Is Local Taxation Predictable? A Machine Learning Approach

This paper investigates determinants of local tax policy, with a particular focus on personal income tax rates in Italian municipalities. By employing seven Machine Learning (ML) algorithms, we assess and predict tax rate decisions, identifying Random Forest as the most accurate model. Results underscore the critical influence of demographic dynamics, fiscal health, socioeconomic conditions, and institutional quality on tax policy formulation. The findings not only showcase the power of ML in enhancing predictive precision in public finance but also provide actionable insights for policymakers and stakeholders, enabling more informed decision-making and the mitigation of fiscal uncertainties.