Modellare Depositi a Vista nel Sistema Bancario Italiano

Modelling sight deposits is a crucial aspect for analysing the risks a bank is exposed to. The object of this thesis is finding the most suitable models to forecast the evolution of the deposit interest rate and of the deposit volume. This is done by first simulating the evolution of two risk factors: the market interest rates and the level of creditworthiness of the Italian banking system, measured by the CDS spread. In particular, for the market rates, the used models are the Hull-White one factors and the two-additive-factor Gaussian model (G2++), both calibrated via Kalman filter representation. In addition, the simulation according to a further model based on PCA is performed. CDS spread, instead, is simulated according to the Cox-Ingersoll-Ross (CIR) model, calibrated via maximum likelihood approach. Finally, two statistical tests are used to find the best linear model to explain the evolution of deposit interest rate and of deposit volume. The tests, test on residuals and Chow test, allow to find a model that does not cause an increase in residuals' magnitude in the out-of-sample set and whose parameters' estimate is stable, when new data become available. This analysis is important in order to identify which are the parameters that most influence the levels of deposit rate and volume. After the choice, the most suitable model is calibrated using the entire data set and the simulation process is performed. In the first chapters of the thesis, the risk factors and the deposit rate and volume are introduced, with a particular focus on the theoretical aspects needed for the calibration and simulation of the processes. In the second part, the results of calibration and simulation processes are presented. Chapter 6 reports instead the comparison between linear models for deposit interest rate and for deposit volume, and the choice of the most suitable ones. In the last Chapter, the obtained results regarding the relationships between variables are stated.