Market, Liquidity and Asset–Liability Risk Analytics

Introduction

Banks and insurers face several interdependent balance-sheet risks. Market prices can move abruptly, securities may become difficult to sell, depositors may withdraw funds, and assets and liabilities may reprice at different times. Risk analytics converts these exposures into measurable sensitivities, loss distributions, liquidity requirements and balance-sheet gaps.

Market risk concerns losses caused by changes in interest rates, exchange rates, equity prices, credit spreads, commodity prices or volatility. Liquidity risk has two dimensions: market liquidity risk—the inability to trade without a material price concession—and funding liquidity risk—the inability to meet payment obligations when due. Asset–liability management (ALM) examines how differences in maturity, repricing, duration, optionality and currency affect earnings, economic value and solvency.

These risks reinforce one another. A market shock may reduce collateral values, trigger margin calls and force asset sales. Forced sales may widen bid–ask spreads and cause further losses. Effective management therefore requires an integrated rather than silo-based approach.

Learning Objectives

After studying this chapter, students should be able to:

1. Calculate and interpret price sensitivities, duration, convexity and basis-point value.
2. Estimate Value at Risk (VaR) and Expected Shortfall (ES) using appropriate assumptions.
3. Explain VaR backtesting and model-validation procedures.
4. design historical, hypothetical and reverse stress scenarios.
5. Calculate market- and funding-liquidity metrics.
6. Apply repricing-gap, duration-gap and cash-flow-ladder analysis.
7. Evaluate the limitations, governance requirements and regulatory implications of risk models.

1. An Integrated View of Financial Risk

A useful risk architecture connects positions, risk factors, cash flows and management responses.

Figure: A market shock can generate valuation losses and cash demands, creating liquidity pressure and further balance-sheet effects.

The central analytical principle is that no single measure gives a complete risk assessment. Sensitivities describe local exposure, VaR and ES summarize probabilistic losses, stress tests examine exceptional conditions, and liquidity and ALM metrics identify survival and structural vulnerabilities.

2. Market-Risk Sensitivity Measures

2.1 First- and second-order sensitivities

For a portfolio value V(x) driven by risk factor x, a small change may be approximated by a second-order Taylor expansion:

Here, δ = ∂V/∂x is delta, measured as currency value per unit change in the risk factor, and γ = ∂²V/∂x² is gamma, measured as currency value per squared unit. The approximation assumes that the derivatives exist and the risk-factor change is not so large that higher-order effects become material.

Delta indicates directional exposure. Gamma captures curvature and is especially important for options. Other option sensitivities include vega, the change in value for a change in implied volatility; theta, the change as time passes; and rho, the response to interest rates. These measures depend on the valuation model and should not be interpreted independently of model assumptions.

2.2 Modified duration, DV01 and convexity

For a fixed-income instrument, modified duration measures the approximate percentage price response to a small parallel change in yield:

In this expression, P is the bond price in currency units, D_mod is modified duration in years, and ∆y is the yield change expressed as a decimal, such as 0.0010 for 10 basis points. The approximation assumes a small, parallel yield movement and unchanged cash flows.

Macaulay duration and modified duration are related by:

Here, D_Mac is Macaulay duration, y is the nominal annual yield and m is the number of compounding periods per year. The formula requires yield and payment-frequency conventions to be consistent.

Dollar value of one basis point, also called DV01 or PV01, is:

DV01 is reported as a positive loss magnitude for a one-basis-point rise in yield. Convexity improves the approximation for larger movements:

The convexity C is expressed in years squared under the conventional yield-based definition. Embedded options can create negative or unstable convexity, making static duration estimates unreliable.

Worked Example: Duration and Convexity

A bond portfolio has a market value of $10 million, modified duration of 4.2 years and convexity of 25 years². For a 50-basis-point increase in yield, ∆y = 0.005:

The estimated change is −2.06875%, giving:

The portfolio DV01 is:

Duration alone would estimate a $210,000 loss; positive convexity reduces the estimated loss to $206,875.

2.3 Key-rate and spread sensitivities

A parallel shift is rarely sufficient for portfolios exposed to yield-curve changes. Key-rate duration measures sensitivity to a change at a selected maturity while other curve points are held fixed or interpolated. Credit-spread DV01 measures the price response to a one-basis-point spread change. Currency and equity exposures may similarly be decomposed by risk factor, region, issuer or sector.

3. Value at Risk

3.1 Definition

Let L denote portfolio loss over a stated horizon. Positive values represent losses. VaR at confidence level α is the loss quantile:

A one-day 99% VaR of $2 million means that, under the model, the one-day loss will not exceed $2 million with 99% probability. It does not state the maximum possible loss or the expected loss after the threshold is exceeded.

A VaR report is incomplete unless it specifies the valuation date, horizon, confidence level, currency, data window, weighting scheme and treatment of nonlinear instruments.

3.2 Parametric VaR

If portfolio returns are normally distributed and portfolio value is V, then loss is approximately normal. With expected return μ_R and return volatility σ_R, parametric VaR is:

Here, z_α is the standard-normal quantile. For a $5 million portfolio with zero expected daily return, daily volatility of 1.2% and z_0.99 = 2.326, one-day VaR is:

Normal VaR is convenient but can underestimate risk when returns are skewed, heavy-tailed, serially dependent or nonlinear. Square-root-of-time scaling, VaR_T ≈ VaR₁√T, is appropriate only under restrictive conditions such as independent, identically distributed changes with stable variance and linear exposure.

3.3 Historical simulation

Historical simulation revalues the current portfolio under observed historical risk-factor changes. If L₁,…,L_n are simulated losses, their empirical quantile estimates VaR. This method does not require normality, but it assumes that the historical window is informative about current risk. It may omit unprecedented shocks and can respond slowly when old observations leave the sample.

3.4 Monte Carlo simulation

Monte Carlo VaR simulates risk factors from a specified stochastic model and revalues the portfolio in each scenario. It can represent nonlinear instruments and complex dependencies, but results depend on distributional assumptions, calibration, correlation modeling and computational accuracy.

4. Expected Shortfall

Expected Shortfall measures the average loss in the worst 1−α probability mass. A quantile-based definition applicable to discrete and continuous distributions is:

For a continuous loss distribution, this equals E[L | L ≥ VaR_α]. With discrete data, simply averaging every observation equal to or above VaR may include too much probability mass; fractional weighting at the quantile boundary may be necessary.

Worked Example: Empirical ES with Exact Tail Mass

Suppose five equally weighted losses, in ascending order, are:

At 80% confidence, the tail probability is 20%. Each observation has 20% probability, so the worst 20% consists only of the $12,000 observation. Therefore:

Under the left-quantile convention, empirical VaR_0.80 is $8,000. Averaging $8,000 and $12,000 would use 40% of the observations and produce $10,000; that is a threshold-tail average, not the standard 80% ES for this equally weighted sample.

For a normal loss distribution with mean μ_L, standard deviation σ_L and standard-normal density φ, ES is:

This formula applies only when losses are normally distributed. Unlike VaR, ES accounts for loss severity beyond the quantile and is subadditive under standard integrability conditions, making it a coherent risk measure.

5. Backtesting and Model Validation

Backtesting compares ex ante risk forecasts with realized outcomes. A VaR exception occurs when actual loss exceeds forecast VaR. If a correct 99% one-day VaR model is evaluated over 250 independent trading days, the expected number of exceptions is 250 × 0.01 = 2.5.

If exceptions are independent and the exceedance probability is p = 1−α, their count X follows a binomial distribution under the null hypothesis:

The Kupiec unconditional-coverage test assesses whether the observed exception rate is consistent with the intended confidence level. The Christoffersen test additionally examines independence, detecting clustered exceptions that may indicate volatility dynamics or model breakdown. Basel’s traditional 250-day traffic-light framework for 99% VaR classifies 0–4 exceptions as green, 5–9 as yellow and 10 or more as red, although applicable supervisory rules should always be checked.

ES backtesting is more difficult because ES is not a single quantile event. Institutions therefore use joint VaR–ES tests, realized tail-loss comparisons, conditional calibration tests and simulation-based diagnostics.

A complete validation program also examines:

• data quality and missing-price treatment;
• valuation-model accuracy;
• parameter stability and benchmark models;
• stress performance and sensitivity analysis;
• assumptions concerning correlations, liquidity and holding periods;
• model changes, overrides and documentation.

Backtesting should not be manipulated by excluding unfavorable observations or redefining losses after outcomes are known.

6. Stress Testing and Scenario Analytics

VaR and ES are estimated from a probability model; stress testing asks what happens under severe but plausible conditions. The principal methods are:

A scenario should specify shocks to market factors, customer behavior, collateral haircuts, margin calls, funding access and management actions. Correlations observed in normal periods should not automatically be retained: crisis correlations, basis movements and market closures may be more relevant.

Scenario design involves three stages. First, define a macro-financial narrative, such as inflation persistence followed by a rapid rate increase. Second, translate the narrative into risk-factor paths—for example, yield-curve shifts, deposit runoff and spread widening. Third, revalue positions and forecast cash flows, capital and liquidity over time.

Management actions must be feasible. Selling assets that become illiquid, raising unsecured funding during a downgrade, or hedging after markets close are not credible assumptions.

7. Liquidity-Risk Analytics

7.1 Market liquidity

Market liquidity concerns the cost and speed of trading. Common indicators include bid–ask spread, market depth, turnover, price impact, days to liquidate and concentration relative to average daily volume.

For a position size Q and average daily volume ADV, a simple participation-based liquidation estimate is:

Here, k is the permitted participation rate, such as 20%. If Q = 1,000,000 shares, ADV = 250,000 and k = 0.20, estimated liquidation time is 20 trading days. This is a planning approximation, not a law: volume and price impact may deteriorate during stress.

Liquidity-adjusted risk can be estimated by adding bid–ask costs and modeled price impact to market-loss estimates. Care is needed to avoid double counting when stressed price data already contain liquidity effects.

7.2 Funding liquidity and cash-flow ladders

A cash-flow ladder places contractual and behaviorally adjusted inflows and outflows into time buckets. For bucket t, the net cash flow and cumulative position are:

Negative cumulative liquidity indicates a funding requirement. Behavioral assumptions are essential because deposits may be withdrawn before contractual maturity, credit lines may be drawn, and loan repayments may be delayed.

The survival horizon is the first date on which available liquidity becomes negative under a specified stress scenario. It should incorporate usable collateral, operational constraints, currency convertibility and settlement timing.

7.3 Liquidity Coverage Ratio

The Liquidity Coverage Ratio (LCR) is:

High-quality liquid assets (HQLA) are included after regulatory eligibility rules and haircuts. Under the Basel framework, recognized inflows are generally capped at 75% of total expected outflows, subject to jurisdictional rules.

Suppose HQLA equals $120 million, 30-day outflows are $180 million and eligible inflows are $70 million. The inflow cap is 0.75 × $180m = $135m, so all $70 million is recognized. Net outflows are $110 million and:

An LCR above 100% indicates that the regulatory stock of HQLA covers modeled 30-day net outflows. It does not eliminate intraday, currency-specific or longer-horizon risk.

7.4 Net Stable Funding Ratio

The Net Stable Funding Ratio (NSFR) promotes structural funding stability over a one-year horizon:

Available and required stable funding are regulatory weighted amounts, not raw accounting totals. If available stable funding is $520 million and required stable funding is $500 million, the NSFR is 104%.

Other useful indicators include uninsured-deposit concentration, secured-funding dependence, asset encumbrance, collateral headroom, funding maturity concentration and intraday liquidity usage.

8. Asset–Liability Risk Analytics

8.1 Repricing-gap analysis

Rate-sensitive assets (RSA) and rate-sensitive liabilities (RSL) are allocated to buckets according to the next contractual repricing date or expected behavioral repricing date. The gap for bucket t is:

For a small parallel rate change, a simple one-period net-interest-income approximation is:

Here, ∆r is the annualized rate change in decimal form and τ is the fraction of a year for which the repricing effect applies. This approximation assumes equal and immediate pass-through to asset and liability rates and ignores basis risk, caps, floors, customer behavior and new business.

If RSA is $420 million and RSL is $500 million, the gap is −$80 million. For a 100-basis-point increase applying for a full year:

A positive gap generally benefits from rising rates under the simple assumptions; a negative gap generally loses. Actual outcomes may differ because deposit rates and loan rates do not reprice identically.

8.2 Duration-gap analysis

Duration-gap analysis estimates the change in the economic value of equity (EVE). Let:

• A = market value of assets;
• L = market value of liabilities;
• D_A^mod = market-value-weighted modified duration of assets;
• D_L^mod = market-value-weighted modified duration of liabilities.

The modified duration gap is:

For a small parallel yield shift ∆y applied consistently to assets and liabilities:

This formula is valid when the durations are modified durations. If Macaulay durations are used and assets and liabilities share a common yield y with annual compounding, the corresponding approximation is:

If asset and liability yields differ, each side must instead be divided by its own 1+y term; using a single denominator would be inappropriate.

Worked Example: Duration Gap

A bank has assets of $1.0 billion, liabilities of $920 million, asset modified duration of 3.8 years and liability modified duration of 2.1 years:

For a 50-basis-point parallel increase:

The positive duration gap means asset value is more rate-sensitive than liability value; rising rates reduce EVE. The result is a first-order estimate and ignores convexity, nonparallel shifts and changing customer behavior.

Figure: Repricing-gap analysis emphasizes near-term earnings, whereas duration-gap analysis emphasizes the present value of balance-sheet cash flows.

8.3 Optionality and basis risk

Deposits without contractual maturity have uncertain effective durations. Borrowers may prepay loans when rates fall, while depositors may migrate to higher-yield products when rates rise. These embedded options make cash flows state-dependent.

Basis risk arises when asset and liability rates reference different benchmarks or reprice by different amounts. Yield-curve risk arises from nonparallel movements. Institutions therefore supplement simple gaps with dynamic simulation, key-rate duration, stochastic prepayment models and multiple yield-curve scenarios.

9. Illustrative Integrated Case: Meridian Bank

Background

Meridian Bank is a fictional mid-sized retail bank. It has rapidly expanded fixed-rate mortgages funded partly by large, rate-sensitive deposits. Management reports adequate capital and a regulatory LCR slightly above 100%.

Problem

The risk committee is concerned about a rapid rate increase accompanied by deposit withdrawals and wider bond spreads. Separate reports show a positive duration gap, a negative one-year repricing gap and concentrated uninsured deposits, but the interactions have not been evaluated.

Analysis

Meridian has assets of $2.5 billion and liabilities of $2.3 billion. Asset and liability modified durations are 4.4 and 2.0 years respectively. Its duration gap is:

A 150-basis-point parallel rate increase produces the first-order EVE estimate:

The one-year RSA and RSL are $900 million and $1.1 billion. Under equal full pass-through, the −$200 million gap implies an approximate annual NII decline of $3 million after a 150-basis-point increase.

The liquidity stress assumes deposit outflows of $210 million, contractual and contingent outflows of $90 million, and eligible inflows of $60 million over 30 days. Net outflows are $240 million. Available HQLA after regulatory haircuts is $228 million, giving a stressed LCR of 95%.

A further concern is market liquidity. Selling bonds to replace deposits generates spread and price-impact losses, while lower bond values reduce collateral capacity. Thus the EVE, earnings and liquidity effects cannot simply be managed independently.

Approach and Solution

Management adopts a coordinated plan:

1. Reduce duration gap through pay-fixed/receive-floating asset swaps or other appropriately governed hedges.
2. Lengthen funding maturities and reduce dependence on concentrated deposits.
3. Increase unencumbered HQLA and establish operational access to secured funding.
4. Reprice selected loans and deposits without assuming complete customer pass-through.
5. Impose limits for EVE sensitivity, NII sensitivity, deposit concentration and survival horizon.
6. Backtest behavioral assumptions and run combined rate, spread and runoff stresses.

Hedging must consider counterparty exposure, collateral calls, accounting treatment and basis risk. A derivative that reduces EVE sensitivity could increase short-term liquidity needs if variation margin becomes payable.

Outcome

After implementing the plan, the bank’s projected EVE sensitivity is reduced, the stressed LCR rises above the internal buffer, and the survival horizon lengthens. The case demonstrates that regulatory ratio compliance at one date is not a substitute for integrated forward-looking analysis.

Learning Points

• Market and liquidity shocks interact through collateral and forced-sale channels.
• Earnings and economic-value perspectives can signal different vulnerabilities.
• Behavioral assumptions can be as important as contractual maturities.
• Risk mitigation may transfer risk rather than eliminate it.

10. Governance, Regulation and Ethical Practice

Risk measurement requires clear ownership across trading, treasury, finance, ALM, model-risk management and internal audit. Boards should approve risk appetite, while independent risk functions monitor limits and challenge assumptions.

Regulatory frameworks increasingly emphasize tail and liquidity risk. The Basel market-risk framework uses Expected Shortfall in internal-model approaches and applies liquidity-horizon adjustments. Basel liquidity standards include the LCR and NSFR. Interest-rate risk in the banking book is commonly assessed through both EVE and NII perspectives under prescribed and internal shock scenarios. Exact implementation varies by jurisdiction.

Ethical analytics requires truthful data selection, transparent assumptions and timely escalation. Analysts should not suppress stress results, selectively change observation windows, or assume unrealistic management actions to remain within limits. Model outputs should communicate uncertainty rather than present a false sense of precision.

Artificial intelligence and machine learning can improve deposit-decay estimates, anomaly detection, scenario generation and intraday liquidity forecasting. However, nonlinear models may be unstable under regime change and difficult to explain. Human oversight, data governance, fairness testing, explainability and challenger models remain necessary.

11. Limitations and Best Practices

Major limitations include nonstationary data, structural breaks, sparse tail observations, model risk, valuation uncertainty, unstable correlations and unobservable behavioral responses. VaR may conceal tail severity; ES may be estimation-sensitive; gap measures aggregate heterogeneous positions; and liquidity ratios may overlook intraday timing or trapped liquidity.

Sound practice therefore includes:

• using multiple complementary metrics rather than one headline number;
• reconciling market, liquidity, capital and accounting data;
• separating contractual from behavioral cash flows;
• testing parallel and nonparallel yield-curve shocks;
• incorporating optionality, collateral and margin requirements;
• validating models independently and documenting limitations;
• setting management buffers above regulatory minima;
• linking limits to credible contingency-funding and hedging actions.

Chapter Summary

Market-risk sensitivities measure local responses to changes in prices, yields and volatility. Modified duration and DV01 estimate first-order fixed-income exposure, while convexity captures curvature. VaR identifies a loss quantile, whereas ES measures average loss over an exact tail probability and therefore reflects tail severity. Backtesting evaluates forecast calibration and exception independence, while stress testing examines severe scenarios not adequately represented by routine probability models.

Liquidity analytics distinguishes the cost of trading from the ability to meet cash obligations. Cash-flow ladders, survival horizons, LCR, NSFR and market-depth measures reveal different aspects of liquidity. ALM uses repricing gaps to estimate earnings sensitivity and modified duration gaps to estimate economic-value sensitivity. Because market, funding and balance-sheet risks interact, institutions need integrated scenarios, robust governance, transparent assumptions and feasible management responses.

Key Terms

Asset–liability management: Coordinated management of balance-sheet earnings, value, funding and liquidity risks.

Backtesting: Comparison of model forecasts with realized outcomes.

Basis risk: Risk arising when related rates or prices do not move together.

Convexity: Second-order sensitivity of a bond’s price to yield changes.

Delta: First-order sensitivity of value to a risk-factor change.

Duration gap: Difference between asset duration and liability duration adjusted for the liability-to-asset ratio.

DV01: Approximate currency-value change from a one-basis-point yield movement.

Economic value of equity: Present value of asset cash flows minus the present value of liability cash flows.

Expected Shortfall: Average loss within the worst specified probability mass.

Funding liquidity risk: Risk of being unable to meet payment obligations when due.

High-quality liquid assets: Regulatory-eligible assets expected to remain liquid during stress.

Liquidity Coverage Ratio: HQLA divided by modeled 30-day net cash outflows.

Market liquidity risk: Risk that a trade cannot be executed promptly without material price impact.

Modified duration: Approximate percentage price sensitivity to a unit change in yield.

Net Stable Funding Ratio: Available stable funding divided by required stable funding.

Repricing gap: Rate-sensitive assets minus rate-sensitive liabilities within a time bucket.

Stress testing: Evaluation of financial effects under severe scenarios.

Survival horizon: Time until available liquidity becomes negative under a specified stress.

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

Review Questions

Short Answer Questions

1. Distinguish market liquidity risk from funding liquidity risk.
2. What information must accompany a reported VaR number?
3. Why does ES provide information that VaR does not?
4. What does a negative repricing gap imply under a simple rising-rate scenario?
5. Why are non-maturity deposits difficult to model in ALM?

Long Answer Questions

1. Compare parametric, historical-simulation and Monte Carlo approaches to VaR.
2. Explain the roles of Kupiec and Christoffersen tests in VaR backtesting.
3. Discuss how stress testing should incorporate market, collateral and funding interactions.
4. Compare the earnings and economic-value perspectives of interest-rate risk.
5. Evaluate the benefits and limitations of LCR and NSFR as liquidity measures.

Analytical / Application Questions

1. A $25 million bond portfolio has modified duration of 5.4 years and convexity of 38 years². Estimate its value change after a 75-basis-point yield increase using duration and convexity.
2. Construct a 30-day cash-flow ladder for a bank using assumed inflows, deposit withdrawals, wholesale maturities, collateral calls and available liquidity. Estimate its survival horizon.
3. A bank has assets of $4 billion, liabilities of $3.7 billion, asset modified duration of 4.8 years and liability modified duration of 2.6 years. Calculate its modified duration gap and approximate EVE change after a 100-basis-point parallel increase. Discuss at least three limitations of the estimate.

Multiple Choice Questions with Answers

Question 1

A bond portfolio is worth $10 million and has modified duration of 4.2 years. What is its approximate DV01?

A. $420

B. $4,200

C. $42,000

D. $420,000

Correct Answer: B. $4,200

Question 2

Five equally weighted losses in ascending order are −$2,000, $1,000, $3,000, $8,000 and $12,000. Using exact empirical tail mass, what is 80% Expected Shortfall?

A. $12,000

B. $10,000

C. $8,000

D. $4,400

Correct Answer: A. $12,000

Question 3

A bank has $120 million of HQLA, expected 30-day outflows of $180 million and eligible inflows of $70 million. What is its LCR?

A. 66.7%

B. 100.0%

C. 109.1%

D. 171.4%

Correct Answer: C. 109.1%

Question 4

Rate-sensitive assets are $420 million and rate-sensitive liabilities are $500 million. Under equal full pass-through, what is the approximate full-year NII effect of a 100-basis-point rate increase?

A. A $0.8 million increase

B. A $0.8 million decrease

C. A $8 million increase

D. No change

Correct Answer: B. A $0.8 million decrease

Question 5

A bank has assets of $1 billion, liabilities of $920 million, asset modified duration of 3.8 years and liability modified duration of 2.1 years. What is the approximate EVE effect of a 50-basis-point parallel rate increase?

A. A $9.34 million increase

B. A $1.87 million decrease

C. A $9.34 million decrease

D. A $18.68 million decrease

Correct Answer: C. A $9.34 million decrease

Question 6

Which statement correctly describes a principal limitation of Value at Risk?

A. It does not indicate the severity of losses beyond its quantile.

B. It always assumes normally distributed losses.

C. It cannot be calculated from historical data.

D. It is identical to the maximum possible loss.

Correct Answer: A. It does not indicate the severity of losses beyond its quantile.

Question 7

Which backtesting procedure specifically examines whether VaR exceptions are clustered over time?

A. Duration-gap analysis

B. Christoffersen independence test

C. Liquidity Coverage Ratio

D. Historical scenario replay

Correct Answer: B. Christoffersen independence test

Question 8

Which measure most directly indicates market liquidity risk?

A. Bid–ask spread

B. Modified duration

C. Liability-to-asset ratio

D. Net interest margin

Correct Answer: A. Bid–ask spread

Question 9

Under the simple repricing-gap model, how does a positive gap generally affect NII when rates rise?

A. NII decreases because liabilities exceed assets.

B. NII remains unchanged in every case.

C. NII generally increases under equal rate pass-through.

D. EVE necessarily increases by the same amount.

Correct Answer: C. NII generally increases under equal rate pass-through.

Question 10

Why might an interest-rate derivative hedge that reduces EVE risk increase funding-liquidity risk?

A. It eliminates all collateral requirements.

B. It can generate variation-margin cash outflows.

C. It prevents market prices from changing.

D. It converts every liability into equity.

Correct Answer: B. It can generate variation-margin cash outflows.