Conditional cash flow modelling
Regulatory Definition
Cash flow modelling under the assumption that the timing and amount of cash flows is dependent on the specific interest rate scenario.
EBA GL/2022/14
What This Actually Means
Your cashflow projections change depending on which rate scenario you're running. In a falling rate scenario, mortgage prepayments accelerate (borrowers refinance), so your asset cashflows shorten. In a rising rate scenario, deposits migrate to higher-paying products, so your liability profile shifts. The model reacts to the scenario.
Where It Matters
Conditional modelling is more sophisticated but the case for it is not clear-cut. There is a genuine trade-off between model complexity and the actionability of the risk metrics it produces.
The core tension: many of the conditional elements — rate-dependent prepayment speeds, deposit migration, volume sensitivity — are non-hedgable interest rate risks. They may be useful for forecasting and business planning, but they are less useful for BAU interest rate risk management because the ALM desk cannot put on a trade to hedge them. A highly conditional model can therefore produce metrics that look comprehensive but are difficult to act on — and deeply difficult to interpret without processes in place to bifurcate the hedgable and non-hedgable components. If an organisation adopts conditional modelling, practitioners need to be able to attribute between these two components — otherwise the metrics conflate risks that require different responses and different ownership, and the model produces numbers that senior stakeholders cannot meaningfully use.
Reliability: conditional models are only as reliable as the relationships they embed. Rate-dependent behavioural assumptions — how prepayments respond to rates, how deposit volumes shift — are estimated from historical data that may not be representative of the current or future environment. In scenarios outside the historical range, these relationships can break down entirely, making the conditional model less reliable precisely when it is most needed.
A valid alternative: running unconditional models for BAU IRRBB management and producing supplementary standalone stress metrics for the conditional elements — stressing prepayment speeds, deposit volumes, and pipeline conversion independently rather than embedding them as rate functions — can give practitioners the best of both. The core metrics remain clean and actionable; the behavioural sensitivities are visible and owned separately without contaminating the hedgable risk picture. The limitation of this approach is that standalone stresses are applied independently and therefore do not capture compounding interaction effects — scenarios where changes in assumptions all occur simultaneously and reinforce or elevate each other. Where those interactions are material, a conditional model or a combined scenario stress that explicitly layers the assumptions together may be needed as a supplement.