No. 1078 - Machine learning methods for the prediction of municipal financial distress: evidence from Italy
This paper compares statistical and machine learning (ML) models to predict financial distress in Italian municipalities, defined by entry into pre-default or default procedures. By combining final accounts and budget forecasts with daily data on treasury advances, it examines the trade-off between predictive accuracy and timeliness to strengthen early-warning systems. Model interpretability also helps identify key risk factors and provide useful insights for public-sector oversight.
Compared with the system established under current regulations, the best-performing ML models more than double the ability to promptly identify municipalities at risk of financial distress, partly by capturing complex interactions among financial indicators. Key risk signals include indebtedness, reliance on treasury advances, and weak investment spending.
Instagram
YouTube
X - Banca d'Italia
Linkedin
RSS
E-mail Alert
