FORECASTING INDUSTRIAL CARBON EMISSIONS AND EVALUATING CARBON TAX SCENARIOS IN NIGERIA: EVIDENCE FROM ECONOMETRIC AND MACHINE LEARNING MODELS

Abstract

Industrial carbon emissions have become a major environmental and fiscal concern in developing economies, particularly in resource-dependent countries such as Nigeria. This study investigates the determinants, forecasting dynamics, and tax implications of industrial carbon emissions in Nigeria using econometric and machine learning approaches. Annual time-series data spanning 1990–2024 were employed, drawing from the World Bank DataBank, Central Bank of Nigeria, National Bureau of Statistics Nigeria, and International Energy Agency. The study applied the Autoregressive Distributed Lag (ARDL) model to estimate long-run relationships between industrial CO₂ (Carbondioxide) emissions, industrial output, energy consumption, urbanization, and environmental taxation proxies. Forecasting performance was compared across ARIMA and Long Short-Term Memory (LSTM) machine learning models. Results reveal that industrial output and fossil-energy consumption significantly increase carbon emissions, while environmental taxation and renewable energy adoption reduce emissions over time. The LSTM model outperformed ARIMA in predictive accuracy, indicating superior capability in capturing nonlinear emission patterns. Carbon tax scenario simulations further reveal that a moderate carbon tax regime could reduce industrial emissions by approximately 8–15% while generating substantial fiscal revenue for green infrastructure financing. The study concludes that Nigeria requires a phased carbon taxation framework integrated with industrial transition policies and renewable-energy incentives to achieve sustainable industrialization and climate commitments.