In-principle decisions on the denominator framework

Ref #7885852 v1.0 MEMORANDUM FOR FSO Committee FROM Financial Policy (Principal authors: Paula Hontalba and Matthew Brunton) MEETING DATE 19 June 2018 SUBJECT In-principle decisions on the denominator framework FOR YOUR Decision We recommend that FSO:  Note that the aim of the Capital Review is to identify the most appropriate framework for setting capital requirements for New Zealand banks, and to address our concerns about the way risks are measured by the IRB banks (these concerns were outlined in the ‘denominator’ consultation paper released in December 2017).  Note that in December 2017, we consulted on a number of reform options relating to the risk measurement framework (denominator). A high level summary of the submissions can be found in Table 1.  Note that we have analysed the submissions and we suggest that there are two available distinct alternatives: (i) remove internal models from the capital framework (i.e. require full standardisation), or (ii) retain internal models but limit IRB modelling (see point below).  We seek the Committee’s decision on the following items: a. Decide on whether to retain internal models in the regime; b. If we retain internal models, then: o Agree to apply a standardised approach for calculating capital for externally rated loans; o Agree to require dual reporting of IRB banks; and o Agree to implement a capital output floor for IRB banks. c. Agree to apply to all banks the Standardised Measurement Approach for operational risk (i.e. remove advanced modelling for operational risk).  Agree that we should publish a summary of submissions and our proposed response on the website. A draft paper on ‘Proposed Response and Summary of Submissions’ (#7532285) has been prepared, and will be revised to reflect FSO’s decisions.  Note that underlying these ‘big picture’ decisions is a need for further detailed work on fine-tuning the IRB and standardised measurement framework, factoring in APRA’s proposed changes to their capital framework (#7507190). Financial Policy will bring subsequent detailed decisions to FSO (or BSG if minor and technical if FSO is comfortable with that).

2 Ref #7885852 v1.0 Principles of the Capital Review

1. The aim of the capital review is to identify the most appropriate capital framework for New Zealand. As we have communicated in the initial ‘Issues’ paper in May 2017 (#6976453), we will pay regard to the following principles in the Capital Review: a. Capital must readily absorb losses before losses are imposed on creditors and depositors. b. Capital requirements should be set in relation to the risk of bank exposures. c. Where there are multiple methods for determining capital requirements outcomes should not vary unduly between methods. d. Capital requirements of New Zealand banks should be conservative relative to those of international peers, reflecting the risks inherent in the New Zealand financial system (country-risk) and the Reserve Bank's regulatory approach (focus on disclosure requirements and market discipline rather than supervisory overlays). e. The capital framework should be practical to administer, minimise unnecessary complexity and compliance costs, and take into consideration relationships with foreign-owned banks’ home country regulators. f. The capital framework should be transparent to enable effective market discipline.
2. The most relevant of these principles for the papers in front of FSO today are the ones that relate to the measurement framework dimensions particularly:  (b) e.g. should reflect exposure risk, not model atifacts;  (c) e.g. providing a more robust mechanism for resetting the gap between standardised and IRB banks, and levelling the playing field for IRB banks with like-exposures;  (e) e.g. removing complexity/increasing simplicity, and aligning with APRA where possible; and,  (f) e.g. dual reporting and floors should materially improve transparency and therefore market discipline. Problem definition / Issues identified
3. In December 2017, we consulted on the measurement framework for risk-weighted assets (i.e. the denominator). In this we outlined several concerns about the way risks are measured and aggregated by IRB banks. In particular: a. There is a lack of transparency about the design and implementation of the banks’ internal models. It is very difficult to know with certainty whether the internal models are accurately capturing risk. This makes it difficult for supervisors and markets to assess bank capital adequacy and compare banks; b. There is a (growing) gap in capital requirements between the IRB and standardised approaches (and between some IRB banks). Preliminary evidence suggests this is not justified by differences in underlying risk; c. There is unnecessary complexity in parts of the framework. In some cases any marginal value in risk identification is questionable (e.g. op risk); and

3 Ref #7885852 v1.0 d. Due to the complexity of internal models, the assessment process is significantly resource-intensive and it is difficult to verify if banks are ‘gaming’ the IRB capital framework. 4. In addition, there have been changes to international capital standards: a. APRA has announced revisions to its capital framework, with some of their proposals differing from the new Basel approach. b. Basel 3 has been finalised. 1 5. As part of the capital review, we looked at the benefits of alignment with these international standards, particularly the proposed changes by APRA.2 Alignment with APRA was supported by three of the IRB banks (and TSB), whilst Kiwibank and Genworth supported alignment with Basel III. 6. FSO has previously expressed its desire to have the Capital Review completed by the end of 2018. However, APRA has indicated to us that they will publish their final proposals, which may differ from their current, in a consultation document for final discussion in the second half of 2019. This may require RBNZ to review proposed changes that emerge from alignment from APRA later than the desired completion date indicated by FSO. 7. It should be emphasised that complete alignment with the APRA framework would be a substantial undertaking. There are significant differences between our current capital framework and APRA’s (see Appendix 1), and in some ways, APRA is departing further from Basel.3 Another thing to note is that if FSO agrees to dual reporting, any changes on the IRB framework can have a flow-on impact on the capital framework for standardised banks (see Paras 27-28). As such, complete alignment with APRA is likely to impose unwarranted costs for all standardised banks. 8. Given that APRA has not yet confirmed their final decisions on their capital framework, further analysis will be needed to assess the merits of increased alignment with APRA. In our view, the decision on the technical details of the capital framework (i.e. asset class framework) does not materially affect the recommendations in this paper, which are more high level, direction of travel type decisions. Reform options and decisions sought from FSO 9. The submissions to our denominator consultation paper have been summarised in Table 1, and a more detailed summary can be accessed at #7532285.

1 Notable changes include: standardisation of operational risk, increased limitations on the IRB approach; and new standardised and internal models approaches to market risk capital requirements. 2 The main advantages to harmonising relate to administrative efficiencies and increased transparency for external parties, including investors. The disadvantages relate to harmonisation without power to influence the framework and the framework not reflecting national circumstances. 3 For a more detailed assessment of APRA’s proposed changes to their capital framework, please refer to ‘APRA’s Proposed Revisions to Capital Framework’ (#7507190).

4 Ref #7885852 v1.0 10. We seek the Committee’s decision on whether to remove or retain internal models in our capital framework. 11. If the Committee decides to continue allowing internal models in the capital framework, our recommendation is to: a. require IRB banks to use the standardised approach for exposures that have an external credit rating; b. introduce an output floor, so that the risk weights used by IRB banks cannot deviate other than by a set percentage from the risk weights that would apply to a standardised bank (this floor could be set at the level of individual exposures, a portfolio of exposures, or at an aggregate level);4 and c. require IRB banks to calculate capital using both their internal models and the standardised framework and publish both results (‘dual reporting’). 12. We also recommend removing the internal models approach for operational risk (Advanced Measurement Approach) and adopting the Basel standardised approach with APRA alterations. 5

13. The remainder of the paper explains the rationale behind these recommendations, and summarises the banks’ views with respect to the proposed reforms. To model or not
14. All of the IRB banks argued against the removal of internal models, noting that the IRB approach provides incentives to improve risk management, and that a less risk sensitive approach may create undesirable incentives (e.g. taking on riskier exposures).
15. Kiwibank and TSB argued for the removal of IRB approach, noting the complexity and opaqueness of IRB models, and the competitive advantage associated with the lower capital outcomes that IRB models produce. Kiwibank suggested that if we choose to retain IRB approach, we should significantly reduce the risk weight differentials between IRB and standardised approach (especially in retail portfolios).
16. Full standardisation has the benefit of simplifying the regulatory framework and is much less resource intensive to supervise than our current framework. Standardisation may come at the cost of less risk-sensitive approach to calculating capital requirements for IRB banks. However, we are unconvinced that bank portfolios differ in key lending areas (particularly retail lending).

4 Imposing an output floor could also limit the variation of risk weights among IRB banks. 5 Currently all banks use a common approach for measuring market risk, and we recommend continuing with this policy.

5 Ref #7885852 v1.0 Measurement of credit risk using internal models 17. If the Committee decides to continue allowing internal models to calculate capital requirements for credit risk, we recommend requiring the standardised approach for all externally-rated exposures (e.g. large corporates, banks, sovereign exposures). 18. In the December 2017 consultation paper, we proposed 5 options that varied in the degree of constraints on IRB modelling, ranging from status quo (Option 1) to removal of the IRB approach (Option 5). 19. We did not find any evidence to support status quo (Option 1) as viable given the changes in international standards and our aforementioned concerns with the current IRB framework. The IRB banks held disparate views about the preferred approach, although three out of four of them expressed some support for adopting Basel constraints on modelling for certain exposures (Option 2).6 20. However, we believe that banks’ internal models would not provide much more information than external ratings where such ratings are available. In other words, the probability of default of a BBB+ corporate exposure would be roughly equivalent for both an IRB and a standardised bank. As such, we recommend requiring standardised approach for all externally-rated exposures (Option 3). This is a slightly more conservative approach than Option 2, and slightly more conservative than Basel 3. 7 21. ANZ and WNZL argued against this Option 3. ANZ argued that as the ‘bearer’ of credit risk, banks are more incentivised to assess credit risk more accurately compared to credit rating agencies. Banks also engage more continuously with their obligors and can respond more quickly to changes in borrowers’ creditworthiness. WNZL argued a similar point, noting that since external credit rating agencies are slow to respond, this could have the effect of concentrating risk due to (assumed) external rating stability. 22. Although both banks make a good point, this issue can be addressed by requiring banks to undertake their own due diligence and to reduce the external rating when the bank considers that it is no longer justified. Dual reporting 23. We propose to require IRB banks to report their capital ratios under both the IRB and standardised approach.

6 ANZ supported Option 2 if status quo was not available. ASB supported standardising all externally￾rated exposures (Option 3) but opposed modelling overlays. Westpac supported Option 2 and standardising sovereign exposures. BNZ suggested a modified version of Option 2, which is to limit modelling to only probability of default and to prescribe a specific form of housing PD model. 7 Under the IRB approach, banks can use internal models to estimate PD (probability of default), LGD (loss given default), and EAD (exposure at default), and PD * LGD * EAD = Expected Loss. Basel 3 proposes to standardise LGD and EAD portfolios for large corporates, banks and other financial institutions, but allows modelling for PD. Our proposed approach will standardised all externally rated exposures (no modelling for LGD, EAD, or PD).

6 Ref #7885852 v1.0 24. We consider this to be the key reform, likely to bring considerable benefits in terms of enhanced transparency and thus improvements in the monitoring of capital adequacy, bank supervision and market discipline. 25. Most of the IRB banks argued against dual reporting, claiming that: i) dual reporting will likely confuse external observers, ii) that it would be onerous to explain the reasons for differences between the two numbers, and iii) it would be costly to produce the additional information. In contrast, Kiwibank, TSB, and WNZL expressed support for dual reporting to enhance transparency. 26. We note that there will be a need to explain the differences between standardised and IRB risk weights. This could be addressed via a Bulletin article or via other forms of communication. A key goal is in fact to stimulate more public debate about the risk weights used by IRB banks. 27. Implementing the standardised calculation will likely require changes to the banks’ data systems. We do not currently possess estimates of additional costs that would arise under dual reporting, but we intend to ask for cost estimates in the Quantitative Impact Study. We will consider the cost estimates in planning the transitional arrangements and timeframes. 28. There will be some flow-on work for Financial Policy arising from this reform. The current IRB and standardised approaches are not currently directly comparable, and further work will be required to increase alignment between the two approaches. For example, the IRB approach uses an estimate of ‘Exposure at Default’ (i.e. at the time the borrower defaults, how much does the borrower owe the bank?) while the standardised approach calculates capital requirements using current borrower exposure (i.e. how much does the borrower currently owe the bank?). 29. To increase the consistency between the IRB and standardised approach, increase the risk sensitivity of the standardised approach, and to reduce the costs of dual reporting for the IRB banks, we will consult on revising the standardised framework to make it more aligned with the IRB framework at a later stage. Output floor 30. To reduce the excessive variability of risk weights and to enhance the comparability of risk-weighted capital ratios, we propose to subject IRB banks to a floor on the risk￾weighted assets. Put differently, the minimum capital requirements of IRB banks will be the higher of (i) risk-weighted assets calculated using the IRB approach, and (ii) a proportion of the risk weighted assets calculated using the standardised approach. The underlying concept here is that IRB banks would not just measure (as proposed in the previous section on ‘Dual Reporting’) but hold capital on a greater-or-equal to basis against a floor. 31. As previously indicated, the calibration of the level of the floor would be considered along with minimum capital ratios. Basel and APRA have set the floor at 72.5%, and we will assess calibration of the floor (e.g. 85% or higher) once we have undertaken

7 Ref #7885852 v1.0 the Quantitative Impact Study.8 We also propose to defer the decision on how the output floor will be implemented (see Appendix 2) once we have received cost estimates. 32. Most of the IRB banks argued that, if an output floor is to be imposed, it should be applied at the aggregate level, arguing that this will be simpler to implement and will be consistent with the approach taken by the Basel Committee and APRA. The smaller banks argued that the underlying risk is more likely to be similar for retail portfolios (e.g. housing), and that a more restrictive floor (i.e. exposure or asset class basis) should be set for these portfolios. 33. We have not yet determined whether the floor will apply on an exposure by exposure basis, at each portfolio (or asset class), or at total risk-weighted assets. Due to averaging effects, where the floor is set can have a significant impact on how binding the output floor is (see Appendix 2). 34. Figure 1 provides a stylised description of the additional costs and robustness associated with the different ways of implementing dual reporting and output floor. Dual reporting on the aggregate portfolio level (which can be likened to ‘best￾endeavours’ basis) leads to the least robust but likely cheapest outcome. Requiring banks to calculate capital on both IRB and standardised basis and hold capital on a greater-or-equal to basis against a floor on each exposure leads to the most robust outcome, but the most costly option for the banks. Figure 1: Dual reporting and output floor Operational risk 35. We propose to adopt most of the new Basel Standardised Measurement Approach for calculating operational risk, replacing the Advanced Measurement Approach available to IRB banks. 36. Similar to APRA, we propose not to adopt the new Basel standard’s loss component (Basel allows banks to incorporate their actual loss experience in the estimate of

8 Note however that although APRA proposed to apply the 72.5% floor on the aggregate portfolio, APRA proposed changes to their standardised framework which results in higher capital requirements (e.g. more prescriptive requirements for mortgages). As a result, APRA’s floor is effectively higher than 72.5% on a Basel equivalent basis, after taking APRA’s changes to the capital framework into account.

8 Ref #7885852 v1.0 operational risk). We do not recommend adopting this feature as historical losses may not be a good indicator of future operational risk. 37. Nearly all of the banks supported the adoption of the new Basel standardised approach, given that it is likely to increase consistency across the banks and given the difficulty of modelling operational risk. 38. This will likely incur only a small cost for banks given the simplicity of the new approach. Next steps 39. As soon as practicable, we will publish the in-principle decisions along with a summary of submissions. 40. The next step of the capital review is designing the Quantitative Impact Study (QIS), where we will seek for banks’ input on the impact of the proposed changes to the capital framework, and ask for cost estimates of applying dual reporting. Output from the QIS will feed into the work being done to set the minimum capital ratios. 41. Subject to FSO’s agreement, we will undertake more detailed analysis of the standardised approach and consult on any changes we consider necessary at a later stage.

9 Ref #7885852 v1.0 Table 1: Summary of submissions and proposed response9 Summary of banks’ submissions Proposed response Credit risk – IRB approach No consensus among the Big 4 on the extent of limiting IRB modelling, while Kiwibank and TSB argued for the removal of the IRB approach. 3 of the 4 Australian banks, along with TSB, argued for alignment with APRA, while Kiwibank and Genworth expressed support for aligning with Basel III. Keep IRB approach but require standardised approach for externally-rated exposures; Propose to add APRA’s customisations unless there are good reasons not to (noting that APRA’s capital framework may change as a result of their capital review), and keep NZ variations when warranted. Dual reporting for IRB banks All of Big 4 except for WNZL did not support dual reporting. Propose to require dual reporting. Risk weight floor for IRB banks Two of the Big 4 supported applying a single floor on the whole portfolio, while WNZL supported applying the floor on a more granular level (asset class). Propose to require output floor. Credit risk – standardised approach Kiwibank, TSB and Genworth supported adopting Basel 3 and keeping the 0% risk weight for sovereigns. At a later stage, consult on increasing alignment of the standardised approach with the IRB approach. Operational risk All of the Big 4 banks supported adopting new Basel 3 standardised framework for operational risk, as well as adopting additional requirements for op risk management. Adopt Basel’s standardised approach but drop the option of using historical loss data to calculate op risk capital (same as APRA’s approach). Market risk Nearly all submitters agreed to keep status quo, with some arguing for adoption of Basel approach at a later stage. Keep status quo and defer review of market risk framework.

9 Please refer to #7532285 for the whole paper on ‘Proposed Response and Summary of Submissions’.

10 Ref #7885852 v1.0 Appendix 1: Comparison of RBNZ and APRA’s IRB capital framework The following figures illustrate the difference between BS2B (RBNZ’s current IRB capital framework) and APRA’s proposed changes to their capital framework (differences are highlighted in yellow). The key portfolios where RBNZ and APRA differs are: farm lending, residential mortgage loans, small and medium enterprise (SME) loans, and commercial property (Income-producing real estate). Figure 1: APRA’s proposed IRB framework

11 Ref #7885852 v1.0 Figure 2: RBNZ's current IRB framework

12 Ref #7885852 v1.0 Appendix 2: Output floor implementation The following three tables illustrate why it matters on where the output floor is set. The example assumes a floor of 100% relative to standardised approach (i.e. an IRB bank’s capital requirement will be the higher of capital requirement under IRB and standardised approach). In this example, applying the floor on each exposure (or on each loan) results in the highest marginal increase in capital requirements.

1. Exposure by exposure Pure IRB Standardised (exposure-by￾exposure) Min capital (exposure basis) Explanation for Column F Credit risk Exposure (A) RW (B) Min capital C = A x B x 8% RW (D) Min capital E = A x D x 8% Min capital F = Max (C, E) Housing Loan A $ 100.0 10% $ 0.8 30% $ 2.4 $ 2.4 Max (0.8,2.4) Loan B $ 100.0 50% $ 4.0 30% $ 2.4 $ 4.0 Max (4,2.4) Total Housing $ 200.0 $ 4.8 $ 4.8 $ 6.4 Subtotal Corporate $ 100.0 150% $ 12.0 100% $ 8.0 $ 12.0 Max (12,8) Op Risk $ 1.0 $ 2.0 $ 2.0 Max (1,2) Market Risk $ 5.0 $ 1.0 $ 5.0 Max (5,1) Total $ 22.8 $ 15.8 $ 25.4 Additional capital due to standardised floor (exposure by exposure) 2.60 = 25.4 – 22.8
2. Asset class Pure IRB Standardised (asset class) Min capital (asset class basis) Explanation for Column F Credit risk Exposure (A) RW (B) Min capital C = A x B x 8% RW (D) Min capital E = A x D x 8% Min capital F = Max (C,E) Housing Loan A $ 100.0 10% $ 0.8 Loan B $ 100.0 50% $ 4.0 Total Housing $ 200.0 $ 4.8 30% $ 4.8 $ 4.8 Max (4.8,4.8) Corporate $ 100.0 150% $ 12.0 100% $ 8.0 $ 12.0 Max (12,8) Op Risk $ 1.0 $ 2.0 $ 2.0 Max (1,2) Market Risk $ 5.0 $ 1.0 $ 5.0 Max (5,1) Total $ 22.8 $ 15.8 $ 23.8 Additional capital due to standardised floor (asset class) 1.00 = 23.8 – 22.8
3. Aggregate Pure IRB Standardised (portfolio basis) Min capital (portfolio basis) Explanation for Column F Credit risk Exposure RW Min capital RW Min capital Housing Loan A $ 100.0 10% $ 0.8 Loan B $ 100.0 50% $ 4.0 Total Housing $ 200.0 $ 4.8 Corporate $ 100.0 150% $ 12.0 Op Risk $ 1.0 Market Risk $ 5.0 Total $ 22.8 $ 15.8 $ 22.8 Max (22.8,15.8) Additional capital due to standardised floor (aggregate) 0.00 = 22.8 – 22.8