Forecasting risk of European carbon emissions trading market with DySco-SKST model

Abstract

Influenced by energy prices, macroeconomic factors and policy factors, the price of the European carbon emissions trading market has changed significantly. Effective measurement of price risk of the European carbon emissions trading market is of practical importance for market participants. Two measures Value at Risk (VaR) and Expected shortfall (ES) are used to assess price risk. Based on the dynamic score (DySco) model and skewed Student-t (SKST) distribution, the DySco-SKST model is constructed and used to predict the price risk. The unconditional/conditional coverage tests, dynamic quantile test, Actual over Expected ratio, mean and maximum Absolute Deviation, quantile loss and FZ loss are used to evaluate the VaR and ES prediction performance. The daily spot closing prices of EUAs from January 3, 2013 to November 23, 2018 are used. The empirical results show that the VaR and ES prediction performance of the DySco-SKST model is better than that of the DySco-N and DySco-ST models. The VaR and ES prediction performance is affected by the parameter re-estimation scheme, but not by the parameter re-estimation frequency. The DySco-SKST model is better at predicting VaR and ES under the rolling window than under the expanding window.