← Module C

CLERK2OFFICER · CAIIB / ABM / C07

Risk Management and Credit Rating

2,938 teaching words · 25 explained MCQs · Applied learning exercise

Start chapter practice ↓

Original study edition R1 · 4 November 2026. Not affiliated with IIBF. Examples and calculations are illustrative. Regulatory references are version-labelled. Works offline except reference links. Independent academic sign-off is not claimed.

1. Credit risk is uncertainty about promised payment

Credit risk is the possibility of loss because a counterparty does not meet its obligations as agreed, or because deterioration in its credit quality reduces the value of the exposure. Default is an important event, but risk exists before default occurs. A borrower can become more vulnerable while continuing to pay on time. The lender must therefore consider both the probability of failure and the financial consequence if failure occurs.

Risk management does not mean eliminating every uncertain loan. It means identifying, measuring, controlling, monitoring and reporting risk within an approved appetite. The process operates at borrower, facility and portfolio levels. A strong borrower can have a poorly structured facility; a well-secured facility can belong to a concentrated portfolio; and an attractive portfolio average can conceal a few large vulnerabilities. This chapter links qualitative judgement with probability, loss severity, ratings and risk mitigation. The numerical examples are deliberately simplified and do not constitute regulatory capital calculations or production credit models. Their purpose is to help students distinguish the components of risk and understand why different controls address different causes of loss. A useful credit decision explains those causes rather than treating a single rating symbol as a complete answer.

2. Identify business, financial and external drivers

Business risk arises from demand, competition, technology, input availability, operating capability and customer concentration. Financial risk includes leverage, liquidity, cash-flow volatility, currency mismatch and repayment structure. Management and governance affect the quality of decisions and information. External conditions such as interest rates, regulation, weather or supply-chain disruption can influence several borrowers simultaneously. These drivers interact rather than operate in separate boxes.

For example, a manufacturer with high debt and one major customer is exposed to both financial leverage and customer concentration. If that customer delays payment, working-capital borrowing rises and interest costs may weaken margins further. A small initial shock can therefore amplify through the financial structure. The analyst should map the causal chain and identify what evidence would show deterioration. Do not assume that every risk factor carries the same weight for every borrower. Export earnings may reduce one currency mismatch while creating dependence on overseas demand. Collateral may reduce loss severity but leave the probability of operating failure largely unchanged. Good identification distinguishes the risk driver, the exposure mechanism and the proposed response. This makes later measurement and monitoring more meaningful than a generic list of risks copied into every appraisal.

3. Probability, severity and exposure are different

Probability of default, or PD, estimates the likelihood of default over a specified horizon under a defined default concept. Loss given default, or LGD, is the proportion of exposure lost after recoveries and relevant costs under the model's convention. Exposure at default, or EAD, estimates the amount outstanding when default occurs, which may differ from today's balance if undrawn commitments can be used. These quantities must be defined consistently.

A simplified expected-loss calculation is PD multiplied by LGD multiplied by EAD. For a one-year PD of 2%, LGD of 40% and EAD of 100 lakh, expected loss is 0.8 lakh. This is an average over the model's possible outcomes, not a prediction that the bank will lose exactly 0.8 lakh on the individual loan. The actual result may be no loss or a much larger loss. If collateral improves recoveries, LGD may fall without changing PD. If a revolving line is likely to be drawn before default, EAD may exceed the current balance. The calculator lets students vary each component independently. It deliberately avoids labelling the result a provision, regulatory capital requirement or accounting impairment amount, because those measures involve additional rules, assumptions and definitions.

Expected loss has three inputsPD 2% × LGD 40% × EAD 100 = expected loss 0.8. Horizon and definitions must match; this is not a provision or capital calculation.Expected loss has three inputsPD2%×LGD40%×EAD100Expected loss = 0.8
PD 2% × LGD 40% × EAD 100 = expected loss 0.8. Horizon and definitions must match; this is not a provision or capital calculation.

4. Expected and unexpected loss serve different purposes

Expected loss describes the mean loss under the chosen model and horizon. Unexpected loss concerns variation around that mean and the possibility of worse outcomes. Pricing and loss planning may consider expected loss, while capital and risk appetite address the ability to absorb adverse outcomes. The exact regulatory and accounting treatments depend on applicable frameworks and should not be inferred from the simplified teaching model.

Imagine one hundred similar loans with low individual default probability. Average losses may be manageable, but a common economic shock could cause many defaults together. The average expected loss does not describe the full tail of that distribution. Correlation, concentration and loss severity influence the adverse outcome. Nor does a low historical loss rate prove that future risk is low: the observation period may exclude a downturn, underwriting may have changed or exposures may not yet have matured. Students should distinguish estimation uncertainty from the underlying variability of losses. A model can be uncertain about PD and still produce a precise-looking number. Responsible use reports assumptions, data limitations and sensitivity. The analytical goal is to support a decision under uncertainty, not to create false confidence through mathematical notation.

5. Ratings summarise a judgement within a scale

A credit rating places an obligor or instrument in a category representing an assessment of credit quality under a defined methodology. Internal ratings support a bank's own decisions; external ratings are issued by rating agencies under their frameworks. The scale, horizon and object of assessment matter. An issuer rating and a particular instrument rating may differ because security, seniority or support affects the instrument's risk. A symbol cannot be interpreted without its definition.

Ratings are opinions or model-supported assessments, not guarantees of repayment. They can change as information or conditions change. A favourable rating should not replace the lender's own appraisal, especially where the facility differs from the rated instrument. Compare ratings only with care: different agencies or internal systems may use different definitions, and a mapping should be validated rather than assumed. Outlooks and watch indicators provide additional information but are not identical to a current rating category. In a case question, ask what is rated, by whom, using which evidence and as of what date. A stale rating can be misleading even if it was well founded when issued. Good credit management uses ratings as part of an evidence system, not as a permission slip to ignore contrary facts.

6. Build an internal rating from relevant evidence

An internal rating methodology may combine financial ratios, business factors, management assessment, account conduct and other information. Some systems use statistical models; others use structured judgement or a combination. Weights and overrides should reflect validated policy rather than individual preference. The methodology should be appropriate for the borrower segment. A large corporate model may be unsuitable for a small seasonal enterprise with different records and risk drivers.

Input quality is central. Financial ratios based on unreliable statements can produce an apparently objective but misleading grade. Qualitative factors also require evidence: 'strong management' should connect to experience, controls and observed behaviour rather than personal confidence. Document the basis for material judgements and any override. An override can improve a model's result when important information is missing from its design, but uncontrolled overrides can undermine consistency. Monitor whether overrides systematically improve or worsen subsequent outcomes. The final grade should be approved and refreshed according to the bank's framework, including material events between routine reviews. Students should understand the logic of combining evidence without assuming that one generic scorecard or set of weights is universally prescribed. The chapter examples illustrate reasoning, not a bank-approved rating engine.

7. Validate models before relying on their outputs

Validation examines whether a model is conceptually sound, implemented correctly and performing adequately for its intended use. It includes review of data, assumptions, estimation, calibration, discrimination and ongoing outcomes. Discrimination concerns the ability to distinguish relatively riskier from safer cases; calibration concerns whether predicted risk levels match observed outcomes under the relevant conditions. A model can rank borrowers reasonably well while systematically understating default probability.

Back-testing compares predictions with later outcomes, but results require interpretation. A small sample may produce unstable default rates, and an unusual economic period may differ from development data. Monitor changes in borrower mix, products, data definitions and approval practices. If a model was trained only on previously approved borrowers, applying it to a broader applicant population can create limitations. Independent challenge should examine these issues and the consequences for decisions. Technology performance also matters: a correct model implemented with the wrong variable units can fail operationally. Keep version records and test changes before deployment. A model is not validated merely because its software runs without error or its historical accuracy looks high. The evidence must support the specific decision for which it is used.

8. Rating migration helps monitor change

Rating migration describes movement between credit grades over time. A migration matrix can show the proportions of borrowers remaining in a grade, improving, deteriorating or defaulting over a defined period. It helps assess portfolio quality and model behaviour, provided grades, periods and populations are consistent. Migration is not necessarily proof of model failure: a useful system should respond to genuine changes in risk.

Suppose a bank sees an increase in downgrades among businesses linked to one commodity. The next step is to investigate the shared driver, exposure size and likely cash effect, not merely report the number of downgrades. A stable grade can also be misleading if reviews are overdue. Track rating age and missing updates alongside migration. Compare borrower counts and exposure-weighted measures, because a few large downgrades may matter more than many small upgrades. A transition estimate based on limited history should not be treated as a timeless probability law. Economic conditions and underwriting standards can change. For examination purposes, understand what each row and column represent and ensure that a row's outcomes reconcile to the stated population. Interpretation should connect the matrix to risk decisions rather than stop at arithmetic percentages.

9. Mitigation should address a specific risk mechanism

Credit-risk mitigation can include better selection, appropriate structure, covenants, collateral, guarantees, diversification, insurance and risk transfer. Each tool has limits and may introduce new risks. Collateral can improve recoveries but requires enforceability and realistic valuation. A guarantee adds a counterparty whose capacity may be correlated with the borrower. Covenants support early intervention but require monitoring and usable remedies. Diversification reduces concentration but cannot eliminate economy-wide shocks.

Match the response to the problem. If the issue is a seasonal cash mismatch, a suitable repayment schedule may help more than additional property. If the issue is unreliable information, independent verification is needed; a higher interest rate does not repair the data. If the issue is customer concentration, monitoring and exposure limits may be relevant, while a guarantee from another company dependent on the same customer may offer weak protection. Record residual risk after mitigation. The phrase 'risk covered' is often too absolute. Ask what happens if the protection fails, is delayed or cannot be enforced. Effective mitigation changes the loss mechanism in a demonstrable way and fits the bank's legal, operational and regulatory framework. It does not merely add reassuring language to an otherwise unsupported proposal.

10. Concentration and correlation shape portfolio loss

Concentration is excessive dependence on a borrower, group, sector, geography, product or other common factor. Correlation describes how outcomes move together. The two concepts interact: a portfolio of many names can remain highly exposed to one shock if those borrowers share a key risk driver. Legal separation does not establish economic independence. Group links and supply-chain relationships deserve attention alongside formal borrower limits.

Consider five exporters serving the same overseas distributor. Their accounts and products may differ, but a payment failure at the distributor can affect all five. Adding a sixth exporter with the same dependency may increase nominal diversification while worsening actual concentration. Portfolio controls should therefore use meaningful groupings and reliable exposure data. Limits must be linked to authority and action when approached or breached. Stress scenarios can test the effect of a common adverse event, including changes in recovery values. Correlation can rise in a downturn when businesses fail together and collateral markets weaken. A model that assumes independence may understate tail risk. Students should be able to explain this qualitatively even without an advanced portfolio model. The key is to ask what common event could damage several apparently separate exposures at once.

11. Credit derivatives transfer defined risk

A credit derivative transfers specified credit risk under a contract without necessarily transferring the underlying loan itself. In a simple credit default swap description, one party pays a premium and receives protection contingent on defined credit events and settlement terms. The exact exposure, reference obligation, maturity and legal definitions are essential. A protection contract that does not match the loan can leave basis risk.

Risk transfer does not mean risk elimination. The protection seller may fail, documentation may be disputed, the contract may mature before the exposure, or the defined event may not match the bank's loss. Counterparty, legal, liquidity and operational risks therefore accompany the transferred credit risk. Actual use must comply with the relevant RBI framework, eligibility and the bank's authority; the conceptual example is not a recommendation to enter a derivative. For learning, compare selling a loan with buying protection: the first changes ownership of the asset under its terms, while the second can leave the asset on the bank's books with a separate contractual hedge. Capital recognition is another question governed by applicable rules. A student should describe what risk moves, to whom, and what residual risks remain rather than simply state that derivatives make lending safe.

12. Credit information improves the evidence base

Credit information systems help lenders understand borrowing, repayment conduct and other relevant information beyond a single account. Credit information companies and regulatory reporting arrangements serve different purposes and should not be treated as one identical database. Information must be accessed, used and reported through authorised processes. Accuracy, identity matching, timely updates and correction of errors are important for both lender decisions and fair customer treatment.

A report showing an overdue amount may indicate genuine stress, an unresolved dispute, stale reporting or a mismatch. The bank should investigate material inconsistencies rather than ignore the report or treat it as infallible. A clean report is also not proof of adequate future cash flow; it primarily reflects the information and history captured. The RBI Commercial Banks Credit Information Reporting Directions, 2025 are included in the reference register, with the observed version identified. Operational reporting intervals and obligations should be checked against the applicable current text rather than memorised from an old circular. In appraisal, combine external information with the borrower's disclosures, financial analysis and account conduct. The best use of credit information is to reveal questions and improve evidence, not to replace the complete lending assessment with a single score.

13. Governance links measurement with decisions

A credit-risk framework needs an approved appetite, policies, delegated authority, independent challenge, reliable information and escalation. Measurement is useful only when it influences decisions. A rating may affect pricing, review frequency, limits or approval level under the bank's policy, but those links must be explicit. Unauthorised exceptions can undermine the framework even when the model itself is sound. Record who approved departures and why.

RBI requirements and the bank's internal standards must be distinguished from illustrative teaching methods. A PD-times-LGD calculation is not automatically the required provision, and an internal score is not automatically an external rating. Model changes should be controlled, tested and monitored. Portfolio reports should reach the appropriate decision-makers with clear explanations of deterioration and uncertainty. Risk and business teams should be able to challenge assumptions without suppressing legitimate lending opportunities or adverse evidence. The quality of governance becomes especially important when conditions change rapidly. A framework that worked in a stable period may need revised assumptions and closer monitoring during stress. Students should connect the organisational process with the numerical measures: risk management is a continuing set of decisions and controls, not a spreadsheet produced once at sanction.

14. Original expected-loss case

Bank Alpha compares two invented one-year exposures, each with EAD of 200 lakh. Borrower X has PD of 1.5% and LGD of 60%, giving expected loss of 1.8 lakh. Borrower Y has PD of 3% and LGD of 25%, giving expected loss of 1.5 lakh. Y has the higher estimated probability of default but the lower expected loss because assumed recovery is better. This does not prove that Y is the preferable loan: uncertainty, concentration, income, capital and operational costs also matter.

Now suppose the collateral supporting Y is closely tied to the same industry downturn that could cause default. Its LGD estimate may be too optimistic under stress. Raising LGD to 50% increases expected loss to 3 lakh. The example demonstrates why inputs should be challenged separately and jointly. It also shows that a rating focused on default likelihood does not fully describe facility loss severity. The interactive calculator reports the expected amount and loss rate under the user's assumptions, with a reminder that realised losses can differ greatly. In a written answer, explain the assumptions and the reason for the result. Do not rank borrowers solely by collateral value or a single expected-loss number without considering the wider credit decision.

Higher default chance is not the whole storyX: 1.5% × 60% × 200 = 1.8. Y: 3% × 25% × 200 = 1.5. If Y LGD rises to 50%, expected loss becomes 3.Higher default chance is not the whole storyExpected loss in lakh1.8X1.5Y base3Y stress
X: 1.5% × 60% × 200 = 1.8. Y: 3% × 25% × 200 = 1.5. If Y LGD rises to 50%, expected loss becomes 3.

15. Revision and common traps

For a risk question, define the exposure and horizon, identify the driver of failure, distinguish probability from severity and quantify only where inputs are justified. Then examine mitigation, residual risk and portfolio dependence. If a rating is supplied, identify whether it applies to the borrower or instrument and whether it is current. If a model result is supplied, ask about data, calibration and intended use. This sequence makes the answer analytical rather than descriptive.

Common mistakes include reading 2% as 2 in a probability formula, treating expected loss as the most likely individual outcome, assuming collateral lowers PD automatically, and equating many borrower names with diversification. Other traps are treating an external rating as a guarantee and assuming a hedge removes counterparty or legal risk. In model questions, high overall accuracy can be misleading when defaults are rare; a model predicting no defaults may appear accurate while identifying none of the risky cases. Focus on the decision and the consequences of errors. Credit risk cannot be reduced to one number, but clear definitions and disciplined arithmetic make judgement more reliable. The purpose of the chapter is to combine that arithmetic with evidence, governance and a realistic understanding of uncertainty.

Expected credit loss illustration

One defined horizon, user-supplied assumptions. This is not an accounting provision or regulatory capital engine.

Change inputs, then calculate.

Practice and review

One answer per question. No negative marking. You can read every explanation even without JavaScript.

Optional browser-local answer saving, with no account synchronisation.

1. Credit risk exists at what stage?
Answer and all option explanations

Correct answer: C. Deterioration can increase risk while payments remain current.

  • A. Risk exists much earlier.
  • B. Secured exposures also carry risk.
  • C. Deterioration can increase risk while payments remain current.
  • D. The bank is exposed during the loan.
2. Which combination can amplify a collection shock?
Answer and all option explanations

Correct answer: D. Delayed cash can increase borrowing and financing pressure.

  • A. The stated customer dependence contradicts independence.
  • B. Branding does not create the financial mechanism.
  • C. Those are not the stated conditions.
  • D. Delayed cash can increase borrowing and financing pressure.
3. PD represents what?
Answer and all option explanations

Correct answer: B. The definition and time period must be specified.

  • A. A probability is not a certain outcome.
  • B. The definition and time period must be specified.
  • C. That relates to recovery and LGD.
  • D. That is an exposure amount.
4. LGD represents what under a stated model?
Answer and all option explanations

Correct answer: D. Its convention should include the defined costs and timing.

  • A. That is PD.
  • B. That is not a loss proportion.
  • C. Pricing and severity are different quantities.
  • D. Its convention should include the defined costs and timing.
5. Why can EAD exceed today's drawn balance?
Answer and all option explanations

Correct answer: C. Future utilisation can increase exposure.

  • A. That is not guaranteed.
  • B. Probability is not a currency amount.
  • C. Future utilisation can increase exposure.
  • D. Security does not itself increase borrowing.
6. PD 2%, LGD 40%, EAD 100 lakh give expected loss of?
Answer and all option explanations

Correct answer: A. 0.02 multiplied by 0.40 multiplied by 100 equals 0.8.

  • A. 0.02 multiplied by 0.40 multiplied by 100 equals 0.8.
  • B. That assumes default is certain.
  • C. That omits LGD.
  • D. That treats percentages incorrectly.
7. What does expected loss on one loan mean?
Answer and all option explanations

Correct answer: D. The actual outcome may be zero or much larger.

  • A. Capital requires a separate applicable framework.
  • B. Expected value is not certainty.
  • C. An average need not be the modal outcome.
  • D. The actual outcome may be zero or much larger.
8. Collateral improves recoveries without changing operations. Which input may primarily fall?
Answer and all option explanations

Correct answer: B. Better recovery can reduce severity if legally and economically effective.

  • A. Exposure is not reduced below zero by collateral.
  • B. Better recovery can reduce severity if legally and economically effective.
  • C. Operating default risk may remain.
  • D. No such pricing consequence follows automatically.
9. Unexpected loss concerns what?
Answer and all option explanations

Correct answer: D. The mean does not describe the entire loss distribution.

  • A. Credit outcomes themselves vary.
  • B. It concerns risk of worse outcomes.
  • C. Neither measure is a guaranteed individual loss.
  • D. The mean does not describe the entire loss distribution.
10. Why can low historical losses mislead?
Answer and all option explanations

Correct answer: A. Observed experience may not represent future conditions.

  • A. Observed experience may not represent future conditions.
  • B. Estimation requires evidence.
  • C. It is useful with context and limitations.
  • D. Finite experience does not establish impossibility.
11. An issuer rating and instrument rating may differ because what?
Answer and all option explanations

Correct answer: D. The object of assessment matters.

  • A. Methodology and rated obligation differ.
  • B. Substantive credit features can differ.
  • C. They are assessments subject to change.
  • D. The object of assessment matters.
12. What is unsafe when using an external rating?
Answer and all option explanations

Correct answer: C. Facility-specific and current evidence still matter.

  • A. Definitions support correct interpretation.
  • B. Currency is relevant.
  • C. Facility-specific and current evidence still matter.
  • D. New evidence can challenge a stale rating.
13. Internal rating qualitative factors should be supported by what?
Answer and all option explanations

Correct answer: B. Labels such as strong management need a basis.

  • A. That cannot be reviewed or reproduced.
  • B. Labels such as strong management need a basis.
  • C. Familiarity can introduce bias.
  • D. Segment and validation requirements differ.
14. What is a sound approach to model overrides?
Answer and all option explanations

Correct answer: B. Overrides can help but also undermine consistency.

  • A. The audit trail should preserve the decision history.
  • B. Overrides can help but also undermine consistency.
  • C. Relevant information may require controlled review.
  • D. That removes governance and traceability.
15. Calibration asks whether what is accurate?
Answer and all option explanations

Correct answer: A. It concerns probability levels, not only ranking.

  • A. It concerns probability levels, not only ranking.
  • B. That is unrelated to risk estimation.
  • C. Technical operation is not sufficient validation.
  • D. Interface size does not establish predictive quality.
16. Discrimination concerns what?
Answer and all option explanations

Correct answer: D. Ranking ability differs from calibration of probabilities.

  • A. No ranking model guarantees outcomes.
  • B. That is not the statistical meaning here.
  • C. That may fail to identify risk.
  • D. Ranking ability differs from calibration of probabilities.
17. Rating migration should be reviewed with what additional information?
Answer and all option explanations

Correct answer: B. Counts alone can conceal material deterioration.

  • A. Visual design does not explain risk.
  • B. Counts alone can conceal material deterioration.
  • C. Stale reviews can hide change.
  • D. Conditions affect transitions.
18. Which mitigation best addresses unreliable financial data?
Answer and all option explanations

Correct answer: D. A higher rate or more labels does not repair evidence.

  • A. The information risk remains.
  • B. That may add another uncertain source.
  • C. Pricing cannot validate inputs.
  • D. A higher rate or more labels does not repair evidence.
19. Many borrowers depend on one distributor. What remains?
Answer and all option explanations

Correct answer: D. Different legal names can share one failure source.

  • A. Economic dependence remains.
  • B. Exposure controls may be relevant.
  • C. The distributor affects cash generation directly.
  • D. Different legal names can share one failure source.
20. Why can correlation worsen during a downturn?
Answer and all option explanations

Correct answer: B. Defaults and collateral values may move adversely at once.

  • A. Distress can reduce recovery markets.
  • B. Defaults and collateral values may move adversely at once.
  • C. That is the opposite mechanism.
  • D. They remain distinct components.
21. A credit derivative primarily does what?
Answer and all option explanations

Correct answer: D. Its scope, counterparty and settlement conditions matter.

  • A. Residual and new risks remain.
  • B. Recognition depends on applicable rules.
  • C. Protection and asset sale differ.
  • D. Its scope, counterparty and settlement conditions matter.
22. Protection matures before the loan. What risk is highlighted?
Answer and all option explanations

Correct answer: A. The later loan period may be unprotected.

  • A. The later loan period may be unprotected.
  • B. Terms define the gap.
  • C. The hedge does not cancel borrower debt.
  • D. Coverage ends before the exposure.
23. A credit report contains an apparent overdue mismatch. Best response?
Answer and all option explanations

Correct answer: B. The report is important but not infallible.

  • A. A mismatch may reflect error or stale data.
  • B. The report is important but not infallible.
  • C. That discards useful evidence.
  • D. Records should be reconciled, not concealed.
24. X has PD 1.5%, LGD 60%, EAD 200; Y has 3%,25%,200. Which expected losses follow?
Answer and all option explanations

Correct answer: D. The products show higher PD need not mean higher expected loss.

  • A. A 25% LGD still represents loss severity.
  • B. That assumes full certain loss.
  • C. Those figures do not apply the stated products.
  • D. The products show higher PD need not mean higher expected loss.
25. Why is always predicting no default a weak rare-event model?
Answer and all option explanations

Correct answer: C. Class imbalance makes accuracy alone insufficient.

  • A. Its decision usefulness must be tested.
  • B. It identifies no actual defaults.
  • C. Class imbalance makes accuracy alone insufficient.
  • D. A prediction is not evidence of certainty.

Ready: 25 questions.

References and conventions

Scope and cited reference pages checked 4 November 2026. Original exposition, branch cases, diagrams and question wording. Apply the cited source scope, subsequent changes and IIBF examination cutoff. Observed source versions are recorded in the reference register. Calculators are educational, not live credit or legal decision engines.

Next chapter →