Environmental impacts of economic growth: A STIRPAT analysis using machine learning algorithms

• Non-linear impact of economic growth on CO 2 emissions, water, forests, biodiversity by income bracket are studied. • ML algorithms: ridge, LASSO, elastic net & kernel regressions are used. • Better institutional quality does not necessarily translate to better environmental outcomes. • GDP and population growth harm all dimensions. Renewable energy aids emissions, water, forests. • U, M-shaped EKCs observed using a novel method for four impacts. This study examines the environmental consequences of economic growth using four key dimensions – carbon dioxide emissions, freshwater availability, forest area and biodiversity, and introduces a novel methodology to verify the environmental Kuznets Curve (EKC) using pointwise derivatives. Built upon an extended STIRPAT (Stochastic Impacts by Regression on Population, Affluence, and Technology) framework, we expand the original model to include sectoral growth, trade openness and quality of government institutions. Given the high dimensionality of the data and potential nonlinearities in relationships, we employ advanced machine learning regression techniques like ridge, LASSO, elastic net and kernel-regularised least squares (KRLS) to perform segmented analyses across low, lower-middle, upper-middle and high-income countries. Our findings reveal complex, income-dependent environmental impacts, with growth in lower-income groups typically worsening environmental quality and resource depletion, while wealthier countries suffer reduced environmental strain. While the inverted U-shaped EKC is observed in most cases, more complex patterns, such as an M-shaped curve, emerge as countries progress economically. This study provides valuable insights to inform targeted, region-specific sustainability strategies.