Research On Financial Risk Measurement Based On Analysis Method Of Mixed Frequency Data

In recent years,with the rapid development of economic globalization,the relationship between various economies has become increasingly close,and the dependence between markets has gradually increased.With the rapid development of financial liberalization,financial innovation is emerging one after another,and the financial system is becoming increasingly vulnerable.All these will lead to the financial crisis.Frequent outbreak of financial crisis also makes people realize the importance of financial risk management.Micro prudential supervision and macro prudential supervision are two important supervision methods in modern financial risk management.They complement each other and are indispensable.Under this two supervisions,financial risk measurement is the most critical.At present,how to provide accurate financial risk measurement is not only an important issue for regulatory authorities,but also an important topic for scholars.In the process of financial risk measurement,due to the inconsistent observation frequency of risk factors from different sources,there is the problem of mixing frequency data,in other words,the data frequency of explanatory variables and explained variables or between explanatory variables is inconsistent.The traditional method to deal with this issue is to aggregate high-frequency data into low-frequency data or interpolate low-frequency data into high-frequency data.But these two methods both will produce large errors,resulting in inaccurate risk measurement results.Therefore,people need to study new methods to measure financial risk.In this regard,this thesis selects the subject of "Research on financial risk measurement based on mixed frequency data analysis ",and introduces the mixed data sampling model(MIDAS)into the both micro prudential supervision and macro prudential supervision.The MIDAS structure can modeling directly use the mixed frequency data without changing the frequency of the original data by using frequency alignment.The MIDAS overcomes the shortcomings of the traditional methods well.Therefore,this thesis uses the MIDAS structure to deal with the mixed frequency data in measuring the financial risk.From the perspective of micro prudential,this thesis constructs ER-MIDAS model and JE-MIDAS model to respectively measure the financial risk indirectly and directly.From the perspective of macro prudential,this thesis proposes DCC-MIDAS-t model for financial systemic risk measure.The detailed researches and main innovations of this thesis are as follows:(1)The Expectile regression mixed data sampling(ER-MIDAS)model is proposed,and a new method to measure VaR and ES is given.In order to make full use of the information contained in high frequency data,this thesis proposes a ER-MIDAS model by introducing the MIDAS structure into Expectile regression model and gives the model expression,parameter selection and model estimation method.Then,the numerical simulation experiment is designed to evaluate the accuracy of ER-MIDAS model in measuring VaR and ES,and compares the performance of it with traditional models:RiskMetrics,GARCH-t,GJRARCH-t,Realized GARCH-t model,and CARE.The results imply that the high-frequency information can help to improve the accuracy of risk measurement.Finally,the weekly risk of Shanghai Composite index,S&P 500 index and FTSE index are measured by ER-MIDAS model.The empirical results show that ER-MIDAS model is suitable for the normal circumstances without outlier events and risk-averse investors.(2)A joint elicitable mixied data sampling(JE-MIDAS)model is proposed,and a new method to measure VaR and ES is given.The joint elicitable regression model can get the estimation of VaR and ES at the same time,which realizes the directly measurement of ES.This thesis establishs a JE-MIDAS model by introducing the MIDAS structure into the joint elicitable regression model.The proposed JE-MIDAS model can directly use high-frequency data to measure low-frequency risk and improve the accuracy of risk measurement.This thesis gives two kinds of model expressions and introduces the methods of parameter selection and model estimation.Through numerical simulation,the JE-MIDAS model is competitive with GARCH-t model,Realized GARCH-t model,JE-AL model and ES-CAViaR model in measuring VaR and ES.The empirical study on the Shanghai Composite index,the S&P 500 index and the FTSE index shows that market volatility will increase the risk.Interest rate has a positive impact on the risks of the S&P 500 index and the FTSE index,while it has a negative impact on the Shanghai Composite index.(3)Modify the distribution assumption of DCC-MIDAS-N model,and propose the DCC-MIDAS-t model to measure the Co VaR and COES.Considering the characteristics of financial time series with sharp peaks and fat tails,this thesis modifies the original normal distribution assumption of DCC-MIDAS model to student t distribution,constructs DCC-MIDAS-t model,and applies the modified DCC-MIDAS-t model to measure CoVaR and COES.Firstly,the estimation method model and the method of measuring CoVaR and COES with the DCC-MIDAS-t are given.Secondly,this thesis measures the systemic financial risk of China's banking industry.It selects the period before and after the crash of China's stock market in 2015 as the research period,and uses macroeconomic variables such as industrial production growth rate,money supply and producer price index to measure the systemic risk of the banking industry.It also discusses the volatility,correlation and risk spillover of banks before,during and after the crisis.This thesis extends the mixed data analysis model to the field of financial risk measurement,develops new risk measurement models,and gives model representation,parameter selection and model estimation,which enriches the theoretical research of financial risk measurement and the application research of mixed data analysis model.At the same time,applying the new mixed data risk measurement model to measure the risk of international important stock indexes and China's banking industry will help investors to grasp the trend of risk and make correct investment strategies.It will also help financial institutions and regulators to improve the ability of risk management and policy-making.