Оценка надежности банка как объекта инвестирования
The approach proposed to classify commercial banks into banks that have a high probability of revoking a license and reliable banks, as well as an information and logical model for identifying a group of banks (or one bank) among reliable banks that are attractive for investment. The probability of license revocation was assessed using a logistic regression model based on database, consisting of 17559 observations for all banks, covering the period from Q1 2012 to Q4 2017. In view of the multicollinearity in data, RIDGE modification was applied with the algorithm for determining the penalty coefficient. In the model, the indicators of volatility of macroeconomic variables, expressed in the standard deviation and variance of the macroeconomic variable within the period under review, were included as regressors. The null hypothesis of statistical zero coefficients at volatility indicators of macroeconomic variables is rejected at various significance levels. The model is built in the R-studio programming environment using the «RIDGE» package. Based on the hierarchical clustering by the Ward method (as a measure of the distance — the square of the Euclidean distance) eleven clusters were obtained. In this paper, a brief description of these clusters is presented on the basis of the absolute mean values of the variables, as well as the relative average values of the bank's financial variables. Using the nonparametric Kruskal—Wallis criterion, which makes it possible to compare the average values for several groups simultaneously, it was found that the financial variables differ significantly at high levels of significance. The results of the cluster analysis can be used to support the investor's decision to select a cluster (or a bank within the cluster) to conduct stress testing of credit risk (as the largest source of losses) in order to invest in banks belonging to the selected cluster that have withstood stress testing. In the future, based on the results of stress testing of credit risk, it is possible to select from the analyzed cluster those banks that are the most resistant to stressful events and can later be viewed by the investor as investment objects.