Definition
A behavioural model describes the gap between a product's contractual run-off and its actual run-off when the client holds an explicit or implicit option. The main ALM cases are:
- non-maturity deposits, with no contractual maturity, a stable share of which stays on the balance sheet for a long time;
- loan prepayments, structural or driven by the gap between client and market rates;
- regulated savings and housing savings products, with their deposit and drawdown options;
- drawdowns on credit lines and off-balance-sheet commitments.
Method
Behavioural modelling combines statistical approaches (regressions on historical data, prepayment rates as a function of the rate incentive) and, for some products, stochastic approaches (simulating rate paths and client decisions). The model produces a run-off profile and, where needed, a pass-through profile, then applied by the cash flow engine.
Each model comes in variants: central calibration, stressed calibration, parameter sensitivity, to measure the effect of assumptions on EVE, NII and liquidity.
Regulatory requirement
The EBA IRRBB guidelines (EBA/GL/2022/14) require behavioural assumptions to be grounded in historical data and analysis, documented, regularly tested and subject to model governance. The standardised approach notably frames the stable share of deposits and its maturity. In liquidity, the run-off assumptions used in the ILAAP must be justified.
Practical challenges
- Short histories or histories dominated by a single rate regime, as the low-rate years and the 2022 rise showed.
- Model governance: versioning, documentation, validation, backtesting, comparison with the validated model.
- Consistent use of the same model in every measure that relies on it.
In RIALTO
This concept is covered by the Quant ALM Insight Core module. The use case 'Stress simulation' shows it in practice.