Forecasting Crude Oil Market Volatility Based On The Time-varying Combined Method And Various Indicators

Energy plays a strategic role in economic development and state security.Particularly,crude oil is the representative one as it is limited and non-renewable.The unexpected fluctuations of crude oil price are bound to have serious impact on national economy,financial markets and even national security.With energy financialization,crude oil and its derivatives obtain much attention from investors and government decision makers.Meanwhile,Oil market volatility is crucial to asset pricing,portfolio selection,and risk management.Therefore,modelling and forecasting crude oil price volatility,and deeply understanding the determinants and predictive indicators of crude oil market are focused by people from various sectors,also these are the challenging issues in the financial academics.In recent years,crude oil fluctuates intensely because of the geo political events,supplydemand imbalance,financial crisis and speculations.In an unusually complex energy system,whether the oil price volatility is predictable? How can we forecast crude oil volatility more accurately and thus hedge price risk? What are the main determinants that affected crude oil market? How about the predictive power of these factors over time? Among a large set of indicators,which ones are the most predictive? This study focuses on these issues and discusses them in detail.The whole study mainly includes three parts: First,how can we improve the forecast accuracy of crude oil volatility with the availabity of high-frequency data? Second,there is a traditional controversy of supply-demand or speculation in determining crude oil price fluctulations.When considering the economic and policy uncertainty,which one has the strongest predictive power? Third,since crude oil system is so complex and affected by a variety of indictors,we select a large set of indicators according to previous theoretical and empirical results.Then,we examine their forecasting performances in crude oil market.The main contents in this study are provided in the following.Chapter 2 considers the time-varying feature of crude oil price and aims to improve the forecast accuracy of crude oil volatility.First,based on the high-frequency data,Chapter 2 uses the heterogeneous autoregressive realized range-based volatility(HAR-RRV)model and its various extensions called HAR-RRV-type models to forecast crude oil volaitlity.Second,since the performances of individual HAR-RRV-type models are unstable over time,we construct combined models with constant weights and combined models with time-varying weights.Finally,under several criteria such as error statistics,the mean mixed statistics(MME),and the model confidence set(MCS)test,the forecasting performances of three types(individual HAR-RRV-type models,combined models with constant weights,and combined models with time-varying weights)are evaluated.Our out-of-sample empirical results show that combined models with time-varying weights can not only generate more accurate forecasts,but also beat individual models and combined models with constant weights.In short,the combined model with time-varying weights performs the best.Chapter 3 aims to identify the most informative determinant in forecasting crude oil market volatility.We use a new GARCH-class model based on mixed data sampling regression and the combined model with time-varying weights constructed in Chapter 2 to examine the predictive power of the determinants.We consider not only the traditional controversial of supply-demand and speculations determining crude oil price fluctuations,but also the economic and policy uncertainty because of the political feature of crude oil.Chapter 3 is the first to integrate both the global economic policy uncertainty(GEPU)indices and several crucial national economic policy uncertainty(EPU)indices with traditional determinants,such as global oil demand,supply,and speculation.Our analysis suggests that the EPU indices comprehensively integrate the information contained in other determinants and they are the most informative and predictive.Specifically,GEPU indices and the U.S.'s EPU index have superior predictive powers for West Texas Intermediate spot oil volatility.The findings highlight the importance of EPU indices,implying that they are the most important determinants to consider in crude oil market.Based on the analysis of Chapter 3,the fourth chapter investigates various indicators in a more sophisticated view.Chapter 4 is the first to explore the effectiveness of a large set of indicators in forecasting crude oil price volatility,and thus it provides an overview of crude oil price fluctuations.Overall,we select forty-five indicators and divide them into three groups: market sentiment indicators(UMS),macroeconomic indicators(MF),and technical indicators(TE).Using an OLS regression,combination forecasts with constant and time-varying weights,and a LASSO regression to forecast oil price volatility,we obtain several noteworthy findings.First,we determine the most effective indicators in forecasting oil price volatility,for example,the uncertainty index is notable in this regard.Second,we observe that in general,combination forecasts and the LASSO regression significantly outperform our benchmark model.In particular,the DMA strategy generates better forecasts than do other strategies.Third,generally,the indicators and the combined and LASSO strategies perform considerably better during recessions than expansions in forecasting oil price volatility.Thus,our study provides evidences regarding which indicators and strategies can improve forecasting accuracy in the oil market.