Study On Financial Market Risk Based On GARCH Models

In this dissertation, the main purpose is exploring risk transmission mechanism in which among measure of risk of financial asset portfolio and financial assets, by means of applying the ARCH (GARCH) model to modeling volatility of such financial assets. For this purpose, I developed a series of single variable ARCH (GARCH) models that can handle better risk assets which follows fat-tail, long memory distribution and with financial leverage feature. In the same time, such single variable models can improve calculation of financial assets value. In another circumstance of ARCH (GARCH) model cluster, multiple variable ARCH (GARCH) models, under the aim of increase accuracy, that I sort its structures, parameter estimates and parameter test methods, it should offer effective analysis instrument of measuring risk of portfolio, predict risk, relevance analysis among financial assets and mechanism of risk transmission.The structure of this dissertation was composed of following chapters:Chapter I is introduction. I defined research question, and summarized background and present situation of this research. In addition, I described theoretical and practical meaning of the research, then I gave out explanation for structural arrangement, and point out innovation of this dissertation.Chapter II recall for the theory of financial risks. Derived from overall reviewing of research achievement and correlated preference, I summarize financial risk defines, financial risk management and measuring method of risk. Especially, it is key of risk management of financial risk that measuring of Value-at-Risk (VaR), and exploit GARCH model calculate VaR is main purpose of this research. In this chapter, also reviews scientific literature of transmission of risk, which was theoretical groundwork of modeling risk transmission by application of GARCH model.Chapter III reviews the theory of the GARCH model as well as emphasizes the development of the GARCH model. There are two parts in this segment. One part of work was categorized single variable GARCH model's transmutation. Another is introducing of multiple variables GARCH model's structure, parameter estimation and testing methods.Chapter IV developed single variable GARCH model and its empirical analysis. Although the series data generated by one single variable GARCH model is fat-tailed, but not good enough to simulate fat-tailed and non-symmetrical feature of financial time series data. To dealing with this problem, I suppose various distribution hypotheses, especially skewed student t distribution, and then combined single variable GARCH model which introduced in previous chapter, empirical analyze VaR of China's stock market.Chapter V is study of selection of multiple GARCH model. Model selection is delicate work. In this segment utilization of data from ten indexes of trade of the HUSHEN300 index, by the"fitting→evaluating→predicting"steps, empirical analyze common GARCH models. It should bring us such benefits that steps of choosing GARCH model.Chapter VI discuss measuring styled portfolio's VaR that based on GARCH model as well as the studying relationship between market risks and dynamic correlation among the portfolios. Here exploit DCC-MGARCH model which one of the multiple variable GARCH models empirical analyze relationship in which between market risk and a few style of investment which exist in the Chinese market.Chapter VII explores risk transmission between stock market and institutional investors. In this chapter use FULLBEKK-T model, build two binary variable VAR (3)-BEKK (1,1,1) models in which provide the volatility spillover of the fund index and the A-stoke index. In the same time, take advantage of time-varying variance and covariance matrix of multi-GARCH model calculate the time-varying correlation coefficient between the Fund's index and the A-stock Index. In the same way, calculate risk factors, and study of risk factors under different market conditions. Chapter VIII is conclusion of dissertation. Here summarizes result of empirical analysis which based on the GARCH model. From the output of calculate VaR, obtained best measuring VaR model of financial assets. Analyzed financial assets volatility transmission mechanism based on multiple GARCH model.In this dissertation may include following innovations:1. Confirm skewed student t distribution hypothesis FIAPARCH model is excellent to fitting financial assets value in which yield follows fat-tailed, non-symmetrical distribution and long memory volatility. Also find out that FIAPARCH model is effective to predicting and measuring VaR of single financial asset.2. First of all, I introduced forms of common used variant multiple variable models, then proposed method of test for evaluate and predict. In advance, follow the " fitting→evaluate→predict " steps to choose MARCH model in which better convergent and good prediction. As a result, be acknowledged that the simply structured model's prediction ability is better, but time-varying correlation coefficient model's prediction ability is not good enough that can reflect dynamic relationship among assets.3. I Exploit DCC-MGARCH model modeling portfolio volatility model, and discuss parameter estimation, test for evaluate and application of VaR. benefit from these work, I obtained higher accuracy prediction of the VaR of portfolio.4. Depends on historical data of relationship of domestic and foreign institutional investors, I study institutional investors who involved in China's stock market, use t distribution hypothesis FULLBEKK model, effort to analyze relationship between index of market and overflow of mean and variance of the index of institutional investors asset. It provide me research point of view that explorer risk of transmission mechanisms between institutional investors and the market.The point of further improvement:Although I choose different style of index of investment and index of fund's assets to do comparative empirical analysis, but subject to length of series dada, the sensitive of data length needs further improvement.