Abstract:
This study develops a framework for forecasting selected balance sheet items of the four largest Maltese core banks, with a particular emphasis on bank profitability. Methodologically, it employs two multivariate time series models, namely a Factor-Augmented VAR (FAVAR) and a Bayesian VAR (BVAR). These models are estimated using macroeconomic series alongside a novel bank-level dataset for selected Maltese core banks compiled by merging data from various sources. Forecasting accuracy is assessed through in-sample forecast evaluation across various data sub-samples. This paper also discusses how these models are used to compute forecasts conditional on a given path for selected macroeconomic variables, providing a comprehensive profitability outlook for the Maltese core banking sector. Key findings are summarized as follows: (i) neither of the two models exhibits a distinct superiority in forecasting performance, yet both demonstrate relative strengths when compared with a naive forecasting model in-sample, (ii) both models generate similar outof-sample forecasts for the period 2023Q4 - 2025Q4.