Research On The Forecasting Modeling And Influences Of International Oil Price Changes Considering Multidimensional Uncertainties

As important industrial raw material,bulk commodity and strategic resource,crude oil has commodity,financial and political attributes.The complex attributes result in that the changes of crude oil prices are influenced not only by fundamental supply and demand factors,but also by many other non-fundamental risk factors(such as speculative activities,investor sentiment,the U.S.dollar index,economic policy uncertainty and geopolitical risks,etc),leading to the ups and downs of international oil prices,which have significant impacts on macroeconomic and financial markets.Therefore,improving the forecasting accuracy of international oil price is beneficial for governments and market participant s to make decision deployment and asset management in advance,so as to prevent and mitigate oil price volatility risks.However,the complex attributes of crude oil lead to that the influencing factors of oil prices are various and time-varying,and the predictors of oil prices is characterized by information spillover,which makes oil price forecasting faced with variable selection uncertainty and model estimation uncertainty.How to construct suitable models to dynamically screen important predictors and extract useful forecasting information of variables is the key to improve forecasting accuracy of oil price.Moreover,in the context of global energy low-carbon transition,the dramatic shocks in oil price may affect the government’s low carbon policy pr omotion and the operating conditions of energy firms,which may in turn lead to government climate policy uncertainty and stock price uncertainty of listed oil and gas firms.Therefore,this dissertation focuses on the forecasting of international oil price changes and its impact on government decisions and enterprise development,and tries to address four key scientific questions: Can the shrinkage methods improve the forecasting accuracy of oil price returns? Can the scaled principal component analysis(SPCA)method improve the forecasting performance of oil price volatility? What is the impact of international oil price changes on government climate policy uncertainty? How do international oil price changes affect the stock price uncertainty of oil and ga s firms?Thus,this dissertation proposes various methods for international oil price forecasting which can address the uncertainties in variable selection and model estimation,and further analyzes the impact of international oil price changes on governme nt climate policy uncertainty and stock price crash risk of oil and gas firms.The main research work and findings of this dissertation are as follows:First,multiple regularized based shrinkage methods are proposed to forecast international oil price returns,thus compensating for the traditional shrinkage method’s inability to dynamically adjust the parameter penalty intensity and produce large estimation bias.The results indicate that,shrinkage methods can significantly improve the forecasting accuracy of oil price returns.Within the one-year horizons,the forecasting results of shrinkage methods with non-convex penalty and Huber loss outperform the benchmark and traditional shrinkage methods,including least absolute shrinkage and selection operator(LASSO),elastic net and ridge regression.For the economic valuation,this study constructs mean-variance portfolios considering transaction costs based on various forecasts,and the shrinkage forecasts derived from non-convex penalties can generate the largest economic profits for investors mostly.Particularly,when crude oil prices fall sharply,the statistical results and economic performance of shrinkage forecasts are distinctly better than those of the benchmark model.In addition,imposing sign restrictions on the estimated coefficients of shrinkage models can further improve forecasting performance for oil price returns in most cases.Second,the supervised learning autoregression(AR)SPCA method is designed to downscale variables for the four main categories of influencing factors(macro fundamentals,technical indicators,market sentiment and market uncertainty)of crude oil price volatility and perform out-of-sample prediction.The proposed method can alleviate the forecasting information loss that may arise from shrinkage models when filtering key variables.The results indicate that,the AR-SPCA forecasts for oil price volatility are significantly more accurate than not only benchmark autoregressive model but also competing models,such as principal component analysis(PCA),LASSO,elasticity net and combination forecasting methods,both in terms of statistical and economic significance.Among the four types of variables,macro-fundamental factors deliver the best statistical forecasts for oi l price volatility,with a 14.14% improvement in forecasting accuracy compared to the benchmark,and volatility forecasts based on technical indicators an d market sentiment yield relatively higher economic values.In the economic recession period,the forecasting performance of each kind of predictors on oil price volatility is better than that in the economic expansion period.Third,the time-varying parametric vector autoregression(TVP-VAR)model combined with the DY index is applied to explore the dynamic time-frequency spillover effects of international oil price shocks on the U.S.climate policy uncertainty(CPU),expanding research perspectives for crude oil market and policy uncertainty.The results indicate that,there exists significant spillover effect between oil price shocks and CPU,with the spillover index ranging from 15% to 35%,and the spillover effect increases when major events occur.Crude oil supply shocks and risk shocks have long-term spillover effects on CPU,while CPU affects oil price shocks on a short-term time scale within six months.The spillover effect of oil price shocks on CPU is asymmetric,with the CPU leading to a short-term increase in crude oil prices,while crude oil price declines due to negative demand shocks and positive supply shocks have long-term effects on CPU.Finally,panel fixed-effect model is constructed to explore the impact of international oil price uncertainty on stock price crash risk of Chinese oil and gas firms,which breaks through the limitations of the commonly used return or volatility to portray the extreme risk of stock price and reveals the mechanism and channel by constructing moderating and medi ating effect tests.The results indicate that,the increase in oil price uncertainty leads to a significant increase in stock price crash risk of oil and gas firms by 20.31% to 35.70% in the current period,but its effect is significantly decreased in the next period.When oil price uncertainty increases,firms with better corporate social responsibility performance have higher stock price crash risk,and the positive moderating effect of environmental corporate social responsibility is stronger,which implies the existence of "green washing" phenomenon in China’s oil and gas industry.Retail investor sentiment is an important intermediary channel through which oil price uncertainty affects the stock price crash risk of oil and gas firms.Increase in oil price uncertainty will lead to decrease in optimistic retail investor sentiment,which in turn will exacerbate the stock price crash risk for Chinese oil and gas firms.In general,this dissertation focuses on the forecasting of international oil price s,and the impact of crude oil price changes on policy decisions and corporate value.It proposes multiple forecasting methods to alleviate uncertainty in variable selection and model estimation,and systematically explores the impact of oil price changes on government climate policy uncertainty and stock price uncertainty of oil and gas firms.On this basis,this dissertation proposes some policy suggestions,including that the policy makers and crude oil market investors can choose suitable forecasting models that can handle variable and model uncertainties,so as to grasp the oil price trend and the key drivers of oil price changes;the government should maintain the consistency and stability of climate policies in the face of fluctuating crude oil prices,and strengthen the information disclosure system and green certification system for the green development of oil and gas industry,so as to promote the scientific and orderly advancement of energy transition.