Explaining Default of Mortgages in An Emerging Market Using SHAP and LASSO

We employ two frameworks for finding the best variables for explaining default of mortgages in an emerging market. Firstly, we use SHapley Additive exPlanations (SHAP) values with five different tree-based machine learning methods. Secondly, we apply the least absolute shrinkage and selection operator (LASSO) with logistic regressions. Our findings reveal that loan-specific, macroeconomic, and demographic variables are important in explaining mortgage default. We share our code for future studies