Fairness Evaluation of Credit Approval Algorithms: Group Variance Analysis after Controlling for Key Variables

With the rapid development of financial technology, algorithmic credit approval systems have been widely adopted, especially in credit. However, the algorithm system is very likely to have systemic injustice. So, this study is concerned about the potential bias in the credit approval algorithm system, investigating the dimensions that bias is most likely to exist in such as gender, education, and home ownership. The author collected the approval rate data from Kaggle to conduct this study. The author used EXCEL to conduct fundamental analysis and use SPSS to conduct further binary logistic regression to test the relationship between the credit approval rate and these variables. The results show that there are no significant biases in the field of gender and education. However, significant disparities were found for home ownership status even after controlling key financial factors. This study helps to evaluate the fairness of the algorithm and thus helps to promote the transparency and fairness of algorithms in the field of credit.