Capital Review: Going concern capital requirements

MEMORANDUM FOR Financial Stability Committee FROM Financial Policy MEETING DATE 22 November 2019 SUBJECT Capital Review: Going concern capital requirements FOR YOUR Information It is recommended that the Committee:

1. Note that Financial Policy used four analytical approaches in our recommended calibration of Tier 1 capital requirements: portfolio risk modelling, international crisis research, ‘optimal’ capital models, and stress tests. This paper summarises feedback received on the first three of these (feedback received on stress tests was presented at the previous meeting).
2. Note that much of the feedback we received, from submitters and the external experts, focussed on the portfolio risk model. Feedback from submitters was that the inputs we used in the model were overly pessimistic: multiplicative conservatism for each input led to an overstatement of the current probability of a crisis and the level of capital needed to meet our soundness objective.
3. Note that Financial Policy has reviewed the inputs to the model and, while we accept that some inputs are pessimistic, in other areas we have not applied conservatism. On balance, we think that a holistic view of this modelling work, as recommended by David Miles, supports our initially recommended range of Tier 1 ratios in relation to our soundness objective.
4. Note that some submitters criticised our interpretation of overseas studies relating to the probability of a banking crisis and capital ratios. In part this relates to the difficulty of translating different measurements of capital ratios to the New Zealand context. While these interpretations can lead to a lower assessment of the probability of a crisis at a given capital ratio, other interpretations are also reasonable, and feedback from the external experts was that our current cost of crisis is at the lower end of plausible ranges in the international literature.
5. Note that Basel Committee researchers recently reviewed a range of “optimal” capital studies. While not pinning down a precise, harmonised optimal capital ratio, the researchers concluded that the benefits of higher capital exceed the costs over a wide range, and that optimal capital ratios are likely to be similar to or higher than those that Basel Committee researchers found in 2010. We continue to consider that our proposed calibration is within the bounds of an optimal range for the setting of capital requirements. Purpose
6. 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.
7. The paper recaps and presents an updated view of the analytical approaches we used to arrive at the 16% Tier 1 proposal, taking into account feedback we received on each of these through public submissions, the external experts’ reports, and internal peer review.

2 3. The question the paper seeks to address is whether, based on the evidence we have gathered since releasing the Consultation Paper in December 2018, we still think that a 16% Tier 1 capital ratio is consistent with a 1-in-200 year risk tolerance for a systemic banking crisis. Overview 4. The risk appetite framework splits the calibration of the Tier 1 capital ratio into two stages: a. Set Tier 1 capital at a level where the probability of a systemic crisis is reduced to no more than 1-in-200 years (0.5%); and b. If there are clear net benefits from further increasing Tier 1 capital beyond that level, then Tier 1 capital requirements should increase further so as to maximise expected output. 5. Four analytical approaches were used to arrive at the proposed Tier 1 capital ratio requirement for large banks (16%): a. Portfolio risk modelling using the Asymptotic Single Risk Factor (ASRF) model, used in the internal ratings based (IRB) approach of Basel II/III. b. International research relating the probability of systemic banking crises to bank capital levels. c. Results from Reserve Bank stress tests, including the 2017 APRA industry stress test, and our own internal stress testing analysis. d. Results from ‘optimal’ capital models, including the Reserve Bank’s in-house model, TUATARA, which was redeveloped in 2016. 6. From the outset, our approach to calibration has focussed on capital requirements for systemically important banks (DSIBs). This reflects our view that the failure of any of the DSIBs would constitute a systemic banking crisis, which we are attempting to reduce to a notional probability of 1-in-200 years. The setting of our risk tolerance and calibration of capital requirements for non-DSIBs, relative to DSIBs, has been addressed in another paper (see #8584098). 7. We have already presented the Committee with our view on the feedback received in relation to stress tests (see #8615131). This paper provides Financial Policy’s view of the feedback received on the other three analytical approaches. Portfolio risk modelling (ASRF) 8. One of the central tools in our analysis was the ASRF model, which IRB banks use to determine their credit risk capital requirements under the Basel framework. 9. This model takes risk characteristics of a credit exposure, including probability of default (PD), downturn loss given default (LGD), and default correlation (R), to estimate a loss distribution for that credit exposure, and the capital a bank would need to cover potential losses on that exposure up to a confidence interval. 10. We used the model to estimate the capital needed for a single representative credit exposure, representing the aggregate of locally incorporated banks’ credit portfolios. 11. Several detailed critiques were submitted on our use of the ASRF model, and the inputs we used. These included: a. The model itself, and our approach to using it (e.g. single credit exposure), is too crude and simplistic to estimate the capital needed to meet the soundness objective for the banking system as a whole.

3 b. Using non-performing loans ratios (a stock measure) to calibrate the PD input of the model (a flow concept) leads to an overstatement of default risk that the Bank has not adequately adjusted for; moreover, the Bank’s compilation of non￾performing loan data placed unwarranted weight on smaller, higher risk institutions, rather than a system-weighted average. c. The LGD input the Bank used was materially higher than other plausible benchmarks, for example outcomes from IRB models (which the Bank has approved) and stress test outcomes. d. Historic data was not adjusted to reflect changes in banks’ credit portfolios over time, overstating current risks. e. The Bank has misunderstood the default correlation input to the model, and putting that interpretation to one side, has used unjustifiably conservative correlation inputs. 12. Taken together, the criticisms in public submissions can be summarised as us having taken overly conservative views on each of the model’s inputs, with the outcome of this multiplicative conservatism being overly pessimistic model outputs. This leads to submitters concluding we both i) overstate the probability of a crisis given current capital ratios, and ii) overstate the capital needed to reduce the probability of a crisis to 0.5%. 13. David Miles spent a large part of his discussions with us, and his final report, examining the ASRF model and our chosen inputs. He observed that: a. Our inputs err on the side of caution, giving NZ a risk profile similar to countries which have experienced materially higher credit losses than NZ. “This might bake into the analysis a degree of caution relative to a neutral assessment of the factors which treats risks as symmetric”. b. But our analysis does set out reasons for why we have chosen conservative inputs, and that our caution “should be balanced against other areas where the RBNZ analysis has chosen values on the less cautious side of a central value”. In particular, he cites our assumption of a bank failing when its capital falls to zero. Miles argues bank failure could justifiably be modelled as occurring at higher capital ratios, as for example banks will likely lose access to funding markets at low single digit capital ratios. c. Based on his own variation to the net benefit calculation, “there are sets of assumptions which are not implausible and which could justify both materially lower and higher levels of bank capital relative to RWA than the RBNZ proposals. To my mind they do not show that the assumptions the RBNZ made in the base case – viewed in their entirety – have skewed the results towards proposals for bank capital that are clearly higher than they should be”. 14. Miles thus concluded that, while our ASRF modelling may on balance have been overly pessimistic, generating a higher level of capital than is needed to meet our soundness criterion alone (i.e. 16% Tier 1 is probably more capital than is needed to achieve a 0.5% probability of a crisis), we have taken more benign views in other components of the net benefit calculation (understating the cost of a crisis, and overstating the lending rate increase). Therefore, a 16% Tier 1 ratio remains within the bounds of an optimal (expected output-maximising) calibration. 15. We have undertaken a detailed review of the inputs used in the ASRF modelling, in light of the range of feedback received through submissions and the external expert reviews (#8500290). Overall, we accept that some of the inputs to the model in our base case are towards the conservative end of the plausible range, but we consider these to be justifiable choices when viewed in their totality. We undertook multiple streams of modelling work using the ASRF model to gain a holistic understanding of its outputs and limitations.

4 16. For example, in one stream various inputs (PD, LGD and R) were materially higher than what some submitters suggested would be justified based on a read of the available data for New Zealand. However, this stream of work did not otherwise make allowance for the practical and conceptual limitations of the model itself, such as the assumption of a normal distribution and that bank failures occur when capital falls to zero. Hence the argument of multiplicative conservatism (through choice of inputs) does not hold. 17. An alternative stream of work used inputs (PD, LGD and R) which were closer to the observed ranges in the data, and submitters’ views, but illustrated that addressing the limitations of the model through applying a higher failure threshold, and a sensitivity test of its distributional assumptions, can lead back to the same range of plausible Tier 1 capital ratio outcomes as in the first stream (the 14.5% to 16% we originally presented to the Committee). 18. To summarise our views on the ASRF modelling: a. While we do not take all criticisms as valid, we accept that inputs to some of the modelling can be seen as overly conservative, but this reflects a consciously cautious approach given the data limitations we faced, and to compensate for limitations of the model itself. b. Alternative calibrations of the modelling, closer to values that submitters would use, would in isolation reduce the level of capital needed to meet the 1-in-200 year soundness objective. c. Our sensitivity testing of the model, taking a holistic view of both input conservatism and the limitations of the model in line with David Miles’ feedback, gives us confidence that the range of capital ratios we originally calculated would meet or exceed a 1-in-200 year soundness objective. International literature 19. Some submitters criticised our interpretation of international literature, noting that if we accounted for cross-country differences and focused on countries that are more relevant for New Zealand, then New Zealand banks’ current capital ratios would be within the range of optimal capital ratios. In particular, submitters observed that: a. Some of the studies that we chose are likely to be less relevant for New Zealand. In particular, some of the stress events that we referenced are less likely to occur in New Zealand, given that these countries often had structural differences (e.g. fixed exchange rates), and their banks had very different risk profiles compared to our ‘vanilla’ New Zealand banks.1 Professor Levine also suggested that we place more emphasis on studies that are more relevant to New Zealand. b. We did not adjust for cross-country differences in risk weights to estimate the relationship between capital and the probability of a banking crisis. Submitters noted that if we accounted for New Zealand banks’ relatively conservative risk weights compared to international peers, then the implied current probability of a banking crisis in New Zealand would be lower than 1-in-200 years. c. In interpreting studies’ cost of financial crises, we did not adjust for the differences in discount factor used. Most of the studies we referenced had a discount rate that is less than Treasury’s recommended discount rate of 6%, which means that if we used Treasury’s discount rate of 6%, then the present value of the cost of crises (i.e. the benefits of higher capital requirements) would be lower.

1 In particular, submitters noted that the following crises are not relevant to New Zealand: South Korea (1997), Iceland (2008), Ireland (2008), Greece (2008), and Cyprus (2011).

5 20. It is true that many banking crises have idiosyncratic circumstances that may not be currently relevant to the New Zealand financial system. New Zealand has a flexible exchange rate, sound institutional arrangements, and good fiscal health, which could make it more resilient to shocks. However, New Zealand’s economy and financial system is less diversified than other peer countries, which increases the risk of contagion and potential cost of crisis. 21. Professor Miles concluded that excluding international evidence would limit our analysis in estimating the cost of crisis for New Zealand, given the lack of crisis episodes in New Zealand. He noted that if we only selected the experience of all countries with similar levels of per capital income as New Zealand, then the banking sector losses that we used in our analysis do not seem pessimistic at all. Overall, Professor Miles concluded that other banking systems do not seem to be so different as to be irrelevant. 22. While New Zealand banks do have relatively simple operating models, this does not make them immune to a banking crisis. As Professor Miles noted, it is unclear whether New Zealand’s performance during the 2007-2008 period illustrates an enduring level of above average banking sector stability, or a period of above average luck. Furthermore, exuberant credit booms often arise because inadequate risk pricing leads to resource misallocation, and more capital can help mitigate this issue. 23. Some submitters have criticised our interpretation of certain ‘optimal capital’ studies, noting that if we accounted for cross-country differences in risk weights, international literature would support capital ratios close to where they are now. This criticism seems fair with regard to a few studies (e.g. the Bank of England paper, Brooke et. al (2015)). 2 24. Some submitters also criticised our interpretation of the Federal Reserve paper, Firestone et. al (2017). Firestone et. al (2017) defined a crisis as occurring when the Tier 1 capital ratio falls below 4.7%, and this threshold was chosen so that the average probability of a crisis is 3.8%, which coincides with the historical frequency of crises in advanced economies from 1988 to 2014 (Laeven and Valencia, 2012). However, the conclusions change if we use a longer time period from the updated Laeven and Valencia database (1980 to 2017), and expand the list of advanced economies (include Singapore and Hong Kong). In this case, the probability of a crisis falls to 2.4% (and the implied current probability of a crisis falls, as well as the level of capital needed to reduce the probability to 0.5%). While this criticism has its merits, it is unclear how much the implied probability of a crisis would fall if we adjust for the country and time period selection. As a general point, this illustrates how sensitive estimates of the probability of a crisis are to study methodologies. 25. As Professor Miles pointed out, while may have been conservative in some of our estimates, this was likely balanced out by the lack of conservatism in some inputs. He

2 In this paper the authors estimated the empirical relationship between capital levels and the probability of individual and systemic banking crises. In the “top down” approach, the authors estimated the probability of a systemic crisis at a given leverage ratio (defined as tangible common equity to tangible assets). The authors calculated a probability of systemic crisis of 0.4% at a TCE/TA ratio of 6.6%. New Zealand banks’ TCE/TA ratio was 7.2% in September 2019. On average across the authors’ “top-down” and “bottom-up” models, a mid-cycle probability of systemic crisis of 0.3% can be expected with a Tier 1 leverage ratio of 6% (we estimate New Zealand banks’ Tier 1 leverage ratio would be around 7% as at September 2019). The authors then converted their leverage ratio-based estimates, to risk-based estimates using the average risk weight of UK banks (37%). Our interpretation of the study was on the basis of these post-adjustment, UK-specific risk-based ratios, whereas arguably we should have used the unadjusted leverage ratio-based estimates which simply represent the empirical relationship between capital and failure probability, although this may not fully account for the risk profile of New Zealand banks.

6 noted that we are likely underestimating the cost of crisis by using a higher discount rate (63% cost of crisis estimate uses a 5% discount rate) and by assuming that the effects of a banking crisis are only moderately permanent. He pointed out that if we used a real discount rate of 3% a year and that the cost of a crisis is 10% of annual incomes, then this generates a cost of crisis of 330% of GDP. Professor Miles further demonstrated that our analysis was robust to input changes, given conservatism in some estimates. ‘Optimal’ capital research 26. Financial Policy rebuilt its ‘optimal’ capital model, TUATARA, in 2016.3 That modelling exercise concluded that a Tier 1 ratio of 18.4%, which corresponded to a probability of crisis of 0.37%, maximised the net benefits of capital (a lower probability of crisis, offset by higher interest rates net of wealth transfers). We caution that the TUATARA model is highly sensitive to parameter inputs, and the precise point estimate of optimal capital it produces has wide confidence intervals around it. Nevertheless, based on the changes we are making to the RWA amounts calculated by IRB banks, an 18.4% Tier 1 ratio translates to approximately 16.4%. 27. In June 2019 the BCBS published an updated assessment of the costs and benefits of bank capital.4 The purpose of the assessment was to reassess the 2010 research that supported the calibration of Basel III, taking into account the range of studies of the impact of changes in bank capital that have been published in recent years. The BCBS paper collated findings from both “fully-fledged” optimal capital models, which work through all of the various components to derive an optimal capital ratio, and piecemeal studies, which look at the effects of particular components of interest (e.g. impact of capital on lending rates, crisis probability, or bank lending volumes). 28. The 2019 BCBS paper finds that the net macroeconomic benefits of capital requirements are positive over a wide range of capital ratios. The authors’ reading of the post-2010 research is that the range of the theoretically-optimal level of capital requirements is likely similar to or higher than the range estimated in the original Basel Committee study (10-15%, measured using tangible common equity / Basel II RWAs). 29. The authors do not attempt to fully reconcile all of the estimates they collate, as these types of studies are highly sensitive to judgements about the economic costs of crises, MM effects, and the specification of the capital ratio (whether risk-based or leverage￾based), among other factors. The (unadjusted) net marginal benefit curves of each of the optimal capital studies considered is reproduced below.

3 See FSO paper (#7286174). 4 https://www.bis.org/bcbs/publ/wp37.htm. This paper was also presented to the Committee in July (#8107513).

7 Figure: Net marginal benefit (discounted, annual GDP bps) of different capital ratios Source: BCBS 2019. 30. As noted above, the chart should not be read as indicating that there is agreement on a particular capital ratio as being optimal. However, by plotting the outcomes of these studies one can infer that there are large gains in increasing capital from low levels, but that the net marginal benefit of higher capital begins to taper off once capital ratios reach a (study-specific) threshold. 31. Overall, we consider the position articulated in the Consultation Paper remains appropriate: that optimal capital modelling exercises show a wide range of plausible ratios, depending on the authors’ assumptions and the inputs to the model, and as such we should not overly rely on these types of models to determine an optimal capital ratio for New Zealand. At best, optimal capital studies give us comfort that the original range for Tier 1 capital ratios we presented to the Committee (14.5-16%) remains within the bounds of an optimal level.

MEMORANDUM FOR FSC FROM Financial Policy (Maisie Prior and Paula Hontalba) MEETING DATE 22 November 2019 SUBJECT The role of Tier 2 in the capital framework FOR YOUR Information Background

1. As part of the Capital Review, we consulted on the broad question on retaining Tier 2 as part of the capital framework. The question we effectively asked was whether it had a role if Tier 1 capital was going to 16%, we did not state a position on how the removal of Tier 2 would affect minimum total capital ratio requirements (i.e. would it drop from 8% to 6% or remain unchanged).
2. This paper discusses the merits of Tier 2 (‘gone-concern’) capital as a portion of the total capital ratio requirement, taking into account the feedback of submitters and the External Experts. This paper assumes that Tier 1 capital is set at 16%. The level of going - concern capital is assessed in a separate paper.
3. We have also made changes to the definition of Tier 2 as part of the Capital Review. In November 2017, we made the in-principle decision to only accept redeemable, non￾contingent preference shares and long-term subordinated debt as Tier 2 capital. Subsequently, we consulted on the suite of capital proposals in December 2018, including the definition of capital instruments. In September 2019, FSC decided to reverse its in-principle decision and accept redeemable non-contingent perpetual preference shares as AT1 capital. This paper therefore assumes that only long-term subordinated debt is accepted as Tier 2 capital. Feedback from submitters and External Experts Submitters’ feedback
4. Around 30 submitters commented on whether Tier 2 should be retained in the capital framework. Banks, law firms, investment banks, asset management firms and some finance professionals provided substantive feedback on the advantages and disadvantages of Tier 2 capital. It is recommended that the Committee:
5. Note that as part of the Capital Review we consulted on the general question of whether Tier 2 capital should remain in the framework. We did not state a position on what impact this would have on regulatory minimum capital requirements (i.e. whether it would be 6% or 8%).
6. Note that this paper summarises the feedback of submitters and External Experts.
7. Note that this paper focuses on the different arguments for retaining or removing Tier 2 as a component of the capital framework, and includes a qualitative cost-benefit analysis. The decision on whether to retain Tier 2 will be made as part of the decision on the capital stack (#8637807).
8. Note that members of the Banking Steering Group supported removing Tier 2 from the capital framework, if Tier 1 capital was set at 16% for D-SIBs.

2 5. Most submitters supported removing Tier 2 from the framework if the Reserve Bank goes ahead with the proposed level of the Tier 1 ratio. However, most submitters also supported retaining Tier 2 in the framework if the Reserve Bank decides to reduce the total CET1 ratio from 13.5%-14.5% to 11-13%. 6. Submitters noted that the main advantages of Tier 2 relate to:  Its cost relative to higher forms of capital. Most submitters noted that Tier 2 costs about one-fifth to a quarter of the cost of CET1 capital, and that therefore this provides a cheaper alternative source of capital.  The benefits of having sufficient gone-concern capital in resolving a failed bank. Submitters also noted that having Tier 2 in the framework could aid bank resolution, since having higher levels of subordinated capital would lessen the extent of potential losses in the event of a bank default.  The impact of having Tier 2 as an alternative investment proposition which could help develop New Zealand capital markets. 7. Some submitters also noted that although Tier 2 is mostly accepted as gone-concern capital, it may have benefits in reducing the probability of bank failure (and therefore should be seen as going-concern capital) due to the monitoring discipline imposed by Tier 2 capital owners. This point was also raised by Professor Ross Levine, and is further elaborated in the next section. 8. There was also general feedback that we should align with other regulators (e.g. APRA) and adopt a larger proportion of gone-concern capital as a proportion of the total capital stack. We note however that Total Loss Absorbing Capacity (TLAC) is not limited to Tier 2 instruments, and that the proportion of a bank’s balance sheet that is subordinated was not the main focus of this Capital Review. 9. Treasury has also made the point that removing Tier 2 capital from the framework could make bail-in more difficult. However, removing Tier 2 capital does not prohibit us from adopting TLAC in the future. We also note that bail-in instruments are not limited to Tier 2 capital, and that the OBR policy already allows us to lower the losses incurred by a failed bank (by applying a haircut on bank liabilities). External Experts feedback 10. Only Professor Ross Levine commented substantively on the role of Tier 2 capital. He asked us to consider the possible influence of Tier 2 capital on the governance of banks and their incentives for taking risk. 11. Professor Levine notes that the Reserve Bank focuses on capital as a cushion against adverse shocks to bank assets. He notes that Tier 2 capital, such as subordinated debt is less effective as a cushion against non-viability because subordinated debt only absorbs losses when the bank is no longer a going concern. 12. He states that if the Reserve Bank expands its focus to consider capital as both a cushion and an incentive device, Tier 2 capital can be a valuable complement to (not a substitute for) Tier 1 capital. He noted that, “To the extent that holders of Tier 2 capital (a) do not believe that the government will bail them out in times of distress and (b) influence bank decision-making either directly or through the price of Tier 2 capital, Tier 2 capital can dampen risk-taking within banks… While it is well-established that diffuse shareholders in widely-held companies typically have strong incentives to increase corporate risk after obtaining funds from debtholders, uninsured debtholders have incentives to limit this risk-shifting. Thus, well-designed Tier 2 capital can enhance incentives and complement Tier 1 capital.”

3 13. Examining the incentive effects in the New Zealand context, holders of currently issued Tier 2 capital are more diverse compared to holders of AT1 capital. Around 80-90% of currently issued AT1 capital are held by the parent or by related entities, whereas only 59% of currently issued Tier 2 capital are held by the parent or related entities. On this basis, Tier 2 capital is more likely to enhance incentives by providing a more diverse set of investors compared to AT1 capital. However, this dynamic could change as a result of the decision to count redeemable perpetual preference shares as AT1 capital, as there may be a broader pool of investors providing oversight of a bank. What is allowed as Tier 2 capital will also change from long-term subordinated debt with non-viability triggers, to simple long-term subordinated debt. The changes to the definition of AT1 and Tier 2 capital could affect their pricing and market demand for these products. 14. An earlier FSC paper that further examines the role of incentives in regards to the Capital Review (#8572613) notes that, “uninsured debtholders are incentivised to limit bank risk-taking and leverage via the price they require for debt. An additional beneficial incentive effect may stem from new investors holding such debt, which would likely be listed locally. However, higher capital requirements could also reduce the motivation of some debtholders (e.g. depositors) to closely monitor banks, although this effect is likely to be less pronounced for more subordinated/junior forms of debt.” 15. It is also unclear how much additional benefit can be derived from Tier 2 issuance (in terms of stronger discipline exerted on bank management) when Tier 1 capital is set at 16%. Given that shareholders are the first set of investors to absorb losses, then equity capital is the most effective form of capital in dis-incentivising excessive risk-taking. If shareholders have sufficient “skin in the game”, it is unclear how much additional benefit can be gained from another set of investors owed 2% of a bank’s risk-weighted assets. Retain or Remove Tier 2? 16. As outlined by the feedback from submitters and Professor Levine, the key benefits of retaining Tier 2 in the framework are that:  It provides gone-concern capital and therefore aids resolution.  Tier 2 capital owners could provide additional monitoring discipline.  Tier 2 is cheaper than equity and therefore more Tier 2 capital is likely to result in lower increases in the interest rate.  Some submitters also noted that Tier 2 provides an additional investment option for retail investors, and that this could help market development. 17. Removing Tier 2 has the benefits of:  Adopting a simpler capital framework, and  reducing the total capital ratio requirement (from 18% to 16% for D-SIBs). 18. Removing Tier 2, along with the proposed changes to the definition of AT1 capital, would result in a simpler capital structure. Many banks also noted that if we require a high Tier 1 ratio, then the likelihood of a bank failure is significantly lower and therefore there is less need for gone-concern capital (and the total capital ratio requirement can be reduced). 19. Given the operational costs involved with managing Tier 2 capital levels, in practice, it is possible that banks may need more Tier 1 capital (or to hold more Tier 2 above the 2% level) if we keep total capital requirements at 18% for D-SIBs (or 16% for non-D-SIBs, assuming a 2% D-SIB buffer).

4 20. In addition, removing Tier 2 would not particularly affect the relatively conservative calibration and structure of our proposed framework, in that we would still be in the top quartile. On the other hand, retaining Tier 2 capital in the framework would be more conventional globally (although not in New Zealand as both the NBDT and bank capital regimes are effectively Tier 1 frameworks). It is also important to note that most of the optimal capital modelling is done on the basis of CET1 or Tier 1 capital, as this is of primary relevance. Qualitative cost-benefit analysis of removing Tier 2 21. At the FSC Chew discussion on 1 November, FSC asked for a cost-benefit analysis of removing versus retaining Tier 2 capital (and keeping total capital ratio at 18% for D￾SIBs). Table 1 provides a qualitative cost-benefit analysis of removing Tier 2 and reducing the total capital ratio to 16% for D-SIBs. 22. The net benefits will be impacted by the change in the definition of Tier 2 capital (from having non-viability triggers to removing them), and the proposed changes to APS 111, which effectively means that any capital issued to by an Australian parent bank to its New Zealand subsidiary could have a diminishing effect on the parent’s CET1 ratio. Table 1: Cost-benefit analysis of removing Tier 2 CBA components Impact of Removing Tier 2 Comments Probability of crisis Zero / marginal  Although Professor Levine and some submitters noted that Tier 2 capital owners could provide additional monitoring discipline and therefore reduce a bank’s risk-taking, we are of the view that the additional incentive benefits of Tier 2 are likely to be marginal once a bank has a high Tier 1 capital ratio (i.e. if shareholders have sufficient “skin in the game”). Furthermore, it is difficult to quantify how much additional incentive benefits there are from having Tier 2 capital compared to other forms of debt.  If we are of the view that Tier 2 has limited incentive benefits, then Tier 2 does not reduce the probability of bank failure and the probability of a crisis. Cost of crisis Slightly higher (removal may increase cost of crisis)  We view Tier 2 as gone-concern capital, and that it aids in the resolution of a bank. Removing Tier 2 from the framework means that $6.9bn (as at June 2019) of subordinated debt capital will not be available to absorb losses at the point of failure. $6.9bn is equivalent to 2.3% of GDP (as at June 2019). However, this number likely materially overstates the impact of Tier 2 capital on the cost of crisis, and even so, this number is small relative to our assumed cost of crisis of 63% of GDP. 1

 Hence, removing Tier 2 from the framework is likely to result in a marginal increase in the cost of crisis.

1 To get the present value of the increase in the cost of crisis, we would need to calculate and discount the amount of output lost in a crisis that results from not having Tier 2 capital to aid in resolution. Hence, 2.3% is likely to overstate the increase in the cost of crisis.

5 Interest rate impact Uncertain Factors that could lower interest rate impact:  Removing Tier 2 capital from the framework means that the total capital ratio of D-SIBs, as proposed, decreases from 18% to 16% (and from 16% to 14% for non-D-SIBs). Lowering the total capital ratio means lower funding costs for the banks, since banks can choose to use cheaper forms of funding in place of Tier 2.  To the extent that Tier 2 may be operationally complex to manage (due to its amortisation over the life of the loan), banks may need more Tier 1 capital (which is more costly than Tier 2) or to hold more Tier 2 than 2% of RWA to meet total capital requirements.  Hence, removing Tier 2 and reducing the total capital ratio could result in lower repricing. Factors that could increase interest rate impact:  The cost of Tier 2 capital for Trans-Tasman banking groups would likely increase if the proposed capital is not APRA compliant (we think it will not be APRA￾compliant). This means that Tier 2 capital issued by the subsidiaries would not count towards the Australian banking group’s total capital ratio.  Proposed changes to APS 111 also means that the cost of all forms of capital (including Tier 2) increases, once the parent has more than 10% capital investment in its banking subsidiary. This means that once the 10% threshold is reached, any additional capital issued to the subsidiary (even if this is AT1 or Tier 2 capital) will be deducted from the parent’s CET1 capital.  To the extent that Trans-Tasman rule changes makes our proposed Tier 2 capital expensive for D￾SIBs, interest rate repricing may be higher. Other factors  Compared to the current definition of Tier 2 capital, the proposed Tier 2 is simpler, given that we propose to remove non-viability triggers. Hence, the proposed Tier 2 capital is probably likely to be cheaper (~1% above benchmark) than the current Tier 2 instrument (2-2.5% above benchmark). There are also other considerations that do not directly factor into the key elements of the Cost-benefit analysis… 23. Another important consideration is how banks, particularly the small domestic banks, would view the removal of Tier 2, as it provides an alternative source of funding for them. However, it is unclear whether the domestic banks would see any benefit in having Tier 2 (subordinated debt) in light of the decision to accept Redeemable Perpetual Preference Shares (RPPS) as AT1 capital. We also note that banks can issue subordinated debt as

6 an alternative source of funding to deposits, it would just not count towards their capital requirements. 24. As part of Phase 2 of the Reserve Bank Act Review, there are questions as to whether the Reserve Bank will be mandated to adopt a bail-in framework, with additional TLAC requirements. Retaining Tier 2 would ensure there is gone-concern capital in the framework, and that the market for subordinated bail-in-able debt instruments remains if a TLAC framework is implemented. In a sense, this would provide some degree of ‘future proofing.’ However the TLAC instruments could be different, and more diverse, than simple subordinated debt (the proposed Tier 2 instrument). 25. An additional consideration is what the removal of Tier 2 could signal about the Reserve Bank’s non-zero failure regime, as the distribution of losses would change if Tier 2 is removed from the framework. From a political economy standpoint, having a class of creditors classified as ‘gone-concern’ reduces the losses that will be borne by other creditors (e.g. depositors), and could make the operationalisation of resolving the bank more politically acceptable. Therefore, keeping Tier 2 could make it more credible that we have a non-zero failure regime. This signalling point would need to be carefully addressed if we removed Tier 2 from the framework. Recommendation 26. The majority of members of the Banking Steering Group favoured removing Tier 2 from the capital framework, noting that the costs of Tier 2 would likely exceed the benefits if we went ahead with the proposed Tier 1 capital requirements (16% for D-SIB banks). 27. The decision to retain or remove Tier 2 depends on the Committee’s weighting of the qualitative benefits and costs of Tier 2. It also, in part, depends on the Committee’s preference for the capital stack options proposed in the forthcoming ‘Capital Stack’ decision paper. 28. Hence, this paper does not provide a recommendation, but presents some of the relevant considerations for the role of Tier 2 as background for the final decisions on the capital stack (i.e. whether the stack includes Tier 2 or not).

MEMORANDUM FOR FSC FROM Financial Policy (Richard Downing) MEETING DATE 22 November 2019 SUBJECT Final Draft Regulatory Impact Assessment and Cost Benefit Assessment FOR YOUR Information It is recommended that the Committee:

1. Note that the Financial Policy has prepared a final draft Regulatory Impact Assessment, which includes a full Cost Benefit Assessment (CBA).
2. Note that the draft incorporates all decisions made to-date.
3. Note that where elements of the CBA are dependent on other FSC decisions scheduled for 22 November, the draft is based on December 2018 proposals.
4. Note that the quantified costs and benefits will be updated to reflect any decisions that differ from the base used in the draft, as per recommendation (3) above.
5. Note that the CBA shows: a. Benefits exceed costs at a 16% capital ratio under a range of alternative specifications of the key inputs. b. The net benefits approach zero as CET1 capital is at 22% and above.
6. Note that the RIA uses Monte-Carlo analysis to illustrate a range of alternative scenarios, rather than constructing specific alternative scenarios. Background
7. 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 our website.
8. The draft RIA consists of two parts. Part One is an analysis of the options considered at each stage of the review. Part Two is a detailed assessment of the quantified and unquantified costs and benefits of the preferred options identified in Part One.
9. Drafts of the CBA have been discussed at FSC meetings on 10 September and 7 November, as well as at ‘chew sessions’ in advance of those FSC meetings. The estimated costs and benefits have continued to be refined over that period, to incorporate subsequent in-principle decisions, as well as to incorporate additional analysis to inform a number of the inputs into the assessment of costs and benefits. Draft RIA shows quantified benefits exceed costs at a 16% capital ratio
10. The numbers presented in this paper incorporate all decisions that have been made to-date. For decisions that are due to be made on 22 November (eg. the final level of the capital ratio) the numbers in this paper incorporate the costs and benefits associated with the December 2018 proposals (eg. capital ratio of 16%). The quantified CBA will be updated to incorporate any changes in quantified costs and benefits that arise from final decisions on 22 November.

3 a. Common Equity Tier 1 capital. b. Additional Tier 1 capital. c. Tier 2 capital. d. Marginal, or return-sensitive, debt funding, i.e. wholesale and term deposits. e. All other debt funding, e.g. other deposits, derivative liabilities. 8. The net interest rate impact of the Capital Review proposals has been modelled using a ‘base rate’ plus margin approach, whereby the return that investors require for each tier of funding is expressed relative to a common benchmark interest rate, which can be thought of as either a swap rate or risk-free rate. Summing up the product of the quantities of funding at each tier, and the computed required returns, allows us to compare the blended cost of funds for different capital stack compositions, resulting in the estimate of a 22.9 basis point increase. A more detailed technical write-up of our interest rate methodology is available in #8628109. Sensitivity to higher interest rate estimates 9. By construction, the CBA is sensitive to the estimate of interest rate impacts. However, we are confident that the central interest rate assumption is robust for a number of reasons. It has been derived from the detailed methodology above, which has been reviewed by Reserve Bank staff from outside the Financial Policy team. It is also consistent with the approach recommended by Dr Cummings, one of the Capital Review External Experts. 10. One of the issues with interest rates being central to the framework is that interest rates move around constantly for a variety of reasons that are well known to the Committee. One recent example is the repricing of risk in the rural sector, which is unrelated to capital requirements. The important things to keep in mind are the economic fundamentals of the capital and interest rate nexus – these are the critical factors, for the purpose of the CBA, not the other possible drivers of change. 11. The approach described above has also been used to consider possible alternative increases in interest rates. In particular, if interest rates were to increase by more than included in the CBA, the costs of higher capital would increase. If interest rates increased by two to three times more than estimated in the base case then the net quantified benefits would likely be close to zero. 12. However, we do not consider a two to three times larger increase in interest rates than the base case to be plausible, as it would require a combination of inputs that are not credible. Such an increase would require all of the following to hold:  The MM offset is 0% rather than approximately 20% in the base case, i.e. investors’ returns are invariant to risk;  The cost of equity for banks is 11% (an equity risk premium of 9.5%), compared with the range of 8% to 8½% cost of equity (recently quoted by the ANZBGL CEO for example); and  Banks hold a voluntary buffer of two percentage points, double what is included in the base case. 13. Such an increase in interest rates would be well outside what we consider to be plausible, based on the economic factors that should underpin interest rates and the contribution of banks’ funding structure to loan pricing. That said, it is possible that the economic fundamentals will not hold if there are other factors that influence the final impacts.

6 Table 2: Technical feedback from Dr Yeabsley Issue Response in detailed RIA/CBA document ‘Wealth transfer’ costs could be overestimated as they do not incorporate New Zealand ownership of shares in foreign banks. Definitive information about NZ ownership of shares in banks is limited. The information we do have suggests that current NZ ownership is low, in the range of around 1% to 5%. Even at 5% this would not have a material impact on the quantified costs and benefits, although it could be more significant in the future should NZ ownership of bank shares grow. For these reasons, this has not been incorporated into the base case, but will be addressed in the narrative of the documents. The benefit of transferring risk of unexpected losses away from NZ depositors and on to foreign shareholders has not been included in the quantified benefits. The current detailed draft RIA covers this benefit in the unquantified costs and benefits section. There has not been sufficient time after receiving Dr Yeabsley’s feedback to analyse this issue in sufficient depth to quantify the impact. Our preliminary assessment is that the impact on quantified benefits will be small, and that this is best managed through the Monte Carlo sensitivity analysis, rather than incorporating the benefits in the base case. This would likely result in a small increase in the number of positive net benefit estimates in the Monte Carlo analysis. We will provide an update about this issue at the FSC Chew Session on 20 November. Next Steps 22. The RIA will be provided to the Minister of Finance on 26 November, alongside a briefing note about the final decisions.

MEMORANDUM FOR FSC FROM Financial Policy (Paula Hontalba) MEETING DATE 22 November 2019 SUBJECT Response to Expert Reports FOR YOUR Information

1. We published the External Experts’ final reports in October 2019, and the External Experts were generally complimentary of the quality of advice underpinning the Capital Review proposals. The External Experts also signalled their general comfort with overall direction proposed in the Capital Review. Nevertheless, each Expert has signalled a few areas where they think more analysis is needed.
2. Financial Policy team has now addressed most of the comments raised by the External Experts, although there are still a few outstanding issues. The outstanding issues mainly relate to further consultation needed on the Countercyclical Capital Buffer (CCyB) framework and the Escalated Supervisory Response (ESR) framework.
3. Professor Cummings also highlighted a few issues that we may wish to monitor. In particular, monitoring the performance of Basel III loss absorption measures, and whether they can be adjusted for New Zealand, as well as assessing the effectiveness of IRB models in predicting unexpected losses. We agree that it makes sense to monitor relevant regulatory and related market developments to inform policy decisions.
4. The table overleaf identifies the Experts’ recommendations and notes their status, from the perspective of the Capital Review. That is, Dr Cummings’ recommendation to investigate whether IRB models provide a more accurate prediction of unexpected losses is closed in the sense that no further work is required for the Capital Review. The decision has been taken to retain IRB in our capital framework. That is not to say that we disagree with the recommendation. IRB performance compared to standardised performance should continue to be monitored and inform decisions about the capital framework.
5. The External Experts’ recommendations have all been helpful, and we think, from the perspective of the Capital Review, have helped improve the final product. This paper recommends that the Committee: i. Note that Financial Policy has addressed most of the points raised by the External Experts (see Annex 1). ii. Note that there are some outstanding issues that we will need to do further work on: a. Countercyclical capital buffer framework (we will consult next year); b. Escalated supervisory response (we will consult next year); and c. Note the importance of monitoring the financial ecosystem and market dynamics more closely, in particular as the capital proposals take effect.

4 increases revised down by around 5 bps as a result. Escalating supervisory response Pro Miles suggests that the RBNZ should make it clear that a temporary dip in the level of CET1 below the (substantially higher) conservation buffer under the new regime is of significantly less concern than a similar size shortfall under the current regime where the conservation buffer is less than one third as large. In 2020, Financial Policy will develop and consult on a detailed proposal regarding the Escalating Supervisory Response as a bank moves further through the buffer. No – consult in 2020 Countercyclical buffer (CCyB) Prof Miles suggests that the RBNZ consider making the CCyB part of the rest of the prudential buffer, to remove direct allocation of part of the buffer to the CCYB. Financial Policy (Prior, 2019) have reviewed this topic in detail (#8544142, #8609793). However, further work and consultation is needed in developing the operational framework of the CCyB. No – consult in 2020

PAPER FOR Financial Stability Committee FROM Susan Guthrie (Financial Policy) DATE 22 Nov 2019 SUBJECT Final decisions: capital stack FOR YOUR Decision We ask that FSC: Note In December 2018 the Reserve Bank announced a proposal to increase bank capital requirements. This proposal followed earlier in-principle capital policy decisions. Together the December reform proposal and the earlier in-principle decisions can be considered one reform package, (the ‘2018 proposal’). Note After deliberation, FSC has made provisional decisions to adjust two aspects of the 2018 proposal: include redeemable perpetual preference shares (RPPS) as AT1 capital and include a DSIB buffer of 2% conditional on the buffer delivering no less than 14% Tier 1 capital for non-systemic banks. The impact of these adjustments is to reduce the estimated lending rate impact of the 2018 proposal by approximately 6 bps with no lessening of the protection provided in terms of a reduced probability of a banking crisis. Note The purpose of this paper is to seek decisions from FSC in relation to the capital ratio and its component parts (the ‘capital stack’). The decisions that need to be made are whether or not to include Tier 2, what level of Tier 1 capital to require of systemic banks, how much of the Tier 1 requirement can be meet with AT1 and (confirming) the DSIB buffer setting. Note The paper proceeds by presenting four capital stack options for consideration by FSC:  Option 1 is the 2018 proposal amended to include RPPS as AT1 and a 2% DSIB buffer. Option 1 is currently the base case for the Cost Benefit Analysis.  Option 2 has no Tier 2, and is otherwise as for Option 1 (RPPS as AT1, 2% DSIB buffer).  Option 3 departs from Option 1 by giving a greater role to AT1 at the expense of CET1 (overall Tier 1 capital requirements remain the same)  Option 4 departs from Option 1 by reducing the overall Tier 1 required of systemic banks from 16% to 15%. Note Option 4 is exactly the same as Option 1 for non-systemic banks because the 14% condition imposed on the DSIB buffer comes into play. That means the only alternatives to Option 1 presented here for non-systemic banks are removing Tier 2 (Option 2) or having AT1 make a larger contribution to Tier 1 capital (Option 3). Note The capital stack options need to be assessed relative to some criteria. Our view is that resilience (i.e. the protection provided against bank failure) and cost are arguably the most important criteria, reflecting the general approach in the optimal capital literature and the approach taken in the cost benefit analysis. However, there are other lenses that can be used too, and we include these, alongside resilience and cost, in a ‘traffic light’ comparison of the capital stack options. Agree To make a provisional selection of the capital stack from Options 1 to 4.

2 2 Background

1. In December 2018 the Reserve Bank announced a proposal to increase bank capital requirements. This proposal followed earlier in-principle capital policy decisions. Together the December reform proposal and the earlier in-principle decisions can be considered one reform package, (the ‘2018 proposal’).
2. Financial Policy has recently further reviewed the literature, considered modelling￾related feedback provided by the external experts and submitters, undertaken a reverse stress-testing exercise, and undertaken transitional analysis (see companion papers provided to BSG and FSC).
3. Based on this analysis Financial Policy continues to view 16% of RWA as an appropriate level of Tier 1 capital for the system as a whole.
4. As well, after considering feedback and Financial Policy’s analysis, FSC has agreed to two provisional adjustments to the 2018 proposal: including redeemable perpetual preference shares (RPPS) as AT1 capital and increasing the DSIB buffer to 2% conditional on the buffer delivering no less than 14% Tier 1 capital for non-systemic banks. The impact of these adjustments is to reduce the estimated lending rate impact of the 2018 proposal by approximately 6 bps with no lessening of the protection provided in terms of a reduced probability of a banking crisis. The ‘capital stack’ needs to be decided
5. The purpose of this paper is to seek decisions from FSC in relation to the capital ratio and its component parts (the ‘capital stack’). The decisions that need to be made are whether or not to include Tier 2, what level of Tier 1 capital to require of systemic banks, how much of the Tier 1 requirement can be meet with AT1 and (confirming) the DSIB buffer setting.
6. The paper proceeds by presenting four capital stack options for consideration by FSC. These options are outlined in Table 1:

3 3 Table 1 7. The four options can be summarised as follows: Option 1 is the 2018 proposal amended to include RPPS as AT1 and a 2% DSIB buffer. Option 1 is currently the base case for the Cost Benefit Analysis. Option 2 has no Tier 2, and is otherwise as for Option 1 (RPPS as AT1, 2% DSIB buffer). Option 3 departs from Option 1 by giving a greater role to AT1 at the expense of CET1 (overall Tier 1 capital requirements remain the same). Option 4 departs from Option 1 by reducing the overall Tier 1 required of systemic banks from 16% to 15%. As a result the CET1 buffer is lower than under Option 1. 8. The four capital stacks are illustrated in Figure 1 (an alternative illustration is provided in Appendix 1).

6 6 assumption would lead to a lower interest rate impact (and higher net benefit) for all the Options, but particularly Option 3 which gives a greater role to AT1. 14. We also assume a required return on equity (market value measure) of 7%. If this is an under-estimate, the lending rate impacts from all options will be higher but the difference between Option 1, on the one hand, and Options 3 and 4 on the other would be greater (both Options 3 and 4 forego some CET1). 15. Not all assumptions are conservative though. We are assuming all banks will issue AT1 if given the opportunity. Some may not, instead opting to meet the higher Tier 1 entirely with CET1, and this would see the lending rate impact of all Options increase somewhat. Resilience and cost 16. In terms of resilience we view RPPS as satisfactory Tier 1 capital providing it is held in modest amounts. However RPPS have some (manageable) residual risk which distinguishes this type of capital from CET1 (hence Option 3 is marked as amber on resilience indicators). 17. In Table 2 we indicate costs by both the expected average lending rate impact and the ease of transition for non-systemic banks. The rationale for including the latter is the more pressure non-systemic banks are under to comply with the capital requirements, the weaker competitive pressures will be in the system. A separate paper on transition impacts has been provide to FSC. ‘Hard’ minimums 18. In the current regime, in order to retain registration, banks need to have Total capital equal to 8% of RWA, Tier 1 capital equal to 6%, and CET1 equal to 4.5%. These are the ‘hard’ capital minimums. 19. Currently 2% of the 8% Total capital minimum can be met with Tier 2 capital, as per Basel III. The issue is what to do in the case where Tier 2 is omitted. Should a hard Total capital minimum be retained? If so, the Total capital minimum would need to be met entirely with Tier 1 capital (meaning a bank could potentially be deemed non￾viable despite having 8% Tier 1 capital). If not, New Zealand banks would just be subject to a requirement to have 6% Tier 1 capital (and 4.5% CET1) in order to retain registration. However omitting a Total capital requirement would mean the regime does not comply with Basel III. 20. A related issue comes up with Option 3. If AT1 is permitted to contribute up to 2.5% of Tier 1 capital, Basel III compliance would imply the ‘hard’ minimum applying to Tier 1 capital would have to increase to 7% (in order to deliver 4.5% CET1 capital). Relatedly, the Total capital hard minimum would have to increase to 9% assuming a 2% role for Tier 2 remains part of the regime.

7 7 Recommendation: R1: We recommend that FSC make a decision on one of the capital stack options above (Options 1 to 4). The options are generally similar in terms of the costs and benefits (there is some minor variation in both costs and benefits, but the options are substantially the same). The distinguishing features of the four options will be weighted by committee members according to the values they place on the various considerations in Table 2. A tractable approach, we believe, is to take the following sequential decisions: Step 1) Decide whether to retain or remove Tier 2; Step 2) Decide based on resilience relativities between the options remaining after Step 1; Step 3) Decide based on cost relativities between the options remaining after Step 2; Step 4) Decide between any remaining options using the additional criteria included in Table 2.

Ref #8677999 v1.6 MEMORANDUM FOR FSC FROM Financial Policy (Matthew Brunton, Maisie Prior, Paula Hontalba) MEETING DATE 22 November 2019 SUBJECT Transitional arrangements for Capital Review proposals FOR YOUR Decision It is recommended that the Committee:

1. Note that several submitters commented on the transitional arrangements proposed in the Capital Review. In general, submissions were in favour of extending the transition period, suggesting that transitional costs associated with the proposals could be large and are underestimated by Reserve Bank estimates.
2. Note that the key considerations for assessing the appropriate transitional arrangements are the economic impacts that may occur if banks restrict credit availability to meet the requirements. As these economic impacts are difficult to measure and highly sensitive to assumptions, our approach is to mitigate these impacts as much as possible and within reason so that they are not material.
3. Note that implementation of the ratio requirements will need to be delayed by one quarter to accommodate the handbook restructure and exposure drafts. This means the capital requirements will come into effect on 1 July 2020 instead of 1 April 2020.
4. Note that based on this analysis, we propose allow seven years for transition of ratio requirements (a change from initial proposals) for all banks. Background
5. As part of the Capital Review, the Reserve Bank proposed to implement all changes to capital requirements over a five year period. Submitters generally opposed the five year transition period, with many arguing for longer transition. Submitters also noted that the consultation did not take into consideration the potential impact of credit rationing during the transition.
6. We have now considered how the transitional arrangements could mitigate any economic impacts that may occur if banks restrict credit availability to meet the requirements, and we recommend a longer transition period to minimise adjustment costs. This paper first discusses submitters’ views, followed by an assessment of potential transitional costs due to credit rationing. Lastly, the paper considers each of the ten most systemic banks’ ability to grow and meet the capital requirements under different transition periods.
7. Our approach is to mitigate any potential credit rationing that may occur through the transitional arrangements by extending (within reason) the timeframe banks have to meet the requirements.
8. It is also important to note that the implementation of capital requirements will need to be delayed by one quarter to accommodate the handbook restructure and exposure drafts. This will be explained in the presentation on the next steps of the Capital Review (#8639968).

2 Ref #8677999 v1.6 Submitter views 5. Overall, submitters were in favour of extending the transition period, suggesting that a transition period of between 7 to 11 years should be sufficient for banks to meet the requirements through profit retention, and to develop the market for the new AT1 instrument. Submitters noted that if a recession occurs within the next five years, then this will impact the ability of banks to organically generate capital through profit retention. 6. Some submitters suggested that the transition should be delayed until the Reserve Bank has finalised all policy proposals, including technical changes to the capital framework (e.g. dual reporting). The IRB banks also asked for the same transition period to be applied to all banks, and to adopt a longer transition period for the change in the scalar and output floor, and the grandfathering of contingent capital instruments. 7. Some submitters also asked for a comprehensive post-implementation review, with some recommending that this is undertaken during the transition period (e.g. during years 2-4), and some recommending that this is undertaken once changes to the capital framework have been fully adopted. 8. Many submitters, particularly from the rural sector, requested for a longer transition period to give them sufficient time to respond to other regulatory changes (e.g. changes to water quality requirements, carbon pricing) and to improve their financial position. Submitters from the rural sector also highlighted that they have already experienced a tightening in credit availability, despite the fact that capital requirements have not increased.1 9. ANZ, NERA, and an individual submitter said that there was a lack of analysis on transitional costs. as a result, cumulative GDP growth over the next decade could be 1-3 percentage point lower. Both ANZ and NERA cited research which indicated that banks respond to higher capital requirements by targeting a reduction in RWA or by reducing credit availability to high risk-weight sectors. Credit rationing and the economic impacts 10. The general discussion around the costs of the capital proposals has largely centred on the interest rate increase during the ‘steady-state’ (which is what we use for the cost￾benefit analysis). However, increasing capital requirements can create costs in the transitional period that are independent of the steady-state costs. This could be the result of banks restricting credit growth rather than dividend payouts in order to meet capital requirements, generally through both increasing price as well as non-pricing factors (such as collateral and lending standards). Banks are able to do this during the transition period because of market frictions that may not fully adjust in the short term. 11. According to the recent Credit Conditions Survey, credit availability has somewhat tightened for agri, commercial property, and corporate loans, mostly due to changes in price (e.g. interest rate and fees). However, we also note that lending standards and payment terms (e.g. maturity and principal repayment requirements) have somewhat tightened for agri. 1 While this is largely related to a repricing of risk, we note that ANZ’s rural capital requirements increased by $151m ($1.9bn RWA) as at 30 June 2019. s(18)(c)(i)

s(18)(c)(i) s(18)(c)(i)

8 Ref #8677999 v1.6 34. Finally, we note that most research identify monetary policy as a potential tool to counteract the negative effects of a reduction in credit supply. We have not incorporated potential monetary policy response in our simplistic analysis. Were credit rationing to occur and create adverse economic impacts, it is likely that monetary policy would be used in such a circumstance.15 This would therefore counter-act some of the down-side risks associated with the transitional arrangements. 35. Nor have we factored in competition from outside the banking sector. We have effectively modelled a closed system, which over the timeframes discussed, does not seem completely realistic (and again, this would be a conservative assumption that overstates the impact). Differential transition periods 36. Domestic banks have suggested that non-D-SIBs should have a longer transition period. Many non-D-SIBs are mutual societies, co-operative, or have unique corporate structures (such as TSB bank, which is owned by a trust). This means they are more constrained in their ability to raise CET1, often being restricted to retaining earnings. 37. However, this needs to be weighed-up against the risks of having less well-capitalised small banks for a longer period. Furthermore, a longer transition for non-D-SIBs beyond seven years, and to a lower Tier 1 ratio than initially proposed, is unlikely to provide much more reprieve for their transition path. Some non-D-SIBs have more underlying issues, such as high expenditure to income, and unsustainable growth trends, which cannot be addressed by extending the transition path (nor would we wish to facilitate unsustainable growth trends). 38. On balance, we believe a seven-year transition period, in combination with a lower Tier 1 ratio requirement for non-D-SIBs (14% rather than 15%), is sufficient for non-D-SIBs. Recommendation 39. Given the uncertainty in estimating transitional costs, our preferred approach is to mitigate potential adjustment pains by lengthening the transition period to seven years for all banks. This should provide banks with ample time to meet the requirements through profit retention whilst also being able to maintain reasonable dividend payouts, and strikes a reasonable balance between costs and benefits of the recommended capital policies. 40. We also recommend delaying the implementation of ratio requirements by one quarter (implemented Q3 2020 instead of Q2 2020) to accommodate enough time for the handbook restructure and exposure drafts. Next Steps 41. Depending on whether FSC agrees with the proposed transition period or settles on an alternative transition period, we may come back to FSC to provide further information if changes are required to accommodate decisions in #8637807. 42. We will update the cost-benefit analysis/ Regulatory Impact Statement to include a section on transitional costs and benefits. We will also incorporate some of the analysis in this paper in the Response to Submissions. 15 This is on the assumption that such a scenario would align with easing monetary policy responses.

10 Ref #8677999 v1.6 Potential bank response: If banks choose to target a certain dividend payout ratio or amount, then they may reduce their lending growth further to accommodate this. While this scenario leverages bank-specific information, it is still highly circumspect and sensitive to underlying assumptions. In this analysis, this would result in a 137 bps fall in credit growth per annum over the transition period. s(18)(c)(i)

12 Ref #8677999 v1.6 Potential bank response: In this analysis, this would result in a 24 bps fall in credit growth per annum over the transition period. s(18)(c)(i)

Ref #8675720 v1.2 MEMORANDUM FOR Financial Policy FROM Charles Lilly DATE 14 November 2019 SUBJECT Capital Review: Funding cost inputs for CBA FOR YOUR Information

1. This memo outlines an approach to determine the funding cost and lending rate impact of the Capital Review decisions for use in the Cost Benefit Assessment (CBA).
2. Using this approach, the estimated lending rate impacts of the four capital stack options being discussed at the 22nd November FSC meeting are reported in Table 1: Table 1: Interest rate impact of different capital stack calibrations (for DSIBs) Option 1 (Dec 18) Option 2 (No Tier 2) Option 3 (More AT1) Option 4 (Lower CET1) CET1 (%) 14.5 14.5 13.5 13.5 AT1 (%) 1.5 1.5 2.5 1.5 Tier 2 (%) 2 0 2 2 Lending rate impact relative to status quo (bps) 22.9 22.4 20.5 17.4

3 Ref #8675720 v1.2 are not publicly traded, we need to estimate a market value for the equity of the New Zealand banks. Conceptual approach 7. Our previous interest rate estimates have been based on simple funding structures: banks are funded with either equity or debt, and we assess the effect of changes in the mix of the two. A benefit of this simple approach is that it allows for the input of an “MM” parameter, as estimated in several overseas studies that use this simple equity/debt funding structure. 8. To refine our estimate, and to allow for exploration of the effects of different mixes of capital types in the capital stack, we need to model an interest rate estimate using a more graduated funding structure, i.e. the various combinations and required returns on: a. Common Equity Tier 1 capital; b. Additional Tier 1 capital; c. Tier 2 capital; d. Marginal, or return-sensitive, debt funding, i.e. wholesale and term deposit funding; and e. All other debt funding, e.g. transaction and savings deposits, derivative liabilities. 9. This paper models the net interest rate impact of the Capital Review proposals using a ‘base rate’ plus margin approach, whereby the return that investors require for each tier of funding is expressed relative to a common benchmark interest rate, which one can think of as either a swap rate or risk-free rate. Summing up the product of the quantities of funding at each tier, and the computed required returns, allows us to compare the blended cost of funds for different capital stack compositions. 10. In practice, banks’ funds transfer pricing methodologies to determine the cost of funds for a given loan take into account more than just the position of a class of funding in the creditor hierarchy, as they also need to manage their liquidity and maturity position among other concerns. However, for the purposes of this pricing approach liquidity or duration effects are not modelled. 11. A downside to this more elaborate approach is that singular MM parameters, as estimated in other studies, are not easy to integrate into the analysis. For example, we do not have empirical estimates of the extent to which required returns on AT1 instruments adjust in response to the quantity of CET1 capital below them in the capital stack. As a result, the analysis relies on some judgemental calibrations. Balance sheet changes 12. Firstly, to use the cost of equity in the analysis, we need to estimate a market value of the equity of New Zealand banks. I use a price-earnings (PE) multiple of 12.5x the trailing twelve months’ net profit after tax, which is consistent with the long run average PE multiple of listed Australian banks. 13. Next, I assign banks’ debt funding to one of two categories – what I call “marginal debt” funding, in which I include term deposits, debt instruments and borrowing; and other debt funding, which includes savings and transaction accounts, derivative liabilities, and all other liabilities. Annualised yields for these two categories in the month of September 2019 were 2.66% and 0.87% respectively. The purpose of splitting banks’ debt funding into these two categories is to reflect that, given a

4 Ref #8675720 v1.2 choice, when required to increase their capital levels banks would prefer to use that capital to retire their most expensive forms of debt. In the analysis, I assume that the net increase in Total capital is associated with a 1-to-1 decrease in marginal debt. 14. Table 3 presents the current and post-Capital Review balance sheet of the aggregate of the 10 largest locally incorporated banks, based on the December 2018 Consultation Paper capital stack. A 1% management buffer of CET1 capital is included. Table 3: Aggregate balance sheet before and after the Capital Review Margins 15. Having determined the relative quantities of the different funding sources before and after the Capital Review, the next step is to work out the relative cost of each source before and after the Capital Review. 16. To set a ‘base rate’, I choose a value so that the spread on marginal debt is roughly consistent with spreads on recent bank senior debt issuances (relative to BKBM), and funding conditions reported in banks’ internal reporting in recent months, such as TD spreads. Assuming a 1.5% base rate provides a spread on marginal debt of 1.16% (2.66%-1.5%), and a spread on other debt of -0.63% (0.87%-1.5%). 17. Table 4 outlines the key margin inputs to the pricing calculation, including an explanation as to how these margins are assumed to adjust after the transition to higher capital levels. Current (September 2019) Change Pro forma post-Capital Review CET1 capital 35,345 +18,503 53,848 Net profit after tax (year to 09/2019) 5,586 Market value of equity 69,830 +18,503 88,334 AT1 capital 6,330 -1,118 5,211 Tier 1 capital 41,675 +17,385 59,060 Tier 2 capital 2,734 +4,214 6,948 Total capital 44,409 +21,599 66,008 Marginal debt (term deposits, debt securities, borrowing) 307,924 -21,599 286,326 Other debt (savings and transaction deposits, derivative liabilities, other) 182,706 0 182,706 Risk-weighted assets 315,355 347,409 CET1 ratio 11.21% 15.50% Tier 1 ratio 13.22% 17.00% Total ratio 14.08% 19.00%

5 Ref #8675720 v1.2 Table 4: Investors’ required returns by funding source, including base rate, assumptions for interest rate estimate Funding source Current Post￾Capital Review Explanation Equity (market value) 8.5% (after corporate tax) 8.25% (after corporate tax) • Current required return implies a 7% equity risk premium (7% + 1.5% base rate), consistent with current Treasury guidance (for firms with equity beta of 1), and comments from ANZBGL CEO who stated ANZBGL’s cost of equity was now assessed at between 8% and 8.5%.5 • RBNZ replication of Cummings and Nguyen (2019) implied a 25bps decline in the equity risk premium given the proposed change in the equity to debt mix. • A 100% MM effect would see the equity risk premium fall by around 140bps.6 • I assume a 25bps decline in the equity risk premium (MM effect of approximately 20%) based on the Cummings and Nguyen replication, although a steeper decline could be justified on several grounds. Additional Tier 1 5.5% 5.5% (after corporate tax) • Current AT1 issuances are assumed to include a margin of 400bps. In September 2019 estimated a margin on Basel III￾compliant AT1s of 375bps.7 • Increases in CET1 capital reduce the expected loss of AT1 instruments. • The trigger for primary loss absorption on RPPS will effectively be 0% CET1 under our proposals, instead of the minimum of 5.125% under current rules, reducing expected loss. • On the other hand, there is likely a reduced pool of investors for the new instrument. There is likely no appetite from parent banks to purchase, given APS111 changes. There is unknown appetite from the current pool of external investors, for example given the novelty of the instrument, lack of Basel III compliance and how this affects investment mandates, and loss of tax deductibility relative to current AT1 debt instruments. • On balance, no change in the required investor return (after corporate tax) is assumed. Tier 2 3.5% 3.0% • Current Tier 2 issuances include a margin of approximately 200- 250bps.8 • Increase in Tier 1 capital reduces the expected loss of Tier 2 instruments, justifying a decrease in margin (I assume -50bps). Marginal debt 2.66% 2.61% • Increase in Total capital reduces the expected loss of senior debt. However, consultation feedback suggests that with no change in issuer credit ratings as a result of the Capital Review, little to no material changes to senior funding spreads should be anticipated. • I assume a modest decline (-5bps). Other debt 0.87% 0.87% • No change in required return as this type of debt is assumed to be relatively inelastic.

5 Transcript of interview with ANZ CEO following FY19 results announcement, October 2019. 6 Based on the figures in Table 3, and assuming a current equity beta of 1, the current asset beta is 1/(1+(1-0.28)7.16)=0.163. Holding this asset beta constant, the decrease in the market leverage ratio from 7.16 to 5.45 as a result of the increase in equity shown in Table 3 suggests the equity beta should fall to 0.8. This implies an equity risk premium of 0.8700bps=560bps. . 8 For example, ANZBGL issued $1750m of domestic Tier 2 (10 year, non-call 5 year) at BBSW+200bps in July 2019. s(18)(c)(i) s(18)(c)(i)

6 Ref #8675720 v1.2 Cost of funds 18. Given the assumed quantities and relative margins for each source of funds, we can compute the change in the total blended cost of funds before and after the Capital Review changes. Tables 5 and 6 report the outcomes, including the grossing-up of for corporate tax of equity funding. Table 5: Total blended cost of funds, status quo Table 6: Total blended cost of funds, post-Capital Review 19. Based on the December 2018 Consultation Paper proposals, this analysis suggests an increase in the blended cost of funds for banks of 20.1bps. 20. Recovery of the increased cost of funds over total interest-earning assets of $499,759m as at September 2019 implies an average increase in lending rates of 22.9bps. Alternative Capital Stack calibrations 21. Since the above calculations are based on a five tier funding structure, we can assess the interest rate impacts of different combinations of instruments in the capital stack. 22. Table 7 compares the estimated impact of the December 2018 proposal with alternative capital stack options canvassed in our recent slidepack #8630886. All options include a voluntary 1% CET1 management buffer on top of the regulatory calibration. Funding source Base rate Required margin (for investor) Grossed for tax Cost of funds (for bank) Quantity ($m) Equity (market value) 7.00% 2.38% 10.88% 69,830 AT1 4.00% 0% 5.50% 6,330 Tier 2 2.00% 0% 3.50% 2,734 Marginal debt 1.16% 0% 2.66% 307,924 Other debt -0.63% 0% 0.87% 182,706 569,525 17,818 3.129% 1.50% + = Blended cost of funds (%) Total equity and liabilities ($m) Blended cost of funds ($m) Funding source Base rate Required margin (for investor) Grossed for tax Cost of funds (for bank) Quantity ($m) Equity (market value) 6.75% 2.31% 10.56% 88,334 AT1 4.00% 1.54% 7.04% 5,211 Tier 2 1.50% 0% 3.00% 6,948 Marginal debt 1.11% 0% 2.61% 286,326 Other debt -0.63% 0% 0.87% 182,706 569,525 18,962 3.330% 1.50% + = Total equity and liabilities ($m) Blended cost of funds ($m) Blended cost of funds (%)

7 Ref #8675720 v1.2 Table 7: Interest rate impact of different capital stack calibrations (DSIBs) Option 1 (Dec 18) Option 2 (No Tier 2) Option 3 (More AT1) Option 4 (Lower CET1) CET1 (%) 14.5 14.5 13.5 13.5 AT1 (%) 1.5 1.5 2.5 1.5 Tier 2 (%) 2 0 2 2 Lending rate impact relative to status quo (bps) 22.9 22.4 20.5 17.4 Lending rate impact relative to status quo (bps) (no 1% voluntary buffer) 17.4 16.8 14.9 11.8 Alternative modelling assumptions 23. Financial System Analysis (FSA) provided feedback on the methodology used here and suggested some slightly different calibrations for the changes in required returns in the post-Capital Review calculation. They suggested that the change in the equity risk premium should be stronger (-50bps instead of -25bps), that Tier 2 would not see as strong a decline (-20bps instead of -50bps), and that senior debt would be unlikely to see a decline (0bps instead of -5bps). Running these numbers through the calculation leads to a net lending rate impact of Option 1 of 20.5bps, 2.4bps lower than my estimate. 24. FSA also suggested using a long run base rate, rather than the 1.5% which is closer to swap rates in recent months. Using the average 10 year swap rate between 2011- 2019 of 3.5%, and FSA’s preferred margin assumptions, the net lending rate impact of Option 1 is 23.2bps. 25. Supervision also provided feedback on the assumed margins, with their view being that the methodology and assumptions are both reasonable and sensible, based on their view of recent funding cost data from banks.