Forecasting private sector bank deposits in Greece: determinants for trend and shock effects

In this work, we employ a Markov regime switching autoregressive model (MS-AR) to project private sector bank deposits in Greece, where low uncertainty regimes are followed by high uncertainty regimes or vice versa. MS-AR models are a well-established method for modelling effectively nonlinear financial time series, which entail non-homogeneous observed data and where the patterns of the entailed temporal dynamics may change over time. Our empirical evidence suggests that such a model has a superior out of sample predictive performance compared to a broadly used time series model. Essentially, our model outperforms in forecasting both the evolution of outflow rates and the absolute level of customer deposit balances over time, while its superiority relative to peer models is even more pronounced during periods of systemic shock in a country. Our model specification can be employed not only in performing baseline forecasting of the deposit growth rate, but also in modelling volatility shocks in the deposit time series. Essentially, our model can be used as an integrated part of a capital or liquidity stress testing framework by supervisory, macroprudential or monetary policy authorities, in assessing the resilience of a certain bank or of the whole banking sector.