Reader question: How can a regulatory algorithm turn a bank’s size, network position and market activity into a single systemic-importance score — and why can that score change even when the bank itself barely changes?
The short answer is that the Basel global systemically important bank, or G-SIB, methodology is a relative indicator-scoring system. For each indicator, a bank’s reported amount is divided by the aggregate amount for the assessment sample and multiplied by 10,000. Indicator scores are combined into five categories, the categories are given equal overall weight, and the resulting score is compared with a cutoff and bucket boundaries. Higher buckets imply higher common-equity Tier 1 loss-absorbency requirements.
What this page does — and does not do
This is a mathematics and algorithms article. It explains how a systemic-importance scoring rule transforms data into a regulatory classification. It does not judge whether any named bank is safe, predict bank failure, recommend bank shares or deposits, or replace the Basel Committee and Financial Stability Board documents. Jurisdictions implement G-SIB requirements through their own legal and supervisory frameworks.
The problem the algorithm is trying to solve
A very large or deeply connected bank can create costs that are not captured by looking only at its own probability of default. If it fails, payment links, funding relationships, securities markets, custody, underwriting, derivatives and cross-border positions can transmit disruption to other institutions and the wider economy.
That means “systemic importance” is not the same variable as “stand-alone riskiness.” A bank can be well capitalised and still be systemically important because replacing its services rapidly would be difficult. Conversely, a smaller risky institution can have a high probability of distress without meeting the G-SIB systemic-footprint threshold.
The five-category architecture
The Basel indicator-based approach organises systemic footprint into five broad categories, each receiving 20% of the overall score:
- Size — the scale of the bank’s overall exposures.
- Cross-jurisdictional activity — claims and liabilities that span borders.
- Interconnectedness — financial-system assets, liabilities and securities outstanding that connect the bank to other institutions and markets.
- Substitutability / financial-institution infrastructure — activities such as payments, custody, underwriting and trading where abrupt withdrawal can be hard for the system to replace.
- Complexity — activities such as OTC derivatives, trading assets and Level 3 assets that can make resolution and risk transfer more difficult.
Most indicators within a category are equally weighted within that category. The substitutability category has a more detailed indicator-weight structure under the current methodology, and its category score is capped at 500 basis points. That cap matters because extraordinarily large payment or custody activity should not mechanically dominate the complete systemic-importance measure.
The core equation: a relative share turned into basis points
For indicator j, the mechanical score can be represented as:
indicator scorej = bank indicator amountj / sample aggregate amountj × 10,000.
Suppose a bank reports €500 billion for an indicator and the assessment-sample denominator for that indicator is €25 trillion. Its score for that indicator is:
500 / 25,000 × 10,000 = 200 basis points.
The number 200 here is a score, not a 2% probability of failure and not automatically a 2% capital surcharge. That distinction is crucial.
From indicators to category scores
Within each category, the relevant indicator scores are combined using the methodology’s weights. The five resulting category scores are then averaged, giving the overall systemic-importance score. The Basel methodology rounds the overall score to the nearest whole basis point for bucketing.
A simplified worked example makes the aggregation visible. Imagine an illustrative bank has category scores of:
- Size: 160 bp
- Cross-jurisdictional activity: 220 bp
- Interconnectedness: 140 bp
- Substitutability: 260 bp
- Complexity: 120 bp
The simple average is:
(160 + 220 + 140 + 260 + 120) / 5 = 180 bp.
Under the Basel cutoff structure, 180 bp is above the 130 bp cutoff and lies in the first G-SIB bucket. This is only an arithmetic illustration; real assessments use the exact reported indicators, official denominators, applicable methodology, data-quality processes and supervisory judgment.
How scores map to buckets and surcharges
The Basel Committee’s current cutoff page sets the G-SIB cutoff at 130 bp and uses 100 bp bucket widths. The published bucket structure is:
- 130–229 bp: 1.0% higher loss absorbency.
- 230–329 bp: 1.5%.
- 330–429 bp: 2.0%.
- 430–529 bp: 2.5%.
- 530–629 bp: 3.5%.
The surcharge is an additional common-equity Tier 1 requirement expressed relative to risk-weighted assets under the applicable framework. It should not be confused with the score itself. A bank with a 180 bp systemic score does not receive a “1.8% surcharge”; it falls into the bucket determined by the published boundaries.
The denominator is part of the algorithm, not background data
The most interesting mathematical feature of the G-SIB score is that it is relative. Each indicator divides the bank’s value by the aggregate value for the assessment sample. Therefore, a bank’s score can move even if its own measured activity is unchanged.
Example: if the bank still has €500 billion of a particular activity but the sample denominator grows from €25 trillion to €30 trillion, its indicator score falls from 200 bp to about 166.7 bp. Nothing about the bank’s numerator changed. The peer system changed around it.
The reverse is also possible. If peer activity contracts, a constant bank can become a larger share of the system and its score can rise. This is a powerful counterexample to the intuition that a regulatory score is always an absolute measurement of the bank alone.
Why the official denominator file matters
The Basel Committee publishes denominator values used in the assessment. For the end-2024 data used in the 2025 assessment cycle, the published file includes aggregate figures for total exposures, cross-jurisdictional claims and liabilities, intra-financial-system assets and liabilities, securities outstanding, assets under custody, payments activity, underwriting, trading volume, OTC derivatives, Level 3 assets and trading/available-for-sale securities.
A reproducible calculation must therefore freeze both sides of the ratio: the bank numerator and the official denominator for the same assessment cycle. Mixing a 2024 bank numerator with a 2023 denominator can produce a plausible-looking but invalid score.
Current-cycle evidence: the 2025 FSB list
The Financial Stability Board published the 2025 list of G-SIBs on 27 November 2025, using end-2024 data. The list contained 29 banks. The set of banks was unchanged from the 2024 list, although some institutions moved between buckets. Under the FSB notice, increases in higher loss-absorbency requirements from that list apply from 1 January 2027. The next annual list is expected in November 2026.
This illustrates another important point: classification is a periodic process tied to a defined data vintage. A live internal dashboard is not the same thing as the official annual G-SIB designation.
Inputs and outputs of a robust scoring implementation
Inputs should include the assessment date, bank identifier, reporting perimeter, every raw indicator, currency and unit controls, the official denominator dataset, methodology version, indicator weights, substitutability cap, cutoff and bucket table.
Useful outputs go beyond “G-SIB: yes/no.” They include each raw numerator, denominator, indicator score, category score, capped versus uncapped substitutability result, overall score, rounded score, bucket, implied surcharge and an audit trail showing exactly which vintage produced the result.
Assumptions hiding inside the score
- Reported indicators are comparable. Definitions and consolidation perimeters must be applied consistently across banks.
- Systemic footprint is adequately represented by the selected indicators. No finite indicator set captures every channel through which distress can propagate.
- Relative shares are meaningful. A bank’s systemic importance is evaluated partly against what the rest of the assessment sample is doing.
- Weights are policy choices. Equal category weights simplify the multidimensional problem but do not prove that each dimension has identical real-world harm.
- Bucket boundaries are deliberately discrete. Two banks one basis point apart can face different surcharge buckets near a threshold.
Evidence polarity: what strengthens or weakens confidence?
Evidence for a score includes reconciled regulatory returns, correct reporting perimeter, official denominators from the matching assessment year, independent recalculation, agreement between published disclosure and internal source data, and stable unit conversions.
Evidence against confidence includes large unexplained revisions, denominator-vintage mismatches, category totals that cannot be reproduced, unit errors such as millions versus billions, missing cross-border entities, a substitutability score that breaches its cap without explanation, or abrupt movements that are unsupported by business activity or peer-denominator changes.
Failure modes and counterexamples
Unit catastrophe: entering millions as billions multiplies a numerator by 1,000. Because the scoring formula is linear in the numerator, the indicator score inherits the same scale error.
Peer-denominator blindness: attributing every score move to the bank itself can be wrong. Decompose the change into numerator and denominator effects.
Threshold illusion: a one-basis-point move across a bucket boundary can change the required surcharge even though the underlying systemic footprint barely moved. That is a property of discretisation, not evidence of a sudden jump in economic danger.
Category averaging hides concentration: two banks can obtain the same overall score with very different structures — one dominated by payments, another by cross-border and derivatives activity. The overall scalar is not a substitute for the category vector.
Mechanical-score absolutism: the framework permits supervisory judgment in the identification process. The indicator score is central, but it is not the only possible information relevant to classification.
Diagnostics: how to test the weak links
- Recompute every indicator score directly from raw numerators and official denominators.
- Run a unit check by reconstructing known public disclosure examples.
- Decompose year-to-year score changes into the bank’s numerator effect and the sample-denominator effect.
- Test values immediately below, at and above 130 bp and every subsequent bucket boundary.
- Verify that the substitutability category cap is applied at the correct stage rather than to unrelated indicators.
- Preserve both unrounded and rounded scores so a rounding rule cannot silently alter the bucket.
- Compare the total score with the vector of category contributions; a plausible total can conceal one impossible component.
What would falsify confidence in the result?
Confidence should be withdrawn if independent recalculation does not reproduce the score; if changing an official denominator leaves the corresponding indicator score unchanged; if a category’s weights do not reconcile to its required contribution; if the same raw data produce different results under the same methodology version; or if the published bucket cannot be recovered from the stored rounded score and cutoff table.
Alternatives answer different questions
G-SIB scoring is not a default-probability model, a market-risk measure, a stress test or a network-contagion simulation. Those tools can complement it. A network model, for example, can represent bilateral exposures and feedback loops explicitly; the G-SIB methodology instead needs a transparent, internationally comparable indicator system that can be applied across many banks.
How this connects to the surrounding mathematics
The systemic score is easier to understand alongside interbank-network contagion, where graph structure makes transmission channels explicit. Once a bank is designated systemically important, connect the result to the broader bank-capital constraint and leverage constraints. In a distress scenario, the institutional logic connects to bank resolution and bail-in waterfalls.
Verification and update triggers
Store the data year, assessment year, denominator release, Basel methodology version, indicator definitions, score calculation, bucket table and publication date. Re-check the implementation when the Basel Committee revises the G-SIB methodology, when denominator data are refreshed, when the FSB publishes the annual list, when reporting definitions change, or when a merger, disposal or structural change materially alters a bank’s perimeter.
Primary and high-quality references
- Basel Committee on Banking Supervision, Global systemically important banks: assessment methodology and indicator resources.
- Basel Committee, G-SIB cutoff score and bucket boundaries.
- Basel Committee, G-SIB denominators.
- Financial Stability Board, 2025 list of global systemically important banks, published 27 November 2025.
- Basel Committee, Basel Framework, SCO40 — global systemically important banks.
Educational boundary: This article explains a public regulatory scoring method. It is not a rating of any bank’s safety, a forecast of failure, a deposit recommendation, or investment advice.