RISK5001 Chap.7 Market and Credit Risk Measurement
Market and Credit Risk Measurement
This chapter interprets Value at Risk as a modelled quantile, uses exceedances for back-testing, separates default probability, exposure and loss severity, and links expected loss with pricing, reserves and economic capital. Every measure has a horizon, unit, confidence level and model boundary that must travel with the number. Validation asks whether realised outcomes behave consistently with those assumptions.
Credit measurement separates the chance of default from the amount exposed and the proportion lost, allowing expected cost to be distinguished from the unexpected variation that capital and limits address.
What this chapter covers
- 01
Market risk measurement
- 02
Value at Risk
- 03
Confidence and horizon
- 04
Back-testing and stress
- 05
PD, LGD and EAD
- 06
Expected and unexpected loss
Worked example · free
Expected loss from credit inputs
- 1LGD = 1 - 25% = 75%.
- 1Expected loss = 0.012 x 0.75 x $9,000,000.
- 1Expected loss is $81,000 and is not a maximum loss.
Key terms
- Exposure at default
- The modelled amount outstanding when a counterparty defaults, including relevant future drawings or movement.
- Economic capital
- Capital allocated to absorb unexpected loss at a chosen confidence level beyond the expected cost.
- Value at Risk
- A modelled loss threshold associated with a stated horizon and confidence level under a specified distribution, data set and position boundary.
- Back-testing
- Comparison of model forecasts with realised outcomes to assess whether the frequency and pattern of exceptions are credible under the stated confidence level.
- Loss given default
- The proportion of exposure not recovered after default, measured consistently with recovery costs, timing and the valuation convention used.
- Expected loss
- The modelled average credit loss over a defined horizon, commonly expressed as probability of default multiplied by loss severity and exposure.
Market and Credit Risk Measurement FAQ
Is Value at Risk a maximum possible loss?
No. It is a threshold for a stated horizon and confidence under model assumptions. Losses beyond the threshold remain possible and can be much larger during stress.
Why separate expected and unexpected credit loss?
Expected loss is a recurring modelled cost suited to pricing and allowance. Unexpected loss describes variability around that mean and informs diversification, limits and economic capital.
What information must accompany a VaR number?
State the monetary unit, holding period, confidence level, portfolio boundary and modelling approach. A threshold without those labels cannot be compared or interpreted. Also note important exclusions, such as liquidity cost or nonlinear positions, because the result is conditional on the data and assumptions rather than a universal description of possible loss.
How are VaR exceptions interpreted?
An exception occurs when realised loss exceeds the forecast threshold. A small number is expected because confidence is below one hundred per cent. Too many, clustered exceptions or exceptions linked to changing regimes can indicate weak calibration or dependence. The analyst compares observed and expected frequency before investigating magnitude and cause.
Why calculate loss given default from recovery?
Recovery and loss severity describe complementary portions of exposure when measured on the same basis. If recovery is forty per cent, loss given default begins at sixty per cent before any convention-specific adjustment. Keeping the percentage as a decimal in the expected-loss formula avoids a common hundredfold error and makes sensitivity analysis straightforward.
How should an expected-loss result be used?
Interpret it as a modelled average cost across comparable exposures or repeated periods, not the certain loss on one facility. It can inform pricing, allowances and portfolio planning. Limits, stress tests and capital remain necessary because realised loss can be zero, concentrated or far above the mean when defaults and recoveries are dependent.
When should model output be escalated without a limit breach?
Escalation is justified when assumptions, data or portfolio structure change enough to weaken the meaning of the reported threshold. Examples include clustered VaR exceptions, rapid rating migration, unusual facility drawings, falling collateral values or a concentration that the aggregate number conceals.
The analyst should describe the signal, its likely transmission and the decision that cannot safely wait for a formal breach. A review might refresh calibration, reduce exposure, tighten approval, add a scenario or increase monitoring.
Early escalation is not a declaration that the model is useless; it recognises that governance must respond when evidence moves outside the conditions under which the measure was judged reliable.
Exam move
Always write VaR with horizon, confidence, units and model boundary. For credit loss, compute LGD from recovery, convert percentages to decimals and finish with the management use of the result. Build a validation table containing forecast threshold, realised outcome, exception flag and an explanation for clusters.
For each credit example, write the units beside probability of default, loss given default and exposure before multiplying. Change one component at a time to see which assumption drives expected loss. Then separate the mean from a stress outcome and explain whether pricing, allowance, limit or capital is the relevant response.
Working through Market and Credit Risk Measurement in RISK5001? Sia is AskSia’s AI Finance tutor — ask any RISK5001 Market and Credit Risk Measurement question and get a clear, step-by-step explanation grounded in how RISK5001 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
