Estimation of default rates using the regression model and forward-looking modeling

Objective: The overall background of this paper is the economic and financial reality around credit risk and default probability estimates made by banks and factored in as a risk element inside predicted losses. This paper describes how to build a regression model for calculating forwardlooking adjustment factors. In practice, determining a link between the default rate curve used to calculate the adjustments for expected losses related to credit risk exposures and macroeconomic factors is what this requirement entails. The default rate curve will be adjusted in accordance with the developed model based on forecasts of macroeconomic factors. Furthermore, forward-looking models include rational expectations, and economic agents understand the correct future values of each variable. The primary objective of the current paper analysis is to recognize statistical relationships between the default rate of a portfolio of physical clients, which is variable dependent, and one or more macroeconomic variable, which are variable independent. This will be accomplished by employing a number of linear rule models and, ultimately, by selecting a single model that will accurately represent the statistical and economic aspects of this relationship. Following that, on the basis of predictions, factors of adjustment k will be calculated over a three-year period, using PD-based historical conditions to calculate provision. The time span for which predictions will be fulfilled is from June 2019 to September 2022. Estimating default rates and discovering relationships between them and various macroeconomic factors can provide an overview of economic reality and reflect the extent to which a person has the necessary resources for the development and capitalization of their heritage. Higher default rates can indicate a lack of individual resources to repay loans, a change in macroeconomic reality that affects population incomes, or, indirectly, an inability to capitalize, and develop heritage. Method: This section will contain the rules for multiple-line regression. It was chosen because it is both robust and interpretable. There are multiple input variables that could have an impact on the outcome, or target variable, according to multiple regression. The classic multiple linear regression must be used here with a modification to make the dependent variable specifics more predictable. Rates by default will be between 0 and 1 (0% and 100%), with the emphasis on the fact that no negative values will be will be that all model predictions will be correct and will have values between 0 and 1. Results: The model selected for predicting default rates is the one made up of the independent variables consumer price index and gross domestic product, as well as the dependent variable DR, for which a cubic transformation was performed. Lag4 and lag1 transformations were used for the two independent variables. The values derived from the projections are reasonable and appropriately capture the trend. Default rates climbed during the pandemic of 2020 as unemployment also increased, people had trouble making loan payments, some of them lost their jobs, and rates then started to decline. Additionally, using these forecasts, the yearly adjustment factors k were derived as a ratio between the average of the forecasted values for the previous four quarters and the average of the historical default rates during the preceding three years: k1 = 0.59, k2 = 0.70, k3 = 0.39. Originality: The original approach entails establishing a correlation between macroeconomic variables that represent potential recent shocks to credit default rates in order to forecast these rates and, based on them, determine adjustment factors for the probability of default using forward-looking modeling.