基于多目标规划的保险公司随机资产负债管理

发布时间：2018-06-14 16:42

本文选题：多目标规划 + 随机情景生成　；参考：《南开大学》2014年博士论文

【摘要】：资产负债管理(ALM)是金融机构经营管理的核心内容,其早期作为重要的风险管理手段,能够管理利率风险、流动性风险或汇率风险,近年来资产负债管理的内容逐渐扩展至资本管理、利润规划与发展等方面,成为金融机构重要的管理手段。然而,由于技术的局限性,传统的资产负债管理大多是基于单目标的最优决策模型,事实上,保险公司经营具有多目标属性,其要实现的目标有多个,单目标资产负债管理技术无法满足保险公司综合性管理需求,无法得到多种目标下的最优决策结果,多目标最优决策模型则可以解决这一问题。此外,确定性管理模型不具备良好的预测能力,而随机管理模型能够根据宏观经济形势、资本市场随机变化、公司自身资产负债价值波动等特征做出有效决策。如果能够将多项目标、资产与负债随机变化等特征同时纳入到决策过程中,便可以使保险公司得到更加全面、科学的资产负债管理决策,显著提高其综合管理能力。论文的主要内容如下： 1、金融机构资产负债管理与多目标规划理论研究与文献综述,梳理了资产负债管理理论与多目标规划理论,分别从传统ALM技术、基于随机规划的ALM和基于随机控制的ALM三个方面进行了分析和总结,并对多目标规划理论研究、应用研究和求解研究三方面进行了文献整理和综述。 2、提供了一个完整通用的保险公司多目标资产负债管理决策系统。首先阐述了保险公司经营的多目标属性,给出资产配置、资本管理、险种结构三种决策的多目标决策过程。其次构建了保险公司资产负债管理框架,指出保险公司资产负债管理与其他金融机构的最大不同之处在于资产和负债的特性,并对其进行分析。最后建立了保险公司多目标资产负债管理决策系统,并重点讨论了公司的目标函数选择和约束条件。 3、随机资产负债情景生成,提出寿险负债情景生成方法。整理现有的一般化情景生成方法并比较各种方法的优劣。利用动态Nelson-Siegel利率期限结构模型和向量自回归模型生成资产收益率情景,给出了保单数量、死亡率、退保率、分红政策的随机变化方程,并在此基础上得到寿险保费、赔款、精算准备金、红利情景生成模型。根据我国宏观经济数据以及保险业经验数据进行资产与负债随机情景生成的实证研究。 4、寿险公司多目标资产负债管理模型。根据目标的重要程度选择利润目标、价值目标、风险目标、偿付能力约束与资产负债价值变化约束作为寿险模型的目标和约束。在多目标规划求解方面,采用遗传算法求解,分别给出了传统不分红保险、理财型分红险、保障型分红险和公司整体层面的资产负债决策结果。在各险种的敏感性分析方面,重点讨论死亡率、退保率、预定利率、费用率等寿险产品定价关键假设,在公司层面敏感性分析上,则主要讨论发展战略、资本和规模增长率的对决策结果的影响。根据我国保险市场上寿险公司的资产规模、盈利能力、资本结构、成长性等指标将保险公司分为不同的类别,根据保险公司的特点以及不同时期的经营目标的偏好,得到最优决策结果并进行分析。 5、财险公司随机资产负债管理。财产保险行业不同于寿险业的主要特点在于负债的短期性和保险资产的高流动性,因此目标函数与约束条件选择了利润目标、风险目标、流动性约束、偿付能力约束以及资产价值变化约束。相比寿险险种,非寿险产品的险种差异较小,因此主要给出公司层面的资产负债管理决策结果。在敏感性分析中主要讨论了赔付率、资本和战略目标对决策结果的影响。论文的主要创新点如下： 1、考虑资产与负债双重随机因素,提出了负债情景生成方法。一般情况下资产情景生成的方法较为成熟,而保险产品包含很多保单选择权及监管制约,导致负债变化具有极大的复杂性。本文在给出保单数量、死亡率、退保率、分红政策的随机变化方程基础上根据保险精算原理得到保费、赔款、精算准备金、红利等负债情景,建立双随机资产负债管理模型。 2、多目标资产负债管理模型克服了单目标资产负债管理的局限性。以往的资产负债管理往往仅追求盈利或风险管理的单一层次管理,并且设定的目标只关注本层级能够实现的目标,范围较窄,本论文将这些目标进行扩展及提升,从更高层次实现保险公司资产负债管理的功能。 3、利用智能算法求解随机多目标资产负债管理模型。从多目标理论而言,多目标模型的求解问题是最关键的问题之一,而理论上多目标规划的结果应该是非劣解集,但由于求解困难较大,所以一般将多目标转换为单目标,但在转换过程中需要各种条件的加入,因此具有很大的主观性和局限性,使结果不具有一般性。非劣解集则解决了该问题,可以使结果应用于更加宽的范围。考虑到决策者在了解偏好的情况下需要得到唯一解,所以本论文的研究中对于非劣解集和唯一最优解都进行了研究。 4、随机资产负债管理模型及其求解结果的应用,给出了具有指导意义的投资原则和决策依据。不同的保险公司发展战略不同,而同一家保险公司在不同的发展阶段战略目标也会不同。本文首次对不同发展战略和不同发展时期的公司ALM决策进行实证研究,得到许多具体而有效的资产负债管理建议,这些建议对于指导管理者得到正确全面的资产负债管理决策极为重要。
[Abstract]:Asset Liability Management (ALM) is the core content of the management of financial institutions. Early as an important means of risk management, it can manage interest rate risk, liquidity risk or exchange rate risk. In recent years, the content of asset liability management has gradually expanded to capital management, profit planning and development and so on, which has become an important management means for financial institutions. However, due to the limitation of technology, the traditional asset liability management is mostly based on the optimal decision model based on single objective. In fact, the insurance company has multiple objectives, and it has many goals to achieve. The single objective asset liability management technology can not meet the comprehensive management needs of the insurance company and can not get the most objective. The multi objective optimal decision model can solve this problem. In addition, the deterministic management model does not have good forecasting ability, and the stochastic management model can make effective decisions based on the macroeconomic situation, the random change of capital market, the value fluctuation of the company's own assets and liabilities, and so on. The random changes of assets and liabilities are incorporated into the decision-making process, and the insurance companies can get more comprehensive and scientific assets and liabilities management decisions and significantly improve their comprehensive management ability.
The main contents of the paper are as follows:
1, the theoretical research and literature review of assets and liabilities management and multi-objective programming in financial institutions, combing the theory of asset liability management and multi-objective programming, analyzing and summarizing the three aspects of the traditional ALM technology, ALM based on random programming and ALM based on random control, and the research on the theory of multi-objective programming, application and research. Three aspects of solution research are reviewed and summarized.
2, it provides a comprehensive and universal multi-objective asset liability management decision system for insurance companies. First, the multi-objective attributes of the insurance company are expounded, and the multi-objective decision-making process of the assets allocation, capital management and the insurance structure of three kinds of decision-making is given. Secondly, the management framework of the insurance company's assets liabilities is constructed, and the assets and liabilities of the insurance companies are pointed out. The biggest difference between management and other financial institutions lies in the characteristics of assets and liabilities, and analyzes them. Finally, the multi target asset liability management decision system of insurance companies is established, and the selection and constraints of the objective function of the company are emphatically discussed.
3, the situation of the random asset liability scenario is generated, and the generation method of the life insurance liability scenario is put forward. The existing general scenario generation method is arranged and the advantages and disadvantages of various methods are compared. The dynamic Nelson-Siegel interest rate term structure model and the vector autoregressive model are used to generate the asset return scenario, which gives the policy number, death rate, reinsurance rate, and dividend policy. On the basis of this, we get the life insurance premiums, indemnities, actuarial reserves, and the dividend scenario generation model. The empirical research on the formation of the random situation of assets and liabilities based on the macroeconomic data and the insurance industry experience data in China.
4, the multi-objective asset liability management model of life insurance company. According to the importance of the target, we choose the profit target, the value target, the risk target, the solvency constraint and the asset and liability value change constraint as the goal and constraint of the life insurance model. In the multi objective programming solution, the traditional non dividend insurance is given by using the genetic algorithm. In the sensitivity analysis of various types of insurance, we focus on the key assumptions of the pricing of life insurance products such as death rate, reinsurance rate, predetermined interest rate, cost rate and so on. In the sensitivity analysis of the company level, the development strategy, capital and scale growth are mainly discussed. According to the assets scale, profitability, capital structure, growth and other indicators of life insurance companies in China's insurance market, the insurance companies are divided into different categories. According to the characteristics of the insurance companies and the preference of the operating targets in different periods, the optimal decision results are obtained and analyzed.
5, the property insurance company's random asset liability management. The main characteristics of the property insurance industry different from the life insurance industry are the short-term liabilities and the high liquidity of the insurance assets. Therefore, the target function and the constraint conditions choose the profit target, the risk target, the liquidity constraint, the solvency contract and the change of the asset value. There is a small difference in the risk of non life insurance products. Therefore, the results of asset liability management decision are mainly given at the company level. In the sensitivity analysis, the impact of the loss rate, capital and strategic objectives on the decision results are mainly discussed.
The main innovations of the paper are as follows:
1, considering the dual random factors of assets and liabilities, a method of generating debt situation is proposed. Under the general situation, the method of asset scenario generation is more mature, and the insurance product contains a lot of policy options and regulatory constraints, which leads to the great complexity of the debt change. In this paper, the number of policies, the mortality rate, the rate of reinsurance and the dividend policy are given in this paper. On the basis of the machine change equation, according to the actuarial principle, we get the debt situation of premium, reparations, actuarial reserves, dividends and so on, and establish a double stochastic asset liability management model.
2, the multi objective asset liability management model overcomes the limitation of the single objective asset liability management. The past assets and liabilities management often only pursue the single level management of profit or risk management, and the goal is only concerned with the goal that can be realized at the level of this level. The functions of assets and liabilities management of insurance companies are realized at different levels.
3, using the intelligent algorithm to solve the stochastic multi-objective asset liability management model. From the multi-objective theory, the solving problem of the multi-objective model is one of the most critical problems, and the result of the multi-objective programming in theory should be a non inferior solution set, but because of the difficulty of solving the problem, the multiobjective is converted into a single target, but in the process of conversion, the multi-objective model can be converted into a single target. It requires a variety of conditions to join, so it has a lot of subjectivity and limitations, so that the result is not general. The set of non inferiority solutions can solve the problem and can apply the result to a wider range. Considering that the decision-maker needs a unique solution in the case of understanding the preference, the study in this paper is not a bad solution set and the only one. The optimal solution is studied.
4, the application of the stochastic asset liability management model and its solution results give the guiding significance of the investment principles and decision-making basis. Different insurance companies have different development strategies, while the same insurance company will have different strategic goals at different stages of development. This paper is the first of the different development strategy and the different development period of the company ALM. A number of specific and effective asset liability management recommendations are obtained by empirical research, which are very important to guide managers to get a correct and comprehensive asset liability management decision.
【学位授予单位】：南开大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：F842.3;F840.4

【相似文献】