Driver-based modelling
Definition
A forecasting approach in which balance sheet volumes, product mix and pricing are modelled as explicit functions of macroeconomic or rate drivers rather than held static or grown by a flat assumption.
What This Actually Means
Instead of assuming your mortgage book stays flat or grows at 3% regardless, you build a relationship: mortgage volumes are a function of the rate level, the spread to SVR, unemployment, and house price growth. When you run a rate shock scenario, the model automatically adjusts new business volumes, deposit mix, and pricing to be consistent with that environment. This is how a truly rate-conditioned dynamic balance sheet is implemented in practice.
Where It Matters
The implementation risk is significant and often underestimated. Driver relationships are typically estimated from historical data — which means they embed the rate environments and competitive dynamics of the past. In a scenario that's outside the historical range (e.g. rates at 5% after a decade near zero), the drivers may break down entirely.
A subtler problem: drivers introduce correlation structure into your NII model. If your mortgage volume driver and your deposit repricing driver are both functions of the same rate variable, your model will show offsetting movements that may look like natural hedging but are really just a modelling artefact. Stress-testing the driver relationships themselves — not just the rate scenarios — is essential but rarely done.
Governance challenge: driver assumptions need to be owned, challenged, and updated. In practice they often get set once during model build and drift out of date as the competitive landscape changes.