The decision
Which deposit assumptions matter most to earnings across the rate paths the bank is planning for, and how would the committee know when those assumptions have stopped being true?
The approach
Segment the deposit book, then hold every behavioral parameter (beta, lag, floor, migration, decay) in a governed assumption register with a source and a bound. Build a scenario engine that links rate paths to segment repricing and earnings, and validate it against observed history and a challenger model. The deliverable is the model, the register, and the validation evidence, in a form the ALM committee and model risk can review.
Known constraints
- Segment definitions may differ across source systems and need reconciling before anything else.
- Observed history may not cover the next rate regime; the register states where assumptions are extrapolated.
- Behavioral assumptions require governance and challenge, so the model is built to be re-estimated, not fixed.