Neural Networks as a Research Tool for Environmental Market Forecasting

This poster presents neural networks as a practical research tool for analyzing and forecasting environmental markets, using U.S. carbon pricing as a case study. Environmental and policy-driven markets often suffer from fragmented data sources, limited historical depth, and nonlinear interactions with macroeconomic forces, making traditional econometric approaches difficult to apply consistently. The study introduces a synthesized United States Carbon Allowance (USCA) index constructed from three major carbon market programs: California’s Cap-and-Trade Program (CARB), the Regional Greenhouse Gas Initiative (RGGI), and the Voluntary Carbon Market (VCA). A feedforward neural network is trained on monthly data from 2015 to 2025, incorporating lagged index values and key macroeconomic indicators, including energy prices, financial markets, inflation, and interest rates. Rather than emphasizing model complexity, this work highlights how neural networks function as a flexible research instrument. The approach integrates heterogeneous datasets, captures nonlinear relationships, and enables forward-looking inference in data-constrained environments. The model achieves strong predictive performance while remaining computationally lightweight, demonstrating that meaningful environmental market insights can be generated without specialized hardware. The poster emphasizes broader research implications and shows how neural networks extend the analytical toolkit available to researchers studying environmental, economic, and policy-linked systems.