Research On Volatility Forecasting Based On European Union Carbon Trading Market

The rapid development of the economy is inseparable from industrial production.However,greenhouse gas emissions from industrial production have brought various severe challenges to the global environment.Recently,it should be noted that the fast evolution of the global economy has accelerated global greenhouse gas emissions,resulting in global warming and various extreme weather.The sustainable development of human society has been threatened unprecedentedly.Therefore,to slow down global climate change,realize a low-carbon economy,and promote sustainable development of human society,the carbon trading market arises at a historic moment.In this context,carbon price fluctuations have undoubtedly received increasing attention.Meanwhile,volatility forecasting is regarded as an important tool,which can grasp market trends,optimize asset allocation and improve risk management.In addition,it is of irreplaceable practical significance in improving the carbon financial system,promoting energy saving and emission reduction of enterprises,and further realizing green,low-carbon and sustainable development.Therefore,how to improve the forecasting accuracy of the price volatility of the carbon trading market is extremely urgent.With the launch of the European Union Emissions Trading(EU ETS)and the European Union Allowance(EUA)futures in 2005,carbon emission trading systems have been gradually established and developed rapidly around the world.This kind of emerging trading system is receiving continuous attention worldwide.Meanwhile,the worldwide largest and most mature carbon emission trading market is in Europe.The EU ETS plays an exemplary role.Therefore,this paper focuses on the EUA futures market and carries out a series of forecasting and application researches on carbon price volatility modeling.This paper aims to solve the following questions.First,considering the historical volatility characteristics of EUA futures prices,is it helpful to forecast its volatility by using its short-term and long-term jump,asymmetry,and extreme observations? Second,considering the potential impact of exogenous shocks on the EUA futures market,can four kinds of exogenous factors,including commodityrelated,bond-related,stock-related,and uncertainty-related,be useful to improve the forecasting performance of EUA futures market price volatility? Meanwhile,which type of factor is more predictive? Third,based on the above-mentioned research finding that the uncertainty-related factors own the highest forecasting ability,can comprehensively apply the information of different countries’ economic policy uncertainty(EPU)to improve the forecasting accuracy of EUA futures market price volatility? Fourth,in the context of the rapid development of artificial intelligence technology,can machine learning methods be combined to further extract the predictive information contained in the categorical EPU index to improve the forecasting accuracy? Finally,does the theoretical research mentioned above have practical application value for asset allocation and risk management? This work contains the following contents.Chapter 1 introduces the research background,research content,research significance and research innovation.Chapter 2 reviews the development history,current situation,and prospects of the carbon trading market,and discusses the potential factors that affect the price fluctuations of the carbon trading market,including self-factors and exogenous factors.Three basic models and four out-of-sample testing methods are also introduced.In Chapter 3,a variety of extended models are constructed by combining the short-term and long-term jump,asymmetry,and extreme observations of the EUA futures market,and the forecasting performance of these factors is further discussed.In terms of asymmetry,the extended model with short-term asymmetry,long-term asymmetry,and long-term leverage has a better forecasting effect.In terms of extreme observations,the extended model focusing only on short-term extreme observations has better forecasting performance.In terms of jump information,both the extended model involving short-term jumps and that involving both short-term and long-term jumps perform better.However,the advantages of the abovementioned extended model are mainly reflected in the low-volatility periods.Chapter 4 provides an in-depth study of EUA volatility forecasting by considering exogenous factors,including bonds,commodities,stock indices,and uncertainties.First,it is found that few single factors can successfully predict EUA price volatility.Then,to improve the predictability,this paper further combines the combined prediction method,diffusion index method,and supervised learning method based on the benchmark model.It is found that neither the diffusion index method nor the combination prediction method can improve the forecasting accuracy.However,the machine learning method has a better prediction effect on EUA volatility.More importantly,the empirical results show that the kind of uncertaintyrelated factors is better than other kinds of exogenous factors in predicting EUA futures volatility in most cases.Chapter 5 further explores the forecasting ability of EPU to forecast EUA market price volatility based on the conclusion of the previous chapter.This section mainly discusses the differential influence of different countries’ EPU information on EUA futures price volatility and further conducts medium-and long-term forecast research on it.The study confirms that different countries’ EPU information has differentiated long-term forecasting ability on EUA futures price volatility.In addition,this chapter further discusses how to integrate different countries’ EPU information to improve the prediction effect.The results show that both the diffusion index model and the combined prediction method can produce a considerable longterm out-of-sample prediction effect and are more effective for EUA volatility prediction in low-volatility periods.Chapter 6 comprehensively considers the different policy-related EPU information,and further discusses the influence of categorical EPU information on the price volatility of the EUA futures market.It is found that categorical EPUs have different prediction abilities for EUA futures volatility,and the prediction ability of a single EPU is not robust.At the same time,the results show that the machine learning method based on Markov-switching technology is more helpful to obtain effective information from multiple classification EPU indices and improve the forecasting accuracy.This chapter also makes a supplementary analysis and finds that the machine learning method based on the MIDAS regression framework is superior to the traditional machine learning model based on the AR regression framework in predicting EUA futures volatility.Chapter 7 takes the empirical results of Chapter 6 as an example to investigate the application of the volatility prediction model in asset allocation and risk aversion.It is found that investors with different risk preferences can obtain relatively considerable investment returns in the carbon futures market by using machine learning methods.More importantly,risk managers can use machine learning methods to better formulate risk hedging strategies for European and American stock markets,industrial and transportation stock markets,energy markets,and Chinese stock markets.