Replication data and code: Spatial spillovers and nonlinear determinants of PM2.5 in the European Union (EU-27, 2000–2020)

This dataset accompanies the article “Spatial Spillovers and Nonlinear Determinants of PM2.5 in the European Union: A Hybrid Spatial Econometric and Machine Learning Approach” (Chovancová & Petruška). It contains a balanced panel of 27 EU member states for 2000–2020 (567 country-year observations) with PM2.5 exposure and eight candidate determinants (energy intensity, renewable energy share, Human Development Index, urbanization, population density, industrial value added, vehicle ownership and consumption-based CO2 emissions). The series are compiled from the World Bank World Development Indicators, Our World in Data and Eurostat; sources and definitions are given in the data dictionary. Vehicle ownership has 12 missing observations. The package also includes the spatial weights matrix (inverse great-circle distance between capital cities), capital coordinates, Stata code for the random forest, SHAP and partial dependence analysis and for the Moran test for spatial dependence, and Python code for permutation Moran’s I and Lagrange multiplier tests. See README.md for contents and instructions.