Banking Stability Multifactor Modelling Using Varma Methodology in the Republic of Armenia

Abstract

The banking stability modelling with VARMA model has been conducted in the given research where 48 factors, representing different sectors of economy, have been selected. After that, by using RIDGE and Lasso regressions, as well as correlation coefficients with the dependent variable, only the most significant factors have been chosen. As the determinant of the banking stability, the banking Z-score has been selected, which is calculated as a country level measure, and not for individual banks. Using these factors, the VARMA model has been estimated, and the best model is identified as the VARMA(2,1). It is proven, that factors like Regulatory capital to riskweighted assets, dollarization rate, exchange rates, CPI, profitability of government bonds, nonperforming loans, ROA, ROC, liquid assets to demand deposits, and other factors selected in the result of model evaluation are good indicators for modelling the banking stability. The model prediction power is tested by splitting data into train and test parts and comparing predicted values with test dataset. The model shows quite small mean squared error and proves to be useful for modelling Armenian banking stability with existing dataset.