Is the energy transition impacting the Eurozone sovereign credit risk? Evidence from Machine Learning

We investigate whether the energy transition can influence sovereign credit risk in the Eurozone using Machine Learning methods. Specifically, we employ Random Forest and Neural Network regression models to analyse the relationship between 10-year government bond spreads and the quantity of renewable energy produced in 10 countries that adopted the Euro as their currency since its inception. To this purpose, we tune the model hyperparameters so that the models forecasts the one-step-ahead bond spreads as accurately as possible, and we use SHapley Additive exPlanation (SHAP) values to examine the relevance of the models features. The results show that Random Forests and Neural Networks display good forecasting performances. As for the energy transition, it shows a moderate influence on the sovereign credit spreads, both when renewable energy is normalized by GDP and when it is considered in terms of rate of change. In addition, higher variations are associated with larger SHAP values, showing that rapid shifts in energy policies may influence the credit spreads. Overall, as may be expected, the energy transition does not play a major role as a driver of the sovereign credit risk; however, it is intersting that it displays some in infuence nonetheless.