The Research On Stock Market Volatility Effect Based On Mixed Frequency Data

Financial market volatility is the most important reference factor in modern financial theory,especially in the field of asset pricing and risk management.It plays a vital role in investors’ investment strategy,risk management and financial supervision.As the most important financial market,the stock market has great influence on promoting the healthy development of the national economy and the integration of the world economy.It also has the funds to raise funds for enterprises,transform business management systems,optimize resource allocation and diffusion risks.At present,the stock market has become an indispensable part of China’s market economy.It plays an irreplaceable role in the operation of the entire financial system,significantly affecting China’s economic development and social stability.In this context,it is especially important to effectively prevent risks and ensure the safe and stable operation of the stock market.Accurately depicting the volatility of the stock market is the premise of effectively preventing risks.A deep understanding of the volatility is crucial to ensuring the healthy and stable development of China’s stock market.The study of volatility began in the 1960 s.Early studies focused on the accurate estimation,characterization and nature of volatility and volatility spillovers from a purely time series perspective.A large number of volatility models emerged,laying the foundation for volatility modeling.The study of volatility began in the 1960 s.Early research is to accurately estimate and characterize volatility and volatility spillovers from the perspective of simple time series.During this period,a large number of volatility models emerged,laying the foundation for volatility modeling,such as ARCH,GARCH and various extended models,SV models,high frequency volatility modeling,and so on.On this basis,scholars further explore the root causes of stock market volatility,especially the correlation between stock market volatility and macroeconomics.In the domestic,VAR,VECM and Granger causality test are used to study the co-integration relationship between stock market and macroeconomic variables.The above research has some shortcomings in the use efficiency of data information: the biggest contradiction is the mismatch between high frequency financial data and low frequency macro data.When discussing stock market volatility and macroeconomic relations,Traditional co-frequency modeling methods often lose the effective information of high-frequency data,which is easy to cause defects such as model misuse and estimation bias.In addition,quarterly GDP information is often abandoned because its frequency is too low,but it is obvious that GDP contains important macroeconomic information.Neglecting GDP information results in incomplete and inadequate research results.Secondly,for the characterization of low-frequency historical volatility,the information used is not sufficient,and further consideration should be given to whether high-frequency data information can be added to improve prediction accuracy.Finally,the long-term and short-term components of stock market volatility and volatility spillovers have different sources of influence,and macroeconomic variables have a significant influence on long-term stock market volatility(correlation)component identification and long-term risk measurement,so one should consider decomposing it and extract long-term components,and add macroeconomic variable information to improve stock market volatility(correlation)prediction accuracy,and explore the influence of macroeconomic variables on long-term volatility(correlation).In order to make up for these shortcomings and make more effective use of data information,this paper introduces the idea of mixed frequency model.The mixed frequency model is a kind of Econometric model that directly use different frequency data in the past ten years.It can make full use of the mixed data information to improve the estimation and prediction accuracy.The application in the domestic economic and financial field is still in its infancy.Therefore,this paper takes the characterization of stock market volatility and volatility spillover as the starting point.Based on the mixed frequency data,this paper uses the statistical method to conduct in-depth research on the volatility and volatility effect of China’s stock market,and provides theoretical and practical basis for the volatility,helping the government to establish and improve risk monitoring system.Specifically,based on mixed frequency data,using the mixed frequency model idea to study how to make full use of all available information to improve existing results,accurately estimate and predict stock market volatility and dynamic correlation coefficient.On the basis,research on three aspects of the stock market volatility effect: First,the high-frequency information is used to improve the forecasting accuracy of low-frequency volatility of the stock market,and the relationship between low-frequency fluctuations and returns of the stock market is studied.Second,the low-frequency information is used to improve the prediction accuracy of high-frequency volatility and volatility spillovers and considering its decomposition effect;the third is to explore the root causes of macroeconomic impacts of stock market volatility and volatility spillovers,especially the important role of GDP.These study comprehensively interpret all aspects of stock market volatility effects,and has important reference value for financial supervisory authorities and market investors.Governments and investors should pay close attention to stock market volatility to control stock market risks.The main work completed in this paper and its conclusions are as follows:First,high-frequency information helps to improve the prediction accuracy of low-frequency volatility and explore the relationship between monthly(week)risk and income trade-off of China’s stock market,namely,mixed ICAPM test.In the past,the GARCH-M model or the realized volatility was used and the research conclusions are not uniform.Ghysels attributed the reason to the fact that the volatility is not accurate enough.With the help of GARCH-M and considering the volatility asymmetry,this paper constructs the asymmetric index Almon-MIDAS-M model,and uses the mixing filter to incorporate the daily income information into the monthly(weekly)volatility,which improves the volatility prediction accuracy and ICAPM model test results.The study found that ⑴ high-frequency information helps to improve low-frequency volatility estimation and prediction accuracy.Asymmetric MIDAS achieved volatility is better than standard MIDAS and GARCH filter volatility;Chinese investors have risk aversion characteristics;⑵Shenzhen market risk premium is significantly positive,however,the risk-return relationship of Shanghai stock market is not significant;⑶unlike the long-term lag period of mature stock markets in Europe and America,the actual lag period of China’s stock market is generally short.Second,Using low-frequency information improve the accuracy of high-frequency stock market volatility prediction,and explore the stock market volatility decomposition and its macroeconomic roots.This article comprehensive use of daily returns,monthly volatility,and monthly CPI,M1,interest rates,consumption,investment,Funds outstanding for foreign exchange,exchange rate and quarterly GDP information,improve existing results.This question is divided into three chapters.First,Chapter 4 discusses the effect of low-frequency macro information on low-frequency volatility,and uses month and quarter data to construct a BMF-VAR model,gives its parameter estimation and impulse response diagram,and discusses the economic significance.Chapter 5 discuss the impact of low frequency macro information on high frequency volatility,and uses the GARCH-MIDAS model,the daily stock yield and monthly macroeconomic variables rate and volatility is modeled together,the macro-low frequency factor is introduced,and the long-term volatility is extracted.The effect of macroeconomic variables on the long-term volatility is studied,and the coefficient and economic significance are analyzed and the prediction accuracy are compared.Chapter 6,using the PCA method synthesize macro-variable information into comprehensive economic indicators,improves insufficient of the current consensus index about the lack of quarterly GDP,inflation,and monetary policy information.Based on the GARCH-MIDAS model,the impact of the overall macroeconomic environment on stock market volatility is discussed.The study found that: ⑴First,the low frequency macro information helps to improve the accuracy of high frequency volatility prediction.and the GARCH-MIDAS model is superior to GARCH model in the sample fitting and out-of-sample prediction,the BMF-VAR model is superior to the quarterly VAR model.⑵Second,the study found that macroeconomic factors have significant explanatory power on China’s stock market volatility.The most important factors are GDP and CPI.⑶The GDP has a positive impact on the stock market volatility.a stable economic environment helps stabilize market expectations,enhance market confidence and reduce fluctuations in the stock market.The performance of the Chinese market is generally in line with this trend,But not as stable as the European and American markets.⑷The interest rate 、CPI 、M1 growth rate and volatility have a positive impact on the stock market volatility.Inflation or its expectations will affect investors’ decision-making.The inflation of the real economy will spread to the capital market,causing the capital market to become bigger fluctuation.Frequent changes in currency liquidity will cause stock market volatility to increase;(5)Macroeconomic fundamentals volatility has a positive impact on stock market volatility,once again demonstrating that a stable economic environment is the basis for a stable stock market.Third,the paper explore the volatility spillover decomposition and its macroeconomic roots,Expand from One-dimensional to Multidimensional.Based on the mixed frequency data,Chapter 7 use DCC-MIDAS model to decompose the dynamic conditions of the mainland and Hong Kong stock markets,extract long-term dynamic correlation trends,and explore the macroeconomic roots of long-term dynamic correlation.The study found that ⑴ the correlation is time-varying.With China’s accession to the WTO and the QFII 、QDII system has been introduced,the linkage and dynamic correlation coefficient between mainland and Hong Kong stock market is Upward trend,but the recent“Shanghai-Hong Kong Stock Connect” policy has not significantly changed its dynamic relevance.⑵The volatility of long-term dynamics-correlation is smaller than short-term,and tightening monetary policy promotes long-term dynamic correlation;inflation or inflation expectations rise or frequency change prompts the linkage;the growth rates of GDP,investment have a positive impact on their linkages,while volatility is a negative one.It shows that when the economic environment in the Mainland is generally good and the development is stable,it will promote a long-term linkage.⑶In addition,the basically same situation reflect the close relationship between the Shanghai and Shenzhen stock markets.The innovations of this paper are mainly reflected in the three aspects:1.Based on the traditional discussion of ICAPM model using GARCH-M family model,this paper introduces the idea of mixing model,and uses mixed MIDAS filtering to incorporate high-frequency daily yield information to construct the volatility of mixing and achieve domestic adjustment.Furthermore,considering the asymmetric characteristics of the stock market,the volatility of the asymmetric index Almon weight mixing has been constructed,and the fitting efficiency is improved compared with the GARCH filter volatility.The asymmetric index Almon-MIDAS-M model can improve the test results of the ICAPM model.2.Based on the GARCH and DCC methods,this paper introduces the idea of mixing filter and component volatility model decomposition,and uses MIDAS filtering to incorporate the low-frequency realized volatility or correlation coefficient into the GARCH or DCC model as a low-frequency factor to construct GARCH-MIDAS and DCC-MIDAS model with the advantage of decomposing stock market volatility(correlation coefficient)into short-term components and long-term components,and can use MIDAS filtering to incorporate macroeconomic variables as low-frequency factors into long-term components,and by increasing low-frequency(macro-variable)data information,the accuracy of high-frequency stock market volatility in the sample fit and out-of-sample prediction is improved.3.This paper has several improvements in the macroeconomic roots of stock market volatility and volatility spillovers.⑴First,this paper focuses on the second-order moment level,that is,the variance level.Previous studies focused on the first-order moment,that is,the mean level.⑵Second,this paper makes full use of GDP information,confirms the important influence of GDP on stock market volatility,and makes up for the lack of discussion on quarterly GDP in previous studies.⑶The third is to build monthly comprehensive economic indicators based on mixed data,especially to add quarterly GDP information,improve the lack of quarterly GDP,inflation and monetary policy information,and based on the GARCH-MIDAS model,the impact of the overall macroeconomic environment on stock market volatility is discussed.⑷The fourth is to examine the characteristics of long-term dynamic correlation and short-term dynamics of domestic and Hong Kong stock makets in stages,and to analyze the long-term dynamics of macroeconomic impacts by adding macro-variable information.The application of DCC-MIDAS model in domestic financial market is rare,its research conclusions have reference value.