Skip to content

GRA · AI and Financial Model Risk

Spatial financeWhere capital meets hazard, exposure and transition

Credit and pricing models

Credit scores, probability of default models and insurance pricing models decide who gets credit or cover and at what price. Machine learning can improve prediction and also introduce opacity, instability and proxy discrimination. These models also feed provisioning and capital. This page sets out how credit and pricing models fail, how their risks are measured, the data they depend on and who decides how they are used.

01Qeyd

Credit and pricing models on the record.

02Pirs

Credit and pricing models: how the risk develops.

Finance-specific automated decisions. The dossier below sets out how the risk develops, what is measured, where the data come from, how it is modelled and who decides.

Mekanîzma

Data reflect the past

Models learn from historical lending and claims that reflect past decisions, economic conditions and exclusions. Applicants unlike past borrowers are scored with less reliable evidence.

Mekanîzma

Performance degrades in new conditions

Models fitted in stable periods can misjudge default or claims when rates, inflation or unemployment shift, which is when accurate risk estimates matter most.

Mekanîzma

Proxies create unfair outcomes

Variables correlated with protected characteristics can produce disparate outcomes without using those characteristics directly. Law and supervisors constrain which factors may be used.

Mekanîzma

Errors reach capital and provisions

Probability of default and loss given default estimates feed expected credit loss provisions and capital. Systematic optimism understates both until losses arrive.

03Exposure, loss and capital

What credit and pricing models means for exposure, loss and capital.

The exposure, loss, scenario and capital views a risk officer, supervisor or ministry of finance would ask for, with the frameworks they report against. Structures follow supervisory and actuarial practice. Values are illustrative: never prices, ratings or advice.

Capital, solvency and fiscal indicators

The indicators a risk officer, supervisor or ministry of finance reads first on credit and pricing models.

NîşaneYekîneWho reads itÇarçove
CET1 ratio% of risk-weighted assetsBanks, supervisorsBasel framework
Rêjeya veşartina lîkîdîteyê% of stressed outflowsBanks, central banksBasel framework
Solvency ratio% of solvency capital requirementInsurers, supervisorsIAIS Insurance Core Principles
Combined ratio% of earned premiumSîgortekarActuarial practice
Valahiya parastinêuninsured share of economic lossInsurers, ministries of financePublished loss statistics
Public debt% ji GDPMinistries of finance, IMFDebt Sustainability Framework
Gross financing needs% of GDP per yearMinistries of finance, investorsDebt Sustainability Framework
Source: the frameworks named in each row. Indicator names, units and frameworks only; values belong to each institution’s own reporting and are never shown here.
NGFS climate scenarios · Network for Greening the Financial SystemIFRS S2 Climate-related Disclosures · ISSBTCFD recommendations · Financial Stability Board TCFDPrinciples for the effective management of climate-related financial risks · Basel Committee on Banking Supervision
04Nîşander

What is measured, and how well.

Each indicator carries its unit, its cadence, the kind of source behind it and the evidence label a value would carry. No values are shown: a programme supplies them under its own authority.

NîşaneYekîneRîtmSource typeNîşan
Discriminatory power of scorecardsGini coefficient or area under the curvesê-mehaneModel monitoringTEXMÎNKIRÎ
Calibration of probability of defaultobserved default rate against predicted, ratiosalaneBack-testingTEXMÎNKIRÎ
Population stability indexdimensionlessmehaneModel monitoringTEXMÎNKIRÎ
Approval rate differences across groupsratio of approval ratesperiodicFair lending analysisTEXMÎNKIRÎ
Overrides of model decisionsshare of decisionsmehaneTomarên biryaranÇAVDÊRÎKIRÎ
Expected credit loss model adjustmentscurrency of post-model adjustmentssê-mehaneFinancial statementsStatîstîkên weşandî
05Çavkanî

Where the evidence comes from.

Named public datasets and institutional records, each with its licence class. Sources by agreement are used only under their terms and never republished.

ÇavkanîWhat it givesLicence class
EBA guidelines on loan origination and monitoringExpectations for credit decision models including automated modelsFree, EU reuse terms
CFPB circulars on adverse action and complex algorithmsUnited States expectations for explaining credit denialsGiştî
HMDA dataUnited States mortgage application and outcome dataGiştî
Bank Pillar 3 credit risk disclosuresProbability of default back-testing and model outcomesPublic records, terms per bank
Actuarial and insurance supervisory publications on pricing modelsPractice and expectations for pricing modelsFree, terms per publisher
06Biryar

Who decides, and what the evidence serves.

AI and Financial Model Risk prepares evidence. The decisions below are taken by the authorities named, under their own rules, and never by the platform.

Kî biryarê didePirsEvidence this page serves
Credit risk and pricing committeesWhich models to use and with what overlaysPerformance, calibration and fairness evidence
Model validatorsWhether models are fit for purposeBack-testing and benchmark comparison
Conduct and fair lending supervisorsWhether model outcomes are lawful and fairOutcome data and adverse action practice
Audit committeesWhether provisioning models and overlays are reasonableModel outputs and adjustment rationale

Outputs this topic produces

  • Guide to credit and pricing model performance measures
  • Method note on testing for proxy discrimination
  • Brief on post-model adjustments and what they reveal
  • Questions for credit committees on model limits
07Core questions

What experts ask.

Credit and pricing questions combine accuracy, fairness and legality.

  • Does the model predict default or loss accurately across all customer groups?
  • Do input variables act as proxies for characteristics that law protects?
  • Can the institution explain to a customer why credit was refused or a price set?
  • How does the model perform through the economic cycle, not only in benign conditions?
  • How do alternative data sources affect accuracy, inclusion and privacy?
08Mechanisms

Mechanisms and drivers.

Credit models estimate the likelihood that a borrower will repay; pricing models estimate expected losses and costs. Both learn from historical data, which reflects past lending practices, economic conditions and social inequalities. A model can be accurate on average while performing poorly for groups underrepresented in the data, or can reproduce historical discrimination through correlated variables.

Credit models are especially exposed to economic regime change: relationships that held in an expansion may break in a downturn. Some jurisdictions classify AI used to assess the creditworthiness of individuals as high-risk, with specific obligations for providers and users.

09Methods and data

How it is studied, and where the methods stop.

Method or dataWhat it showsSînor
Discrimination and calibration testingAccuracy and calibration across groupsProtected attributes may not be recorded
Fairness metricsDisparities in outcomes or errorsDifferent metrics can conflict
Explanation methodsMain drivers of individual decisionsApproximate explanations of complex models
Through-the-cycle backtestingPerformance across conditionsFew downturns in the data
10Rêbaz

How it is studied, and what the methods cannot say.

11Pirs

Pirs û bersiv.

What does a Gini coefficient measure for a scorecard?
How well the score ranks borrowers who later default above those who do not; it measures ranking power, not whether predicted default rates are accurate.
What is a post-model adjustment?
A management overlay applied to model output, often for expected credit loss, when the model is judged not to capture current conditions.
Can a model discriminate without using protected characteristics?
Yes. Variables that correlate with protected characteristics can produce disparate outcomes, which is why fairness testing looks at outcomes as well as inputs.
Does alternative data improve inclusion?
It can help people with thin credit files, but it also raises privacy, accuracy and fairness questions that need testing.
Does the platform assess lenders’ models?
No. GRA does not assess, validate or approve any model.
rêzeya Pergalên StratejîkTevkarî bike

Ji vê rûpelê delîl bidin.

Reports weşana sê mehane ya Konsorsiyûmê ye. Tiştê ku tu li ser vê mijarê dizanî dikare di çapa pêş de bibe delîl, dosya an rastkirin, bi navê te û di bin CC BY 4.0 de weşandî li cihê ku karê te bi xwe be.

Global Risks AllianceAI and Financial Model RiskMethods and TechnicalExponential Technology
Xetera AI û modelên darayî

Bring an exposure or capital question to the Alliance.

Agahî û pirs îro vekirî ne. Pirsek referansekê distîne û gava piştî wê dide nasîn; serlêdan, drav, veqetandina hesabkirinê û tomarkirin tenê dema ku tîm, mercên wê û kapasîteya vekolînê hebin vedibin.