Risk Management

Market Risk

Read at your own pace. Use chapter practice to check your understanding.

Chapters and chapter contents
100%
Practise this chapter ↓

Last updated · By the Clerk2Officer editorial team · Report a correction

1. Identify the variable that can move

Market risk arises when changes in market variables adversely affect the value or earnings of a position. For a bank, relevant variables include interest rates, exchange rates, equity prices, commodity prices and related spreads or volatilities. Begin with the position and its sensitivity. A fixed-rate bond generally loses value when its required yield rises; a long dollar position gains in rupees when the dollar strengthens. These directional relationships are more useful than memorising a risk label without understanding the cash flows.

Consider fictional Meridian Bank holding INR50 crore of fixed-rate bonds and a net long USD1 million position. A rise in domestic yields can hurt the bonds, while an increase in INR per USD benefits the currency position. The combined outcome depends on both movements and position sizes. A report showing only net profit can hide a large loss in one activity offset by an unrelated gain in another. Risk identification should therefore preserve meaningful drivers before aggregating them.

Market risk can exist even when a counterparty performs every contractual obligation. The bond issuer may pay in full while the market value falls. Credit-spread changes can also affect price before default occurs. Management must distinguish price sensitivity, issuer credit quality and settlement obligations while recognising their interaction. A hedge selected for the wrong driver can leave the main exposure untouched. For example, a currency hedge does not eliminate the interest-rate sensitivity of a foreign-currency fixed-rate bond.

2. Organise the dealing control framework

The board-approved framework should connect permitted activities, risk appetite, limits, valuation and escalation. Senior management needs a clear view of which instruments the bank can understand, price and control. Product approval should precede dealing and consider legal documentation, accounting, settlement, systems, staff skills and stress behaviour. An instrument should not become acceptable merely because a dealer can obtain a market quote. The bank must also be able to verify the quote and manage the exposure after execution.

Front-office staff execute authorised business; independent risk oversight monitors exposures and challenges measurements; settlement and accounting functions confirm and reconcile transactions. The actual organisational labels can vary, but conflicting duties need effective separation. Independent price verification matters because a trader's incentive may conflict with conservative valuation. Reconciliations should detect missing, duplicated or amended trades. A sophisticated VaR model is unreliable if the position file excludes a material contract or uses an incorrect notional amount.

Limits should match the risk. Notional limits constrain size, sensitivity limits constrain response to a specified movement, VaR limits constrain a modelled quantile and stop-loss limits respond to realised or marked losses under defined rules. None is a complete substitute for the others. The framework should specify monitoring frequency, breach authority and required action. A limit increase after a breach should require genuine review, rather than becoming the routine means of avoiding an inconvenient reduction in exposure.

From position to actionDirection and units come first. VaR is a quantile, not maximum loss. Review limits and feasible hedges.From position to action1Identify the price driverDirection and units come first.2Measure and stressVaR is a quantile, not maximum loss.3Control and explainReview limits and feasible hedges.
From position to action. Direction and units come first. VaR is a quantile, not maximum loss. Review limits and feasible hedges.

3. Measure bond sensitivity with duration

Modified duration approximates the percentage price change of a bond for a small change in yield. The first-order relationship is percentage price change approximately equal to minus modified duration multiplied by the yield change expressed as a decimal. For a bond portfolio worth INR50 crore with modified duration 4, a 0.50 percentage-point yield rise is 0.005 in decimal form. The approximate price change is minus 4 times 0.005, or minus 2%, equivalent to an INR1 crore loss.

One basis point is 0.01 percentage point, or 0.0001 as a decimal yield change. DV01 expresses the approximate monetary value change for one basis point. In the example, DV01 is INR50 crore times 4 times 0.0001, or INR0.02 crore: INR2 lakh. State whether DV01 is reported as a positive magnitude or a signed sensitivity. The calculator reports the magnitude and separately explains the sign of the price change for the supplied yield shock.

Duration is a local approximation. Large changes, non-parallel yield-curve movements and embedded options can make actual results differ materially. Convexity provides a second-order adjustment for suitable instruments, but requires a consistent definition and units. Callable bonds can behave differently as the likelihood of exercise changes. Use duration to understand and compare first-order exposure, then apply appropriate revaluation or scenario methods when the position is nonlinear or the proposed shock is too large for a simple approximation.

4. Read foreign-exchange exposure correctly

A net long foreign-currency position benefits from appreciation of that currency against the reporting currency, while a short position benefits from depreciation, ignoring other effects. If Meridian is long USD1 million and INR per USD rises from 84 to 85, the simple translation gain is INR1 million, or INR10 lakh. The direction reverses for a short dollar position. Always state the quotation convention because an increase in a reciprocal quote means the opposite economic movement.

Netting assets and liabilities in the same currency can reduce exposure, but dates and contractual conditions still matter. A dollar receivable in six months does not automatically fund a dollar payment tomorrow. A forward can hedge an exchange-rate movement while introducing counterparty, liquidity or settlement requirements. If the underlying amount or date changes, the hedge may become imperfect. Monitoring should therefore track the underlying exposure alongside the derivative rather than treating a signed hedge contract as permanent proof of protection.

Basis risk arises when the hedge and exposure do not move identically. A proxy currency hedge, a benchmark mismatch or different maturities can leave residual sensitivity. Over-hedging can create a new speculative position. The hedge decision should specify the objective, amount, horizon, permitted instruments and effectiveness monitoring. A favourable derivative result alone is not evidence of success; it should be assessed together with the underlying loss or gain and the costs incurred to maintain the hedge.

5. Interpret value at risk precisely

Value at risk, or VaR, is a loss quantile for a specified horizon and confidence level under a stated model. A one-day 99% VaR of INR2 crore means that the model assigns approximately 1% probability to a loss exceeding INR2 crore over one day. It is not a maximum possible loss, a promised limit or the expected loss on the worst day. The phrase is incomplete without the confidence level, horizon, portfolio definition and measurement method.

For a simplified zero-mean normal-return model, VaR equals portfolio value multiplied by return volatility multiplied by the normal quantile. With INR100 crore value, daily volatility of 1% and a 99% quantile of approximately 2.32635, one-day VaR is about INR2.32635 crore. Volatility entered as 1% must become 0.01 in the formula. The chapter calculator supports specified confidence choices and reports its assumptions; it does not estimate volatility from prices or claim to implement a regulatory capital model.

Under independent, identically distributed returns with constant volatility, a multi-day volatility estimate scales by the square root of the number of days. Ten-day VaR in the same model is therefore one-day VaR multiplied by the square root of ten, approximately INR7.3566 crore. Multiplying by ten would be a different assumption. Serial dependence, changing volatility, illiquidity and nonlinear positions can invalidate the square-root shortcut, so the convenience of the formula should not be mistaken for universal accuracy.

6. Compare VaR methods and expected shortfall

Parametric methods use a specified distribution and estimated parameters. Historical simulation applies observed historical movements to a current portfolio. Monte Carlo simulation generates movements from a model and revalues the positions. Each method depends on choices about data, scenarios, valuation and horizon. Historical simulation is not assumption-free: the chosen history may omit the next crisis. Monte Carlo is not automatically superior: simulating many paths from a poor model can produce a very precise description of the wrong distribution.

Expected shortfall describes the average loss in the selected tail, under an appropriate definition and model. It provides information about losses beyond a quantile that VaR alone does not reveal. In a continuous zero-mean normal model, expected shortfall at confidence c equals value times volatility times the standard-normal density at the quantile divided by one minus c. At 99%, the multiplier is approximately 2.6652, larger than the 2.32635 VaR multiplier for the same one-day inputs.

This distinction matters when two portfolios share the same VaR but have different extreme losses. A portfolio with rare catastrophic outcomes may have a much larger expected shortfall. Neither measure removes model risk or the need for stress testing. This chapter uses expected shortfall as an explanatory internal-risk concept and does not assert that its simple normal calculator is the applicable Indian trading-book capital method. Regulatory approaches require their own scope, definitions, calibration and implementation conditions.

7. Diversification depends on relationships

For two exposures with loss standard deviations s1 and s2, a simplified portfolio standard deviation is the square root of s1 squared plus s2 squared plus twice correlation times s1 times s2. With standard deviations of INR3 crore and INR4 crore, zero correlation produces INR5 crore; perfect positive correlation produces INR7 crore. At perfect negative correlation the result is INR1 crore. This is a variance calculation, not a claim that real-world positions can always be perfectly hedged.

Correlation estimates depend on the observation period and variables selected. Correlations can rise in stress, particularly when investors sell different assets together to obtain cash. A diversification benefit estimated during calm markets may therefore overstate protection in an adverse scenario. Risk reports should show important concentrations and factor sensitivities alongside aggregate VaR. Otherwise a low aggregate number can conceal substantial exposure to one common driver or an unstable offset between positions that may break down when most needed.

Diversification also has operational and legal limits. Positions in different entities may not be freely transferable, and offsetting market sensitivities can still create separate collateral calls or settlement demands. A bank should not net exposures merely because their estimated prices move in opposite directions. Establish the purpose of aggregation and the enforceability of any contractual netting. Economic offset, accounting offset and regulatory recognition are related but distinct concepts, each with its own requirements and practical implications.

8. Backtesting and stress testing challenge the model

Backtesting compares forecasts with realised outcomes under a defined profit-and-loss measure. A 99% one-day VaR model would, over a sufficiently representative sample, be expected to have exceptions on roughly 1% of observations. That does not mean exactly one exception in every hundred days. Investigate both the number and pattern of exceptions, data quality and differences between actual trading outcomes and the portfolio assumed by the forecast. A model can be misused even when its mathematical implementation is correct.

Stress testing examines specified adverse movements, including events beyond the routine model window. A useful scenario might combine a yield rise, currency depreciation, widening bid-ask spreads and higher collateral demands. Explain why the movements occur together and revalue the positions consistently. Adding unrelated worst observations can produce an incoherent scenario; relying only on recent mild observations can be too optimistic. The aim is a relevant challenge to the portfolio, not merely the largest number obtainable from arbitrary inputs.

Reverse stress testing identifies conditions that would breach a critical loss or capital threshold. It can reveal that a seemingly diversified portfolio depends on one fragile market relationship. Management should then consider exposure reduction, different hedges, additional liquidity or revised limits. Actions must be feasible in the scenario. Assuming that a large position can be sold instantly at yesterday's mid-market price ignores the very liquidity deterioration that may make the stress severe.

Normal one-day tail measuresINR crore; value 100, volatility 1%, zero mean. Neither tail measure is maximum loss. [('Standard deviation', 1), ('99% VaR', 2.32635), ('99% expected shortfall', 2.66521)]Normal one-day tail measuresStandard deviation: 199% VaR: 2.3263599% expected shortfall: 2.66521
INR crore; value 100, volatility 1%, zero mean. Neither tail measure is maximum loss.

9. Trading liquidity changes achievable value

A quoted price may apply only to a small trade. The bid-ask spread, market depth, concentration and time needed to liquidate determine the cost of reducing a position. Market liquidity can disappear quickly after an issuer problem or broad market disruption. Valuation and liquidity management should therefore distinguish an observed quote from an executable price for the bank's entire holding. A model based on small daily price changes can understate the loss incurred during a forced large sale.

Suppose a portfolio is marked at INR30 crore, but a stressed sale is assumed to require a 4% discount plus INR0.10 crore execution costs. Estimated proceeds are INR28.70 crore, and the reduction from marked value is INR1.30 crore. This is an explicit scenario, not a forecast of an actual market quote. If the same securities are already pledged, they may not be available for immediate sale. Asset availability matters as much as the estimated haircut.

Funding and trading liquidity interact through margin calls and collateral. A hedge that reduces eventual price risk may require cash today when its market value moves adversely. The bank needs resources to maintain the hedge until the underlying benefit is realised. An otherwise sensible hedge can be disrupted if funding assumptions fail. For this reason, treasury and risk teams should review collateral terms, settlement dates and stressed cash demands together with market-value sensitivities.

10. Design mitigation and reporting around decisions

Market-risk responses include reducing positions, changing maturities, diversifying drivers and using suitable derivatives. Each response has costs and residual risks. A futures hedge can leave basis risk; an option requires a premium and changes sensitivity as prices move; a swap introduces counterparty and collateral considerations. The objective should be explicit: stabilise earnings, protect economic value, cap a loss or support customer business. Without an objective, it is impossible to judge whether the hedge remains appropriate after conditions change.

Reports should show positions, sensitivities, VaR or other approved measures, stress results, realised and unrealised losses, liquidity assumptions and breaches. Explain material daily movements rather than presenting tables alone. A falling VaR could reflect reduced positions, lower measured volatility or a data error; those explanations have different implications. Independent review should challenge unusually favourable results as well as obvious losses. Good risk reporting makes uncertainty visible and supports a timely decision by someone with clear authority.

For revision, solve the market case in order: identify the variable and position direction, establish units, select the appropriate sensitivity or model, calculate the result and explain what is excluded. A VaR number is a quantile, duration is an approximation and a hedge is a conditional relationship. The strongest answer gives both the numerical result and its interpretation. That combination prevents common errors such as calling VaR a maximum loss or treating a long foreign-currency position as harmed by that currency's appreciation.

11. A dealer's integrated morning review

Imagine that Meridian's morning report shows bond DV01 of INR2 lakh, a long USD1 million position and a one-day VaR of INR2.33 crore. The dealer proposes doubling the bond holding because yesterday's market was quiet. Independent risk should examine the new sensitivity, remaining limit headroom, financing capacity and the effect of a stressed yield movement. Yesterday's quiet return does not establish that doubling today's exposure is prudent. Position size and the reliability of the volatility estimate both matter.

If the holding doubles with duration unchanged, DV01 doubles to INR4 lakh. A 50-basis-point parallel yield rise then implies an approximate INR2 crore price loss. That calculation is a useful first check, but the team should also consider concentration, non-parallel shifts and execution costs. If the dealer plans to hedge with a different maturity, the residual curve exposure should be shown. A nominally equal hedge amount does not necessarily provide an equal and opposite sensitivity.

The decision should record approval, conditions and monitoring. If market depth deteriorates or the hedge basis changes, the plan may need revision before a hard loss limit is reached. This is the practical connection between identification, measurement, control and reporting. The tools below allow you to vary the assumptions and observe the direction and size of the results; the written explanation remains essential because a calculator cannot decide whether the model fits the actual instrument.

Open calculators and interactive practice

LEARN THROUGH EXPLANATIONS

Chapter practice

This practice is separate from timed mock tests. Answers and explanations appear after you choose an option.

Your reading progress

Sign in to save your progress across devices.

Browse published timed tests for this paper