Nonlinear dynamics of energy prices, renewable transition, and emission intensity in the European union: evidence from a deep neural network panel model

The European Union has committed to cutting its greenhouse gas emissions by at least 55 per cent before 2030 and to be climate-neutral by 2050. In this regard, the EU's green transition agenda focuses on renewable energy growth, carbon pricing, green innovation, and energy efficiency. The present research is inspired by the fact that the traditional linear panel frameworks cannot explain the nonlinear, dynamic, and cross-country relationships in which renewable deployment and technological change expose energy prices and the level of emission-reduction efficiency in the EU. We test the hypothesis that the share of renewable energy, the price of carbon, green investment, and green innovation have a mitigating impact on emission intensity, and that the dependence on fossil fuels and concentrated industrialization aggravate it. We also assume that such relationships are nonlinear and both cross-national and time-varying. Our methodology combines a deep neural network panel model with fixed effects and SYS-GMM, along with CS-ARDL and common correlated effects estimators, within an extended STIRPAT framework to study 25 EU countries during 2005-2023. The findings indicate that increased renewable energy, a higher carbon price in the ETS, green investment, and green innovation always reduce emissions intensity. In contrast, an increased share of fossil fuels and a more concentrated industrial structure increase it. The deep neural network model outperforms the benchmark estimators and reveals more nonlinear interactions among policy and technology, as well as among the market variables.