Bayesian assessment of ecological footprint drivers in Finland: a model averaging approach under structural and model uncertainty

This study employs Bayesian regression and Bayesian Model Averaging (BMA) to analyze the determinants of the ecological footprint (EF) in Finland from 1990 to 2023. Using Hamiltonian Monte Carlo (HMC) sampling and model averaging through the Bayesian Adaptive Sampling (BAS) algorithm, the paper accounts for parameter uncertainty, nonlinearity, and multicollinearity in environmental data. The analysis includes key predictors such as GDP, renewable energy consumption (REN), foreign direct investment (FDI), urbanization (URB), and innovation (measured by PA), alongside interaction term FDI⋅URB. Posterior estimates reveal a positive association between GDP and EF, while REN is robustly linked to a reduction in EF. The squared GDP term shows only weak support, leaving the existence of an Environmental Kuznets Curve (EKC) pattern in Finland uncertain. The interaction FDI×URB is negative, suggesting that foreign investment may exert a less harmful ecological impact in highly urbanized settings, though this effect is not robust. Bayesian diagnostics confirm model convergence and predictive reliability, supported by low LOOIC and WAIC values. Model comparison using Bayes Factors shows no substantial evidence favoring complex specifications over simpler ones. The BMA results identify REN as the most robust determinant, whereas GDP and its squared term display only moderate inclusion probabilities, with all other predictors showing weak empirical support. This study provides methodological and policy-relevant contributions by integrating advanced Bayesian modeling with environmental macroeconomics, offering robust insights into sustainability transitions in Finland.