Forecasting Sovereign Risk in the Euro Area via Machine Learning

Abstract

We test the usefulness of Machine Learning (ML) for sovereign risk assessment and pricing in the euro area along two important dimensions: i) their predictive accuracy compared to traditional econometrics methods and, ii) their assessment on what are the most important economic factors behind market perception of sovereign risk. We find that ML techniques can capture the dynamics inherent in market assessment of sovereign risk in a far more efficient way than traditional econometric models, both in a cross section and time series setting. Moreover, we show that public sentiment about financial news, redenomination fears and the degree of hawkishness/dovishness expressed in the ECB president speeches, rank high as contributors for sovereign spreads. We also confirm that macroeconomic and global financial factors affect sovereign risk assessment and the respective formation of sovereign spreads.

Keywords: Sovereign Risk, Machine Learning, forecasting, euro area, Google Trends, Text Mining, XGBOOST, Support Vector Machines, Neural Networks, Random Forests.

JEL Classification: G01, G21, C53