Not all bank capital is equal, even when it all shows up on the same balance sheet. CET1, AT1, and Tier 2 rank on one key property: how reliably each absorbs losses while the bank keeps running. Get that ranking wrong, and you'll misread every capital ratio you look at afterward, including why a 14% CET1 ratio didn't save Credit Suisse. Here's the stack, properly explained.
CET1 vs AT1 vs Tier 2 is a ranking on one property: how reliably the instrument absorbs losses while the bank is still operating.
CET1 is common equity and retained earnings, and absorbs anything, at any time.
AT1 is perpetual subordinated instruments that convert or write down on a trigger and are meant to absorb losses in a going concern.
Tier 2 is dated subordinated debt that absorbs losses only in insolvency or resolution.
Everything else in the numerator, including the deduction regime that makes it credible, follows from that ordering.
The rest of this piece is that ordering, worked through in detail.
Capital is loss-absorbing funding, not a pile of cash set aside somewhere. It is a liability-side concept: the portion of the balance sheet financed by claims that can take losses without the bank defaulting. The tiers exist because that property comes in degrees.
Basel expresses the degrees as three properties, present in descending strength across the tiers:
permanence (no maturity, no incentive to redeem),
payment flexibility (distributions fully discretionary and non-cumulative),
and subordination (position in the loss queue).
CAP10 in the consolidated framework turns those properties into eligibility criteria: fourteen for CET1, sixteen for AT1, and ten for Tier 2. The original 2011 text ran 14/14/9; the point-of-non-viability and special-purpose-vehicle issuance requirements were later folded into the numbered lists.
The other distinction to hold onto is going concern versus gone concern. Going-concern capital absorbs losses while the bank keeps operating, so its purpose is to make failure unlikely. Gone-concern capital absorbs losses after failure, so its purpose is to make failure survivable. CET1 is unambiguously the first. Tier 2 is unambiguously the second. AT1 was designed to be the first and, in the one systemic test to date, behaved like the second, which is why the tier remains the most contested part of the numerator.
CET1 is common shares meeting all criteria, share premium, retained earnings, accumulated other comprehensive income and disclosed reserves, plus qualifying minority interest, minus the regulatory adjustments below.
Two of the fourteen criteria do quiet but important work. Distributions must be paid only out of distributable items, and must never be linked to the amount paid in at issuance, which blocks the dividend-guaranteed instruments that used to be dressed as equity. And the instrument must take “the first and proportionately greatest share of losses”, which is the criterion that stops financial innovation at the bottom of the stack.
CET1 matters disproportionately because it is the numerator of almost everything that binds. The 4.5% Pillar 1 minimum is CET1. Every buffer is met in CET1. The maximum distributable amount trigger is measured in CET1. In practice, a bank’s CET1 ratio is the constraint it manages to, and the Tier 1 and total capital minimums are rarely the binding ones.
Qualifying minority interest carries a rule worth knowing: capital a consolidated subsidiary issued to third parties counts only as far as it covers that subsidiary’s own requirements, and the surplus is excluded. Without that rule, a group could manufacture consolidated capital out of shares that could not absorb losses where the group actually needs them to.
AT1 must be perpetual. No maturity, and no incentive to redeem, which makes coupon step-ups disqualifying. It is callable no earlier than year five, and then only with supervisory approval plus either replacement or a demonstration of headroom. Coupons are fully discretionary and non-cumulative, payable only from distributable items. Canceling an AT1 coupon is not a default and must not trigger anything else, which is the entire point of the instrument and also the reason banks defend their MDA headroom so fiercely: the rulebook treats a skipped coupon as normal, and the market does not.
Instruments classified as liabilities for accounting purposes must convert to equity or be written down when the group’s CET1 ratio falls below a trigger of at least 5.125%. Issuers may set higher triggers and often do, with 7% “high-trigger” instruments dominating the Swiss market. Write-down may be permanent or temporary depending on the contract and the jurisdiction.
Then comes the layer that overrides all of it. Since January 2013, every AT1 and Tier 2 instrument must additionally absorb losses at the point of non-viability (PONV): write-off or conversion when the relevant authority determines the bank would otherwise fail, or that public support is required. That obligation can be met by a contractual term or by a statutory resolution regime recognized as achieving the same effect.
PONV, not the contractual trigger, is where loss absorption has actually happened. Credit Suisse’s AT1 was extinguished by a decree invoking both the instruments’ viability-event clauses and emergency-ordinance authority, with reported CET1 above every contractual trigger in its instruments, and the Swiss courts have since found the decree wanting at first instance.
The design lesson supervisors drew is that mechanical triggers are set so low that non-viability arrives first, so the going-concern label on AT1 describes an intention rather than an observed behavior. Four reform options are on the table and none has been adopted: higher triggers so that conversion happens while the bank is genuinely a going concern, mandatory conversion rather than write-down structures, coupon-stopper redesign, and honest reclassification of AT1 as gone-concern capital.
Tier 2 requires an original maturity of at least five years, with recognition amortizing straight-line over the final five years, so a bond bought at ten years contributes fully for five and then declines. It must be subordinated to depositors and general creditors, and carry no acceleration rights other than in liquidation. Cumulative coupons are permitted.
Tier 2 protects depositors and senior creditors in insolvency or resolution and absorbs nothing while the bank operates. It makes no claim to do otherwise, and that is the whole of its design.
|
|
CET1 |
AT1 |
Tier 2 |
|
Typical instrument |
Common shares, retained earnings, AOCI |
Perpetual subordinated notes |
Dated subordinated debt |
|
Maturity |
None |
Perpetual, call from year 5 with approval |
Minimum 5 years original |
|
Coupon or dividend |
Fully discretionary |
Fully discretionary, non-cumulative |
Cumulative permitted |
|
Loss absorption in a going concern |
Continuous |
On trigger, minimum 5.125% CET1 |
None |
|
Loss absorption at PONV |
By definition |
Yes, contractual or statutory |
Yes, contractual or statutory |
|
Capital recognition |
Full |
Full while eligible |
Amortizes over final 5 years |
|
Eligibility criteria (CAP10) |
14 |
16 |
10 |
|
Counts toward |
4.5% CET1, 6% Tier 1, 8% total |
6% Tier 1, 8% total |
8% total only |
An eligibility list alone would not make the numerator credible, because some assets are worth nothing precisely when capital is needed. The deduction regime (CAP30) applies almost entirely at CET1 level, and its design principle is a single sentence: anything whose value is contingent on the bank’s own survival or future profits cannot count as protection against its failure.
Fully deducted: goodwill and other intangibles, net of associated deferred tax liabilities; deferred tax assets dependent on future profitability arising from carry-forwards; any shortfall of provisions against expected loss under IRB; defined-benefit pension fund assets; gains on sale from securitisation; and own-credit gains on fair-valued liabilities. The full deduction schedule, including prudent valuation and the treatment of holdings in other financial institutions, has its own piece. What matters for the stack is the direction: every deduction lands on CET1, the top tier, not on the tier that generated the exposure.
Holdings of other financials’ capital instruments are treated in two buckets, to prevent double leverage. Without the rule, two banks could each “raise” capital by buying the other’s shares, creating recorded capital with no new loss absorbency in the system.
Non-significant investments (10% or less of the issuer’s common shares): aggregate holdings above 10% of the bank’s CET1 are deducted pro-rata by tier, the “corresponding deduction” approach.
Significant investments (more than 10%): non-CET1 instruments are fully deducted, while common shares enter the threshold basket alongside mortgage servicing rights and temporary-difference DTAs. Each of the three is capped at 10% of CET1 and the three together at 15%, with amounts below the caps risk-weighted at 250% rather than deducted.
National variation in deductions is among the quietest and most material divergences in the framework, so a cross-border comparison of headline CET1 ratios without a look at the deduction schedules is not a comparison.
Three seams are worth tracking, because each is live.
AOCI. Basel puts unrealized securities gains and losses in CET1 through accumulated other comprehensive income. The US lets banks below its top two categories elect out, and what that election did at SVB is the reason this deviation is the one with a demonstrated failure attached to it.
AT1 design. The going-concern label survived its first real test in name only, and the reform options remain options.
Insurance participations. The EU’s “Danish compromise”, risk-weighting rather than deducting insurance subsidiaries under conglomerate rules, is a material and RCAP-flagged softening that CRR3 left in place.
The tiers rank loss absorbency in a going concern: CET1 absorbs continuously, AT1 on a trigger, Tier 2 not at all.
CET1 is the tier that binds. It is the numerator of the 4.5% minimum, of every buffer, and of the MDA trigger. The distinction between an unusable minimum and a usable buffer is where that bites.
AT1’s contractual trigger sits at a minimum of 5.125% CET1, but PONV has been where loss absorption actually occurs, which is the tier’s central design problem.
Tier 2 is gone-concern capital and does not pretend otherwise; recognition amortizes over its final five years.
The deduction regime is not administrative detail. It is what stops the numerator from counting assets that vanish in failure, and it is where jurisdictions diverge most quietly.
Above the three tiers sits a separate resolution layer, TLAC and MREL, sized to fund failure rather than prevent it.
For the complete picture, including how TLAC and MREL sit above this stack, how the EU, UK and US have each implemented these deduction rules, and what the 2023 failures revealed about AT1 design, read the full guide, "Capital Regulation: A Complete Reference — From Zero to Expert: Basel, the EU, the UK and the US Frameworks."
To see how Mirai Regulatory Reporting tracks CET1, AT1, and Tier 2 eligibility, deductions, and buffer distance consistently across Basel, EU, UK, and US rulebooks, learn more here.