Forecasting unemployment in the euro area with machine learning

Unemployment has a direct impact on public finances and yields serious sociopolitical implications. This study aims to directionally forecast the euro-area unemployment rate. To the best of our knowledge, no other studies forecast the euro-area unemployment rate as a whole. The data set includes the unemployment rate and 36 explanatory variables, as suggested by theory and the relevant literature, spanning the period from 1998:4 to 2019:9 in monthly frequency. These variables are fed to three machine learning methodologies: decision trees (DT), random forests (RF), and support vector machines (SVM), while an elastic-net logistic regression (logit) model is used from the area of econometrics. The results show that the optimal RF model outperforms the other models by reaching a full-dataset forecasting accuracy of 88.5% and 85.4% on the out-of-sample.