| Finance is the core operation of market economy, the sustained and healthy development of market economy is inseparable from the stability of the financial system. How to build the integrity of the financial system and guard against financial risks has been an important issue of academic research.2007--2009 outbreak of the US subprime mortgage crisis, soon evolved into a global financial crisis. Compared with the previous financial crisis, the scope of the financial crisis is broader and deeper, killing since the 1930s "Great Depression" the most serious global financial and economic crisis. International academic community generally agreed that lack of accumulation and macro-prudential supervision of systemic financial risk is one of the important factors that led to the severe financial crisis. After the crisis, the world’s financial regulators have introduced measures to strengthen supervision of the financial system systemic risk. In China, high homogeneity of financial institutions, while increasing the degree of financial innovation and liberalization of lending to each other, secured between financial institutions, mutual financing facility so intricate relationship between financial institutions, financial institutions, financial Contact between departments more closely, and the market is still a lack of adequate risk control tools and trading mechanisms, regulation of financial and technical means backward. so the risk of facing the future of China’s financial system is also gradually increasing.Network structure model is quite cutting-edge research and concern systemic risk approach. Network structure model IMF (2009) proposed for the main bank, its network based on mutual exposures and transactions on the basis of inter-bank balance sheets, depending on the circumstances and inter-bank network simulation shape the market risk of cross infections, which estimates the accumulation of systemic risks in each bank network. In contrast, in the entire financial system, the bank’s security than other non-bank brokers and other financial institutions, non-bank financial institutions can not pass between the banks’balance sheets and transactions join the network structural model. At the same time, in practice, Barra factor model (2010, Nielsen & Bender) using at least 20 industrial factors trying to get together in the same industry fluctuations, as a company in the same industry tend to fluctuate together. However, not all companies in the same industry in the same direction are fluctuations or fluctuations in accordance with the same amplitude. Therefore, further study towards a common corporate level fluctuation is necessary.On the other hand, some studies have found that the residual asset pricing income and residual income stocks often linked to other stocks, indicating that interaction or independent structure between the companies. But usually as the residual white noise, to eliminate residual risk through hedging, there is little research to study the interaction between the company by mutual association between residual income, in reality, a residual risk can not be ignored risk, interconnected between the residual income can provide a study of risk research ideas in the spread between the companies. Therefore, this article mainly from the perspective of the residual income interrelated to study the situation of China’s financial network risk of infection.Research Framework herein by reference Yi-An Chen (2014) Risk factors for network theory and its empirical research during the US subprime crisis to do, the risk of network theory to the Chinese market, and the stock market crash occurred in June 2015 for experience research demonstrates the effectiveness of risk theory network in the Chinese market, and the role of risk in this network played in the crash. The specific method is to construct a network capable of capturing network risk factor in a simulated securities portfolio. As used herein, the central measure the number of network structure to the right network factor weights securities portfolio. The structure of each measure by the central node interconnected to calculate the score for each node. This article will involve two calculation methods center Measure:Measure the degree of centrality and eigenvector centrality measure. Eigenvector centrality measure compared to measure the degree of centrality is an upgrade, we use eigenvector centrality measure, because it calculates a score for each node by the recursive method, if a node around it has a high score, then to give it a high score. When network risk factor reference is constructed by the least squares estimation (LSE) for each stock can be drawn with respect to the risk of cyber-risk exposure. Based on this information, the residual income Fama-French three-factor model obtained can be decomposed into individual network risk earnings and net earnings risk, then the risk factor models can easily be broken down into the general form of the measurement of risk, such as standard deviation, VaR and ES.Based on the above method, the whole article is divided into five chapters, the specific arrangements, Chapter 1 is an introduction, mainly introduces the research background, research significance, research methods, research and innovation. Chapter 2 is the literature review, starting with the proposed Fama-French three-factor model, the lack of verification of several aspects of the three-factor model, the domestic three-factor model as well as some of the study on the effectiveness of the three-factor model of previous literature reviewed; secondly for some research on systemic risk and the risk of infection were analyzed.Chapter 3 of the study design, in this part of the thesis is mainly to build the model, first introduced Fama-French three-factor model of obtaining residual income, followed by the construction of a network of risk factors detailed analysis, followed by the weighting function and two kinds of network-centric approach (degree centrality and eigenvector centrality) network architecture are discussed; then factor model of risk analysis decomposed by a variance-covariance matrix and risk; and finally to reflect the network risk time-varying characteristics of the above added to the base model dynamic conditions correlation model (DCC) will extend the model to the conditions of the network risk model.Chapter 4 of the empirical analysis, we selected data is relatively complete 36 financial institutions in weekly data January 1,2011 to December 31,2015 from a research sample, since missing data prior empirical research with multiple interpolation the data processing method filled. In the empirical research, first through the network model to calculate the sensitivity of the risk factor risk factor reference networks and financial institutions to network risks; secondly decomposition (Std, VaR and ES) variance and risk of 36 financial institutions carried out during the stock market crash risk analysis; again with a cumulative gain of CoStd, Co VaR CoES and return to measure the effectiveness of network risks; Finally, dynamic conditions correlation model (DCC) to calculate the adjacency matrix between nodes, the above model will be extended to the conditional empirical model the study. Chapter 5 is the conclusion and prospect, based on empirical results is the conclusion to give policy recommendations, and noted some shortcomings thesis and prospects for further researchChapter 4 of the empirical analysis, we selected data is relatively complete 36 financial institutions in weekly data January 1,2011 to December 31,2015 from a research sample, since missing data prior empirical research with multiple interpolation the data processing method filled. In the empirical research, first through the network model to calculate the sensitivity of the risk factor risk factor reference networks and financial institutions to network risks; secondly decomposition (Std, VaR and ES) variance and risk of 36 financial institutions carried out during the stock market crash risk analysis; again with a cumulative gain of CoStd, Co VaR CoES and return to measure the effectiveness of network risks; Finally, dynamic conditions correlation model (DCC) to calculate the adjacency matrix between nodes, the above model will be extended to the conditional empirical model the study. Chapter 5 is the conclusion and prospect, based on empirical results is the conclusion to give policy recommendations, and noted some shortcomings thesis and prospects for further researchCompared with previous studies, this article might have the following novelties (1) to study the relationship between interrelated residual income, this paper centrality index method to construct a network of risk factors of Fama-French three-factor model residual income not yet pricing analysis. (2) with CoStd, Std, and CoES Co VaR percentage of total VaR and ES to measure the contribution of the network risk, and with CoStd, Co VaR and CoES regression explain ScS network risk for yield strength through cumulative revenue. (3) In order to make the model with time-varying characteristics, adding dynamic conditions associated model (DCC) in the model to estimate the adjacency matrix, and using maximum likelihood estimation to estimate the sensitivity of the risk factor of the network.However, due to my lack of academic ability and other reasons, the paper still has the following disadvantages:(1) Although the data has been selected considering the availability and integrity of data, but due to the late development of China’s financial market, most the stock market too late, so the sample does not contain all the financial stocks, nor to select the appropriate time interval. (2) completion of the data, although the method has been used to replace multiple missing data interpolation, but the lack of reasons for missing data and other analysis, data interpolation with a certain degree of subjectivity. |
PDF Full Text Request