Stock Market Volatility Modeling And Forecasting Based On LASSO And Markov Regime Switching

Amidst the backdrop of globalization,the full flow of international capital has made the world of today increasingly complex and interconnected.The Coronavirus Pandemic(Covid-19)has increased the complexity of the world’s economic interdependencies.In the post-pandemic era,the correlated effects of capital flight,falling exchange rates,and commodity price volatility have increased uncertainty about the future development of financial markets globally,which has become one of the biggest barriers to global economic recovery.Throughout the world,policymakers,market investors,and academic researchers have been observing the ups and downs of financial markets in order to restart the economy and restore production.To prevent monetary risks and to aid financial market participants,market policy makers,and corporate managers in utilizing volatility models effectively.It has become imperative to increase the ability of volatility models to capture and predict the volatility of real financial markets.There are several national stock markets providing a wide variety of trading platforms for investors and financiers,as well as being closely tied to global economic fluctuations.Further understanding of volatility patterns in developed and emerging markets is of critical importance for financial decision making and portfolio management.Furthermore,the modeling and forecasting of stock market volatility has become an important branch of financial econometrics.Healthy market development is crucial for growing the economy and optimizing the industrial structure both theoretically and practically.Furthermore,academics and practitioners have been discussing how to improve both the accuracy and the applications of forecasting stock market volatility.Computer technology advancements and the expansion of the Internet of Things industries have provided a strong platform for financial markets to benefit from.Access to market data has been greatly enhanced.The advent of massive amounts of data,both macro and micro,as well as low-frequency and high-frequency data,has also presented challenges for academic researchers.There has been a surge in research examining how to use varied data to effectively predict risk.Despite considerable efforts made by econometricians to determine how to accurately assess market price volatility,no consensus exists regarding which indicators are most accurate and influential.Selecting appropriate indicators for the Chinese stock market remains an issue that is worth considering.That is the subject of this dissertation as well.To begin,the heterogeneous autoregressive realized model(HAR-RV)is used as a benchmark.In the study,a mixed-frequency model(MIDAS)is utilized for constructing MIDAS-RV-GARCH,a model for forecasting volatility that is tested with data from SSEC(China),FCHI(France),FTSE(UK),GDAXI(Germany),and SMSI(Spain).In order to account for multiple influencing variables that emerge along with the voluminous information flow,we then combine the sparse method LASSO model with the mixed frequency model(MIDAS).The resulting mixed frequency model is MIDAS-RV-GARCH-LASSO.Finally,volatility is susceptible to sudden changes in structure caused by a variety of factors,such as economic cycles,significant events,and economic policy instability.In this study,we use a Markov model to attempt to objectively represent the true state of volatility in the actual market.Several empirical findings can be summarized as follows.First,this paper uses GARCH to model the residuals of the mixed-frequency model MIDAS-RV,and proposes a new MIDAS-RV-GARCH model to forecast Chinese stock market volatility.Our research compares the benchmark models HAR-RV,HAR-RV-GARCH with the MIDAS-RV Model and its extended model MIDAS-RV-GARCH.The MIDAS-RV-GARCH model is found to be more predictive than the existing models.This model has more predictive capabilities than existing models,and it is more effective at capturing the complex and changes of the Chinese stock market.Second,this study introduces the sparse model LASSO into the MIDAS-RV-GARCH model to accurately characterize the Chinese stock market volatility model,and proposes the model MIDAS-RV-GARCH-LASSO with a LASSO model.Based on the empirical results,the model is not only capable of portraying and predicting the volatility of the Chinese stock market,it can also select the variables that can more effectively affect the Chinese stock market under changing market conditions.Third,the Markov regime switching is introduced into the above model.In other words,the new model contains mixed frequency,sparse model and regime switching.As assumed by the Markov switching model,volatility mechanisms are not easily observable and that data features belong to different mechanisms with different probabilities,switching between them as necessary.The empirical results demonstrate that the FTP-MIDAS-RV-GARCH-LASSO model has a higher econometric advantage over the existing models and their extensions in portraying and forecasting the Chinese stock market.