Between Risk Weights and the Leverage Ratio

Since the Great Financial Crisis, Basel III has been the central regulatory response to the question of how banks can be made more resilient to losses and systemic shocks. The reform package did more than raise minimum ratios. It tightened the definition of eligible capital, established additional buffers, introduced a non-risk-based leverage measure as a safeguard and imposed specific surcharges on global systemically important banks. The objective was clear: the greater the damage that the failure of an institution could inflict on the financial system and the real economy, the lower the probability of such a failure should be.

The study "On the comparison of capital requirements for global systemically important banks" by Patrizia Baudino, Jonathan Beissinger, Renzo Corrias, Mathias Drehmann, Egemen Eren, Burcu Erik and Nikola Tarashev addresses a pivotal issue in the current regulatory debate. Increasingly, banks are said to face a competitive disadvantage because their capital requirements appear higher than those of foreign rivals. Such comparisons look straightforward: place the required CET1 or total capital ratios side by side. According to the authors, however, this is precisely where the analytical trap lies. A ratio has both a numerator and a denominator. The denominator – risk-weighted assets – can vary substantially because of different model rules, supervisory practices and balance sheet structures.

Basel III is an international minimum standard, not a completely uniform code of law. National authorities may impose stricter requirements and adapt principles-based elements to the structure of their domestic banking systems. This flexibility is not a defect but part of the framework's design: it is intended to ensure that comparable risks are adequately capitalised even when business models, market structures and macro-financial vulnerabilities differ. At the same time, it creates a variety of regulatory capital stacks that can make international rankings and political slogans misleading.

An Unusually Broad and Laboriously Harmonised Data Set

The FSI analysis is based on a newly constructed data set covering 29 global systemically important banks across seven jurisdictions: the European banking union, Canada, Switzerland, China, Japan, the United Kingdom and the United States. The observation period runs from 2014 to 2025, covering the decisive phase of Basel III implementation. At the end of 2025, the institutions had consolidated assets of approximately USD 78.4 trillion – around 70% of the consolidated assets of the global banking system.

The particular value of the data set lies less in the number of institutions than in the granularity of the information. The authors collected data from annual reports, Pillar 3 disclosures, supervisory publications and capital market sources and organised them within a common framework. This was necessary because publicly available information is neither centralised nor consistently structured or machine-readable. That finding is itself relevant for regulation: market transparency and comparability are objectives of Pillar 3, yet the practical accessibility of the data remains limited.

The analysis covers both risk-based requirements and leverage ratio requirements. It decomposes the capital stacks into minimum requirements, legally binding buffers, bank-specific Pillar 2 add-ons and non-binding supervisory expectations. This makes visible why two banks reporting similar total capital ratios may nevertheless face very different regulatory positions.

Basel III Has Measurably Strengthened Resilience

The historical finding is initially unambiguous. Since the beginning of the implementation phase, both the regulatory requirements and the actual capital ratios of G-SIBs have increased. Crucially, the rise is visible not only in total capital but also in Common Equity Tier 1 – the highest-quality form of capital and the component that absorbs losses most directly. Banks therefore hold more equity that can absorb losses while they remain a going concern, before creditors or public support mechanisms need to be called upon.

The development has not been uniform. Requirements vary materially across institutions because national implementation, systemic importance, business models and idiosyncratic risks differ. Interestingly, actual capital ratios are more homogeneous at the end of the observation period than the regulatory requirements themselves. Banks appear to benchmark not only against legal thresholds but also against peers, rating agencies, investors, internal risk limits and explicit or implicit supervisory expectations.

Figure 01: Regulatory capital requirements and actual capital levels of G-SIBs [Source: Patrizia Baudino / Jonathan Beissinger / Renzo Corrias / Mathias Drehmann / Egemen Eren / Burcu Erik / Nikola Tarashev (2026): On the comparison of capital requirements for global systemically important banks, FSI Insights on policy implementation, No 76, July 2026, Graph 1, p. 5]

The Capital Stack: A Common Foundation, National Superstructures

The Basel foundation is clearly defined. Internationally active banks must maintain at least a CET1 ratio of 4.5%, a Tier 1 ratio of 6% and a total capital ratio of 8% of risk-weighted assets. On top of this sits a 2.5% CET1 capital conservation buffer. The countercyclical capital buffer can be built up depending on the credit cycle and released during periods of stress. G-SIBs are also subject to a systemic surcharge calibrated according to the global importance of the institution concerned.

Beyond this core, the systems diverge. The European banking union uses, among other instruments, bank-specific Pillar 2 requirements and a systemic risk buffer. Canada applies a non-binding Domestic Stability Buffer on a system-wide basis to systemically important institutions. Under its too-big-to-fail regime, Switzerland has introduced specific CET1 and AT1 buffers and does not allow Tier 2 capital to meet certain G-SIB requirements. China imposes a higher minimum CET1 ratio of 5%. The United Kingdom combines binding Pillar 2A requirements with a non-binding Pillar 2 buffer. The United States uses a Stress Capital Buffer and a domestic G-SIB methodology whose Method 2 generally produces higher surcharges than the international Basel methodology.

The number of elements making up the overall capital stack ranges from four to eight across the jurisdictions studied. This illustrates why an isolated comparison of individual buffers can lead to false conclusions. A lower surcharge in one category may be offset by a higher requirement elsewhere, stricter capital quality rules or more conservative risk weights.

From Just Under 12% to Almost 17% 

For year-end 2025, the study identifies substantial differences in average total capital requirements. Including non-binding buffers, the figures range from around 11.9% in Japan and China to 16.8% in the United Kingdom. The European banking union stands at almost 15.9%, Canada at about 15.1% and Switzerland at just under 15%. For US G-SIBs, the figures are approximately 14.0% under the standardised approach and 13.7% under the advanced approach.

At first glance, this range appears to permit a clear ranking. The study cautions against such an interpretation. The total ratio includes capital instruments of different quality. CET1 absorbs losses most reliably, whereas Additional Tier 1 and Tier 2 instruments have different loss-absorption mechanisms and trigger points. When capital stacks are grouped by capital quality, CET1 requirements are more closely aligned: including non-binding buffers, they average from just over 8% to just under 12%. The strongest international convergence is therefore found in the highest-quality form of capital.

Table 01: Weighted averages as percentages of RWA or total assets [Source: Patrizia Baudino / Jonathan Beissinger / Renzo Corrias / Mathias Drehmann / Egemen Eren / Burcu Erik / Nikola Tarashev (2026): On the comparison of capital requirements for global systemically important banks, FSI Insights on policy implementation, No 76, July 2026, Table 1, p. 8]

Why Actual Ratios Converge More Strongly Than the Rules

The chart of capital stacks reveals a striking pattern: the required regulatory ratios and their composition differ considerably, while the actual total capital ratios of G-SIBs are comparatively close together. The authors interpret this as the result of several adjustment mechanisms. Banks deliberately hold a management buffer above legal thresholds to absorb unexpected losses, earnings volatility, changes in RWA and regulatory model adjustments.

Another factor is the concern about restrictions on distributions. If a binding buffer threshold is breached, many jurisdictions automatically restrict dividends, share buybacks, variable remuneration or coupon payments on AT1 instruments. The maximum distributable amount threshold therefore acts as an economically hard boundary, even though capital buffers are formally intended to be usable in stress. Banks avoid moving too close to this threshold because even the prospect of distribution restrictions can undermine market confidence and raise funding costs.

Non-binding supervisory expectations also have real effects. Pillar 2 Guidance in the banking union, the Pillar 2 Buffer in the United Kingdom and the Domestic Stability Buffer in Canada do not immediately trigger automatic distribution restrictions. Nevertheless, they influence capital planning, supervisory dialogue and market expectations. In Japan, similar signals are conveyed more through stress-test benchmarking and supervisory communication. What does not appear as a binding buffer in a public comparison table may therefore still form part of the effective capital target in practice.

Figure 02: Risk-based total capital ratios - regulatory capital stacks and actual capital levels [Source: Patrizia Baudino / Jonathan Beissinger / Renzo Corrias / Mathias Drehmann / Egemen Eren / Burcu Erik / Nikola Tarashev (2026): On the comparison of capital requirements for global systemically important banks, FSI Insights on policy implementation, No 76, July 2026, Graph 2, p. 12]

The Denominator Matters: Risk-weighted Assets as a Comparability Problem

The study's most important analytical contribution is the connection it draws between capital ratios and risk weights. A risk-based capital ratio relates regulatory capital to risk-weighted assets. If RWA density – the share of RWA in total assets – increases, the absolute capital requirement rises proportionately for a given ratio. Conversely, a low RWA density can qualify the apparent severity of a high percentage requirement. Looking only at the capital ratio therefore reveals only half of the regulatory burden.

The differences are substantial. At the end of 2025, the average RWA density of Chinese G-SIBs was almost 55% of total assets and that of US G-SIBs around 45%. In the European banking union, Canada, Switzerland, Japan and the United Kingdom, the figures were mostly between approximately 26% and 31%. These differences may reflect genuinely different portfolios and risk profiles. But they also reflect methodological choices: standardised approaches typically produce higher risk weights than internal models, and national authorities restrict model use to different degrees.

At the jurisdictional level, the study finds a negative relationship between RWA density and CET1 requirements. Jurisdictions with higher risk weights tend to have lower capital ratio requirements, while systems with lower RWA densities more often use higher ratios or Pillar 2 add-ons. This is not proof of deliberately coordinated compensation, but it is strong evidence that supervisory systems do not calibrate the numerator and denominator independently.

Figure 03: Average RWA density over time and in relation to CET1 capital requirements [Source: Patrizia Baudino / Jonathan Beissinger / Renzo Corrias / Mathias Drehmann / Egemen Eren / Burcu Erik / Nikola Tarashev (2026): On the comparison of capital requirements for global systemically important banks, FSI Insights on policy implementation, No 76, July 2026, Graph 3, p. 14]

Internal Models: More Precise Risk Measurement or a Source of Regulatory Dispersion?

Basel III generally permits two routes for calculating risk-weighted assets. Standardised approaches assign regulatory risk weights. Internal models use banks' own data and parameter estimates, provided that strict eligibility criteria are met and supervisory approval is obtained. Models can capture risk more precisely, but they also create model risk and room for discretion that may generate unwarranted variation across banks.

National frameworks place different emphasis on these approaches. US G-SIBs use internal models for part of their risks, but in practice most are constrained by the standardised approach. For credit risk, this leads to a materially higher RWA density than would result from approved internal models. Chinese G-SIBs mainly use the Foundation IRB approach: banks estimate probability of default themselves, while other parameters are set by regulators. This structure is more conservative than fully model-based Advanced IRB methods and contributes to China's high RWA densities.

When the effects of national model restrictions are removed counterfactually and similar measurement approaches are compared, part of the observed dispersion narrows. This supports the view that international differences are shaped not only by the economic risk of portfolios but also by measurement methodology. Caution remains necessary, however: without identical reference portfolios it is impossible to determine precisely how much of the difference reflects business models, credit quality, asset mix or regulatory model policy.

Figure 04: Use of internal models by G-SIBs and credit risk RWA density before model restrictions [Source: Patrizia Baudino / Jonathan Beissinger / Renzo Corrias / Mathias Drehmann / Egemen Eren / Burcu Erik / Nikola Tarashev (2026): On the comparison of capital requirements for global systemically important banks, FSI Insights on policy implementation, No 76, July 2026, Graph 4, p. 16]

The Output Floor Is Intended to Limit Model Benefits

The final Basel III reforms respond to unwarranted variability in internal models through the output floor. Once fully implemented, a bank's total RWA may not fall below 72.5% of the RWA that would result from the standardised approaches. The floor therefore limits the capital benefit of internal models without abolishing them. It is intended to improve comparability and prevent model-driven low risk weights from undermining the credibility of capital ratios.

However, jurisdictions are introducing the output floor at different times and under different transitional arrangements. The banking union and Switzerland have begun phased implementation; in the United Kingdom, implementation is scheduled for 2027 and is intended to rise to 72.5% by 2030. US reform proposals instead envisage a more standardised single stack without adopting the Basel output floor in the same form. During the transition, international comparability may therefore initially deteriorate before stronger alignment becomes possible.

The Leverage Ratio: A Simple Backstop with a Complicated Implementation

Risk-based capital ratios are sensitive to modelling assumptions. Basel III therefore introduced the leverage ratio as a non-risk-based safeguard. It relates Tier 1 capital to a broad exposure measure covering on-balance-sheet positions, derivatives, securities financing transactions and off-balance-sheet commitments. The international minimum is generally 3%. G-SIBs face an additional buffer that, under the Basel standard, equals half of the risk-based G-SIB surcharge.

Even for this apparently simple instrument, the study finds substantial national differences. China requires a minimum leverage ratio of 4% for G-SIBs. Switzerland adds systemic buffers and institution-specific requirements to the Basel minimum. The United Kingdom excludes central bank reserves from the exposure measure but partly compensates for this with a higher minimum and additional buffers. Japan follows a similar approach. In the United States, a flat 5% Enhanced Supplementary Leverage Ratio applied until March 2026; since April 2026, the buffer has been linked more closely to the Basel G-SIB surcharge.

The leverage ratio is therefore neither entirely model-free nor free from national design choices. Although it dispenses with risk weights, jurisdictions differ in their definitions of the exposure measure, netting rules, treatment of central bank reserves and required capital quality. The study also shows that actual leverage ratios converge less strongly than risk-based capital ratios. This underlines that the leverage ratio functions as a genuinely complementary backstop rather than merely reproducing the same information in a simplified form.

Why the Uneven Playing Field Narrative Falls Short

The question of a level playing field is politically legitimate. Higher capital requirements can affect banks' cost of equity, business models and pricing. The study shows, however, that claims of competitive disadvantage are analytically robust only when all components are considered together: required ratios, capital quality, RWA measurement, the leverage ratio, binding distribution thresholds and non-binding supervisory expectations.

A jurisdiction with a high headline capital ratio may simultaneously permit low RWA densities. Another may require lower ratios but use more conservative standardised approaches. A third may rely on non-binding buffers that are nevertheless economically effective. Requirements for subsidiaries in host jurisdictions can also raise the consolidated capital planning needs of international groups. A ranking based on a single ratio therefore does not measure the full regulatory capital burden.

Comparing actual capital levels does not provide a complete answer either. Similar reported ratios may reflect different distances to minimum thresholds, business models or market expectations. They may also indicate that international investors, rating agencies and supervisors informally promote some degree of convergence. The decisive question is therefore not which jurisdiction reports the highest percentage, but whether comparable risks are backed by comparable loss-absorbing capacity.

Regulatory Modernisation: Simplify Without Diluting Protection

Several jurisdictions are reviewing their capital frameworks. The United States has proposed refining and generally reducing its domestic G-SIB surcharge and replacing the existing dual structure of standardised and advanced approaches with a more unified approach for the largest banks. The already effective reform of the US leverage ratio is intended to reinforce its role as a backstop and prevent it from becoming the binding primary requirement too often.

In the United Kingdom, final Basel III implementation has been postponed to 2027. At the same time, a system-wide Tier 1 benchmark has been lowered from around 14% to around 13%, justified by improved risk measurement and lower systemic importance of individual banks. In the European Union, policymakers are discussing simplification of the complex buffer architecture, greater harmonisation and possible changes to AT1 instruments. Switzerland, in turn, has proposed stronger CET1 backing for foreign participations at parent-company level following the experience of Credit Suisse.

These reforms may improve transparency and manageability, but they also change the basis of comparison. Simplification is not automatically deregulation, and a lower ratio does not necessarily imply lower resilience if risk weights or measurement methods become more conservative at the same time. Conversely, formal simplification can weaken loss-absorbing capacity if capital instruments or buffers are removed without equivalent replacement. International dialogue therefore remains essential to avoid regulatory fragmentation and competition through lower standards.

What the Study Delivers 

The FSI analysis provides a much stronger empirical foundation for international comparison. For the first time, it documents numerous components of capital stacks in a harmonised format over 11 years. Yet it is not a final measure of the total regulatory burden. Resolution requirements and liquidity rules are deliberately excluded. Not all supervisory expectations are publicly disclosed; for some Pillar 2 components, averages or approximations must be used.

Nor can the correlation between RWA density and capital ratios establish clear causality. Lower RWA densities may reflect less risky portfolios, more permissive models or both. Higher capital ratios may deliberately compensate for model risk, but they may also be driven by other systemic or microprudential factors. Perfect comparability would require identical portfolios to be assessed under every national rulebook and supervisory practice – a counterfactual exercise that cannot be fully conducted in reality. This is precisely why the study's central methodological message matters: individual metrics must not be interpreted in isolation. A bank's capital position is the outcome of a system comprising the numerator, denominator, capital quality, buffers, thresholds, model approvals and supervisory expectations. Any political or economic assessment that singles out one component risks creating false precision.

Conclusion

Basel III has materially strengthened the resilience of global systemically important banks. Its common minimum architecture prevents a regulatory race to the bottom and provides the basis for international market and supervisory discipline. National flexibility has simultaneously created a complex mosaic. Different buffers, Pillar 2 instruments, model rules and leverage ratio approaches are often functionally connected and may offset one another.

The study therefore provides an important corrective to the competitiveness debate. A high regulatory ratio is not automatically stricter, and a low ratio is not automatically more lenient. Only the combination of the capital ratio and RWA density reveals the absolute capital need; only consideration of capital quality shows the effective loss-absorbing capacity; and only attention to distribution thresholds and supervisory expectations identifies the operative management target.

For supervisors, policymakers and market participants, the implications are clear: data should be published in a more centralised, consistent and machine-readable form. Reforms should make capital stacks easier to understand without weakening resilience. And international comparisons should focus less on producing a single league table and more on functional equivalence. In bank capital regulation more than in many other fields, the same percentage can have a completely different economic meaning.

Bibliography and Further Reading: 

• Baudino, Patrizia / Beissinger, Jonathan / Corrias, Renzo / Drehmann, Mathias / Eren, Egemen / Erik, Burcu / Tarashev, Nikola (2026): On the comparison of capital requirements for global systemically important banks. FSI Insights on policy implementation, No 76, Bank for International Settlements, July 2026.