Stress testing and capital planning — closing the institutional silo

The institutional separation between stress testing and capital planning is one of the more durable structural problems in bank risk management. It has a natural explanation. Stress testing grew from the risk function — it was, and remains, a risk measurement and monitoring discipline. Capital planning grew from finance — it is an accounting, forecasting, and regulatory capital optimisation discipline. The two functions use different models, different data pipelines, different planning horizons, and different governance forums. In many institutions they sit in entirely separate organisational silos.

The regulatory framework has been explicit, for more than a decade, that this separation is not acceptable in practice. BCBS Pillar 2 guidance and its national implementations across jurisdictions uniformly require that stress testing be embedded in capital planning — that the outcomes of stress exercises shape the institution’s view of its capital requirements under adverse conditions, not merely sit beside it as a separate regulatory deliverable. The CBUAE ICAAP framework, in force since April 2021, makes the integration requirement concrete.

The gap between the requirement and institutional practice remains, in many banks, substantial.

What the CBUAE framework requires

The CBUAE ICAAP Pillar 2 Guidance addresses the integration of stress testing and capital planning directly in paragraphs 112 and 120 of Section VII.

The operative word is incorporated. Not referenced, not appended, not included in a separate chapter. The stress test outcome should be visible inside the capital plan itself — as one of the scenarios against which capital adequacy is assessed, and as a binding input to the capital buffer and capital action decisions in the plan.

Paragraph 120 reinforces this through the use test. The requirement is that stress tests support decision-making throughout the bank effectively, at every level from the trading desk to the board. A stress testing programme whose outputs circulate in the risk function and appear in the ICAAP submission, but which do not visibly influence credit limits, origination decisions, treasury positioning, or capital allocation, fails the use test regardless of the technical quality of the scenarios.

The BCBS Stress Testing Principles (October 2018) frame the same requirement in terms that apply universally: stress testing should be an integral part of capital planning, should guide the setting of the bank’s risk tolerance, and should have a meaningful impact on business decisions, with the board and senior management engaged in evaluating results and their impact on the risk profile.

Where the silo breaks the integration

The failure mode is not usually visible in the ICAAP submission itself. The submission typically contains both a stress testing chapter and a capital planning chapter, and the stress chapter will reference the capital impact of scenarios under the stress. The failure tends to appear in the mechanics — in what actually happened when the two chapters were written, and in whether the numbers in the capital chapter reflect the numbers in the stress chapter or were produced independently.

The most common pattern is a capital plan built on business-plan assumptions, and a stress testing programme whose outputs stop at the risk dashboard.

In our engagements, the most common pattern is as follows. The finance function builds the capital plan from business-plan projections — revenue forecasts from the business lines, cost trajectories from finance, loan growth from the commercial team. These projections are optimistic in the sense that they reflect management’s plan rather than a statistically neutral forecast. The risk function independently runs stress scenarios — regulatory scenarios, or scenarios constructed internally — and produces capital impact estimates for each. The two outputs are then combined in the ICAAP by inserting a section that shows the capital impact under each stress scenario. But the capital plan itself — the document that drives actual capital decisions, dividend policy, issuance plans, and regulatory buffer management — has not been revised in light of the stress outputs. It reflects the business-plan base case.

The governance of the two processes reinforces the silo. The capital plan is typically approved through the finance committee or ALCO, on a planning timeline that follows the budget cycle. The stress testing programme runs on a separate timeline, driven by regulatory requirements and the ICAAP submission calendar. In institutions where the two timelines do not synchronise, the outputs of stress testing arrive either too early or too late to genuinely inform the capital plan decisions.

The management actions assumption

A specific technical failure that tends to appear at the intersection of stress testing and capital planning is the treatment of management actions.

Capital plans typically include management actions that would be available in stressed conditions — dividend suspension, rights issuance, balance sheet reduction, portfolio de-risking. These actions improve the capital position under stress and are, in principle, available to management. Including them in the capital plan is legitimate. The question is whether they have been stress-tested.

Under a severe macro stress scenario, dividend suspension may be constrained by legal obligations or investor agreements. Rights issuance may be unavailable if market conditions are part of the stress. Balance sheet reduction may generate losses if asset prices are stressed. Portfolio de-risking may conflict with other portfolio management objectives. Each of these management actions is available in normal conditions and potentially unavailable or less effective in the specific stress conditions the capital plan is meant to address.

What functional integration looks like

In the institutions where stress testing and capital planning are genuinely integrated rather than merely co-located in the ICAAP submission, a small number of structural features tend to be present.

The capital planning process runs three tracks — base, adverse, and severe — where the adverse and severe tracks are constructed from stress test outputs, not from independently produced pessimistic forecasts. The risk function and the finance function share a common projection model, or at minimum a transparent mapping between the risk stress outputs and the finance capital model. The results of each track are presented to the board’s risk committee and finance committee jointly, not sequentially.

The management actions in the capital plan are tiered by scenario. Actions that are reliably available under all conditions appear in all three tracks. Actions that depend on favourable market conditions are available only in the base track. Actions that require exceptional board authority or regulatory approval are presented separately, with a note on the governance timeline required to execute them. The capital plan’s adequacy under each track reflects the constrained set of available actions, not the full set.

The timing of stress testing and capital planning is synchronised. This typically means the stress testing exercise runs in the quarter before the capital planning cycle, so that its outputs are available before the capital plan decisions are made rather than after. Where the two cycles run in parallel and the ICAAP is assembled by merging the outputs, the integration tends to be presentational rather than substantive.

The use test is operationalised through explicit linkages to limit-setting and origination decisions. If a stress scenario identifies a particular portfolio segment as a concentration risk under adverse conditions, the capital plan should show either that the segment’s limits are being tightened or that capital is being allocated specifically to cover the elevated risk. Where neither is shown, the use test question — whether stress test results are actually influencing business decisions — has not been answered.

The supervisory test

The supervisory review of the ICAAP increasingly focuses on the integration question. Where the risk chapter and the capital chapter are internally consistent — where the capital adequacy conclusion follows demonstrably from the stress scenario outcomes — the submission tends to generate less friction. Where the two chapters can be read in isolation without reaching the same conclusions, the supervisory dialogue tends to probe the disconnection directly.

The CBUAE’s §120 use test is the supervisory anchor for this. If the board and senior management cannot point to specific business decisions that were informed by recent stress test outputs, the use test has not been met. The ICAAP document demonstrates capability; the supervisory dialogue tests whether the capability was actually used.

For a detailed treatment of integrated stress testing and capital planning architecture — including the scenario construction methodology, management actions tiering, and the governance structures that make integration operationally sustainable — see our white paper in the Library.