Integrating Home Energy Efficiency into Credit Risk Models: A study with real-world data from European Datawarehouse(EDW)

This thesis analyzes the relationship between real estate energy efficiency (EE) and mortgage credit risk using a large dataset from the European Datawarehouse (EDW). Through a dual approach combining interpretable statistical models (Cox survival analysis, logistic regression) and advanced machine learning (LightGBM), it is demonstrated that EE is a robust and statistically significant predictor of default. The results are clear: properties with low energy efficiency have a 43% higher risk of default compared to those with high efficiency, even after controlling for traditional risk factors. Integrating the EE variable significantly improves model performance, with the LightGBM algorithm achieving a much higher predictive accuracy (AUC of 0.83) than logistic models (AUC of 0.69), proving its ability to capture complex non-linear relationships. Therefore, this research provides strong empirical evidence that including energy efficiency in credit risk models is a pragmatic strategy to improve the accuracy of assessments, aligning financial incentives with sustainability goals. ​