Determining the factors for individual credit approval by applying logistic regression and hierarchical logistic regression

Abstract

There has been a rapid increase in applications made to lending institutions due to the inadequacy of individuals' savings to meet their growing demands and needs. However, specific criteria are expected from individuals for the approval of these applications. Various methods have been developed to accurately determine these criteria and facilitate the loan approval process efficiently and promptly. Despite being the most used method, binary logistic regression has its limitations. These include the simultaneous inclusion of all variables in the analysis, the oversight of control variables that could be determined by decision-makers, and the evaluation of each event within the group of the same variables. In an attempt to address these limitations, hierarchical logistic regression analysis was performed alongside binary logistic regression analysis. The study incorporated variables indicating individual characteristics, such as age, socio-demographic characteristics like occupation and account status, and loan attributes, including duration and installment rate. The study obtained 1000 loan applicants, and the models were evaluated based on the goodness of fit results and the explanatory power of the models. Although the study produced similar results in determining the influential variables and evaluating the models, it highlighted that hierarchical logistic regression provides more flexibility to decision-makers by allowing the evaluation of nested models formed by sub-variable groups separately. This approach can offer valuable insights to decision-makers in lending institutions, as it allows for the separate evaluation of various sub-variable groups, contributing to a comprehensive understanding of the key determinants of loan approval. By prioritizing specific variables or groups of variables, this methodology facilitates a more comprehensive analysis of the complex relationships between different factors, leading to more informed and data-driven decision-making. It is crucial, however, to ensure the appropriate application of the methodology and the continuous validation and testing of the hierarchical logistic regression models with new datasets to maintain their effectiveness and relevance in the dynamic lending landscape.