Decarbonization Under Carbon Policy and Pricing Shifts: A Machine Learning-Driven Synthetic Control and Panel Data Analysis

This study evaluates the effectiveness of carbon pricing instruments – Carbon taxes and Emissions Trading Systems (ETS) – in reducing CO2 emissions per capita across 152 countries from 1990 to 2023 using dynamic panel regressions with fixed effects. The analysis estimates both short-run and long-run effects while controlling economic structure, energy mix, and environmental policy stringency (EPS). The results show that ETS policies yield the strongest long-term reduction in emissions at approximately 7.35%, followed by carbon taxes at 5.6%, and EPS regulatory measures at 6.35%. However, when all three instruments are modeled together, the marginal effects of pricing mechanisms weaken, suggesting possible policy interaction or diminishing returns. To reinforce causal identification, the Synthetic Control Method (SCM) is applied to the EU, revealing a growing gap between actual emissions and those of a synthetic counterfactual since the implementation of carbon pricing policies. This machine learning model is used to forecast emissions through 2029, further projecting sustained divergence under the current policies. These findings offer robust evidence that carbon pricing works best when paired with regulatory rigor and institutional capacity. This study contributes a novel empirical approach by integrating econometric modeling, SCM, and predictive analytics to assess the real-world performance of climate policy and instruments, offering insights for policymakers seeking cost-effective paths toward decarbonization.