UK Carbon Price Dynamics: Long-Memory Effects and AI-Based Forecasting

This study examines the price dynamics of the UK Emission Trading Scheme (UK ETS) by integrating advanced computational methods, including deep learning and statistical modelling, to analyze and simulate carbon market behaviour. By analyzing long-memory effects and price volatility, it assesses whether UK carbon prices align with theoretical expectations from carbon pricing mechanisms and market efficiency theories. Findings indicate that UK carbon prices exhibit persistent long-memory effects, contradicting the Efficient Market Hypothesis, which assumes price movements are random and fully reflect available information. Furthermore, regulatory interventions exert significant downward pressure on prices, suggesting that policy uncertainty disrupts price equilibrium in cap-and-trade markets. Deep learning models, such as Time-series Generative Adversarial Networks (TGANs) and adjusted fractional Brownian motion, outperform traditional approaches in capturing price dependencies but are prone to overfitting, highlighting trade-offs in AI-based forecasting for carbon markets. These results underscore the need for predictable regulatory frameworks, hybrid pricing mechanisms, and data-driven approaches to enhance market efficiency. By integrating empirical findings with economic theory, this study contributes to the carbon finance literature and provides insights for policymakers on improving the stability and effectiveness of emissions trading systems.