【作者】 刘艳清；

【导师】 马树才；

【作者基本信息】 辽宁大学 ， 统计学， 2013， 博士

【摘要】 近年来，随着我国金融服务业全球化以及利率市场化进程的不断推进，银行与证券业职能相互交融，市场规模迅速膨胀，金融风险结构日趋复杂，市场风险测量难度不断加大。金融创新进一步加剧了金融机构风险管理的难度。如何充分利用现有信息，准确地测度金融风险，进而对金融风险进行有效控制和管理，对于我国金融服务业的健康发展具有重大的理论意义和现实指导意义。风险管理（Risk Management）是指对风险进行辩识、测量（包括预测）以及对风险优先处理次序进行排序的过程，通过对资源的调整和应用，实现对不利事件影响的最小化，并进行相应的监督和控制，从而达到规避风险的目的。金融风险则是指来自金融市场的不确定性所导致的风险。按照不确定性来源不同，金融风险粗略地可以划分为：市场风险、信用风险、操作风险和流动性风险等。金融风险有效控制与管理的核心问题是如何对风险进行测量。现有的文献对如何测量金融风险提出了许多不同的风险测量方法和风险预测模型，但遗憾的是，近几年发生的重大金融事件，很少能够通过现有的模型加以预测。究其原因，主要是市场波动原因复杂，许多模型的假设条件得不到满足。另一个不可否认的原因是，任何一种统计方法或计量模型均存在一定的局限性。因此，探索新的、科学合理的统计模型来预测和防范可能遇到的风险，就显得十分必要。本文正是在上述背景下展开研究，探索贝叶斯统计对金融市场风险测度分析的模型方法，并据此对我国金融市场中的三类主要风险：市场风险、信用风险及操作风险进行实证分析，并针对我国金融市场风险成因进行分析，进而提出防范化解金融风险的对策建议。本文基于贝叶斯统计对我国金融市场若干风险进行测度分析是金融风险测度分析方法上的研究与探索。主要内容及主要结论为：第一部分（第2章），系统地分析了中国金融市场及市场风险构成、资本收益率分布特征，阐明市场风险、信用风险及操作风险的收益率分布均具有尖峰厚尾特征，详细介绍了测度金融市场若干风险的度量方法，尤其是对VaR的半参数法和极值理论的论述，为后续研究奠定理论基础。第二部分（第3章），金融市场风险测度分析理论基础，在现有的模型假定基础上提出度量金融市场若干风险的贝叶斯模型假定，包括度量市场风险的GARCH-POT模型、度量信用风险的SW-GARCH-POT模型、度量操作风险的Weibull-POT模型，并从贝叶斯统计角度对模型的适用性进行论证，为后续风险度量提供量化工具。第三部分（包括第4,5,6三章），在基础理论和模型方法的支持下对具体的金融市场风险进行实证分析。包括第4章贝叶斯GARCH-POT模型对股票市场风险、第5章贝叶斯SW-GARCH-POT模型对信用风险、第6章贝叶斯Weibull-POT模型对操作风险的实证分析。主要结论是，目前我国股票市场风险上证综指VaR0.99值为0.0354261，商业银行信用风险VaR0.99值为41.11，操作风险VaR0.99值为3934.43；同一置信水平下贝叶斯统计方法测算的VaR较经典统计方法测算的VaR值偏大；贝叶斯模型方法解决了现有模型方法信用调级的繁琐以及金融市场风险度量损失数据缺失问题；贝叶斯风险测度分析方法具有一定的科学性和借鉴性。第四部分（包括第7、8两章），对金融市场若干风险的主要成因进行分析，得出市场机制不健全和政策性因素是影响各类风险的主要原因；并提出防范化解金融市场风险的对策建议。论文研究主要采用理论分析和实证分析相结合、定性分析和定量分析相结合的研究方法，尤其注重贝叶斯统计方法的运用。理论分析综合运用了金融学、金融统计学、计量经济学、贝叶斯统计学的相关理论。实证分析以贝叶斯方法为主。贝叶斯方法的应用，一弥补了现有传统度量方法低估风险的不足，二解决了金融风险分析过程中数据缺失问题。模型中参数估计采用MCMC方法中的Gibbs抽样；数值分析和实证分析的计算采用WinBUGS软件、SAS软件以及Eviews5.1、6.0软件。论文主要创新点体现在：1.股票市场风险度量中的贝叶斯GARCH-POT模型应用属于首创。股市风险测度的贝叶斯分析方法弥补了经典统计方法低估市场风险的不足。2.贝叶斯SW-GARCH-POT模型对商业银行信用风险测度分析是信用风险测度方法上的探索。将信用等级的跃迁以转移变量的形式引入SW-GARCH模型实现对信用风险的测算是一种新的尝试。3.对商业银行操作风险的测度分析中，Weibull分布选择有别于现有其他研究成果。实证结果表明，风险测度的贝叶斯分析方法是可行的，具有可操作性。4.系统应用贝叶斯统计方法度量金融风险在国内还不多见，本文研究为风险度量提供了方法上的选择。在一定意义上弥补国内该领域的欠缺。论文不足之处：1.由于时间所限，金融市场流动性风险未作出分析，有待于后续进一步研究。2.实证分析只测算了风险的VaR值，未对防范风险应配置相应的经济资本作出估算，有待于进一步完善。3.由于本人掌握的知识有限，先验分布的选取借鉴于前人已有的研究成果，这本身就是贝叶斯统计的难点所在，也是本人的研究努力方向。更多还原

【Abstract】 In the past few years, the globalization of financial service industry and thepromotion of the interest rate liberation have been blending the functions of banks andsecurities, enlarging the market size, complicating the financial risk structure andarousing difficulties of measuring market risks. The difficulties are intensified offinancial agencies managing risks by financial creation. Therefore, significant in boththeory and reality to the development of financial service industry is how to make fulluse of available information to measure the financial risks, and then to make an effectivecontrol and management of financial risks.Risk management is a process, in which the risks are identified, measured andpredicted, and arranged in the order of priority, and is aimed to minimize the unfavorableeffect in the way of adapting and applying the resources, make proper supervision andmanagement, and avoid the risks. Financial risks arises from the uncertainty of financialmarket, and due to the different sources, they are generally classified into the followinggroups: market risks, credit risks, operational risks, liquidity risks and so on.The critical step in the effective control and management of financial risks is how tomeasure the risks. Although the available documents offer to measure risks in manydifferent approaches and models, recent important financial crises have rarely beenforecast in the models available. On the one hand, the intricate causes behind the marketups and downs are not easy to be found, and therefore, the assumptions of the models cannot be met; on the other hand, the undeniable is that any statistic approach or measuringmodel exists some limitations. Hence, it is necessary to explore new and scientificstatistic models to predict and avoid the potential risks. In the above background, thethesis is to explore the model of Bayesian statistics measuring the risks in financialmarket, on the basis of which three kinds of risks are to be analyzed, namely, marketrisks, credit risks, and operational risks, to analyze the causes of the risks in China’sfinancial market, and to advance the suggestions and strategies to avoid and solve therisks.The thesis, composed of four parts, measuring and analyzing the risks in China’sfinancial market, is to study and explore the financial risk-assessing-and-analyzing approaches.In Chapter Two, as the first part of the thesis, a systematic analysis is made aboutthe construction of China’s financial market, the characteristics of diverse risks infinancial market, and the features of yield rate distribution; an interpretation is made thatany yield rate distribution of the three, such as the market risks, the credit risks, and theoperational risks, is characterized by a steep kurtosis and a buffering skewness, butbuffers in a different direction. Besides, the theories are introduced about how to assessthe risks in financial market, and the semi-parameter based on EGACH-VAR andExtreme Value theory are specifically illustrated, which bases the succeeding study.In Chapter Three, as the second part, the Bayesian Hypothesis is advanced to assessrisks in financial market on the base of existing model hypothesis, such as BayesianGARCH-POT model to assess market risks, Bayesian SW-GARCH-POT model to assessoperational risks; the applicability of the model is illustrated from the Bayesian Statisticpoint of view, to offer the quantizing tool for the succeeding risks.In Chapter Four, Chapter Five and Chapter Six, as the third part, the empiricalanalyses of the specific financial market risks are made based on the fundamentaltheories and models, including the analyses of stock market risks by BayesianGARCH-POT model in Chapter4, the analyses of credit risks by BayesianSW-GARCH-POT model in Chapter5, and the analyses of operational risks by BayesianWeibull-pot model. A conclusion come from these analyses that the VaR of the risks ofChina’s stock market is0.0041, of China’s commercial banks is0.056, and of operationalrisks is0,032. The problem is solved of the complexity of the existing model adjustingthe grade of credit and of the loss of data, and meanwhile, the problem of data invalidityis avoided aroused by the traditional measuring approach.In Chapter Seven and Chapter Eight, as the part four, the main causes of the risks ofChina’s financial market are analyzed on the base of empirical results, and the conclusionis drawn that the imperfect market mechanism and policy factors are the main elementsto influence the specific market risks.Methodology:The thesis is to adopt the approach of the combination of theory analysis andempirical test, and the combination of qualitative analysis and quantitative analysis, andespecially focus on the application of Bayesian Statistics. In the theory analysis, the theories of finance, finance-statistics, econometrics and Bayesian Statistics are applied ina comprehensive way. Empirical analysis mainly adopt the Bayesian Statistics, whichovercomes the disadvantage of traditional measuring approach underestimating the risks,and solves the problem of loss of data in measuring the financial risks. The application ofGibbs sampling works out the problem of Bayesian Statistics financial risks in thecomputing process.The analyses made in the thesis are favored by the WinBUGS software,SASsoftware and Eview5.1,6.0software.Innovations:First, Bayesian GARCH-POT model is applied to assess the risks of stock marketfor the first time, perfecting the classic statistics in underestimating the market risks.Second, the empirical analysis made by Bayesian SW-GARCH-POT model ofcommercial banks’ credit risks explores the credit risks. The new attempt is to analyze, inthe form of variables, the credit risks.Third, in the analyses of commercial banks’ operational risks, the results with thechoice of Weibull distribution differs from those with others.The study shows that it is feasible and operational to measure risks by means ofBayesian Statistics of Weibull.Fourth, rarely seen is to apply Bayesian Statistics to the analyses of financial marketrisks, and therefore, the thesis is to offer the choices of approaches to assess risks, which,in a way, better the research conditions of this field.The thesis needs to be improved in the following:First, the shortage of time leads to the fact that the liquidity risks in the financialmarket are not analyzed, which will be further studied in the future research.Second, the empirical study only measures VaR, and the evaluation is not made ofcapital that is used to avoid risks, and this will be perfected in the further study.Third, the difficulty, consisting of the accumulation of the knowledge and thereference to the previous academic study in the choice of empirical distribution, is thefurther study in the future.更多还原