Decoupling Economic Growth from CO2 Emissions in Honduras: A Longitudinal Machine-Learning and Econometric Assessment of Low-Carbon Development, 1990–2023

Decoupling economic growth from CO2 emissions is a key challenge for developing economies, with limited evidence for small Central American economies. This study evaluates Honduras during 1990–2023 using GDP per capita, CO2 emissions per capita, energy intensity, and renewable energy share from World Bank Indicators and the Global Carbon Project. The methodology integrates k-means clustering, PELT structural break detection, the Tapio decoupling index, Environmental Kuznets Curve (EKC) modelling, Granger causality, and Random Forest analysis. Since energy-intensity data are available only from 2000, a full GDP-CO2 series (1990–2023, n = 34) is distinguished from a complete four-variable panel (2000–2021, n = 22). Clustering identifies two structural regimes rather than three (silhouette 0.490 vs. 0.476). EKC results support an inverted-U relationship (β2 = −3.012, p < 0.001; adjusted R2 = 0.766), with an estimated turning point of USD 2388 (95% CI: USD 2156–2644, delta method), near the upper boundary of the estimation sample, which Honduras’ 2023 GDP per capita (USD 2527) marginally exceeds as an out-of-sample extrapolation. However, Granger tests find no significant temporal precedence between GDP, renewable share, and CO2 emissions (all p > 0.05), and Random Forest shows GDP per capita (%IncMSE = 31.47) vastly outweighs renewable share (%IncMSE = 0.16). Honduras has likely crossed the EKC threshold, but evidence does not support a renewable-driven decoupling.