Stress Testing, Scenario Analysis and Crisis Management

Introduction and strategic context

Stress testing is the disciplined examination of how a bank’s financial position, risk profile and capacity to continue operating would change under unusually adverse, but plausible, conditions. Scenario analysis provides the narrative and quantitative assumptions for that examination. Crisis management converts the results into decisions before, during and after a period of financial distress.

These activities are strategically important because historical risk measures are not sufficient when relationships change rapidly. A bank may appear well capitalised under current conditions but become vulnerable when unemployment rises, property prices fall, funding markets close and depositors withdraw cash simultaneously. Enterprise-wide stress testing therefore connects credit, market, operational, liquidity, interest-rate and strategic risks rather than analysing each risk in isolation.

A credible programme has four purposes:

1. to identify vulnerabilities and concentrations;
2. to estimate losses, capital depletion and liquidity needs;
3. to test the adequacy of risk appetite, limits and contingency plans; and
4. to support capital planning, recovery planning, supervision and board decisions.

Stress testing is not a prediction of the future. It is a conditional analysis: if specified assumptions occur, what consequences follow, and what management actions remain available?

Learning objectives

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

• distinguish sensitivity analysis, scenario analysis, stress testing and reverse stress testing;
• design an enterprise-wide stress-testing framework;
• translate macroeconomic scenarios into credit, market, operational, capital and liquidity impacts;
• evaluate second-round effects, management actions and model uncertainty;
• interpret stress results for limits, capital planning and risk appetite;
• formulate recovery, resolution and crisis-communication arrangements; and
• assess the systemic implications of bank distress.

1. Core concepts and terminology

A sensitivity test changes one variable, such as the policy rate or default rate, while holding most other assumptions constant. It is useful for isolating exposure but may be unrealistic when economic variables move together.

A scenario analysis changes several variables according to a coherent economic narrative. A recession scenario might combine lower output, higher unemployment, falling collateral values, wider credit spreads and deposit outflows.

A stress test applies a severe scenario to financial statements, risk models and management processes to assess resilience. It normally includes projected profit and loss, balance-sheet changes, risk-weighted assets, capital ratios and liquidity metrics.

A reverse stress test begins with a defined failure or near-failure outcome—such as breaching a minimum capital ratio or exhausting a survival liquidity horizon—and asks which combinations of events could produce it. Reverse tests are especially useful for identifying hidden concentrations and designing recovery triggers.

A baseline scenario represents the institution’s central planning view. An adverse scenario is materially worse than baseline but plausible. An extreme scenario is highly severe and may have low probability, but it remains useful for examining the outer boundary of resilience.

Top-down stress testing uses portfolio-level data and supervisory or macroeconomic models. Bottom-up stress testing requires business units to project exposures, losses and management responses. Effective programmes reconcile both approaches.

2. The architecture of an enterprise-wide stress test

2.1 Governance and accountability

The board approves the risk appetite and challenges the severity, assumptions and implications of scenarios. Senior management owns the programme and ensures that results influence decisions. The chief risk officer normally coordinates methodology, while finance, treasury, economics, business units, operations, compliance and technology provide specialist input. Internal audit independently assesses governance, data, models and controls.

A written stress-testing policy should specify:

• objectives, scope and minimum frequency;
• scenario types and severity standards;
• responsibilities and approval rights;
• data and model requirements;
• treatment of management actions;
• escalation thresholds;
• reporting to the board and supervisors; and
• documentation and independent validation.

Stress testing should be integrated with the internal capital adequacy assessment process, liquidity adequacy assessment, recovery plan, budget, risk appetite and resolution planning. A test that produces a report but does not alter limits or decisions is largely a compliance exercise.

2.2 Scope and risk taxonomy

The perimeter should cover all material legal entities, jurisdictions, products and risk types. Important channels include:

• credit risk: probability of default, loss given default, exposure at default, provisions and migration;
• market risk: interest rates, foreign exchange, equity prices, commodities, credit spreads and valuation adjustments;
• interest-rate risk in the banking book: repricing, duration, deposit behaviour and net interest income;
• liquidity risk: deposit withdrawals, wholesale funding rollover, collateral calls and asset-sale discounts;
• operational risk: fraud, cyber incidents, outages, misconduct, legal costs and business interruption;
• strategic risk: declining volumes, margin compression, reputational damage and ineffective responses; and
• capital and leverage risk: changes in earnings, provisions, risk-weighted assets and available capital.

The test must capture interactions. For example, a property-price decline can increase defaults, reduce collateral recovery, raise provisions, increase risk weights, trigger margin calls and weaken market confidence.

2.3 Scenario construction

A scenario has two connected elements: a quantitative path and a qualitative narrative. Typical variables include real GDP growth, unemployment, inflation, policy and market interest rates, exchange rates, property prices, equity prices, credit spreads, commodity prices and funding conditions.

Scenario design should satisfy four criteria:

1. Internal consistency: variables should describe the same economic story.
2. Materiality: the scenario should affect the bank’s important exposures.
3. Plausibility: assumptions should be severe but defensible.
4. Transparency: users should understand how assumptions were selected.

Historical episodes provide evidence, but a purely historical scenario may omit new vulnerabilities. Hypothetical scenarios can examine climate transition, cyber disruption, geopolitical fragmentation or a sudden loss of confidence. A useful suite combines historical, hypothetical, idiosyncratic, market-wide and reverse stresses.

2.4 Translating macroeconomic variables into risk drivers

The translation process usually has four stages:

1. define the macroeconomic path;
2. estimate changes in risk drivers;
3. calculate losses, income, capital and liquidity effects; and
4. aggregate results while preventing double counting.

For a credit portfolio, expected loss is commonly represented as:

where EL is expected loss, PD is probability of default, LGD is loss given default and EAD is exposure at default. If a portfolio contains different facilities, the aggregate expected loss is:

The expression assumes that the relevant parameters are consistently defined over the same horizon and that expected loss is measured before any separate provision or tax treatment. It is not a complete measure of unexpected loss, capital requirement or economic loss.

A scenario may increase PD through unemployment and lower house prices, increase LGD through weaker collateral recoveries, and increase EAD through drawn credit lines. Correlation and concentration effects must be considered because defaults may become highly dependent during a recession.

For a simplified net-interest-income projection:

This identity is valid for the defined accounting period, but a realistic stress test also models repricing lags, floors, hedges, non-performing assets, deposit beta, behavioural maturities and volume changes.

Capital ratios are often expressed as:

If starting common equity tier 1 capital is 8.0 billion monetary units, stressed losses reduce capital by 1.5 billion, and risk-weighted assets rise from 60.0 billion to 66.0 billion, then:

This calculation assumes that losses directly reduce CET1 capital and that no tax, dividend, issuance, valuation or regulatory adjustment is separately included. In practice, capital projections must incorporate retained earnings, deferred tax effects, distributions, capital actions, regulatory filters and changes in risk weights.

Liquidity analysis distinguishes contractual cash flows from behavioural and stressed cash flows. A simplified net liquidity gap is:

A negative gap does not automatically imply failure if the institution holds eligible liquid assets or has reliable contingent funding. Conversely, positive contractual cash flow may be unusable if assets are encumbered or inflows are uncertain. Stress tests therefore examine liquidity buffers, monetisable collateral, haircuts, intraday needs, central-bank eligibility and funding concentration.

3. From scenario to enterprise impact

3.1 Credit losses and provisioning

Credit stress models may use transition matrices, satellite models, cash-flow analysis, expert overlays or combinations of these methods. A satellite model links macroeconomic variables to portfolio risk drivers. For example, a mortgage model may relate default rates to unemployment, interest-service burdens and property prices. Corporate analysis may additionally consider sector sales, leverage and refinancing needs.

Provisioning projections should distinguish performing assets, significant increases in credit risk and defaulted exposures. A stress scenario may also cause modelled provisions to rise more slowly than actual losses if data are stale or collateral assumptions are optimistic. Overlays may be justified, but they require evidence, approval and subsequent back-testing.

3.2 Market and interest-rate effects

Market stress should revalue trading and investment positions using shocked prices, curves, spreads, volatilities and correlations. It should include basis risk, liquidity discounts, valuation uncertainty and counterparty effects. A parallel rate shock is only one possibility; steepening, flattening, inversion and foreign-currency basis shocks may be more relevant.

For banking-book interest-rate risk, the institution should examine both earnings and economic value. A simplified duration relationship is:

where ∆P is the approximate price change, D is modified duration in years, P is current price or market value, and ∆y is the change in yield expressed as a decimal. This linear approximation assumes a relatively small yield change and omits convexity. It is therefore inappropriate as the sole method for large shocks or instruments with significant optionality.

3.3 Liquidity and funding stress

Liquidity scenarios should separate institution-specific events from market-wide events. An idiosyncratic event may involve reputational damage and rapid uninsured-deposit withdrawals. A market-wide event may close wholesale markets and reduce the liquidity value of securities. The combined scenario is often the most dangerous because the bank loses funding precisely when asset-sale discounts increase.

Important assumptions include withdrawal rates by depositor type, rollover rates, collateral haircuts, margin calls, intraday settlement needs, committed facilities, central-bank access and the speed of management actions. Liquidity stress should be tested daily or more frequently over the early crisis horizon, rather than only at a monthly reporting date.

3.4 Operational, legal and reputational effects

Operational stress is not merely an additional loss percentage. A cyberattack may prevent payments, generate regulatory penalties, create customer withdrawals and impair the bank’s ability to execute recovery actions. A model should therefore combine direct financial losses with service disruption, remediation costs, legal exposure and reputational effects.

4. Reverse stress testing and decision use

Reverse stress testing begins by defining a failure state. Examples include falling below a capital threshold, exhausting the liquidity buffer, losing access to secured funding or becoming unable to deliver critical payment services. The analyst then searches for combinations of shocks that reach that state.

A reverse test may reveal that no single risk is fatal, but a combination of moderate events is. For example, a 12% fall in property prices, a 7% deposit outflow and a two-notch deterioration in wholesale funding may breach recovery triggers when occurring together. The result should not be interpreted as a forecast. Its value lies in identifying concentrations, dependencies and early-warning indicators.

Stress results should lead to concrete decisions:

• reduce sector, geography or counterparty limits;
• revise underwriting standards and collateral margins;
• increase liquid assets or lengthen funding maturity;
• retain earnings or raise capital;
• diversify deposits and wholesale funding;
• pre-position collateral and operational capacity;
• amend recovery triggers; and
• improve data, models and crisis communication.

Management actions must be classified as automatic, feasible but uncertain, or unavailable under stress. Double counting is a common error: a test should not assume both lower lending volumes and a large asset sale if the latter depends on maintaining lending capacity.

5. Crisis management, recovery and resolution

5.1 Crisis-management structure

A crisis-management framework defines escalation levels, decision rights, information flows and communication protocols. A crisis committee typically includes the chief executive, risk, treasury, finance, operations, legal, communications, technology and relevant business leaders. It should operate from a tested playbook with deputies and succession arrangements.

Early-warning indicators may include abnormal deposit outflows, rising collateral calls, declining liquidity ratios, market-price deterioration, payment failures, cyber alerts, social-media activity and adverse media coverage. Indicators should have thresholds, owners and prescribed actions.

5.2 Recovery planning

Recovery planning concerns actions that restore viability while the bank remains a going concern. A recovery plan should identify:

• credible recovery options;
• trigger indicators and escalation processes;
• estimated execution time;
• legal, operational and market constraints;
• dependencies between options;
• communication requirements; and
• minimum information needed for decision-making.

Options may include asset sales, liability management, capital raising, dividend suspension, expense reduction, collateral mobilisation, business disposals and emergency funding. Credibility depends on execution speed, market capacity and the absence of mutually conflicting assumptions.

5.3 Resolution planning

Resolution applies when recovery is no longer credible or the institution is failing or likely to fail. Its objective is generally to preserve critical economic functions while allocating losses according to the applicable legal framework and avoiding disorderly contagion. Resolution planning examines separability of businesses, continuity of access to payment systems, valuation, operational service agreements, data availability and loss-absorbing capacity.

Recovery and resolution should not be confused. Recovery seeks to restore the bank; resolution manages failure. A bank can have a strong recovery plan but still require resolution readiness because some crises evolve too quickly for private recovery actions.

5.4 Crisis communication

Communication is a risk-control activity, not merely a public-relations function. Messages should be accurate, timely, coordinated and appropriate to each audience: employees, depositors, investors, regulators, counterparties and the media. The bank should identify authorised spokespeople, prepare holding statements and establish verification procedures.

Communication must avoid misleading reassurance, premature disclosure of uncertain information and inconsistent statements across jurisdictions. Confidential supervisory or recovery information must be handled according to law. Silence can intensify rumours, but uncontrolled disclosure can also accelerate withdrawals.

6. Systemic risk and macroprudential implications

A bank’s stress test may understate damage if it ignores the financial system. Simultaneous deleveraging can depress asset prices, reduce credit supply and amplify losses across institutions. Common exposures, interconnected derivatives, payment systems, correspondent banking and shared funding markets create contagion channels.

System-wide analysis therefore considers:

• common borrower and sector exposures;
• correlated liquidity outflows;
• fire-sale externalities;
• network contagion;
• procyclical margin and collateral requirements; and
• feedback between sovereign and banking risk.

Macroprudential authorities may use countercyclical buffers, sectoral capital requirements, loan-to-value limits, liquidity measures or supervisory guidance in response to system-wide vulnerabilities. Individual-bank resilience and financial-stability resilience are related but not identical objectives.

7. Data, models, technology and model risk

Stress testing requires granular, timely and reconciled data. Essential attributes include facility identifiers, obligor sector, geography, collateral, maturity, currency, repricing date, counterparty, legal entity and encumbrance status. Data lineage should show how source records become reported results.

Model risk arises from incorrect specification, poor data, unstable relationships, parameter uncertainty and inappropriate use. Models calibrated on calm periods may fail during structural breaks. Independent validation should assess conceptual soundness, performance, sensitivity, limitations, implementation and governance.

Technology can improve scenario generation, intraday liquidity monitoring, natural-language extraction of risk information and aggregation across entities. Machine-learning models may identify nonlinear relationships, but explainability, stability, bias, data drift and governance remain essential. Artificial intelligence should support—not replace—expert challenge and accountable decisions. Synthetic scenarios generated by AI must be economically coherent, documented and subject to human validation.

8. Illustrative integrated case: Meridian Bank

Background

Meridian Bank is a fictional mid-sized bank with a large residential-mortgage portfolio, commercial real-estate lending and moderate dependence on wholesale funding. Its starting CET1 capital is 8.0 billion monetary units and starting risk-weighted assets are 60.0 billion.

Problem

The board requests an enterprise-wide stress test involving a recession, falling property prices, higher unemployment, wider credit spreads and deposit outflows. Management wants to know whether existing limits and recovery triggers remain credible.

Analysis

Under the adverse scenario, credit losses are estimated at 1.5 billion. Net interest income and fee income fall by 0.6 billion, while operating and restructuring costs rise by 0.3 billion. Before tax, the total earnings impact is therefore:

For simplicity, assume the bank has no offsetting tax benefit, dividend payment or capital raising during the stress horizon. CET1 capital falls from 8.0 billion to 5.6 billion. Risk-weighted assets rise to 66.0 billion because of credit deterioration and portfolio migration.

The liquidity analysis identifies a stressed 4.2 billion cash outflow over the first month, against 3.5 billion of immediately available liquid assets. However, 1.0 billion of high-quality collateral can be mobilised subject to a 20% haircut, producing 0.8 billion of usable funding. The remaining shortfall is:

Thus, the bank faces a 0.1 billion residual shortfall under the stated assumptions. This result depends critically on collateral eligibility and timely mobilisation; it is not evidence that funding is guaranteed.

The reverse stress test shows that a simultaneous 10% uninsured-deposit run, closure of unsecured wholesale markets and a further 15% collateral haircut would breach the bank’s internal liquidity recovery trigger. The test also shows that commercial real-estate concentration, rather than the mortgage portfolio, is the principal driver of capital vulnerability.

Approach and solution

The board approves a reduction in commercial-real-estate origination limits, a higher liquidity buffer, earlier escalation for deposit outflows and pre-positioning of collateral. Treasury conducts a mobilisation exercise, while operations tests payment continuity. Finance revises the capital plan to retain earnings and suspend discretionary distributions if the trigger is reached. Communications prepares separate depositor, investor and employee messages.

Outcome and learning points

The test does not predict a crisis. It demonstrates that capital remains positive but liquidity is more time-critical than solvency. It also identifies a concentration that aggregate capital ratios obscure. The principal learning points are that enterprise-wide effects matter, liquidity actions must be operationally tested, and reverse stress testing can reveal combinations of events that ordinary scenarios overlook.

9. Limitations, ethics and best practice

Stress tests are conditional, model-dependent and subject to uncertainty. They may omit unknown risks, underestimate behavioural responses, rely on stale data or assume management actions that are politically or operationally impossible. Scenario severity can also be manipulated to produce comfortable results.

Ethically, institutions must not conceal material weaknesses, selectively report favourable scenarios or use stress testing to justify predetermined decisions. Boards should challenge assumptions, require disclosure of uncertainty and ensure that vulnerable customers are considered when recovery actions affect credit access, fees or branch services.

Best practice includes:

• combining historical, hypothetical and reverse scenarios;
• modelling capital and liquidity jointly;
• testing legal entities and critical operations;
• challenging correlations and management actions;
• using sensitivity ranges rather than false precision;
• documenting data lineage and model limitations;
• conducting regular dry runs and crisis simulations;
• linking results to limits, capital and funding decisions; and
• reporting both numerical outcomes and the conditions under which they fail.

Chapter Summary

Stress testing is an enterprise-wide resilience tool that examines the consequences of adverse but plausible conditions. Sensitivity analysis isolates variables, scenario analysis creates coherent narratives, stress testing evaluates institutional consequences and reverse stress testing works backward from failure. A robust framework integrates credit, market, interest-rate, liquidity, operational, strategic and capital risks.

Macroeconomic assumptions must be translated carefully into probability of default, loss given default, exposure, earnings, valuation, funding and capital effects. Results are useful only when data, models, correlations and management actions are credible. Stress testing should influence limits, capital planning, liquidity buffers, recovery triggers and operational preparation.

Crisis management requires early-warning indicators, clear governance, recovery options, resolution readiness and disciplined communication. Finally, systemic analysis is essential because correlated exposures, fire sales and funding contagion can make the total impact greater than the sum of individual bank losses.

Key Terms

Stress testing: Conditional assessment of resilience under adverse assumptions.

Scenario analysis: Analysis using a coherent set of economic or institutional assumptions.

Sensitivity analysis: Test that changes one principal variable while holding others broadly constant.

Reverse stress test: Analysis beginning with a failure outcome and identifying causes.

Baseline scenario: Central planning or expected economic path.

Adverse scenario: Severe but plausible deterioration from baseline.

Risk driver: Variable through which conditions affect exposure or loss.

Probability of default: Likelihood that an obligor defaults over a specified horizon.

Loss given default: Percentage of exposure lost after recovery and collateral realisation.

Exposure at default: Expected amount outstanding when default occurs.

Risk-weighted assets: Regulatory measure of exposures adjusted for risk.

CET1 ratio: Common equity tier 1 capital divided by risk-weighted assets.

Liquidity gap: Difference between available inflows and expected outflows over a period.

Recovery plan: Arrangements for restoring viability while the bank remains a going concern.

Resolution: Formal management of a failing bank to preserve critical functions and contain disruption.

Management action: Decision intended to reduce losses or restore resilience under stress.

Model risk: Risk of adverse consequences from incorrect or misused models.

Systemic risk: Risk that distress spreads across the financial system or impairs its functioning.

Fire sale: Rapid asset disposal at a price materially below fundamental value.

Crisis communication: Controlled provision of accurate and timely information during distress.

Review Questions

Short Answer Questions

1. Distinguish stress testing from reverse stress testing.
2. What are the three components of the simplified expected-loss expression PD × LGD × EAD?
3. Why should liquidity stress testing consider collateral haircuts?
4. State two responsibilities of the board in a stress-testing programme.
5. Explain why a historical scenario may be insufficient on its own.

Long Answer Questions

1. Design an enterprise-wide stress-testing governance framework for a multinational bank.
2. Explain how a recession scenario can affect credit losses, capital and liquidity simultaneously.
3. Critically evaluate the use of management actions in stress-test projections.
4. Compare recovery planning with resolution planning.
5. Discuss the role of stress testing in risk appetite, capital planning and regulatory supervision.

Analytical / Application Questions

1. A bank has CET1 capital of 10 billion, risk-weighted assets of 80 billion and stressed losses of 2 billion. Risk-weighted assets increase to 90 billion. Calculate the simplified stressed CET1 ratio and interpret the result.
2. Construct a reverse stress test for a bank dependent on uninsured deposits and short-term wholesale funding. Identify at least four indicators and three possible recovery actions.
3. A bank’s credit losses are moderate, but deposit outflows and collateral haircuts create a liquidity shortfall. Explain why the bank may face crisis before becoming balance-sheet insolvent, and propose an integrated response.

Multiple Choice Questions with Answers

Question 1

Which statement best describes a reverse stress test?

A. It changes only one macroeconomic variable.

B. It begins with a failure outcome and identifies combinations of events that could cause it.

C. It estimates the most likely baseline forecast.

D. It measures only historical trading losses.

Correct Answer: B. It begins with a failure outcome and identifies combinations of events that could cause it.

Question 2

A portfolio has PD of 2%, LGD of 40% and EAD of 50 million monetary units. Using EL = PD × LGD × EAD, what is expected loss?

A. 0.04 million

B. 0.2 million

C. 0.4 million

D. 4.0 million

Correct Answer: C. 0.4 million

Question 3

If CET1 capital is 6 billion and risk-weighted assets are 75 billion, what is the simplified CET1 ratio?

A. 6%

B. 8%

C. 10%

D. 12.5%

Correct Answer: B. 8%

Question 4

Which is most important when constructing a coherent macroeconomic stress scenario?

A. Ensuring that all variables move independently

B. Using only the most recent quarter’s data

C. Ensuring that quantitative assumptions support a consistent economic narrative

D. Avoiding hypothetical scenarios entirely

Correct Answer: C. Ensuring that quantitative assumptions support a consistent economic narrative

Question 5

Which action is primarily a recovery action?

A. Liquidating the bank through a resolution authority

B. Raising capital while the bank remains a viable going concern

C. Ignoring early-warning indicators

D. Removing all liquidity reporting

Correct Answer: B. Raising capital while the bank remains a viable going concern

Question 6

Why can a bank experience a liquidity crisis before balance-sheet insolvency?

A. Liquidity depends on immediate funding and asset monetisation, not only accounting net worth.

B. Solvency ratios always exclude capital.

C. Liquidity crises occur only when profits are positive.

D. A solvent bank cannot lose deposits.

Correct Answer: A. Liquidity depends on immediate funding and asset monetisation, not only accounting net worth.

Question 7

Which limitation is associated with the duration approximation ∆P ≈ −D × P × ∆y?

A. It is exact for all yield changes and all securities.

B. It ignores the relationship between price and yield.

C. It is a linear approximation and may be inaccurate for large shocks or significant optionality.

D. It can be used only for ordinary shares.

Correct Answer: C. It is a linear approximation and may be inaccurate for large shocks or significant optionality.

Question 8

Which governance practice most strengthens stress-testing credibility?

A. Allowing business units to approve their own unchallenged assumptions

B. Linking results to limits, capital planning and board challenge

C. Reporting only the most favourable scenario

D. Excluding operational risk from enterprise analysis

Correct Answer: B. Linking results to limits, capital planning and board challenge

Question 9

What is a systemic-risk channel in a banking crisis?

A. A bank’s private employee appraisal process

B. A fire sale that depresses asset prices for multiple institutions

C. A routine change in branch opening hours

D. A single unaffected customer account

Correct Answer: B. A fire sale that depresses asset prices for multiple institutions

Question 10

Which statement about crisis communication is most appropriate?

A. It should be accurate, coordinated, timely and appropriate to each audience.

B. It should always disclose unverified rumours immediately.

C. It is unrelated to liquidity risk.

D. Only marketing staff should control it.

Correct Answer: A. It should be accurate, coordinated, timely and appropriate to each audience.