The Risk Was There Before the Indicators Could See It
The transaction meets policy. The borrower is current. Capacity remains under the limit. The expected return exceeds the required hurdle rate. Nothing in the credit file compels the bank to decline it.
And yet, the decision may be committing capital against a future that is no longer probable.
A bank does not begin to deteriorate when the deterioration appears in its indicators. By then, the decision that caused it may have been accumulating on the balance sheet for years.
A bank is an allocator of scarce capital, constrained by risk, liquidity, solvency, and profitability. That definition raises a fundamental question: what is the risk function actually for?
The conventional answer lists familiar responsibilities: identifying exposures, controlling limits, estimating losses, monitoring arrears, and reporting exceptions. All are necessary. None defines the function’s institutional purpose.
Every banking decision rests on conditions that the underlying file rarely presents as bets. A loan assumes that cash flow will continue. Pricing assumes that the margin will absorb the loss. Collateral assumes that a buyer will exist when enforcement becomes necessary. Funding assumes that depositors will remain. Diversification assumes that different exposures will not react to the same underlying factor.
Compliance with a limit does not prove that the decision retains its economic quality. By the time the loss reaches the financial statements, the capital has already been committed and the available corrective options have narrowed.
IFRS 9 acknowledged this gap by replacing the incurred-loss approach with an expected credit loss model that incorporates reasonable and supportable forward-looking information. Yet a model can include future scenarios and still remain a reporting exercise, disconnected from the decisions shaping the balance sheet.
The risk function does not exist to document losses. It exists to challenge the assumptions under which the bank allocates capital.
The Institutional Safety of Explaining Too Late
The committee receives twenty pages on the existing portfolio and two on the quality of recent origination. It knows delinquency by product, sector, and branch, but may not know how many borrowers remain current only by taking on new debt. It can see the remaining headroom under a limit, but not necessarily the common factor connecting exposures classified as unrelated.
The board receives an accurate snapshot of the balance sheet. Less often does it receive its trajectory: which assumptions have started to weaken, which portfolio vintages are deviating from expected performance, and how much profitability would disappear under less benign conditions.
The explanation is institutional. Historical evidence can be verified. Anticipation requires judgment about scenarios that remain open to competing interpretations.
The earlier the warning, the less conclusive the evidence supporting it will appear. If risk intervenes and the loss never occurs, the organization may conclude that the caution was unnecessary. If risk waits and deterioration materializes, the causes can always be reconstructed with evidence that has become incontrovertible.
This creates a dangerous asymmetry. Anticipation can look like an error, while a late explanation retains the appearance of professional precision.
Regulatory reporting, model validation, and limit monitoring reinforce this tendency. These activities are indispensable, but they can consume the capacity that should be devoted to challenging the organization’s assumptions about the future.
The prudential framework points in a different direction. Basel requires internal capital assessments to incorporate stress testing and examine how risks interact under adverse conditions. The Financial Stability Board’s Principles for an Effective Risk Appetite Framework connect risk appetite with strategy and financial and capital planning.
Risk functions converge toward the past because documenting it is institutionally safer than challenging the future.
When the Balance Sheet Tells the Truth Too Late
A recent portfolio vintage may deteriorate without moving the consolidated delinquency ratio if new lending expands the denominator faster than the problem grows. The headline indicator remains stable while the quality of the decisions entering the balance sheet weakens.
A borrower may also remain current through repeated refinancing, asset sales, or extensions. Payment performance is preserved, but underlying repayment capacity continues to erode. The bank may interpret stability where there is only latency.
The same problem arises with concentration. A portfolio may be distributed across construction, retail, transportation, and services while remaining dependent on a single variable: public expenditure, tourism, remittances, exchange rates, or the price of a commodity.
Sector classifications separate the exposures. Stress reveals what they share.
Historical data describe how those exposures behaved under a particular set of conditions. The risk function must identify the conditions supporting each decision, how far those conditions could move, and what that movement would mean for safety, profitability, liquidity, and capital.
Looking forward does not mean predicting the future. It means testing today’s decisions against futures other than the one expected.
Silicon Valley Bank provides an extreme illustration. Its assets grew from approximately USD 71 billion to more than USD 211 billion between 2019 and 2021. Beneath that growth sat three critical assumptions: deposits would remain, assets could be held to maturity, and interest-rate risk would remain manageable.
The Federal Reserve’s April 2023 review concluded that management changed risk-measurement assumptions to reduce the amount of risk being reported rather than addressing the underlying exposures. When those conditions failed simultaneously, time compressed. Approximately USD 42 billion left the bank on March 9, 2023, and management expected another USD 100 billion in withdrawal requests the following day.
The balance sheet did not become fragile in twenty-four hours. Twenty-four hours was how long it took for the capacity to conceal that fragility to disappear.
Capital Is Consumed Before the Loss Appears
The cost of a backward-looking risk function begins before the accounting loss is recognized. Its most consequential form is the gradual loss of reversibility.
Before origination, the bank can adjust the price, shorten the tenor, require additional collateral, reduce the exposure, or decline the transaction. Once the position enters the balance sheet, the range of available options begins to narrow. When deterioration materializes, the bank may have to accelerate provisions, restrict new lending, raise capital, or reduce exposures under unfavorable conditions.
The institution stops choosing among opportunities and starts reacting to constraints.
During the period of latency, capital remains tied up in assets whose economic return has not yet been tested across the cycle. Liquidity is committed and the capacity to fund better opportunities declines. This opportunity cost remains hidden because the assets continue to generate revenue and the loss has yet to appear.
In its Report on the 2023 Banking Turmoil, the Basel Committee described the episode beginning in March 2023 as the most significant system-wide banking stress since the global financial crisis in terms of scale and scope. The review highlighted governance weaknesses and failures to understand the interaction among liquidity, interest-rate, and concentration risks.
Waiting for certainty progressively reduces the institution’s capacity to respond.
