Market Risk Analytics

Introduction and context

Market risk is the possibility that changes in observable market prices or rates will reduce the value of a financial position or adversely affect income. The principal sources are interest rates, foreign-exchange rates, equity prices and commodity prices. Banks, investment firms, manufacturers, importers, exporters, pension funds and governments all face market risk, although their exposures and risk-management objectives differ.

Market risk analytics converts uncertain market movements into measurable exposure. It supports pricing, hedging, capital allocation, risk limits, stress testing and regulatory reporting. Measurement does not eliminate risk. Rather, it makes assumptions explicit and provides a disciplined basis for decisions.

Learning objectives

By the end of this chapter, you should be able to:

• distinguish market risk from credit, liquidity and operational risk;
• measure interest-rate exposure using duration, modified duration, convexity and key-rate sensitivities;
• explain foreign-exchange, equity and commodity exposure;
• calculate and interpret Value at Risk (VaR) using historical, parametric and Monte Carlo methods;
• identify model assumptions, data requirements and limitations; and
• recommend appropriate controls, validation and governance practices.

1. Foundations of market risk

A portfolio’s market value can be represented as a function of risk factors:

Here, V is portfolio value and rᵢ are risk factors such as yields, exchange rates, equity prices or commodity prices. For a small movement, the change in value can be approximated by a first-order Taylor expansion:

The first derivative is a sensitivity or “Greek.” The approximation is useful for small changes, but it becomes less reliable for large movements or instruments with material non-linearity, such as options.

Market risk differs from:

• Credit risk: loss caused by a counterparty’s failure to perform.
• Liquidity risk: inability to transact or fund positions at reasonable cost.
• Operational risk: loss from failed processes, people, systems or external events.
• Model risk: loss caused by inappropriate models, data or implementation.

A sound risk framework identifies positions, maps them to risk factors, measures exposure, compares results with limits, investigates exceptions and reports decisions to responsible managers.

2. Interest-rate risk analytics

Interest-rate risk arises because the present value of future cash flows changes when discount rates change. For a fixed-rate bond, rising yields normally reduce price, while falling yields increase price.

2.1 Bond valuation

For a bond with coupon C, face value F, yield per period y and n periods:

P is the present price. The formula assumes known cash flows, a constant yield for discounting and consistent payment and compounding periods. If annual coupon rate is c and face value is F, then C = cF when payments are annual.

2.2 Macaulay and modified duration

Macaulay duration is the weighted average time to receive the bond’s cash flows:

CFₜ is the cash flow at time t, measured in periods. Macaulay duration is expressed in periods or years. Modified duration estimates percentage price sensitivity to yield:

where m is the number of coupon periods per year and y is the annual nominal yield. For a small change in annual yield ∆y, with consistent units:

The negative sign reflects the inverse price–yield relationship. If yield changes by 0.01 rather than 1, the result is a decimal percentage change. Duration is most useful for option-free, fixed-income instruments and relatively small, approximately parallel yield movements.

Worked example: duration sensitivity

Suppose a bond is worth 1,000 monetary units and has modified duration of 4.8 years. If its yield rises by 0.50 percentage points, then ∆y = 0.005. The estimated price change is:

∆P/P ≈ −4.8 × 0.005 = −0.024

Thus, the estimated loss is 2.4%, or approximately 24 monetary units, and the new value is approximately 976. The estimate ignores convexity and assumes the relevant yield curve movement is adequately represented by one yield.

2.3 Convexity

Duration gives a linear approximation. Bond prices generally have curvature with respect to yield, so a second-order approximation is preferable for larger movements:

Cₓ is convexity, measured consistently with the yield convention. For a discretely compounded bond, a common approximation is:

This expression assumes yield per period and unit-period changes; scaling must be adjusted when coupon frequency differs. Convexity is normally positive for a plain option-free bond, meaning the actual price increase when yields fall tends to exceed the duration-only estimate, while the actual price decline when yields rise tends to be smaller. Callable bonds and other instruments with embedded options can exhibit negative or changing convexity.

Using the previous example, suppose convexity is 35. For ∆y = 0.005:

∆P/P ≈ −0.024 + 0.5 × 35 × 0.005² = −0.0235625

The estimated loss is therefore about 2.356%, or 23.56 monetary units, rather than 24. The difference is small for this movement but becomes more important for larger shocks.

2.4 Yield-curve and basis risk

A parallel shift assumption is often unrealistic. Short- and long-term rates may move differently, producing yield-curve risk. A portfolio can be protected against a parallel shift yet remain exposed to steepening, flattening or curvature changes.

Key-rate duration measures sensitivity to a movement at a particular maturity while other maturities are held approximately constant. It is useful for mapping a portfolio to 2-year, 5-year, 10-year and other curve points. Basis risk occurs when the hedge and exposure respond to different rates, such as a loan priced from one benchmark and a swap linked to another.

For floating-rate positions, analysts examine reset frequency, reference rate, spread, repricing gap and optionality. A maturity mismatch between assets and liabilities creates net interest-rate exposure even when both instruments are floating-rate.

3. Foreign-exchange risk

Foreign-exchange (FX) risk is the possibility that exchange-rate movements change the domestic-currency value of assets, liabilities, revenues or costs.

Three forms are commonly distinguished:

• Transaction exposure: a contractual foreign-currency receivable or payable.
• Translation exposure: accounting conversion of a foreign subsidiary’s statements.
• Economic exposure: the longer-term effect of exchange rates on competitiveness, prices and cash flows.

If an entity expects to receive USD 2 million and the domestic-currency value of one USD falls from 1.10 to 1.06, the domestic value falls from 2.20 million to 2.12 million. The exposure is therefore a potential 80,000 loss, before considering hedges and transaction costs.

For a foreign-currency asset A valued in foreign currency and exchange rate S defined as domestic currency per unit of foreign currency:

For small movements, the change is:

The definition of S must be stated clearly. Reversing the quotation convention reverses the interpretation of appreciation and depreciation.

FX forwards, futures, swaps and options can reduce exposure, but they introduce counterparty, basis, liquidity and rollover risks. A hedge should be evaluated against the underlying exposure’s amount, timing, currency and cash-flow uncertainty.

4. Equity and commodity risk

Equity risk

Equity exposure may be measured through market value, beta, delta and factor sensitivities. For a diversified portfolio, a simple one-factor model is:

Rₚ is portfolio return, Rₘ is market return, α is the intercept, βₚ measures sensitivity to the market and εₚ is idiosyncratic return. Beta is estimated statistically and can change over time; it is not a guaranteed hedge ratio.

For a portfolio with value V and beta β, an approximate market-factor loss from a market return Rₘ is:

Sector, country, size, value, momentum and interest-rate factors may explain additional variation. Options require delta, gamma, vega and perhaps theta analysis because their values are non-linear and depend on implied volatility and time.

Commodity risk

Commodity exposure affects producers, processors, transport firms and consumers. It may arise from crude oil, natural gas, metals, agricultural products or energy inputs. Important dimensions include grade, location, delivery date, storage, transportation and quality. A futures hedge may not perfectly offset a physical exposure because of basis risk between the physical price and the futures price.

Commodity positions are often assessed through spot price, futures-curve and spread sensitivities. Contango and backwardation affect the cost or benefit of rolling futures contracts. Limits should consider not only price volatility but also concentration, market depth, margin requirements and delivery obligations.

5. Value at Risk

Value at Risk estimates a loss threshold over a specified holding period and confidence level. A statement such as “one-day 99% VaR is 500,000” means that, under the model and data assumptions, the loss is expected to exceed 500,000 on approximately 1% of comparable days. It does not mean the maximum possible loss is 500,000.

VaR must always specify:

1. portfolio and valuation time;
2. horizon, such as one trading day;
3. confidence level, such as 95% or 99%;
4. currency and whether the measure is absolute or relative;
5. treatment of fees, liquidity and financing; and
6. model, data window and revaluation method.

5.1 Historical simulation

Historical simulation applies observed historical risk-factor changes to today’s portfolio. The process is:

1. collect clean historical prices or returns;
2. calculate changes over the chosen horizon;
3. revalue the current portfolio under each historical scenario;
4. rank the resulting profit-and-loss outcomes; and
5. select the appropriate lower-tail percentile as VaR.

For a loss variable L, empirical VaR at confidence c is the c-quantile of losses, subject to the chosen quantile convention. Historical simulation does not require a normal distribution, but it assumes the historical sample is informative about future conditions and that the portfolio can be revalued consistently.

5.2 Parametric VaR

For a linear portfolio whose return is approximately normally distributed, one-day VaR at confidence c is:

where V is portfolio value, σ_V is the standard deviation of portfolio return, and z_c is the standard-normal quantile. If the mean return μ is included, a loss measure is often written:

For short horizons, μ is commonly set to zero because estimating expected return is uncertain and its effect is small. The formula is appropriate only when units are consistent and the linear-normal assumptions are reasonable.

For multiple risk factors, portfolio variance is:

w is the vector of monetary or return exposures, depending on the formulation, and Σ is the covariance matrix. The resulting variance must be non-negative in a valid covariance matrix. Correlations reduce or increase risk depending on their signs and magnitudes.

#### Worked example: parametric VaR

A portfolio is worth 5,000,000. Its daily volatility is 1.2%, and the 99% normal quantile is approximately 2.326. Ignoring mean return:

VaR₉₉ = 2.326 × 0.012 × 5,000,000 = 139,560

The estimated one-day 99% VaR is 139,560 monetary units. This is a model-based threshold, not a forecast of the worst possible loss.

A square-root-of-time scaling rule is sometimes used:

It assumes independent, identically distributed returns with stable volatility and a linear portfolio. Volatility clustering, autocorrelation, non-trading days and nonlinear instruments can make the rule unreliable.

5.3 Monte Carlo simulation

Monte Carlo simulation specifies distributions and dependence structures for risk factors, generates many scenarios, reprices the portfolio and estimates a loss percentile. It can represent non-normal distributions, stochastic volatility, correlations and option non-linearity more flexibly than a simple parametric method.

Its weaknesses include model specification risk, computational cost, random sampling error and dependence on calibration data. More simulations reduce sampling error but do not correct a flawed economic model.

5.4 Expected shortfall

Expected Shortfall (ES) measures the average loss beyond the VaR threshold. At confidence c:

where L is loss and E denotes expectation. ES provides information about tail severity and is generally more sensitive than VaR to extreme losses. In practice, the estimate depends strongly on the number and quality of tail observations and on the treatment of illiquid or stressed positions.

6. Comparison and model choice

No single method is universally superior. Analysts often compare methods, perform stress tests and investigate disagreements rather than selecting a method solely because it produces the lowest number.

7. Stress testing, backtesting and governance

Backtesting compares predicted VaR with actual or hypothetical outcomes. If a 99% one-day VaR model produces many more exceptions than expected, the model, data, valuation or implementation should be investigated. A small number of exceptions does not prove that a model is correct; an unusually calm sample can conceal weaknesses.

Stress testing evaluates specified extreme but plausible scenarios, such as a sharp yield-curve steepening, currency devaluation, equity crash, commodity supply shock or simultaneous market dislocation. Historical episodes can be replayed, but hypothetical scenarios are also necessary because future crises will not exactly reproduce the past.

Good governance includes:

• independent model validation;
• documented assumptions and data lineage;
• approval of risk-factor mappings and limits;
• separation of front-office, risk and control responsibilities;
• regular performance monitoring and backtesting;
• escalation procedures for limit breaches;
• version control and reproducible calculations; and
• clear communication of uncertainty to decision-makers.

Regulatory capital frameworks may prescribe particular risk measures, stress periods, liquidity horizons and modellability requirements. Regulatory compliance does not replace internal risk management: a regulatory measure may be designed for capital adequacy rather than daily trading decisions.

8. Applied case: illustrative multi-asset treasury portfolio

Background

A fictional manufacturing company holds a fixed-rate bond portfolio worth 20 million, expects to receive USD 3 million in 60 days, and forecasts that energy purchases will cost 1 million over the next quarter. Management is concerned about rising rates, a weaker USD and higher energy prices.

Problem

The treasury team must identify material exposures and select suitable analytics without assuming that one aggregate VaR number captures every risk.

Analysis

The bond portfolio has modified duration of 5.2 and convexity of 40. A parallel yield increase of 75 basis points gives:

∆P/P ≈ −5.2(0.0075) + 0.5(40)(0.0075²)

The first term is −0.039; the convexity adjustment is 0.001125. The estimated loss is 3.7875%, or approximately 757,500 on 20 million. Key-rate analysis is then required to determine whether the portfolio is especially exposed at the 5-year or 10-year point.

The USD receivable is an asset in USD. If the domestic currency per USD falls, its domestic value decreases. The company could sell USD forward for the expected amount, while allowing for uncertainty in the receipt date and amount.

The energy exposure is not necessarily identical to a single futures contract. The team must compare the purchase location, grade, delivery month and benchmark price with available contracts. A futures hedge can reduce price risk but leave basis and volume risk.

Approach and solution

The team establishes separate limits for interest-rate DV01, FX notional and commodity price exposure. It calculates historical and parametric VaR for the combined portfolio, then runs stress scenarios: a 100-basis-point rate rise, a 10% currency movement and a 30% energy-price increase. It evaluates hedge effectiveness using cash-flow timing and basis analysis rather than notional amounts alone.

Outcome

The analysis shows that aggregate VaR is moderate because some exposures offset statistically, but the stress tests reveal a substantial energy-price loss. Management therefore approves a partial commodity hedge, a forward hedge for a high-confidence portion of the USD receipt and a curve-aware interest-rate hedge.

Learning points

Diversification can reduce measured VaR without removing concentrated stress exposure. Separate risk-factor sensitivities, scenario analysis and liquidity assessment are essential. Hedge ratios must reflect economic exposure, not merely accounting labels or gross notional value.

9. Limitations, ethics and contemporary practice

Market-risk models are conditional estimates. They can fail during structural breaks, regime changes, illiquidity, gaps, market closures and correlated crises. Historical data may omit the very event management most needs to understand. Correlations can rise sharply during stress, and option values may change nonlinearly.

Ethically, analysts should not select a methodology because it produces a favourable risk number, omit inconvenient scenarios or present model outputs as certainty. Material assumptions, uncertainty ranges and known limitations should be disclosed. Risk reports should be understandable to managers who are accountable for decisions.

Modern systems use cloud computing, automated data pipelines, real-time pricing and machine-learning techniques for anomaly detection, volatility forecasting and scenario generation. These tools can improve speed and pattern recognition, but they introduce explainability, data-quality, drift and governance risks. A machine-learning forecast should be benchmarked against transparent models, validated out of sample and subject to human oversight. Automation must not remove responsibility for reviewing unusual results.

Best practice is therefore a layered approach: sensitivity measures for explanation, VaR for standardized monitoring, stress testing for non-linear and extreme outcomes, Expected Shortfall for tail severity, and qualitative judgement for risks that data cannot capture.

Chapter Summary

Market risk analytics measures the effect of interest-rate, FX, equity and commodity movements on portfolio value. Bond duration provides a first-order estimate of price sensitivity, while convexity improves the estimate for larger yield movements. Key-rate measures recognize that yield curves do not move in parallel. FX exposure depends on currency denomination and quotation convention; equity exposure can be assessed through beta and factor sensitivities; and commodity exposure requires attention to basis, location, quality and futures-curve effects.

VaR is a specified loss percentile, not a maximum loss. Historical simulation preserves empirical movements, parametric VaR is efficient under linear-normal assumptions, and Monte Carlo simulation offers flexibility at the cost of model and computational complexity. Expected Shortfall describes losses beyond VaR. Backtesting, stress testing, validation, governance and ethical disclosure are necessary because every model is conditional and imperfect.

Key Terms

Market risk: Potential loss from adverse movements in market prices or rates.

Risk factor: A market variable that influences portfolio value.

Duration: A measure of bond-price sensitivity to yield.

Macaulay duration: Present-value-weighted average timing of cash flows.

Modified duration: Approximate percentage price sensitivity to yield.

Convexity: Curvature of the price–yield relationship.

DV01: Approximate change in value for a one-basis-point yield movement.

Key-rate duration: Sensitivity to a particular maturity on the yield curve.

Basis risk: Risk that a hedge and exposure do not move identically.

Foreign-exchange risk: Risk from changes in currency exchange rates.

Transaction exposure: Risk from contracted foreign-currency cash flows.

Beta: Sensitivity of an asset or portfolio to a market factor.

Commodity risk: Risk from changes in commodity prices and related basis factors.

Value at Risk: Loss threshold at a stated horizon and confidence level.

Historical simulation: VaR method using observed historical scenarios.

Parametric VaR: VaR based on distributional and covariance assumptions.

Monte Carlo simulation: Scenario generation from specified stochastic models.

Expected Shortfall: Average loss beyond a VaR threshold.

Backtesting: Comparing model forecasts with subsequent outcomes.

Stress testing: Evaluating outcomes under extreme but plausible scenarios.

Review Questions

Short Answer Questions

1. What is market risk, and how does it differ from liquidity risk?
2. Why does a bond’s price generally fall when its yield rises?
3. Define modified duration and state its principal limitation.
4. What is the difference between transaction and translation FX exposure?
5. Why is a VaR estimate incomplete without a horizon and confidence level?

Long Answer Questions

1. Explain duration and convexity, including their use in interest-rate risk management.
2. Compare historical, parametric and Monte Carlo VaR methods.
3. Discuss yield-curve risk, key-rate duration and basis risk.
4. Explain how governance and backtesting improve the reliability of market-risk models.
5. Evaluate the role of stress testing alongside VaR and Expected Shortfall.

Analytical / Application Questions

1. A bond portfolio worth 8 million has modified duration 6.1. Estimate its percentage and monetary change for a 40-basis-point yield increase, ignoring convexity. Interpret the result.
2. A portfolio worth 2 million has daily volatility of 1.5%. Using a 95% normal quantile of 1.645 and zero mean return, calculate one-day parametric VaR and state two assumptions.
3. A company expects a foreign-currency receipt but also has a foreign-currency payable in the same month. Design an analysis to determine the net hedge requirement, including timing, exchange-rate quotation and basis risks.

Multiple Choice Questions with Answers

Question 1

A bond has modified duration of 4 and its yield rises by 0.25 percentage points. Ignoring convexity, the estimated percentage price change is:

A. −1.00%

B. −0.25%

C. +1.00%

D. +0.0625%

Correct Answer: A. −1.00%

Question 2

Which VaR method directly applies observed historical risk-factor changes to the current portfolio?

A. Parametric simulation

B. Historical simulation

C. Duration matching

D. Factor regression

Correct Answer: B. Historical simulation

Question 3

A portfolio worth 1,000,000 has daily volatility of 2% and a 99% normal quantile of 2.326. Ignoring mean return, its parametric VaR is:

A. 23,260

B. 46,520

C. 232,600

D. 2,326,000

Correct Answer: B. 46,520

Question 4

Convexity improves a duration estimate because it captures:

A. Counterparty default probability

B. Curvature in the price–yield relationship

C. Foreign-currency translation

D. Trading volume

Correct Answer: B. Curvature in the price–yield relationship

Question 5

If an exchange rate is quoted as domestic currency per unit of foreign currency, a fall in the rate reduces the domestic value of:

A. A foreign-currency asset

B. A domestic-currency asset

C. A domestic-currency liability only

D. All positions equally

Correct Answer: A. A foreign-currency asset

Question 6

Expected Shortfall measures:

A. The average return of the portfolio

B. The average loss beyond the VaR threshold

C. The maximum possible loss

D. The standard deviation of interest rates

Correct Answer: B. The average loss beyond the VaR threshold

Question 7

Which risk is especially associated with a futures hedge that does not exactly track a physical commodity exposure?

A. Basis risk

B. Translation risk

C. Settlement-free risk

D. Duration risk

Correct Answer: A. Basis risk

Question 8

For a linear portfolio with factor exposure vector w and covariance matrix Σ, portfolio variance is represented by:

A. w + Σ

B. wᵀΣw

C. Σ/w

D. wᵀ + Σᵀ

Correct Answer: B. wᵀΣw

Question 9

A 99% one-day VaR model produces substantially more exceptions than expected. The most appropriate first response is to:

A. Ignore the exceptions because VaR is only an estimate

B. Investigate data, valuation, assumptions and implementation

C. Automatically reduce the confidence level

D. Delete the largest losses from the sample

Correct Answer: B. Investigate data, valuation, assumptions and implementation

Question 10

Which instrument characteristic most clearly makes a simple linear VaR approximation less reliable?

A. A fixed notional amount

B. An option with materially nonlinear value

C. A fully collateralized cash balance

D. A known coupon payment

Correct Answer: B. An option with materially nonlinear value