Bank Stress Tests | Asking the Balance Sheet to Survive a Worse Future

A stress test is not a prediction

The bank is not claiming that the adverse scenario will happen exactly as written. It is asking a different question: if something this severe happened, what would the balance sheet do?

This distinction matters. A forecast seeks the most plausible future. A stress scenario deliberately explores difficult futures that may be less likely but more damaging.

forecast: what do we expect? stress test: what must we survive?

The scenario must be severe enough to change behaviour

If every borrower continues paying normally and every funding market stays open, the exercise reveals little. Useful stress tests combine adverse variables that affect the bank through several channels.

• higher unemployment can weaken household repayment;
• falling property prices can reduce collateral recovery;
• higher rates can increase floating-rate borrower stress;
• lower rates can compress some bank margins;
• wider credit spreads can reduce security values;
• foreign-exchange moves can damage unhedged positions;
• deposit outflows can create liquidity pressure;
• market closure can make refinancing harder.

The scenario should be internally coherent enough to trace how one shock becomes another.

Stress starts outside the bank and enters through specific transmission channels

macro shock → borrower and market behaviour → credit and market losses → income changes → provisions → capital depletion → funding confidence → management response.

A scenario is useful only when the bank can explain how the external event reaches internal numbers.

Credit losses are often the largest pathway

A recession can reduce wages, sales and property values. Borrower default probabilities rise. Recovery rates can fall. More loans become delinquent or non-performing. Provisions increase.

Those provisions reduce earnings. If losses exceed current earnings, they reduce retained earnings and therefore capital.

The stress test follows that movement all the way from a weaker economy to the loss-absorbing layer described in Bank Equity and Retained Earnings.

Collateral can fail at the same time as borrowers

In a property downturn, borrowers can default because income has weakened while property collateral also falls in value. The bank therefore loses both the primary repayment source and part of the secondary recovery route.

A strong stress test captures this wrong-way relationship instead of assuming historical collateral recovery remains unchanged.

Read Why Collateral Does Not Repay a Loan for the recovery boundary.

Net interest income can help or hurt under stress

Interest-rate changes affect loan yields and funding costs on different clocks. A rising-rate scenario can initially raise interest income for an asset-sensitive bank, then increase deposit costs and borrower defaults later.

A falling-rate scenario can reduce borrower payments but compress margins if asset yields fall faster than funding costs.

Stress testing therefore links credit risk and interest-rate risk rather than treating them as independent modules.

Market losses can hit capital before loans default

Securities and trading positions can lose value quickly when yields, spreads, currencies or equity prices move. Some changes affect profit immediately; others can affect other components of equity depending on accounting classification.

The stress test maps those valuation effects into capital rather than assuming all bank losses arrive slowly through loan defaults.

Operational losses belong in severe scenarios too

A cyberattack, major fraud, legal judgment or technology outage can create losses during the same period as financial stress. The bank’s crisis capacity can also be reduced because staff and systems are already under pressure.

A stress framework that considers only credit and market loss can miss the operational events that make a difficult scenario harder to manage.

Capital ratios move through both numerator and denominator

Stress can reduce the numerator because losses reduce capital. It can also change the denominator because borrower risk rises, ratings deteriorate or portfolio composition changes, increasing risk-weighted assets.

capital ratio under stress = lower capital ÷ potentially higher RWA.

This double movement can make capital ratios fall faster than a simple loss calculation suggests.

Read Risk-Weighted Assets for the denominator mechanism.

The leverage ratio provides another stress lens

If stress losses reduce Tier 1 capital while broad exposure remains large, the leverage ratio deteriorates even if risk weights behave unexpectedly.

This provides a useful cross-check on model-based capital outcomes.

Read The Bank Leverage Ratio.

A worked miniature

A bank begins with S$12 billion of eligible capital and S$100 billion of RWA. Its starting risk-weighted capital ratio is 12 per cent in this simplified example.

Under stress, credit and market losses reduce capital by S$3 billion. At the same time, borrower deterioration raises RWA to S$110 billion.

The stressed ratio becomes roughly S$9 billion divided by S$110 billion, or about 8.2 per cent.

The purpose of the example is not to judge a regulatory threshold. It is to show how a stress test changes both sides of the capital equation.

Management actions can change the outcome

A bank facing stress may reduce dividends, cut buybacks, raise capital, slow lending, sell assets, hedge risk, reduce expenses or change funding strategy.

Stress tests can model some management actions, but they should remain credible. It is weak analysis to assume the bank will sell the same assets everyone else plans to sell at normal prices, or raise unlimited new capital during market panic.

The action must survive the world in which the stress scenario occurs.

Static and dynamic balance sheets answer different questions

A static stress test can hold much of the balance sheet constant to isolate the effect of the shock. A dynamic test can allow new business, repayments, management actions and portfolio changes over time.

Static approaches are easier to interpret. Dynamic approaches can be more realistic but introduce more assumptions.

There is no universal best choice; the design should fit the question being asked.

Bottom-up and top-down tests provide different challenges

In a bottom-up exercise, individual banks often apply the scenario using detailed internal data and models under prescribed assumptions. In a top-down exercise, supervisors or central teams can apply more standardised models across institutions.

Bottom-up methods can use richer bank-specific information. Top-down methods can improve comparability and provide an independent challenge.

Using both perspectives can reveal where internal models are unusually optimistic or where standard models miss bank-specific structure.

Sensitivity analysis isolates one variable

A full scenario can change many variables at once. Sensitivity analysis asks what happens if only one driver changes sharply—for example, property prices fall 30 per cent, interest rates rise rapidly, or one large counterparty defaults.

This helps management understand which exposures dominate the outcome.

The narrow test and the broad scenario serve different purposes.

Reverse stress testing starts from failure and works backward

Instead of asking what a given scenario does to the bank, reverse stress testing asks what combination of events would make the business model unviable or breach a critical survival boundary.

The bank then works backward to identify the assumptions and concentrations that create that failure.

This is especially useful for discovering risks that ordinary scenario design may not imagine.

Liquidity stress testing must remain connected to capital stress

A bank can pass a solvency stress and still run out of cash. It can also face liquidity pressure that forces asset sales, creating capital losses that were absent in the original solvency scenario.

Capital and liquidity tests therefore should inform one another.

The liquidity lane begins with What a Bank Liquidity Buffer Is Actually For and The Contingency Funding Plan.

Stress testing is only as good as the data and models behind it

If loan data is incomplete, collateral values are stale or model relationships fail outside historical ranges, the stress result can create false precision.

The bank should therefore understand uncertainty around the result, use conservative overlays where justified and compare model output with expert challenge.

A precise number is not automatically a reliable number.

Stress testing can change strategy before the crisis

If a scenario shows that one sector consumes too much capital under stress, the bank can reduce concentration, tighten underwriting, raise prices or increase buffers before losses arrive.

This is the highest value of stress testing. It is not to produce a report after the balance sheet is built. It is to change the balance sheet while choices are still available.

A passing result is not a guarantee

The real future can be worse, different or faster than the tested scenario. New risks can appear. Customer behaviour can change. Models can fail.

A stress-test pass therefore means the bank remained above the relevant survival boundaries under the tested assumptions. It does not mean the institution cannot fail.

A failing result is not automatically a prediction of failure

If a bank falls below a capital or liquidity boundary in a deliberately severe scenario, the exercise has identified a vulnerability. Management or supervisors can then require more capital, less risk, stronger funding or other remediation.

The test is useful precisely because it can fail on paper before the bank fails in reality.

Stress testing turns capital from a static number into a survival path

Article 41 showed where capital begins. Article 42 showed how risk changes the denominator. Article 43 added a leverage backstop. Stress testing combines them over time.

starting capital → adverse losses → weaker earnings → changing RWA → changing leverage → management actions → ending capital and viability.

The final number matters. The path that produced it matters more.

Four misconceptions to remove

Misconception	Better model
“A stress test predicts the next crisis.”	It tests resilience under specified adverse assumptions rather than forecasting one exact future.
“Only credit losses matter.”	Interest income, markets, operations, funding, RWA and management actions can all change the result.
“Passing means the bank cannot fail.”	A pass applies only to the scenarios, models and assumptions tested.
“A failed stress test means the bank will fail.”	It identifies vulnerability early enough for capital, funding or risk remediation.

A mastery test

1. How is a stress scenario different from a forecast?
2. How can stress reduce both the capital numerator and worsen the RWA denominator?
3. Why should collateral values deteriorate alongside borrower quality in some scenarios?
4. Why must management actions be credible inside the stressed world?
5. Why is a passing stress test not a guarantee of safety?

If those answers connect, stress testing becomes visible as institutional rehearsal: the bank deliberately walks its balance sheet through a worse future while there is still time to repair what would otherwise break.

---