基于面板数据Logit模型的系统性金融危机预警研究

发布时间：2018-04-19 21:25

本文选题：系统性金融危机预警 + 面板数据Logit模型　；参考：《西南财经大学》2012年硕士论文

【摘要】：从1825年的英国经济危机到2010年的欧洲债务危机,在全球经济发展过程中,金融危机总是如影相随。随着全球一体化进程的推进,金融机构间的关系日益密切,金融体系中潜在的系统性风险不断积聚,金融危机相比以前更具有系统性危机的特征。据国际货币基金组织(IMF)报告整理,1980-2002年间就有93个国家发生了114次系统性金融危机(Caprio、Klingebiel和Leaven,2003)。2007年爆发的次贷危机最终演变为全球金融危机,2010年爆发的欧洲债务危机致使许多国家陷入债务泥潭,这些现象再一次引起人们对系统性金融危机的担忧。在金融业高速发展的今天,系统性危机已成为全球金融体系发展中的新威胁。如何防范系统性风险、避免系统性危机的爆发已经引起人们的广泛关注。同时,在国际金融不稳定因素逐渐增加的大环境下,我们又不得不思考：金融业日益壮大的中国如何防范于未然? 金融危机预警研究一直是国际社会关注的焦点,已有研究主要集中在基于危机发生根源界定的货币危机或银行危机。对系统性金融危机的研究多是基于定性角度,如界定、成因、监管等,定量角度的研究甚少。本文在借鉴已有金融危机预警研究成果的基础上,尝试构建系统性金融危机预警模型。模型构建过程中,主要遇到两个问题,一是如何将不同时点、不同国家爆发的系统性危机事件综合在一起运用?二是危机作为被解释变量如何表示?为此,引入面板数据Logit模型。面板数据既考虑了时间因素的影响,又考虑截面差异的影响,将其引入可以解决第一个问题；Logit模型是目前较为主流的危机预警方法,该模型在构建过程中将危机定义为0-1变量,并能够将危机发生的可能性综合为一个有效的概率,使得危机发生的可能性成为一个明确的概念,将其引入可以解决第二个问题。模型构建的前提是预警指标的选取,通过典型的危机事件分析法和已有研究资料借鉴法相结合的方式获取。与已有预警研究中存在同样的问题,某些预警指标间具有较高的相关关系,进而影响模型的准确度。为此,进一步引入全局主成份分析法,对原始预警指标提取公共因子,并将这些不相关的公共因子作为新的变量用于构建Logit模型。运用公共因子构建模型仅是出于消除指标间高度相关关系的影响,这一新的模型是否好于原始模型尚需要进一步的验证,检验分为样本内检验和样本外检验。最后,将效果较好的模型应用于我国,监测我国近年来的系统性风险状况。各章节主要内容如下：第一章为绪论。遵循研究背景、研究价值、研究思路、研究内容、研究方法这一路线,依次展开说明。第二章为述评与界定,分为三个部分。首先阐述金融安全、金融危机与系统性金融危机的定义以及三者之间的联系。其次,综述金融危机理论中的代表性成果——四代危机理论。四代危机理论虽是基于事后研究,但较好的解释了重大危机事件产生的原因,而这些危机事件大多数属于系统性金融危机。最后,对已有的预警研究成果进行综述,得到在预测危机发生概率方面,面板数据Logit模型较为合适。第三章为典型剖析：共性与表征。通过分析典型的系统性金融危机事件,发现这些危机事件存在的共同特点：金融风险积聚的过程中伴随着经济体的某些指标的异常变动。这些异常指标为模型构建过程中预警指标的选择提供事实依据。第四章为寻证与建模。首先对样本来源、指标选取及预警方法的选择进行说明。其次,鉴于某些预警指标间存在较高的相关关系,运用全局主成份分析法提取公共因子,将这些不相关的公共因子作为新的变量。为比较两种模型的效果,本文分别构建了基于原始预警指标的Logit模型和基于公共因子的Logit模型。第五章为检验与应用。模型效果的好坏还需要进一步检验,模型检验分为样本内检验和样本外检验。样本内检验的途径是将样本回代到模型中预测危机发生概率,从而判断模型的准确度；样本外检验的途径是将次贷危机作为样本对模型进行衡量。检验结果得到：基于公共因子的Logit模型效果要好于基于原始预警指标的Logit模型。最后,将检验效果较好的模型应用于我国,得到我国近年来整体风险不大,短期内发生系统性危机的可能性较小,但易受外部环境的影响。第六章为结论与展望,在总结全文的基础上,给出政策含义并指出研究存在的不足和以后需要进一步研究的地方。与以往的相关研究相比,本文较有新意的地方体现在以下几个方面：第一,以系统性金融危机为研究的切入点,构建系统性金融危机预警模型；第二,运用面板数据建立危机预警模型,既考虑了时间因素,又考虑了截面差异；第三,引入全局主成分分析法,对原始预警指标提取公共因子,运用不相关的公共因子构建Logit模型,且基于公共因子的Logit模型效果好于基于原始预警指标的Logit模型。由于学术水平、文献资料、数据来源的有限,本文尚有较多不足之处。尤为值得关注的问题有：第一,运用典型的危机事件分析法和已有研究资料借鉴法相结合的方式获取危机预警的核心指标,并未将全部影响因素涵盖进去,是否存在更为合适的系统性危机预警指标体系需要进一步研究。第二,模型估计的准确度并没有达到理想的状态,需要进一步优化,以提高其预测精度。第三,通过对中国的实践可知,后金融危机的大背景对模型产生了干扰。如何排除这种干扰因素,运用经济体本身存在的不稳定因素判断其系统性风险将是今后研究中值得进一步探讨的地方。
[Abstract]:From the British economic crisis of 1825 to the European debt crisis in 2010, the financial crisis has always followed in the process of global economic development. With the advancement of the global integration process, the relationship between financial institutions is increasingly close, the potential systemic risks in the financial system are constantly accumulating, and the financial crisis has a more systematic crisis than before. According to the International Monetary Fund (IMF) report, 93 countries had 114 systemic financial crises in 1980-2002 years (Caprio, Klingebiel and Leaven, 2003), and the subprime crisis that broke out in.2007 was finally transformed into the global financial crisis, and the European continent debt crisis in 2010 caused many countries to fall into the mire of debt. As once again, people are worried about the systemic financial crisis. Today, with the rapid development of the financial industry, the systemic crisis has become a new threat to the development of the global financial system. How to prevent systemic risk and avoid the outbreak of systemic crisis has aroused widespread concern. In the big environment, we have to think: how can China's financial industry grow stronger?
The research of financial crisis early-warning has always been the focus of attention of the international community. The existing research is mainly focused on the monetary crisis or bank crisis based on the definition of the root of the crisis. The research on the systemic financial crisis is based on the qualitative angle, such as the definition, the cause, the supervision and so on, and the quantitative angle is seldom studied. This paper is a reference to the existing financial crisis. On the basis of the results of the police research, we try to build a systematic financial crisis early warning model. In the process of building the model, there are two main problems. One is how to integrate the systemic crisis events in different countries together? Two is how the crisis is expressed as an explanatory variable? To this end, the panel data Logit model is introduced. The board data not only considers the influence of the time factor, but also considers the influence of the cross section, and introduces it into the first problem. The Logit model is the mainstream crisis early warning method at present. The model defines the crisis as the 0-1 variable in the construction process, and can integrate the possibility of the crisis to be a valid probability, making the possibility of the crisis to be a valid probability. The possibility of the crisis is a clear concept, which can be introduced into second problems. The premise of the model construction is the selection of early warning index, which is obtained through the combination of the typical crisis event analysis method and the existing research data reference method. There are the same problems with the existing early warning research, and some early warning indicators have the same problem. There is a higher correlation, which affects the accuracy of the model. Therefore, the global principal component analysis (PCA) is introduced to extract the public factors from the original early warning index, and the unrelated public factors are used as new variables to build the Logit model. The construction model of the public factor is only to eliminate the high correlation between the indexes. Whether the new model is better than the original model needs further verification, the test is divided into the sample test and the outside sample test. Finally, the better model is applied to our country to monitor the systemic risk status of our country in recent years. The main contents of each chapter are as follows:
The first chapter is the introduction, which follows the research background, research value, research ideas, research contents and research methods.
The second chapter is a review and definition, which is divided into three parts. First, it expounds the financial security, the definition of the financial crisis and the systemic financial crisis and the connection between the three. Secondly, it summarizes the representative results of the financial crisis theory, the four generation of crisis theory. The four generation of crisis theory is based on the ex post study, but it is a good explanation for the major crisis. The cause of the events, and most of these crisis events belong to the systemic financial crisis. Finally, the existing early warning research results are reviewed, and the panel data Logit model is more appropriate in predicting the probability of the crisis.
The third chapter is a typical analysis: generality and characterization. Through the analysis of typical systemic financial crisis events, the common characteristics of these crisis events are found: the process of financial risk accumulation is accompanied by abnormal changes in some indicators of the economy. These abnormal indicators provide facts for the selection of early warning indicators in the process of model construction. According to.
The fourth chapter is to find the evidence and modeling. First, it explains the selection of sample sources, index selection and early warning methods. Secondly, in view of the high correlation between some early warning indicators, the use of global principal component analysis is used to extract public factors and make these unrelated public factors as new variables. This paper compares the effect of the two models. The Logit model based on the original warning index and the Logit model based on common factors are constructed respectively.
The fifth chapter is the test and application. The model effect needs further inspection, the model test is divided into the sample test and the outside sample test. The way to test the sample is to replace the sample into the model to predict the probability of the crisis, so as to judge the accuracy of the model; the way of the sample inspection is to take the subprime crisis as the sample pair model. The result of the test is that the results of the Logit model based on public factors are better than the Logit model based on the original early-warning index. Finally, the model with better results is applied to our country, and the overall risk of our country is not large in recent years, and the possibility of a systemic crisis in the short term is small, but it is easily affected by the external environment.
The sixth chapter is the conclusion and outlook. On the basis of summarizing the full text, the paper gives the policy implications and points out the shortcomings of the study and the future need for further research.
Compared with the previous research, the new places are embodied in the following aspects: first, a systematic financial crisis early warning model is constructed with the systematic financial crisis as the breakthrough point. Second, using panel data to establish a crisis early warning model, considering both the time factor and the cross section difference; third, the introduction of the financial crisis early warning model. Global principal component analysis (PCA) is used to extract public factors for original early-warning indicators, and the Logit model is constructed with unrelated public factors, and the Logit model based on public factors is better than the Logit model based on the original early warning index.
Because of the academic level, the literature and the limited sources of data, there are still many shortcomings in this paper. The problems of particular concern are as follows: first, the core indicators of the crisis early-warning are obtained by using the typical crisis event analysis method and the existing research data reference method. For the appropriate systematic crisis early warning index system need further study. Second, the accuracy of the model estimation has not reached the ideal state, needs further optimization to improve its prediction accuracy. Third, through the practice of China, the background of the post financial crisis has caused interference to the model. How to exclude such interference factors, It is worthwhile to further explore the systematic risk in the future by using the unstable factors existing in the economy itself.

【学位授予单位】：西南财经大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F224;F831.59

【引证文献】