China's carbon trading pilot programs are proceeding gradually, and play a positive role in promoting the development of a domestic carbon market. Because of the late start, deficiencies still exist in the understanding, modeling, and exploration of China's carbon trading pilot market. Our goal is to model the carbon price dynamics and explore the regularities of seven city-based carbon trading pilots. Based on the daily carbon price time series, we analyzed the behavior of the carbon log-returns from 2013 to 2018, modeled them with the ARMA–EGARCH–SGED process, assessed the impact of news on volatility, and carried out the backtesting Value-at-Risk (VaR) analysis. Statistical analysis shows that the log-returns are skewed kurtosis, fat tail, non-normal distribution, and have significantly volatility clustering. The estimated models perform well and capture the asymmetric impact of news on volatility. Positive news increases bigger volatility than a negative one, except for Hubei and Chongqing. The Shenzhen pilot shows the best performance in resisting markets’ risk, while lower-capability in that are found in Hubei and Tianjin. Moreover, the estimated models for seven pilots have adequate capital to cope with large unexpected losses and VaR estimates give a good prediction of market risk at 99%, 97.5%, and 95% confidence level.

This study was published on Physica A: Statistical Mechanics and its Applications (2020):1-11.