capital requirements to is significantly further into the tail of the loss distribution than what we attempt to model in our stress tests, which could be characterised as, for example, a 1-in-50 or 1-in-75 year scale event. The future loss events we are trying to mitigate with our regulatory capital requirements, to a 1-in-200 standard, are

2 The role of stress tests in the calibration of capital requirements (#7157408). 3 Dunstan, A., “The Reserve Bank’s philosophy and approach to stress testing”, RBNZ Bulletin July 2018. 4 For example, failing banks are often taken over by stronger competitors, or broken up into good and bad banks, resulting in a discontinuity in crisis banks’ financial statements.

4 neither easy to translate from foreign experiences or from the few data points we have in New Zealand. Reverse stress tests 14. Given the difficulty of calibrating capital requirements using specific stress scenarios, an alternative is to assess the reasonableness of the level of resilience that a given capital ratio would provide. In other words, at our proposed 16% Tier 1 capital ratio, what is the scale of the loss event we would need to see before a SIB would be at the point of failure? Is that loss event beyond the point of credibility? 15. We use a small-scale reverse stress test model to assess the scale of losses a SIB would need to incur to either breach its minimum capital requirements or exhaust all of its Tier 1 capital, over a 5 year horizon.5 The model simulates the income statement and capital position of the representative SIB, taking into account changes in NIM, operating expenses, RWAs and impairment losses. The model does not specify the macroeconomic scenario needed to generate this scale of losses (e.g. the combination of unemployment and house price declines), as doing so would be a fairly speculative exercise. Rather, it simply calculates the quantum of losses needed to result in a breach or failure. 16. To start with, we estimate that, following the Capital Review’s proposed IRB changes, and allowing for a 20bps increase in NIM to reflect a decreased reliance on debt funding and some repricing, the four large banks’ PPNR will settle at around 2.5% of recalibrated RWA. 17. To stress PPNR, we assume a 100bps decline in NIM phased in over the first two years of the scenario. This is a highly adverse assumption, although not implausible for a systemic shock given low and declining interest rates,6 nor for a ‘name’ crisis. As one point of comparison, we estimate that the NIM for an aggregate of the large Irish banks declined 96bps between 2008 and 2012. Our assumed decline of 100bps would see the PPNR to RWA ratio decline to around 1.2%. Less adverse NIM assumptions in the model would increase PPNR, thereby increasing banks’ ability to absorb losses in the stress scenario. 18. Table 1 presents a summary of the scenarios from the model: the scale of losses that would be needed to either breach minimum capital requirements, or completely deplete Tier 1 capital, using two starting points: a. The December 2018 proposals b. Capital levels of the large banks as at September 2019. 19. Moving from left to right, higher initial capital levels increase PPNR (through a higher assumed initial NIM), and allow banks to absorb materially larger levels of

5 We assume a starting buffer ratio of 11% under our proposals (10% regulatory and 1% management buffer) and that banks make full use of the allowance for 1.5 ppts of AT1 and 2 ppts. of Tier 2 capital at the start of the scenario. The model assumes 20% RWA growth, 10% credit growth, and 10% operating expense growth over 5 years, 75% of impairment losses are realised in the first 3 years of the scenario, and that banks pay a dividend in the first year equivalent to 75% of the prior year’s earnings. These assumptions are consistent with the 2017 APRA-RBNZ stress test. The calculations are based on an aggregate of the four large banks, using data for the year to September 2019. 6 Around one third to 40% of the large banks’ net interest income is derived from the discount they apply to deposits (transactional and call) relative to a benchmark funding cost curve. As interest rates decline this source of interest rate margin would diminish.

6 23. In figure 1C we plot the cumulative impairment losses of the major Irish banking groups, and HBOS in the UK, relative to their end-2008 RWA. 8, 9 At our proposed capital levels the reverse stress test model suggests the four SIBs would be able to absorb losses of around 15% of RWA before breaching a minimum requirement, or 23% of RWA before exhausting all Tier 1 capital. As shown in the chart, several of the Irish banks, and HBOS, incurred losses on this scale. Figure 3 shows our proposed capital levels would have been insufficient to prevent failure in the face of Anglo-Irish and Ulster Bank Ireland’s loss experiences.10 24. A further comparison can be made with the New Zealand banks’ losses in the late 1980s and early 1990s, shown in figure 1D. BNZ and NZI Bank incurred the greatest losses in this period, primarily on corporate lending. In both cases the banks required capital injections from their respective shareholders, and saw significant losses of market share (and eventual closure in the case of NZI Bank). 25. Our reverse stress test model suggests that our proposed levels of Tier 1 capital would be just enough for a SIB today to be able to absorb a BNZ or NZI Bank scale credit loss and meet their regulatory minimum requirements without seeking recapitalisation.

8 Note that end-2008 RWAs are on a Basel II basis but include a floor equivalent to 90% of Basel I RWAs. 9 While the Irish banks are perhaps an extreme case, they serve a useful purpose of illustrating both the scale of the type of event we are calibrating our capital requirements to, and that these loss events can and do happen. 10 In the case of Ulster Bank Ireland, RBS’s retail and commercial-focussed subsidiary in the Republic of Ireland, we estimate a Tier 1 ratio in the order of 70% would have been needed at end-2008 to see that bank through the losses it subsequently incurred without breaching a minimum capital requirement. Ulster Bank Ireland’s survival to the present day was dependent on the multiple capital injections it received from parent RBS throughout this time.