【作者】 黄苒；

【导师】 唐齐鸣；

【作者基本信息】 华中科技大学 ， 数量经济学， 2011， 博士

【摘要】 随着世界经济的发展和经济全球化的推进,金融业已经成为各国经济的核心。但是,金融业的发展和国际化就像一把双刃剑,为给各国经济发展注入活力和能量的同时,也带来了风险和冲击。特别是像中国这样的发展中国家和一些新兴国家,它们的金融行业起步和发展较晚,金融体制建设相当不完善,参与制定国际金融规则的程度较低,对许多新兴金融工具还不熟悉,因而尤其容易受到国际金融风险的冲击,遭受的损失也会更大。自上世纪九十年代以来,亚洲金融危机、俄罗斯金融危机及源于美国次贷风波的全球金融危机等都给这些国家带来了巨大的冲击,进一步加深了各国金融市场的不完全特征和不稳定性,资产收益的正常波动和异常跳跃双重性越来越凸显。因此,针对资产收益跳跃风险的研究已逐渐成为金融领域的重点和难点问题之一。随着中国经济对外开放不断深入和金融行业持续发展,越来越多的公司通过上市来融通资金,成为中国经济发展的重要参与者和受益者。但面对国际金融风险频繁显现的外部环境,以及地震、雪灾、洪涝等突发事件时有发生的内部环境,许多上市公司难以避免的遭受到冲击,经营受到影响,财务状况恶化,导致资产出现大幅跳跃的频率增加。因此,在新的背景下如何对中国上市公司的资产收益特征进行解释?如何对其信用状况进行度量?如何对引入跳跃行为后的资产收益变化过程进行建模?如何分析跳跃行为对公司资产总体风险的影响?以及如何完善信用风险评估体系等问题对管理和控制上市公司的风险十分必要。近年来,国外的信用风险研究也越来越强调跳跃因素的重要性。不过,这些研究大多利用股票期权等金融衍生品交易数据来剥离跳跃风险。而中国金融市场以股票市场和债券市场为主,金融衍生品市场起步较晚,数据量十分有限,因此,利用股票期权等衍生品的市场数据来分析违约风险中的跳跃变化并不十分可行。另外,我国债券市场的规模还很小,各种配套设施和系统也很不完善,公司的违约风险收益和信用等级挂钩,由评级机构利用公司可获得的有限财务数据和信息资源评估出信用等级。但这些资源往往不能真实、及时和全面的涵盖各种外来信息。因此,也需要将时刻变化的权益市场数据和信息纳入债券信用评估体系,以改善其信用评估能力。这种结合时刻变化的市场数据的方法可以提高模型的信用风险识别能力,对我国银行和上市公司构建内部信用评级模型,研究公司可违约债券定价以及提高金融风险预警能力都有重要的意义。基于这些研究目的,本文构建了基于跳跃行为的“权益价值变化→资产价值变化→违约度量”的研究体系,分析跳跃因素引发的资产价值和整个资本结构的变化。利用带有跳跃因子的期权定价模型建立权益价值和资产价值之间的非线性关系,将跳跃行为对公司权益价值的影响过渡到其对资产价值的影响,进而度量跳跃行为对违约风险的影响。在研究跳跃行为对股票资产收益过程的影响时,本文构建了门限效应下状态变量依赖自回归跳跃强度GARCH模型(TSD-ARJI-GARCH)。纳入了状态变量对个股跳跃强度预期的显著影响和门限效应,综合考虑了跳跃行为的时变性、集群性、状态依赖性以及正常波动的集群性、不对称性等特征,并从实证的角度探讨了ST股和非ST股间跳跃行为的差异以及交叉持股公司之间股票大幅波动的传染性。通过与现有的模型进行对比分析发现,TSD-ARJI-GARCH模型无论是在拟合优度方面还是在预测效果方面都具有更多的优势。它能更好的描述跳跃行为的时变特征和集聚效应,能较好的反映状态变量信息的变化对跳跃行为的影响,克服了ARJI-GARCH模型对跳跃行为过度识别的问题,能有效区分ST公司和非ST公司跳跃行为的差异。在考量跳跃因素对违约风险的影响时,以中国上市公司为考察样本,在资产价值变动本身不可观测的情况下,利用权益市场数据所蕴含的信息间接度量含有跳跃风险的资产价值变化过程及违约概率。考察了突发事件等宏观、系统因素和个体的异质性对跳跃行为的影响。比较了跳-扩散模型和纯扩散模型在违约风险度量方面的差异。结果表明,2007-2008年全球金融危机、雪灾和地震等突发事件明显引发了所有样本公司的异常跳跃式变动。在此期间,上市公司的违约风险大幅增加,跳跃风险不仅广泛存在,有很强的系统性,也在很大程度上受到上市公司异质性的影响,不同的个体表现出不同的跳跃特征。考虑了跳跃因素后,公司的违约风险可以进一步区分为跳跃式违约风险和连续变动式违约风险。在短时间内或者是资产价值负债率比较小的情况下,跳跃风险占主导,它的存在使得公司的违约概率高于纯扩散模型估计的违约概率。而在长时间内或资产价值负债率比较高的情况下资产价值的扩散波动风险占主导,它使得公司的违约概率低于纯扩散模型所预示的违约概率。更多还原

【Abstract】 Along with the development of world economy and the economic globalization, the financial industry has already been the core of every country’s economy. However, its development and internationalization brings not only the dynamics and energy, but also the risk and impact. Especially, the developing countries including China and some emerging countries, which boast a short history of financial development and relatively imperfect financial system, have fewer opportunities to join in formulating the international financial principles and are unfamiliar with the emerging financial instruments, are more like to be hit by international financial risks and suffer more losses. Since 1990s, various financial crises, including Asian financial crisis and the global financial crisis resulted from US sub-prime mortgage crisis, have caused mass impacts on these markets and further strengthened their incompletion and instability. The assets in these markets enjoy more obvious double features of normal volatility and jump. Therefore, the research upon the jump risk of asset returns has already become one of the important and difficult issues in financial field.Along with China’s economic development and open to the outside, more and more companies select the road of listing on the market to obtain the capital financing. Until now, the listed companies have already become the important participants and beneficiaries of China’s economic development. However, facing the external environment of frequent international financial risks and internal environment of various sudden events such as earthquake, disaster of snowstorm and floodwater, many listed companies unavoidably suffer impacts and losses, which cause the increase of the frequency of jump in their assets returns. Therefore, the issues, including how to describe the asset return features and measure credit risk of China’s listed companies in new context, how to model the asset return change process with jump and how to improve the credit risk evaluating system, are very important for managing and controlling the risks of listed companies.Recently, the researches about credit risk in foreign countries have paid more and more attention to the jump factors. However, most of them try to isolate the jump risk from traded data of financial derivatives such as stock option and credit swap. But, China’s derivative market only has a very short history and owns small amount of traded data, so this scheme is unworkable when analyzing the jump change in the default risk of companies in China. On the other side, China also features a relatively smaller scale of bond market, facilities and system of which are imperfect. The default risk premium of bond is only linked to the credit grade, which is evaluated by credit rating agency referring to the obtainable limited financing data and resources. But these data and resources usually cannot wholly, actually and immediately reflect various external information, so it also needs to introduce the changing stock market data and information into credit rating system of bonds to improve its rating ability. This method not only helps to elevate the credit-risk indentifying ability of model, but also plays important role in constructing internal credit rating model for banks and listed companies, pricing the defaultable bond and improving the early-warning ability.For this research purposes, the whole jump-behavior based research system of "claim asset change process→total asset change asset→default risk" measurement, which analyzes the change of asset value and total capital structure caused by jump factors, is completely illustrated and constructed. It establishes the non-linear relationship between claim value and asset value on the basis of option pricing model with jump and transfers the impact of jump on the claim asset to the total asset so as to introduce jump factors into default risk measurement.When studying the jump behavior of stock return, it constructs the threshold-based state-dependent autoregressive jump intensity (TSD-ARJI) GARCH model, which taking the significant influence and the threshold effect of state-variables on the expected jump intensity into account and comprehensively analyzing time variation, clustering and state-dependence of jump behavior as well as non-symmetry and clustering of normal volatility. Then, it applies the model to contrast the risk components of ST and non-ST stocks and discuss the volatility contamination between cross-shareholding companies. Both goodness of fit and forecast evaluation show TSD-ARJI-GARCH model performs better in contrast to the existing models. It can better describe the time variation and clustering of jump behavior, more accurately reflect the influence of the information change of state variables on the jump behavior, conquer the over-identification of ARJI-GARCH upon the jump point and more effectively discriminate the jump features of ST and non-ST stocksWhen measuring the impact of jump factors on default risk, it takes the listed companies in China as the samples and uses information contained in the claim market to indirectly measure the asset value change with jump and the default probability. It not only studies the influence of macro and systematic factors and sudden events on jump behavior, but also analyzes the unique jump features of individual stock caused by idiosyncrasy. Then, it contrasts jump-diffusion models and pure diffusion models in measuring default risk. The empirical research results show the global financial crisis, disaster of snowstorm and earthquake in 2007 and 2008, obviously induced abnormal jumps of all stocks. In this period, default risk of listed companies increased significantly. The jump risk not only presented extensively and systematically, but also was influenced by idiosyncrasy. Thus, each company featured unique jump behavior and default risk. After taking the jump factors into consideration, the default risk can be divided into jump-style default risk and continuity-style default risk. In short term or when the ratio of asset value to reliability is relatively lower, the jump part dominates the whole risk to make the default probability larger than that estimated by pure diffusion model. On the contrary, in long term or when the ratio of asset value to reliability is relatively higher, the diffusion part dominates the whole risk to make the default probability smaller than that estimated by pure diffusion model.更多还原