Modeling and liquidity risk management of bank deposit volumes

Abstract

This dissertation focuses on the modeling and liquidity risk management of bank deposit volumes. Bank deposits have no specific contractual maturity, and individual clients can always add or withdraw at any time at no penalty. Moreover bank deposits make up a considerable part of the funding of a bank and while they have no stated maturity, they are a source of stable funding. The bank deposit volumes in this dissertation are modeled using a part-regression part-diffusion model based on the one presented by Kalkbrener and Willing (2004). The regression part determines the growth and the diffusion part models the correlated residuals of the detrended data set. We use maximum likelihood estimation techniques to extract the model parameters of the diffusion part, and we also apply the necessary bias-correction to these parameters. In addition, three liquidity risk management methods on these products are also examined, where the estimation of the core level that determines the stable part and the volatile part is carried out. This includes also comparison of the performance of these three methods. Possible improvements to this study are also suggested.