Research On Measure And Management Of Dynamic Financial Risk

Accompanied by financial risk all time, financial market has made great progress. It is common problem on how to measure the change and character of financial risk and how to avoid the risk in terms of theory and practice. In this dissertation, the technique of generalized autoregressive conditional density modeling is discussed in detail. After that, its applications in measure and management of financial risk are also studied primarily. The main work and innovations of the dissertation include:1. In the area of volatility modeling about higher moments, GARCD model is discussed comprehensively and deeply. In the dissertation, the system of univariate GARCD has been established preliminarily. To simulate the character of multi-asset, multivariate GARCD model is proposed. Based on Copula technique, the estimation methods for parameters of multivariate GARCHSK model and multivariate GARCD model is analyzed in detail.2. In aspects of definition, formation and character of higher moments risk, the sense of higher moments risk has been demonstrated. The cause of kurtosis is explained by the information flows reach the market in a non-linear fashion. Moreover, the skewness risk is cased by the asymmetry of financial information response. In the end, the relation between higher moments and the first two moments are analyzed in time domain and frequency domain respectively.3. As far as dynamic financial risk spillover is concerned, univariate factor GARCD-JSU model is proposed to describe higher moments risk spillover in worldwide and regional wide. The empirical research is made through multivariate conditional Copula-GARCHSK model and Copula-GARCD model.4. Based on VaR, methodologies to financial management has been studied in detail. Under the framework of the first two moments, dynamic VaR and CVaR are defined, and the model for dynamic portfolio selection is proposed. Under the framework of higher moments, the definition of higher moments VaR is suggested, and the measure of HVaR is demonstrated through JSU distribution.5. Based on PGP technique and utility theory, two kinds of models for dynamic portfolio selection under higher moments are established respectively. In the dissertation, these models are compared with each other from theoretical and empirical view. We find that these models can be solved by the function"fmincon", a tool for nonlinear optimization, in MATLAB software. 6. To our knowledge, high frequency data contains more information than low frequency data. Realized volatility has been extended to realized higher moments. And then, the dynamic portfolio selection model is proposed to disperse the higher moments risk. In the end, the empirical results show that realized higher moments is a good method for risk measure.The research is sponsored by National Natural Science Foundation of China: Research on Long Run Eqilibrium in Multivariate Moments Series and Avoiding Tactics of Dynamic Financial Risk (No. 70471050).