Credit Risk Measurement and Management
Credit risk management separates probability of default, exposure at default and loss given default, then considers concentration, dependence and mitigation. Expected loss is an average; capital and tail measures address adverse outcomes beyond that average.
The topic at a glance
| Concept | How it works | Watch for |
|---|---|---|
| Expected loss | For a simple exposure, EL = PD × LGD × EAD over a stated horizon. | Using PD alone as the expected loss rate. |
| Portfolio risk | Assess concentration and joint defaults, not only average default rates. | Assuming many loans guarantee diversification. |
| Mitigation | Evaluate recovery, collateral, guarantees and enforceable netting. | Counting collateral at its face value regardless of stress or legal access. |
01Expected loss
How it works: For a simple exposure, EL = PD × LGD × EAD over a stated horizon.
Watch for: Using PD alone as the expected loss rate.
02Portfolio risk
How it works: Assess concentration and joint defaults, not only average default rates.
Watch for: Assuming many loans guarantee diversification.
03Mitigation
How it works: Evaluate recovery, collateral, guarantees and enforceable netting.
Watch for: Counting collateral at its face value regardless of stress or legal access.
Original worked example
Expected loss on a loan
A one-year loan has EAD of 5 million, PD of 2% and LGD of 40%. Treat EAD and LGD as fixed in this simplified example.
- STEP 1
Loss conditional on default = 5,000,000 × 40% = 2,000,000.
- STEP 2
Expected loss = 2% × 2,000,000 = 40,000.
- STEP 3
A recovery rate of 60% is consistent with LGD of 40%, before costs and discounting.
Expected loss is 40,000. The loss in a default event is 2 million, not 40,000.
Illustrative assumptions and calculations for learning. Not observed market data or an official exam item.
How the ideas fit together
Before you start: Default probabilities, recovery rates and contractual exposure.
Separate the components of expected loss
EL = PD × LGD × EAD, in a simplified fixed-input model
PD refers to default over a specified horizon, LGD is the proportion of exposure lost at default and EAD is exposure when default occurs. Collateral and seniority affect recovery; undrawn commitments can affect exposure. Multiplying unconditional averages is not generally sufficient when default, recovery and exposure are dependent.
Move from an individual borrower to a portfolio
Concentration changes the shape of the portfolio loss distribution even if average expected loss is unchanged. Borrowers can default together because they share sectors, regions or funding conditions. Credit migration can also create valuation losses without default. Derivative counterparty risk adds exposure that varies with market prices, which is why notional alone is inadequate.
Dependence changes the tail
A loan book can have the same expected loss under two dependence assumptions and very different severe-loss outcomes. Common sectors, countries or collateral values create joint exposure. Distinguish default-only models from models that also include rating migration and spread changes; the latter can show losses without a default.
Work through the topic
Connect borrower analysis, exposure, recovery and portfolio dependence to credit decisions. Use the focused credit and CVA lessons for the detailed distribution and derivative calculations.
Separate default probability, exposure and loss severity
Credit loss depends on whether a counterparty fails to perform, how much is exposed when that happens and what is ultimately recovered. Probability of default describes the likelihood over a specified horizon. Exposure at default describes the amount at risk in the relevant default state. Loss given default describes the fraction not recovered, using a stated valuation and cost convention. These quantities must refer to a compatible horizon and exposure definition.
In a simplified one-year loan example, PD is 2%, exposure is 5 million and LGD is 40%. Expected loss is 0.02 × 5 million × 0.40 = 40,000 when the supplied inputs support that product. This is a mean under the model, not a prediction that the borrower will lose exactly 40,000 next year. The actual outcome could be no default or a much larger loss.
The simple product can hide dependence. A borrower may draw an unused credit line as its condition worsens, increasing exposure near default. Recoveries can also deteriorate during a broad downturn. When exposure and severity vary with default conditions, model their joint relationship or use inputs defined appropriately for those conditions. Multiplying unrelated unconditional averages may understate loss.
Move from a borrower to a portfolio
Expected losses add across positions when each expectation is defined consistently. Tail losses depend on the joint distribution. Lending to many companies does not guarantee diversification if their revenues rely on the same property market, commodity or funding source. A concentration review should therefore consider economic drivers as well as legal borrower names and industry labels.
Imagine ten equal loans, each with the same PD and LGD. Under an independence assumption, many simultaneous defaults may be rare. Under a shared severe economic factor, several borrowers can fail together. The sum of individual expected losses can remain unchanged while the portfolio’s upper-tail loss increases. This is why a mean-loss report cannot replace a concentration or capital assessment.
Migration risk adds another dimension. A borrower can weaken and a traded credit instrument can fall in value before default. A mark-to-market framework may include rating or spread transitions, while a default-only model omits them by design. Establish whether the question asks about default loss, changes in economic value or regulatory capital before choosing an approach.
| Decision | Information to bring forward | Question to challenge |
|---|---|---|
| Approve a borrower limit | Repayment capacity, PD, terms and exposure | What changes under a downturn? |
| Assess loss severity | Collateral, priority, enforceability and recovery costs | Can value be realized when needed? |
| Control concentration | Common drivers and joint default behavior | Which names fail in the same scenario? |
| Price or value the exposure | Expected loss, funding, capital and market inputs | Are physical and pricing probabilities being mixed? |
Credit decisions need more than one loss number
- Approve a borrower limit
- Information to bring forward: Repayment capacity, PD, terms and exposure
- Question to challenge: What changes under a downturn?
- Assess loss severity
- Information to bring forward: Collateral, priority, enforceability and recovery costs
- Question to challenge: Can value be realized when needed?
- Control concentration
- Information to bring forward: Common drivers and joint default behavior
- Question to challenge: Which names fail in the same scenario?
- Price or value the exposure
- Information to bring forward: Expected loss, funding, capital and market inputs
- Question to challenge: Are physical and pricing probabilities being mixed?
Evaluate mitigation and monitor the residual exposure
Collateral can reduce loss severity or current unsecured exposure, but its effectiveness depends on valuation, haircuts, access and realization. A guarantee adds reliance on another obligor. Netting reduces exposure only within an applicable, enforceable arrangement. These mitigants alter the loss mechanism; they do not justify deleting every exposure from the risk report.
For an original collateral example, a loan balance is 1 million and eligible collateral has market value 1.1 million. A 20% haircut gives recognized collateral of 880,000 under the assumed arrangement, leaving 120,000 uncovered before other adjustments. If the collateral value falls to 900,000, recognized value becomes 720,000 and the uncovered amount rises to 280,000. The headline market value was not a fixed protection amount.
Monitoring should connect early warnings to actions: review utilization, covenant compliance, borrower cash flows, collateral concentrations and changes in the repayment environment. State whether the response is a limit reduction, additional security, closer review or another permitted action. A report that records deterioration without an owner or decision trigger provides information without completing the management process.
Is expected loss the same as credit VaR?
No. Expected loss is a mean. Credit VaR is used for a selected loss quantile, or sometimes for that quantile less expected loss, depending on the convention. Always state which definition is being used before comparing capital or risk numbers.
Can collateral remove credit risk entirely?
It can reduce a specified exposure, but price changes, liquidation costs, legal uncertainty and timing gaps can remain. The collateral may also lose value in the same conditions that weaken the borrower. Review these dependencies explicitly.
Try a changed assumption
With EAD 5 million and PD 2%, what happens to EL if recovery rises from 60% to 80%?
Show answer and reasoning
LGD falls from 40% to 20%, so EL falls from 40,000 to 20,000 under the same simplified assumptions. Recovery and LGD are complements, not interchangeable inputs.
Original learning check. No email or account needed.
Is expected loss the same as credit VaR?
No. Expected loss averages the loss distribution; VaR identifies a chosen quantile. Some conventions subtract expected loss to report unexpected loss, so check the definition.
Sources and scope
Links checked 2026-10-10. GARP establishes exam scope; the technical references support the methods identified below. These are selected study notes and original examples, not complete curriculum coverage or official exam questions.
- GARP FRM study materials and 2026 learning objectives
Exam-domain scope and the official route to the annual learning objectives. This page does not certify coverage of every reading.
- MIT OpenCourseWare: counterparty credit risk
Exposure, credit valuation adjustment and counterparty-risk modelling.
- MIT: Value at Risk Models, lecture 7
Loss quantiles, estimation methods and the limitations of VaR.
- MIT OpenCourseWare: portfolio theory
Portfolio variance, covariance and diversification.
Prepared by MiloPrep editorial team. MiloPrep is independent of GARP. References do not imply endorsement or independent review of these notes.