Stress testing examines how an institution could be affected by adverse conditions and whether its planned responses would be sufficient. It is not a prediction of the next crisis. Its value lies in exposing vulnerabilities, challenging assumptions and informing decisions before those decisions become urgent.
Capital and liquidity need to be considered together. Credit losses can reduce capital and weaken confidence, while funding pressure can force asset sales or constrain new business. A model that studies these effects in isolation may miss the feedback between them.
The Basel stress-testing principles provide an international reference for objectives, governance, methodology and use. [1] Banks and NBFCs should apply their own relevant domestic requirements and proportionate internal framework. The examples below are hypothetical and are intended to explain analytical design, not prescribe regulatory scenarios or minimum outcomes.
Begin with the management question
A stress test should answer a defined question. Can the institution sustain its business plan under lower collections? How exposed is it to one sector? How long could it operate if a funding source became unavailable? A precise question helps determine the scope, horizon and required detail.
Distinguish sensitivity analysis from a scenario. A sensitivity changes one factor to understand its effect. A scenario combines related developments into a coherent adverse environment. Both are useful, but they support different conclusions.
Identify the decision the result could influence. It might concern capital planning, liquidity resources, concentration limits or contingency actions. A technically elaborate exercise with no connection to decisions can consume substantial effort without improving risk management.
Establish a credible starting position
The starting balance sheet, capital position and cash-flow profile should reconcile to the institution's approved records. Document the reference date, perimeter and any adjustments. An unexplained difference in the starting point weakens every later result.
Include material off-balance-sheet commitments and contingent exposures where relevant. Consider whether the model captures the entities, currencies and products that matter to the question. Exclusions should be explicit, with their possible effect on interpretation.
Record data limitations before running the scenario. If a portfolio lacks reliable maturity or classification information, the institution may need a conservative treatment or a targeted improvement. Do not present a precise result as fully reliable when the underlying exposure is poorly understood.
Design a coherent adverse scenario
Choose drivers linked to the institution's vulnerabilities. Relevant variables can include economic activity, interest rates, asset prices, collection performance and funding availability. Explain why the chosen combination is plausible for the purpose of the exercise and sufficiently challenging.
Specify timing. A sudden funding shock and a gradual deterioration in credit performance produce different demands on management. The scenario should show when losses, cash shortfalls and other pressures emerge, not only the final position.
Avoid selecting assumptions solely to produce an acceptable outcome. Document the rationale, alternative scenarios considered and challenge by relevant specialists. A scenario should help reveal weaknesses, including uncomfortable results that require action.
Map drivers to financial effects
Translate the scenario into transmission channels. A weaker sector may increase arrears, reduce recoveries and affect collateral values. Higher interest rates may change funding costs, borrower affordability and the value of assets. Reduced market confidence may affect renewal of borrowings.
Make these relationships explicit. Where models are used, document their inputs, calibration and limitations. Where expert judgement is necessary, record the reasoning and the person responsible. Neither modelling nor judgement should operate as an unexplained black box.
Check consistency between teams. Credit, finance and treasury may use different assumptions about the same borrower population or economic driver. A combined stress test needs a shared narrative and reconciled treatment of overlapping effects.
A worked example: losses and delayed collections
Suppose a hypothetical lender begins with eligible capital of 150 units and a planned cash surplus of 20 units over a defined period. A scenario generates additional credit losses of 18 units and delays 35 units of expected receipts. The simplified capital position falls to 132 units before other effects.
The liquidity impact is not simply the 18-unit accounting loss. Delayed receipts create a separate cash-flow pressure, potentially turning the planned surplus into a shortfall. Some delayed amounts may ultimately be recovered, but the institution still needs cash before that happens.
The example shows why capital and liquidity outputs should be linked without being treated as identical. A loss estimate, a cash-flow forecast and a regulatory ratio each measure a different aspect of the scenario.
Challenge management actions for feasibility
Management actions can improve a stressed position, but they should not be assumed to occur instantly or without cost. Capital raising, asset sales, funding draws and reductions in new business have timing, approval and market dependencies.
Show results before and after proposed actions. This allows decision-makers to see the underlying vulnerability and the extent to which the outcome depends on execution. Include the period before an action becomes effective, because that may contain the most severe pressure.
Avoid double counting. Selling an asset may generate liquidity while reducing income or crystallising a loss. Drawing a facility may require collateral already assumed available for another action. The model should reflect these interactions consistently.
Use reverse stress testing to expose assumptions
Reverse stress testing starts with a defined adverse outcome and asks what combination of events could produce it. The purpose is to identify vulnerabilities and implausibly strong assumptions, rather than to claim that the outcome is likely.
For example, an institution might examine what collection delay and funding reduction would make its contingency resources insufficient. The exercise can reveal that the main vulnerability is operational access to funding rather than the nominal size of the available facilities.
Discuss the plausibility and implications of the identified conditions. If the outcome requires several extreme assumptions, that is informative. If it arises from one moderate but concentrated disruption, management may need to reconsider limits, resources or contingency arrangements.
Validate the exercise and its limitations
Validation should examine data, logic, assumptions and the reasonableness of results. Independently recompute selected calculations and test sensitivity to important parameters. Large or counterintuitive movements need explanation.
Compare model behaviour with relevant experience while recognising that historical events may not represent the chosen scenario. A model can fit the past and still fail to capture a future structural change. Expert challenge remains important.
State what the exercise does not capture. Examples may include feedback from market reactions, operational constraints or limited data on a new product. Limitations should influence how the results are used, not appear only in a technical appendix that decision-makers never read.
Present results as a decision narrative
A useful report explains the scenario, main drivers, point of greatest pressure and actions available. It should connect the numbers to business implications. Identify which exposures contribute most and which assumptions materially change the conclusion.
The Basel liquidity-risk guidance provides complementary context for contingency planning and stress-related funding analysis. [2] In practice, the stress report should align with the institution's actual funding plan and tested operational capabilities.
Record decisions and follow-up actions. If management accepts a vulnerability, document the authority, rationale and monitoring. If it requests mitigation, define an owner and a measurable outcome. Repeating the test without tracking decisions limits its value.
Make uncertainty visible through ranges
A single stressed result can imply more certainty than the method supports. Where important assumptions are uncertain, show a reasoned range or sensitivity around them. For example, alternative collection-delay assumptions can reveal whether the conclusion depends on a narrow estimate.
Explain which uncertainties matter most. Management does not need every possible model variation, but it should understand the assumptions that could reverse a decision. A small change in one parameter may be more consequential than large changes in several others.
Keep the scenario definition separate from parameter uncertainty. The scenario describes the adverse environment; the sensitivity explores uncertainty in how the institution responds to it. Clear labelling prevents readers from mistaking different analytical purposes.
This approach supports better decisions without pretending that uncertainty can be removed through additional decimal places.
Make stress testing part of ordinary planning
Refresh scenarios when the business model, portfolio or operating environment changes materially. A test designed for an earlier funding mix may no longer address the institution's main risk. Review assumptions after incidents and significant forecast errors.
Use results in product planning, concentration discussions and resource decisions where relevant. Stress testing should challenge the business plan rather than sit beside it as an unrelated exercise. It can reveal which growth assumptions depend on continued favourable conditions.
An effective stress test produces a clear explanation of how pressure develops and what the institution can do about it. Reliable starting data, coherent scenarios, realistic actions and independent challenge make the result useful even when the future unfolds differently from the scenario imagined.
PRIMARY REFERENCES
Sources & further reading
Primary-source references checked on 12 October 2026. Follow the relevant authority’s current text for applicability, amendments and commencement dates. International standards provide context and are not automatically identical to domestic requirements.