我国财险公司偿付能力预警机制研究

发布时间：2018-06-08 19:04

本文选题：财产保险 + 偿付能力　；参考：《浙江工商大学》2013年硕士论文

【摘要】：自上世纪80年代中国保险业务重新恢复发展以来,我国的保险业一直处于快速发展阶段。截至2010年保险行业总资产达到4.9万亿,我国已经成为全球最重要的新兴保险大国。作为国民经济的重要组成部分,保险业的稳定发展对我国经济的良好发展起着重要的作用。保险公司偿付能力充足性,不仅影响保险公司的持续经营,还会影响中国保险业和金融市场的稳定发展。目前我国建立了符合中国国情的以偿付能力、市场行为、公司治理为三大支柱的保险监管体系,而偿付能力监管则居于监管体系的核心地位。但是,截至2010年全国仍至少有5家财险公司偿付能力不足。如何发挥偿付能力监管在风险防范中的核心作用,是我国保险业监管的一项重要工作。建立一套灵敏的偿付能力预警体系对完善我国的偿付能力监管体系具有重大意义。本文对国内外保险偿付能力影响因素和偿付能力预测的研究文献进行了回顾和总结,并比较了几种典型的偿付能力影响因素和偿付能力预测的计量模型。实证过程选取了主成分分析法对影响偿付能力的影响因素进行分析,采用BP神经网络对偿付能力进行预测。BP神经网络模仿、简化和抽象生物大脑神经系统,能够自身适应环境、总结规律、完成某种运算,有着传统统计方法无法比拟的适应性、容错性及自组织性等优点。但是在BP神经网络预测中,使用的指标并不是越多越好,过多的指标会造成BP神经网络在学习过程中受到过多的噪声干扰,并且会由于隐含层过多,造成训练过度,从而影响预测的精度。在实证过程中,为了全面反映财险公司的财务状况,选取的财务指标相对较多,并且指标之间存在一定的相关性,反映的信息在一定程度上有重叠。因此在用BP神经网络进行预测之前,先利用主成分分析把多指标转化为少数几个综合指标。本文选取了33家在2007—2010年具有完整财务报表的财产保险公司,根据其财务报表计算了衡量保险公司偿付能力的13个财务指标,利用主成分分析法得到6个主成分。本文把样本分为训练组和检验组。把主成分分析法得到的6个主成分作为输入变量,以偿付能力充足率作为输出变量,对BP神经网络进行了训练。之后以偿付能力充足率100%作为判定偿付能力是否充足的标准,利用BP神经网络对样本公司未来一年和未来两年的偿付能力进行预测研究。通过实证研究证明BP神经网络对偿付能力不足的保险公司预测正确率达到90%以上预测效果显著。最后,根据实证结果,给出了完善BP神经网络模型在我国财产保险偿付能力预测运用的若干建议。
[Abstract]:Since the reinstatement and development of China's insurance business in 1980's, China's insurance industry has been in the stage of rapid development. By 2010, the total assets of insurance industry has reached 4.9 trillion, China has become the most important emerging insurance country in the world. As an important part of the national economy, the steady development of the insurance industry plays an important role in the good economic development of our country. The adequacy of the solvency of an insurance company not only affects the continuing operation of the insurance company, but also affects the steady development of the insurance industry and the financial market in China. Corporate governance is a three-pillar insurance regulatory system, while solvency regulation is at the core of the regulatory system. However, as of 2010, there are still at least five financial insurance companies underpaid. How to play the central role of solvency regulation in risk prevention is an important work of insurance supervision in China. The establishment of a sensitive solvency warning system is of great significance to the improvement of China's solvency supervision system. This paper reviews and summarizes the research literature on the influencing factors and solvency prediction of insurance solvency both at home and abroad. Several typical influencing factors of solvency and the econometric models of solvency prediction are compared. The empirical process selects the principal component analysis method to analyze the influencing factors of solvency. BP neural network is used to predict the solvency. BP neural network is used to simulate and simplify and abstract the biological brain neural system, which can adapt itself to the environment. Summing up the rules and completing some operations have the advantages of adaptability, fault-tolerance and self-organization, which can not be compared with the traditional statistical methods. However, in the prediction of BP neural network, the more indexes are used, the better. Too many indexes will cause too much noise interference in the learning process of BP neural network, and it will cause excessive training because of too many hidden layers. Thus, the accuracy of prediction is affected. In the empirical process, in order to fully reflect the financial situation of property insurance companies, the financial indicators selected are relatively large, and there is a certain correlation between the indicators, and the information reflected is overlapped to a certain extent. Therefore, before using BP neural network to predict, the principal component analysis is used to transform multiple indexes into a few comprehensive indexes. In this paper, 33 property insurance companies with complete financial statements in 2007-2010 are selected. According to its financial statements, 13 financial indexes are calculated to measure the solvency of insurance companies, and six principal components are obtained by principal component analysis. The sample is divided into training group and test group. Taking the six principal components obtained by principal component analysis as input variables and solvency adequacy as output variables, BP neural network is trained. Then the solvency adequacy ratio of 100% is taken as the criterion to judge whether the solvency is sufficient or not. The BP neural network is used to predict the solvency of the sample company in the next year and the next two years. Through empirical research, it is proved that BP neural network has a remarkable effect on the forecasting accuracy of insurance companies with insufficient solvency to more than 90%. Finally, based on the empirical results, some suggestions on the application of BP neural network model in the prediction of property insurance solvency in China are given.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F842.3;F224

【参考文献】