Credit Ratings, Scoring and PD, LGD or EAD Parameters - Article

Introduction

Banks need a systematic way to distinguish a strong borrower from a weak borrower and a low-risk credit facility from a high-risk facility.

This is where credit ratings, credit scoring and quantitative credit-risk parameters become important.

The current CAIIB BFM syllabus includes credit-risk concepts such as rating methodology, credit ratings, default probabilities, risk migration and related credit-risk measurement under Module B.

The concepts can be understood through three basic questions:

• Credit rating or score: How risky is the borrower or exposure?

• PD: What is the likelihood of default?

• LGD and EAD: If default occurs, how much exposure exists and how much of it may ultimately be lost?

Together, these concepts convert qualitative credit judgement into a more structured assessment of risk.

Under the Basel Internal Ratings-Based framework, important risk components include Probability of Default (PD), Loss Given Default (LGD), Exposure at Default (EAD) and, for relevant exposures, effective maturity.

Learning Objectives

After studying this topic, you should be able to:

• explain the purpose of a credit rating;

• distinguish credit rating from credit scoring;

• differentiate internal and external credit ratings;

• interpret PD, LGD and EAD correctly;

• understand borrower rating and facility rating;

• explain how collateral mainly affects loss severity rather than borrower default probability;

• understand rating migration and model overrides; and

• avoid common CAIIB traps involving credit-risk parameters.

Understanding Credit Rating

A credit rating represents an assessment of the creditworthiness or credit risk of a borrower, counterparty, facility or debt obligation.

A rating places the assessed exposure into a risk grade or category.

For example, a bank's internal system may classify borrowers into grades such as:

Grade 1 → Grade 2 → Grade 3 → Grade 4 → High Risk → Default

The actual names and scales differ between banks.

The key principle is:

Better rating = lower assessed credit risk; weaker rating = higher assessed credit risk.

A rating is therefore not simply a description. It helps create a consistent risk language across the bank.

What Factors May Influence a Credit Rating?

Depending on the type of borrower and rating system, the assessment may consider:

Financial Factors

• profitability;

• leverage;

• liquidity;

• cash flow;

• debt-servicing capacity;

• net worth;

• working-capital position; and

• financial trends.

Business Factors

• industry prospects;

• market position;

• competition;

• customer concentration;

• supplier dependence;

• operating efficiency; and

• sensitivity to economic conditions.

Management Factors

• experience;

• competence;

• integrity;

• governance;

• succession arrangements; and

• strategic capability.

Conduct and Behaviour

The bank may also consider:

• repayment record;

• account conduct;

• overdue position;

• covenant compliance;

• previous defaults; and

• relationship history.

A sound rating system therefore combines relevant quantitative and qualitative information.

Internal Credit Rating vs External Credit Rating

Credit ratings may be broadly divided into internal and external ratings.

Under RBI's Basel III capital framework, eligible external credit ratings remain relevant for regulatory risk weighting of specified exposures. RBI's June 9, 2025 circular also updated the treatment of an accredited domestic External Credit Assessment Institution while leaving the other external-rating provisions unchanged.

An external rating does not replace a bank's own credit appraisal. The bank remains responsible for deciding whether the exposure is acceptable.

Borrower Rating and Facility Rating

This distinction is especially important.

Borrower Rating

A borrower rating primarily answers:

How likely is this borrower to default?

The Basel framework defines a borrower grade as an assessment of borrower risk based on specified rating criteria, from which PD estimates are derived.

Therefore, borrower rating is closely linked with default risk and PD.

Facility Rating

A facility rating considers characteristics of a particular credit facility.

Examples include:

• collateral;

• seniority;

• guarantee;

• facility structure;

• recovery prospects; and

• other factors affecting the loss if default occurs.

The same borrower may have two facilities with different recovery prospects.

For example:

• Facility A may be strongly secured and senior.

• Facility B may be unsecured and subordinated.

The borrower's likelihood of default may be broadly the same for both facilities, but the loss after default can be different.

This leads to an important relationship:

Borrower risk → mainly connected with PD

Facility characteristics → strongly influence LGD

What Is Credit Scoring?

Credit scoring uses specified borrower and transaction characteristics to generate a numerical score or model-based risk assessment.

For example, a scoring system may consider information relating to:

• income;

• occupation;

• repayment history;

• existing debt;

• utilisation;

• age of credit relationships;

• collateral or loan-to-value;

• delinquency history; and

• other relevant variables.

The exact variables depend on the model and type of credit.

Credit scoring is especially useful where a bank handles a large number of relatively standardised exposures, such as:

• personal loans;

• credit cards;

• vehicle loans;

• housing loans; and

• other retail facilities.

Basel permits credit-scoring models and other statistical or mechanical methods to support rating assignments and estimation of credit-risk characteristics. However, it also requires adequate human oversight because a model may not capture every material fact.

Credit Rating vs Credit Score

The distinction is not absolute. A numerical score may be converted into a rating grade.

For example:

Score range → Risk band → Internal rating

Therefore, scoring and rating may form different stages of the same risk-classification process.

Credit Score of a CIC Is Not the Same as the Bank's Internal Score

A common practical confusion is between:

• a credit score generated by a Credit Information Company (CIC); and

• a bank's own internal credit score or rating.

These are not identical.

RBI has clarified that it does not prescribe the methodology used by CICs to calculate their credit scores; the scoring models used by individual CICs are proprietary.

A bank may use credit-bureau information as one input but should apply its own credit policy and underwriting requirements.

Probability of Default — PD

Probability of Default (PD) represents the likelihood that a borrower will default over a specified period.

Under the current Basel IRB framework, PD for corporate, sovereign and bank exposures is generally expressed as the one-year probability of default associated with the relevant internal borrower grade. A borrower already classified in a default grade has a PD of 100% for this purpose.

Simple Interpretation

If a particular rating grade has a higher PD than another grade, borrowers in that grade are considered more likely to default.

Therefore:

Better credit quality → Lower PD

Worsening credit quality → Higher PD

PD primarily represents default frequency or likelihood, not the amount of money that will be lost.

PD Is Not the Same as Historical Default Rate

This is an important conceptual distinction.

A default rate is an observed historical outcome for a group of borrowers.

PD is an estimate of future default likelihood.

Basel specifically distinguishes the realised default rate as an ex-post measure from PD as a risk parameter.

Loss Given Default — LGD

Loss Given Default (LGD) represents the proportion of an exposure that the lender expects to lose if default actually occurs.

LGD therefore answers:

If the borrower defaults, how severe could the loss be?

LGD is influenced by factors such as:

• collateral;

• guarantees;

• seniority of claim;

• recovery costs;

• time taken for recovery;

• enforceability;

• economic conditions; and

• recovery value.

Higher Recovery Means Lower LGD

If a bank can recover a large part of its exposure after default, its LGD is lower.

If recovery is poor, LGD is higher.

Therefore:

Higher recovery rate → Lower LGD

Lower recovery rate → Higher LGD

Why Security Mainly Affects LGD

Suppose the same borrower has:

• one fully secured facility; and

• one unsecured facility.

The borrower's basic probability of default may be similar for both.

But if default occurs, the secured facility may experience a smaller loss because of recoverable collateral.

Thus, security often affects LGD more directly than PD.

Exposure at Default — EAD

Exposure at Default (EAD) represents the amount the bank expects to be exposed to when the borrower defaults.

Basel defines EAD for an on- or off-balance-sheet item as the expected gross exposure of the facility upon default. For an on-balance-sheet exposure, the estimate cannot generally be below the current drawn amount, subject to specified treatment.

EAD is particularly important for facilities where the amount outstanding can change before default.

Examples include:

• cash credit;

• overdraft;

• revolving credit;

• credit cards;

• undrawn loan commitments; and

• other facilities where additional utilisation may occur.

Why Current Outstanding and EAD May Differ

Suppose a borrower has not yet fully utilised a sanctioned revolving facility.

If the borrower's financial position deteriorates, it may draw more funds before eventually defaulting.

Therefore:

Current outstanding ≠ necessarily EAD

EAD attempts to estimate the exposure at the point of default, not merely today's utilisation.

PD, LGD and EAD Together

These parameters describe three different dimensions of credit loss.

A bank may therefore have:

• high PD but low LGD — weak borrower but strong recovery protection;

• low PD but high LGD — strong borrower but largely unsecured facility;

• low current exposure but high EAD potential — large undrawn revolving commitment.

These combinations explain why no single parameter gives the complete credit-risk picture.

Expected Loss Relationship

Expected Loss = PD × LGD × EAD

This formula expresses the average credit loss associated with default likelihood, loss severity and exposure at default.

Refer to the Formula Section for a detailed explanation, variables, and solved examples.

Rating Migration

Credit quality is not static.

A borrower's rating may:

• improve;

• remain unchanged;

• deteriorate; or

• move into default.

For example:

Grade A → Grade B → Grade C → Default

A movement from a better rating to a weaker rating is generally called a rating downgrade or adverse migration.

A movement to a better grade is an upgrade.

Changes in rating may affect:

• PD;

• pricing;

• monitoring frequency;

• credit limits;

• approval authority;

• provisioning or expected-loss assessment where relevant;

• risk-weighted assets under applicable frameworks; and

• portfolio risk.

Transition Matrix

A transition matrix shows how borrowers migrate between ratings over a period.

Conceptually, it helps answer questions such as:

• What proportion remained in the same grade?

• How many were upgraded?

• How many were downgraded?

• How many moved to default?

It is useful for:

• analysing rating stability;

• estimating default behaviour;

• validating rating systems; and

• portfolio monitoring.

A transition matrix describes movement between risk grades. It should not be confused with PD itself.

Retail Pooling and Credit-Risk Parameters

Under the Basel IRB framework, retail exposures are generally grouped into pools containing sufficiently homogeneous exposures.

Relevant factors may include:

• borrower characteristics;

• product characteristics;

• collateral;

• loan-to-value;

• guarantees;

• seniority; and

• delinquency status.

For each relevant retail pool, banks using the framework need quantitative loss characteristics such as PD, LGD and EAD.

This reflects an important distinction:

Corporate approach: often focuses heavily on individual borrower grades.

Retail approach: often uses segmentation, scoring and homogeneous pools.

Rating Systems Need Validation

A model is not reliable merely because it generates a number.

A sound rating or scoring system requires:

• suitable data;

• clear rating criteria;

• consistent application;

• independent validation;

• periodic review;

• monitoring of actual outcomes;

• model-performance testing; and

• controls over overrides.

Basel requires rating definitions to be clear enough to produce meaningful risk differentiation and consistent assignment of similar borrowers or facilities.

Model Risk

Credit models can produce wrong results because of:

• poor-quality data;

• outdated relationships;

• missing risk factors;

• inappropriate assumptions;

• biased sample data;

• structural economic change; or

• incorrect use of the model.

Therefore, the model's output should not be accepted blindly.

Basel notes that credit-scoring models use only part of the information available and therefore requires sufficient human judgement and oversight to consider relevant information outside the model.

Rating Overrides

Sometimes authorised credit officials may have material information not adequately captured by the model.

A rating override may then be justified.

However, an override should be:

• permitted by policy;

• properly authorised;

• supported by reasons;

• documented;

• monitored; and

• reviewed for performance.

Basel requires banks to clearly define when overrides are permitted and to monitor cases where human judgement changes model-generated results.

Practical Banking Application

Consider two borrowers.

Borrower A

• financially strong;

• stable business;

• low internal PD;

• unsecured working-capital facility.

Borrower B

• financially weaker;

• higher PD;

• facility supported by strong and enforceable collateral.

Borrower A may have a lower probability of default, but if it defaults, the unsecured structure could produce relatively high loss severity.

Borrower B may have a higher probability of default, but effective collateral may reduce the loss after default.

This illustrates why:

PD does not equal LGD.

Now assume Borrower B also has a large unused revolving limit.

If financial stress occurs and the borrower draws this limit before default, EAD can increase.

Thus, a complete risk assessment must consider:

Default likelihood + Recovery severity + Exposure at default

rather than relying only on the credit rating.

CAIIB Exam Focus

Remember these important points:

• Credit rating places a borrower or exposure into a risk grade.

• Credit scoring commonly produces a numerical risk score.

• A score can be converted into a rating grade.

• External rating and internal bank rating are different.

• A bureau credit score is not the same as a bank's internal rating.

• Borrower rating primarily reflects borrower/default risk.

• Facility rating captures characteristics that influence transaction loss.

• PD represents probability of default, not expected percentage loss.

• PD is commonly associated with the borrower grade.

• LGD measures the proportion lost if default occurs.

• Recovery and collateral strongly influence LGD.

• EAD represents the expected exposure at the time of default.

• EAD may exceed current outstanding for revolving or undrawn facilities.

• PD, LGD and EAD are separate parameters.

• Rating migration means movement between risk grades.

• A downgrade usually indicates deterioration in credit quality.

• Transition matrices analyse movements among rating grades.

• Credit-scoring models require validation and human oversight.

• Model overrides should be controlled and documented.

• An external rating does not eliminate the bank's responsibility for independent credit appraisal.

Common Exam Traps

Trap 1: “PD indicates the amount the bank will lose if default occurs.”

Incorrect. PD indicates the likelihood of default. LGD relates to loss severity after default.

Trap 2: “LGD is the same as probability of default.”

Incorrect. A borrower may have a low PD but a high LGD if the exposure has poor recovery protection.

Trap 3: “EAD is always equal to today's outstanding balance.”

Incorrect. For facilities with undrawn or revolving components, exposure may increase before default.

Trap 4: “Strong collateral necessarily reduces the borrower's PD.”

Incorrect. Collateral primarily affects potential recovery and therefore LGD. It does not automatically make the borrower financially stronger.

Trap 5: “A high credit score guarantees sanction.”

Incorrect. Credit scores are inputs into underwriting. Income, repayment capacity, policy, exposure and other risks remain relevant.

Trap 6: “External and internal credit ratings are identical.”

Incorrect. They may use different information, methodology, purpose and rating scales.

Trap 7: “A borrower rating and facility rating must always be identical.”

Incorrect. Borrower rating reflects default risk, while facility characteristics may produce different loss severity.

Trap 8: “A downgrade means that default has already occurred.”

Incorrect. A downgrade normally indicates increased credit risk. Default is a separate and more severe credit state.

Trap 9: “A mathematical credit-scoring model does not require human judgement.”

Incorrect. Models can omit material information or become unreliable. Appropriate oversight, validation and judgement remain necessary.

Trap 10: “An override means the rating model has failed.”

Not necessarily. A controlled override may incorporate material information not captured by the model. The danger lies in uncontrolled or unjustified overrides.

Memory Technique

Remember P–L–E:

P — PD = Probability: Will default happen?L — LGD = Loss: How much may be lost after default?E — EAD = Exposure: How much will be outstanding at default?

For rating and scoring, remember:

Score = Number

Rating = Grade

PD = Default likelihood

LGD = Loss severity

EAD = Exposure size

These distinctions solve many statement-based CAIIB questions.