Italian NEET rates in the 2012–2021 years: comparing different forecasting techniques

Abstract For many years, Italy has reported the highest NEET rates among EU countries—that is, the share of young people Not in Employment, Education, or Training. Between 2012 and 2021, several events influenced Italy’s NEET rates, including the recovery from the 2007–2010 financial and economic crisis, the COVID-19 pandemic, and recent labour market reforms, such as the Youth Guarantee and the Citizenship Income. Furthermore, the digital and green transitions have significantly reshaped labour market dynamics, affecting both how individuals work and how they interact with one another. Although NEET rates have declined across Europe, the reduction in Italy has been less marked compared to the EU-27 average. Understanding the determinants of these dynamics and producing accurate forecasts is essential, as it can help identify best practices to accelerate NEET reduction. Traditional statistical analyses can only partially meet these needs, due to the complexity of the phenomenon and the underlying assumptions, such as the linearity relationships. In this paper, in addition to employing a modern time series approach, we adopted three advanced statistical techniques from the field of machine learning: regression trees, random forests, and XGBoost algorithms. The findings provide a clear identification of the main NEET related factors, while the analysis of the predictive accuracy of the models in an out-of-sample framework reveals heterogeneous results, with the random forest performing slightly better overall.