Report for the RBNZ 2025 Review of key capital settings

1 Report for the RBNZ 2025 Review of key capital settings John Vickers, 9 December 2025* Introduction This Report on the review of key capital settings for New Zealand’s banks has two parts. Section A discusses current policy, as set following the 2019 review, and examines the two proposed reform options in broad term against that background. They both involve a substantial reduction in CET1 capital relative to the 2028 destination of current policy. Option 2 has a greater reduction in CET1 capital than Option 1 but has significant LAC for Group 1 institutions. A variation on Option 2 with 1% more CET1 capital – which I term “Option 2+” below – is also discussed. Section B goes into more detail on the following issues:

1. AT1 capital
2. Risk weights
3. Counter-cyclical capital buffer
4. Competition and proportionality. The concluding section summarizes my overall view that, while risk weight granularity can be improved, Option 1 does not appear superior to current policy, and that Option 2 would require great confidence that LAC would work if circumstances required, and investment in institutional capacity. Option 2+ is in my view clearly superior to Option 2, but it still requires belief that LAC would very probably work. Section A: Current Policy and Options 1 and 2 Following the 2019 review, the RBNZ adopted a policy on bank capital that differs from the international norm, particularly in respect of its focus on CET1 (rather than less surely loss-absorbing capital) and its discipline on risk weights. Regarding international comparisons it is however important to note that NZ does not have a

I am very grateful to RBNZ staff for the open and constructive way that they have engaged with us, the external advisers, and for the RBNZ’s commitment to the transparency of our views. Working with Thorsten Beck and Elena Carletti has also been an education and a pleasure.

2 leverage ratio, whereas most other countries do, and that additional Pillar 2 requirements are not applied in NZ. So the difference should not be exaggerated. Capital levels were calibrated with the aid of a framework for cost-benefit analysis (CBA) in the 2019 review, and it is important to ask at the outset (i) whether that framework was flawed (e.g. too risk-averse) to begin with, and/or (ii) whether events since 2019 justify a significant recalibration. The question of risk appetite On (i), the 2019 review has been criticized for the “crisis-once-in-200-years” characterization of optimal capital. My understanding is that this was not a policy choice imposed by the RBNZ but a policy tested objectively against alternatives by way of the CBA. The capital level assessed as best for NZ economic wellbeing turned out to correspond roughly to 1-in-200. I have no grounds to think that the CBA was tuned to bring about that result. The CBA does not itself incorporate risk aversion. Indeed, it is based on a risk￾neutral balancing of gains and losses of NZ output – in other words zero risk aversion. Against this background it is unclear what a greater risk appetite would involve. Of course, there is room for debate about the estimates of the costs and benefits of capital that the CBA (and its sensitivity analyses) apply, but that is not a question of risk appetite. A puzzle about the 2019 analysis is its treatment of gains/losses flowing from NZ to Australia. Higher equity capital is evidently not what the Australian-owned banks want, but it is assumed to increase transfers to Australia in the CBA. Relatedly, the analysis assumes that all cost changes are passed through to borrowers, but with imperfect competition one would expect imperfect passthrough. With less-than-full cost passthrough, a reduction in capital requirements has a smaller impact on rates paid by NZ borrowers, and increases the transfer from NZ to Australia. Below I discuss the latest CBA results, including this issue. Events since 2019 Since 2019 there have been some positive developments in NZ financial regulatory law, policy and institutions that diminish risk somewhat. In my view, however, they do not significantly change the balance of costs and benefits for capital levels. On the other hand, macroeconomic and fiscal risks have risen. As a relatively small open economy, heightened uncertainty about the global trading system has elevated risks to NZ and its financial system. And the pandemic has reduced fiscal capacity,

3 though here NZ compares well to many advanced economies. If anything, the risk outlook has probably worsened since 2019. Overall, therefore, I would not say that the 2019 framework overstates risks as they appear today. Bank capital and growth: general observations Although it is often asserted and believed that higher bank capital has negative effects on economic growth, that view lacks a strong foundation. There is good reason to believe that a steep ramping up of capital requirements may be negative for lending and growth, but much less reason with respect to steady-state or with a gradual transition to higher capital levels, such as NZ’s 2019-28 programme. Indeed, the widely cited paper by Gambacorta and Shin (2018) of the BIS finds a modest positive relationship between bank capital and growth.1 More equity reduces bank borrowing costs, in line with the Modigliani-Miller (MM) proposition, and improves the bank lending channel of monetary policy. Moreover, banking crises can have huge negative effects on economic growth, as many economies have seen following the crisis of 2008. While the pre-2008 growth trajectory might well have softened even in the absence of that crisis, there are powerful reasons to believe that it has depressed output, and worsened fiscal positions, to a very serious extent. Cost-benefit analysis The CBA is a structured way to explore the tradeoff, if capital requirements are eased, between (i) some positive effect on growth from lower bank funding costs (as the MM effect on those costs is considered to be partial, not full), and (ii) the periodic output hit from greater probability (so more than 1-in-200) of crises. Further to the point above, the crisis costs used to calibrate the CBA do not look exaggerated. Before examining the CBA, it is worth asking why MM effects on bank funding costs might be less than full. One line of response is that reducing bank capital requirements lowers funding costs for banks, but not for the economy as a whole. Depending on tax codes, debt funding can have tax advantages for banks (and corresponding disadvantages for the Treasury). And lower capital might increase the prospect of taxpayer bail-out in the event of trouble, in which case the implicit taxpayer subsidy is higher when capital is lower. In other words, there might be a 1 Gambacorta, L. and H. Shin (2018), ‘Why bank capital matters for monetary policy’, Journal of Financial Intermediation, 35: 17-29.

4 wedge between the interests of the banks, especially their equity holders, and those of bank creditors and society generally. Option 1 Option 1 is estimated to reduce total capital by about 13% relative to current policy, and Tier 1 capital by 18% as AT1 is removed. This reduces estimated bank funding costs (given the partial MM offset of just 36%) but worsens the expected output loss from bank failures. In the central estimate, the net effect on GDP relative to current policy is negative: a drop of 0.18% of GDP. The estimated negative effect from greater risk of bank failure (-0.27% of GDP) is three times larger than the positive 0.09% from lower lending rates. The estimated change in wealth transfer to foreign owners, noted above, reduces the loss to 0.11% of GDP. A point made in support of the (arguably counter-intuitive) transfer effect is that Australian banks and the NZ economy both benefit from lower capital requirements in stable periods, whereas costs fall on NZ in the event of a crisis, the risk of which is greater when capital levels are reduced. However, less-than-full passthrough modifies this effect. As noted above, it both diminishes the reduction in NZ borrowing rates from lower capital requirements, and leads to some transfer to Australia, relative to the base case. Less-than-full passthrough presumably affects not only the difference between current policy and the Options, but also the optimal capital level in the first place. Sensitivity analysis is presented for four adjustments, including to the MM offset, passthrough and the estimated cost of a crisis. These give a range for the overall estimated effect of between -0.06% and -0.18% of GDP. Thus, the net benefit of Option 1 relative to current policy is negative in all four sensitivity scenarios as well as the base case. Of course, one could nudge the effect into positive territory by shutting down the MM effect and reducing the estimate of crisis cost enough, but I don’t see a principled basis for doing that. Or one could reject the CBA framework, or CBA analysis generally, but I would not go that route either because the framework looks reasonable. In sum, in overall terms current analysis calls Option 1 seriously into question compared with unchanged policy.

5 Option 2 Option 2 has CET1 capital almost 12% lower than Option 1 but adds substantial LAC – debt intended to allow bail-in if the bank ceases to be economically viable. Thus Option 2 places much more reliance on lower-tier capital than current policy or Option 1. A major question is whether LAC would in fact bear loss in a crisis. If not, then public bail-out will happen instead of private sector bail-in. In my view the CBA takes quite an optimistic view of this. Nonetheless, it estimates a roughly zero effect on NZ GDP compared with current policy. However, Option 2 is estimated to reduce transfers to Australia by 0.12% of GDP, so NZ gains by that amount. (This highlights how important the international transfer effect is in the analysis.) Again, there is sensitivity analysis around these figures. A major question for the RBNZ as it considers Option 2 is how confident to be that LAC would absorb loss in the event of a crisis. There is limited evidence on which to base this judgement because LAC essentially did not exist at the time of the 2008 crisis. There was then a coherent and established system for bondholders and creditors to absorb losses – bankruptcy (or Chapter 11 reorganization in the US or similar). But it didn’t work, and the reasons why are instructive for the assessment of LAC. Policymakers in 2008, given the crisis that they faced, especially after the Lehman bankruptcy, generally regarded the prospect of major banking institutions entering bankruptcy as intolerable. They averted that, often at large cost in terms of public funds. In short, bankruptcy was not a credible mechanism for banks in those systemic crisis conditions, so there was bail-out rather than bail-in. While there is a better prospect of LAC working in the future than bankruptcy procedures did then, there is still the question of whether the regulatory authority would trigger conversion or impose losses on holders of LAC in crisis conditions. Doing so might, for example, heighten contagion risks that the regulator is concerned to dampen. Nobody knows, and it might well depend on the type of crisis. A prerequisite for the credibility of bail-in is investment in regulatory and supervision capacity, which seems to be in its early days at the RBNZ. The events of 2023, both in the US and Switzerland, are not entirely encouraging about resolution necessarily working as planned. The US Fed departed from orthodoxy in respect of both the scope of deposit guarantee and liquidity support even though the regional bank problems were not obviously systemic. This is not to criticize the Fed, but to note how unexpected circumstances can require departure from pre-existing plans. Likewise, the Swiss authorities wrote down to zero the AT1

6 bonds in Credit Suisse, departing from the normal hierarchy of loss-bearing, a matter being litigated in the courts. This illustrates another question for non-CET1 capital, whether going- or gone￾concern. Even where the regulatory authority does invoke powers to make bondholders bear loss, there can be dispute about whether their application was properly done. Moreover, the shadow of possible dispute might be a factor in the regulator’s decision on whether to trigger loss in the first place. The central estimate of the CBA, however, is positive about Option 2. The LAC is assumed to work sufficiently well that, relative to current policy, the output loss from bank failures is quite small. The gain from lower lending rates is larger than in Option 1, as is the wealth transfer from Australia to NZ (which is most of the estimated gain). I have two concerns about this analysis, in addition to the points above about wealth transfer and passthrough. First, it is important to consider why LAC is lower-cost funding for banks. Tax is one factor, but another is the chance in an adverse situation that public bail-out will save debtholders from being bailed in. This highlights that there is some tension between LAC being both inexpensive and loss-absorbing with high probability. Second, I do not have confidence that LAC would absorb loss with high probability, even when internal to the banking group. As the Review document rightly said (Box B): “However, the existence of internal LAC instruments does not guarantee a successful recovery or resolution. The process is still a complex one”. The fact that much of the NZ banking system is under Australian ownership is part of the complexity. In some circumstances that could work to NZ’s advantage, for example if the Australian authorities stabilize a developing crisis before there is major risk to banks operating in NZ. But in other circumstances it could sharpen regulatory dilemmas for the NZ authorities because there might be strong Australian resistance to losses being imposed on their assets. Moreover, under the SPE principle the point of entry for the relevant banks is in Australia, not NZ. An exchange of letters with APRA, and “pre-positioning” of LAC, might give a degree of comfort, but I would not rely on it. No matter how good the working relationships are between the NZ and Australian authorities in normal conditions, times of crisis might be different. Much would depend on the type of crisis, which is unknowable in advance.

7 Option 2 with more CET1 A variant of Option 2 with 1% more CET1 has been analysed. I refer to this as Option 2+. In my view this is clearly superior to Option 2 in basic form. The central estimate in the CBA accordingly has Option 2+ ahead of Option 2, though not by much. It is notable that Option 2+ is the only one presented to us that is positive for expected NZ GDP compared with current policy. Option 2+ would nonetheless reduce CET1 capital, by about 9% for Group 1 institutions, and therefore places reliance on LAC. This would involve significantly less risk than Option 2, but a risk all the same. Provisional conclusion on Options 1 and 2 Of the two initial proposals, Option 2 involves a major shift towards reliance on the loss-absorbency of LAC in the event of a crisis. That appears to be quite a high-risk strategy, and I would not recommend it over Option 1. That said, it is not at all clear that Option 1 is better for NZ than current policy in overall terms. I would say the same for Option 2+, which would however be less risky than Option 2 and in my view clearly superior to it. Section B: Further issues Removing AT1 capital The proposal is to remove the current requirement relating to AT1 capital (including perpetual preference shares) and replace it with requirements for additional CET1 and/or Tier 2 capital. The reasons given are that AT1 capital is hard to issue in NZ conditions, and that international experience has shown that it might not act to stabilize a crisis. This last observation is directly relevant to the appraisal of Option 2 because doubts about the effectiveness of AT1 might well extend to LAC. I agree that the “removal of AT1” is a sensible simplification of regulation, subject to careful transitional arrangements. The question is what to put in its place. I put “removal of AT1” in inverted commas because there is no obligation on banks to use it. Rather, they may use AT1 to meet part of their Tier 1 capital requirement. One way to dispense with AT1 would be simply to remove that option and leave the Tier 1 requirement the same. Then the banks would have to meet all the Tier 1 requirement with CET1. The proposal, however, is to reduce the overall Tier 1 requirement and increase required Tier 2 capital.

8 The appropriate blend of CET1 and Tier 2 returns us to the wider question of current policy versus Options 1 and 2 that was discussed in Section A above. For the reasons given there, I would favour replacing AT1 more with CET1 than with lower-tier instruments. Risk weights While the proposed move to “more granular” risk weights appears reasonable in light of historical loss data, it entails an overall reduction of 7% in RWA compared to current policy. In other words, the risk weight adjustments would on average reduce by 7% the equity capital required for the same asset exposures, which is a sizeable reduction. For the Group 2 institutions this reduction is estimated at 15.3% because the proportionate reductions are greater for their types of lending (e.g. low￾LVR agricultural and residential mortgage lending). A further reduction in total RWA might flow from recommended further work on risk weights. An alternative way to be “more granular” would be to increase some risk weights while reducing others to achieve the desired relativities without reducing overall equity capital relative to assets. The downward-only approach to risk-weight adjustments – and further downward than in the consultation document – should at least add to caution about reducing equity capital requirements, relative to current policy, on other fronts. It should also be kept in mind that the relevant risk relates not to individual loan types but to institutions’ asset portfolios as a whole, and that risk experience (e.g. correlations) from normal periods might be an unreliable guide to crisis conditions. Moreover, lowering risk weights increases the case for a leverage cap alongside requirements based on RWAs. The NZ approach lacks this but is to be commended for its relatively disciplined approach to how much internal ratings-based (IRB) approaches can reduce risk weights below standardized measures. In particular, I support maintaining the 85% output floor and 1.2 scalar in NZ circumstances. Counter-cyclical capital buffer Current policy is for the long-run counter-cyclical capital buffer (CCyB) of CET1 to be 1.5% of RWAs by 2028, but it isn’t yet in place. The proposal is to reduce the target of 1.5% to 1% of RWAs. Even if 1.5% were maintained, risk weight reduction would already imply some diminution of the CCyB relative to overall exposures. The CCyB, once in place, allows macroprudential policy to ease capital requirements safely in a period of stress, so long as underlying prudential buffers are sound. A

9 long-run setting of 1% of RWAs does not give much room for this policy tool to work, and there appears to be a good case for remaining with the current policy target of 1.5% or even increasing it. For comparison, the Bank of England currently has a (neutral) setting of 2%. Competition and proportionality Let me make some general remarks about relationships between competition and bank capital regulation before commenting on specific proposals. The first is that banking crises can be very damaging to competition as well as to the economy generally. The UK in 2008 and Switzerland in 2023 are obvious illustrations. Regulation for financial stability therefore has some benefit in terms of competition insurance. Second, unsolved too-big-to-fail problems distort competition because bigger banks in effect have a larger implicit contingent subsidy from public funds. This is another respect in which financial stability regulation can be pro-competitive. Loosely speaking, it helps level the playing field. Third, financial stability can be detrimental to competition if it gets relative requirements out of proportion as between different types of financial institution. For example, very high capital requirements on banks and laxity toward non-bank financial intermediaries supplying substitute services could distort competition in favour of the latter. But the NZ financial sector appears bank-dominated, so this is not a live issue, and anyway the right approach to it is to regulate NBFIs properly, not to jeopardize financial stability by diluting bank regulation. Perhaps the most difficult question to judge, especially in NZ circumstances, is relative requirements on different types of deposit-taking institution. On risk weights, the NZ discipline on the extent to which IRB approaches can reduce capital requirements prevents one way in which the larger institutions can gain regulatory advantage. As noted above, the proposed “more granular” approach to risk weights reduces capital requirements more for Group 2 than Group 1 institutions, but the risk weight difference for low LVR mortgages is still striking. The difference between Options 1 and 2 has potentially large implications for competition because the move from CET1 to LAC in Option 2 applies to Group 1 only. Consultation responses reflected this, with Group 1 being pro-LAC and some of Group 2 being anti-LAC.

10 If LAC is surely loss-absorbent, there is not a competition disadvantage from Option 2. However, if LAC is not surely loss-absorbent, then that counts against Option 2 in competition, as well as other, terms because it implies an implicit contingent subsidy to Group 1 institutions. In other words, financial stability and competition considerations are not in tension on this issue. The core judgement is about how reliable a loss-absorber LAC would be in a crisis. While competition from smaller deposit-takers is clearly important, in a concentrated banking market such as NZ’s, the intensity of competition between the large banks is a key issue for how well the market serves its customers. Bank capital regulation does not strongly influence that dimension of competition. Much more on point are policy measures to improve the ability of customers to compare rival offerings and to switch between providers, such as open banking. Conclusion The most fundamental question in this review of capital settings concerns the reliability of LAC as a loss-absorber in the event of crisis. Would it absorb loss surely, and without adding to economic disruption? In my view the limited international experience to date with bail-in does not give a basis for a sufficiently high degree of confidence that it would work as designed. The design, moreover, is incomplete, and the necessary regulatory infrastructure has not yet been built. I accept that the Australia-NZ relationship might improve the chances that LAC would work well, but the fact that the largest NZ banks are under Australian ownership could on the other hand cause difficulties in some circumstances. In short, Option 2 would be a risk. On current evidence I would not take it. Option 2 with 1% additional CET1 capital (which I have called “Option 2+” above) would also be a risk, but significantly less of one. Option 1 is safer, but in aggregate terms does not look superior to (the 2028 destination of) current policy because its significant reduction of equity capital is estimated to have greater expected cost than benefit. Part of that reduction comes from the application of more granular (and on average lower) risk weights, for which a good case has been made. It also relates to the proposed form of removal of AT1, and to reduction of the planned counter-cyclical capital buffer. I do not recommend reduction of the long-run counter-cyclical capital buffer, and I would encourage consideration of a more equity-oriented replacement for AT1. Then overall steady-state equity capital could be maintained while adjusting relative risk weights and removing AT1.

11 The policy set following the 2019 review still has considerable merit for New Zealand, especially in a world where risks have not diminished. While its implementation can certainly be improved, I would not recommend significant reduction of its overall equity capital requirements.