Influences of sentiment from news articles on EU carbon prices

Abstract

In this paper, we calculate the carbon tone index that reflects sentiment in news articles through a self-built dictionary. We study the effect of carbon tone index on carbon price return in the period from September 19, 2017 to October 9, 2020. In addition, we employ the Latent Dirichlet Allocation (LDA) method to explore the differential influences of different topic carbon tone indexes on carbon prices. The market confidence boosted by the approved Market Stability Reserve (MSR) policy led to a continuous increase in volume in 2018. Using two subsample periods divided by the implementation of MSR, we explore the problem whether the increased high volume changed the speed of information absorption in carbon market. Quantile regression with control variables (coal, oil, natural gas, electricity and stock prices) is used to test the robustness of the estimated results. The empirical results show that carbon tone index is closely associated with changes in carbon prices and the efficiency of carbon market is improved after MSR. Finally, we use all carbon tone indexes at the 10% significance level in eight predictive models and show the economic value of the optimal predictive model. In summary, carbon tone index has a strong predictive power for carbon prices.