Default Prediction of a Swedish Mortgage Portfolio using Logistic Regression

Abstract: This thesis was conducted to investigate what factors are important for a financial institute when predicting the risk of default for a Swedish mortgage portfolio. The applied method was logistic regression analysis and the data used in the thesis was received from a Swedish financial institute. Many of the conducted studies assessing the risk of default only considers five to ten covariates. This thesis started by 29 covariates, ending up in a final model with 16 covariates included. The most important covariates were shown to be pressure of payments, the sum of assets and the time as customer at the financial institute. The derived final model showed a high predictive ability and provides insight of significant drivers of default for a Swedish mortgage portfolio. Considering the alarming housing market in Sweden, and the subprime crisis in the U.S 2008, the subject is highly relevant.