Exploring Factors Affecting Bitcoin Tail Risk Based On Adaptive-LASSO CAViaR And Forecasting Tail Risk

With the increasing application of blockchain,the types of encrypted digital currencies are also emerging one after another.Encrypted digital currencies are deeply valued by investors due to their decentralization,low transaction costs,and high privacy.In the past ten years,the attention to encrypted digital currency has continued to rise with the advent of the cryptocurrency wave.With the inpouring of speculators into the encrypted digital currency market,prices have fluctuated violently,and the potential financial bubble has also intensified,which makes identifying the risks behind the encrypted digital currency critical.This study,represented by Bitcoin in encrypted digital currency,studies the potential factors affecting the tail risk of Bitcoin based on VaR and builds the Adaptive-LASSO CAViaR framework to predict the VaR of Bitcoin.This study intercepts the daily price data of Bitcoin from the beginning of 2013 to the end of 2021,integrates the existing literature on the factors affecting the tail risk of Bitcoin,and adds a total of 37 variables such as economic policy uncertainty,Baidu Google search engine index,etc.To improve the set of potential variables that affect Bitcoin risk,and explore the impact on Bitcoin tail risk from seven aspects: macro indexes,policy uncertainty,commodities indexes,stock indexes,search trend indexes,market sentiment,and Bitcoin’s own variables.The CAViaR symmetric absolute model and asymmetric slope model proposed by Engle and Manganelli(2004)were used as benchmark regressions respectively,and the effects of different variables on the Bitcoin risk level were studied under three confidence levels of 90%,95%,and 99%.Using the in-sample data,the first step is to filter variables by adding each variable to the significance level of the observed variables in the benchmark CAViaR regression.The second step is to remove the insignificant variables from the latent variable set for joint regression.AdaptiveLASSO CAViaR takes a data-driven approach to the variables of joint regression to screen more influential variables.Furthermore,using out-of-sample data to test the prediction performance of the three methods for future Bitcoin VaR.The research results show that in the symmetric absolute model and the asymmetric slope model at the 90% and 95% confidence level,the risk prediction performance of Adaptive-LASSO CAViaR on the Bitcoin tail is significantly better than that of the benchmark CAViaR model as well as the variable set added CAViaR model.CAViaR model.At the 99% confidence level,the prediction performance of Adaptive-LASSO CAViaR is on par with the benchmark CAViaR model,and better than the CAViaR model with a variable set added.Adaptive-LASSO CAViaR effectively screened out 8 variables with greater influence from the latent variables.Through the analysis of these variables,it was found that as the economic policy uncertainty of China and the United States has decreased,the consumer sentiment in the US market has increased,the capital market become stable,and capital will flow from the crypto market to the traditional market,resulting in an increase in the tail risk of Bitcoin and downward pressure on the price.When events related to Bitcoin occur,the changes in the Google search index and Baidu search index represent an increase in attention to Bitcoin.The increasing concentration on Bitcoin will lead to the participation of speculators,and also increase the uncertainty of the Bitcoin market.Bubble risk accumulates.