Basel III

The committee, the city, the building

"Basel" is not a number, an acronym, or a financial concept. It is a small Swiss city of about 175,000 people on the Rhine — and the headquarters of the Bank for International Settlements (BIS), the world's oldest international financial institution. In 1974, after the failures of Bankhaus Herstatt in Germany and Franklin National in the United States triggered cross-border settlement chaos, the central bank governors of the G10 created a standing committee of bank supervisors that met four times a year at the BIS. That body — formally the Basel Committee on Banking Supervision (BCBS) — is the author of every "Basel" framework.

The BCBS has no legal authority over any bank, anywhere. Its accords are recommendations. They become binding when individual jurisdictions — the EU as a bloc, the US through its prudential regulators, the UK through the PRA, Japan through the FSA — pass them into domestic law. The framework's influence comes not from coercion but from the fact that international correspondent banking, dollar clearing, and the membership of any global G-SIB in the Basel architecture all depend on the home regulator implementing the rules. A bank in a non-implementing jurisdiction can exist, but it cannot easily participate in global finance.

A brief history: I, II, III, IV

The Basel framework has evolved across four (informally five) generations.

Accord	Year published	Headline rule	What it added
Basel I	1988	Capital ≥ 8% of RWA	First common standard. Crude risk weights (0/20/50/100%). Covers credit risk only.
Basel II	2004	Three pillars: capital, supervision, disclosure	Internal-model risk weights. Operational-risk capital. Supervisory review (Pillar 2) and market discipline (Pillar 3).
Basel 2.5	2009	Trading-book fixes	Stressed VaR, incremental risk charge, securitisation re-weighting. A patch to Basel II during the 2008 crisis.
Basel III (core)	2010–2013	Higher and better capital, plus liquidity	CET1 4.5% + 2.5% buffer. Leverage ratio. LCR and NSFR. Countercyclical and G-SIB buffers.
Basel III "Endgame" / IV	2017 (phased 2023–2028)	72.5% output floor on internal models	Standardised approach for credit risk, operational risk, market risk (FRTB). Caps internal-model RWA savings.

Basel I was elegantly simple but blunt: a corporate loan to General Electric and a corporate loan to a marginal junk-rated firm received the same 100% risk weight. Basel II tried to fix that by letting big banks use their own statistical models to assign more granular weights. Then 2008 happened — and the internal models had failed catastrophically. Subprime CDO tranches with AAA ratings carried a 20% risk weight under Basel II right up until the day they were marked down to nothing. The supervisory community had a chastening conclusion: not only did banks not hold enough capital in 2007 (around 4% common equity at major US banks), the capital they did hold was the wrong kind, and the risk weights they used to claim it was enough were wrong too.

Basel III is the answer to all three failures: more capital, better capital, harder-to-game capital ratios.

The capital stack

A bank's regulatory capital under Basel III is layered into three tiers, in descending order of loss-absorbing quality:

• Common Equity Tier 1 (CET1). The purest form: paid-in common shares and retained earnings, minus certain deductions (goodwill, deferred tax assets above thresholds, minority-interest add-backs). CET1 absorbs losses first, before any other claim on the bank. Minimum: 4.5% of RWA.
• Additional Tier 1 (AT1). Perpetual contingent-convertible bonds ("CoCos") that convert to equity or are written down when CET1 falls below a trigger (typically 5.125% or 7%). Together with CET1, they must total ≥ 6% of RWA.
• Tier 2. Subordinated debt with at least five years to maturity. Absorbs losses in resolution but not as a going concern. Tier 1 + Tier 2 must total ≥ 8% of RWA — preserving the original Basel I headline number.

Then come the buffers, all met with CET1:

• Capital conservation buffer: 2.5%. Always required. Falling below it triggers automatic restrictions on dividends, share buybacks, and discretionary bonus payments — the bank must rebuild before returning capital.
• Countercyclical capital buffer: 0–2.5%. Set by national authorities to lean against credit booms. The UK, Switzerland, and several Nordic countries have used it actively; the Fed (until very recently) has set it at zero.
• G-SIB surcharge: 1.0–3.5%. An extra CET1 requirement scaled to systemic importance. As of the most recent FSB list, JPMorgan sits at the top with a 2.5% surcharge; HSBC and Citigroup at 2.0%; most other G-SIBs at 1.0–1.5%.

For a 2.5% G-SIB at full Basel III, the effective CET1 minimum is 4.5% + 2.5% + 0% (countercyclical, if neutral) + 2.5% (G-SIB) = 9.5%. That is more than four times the effective ~2% CET1 floor banks operated under in 2007.

Risk-weighted assets — what the percentages multiply by

Capital ratios are expressed as a percentage of risk-weighted assets (RWA), not total assets. A loan's RWA equals its book value times a risk weight that depends on the obligor and collateral. Under the standardised approach, common weights look like this:

Exposure	Standardised risk weight
Cash, OECD sovereign debt	0%
OECD bank claims (≤ 3 months)	20%
Residential mortgage (well-secured)	35–50%
Investment-grade corporate	50–100%
Speculative-grade corporate	100–150%
Past-due / impaired	150%
Securitisation (subordinated)	up to 1,250% (i.e. dollar-for-dollar capital)

Big banks have, since Basel II, been allowed to use Internal Ratings-Based (IRB) models to compute their own weights. The catch is that those models have to be supervisor-approved and back-tested — and as of Basel III Endgame, their output is floored: total IRB RWA must be at least 72.5% of what the standardised approach would yield.

Worked example: a stylised mid-size bank

Consider a mid-size commercial bank with the following book:

Cash / Treasuries           : $20bn  × 0%   =     $0bn  RWA
Prime residential mortgages :  $40bn × 50%  =   $20bn  RWA
Corporate loans (IG)        :  $25bn × 100% =   $25bn  RWA
Commercial real estate      :  $10bn × 100% =   $10bn  RWA
Speculative-grade loans     :   $5bn × 150% =  $7.5bn  RWA
                                              ─────────
                            Total RWA       =  $62.5bn

The bank's required minimum CET1, assuming a 1.5% G-SIB surcharge and a 0% countercyclical buffer, is:

CET1 minimum  = 4.5% + 2.5% (conservation) + 1.5% (G-SIB) = 8.5%
              = 8.5% × $62.5bn = $5.31bn CET1

Tier 1 minimum (CET1 + AT1)    = $5.31bn + (6%–4.5%) × $62.5bn = $6.25bn
Total capital (T1 + T2)        = ≥ 8% × $62.5bn + buffers     = $8.0bn (pre-buffer)

That sets the regulatory capital floor. But Basel III also imposes a parallel leverage check that ignores RWA entirely. Total on- and off-balance-sheet exposure for this bank is, say, $110bn. The leverage ratio requires Tier 1 ≥ 3% × $110bn = $3.3bn. The risk-weighted ratio is the binding constraint here, but for very-low-risk-weight banks (mortgage-heavy, sovereign-heavy), the leverage ratio is what bites.

Liquidity: the two new ratios

Basel I and Basel II had no liquidity rule at all — they were entirely about capital. The 2008 crisis demonstrated this gap was lethal: Bear Stearns and Lehman Brothers were not insolvent under Basel II ratios when they collapsed; they were illiquid. Funding evaporated faster than capital could absorb the consequences. Basel III responds with two ratios calibrated to different time horizons.

Liquidity Coverage Ratio (LCR)

The LCR forces banks to hold enough High-Quality Liquid Assets (HQLA) to cover 30 days of stressed net cash outflows. The formula is:

LCR  =  Stock of HQLA  /  Total net cash outflows over the next 30 calendar days
     ≥  100%

HQLA is tiered: Level 1 (central-bank reserves, cash, sovereign debt of strong issuers) counts at 100% of value; Level 2A (high-grade covered bonds, certain corporates) at 85%; Level 2B (lower-grade corporates, certain equities) at 50–75% with caps. Outflows in the denominator are computed under prescribed run-off rates: insured retail deposits run off at 3–10%, unsecured wholesale funding from non-financial corporates at 40%, unsecured funding from financial institutions at 100%.

Net Stable Funding Ratio (NSFR)

The NSFR works on a one-year horizon. It requires:

NSFR  =  Available Stable Funding (ASF)  /  Required Stable Funding (RSF)  ≥  100%

ASF weights long-dated and equity-like funding most generously: regulatory capital at 100%, term debt > 1 year at 100%, stable retail deposits at 95%, less-stable retail deposits at 90%, non-financial corporate deposits at 50%, financial-institution funding < 6 months at 0%. RSF weights illiquid and long-dated assets most heavily: long-dated unencumbered loans at 65–85%, illiquid securities at 85%, encumbered assets at 100%. The ratio forces the maturity profile of a bank's funding to roughly match the maturity profile of its assets — no more financing 30-year mortgages with overnight repo.

The leverage ratio — the safety net under the model

Beneath every risk-weighted ratio sits a flat, non-risk-weighted floor: Tier 1 capital divided by total exposure (assets plus off-balance-sheet derivative exposure, securities-financing transactions, and a portion of unfunded commitments). The minimum is 3% — about 33× leverage on Tier 1. G-SIBs face a higher requirement: 1.5× the G-SIB capital surcharge added on top, so a 2.5% G-SIB faces a 4.25% leverage ratio.

The leverage ratio exists because the risk-weighted system is too easy to game. In 2007, Lehman Brothers reported a Tier 1 risk-weighted ratio of 11% — well above the requirement of the day — while running 30× leverage on a 4% pure-equity base. Risk weights claimed the bank was safe; the unweighted ratio would have screamed otherwise. Basel III makes both signals visible at the same time, and a bank must satisfy whichever is tighter.

Basel III Endgame and the output floor

The 2017 BCBS finalisation — informally Basel IV — was not a fourth accord in the lineage but a tightening of Basel III's internal-model regime. Its centrepiece is the output floor: a bank's total RWA, no matter how favourable its IRB models, must be at least 72.5% of what the standardised approach would compute. The motivation came from supervisory studies in the mid-2010s that ran identical hypothetical portfolios through different European banks' internal models. The capital requirements that came out varied by factors of two or three across banks — a result inconsistent with any reading of the regulation as having an objective meaning.

Endgame also rewrote the standardised approach for credit risk (more granular weights for residential mortgages, unrated corporates, specialised lending), abolished the Advanced Measurement Approach for operational risk in favour of a single Standardised Measurement Approach, and finalised the Fundamental Review of the Trading Book (FRTB) for market risk. The phase-in began January 2023 (50% floor) and steps up to 72.5% by January 2028.

Implementation: where we are in 2026

Jurisdiction	Status (2026)	Notes
EU (CRR3/CRD6)	In force from January 2025	Output floor phasing in; some carve-outs for unrated corporates and equity holdings.
UK	In force from July 2025	PRA implementation, very close to EU; SME support factor retained.
Japan, Australia, Canada, Singapore	Largely implemented	Phased earlier; comprehensive coverage.
United States	Materially rolled back in 2025	Trump-era proposals halve the capital increases drafted under the Biden-era 2023 NPR. Output floor and FRTB scaled back; mid-size banks largely exempted.
China	Partial	Capital framework implemented; LCR/NSFR observed; G-SIB designation accepted for ICBC, CCB, ABC, BoC.
India	Phased	Stricter on capital (RBI domestic add-ons) but slower on FRTB.

Critiques and the cost of capital

The Basel III framework has been criticised from both directions throughout its life.

• "Capital is too expensive — it crowds out lending." The industry-standard view: every extra percentage point of CET1 raises the marginal funding cost of a loan and shifts credit out of the regulated banking system into less-supervised non-banks (private credit, money-market funds, fintech lenders). Empirically, the post-2010 share of US commercial credit held outside the regulated banking system has grown substantially.
• "Capital is too cheap — banks pretend they don't have it." The Modigliani–Miller defence (Admati, Hellwig, and others): the cost of equity to a bank is endogenous to its leverage. A safer bank attracts cheaper equity. The "high cost of capital" argument is partly an artefact of subsidies (deposit insurance, implicit too-big-to-fail support) that lower the apparent cost of debt funding.
• "Internal models are more accurate than the standardised approach." Defenders of unconstrained IRB argue the output floor punishes banks with genuinely lower-risk books. Defenders of the output floor reply: the 2010s benchmarking exercises showed internal models produced wildly inconsistent results for the same portfolios, regardless of risk.
• "Basel III missed the 2023 regional-bank crisis." SVB, Signature, and First Republic failed in March 2023 with rates-driven losses on hold-to-maturity Treasuries — and SVB had been exempted from LCR and the most stringent Basel III requirements because it sat below the US tailoring threshold. Critics on both sides agreed: the rules were either applied too narrowly (pro-regulation view) or had a structural flaw in AT1 design (Credit Suisse AT1 writedown spooked markets).

The defenders' summary is simple: the 2008 crisis cost the global economy somewhere between $6 trillion and $14 trillion in lost output, depending on the methodology, plus untold political damage from bailouts. The marginal cost of carrying a few percentage points more capital is a rounding error against that.

Common pitfalls in reading Basel

• Confusing "capital" with "cash." Capital is not money sitting in a vault. It is the difference between assets and liabilities — equity. A well-capitalised bank can still be illiquid; the liquidity ratios exist precisely because capital alone does not pay the next day's withdrawals.
• Reading the headline 8% as a fixed maximum. The legacy 8% total-capital ratio is the floor before any buffer. With conservation (2.5%), countercyclical (up to 2.5%), and G-SIB (up to 3.5%) buffers, the effective total-capital requirement for a top G-SIB approaches 16.5% — twice the headline.
• Treating RWA as objective. RWA depends on the regulatory approach (standardised vs IRB), the supervisor's blessing of the model, and exposure-class definitions that vary across jurisdictions. The output floor was created specifically because RWA was demonstrably not a stable, comparable number.
• Confusing Basel III with Dodd-Frank. Dodd-Frank is US legislation; Basel III is the international template. The US implements Basel III through a separate set of rules (the Reg Q capital rules, the Liquidity Coverage Ratio rule, the FBO and tailoring rules) that are not identical to the BCBS standard, and may diverge further under the 2025 rollback.
• Forgetting the bank-specific Pillar 2 add-ons. The Pillar 1 minimum is what the rulebook prescribes for everyone. Supervisors can — and do — add bespoke Pillar 2 requirements on top of an individual bank's number, particularly for concentration, interest-rate, or model risk that Pillar 1 does not capture.