Stress Testing, Recovery and Crisis Management

Introduction and context

Banks are exposed to interacting risks rather than isolated events. A recession can increase defaults, reduce collateral values, widen funding spreads, weaken fee income and create deposit outflows at the same time. Stress testing provides a disciplined way to examine such combinations before they occur. Recovery and crisis-management arrangements then convert analytical results into decisions, actions and communications.

Stress testing is not a prediction of the most likely future. It is an analysis of how a bank’s financial position, risk profile and viability would respond to specified adverse assumptions. A sound programme combines quantitative modelling, expert judgement, governance and operational testing.

Learning objectives

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

1. Distinguish sensitivity, scenario, reverse stress and capital-planning tests.
2. Construct severe but plausible scenarios and identify transmission channels.
3. Calculate and interpret effects on capital, liquidity, earnings and solvency.
4. Design recovery, resolution, capital-contingency and crisis-communication plans.
5. Evaluate model, governance, behavioural and ethical limitations.
6. Make documented decisions under uncertainty and time pressure.

1. The architecture of stress testing

1.1 Sensitivity tests

A sensitivity test changes one risk driver, or a small number of closely related drivers, while holding other assumptions constant. Examples include a 200-basis-point increase in interest rates, a 10% fall in property prices or a specified deposit outflow. Sensitivities are useful for identifying exposure concentrations and nonlinear thresholds, but they can underestimate danger because real crises involve correlated changes.

A simple loss sensitivity can be represented as:

where ∆P is the change in profit or capital, ∆x is the change in a risk factor and β is the estimated exposure coefficient. This linear form is an approximation. It is inappropriate where defaults, margin calls, option positions or liquidity constraints create nonlinear effects.

1.2 Scenario analysis

Scenario analysis changes several variables together according to a coherent economic or operational narrative. A recession scenario might combine GDP contraction, unemployment, house-price declines, higher default rates, market spread widening and deposit withdrawals. Scenarios may be:

• Historical, reproducing or adapting a past crisis;
• Hypothetical, built around a plausible future shock;
• Macroeconomic, linking economic variables to portfolio outcomes;
• Reverse, starting with failure or a breached limit and working backwards.

The scenario must specify its horizon, frequency, starting balance sheet, management actions and whether assumptions are static or dynamic. Without these definitions, two apparently different results may not be comparable.

1.3 Reverse stress testing

A reverse stress test asks: what combination of events would make the institution breach a critical objective? The objective could be a minimum CET1 ratio, liquidity survival horizon, risk appetite limit or operational viability condition. It is especially useful for discovering vulnerabilities that ordinary forecasting overlooks.

Reverse stress testing does not claim that the identified pathway is the only route to failure. It identifies at least one sufficiently severe pathway and tests whether preventive controls or recovery options interrupt it.

1.4 Integrated stress testing

An integrated test links solvency, liquidity, market, operational and reputational effects. For example, losses reduce capital; a lower credit rating increases wholesale funding costs; uncertainty accelerates deposit outflows; asset sales crystallise valuation losses; and management actions may have legal or market consequences.

A useful distinction is:

2. Designing a credible stress test

2.1 Define purpose, scope and risk appetite

The test should begin with a decision question: for example, “Can the bank remain above its internal capital target after a two-year recession and a 15% deposit outflow?” The scope should identify legal entities, portfolios, currencies, material risks, time horizon and reporting frequency.

Risk appetite translates strategy into boundaries. It may include a minimum capital ratio, maximum concentration, liquidity survival period, tolerance for service interruption and escalation triggers. Regulatory minima are constraints, not necessarily adequate management targets. Buffers are needed because estimates are uncertain and recovery actions take time.

2.2 Build the scenario narrative

A credible scenario has an initiating event, propagation mechanism, duration and endpoint. Consider:

1. Initial shock: for example, a sovereign repricing or cyber incident.
2. First-round effects: market losses, defaults or payment disruption.
3. Second-round effects: funding withdrawal, collateral calls or rating downgrade.
4. Management response: asset sales, pricing changes, capital raising or restrictions.
5. Feedback: whether actions reduce or intensify the stress.

“Severe but plausible” is a judgement, not a precise probability label. Plausibility should be supported by historical ranges, expert evidence, scenario workshops and sensitivity analysis around disputed assumptions.

2.3 Translate drivers into financial outcomes

Credit losses can be approximated using exposure at default, probability of default and loss given default:

EAD is the exposure at default, PD is the probability of default over the selected horizon and LGD is the proportion of exposure lost after recoveries and collateral. The formula assumes compatible horizons and definitions. It is an expected-loss approximation, not a complete stress-loss model: default correlation, migration, discounting, drawn commitments and model uncertainty may need separate treatment.

For a portfolio divided into segments i:

Capital ratios are generally expressed as:

If a shock changes both numerator and denominator, the post-stress ratio must be recalculated rather than estimated from the capital change alone. For example, if CET1 falls from £500 million to £380 million and RWA rises from £4,000 million to £4,200 million:

A liquidity cash-flow gap over a period is commonly represented as:

The terms must be defined consistently. “Monetisable funding” should include only resources that are legally available, operationally accessible, unencumbered or eligible, and usable within the relevant time period. A central-bank facility that is theoretically available but requires collateral, approval or time is not equivalent to immediately deployable cash.

2.4 Model management actions explicitly

Possible actions include reducing lending, retaining earnings, raising equity, issuing secured funding, selling liquid assets, using central-bank facilities, repricing products and restricting dividends. Each action needs assumptions about execution time, capacity, legal authority, market impact, collateral and effect on customers.

Actions must not be counted twice. For example, the same securities cannot simultaneously support a liquidity buffer, collateralise emergency borrowing and be sold for cash. Haircuts, transaction costs and price impact should be applied where relevant.

3. Interpreting results

Stress results should be read as ranges and decision signals, not as precise forecasts. Key questions include:

• Which portfolios generate the largest losses?
• Which assumptions drive the result?
• Does the bank breach regulatory requirements, internal limits or recovery triggers?
• How quickly does liquidity deteriorate?
• Are management actions feasible under the same conditions?
• Are effects concentrated in one legal entity, currency or operational service?

A decomposition improves interpretation:

This is an analytical bridge, not a universal accounting identity. Items must be defined so that they do not overlap, and tax, valuation adjustments and regulatory deductions may require separate treatment.

Threshold analysis is valuable. If the capital buffer above a management floor is £120 million, a test producing £90 million of losses does not automatically imply failure; the outcome depends on earnings, RWA movement, deductions and available actions. Conversely, a capital surplus may coexist with a liquidity crisis because solvency and liquidity are related but distinct conditions.

4. Worked reverse-stress application: Northbridge Bank

The following is an illustrative case, not a real institution.

Background

Northbridge Bank has £500 million of CET1 capital and £4,000 million of RWA, giving an initial CET1 ratio of 12.5%. Its board defines £240 million of CET1 as a critical capital floor for the purpose of a reverse stress test. The bank also maintains liquid assets, but not all are immediately usable.

Its liquidity inventory is:

Thus, the case deliberately distinguishes the £900 million total liquid-asset stock from £500 million immediately available liquidity. The £400 million difference is not counted as immediate cash support.

Problem

Management wants to know what additional losses, beyond an initial recession shock, would reduce CET1 to the £240 million critical floor, and whether a plausible liquidity event could prevent timely recovery actions.

Initial shock

The first scenario produces:

• £110 million of credit losses;
• £70 million of market losses;
• £40 million of operational and conduct losses;
• £60 million of after-tax operating deterioration.

Total CET1 reduction is:

Therefore:

This already falls below the £240 million critical floor. The reverse-stress conclusion is therefore that the initial scenario itself breaches the floor; no further loss is required to reach or cross it. If management instead wishes to analyse the position after a different initial loss of £160 million, then CET1 would be £340 million and the remaining loss capacity to the £240 million floor would be exactly £100 million—not £200 million.

To illustrate a coherent incremental pathway, suppose the initial shock is revised for the specific reverse test to a total loss of £160 million, leaving £340 million CET1. The remaining amount is:

One possible pathway is £60 million of additional credit loss plus £40 million of fire-sale loss:

This is the correct boundary calculation. An additional £100 million credit loss plus £100 million fire-sale loss would be a £200 million incremental loss and would reduce CET1 to £140 million, not £240 million.

Liquidity overlay

Assume the same stress generates a 15% run-off of £2,000 million of relevant deposits:

Other net cash outflows are £120 million and contractual inflows are £80 million. The immediate net cash need is:

The bank has £500 million of immediately available liquidity, so the initial unfinanced cushion is:

The £900 million total liquid-asset figure cannot be used to claim a £560 million cushion because £400 million is restricted or encumbered. If the bank sells assets to meet the £340 million need and suffers £40 million of fire-sale loss, that loss belongs in the capital calculation. The case therefore demonstrates how liquidity actions can worsen solvency.

Approach and outcome

The board should record two findings. First, the initial £280 million loss scenario breaches the £240 million floor immediately. Second, under the alternative £160 million initial-loss calibration, a further £100 million loss reaches the floor, with a feasible decomposition of £60 million credit loss and £40 million fire-sale loss. The liquidity position remains positive on the stated assumptions, but the margin is sensitive to haircuts, timing, additional withdrawals and operational access.

Recovery triggers should activate before the floor is reached. Possible actions include suspending distributions, retaining earnings, increasing secured funding, preparing an equity issue, reducing asset growth and communicating liquidity strength without making misleading claims about restricted assets.

Learning points

1. Reverse stress testing requires a clearly defined endpoint.
2. Loss components must add arithmetically to the stated capital outcome.
3. Total liquid assets are not the same as immediately deployable liquidity.
4. Fire sales can create both liquidity and capital effects.
5. A breach should be linked to pre-authorised actions, not merely reported.

5. Recovery planning

A recovery plan describes how a viable bank can restore its financial position after severe stress. It is different from a resolution plan, which addresses failure or likely failure when recovery is not credible or sufficient.

A recovery framework normally contains:

• Indicators: capital, liquidity, market confidence, profitability and operational metrics;
• Triggers: quantitative thresholds and qualitative warning signs;
• Options: capital raising, asset sales, funding changes, business reduction and restructuring;
• Impact assessment: timing, capacity, cost, customer effects and execution risk;
• Governance: decision rights, escalation and board oversight;
• Communication: internal, supervisory, investor, customer and media messages;
• Testing: simulations, dry runs and periodic updating.

Triggers should be graduated. Early-warning indicators allow prevention; recovery indicators require formal escalation; severe indicators may require supervisory engagement and preparation for resolution. A trigger is not itself a prediction of failure. It is a control point for action.

6. Resolution and capital contingency

Resolution planning protects critical economic functions while allocating losses to shareholders and creditors according to the applicable legal framework. It considers separability of businesses, continuity of payment and deposit services, access to funding, valuation, data quality, legal-entity structure and operational dependencies.

Capital contingency planning should specify:

1. The minimum capital buffer and trigger points;
2. Instruments that could be issued and their investors;
3. Documentation, approvals and realistic execution time;
4. Effects on ownership, control and market confidence;
5. Restrictions on dividends, bonuses and discretionary growth;
6. Consequences if private capital is unavailable.

A capital action that takes six months cannot be treated as an immediate response to a two-week liquidity crisis. Contingency plans therefore require a time axis and prioritisation.

7. Crisis governance and decision-making

Crisis decisions are made with incomplete, conflicting and rapidly changing information. Effective governance establishes an incident-management structure with a clear accountable executive, risk and treasury leads, legal and compliance advice, operational continuity leadership and a board escalation route.

A decision log should record the time, information available, options considered, decision owner, rationale, assumptions, conflicts of interest and review date. This supports accountability and later learning. Pre-agreed authorities reduce delay, but exceptional actions should still be documented and legally reviewed.

Decision quality can be improved by separating:

• facts from assumptions;
• known uncertainties from unknown uncertainties;
• reversible from irreversible actions;
• immediate stabilisation from longer-term restructuring.

The bank should use an information cadence: rapid initial situation reports, scheduled updates, independent challenge and explicit confidence levels. Excessive model precision can be dangerous when input data are unreliable.

8. Crisis communication and ethics

Communication is a risk-control activity. Messages should be timely, accurate, consistent across channels and proportionate to confidentiality obligations. They should explain what is known, what is being done and when further information will be provided.

Management must not describe restricted assets as immediately available, imply guarantees that do not exist or selectively disclose favourable stress results. Misleading reassurance may intensify later loss of confidence. Communications should consider vulnerable customers, accessibility, language, data privacy and the risk of discriminatory service restrictions.

Ethical stress testing also requires attention to distributional effects. A recovery action such as rapid withdrawal of lending may disproportionately harm small businesses or low-income borrowers. Such effects do not necessarily prohibit the action, but they should be identified, governed and mitigated where possible.

9. Technology, data and artificial intelligence

Stress testing depends on accurate exposure, collateral, maturity, liquidity and legal-entity data. Data lineage should show the source, transformation, owner, date and validation status of each material input. Aggregation systems must avoid double counting and should reconcile to the general ledger and regulatory returns where appropriate.

AI and machine-learning tools may assist with early-warning signals, document analysis, scenario generation or anomaly detection. However, they should not replace governance or expert challenge. Important controls include:

• explainability of material outputs;
• testing for bias and instability;
• protection of confidential data;
• human approval of decisions;
• monitoring for model drift;
• fallback procedures when systems fail.

Generative systems can produce plausible but unsupported scenarios. Every machine-generated assumption requires independent validation and documented provenance.

10. Limitations and contemporary developments

Stress tests have several limitations. Models are calibrated on incomplete historical data; correlations may change in crisis; management actions may fail because all institutions act simultaneously; accounting and regulatory rules may alter measured capital; and reputational effects are difficult to quantify. Scenario uncertainty should therefore be expressed through ranges, alternative narratives and sensitivity to assumptions.

Contemporary programmes increasingly examine climate-related transition and physical risks, cyber and third-party concentration, intraday liquidity, crypto-asset exposures, non-bank financial intermediation and cross-border contagion. These risks often have long horizons, feedback effects and limited historical observations. They require exploratory scenarios in addition to conventional capital exercises.

Best practice is to maintain a living programme: update scenarios, back-test selected relationships, challenge assumptions, test operational access to liquidity, rehearse communications and close identified weaknesses. A stress test that is never connected to decisions is an analytical report rather than a risk-management control.

Chapter Summary

Stress testing evaluates how a bank responds to specified adverse conditions. Sensitivity tests isolate drivers, scenario tests combine coherent shocks, and reverse stress tests identify pathways to a defined failure or limit-breach condition. Credible design requires a clear purpose, scope, horizon, narrative, data foundation, behavioural assumptions and management actions.

Capital analysis must distinguish changes in CET1 from changes in RWA and must show all loss components consistently. Liquidity analysis must distinguish total liquid assets from resources that are unencumbered, operationally accessible and available within the relevant time. The Northbridge application demonstrates why arithmetic reconciliation and precise liquidity terminology are essential.

Recovery plans provide graduated indicators, triggers and feasible options for restoring viability. Resolution plans address failure and continuity of critical functions. Capital contingency, crisis governance, decision logs and honest communication convert analysis into effective action. Model uncertainty, operational constraints, ethics, data quality and technology risks must be actively governed.

Key Terms

Stress testing: Structured analysis of performance under adverse assumptions.

Sensitivity test: Test changing one or a few risk drivers while holding others broadly constant.

Scenario analysis: Assessment using a coherent combination of adverse assumptions.

Reverse stress test: Analysis beginning with a failure condition and working backwards to identify causes.

Severe but plausible: A stress level that is highly adverse but credible given evidence and expert judgement.

CET1 capital: Highest-quality regulatory capital, principally common equity and retained earnings subject to regulatory adjustments.

Risk-weighted assets (RWA): Assets and exposures weighted according to regulatory risk measures.

CET1 ratio: CET1 capital divided by RWA, expressed as a percentage.

Expected loss: A credit-loss estimate commonly approximated by EAD × PD × LGD.

Exposure at default (EAD): Amount expected to be outstanding when default occurs.

Probability of default (PD): Likelihood of default over a specified horizon.

Loss given default (LGD): Proportion of exposure not recovered after default.

Liquidity: Ability to meet obligations when due without unacceptable loss.

Encumbrance: Restriction that prevents or limits an asset’s use for another purpose.

Recovery plan: Documented set of actions intended to restore a viable institution.

Resolution: Managed treatment of a failing institution under an applicable legal framework.

Recovery trigger: Threshold requiring escalation or implementation of a recovery action.

Capital contingency: Preplanned means of preserving or replenishing capital.

Crisis communication: Governed communication during a severe disruptive event.

Model risk: Risk that a model or its use produces an inaccurate or inappropriate result.

Review Questions

Short Answer Questions

1. What is the principal difference between a sensitivity test and a scenario test?
2. Why is a reverse stress test useful even when the failure pathway is unlikely?
3. Define immediately deployable liquidity.
4. State the formula for the CET1 ratio and define its numerator and denominator.
5. Why must recovery actions be assessed for execution time and capacity?

Long Answer Questions

1. Explain how a bank should design a severe-but-plausible integrated stress scenario.
2. Discuss the relationship and differences between recovery planning and resolution planning.
3. Explain how liquidity stress can create additional solvency losses through asset sales.
4. Evaluate the governance and ethical requirements for crisis communication.
5. Discuss the principal limitations of model-based stress testing and methods for addressing them.

Analytical / Application Questions

1. A bank has £600 million of CET1. A stress causes £180 million of credit losses, £70 million of market losses and £50 million of operational losses. Calculate remaining CET1 and determine the additional loss that would reduce it to a £250 million floor.
2. A bank reports £1,000 million of liquid assets, of which £350 million is encumbered. It faces £420 million of net cash outflows. Calculate immediately deployable liquidity and the initial surplus or shortfall, ignoring new funding and haircuts.
3. Design a reverse stress pathway for a bank whose critical condition is a CET1 ratio below 8%. Identify at least three linked shocks, two feedback effects, management actions and one reason the pathway might be unreliable.

Multiple Choice Questions with Answers

Question 1

Which test changes one principal risk driver while holding other assumptions broadly constant?

A. Sensitivity test

B. Reverse stress test

C. Resolution test

D. Recovery-capacity test

Correct Answer: A. Sensitivity test

Question 2

A bank has £500 million of CET1 and £4,000 million of RWA. What is its CET1 ratio?

A. 8.0%

B. 12.5%

C. 80.0%

D. 125.0%

Correct Answer: B. 12.5%

Question 3

After an initial loss, CET1 is £340 million and the critical floor is £240 million. What additional total loss reaches the floor?

A. £60 million

B. £100 million

C. £240 million

D. £580 million

Correct Answer: B. £100 million

Question 4

Northbridge’s £900 million liquid-asset stock includes £400 million that is encumbered or restricted. What amount is immediately deployable in the stated case?

A. £400 million

B. £500 million

C. £900 million

D. £1,300 million

Correct Answer: B. £500 million

Question 5

If EAD is £20 million, PD is 5% and LGD is 40%, the approximate expected loss is:

A. £0.04 million

B. £0.40 million

C. £4.00 million

D. £16.00 million

Correct Answer: B. £0.40 million

Question 6

Which statement best describes a recovery plan?

A. It predicts the most likely macroeconomic outcome.

B. It specifies actions to restore a viable bank under stress.

C. It replaces all regulatory capital requirements.

D. It guarantees that no creditor will incur losses.

Correct Answer: B. It specifies actions to restore a viable bank under stress.

Question 7

Why should a bank not count an encumbered security as immediately available liquidity?

A. It has no accounting value.

B. Its use may be legally or operationally restricted.

C. It is always a credit loss.

D. It automatically increases RWA.

Correct Answer: B. Its use may be legally or operationally restricted.

Question 8

A bank has £500 million of immediately available liquidity and a calculated immediate cash need of £340 million. Ignoring further effects, its initial cushion is:

A. £160 million

B. £340 million

C. £500 million

D. £840 million

Correct Answer: A. £160 million

Question 9

Which item most directly improves the auditability of a crisis decision?

A. A decision log recording information, options and rationale

B. An unrecorded verbal instruction

C. A public statement before verification

D. Deletion of superseded assumptions

Correct Answer: A. A decision log recording information, options and rationale

Question 10

What is the central question in a reverse stress test?

A. What outcome is most probable next quarter?

B. What combination of events would breach a defined critical condition?

C. How can every risk be eliminated?

D. Which model produces the highest capital ratio?

Correct Answer: B. What combination of events would breach a defined critical condition?