Research On Identification And Forecast Of Crude Oil Price Influencing Factors Based On Dynamic Model Average

As a source of various industrial products and a strategic resource for countries,crude oil price fluctuations will have a huge impact on the economies of countries around the world.For example,on April 20,2020,West Texas light crude oil,one of the most important pricing standards for international oil prices,fell below negative values for the first time,falling as low as-$37 per barrel.Its price collapse drove the crude oil market down,causing global oil prices to plummet.The fall in oil prices dragged down global stock markets,especially the energy sector,and the fall in energy prices further exacerbated the shocks in global stock markets.Of course,China,the world's top crude oil importer,was inevitably hit by the drop in oil prices.First,the BOC Crude Oil Treasure,which is linked to the West Texas Light Crude Oil futures contract,caused its investors to lose 30 billion yuan,greatly dampening Chinese investors' confidence.Second,the fall in crude oil led to a drop in the prices of various domestic products,bringing down China's CPI and PPI indices and leading to a decline in confidence in consumption and production.However,with the automatic production cuts and demand repair by U.S.crude oil companies and the fall of the U.S.dollar index in late May,international oil prices fell back significantly and the rise in oil prices contributed to the improvement of China's economy.Therefore,identifying the key influencing factors of international crude oil price fluctuations,constructing a crude oil price forecasting model and analyzing crude oil price trends can not only enrich the theories related to crude oil price formation mechanism,but also provide new solutions for the forecasting model of international crude oil prices.It also provides a possible basis for China to predict the international crude oil price trend,avoid the world economic changes brought by crude oil prices,and formulate policies and regulations to avoid risks.The mechanism of international crude oil price formation is more complex and influences a variety of factors.Therefore,using the existing literature,firstly,fundamental factors such as supply and demand,inventories,and imports are identified.Secondly,with the development of financial markets,stock price indices,U.S.dollar indices,and technical indicators also affect crude oil prices.Then,with the development of big data as well as programming techniques,economic activity,policy uncertainty,geopolitics,and investor concerns can be measured,specifically as being global economic activity,economic policy uncertainty,geopolitical risk,and search indices.Finally,the classical safe-haven assets such as gold,as well as digital currencies,which were born with the development of computer technology and the Internet,have increased their influence on crude oil prices.Meanwhile,combining the existing literature,it is found that traditional forecasting models,mostly static,do not allow variables and variable coefficients to change over time and cannot use a larger number of variables.To overcome these problems,this paper uses a dynamic model averaging approach to analyze crude oil price trends and changes in factors affecting oil prices.In addition,in order to compare the dynamic forecasting effect of the classical safe-haven asset gold and the "new gold" digital currency on crude oil prices,the dynamic model averaging method is used to estimate their forecasting probabilities and analyze them in conjunction with the return on crude oil.The empirical analysis leads to the following conclusions: first,with the same predictor variables,whether gold or digital currency is added,the dynamic model selection model with forgetting factor ? and ? of 0.99 is the optimal model compared with the traditional forecasting model.Second,the forecasting effects of both the dynamic model average model and the dynamic model selection model are influenced by the forgetting factor,but the results of the optimal forgetting factor values for the dynamic model average model and the dynamic model selection model are not the same.Third,the number of predictor variables required to forecast crude oil prices varies dynamically based on the same predictor variables,whether gold or digital currency is added.Fourth,traditional predictor variables are still good predictors,but the predictive effectiveness of technical indicators has largely trended upward since2015,with increasingly better predictions.In addition,gold and digital currencies also have some predictive effect.Fifth,gold is more suitable for predicting the case of rising oil prices,and digital currency is more suitable for predicting the case of falling oil prices.Finally,the above empirical results are combined with corresponding suggestions from three perspectives: policy makers,crude oil-related enterprises and market investors,respectively.