The relationship between the contaminating industries and the European carbon price, machine learning approach

This study explores the bidirectional relationship between the industrial stock returns of greenhouse gas (GHG) emitting industries in the European Union (EU) and the European Union Allowance (EUA) price for the period of 2015-2022 by applying novel supervised machine learning (ML) models combined with a granger vector auto regression (VAR) test. Research findings present bidirectional causality and factor significance scores for each industry index in addition to performance metrics of the applied ML models. Results identify the chemicals, renewable energy, meat, electricity, iron&steel, transportation, construction, manufacturing, and cement industries as the most significant factors in explaining the EUA price volatility, while the Extra Trees (ET), KNearest Neighbor (KNN), Recurrent Neural Networks (RNN), and Random Forest (RF) models are determined to be the top performers. From the political perspective, research outputs conclude that the current carbon price is below the marginal decarbonization cost and fails to stimulate the technological decarbonization investments for the majority of the contaminating industries of the EU.