【作者】 朱少杰；

【导师】 张庆洪；

【作者基本信息】 同济大学 ， 技术经济及管理， 2007， 博士

【摘要】 自20世纪初期以来,伴随着汽车的发明及推广应用,机动车辆保险也经历了一个从无到有、快速发展的过程,已成为经济社会中最为重要的险种之一。车险业务经营效益的好环将直接关系到财险公司的成败。保费是保险企业的生产过程或生产增值链中必备的现金流入,相应地,保费制度成为保证保险企业的偿付能力的关键要素。奖惩系统(Bonus-Malus System,BMS)作为经验费率制度的应用分支,是车险费率制度中不可或缺的组成部分,又是反映投保人和保险人之间的保险交易关系的重要载体。精算师在实践中已总结出构造最优BMS的一整套方法,但忽略了投保人在BMS条件下的行为偏好及其对BMS运作的潜在影响。出于上述考虑,本文突破已有的BMS研究的精算框架,从投保人和保险人之间行为博弈的视角全新梳理车险奖惩制度,在寻求精算学和经济学的嫁接思路的同时,试图为我国机动车辆保险的实践提供理论启示。围绕研究意图,本文首先在文献综述的基础上,确立起行为博弈的分析视角;解析费率分级机制的两个层面:先验分级和BMS,重点归纳国际上BMS的运作模式,并附以对我国车险费率分级效率的评价;以实践认识为铺垫,依托经济学理论解释和论证BMS介入车险业务的合理性。至此,着手讨论需要解决的相关问题。出于优化交易利益的考虑,探讨先验分级和奖惩系统的整合运作问题;着重分析BMS框架下投保人可能表现出的信息不对称问题(动态道德风险和奖金饥渴症),并给出相应的治理思路。按照上述的行文思路,各章研究得出的主要结论如下:第1章综述了国内外BMS研究的文献,指出纯精算角度研究BMS存在的缺陷,即忽略了投保人的行为偏好可能带给BMS精算的变化风险。为此,提出运用行为博弈的分析方法来协调投保人和保险人之间的交易关系,在此基础上给出本文需要解决的问题。第2章归纳了车险费率分级机制的两个层面的实践经验,实证考察我国的车险费率分级操作,分析表明先验分级效率不高,BMS在奖惩力度、费率差异化等方面存有不足,但具备快速收敛于稳态概率分布的能力。第3章运用经济学理论系统评价经验估费制度。依据静态福利分析,BMS不能改进固定保险范围的强制车险的市场效率,而对于可变保险范围的商业车险能起到优化保险市场资源配置效率的作用。此外,BMS对于信息不对称条件下投保人可能表现出来的逆向选择、道德风险和保险欺诈等机会主义行为具有抑制作用。因此,在车险业务中引入BMS具有很好的实践意义。第4章探讨车险费率分级机制的两个层面如何整合运作的问题。首先,从理论上揭示出两者之间应有着交替、互斥的关系;其次,说明现有的最优BMS精算模型存在的问题,即割裂了两个层面的内在联系,造成投保人面临双重惩罚或奖励的可能;由此,提出一个整合先验分级和BMS的保险定价模型,使得BMS仅针对先验分级后残余的风险异质性区分初始保费;最后,基于实际数据对比两类精算模型,验证了整合效果。第5章讨论BMS条件下动态道德风险的两种表现形式:出险依赖效应和合约时间效应,并给出动态道德风险的规范策略。当保费仅占投保人的收入的极小比重时,出险依赖效应成立。即存在道德风险时,随着事故的边际成本递增(即续期保费增加),投保人的驾车行为会更加谨慎。非参数假设检验揭示出我国车险市场上存在合约时间效应。即在动态激励条件下投保人的行为偏好呈现出非平稳性的特点,越是临近保险契约的调整时点,驾车行为越趋于谨慎。第6章研究奖金饥渴症的产生根源及其治理机制。奖金饥渴症表现为投保人为了获得优惠费率而自担小额损失。奖金饥渴症截断投保人的真实损失分布,增大了保险人的定价风险,并会使BMS陷入财务失衡的困境。BMS条件下投保人的索赔策略可看作自留损失和保险索赔的融资成本比较的结果。奖金饥渴症的成形基础是保险索赔的融资成本高于自留损失的融资成本。设置合理的契约免赔额和优化BMS奖惩幅度都是降低BMS条件保险索赔的融资成本的可行途径,进而对奖金饥渴症起到抑制效果。第7章总结本文的研究工作,联系我国车险业务的实践状况,给出具体的启示或建议,最后指出本文研究的不足和后续研究的方向。本文的主要创新之处如下:1.改进Chiappori & Salanié(2000)的计量模型,构建多元Probit模型评价车险先验分级的效率,结合我国车险业务的保单数据开展实证分析。一方面,多元Probit模型可以对先验分级变量的分级效果作出统计检验,对于筛选适宜的先验分级变量、定义分级变量的取值类别都有一定的指导作用:另一方面,借助于判断先验分级之后车险业务中是否残存逆向选择,多元Probit模型对先验分级效率的全局评价显得更加稳健,实证分析的结果验证了这一点。2.纠正现有的最优BMS精算模型的缺陷,提出整合先验分级的BMS精算模型。理论分析表明,先验分级和BMS应表现为交替、互斥的关系。现有的最优BMS精算模型是基于保单组合的风险异质性而不是先验分级后残余的风险异质性来推算奖惩系数,不可避免地对投保人实施了双重惩罚或者奖励;而整合先验分级的BMS精算模型仅以残余的风险异质性为基准调整初始保费,实证研究证实改进模型起到了理顺两个分级环节的运作关系的效果。3.梳理动态道德风险的两种表现形式:出险依赖效应和合约时间效应,并给出各自的规范策略。构建博弈分析模型,讨论保费水平的变动与投保人谨慎程度的投入之间的关系,表明BMS框架下存在出险依赖效应,即出险概率与索赔成本之间有着负向联系。递增的免赔额体系可用于强化出险依赖效应,使得投保人的风险属性相对稳定。设计非参数假设检验方法,揭示出我国车险市场上存在着合约时间效应。并且,对保险期和考察期进行错位安排是缓和合约时间效应的可行思路。4.结合奖金饥渴症的产生根源,归纳奖金饥渴症的治理思路:确定最优契约免赔额和优化BMS的奖惩幅度。两种治理思路具有相同的出发点,即控制保险赔付的损失融资成本,使得投保人自担损失的融资成本尽可能地高于保险赔付的损失融资成本,以此削弱奖金饥渴症产生的根基。具体来看,契约免赔额d和投保人的自担免赔额z(d)之间满足z′(d)=-1时达最优;而基于指数损失函数厘定的BMS系数有助于优化设置等级保费之间的奖惩幅度。更多还原

【Abstract】 Since the early days of the twentieth century, motor insurance has experienced an emerging and surging process of development along with the invention and popularization of automobile, becoming one of the important policies in the social economy. Whether a property insurance company can be successful or not, which rests with the performance of the motor insurance line. As premium represents essential cash flow for the operating process or value-added chain of the insurance company, premium scheme is the key element to ensure its solvency. Being an applied branch of experience rating, Bonus-Malus system (BMS) is an indispensable component to the premium scheme of motor insurance, which also is an important carrier to reflect the transaction relationship between the insured and the insurer. In practice, actuaries have summarized a set of methodologies to construct the optimal BMS, but they don’t take into account the behavioral performance of the insured and its impact on the operation of BMS. Based on the above consideration, the paper breaks through the existed framework to study BMS, while reviewing it from a new angle upon the behavioral game between the two parties. The paper attempts to provide theoretical directions for the practice of Chinese motor insurance industry, meanwhile seeking out proper ways to graft economics to actuarial science.Following the research purpose, firstly the paper summarizes the relative literatures, to establish the analytical perspective of behavioral game. Decomposing the premium classification mechanism into two layers (i.e. ex ante classification and BMS), it pays special attention to conclude the operating modes of BMS internationally, combined with assessing the premium classification efficiency of Chinese motor insurance. Backed with the full acquaintance of BMS practice, it relies on economics to verify the rationality of introducing BMS into motor insurance. Thus, the paper comes to discuss certain problems need to be settled. In order to optimize the transaction interests, it probes into the problem how to integrate ex ante classification with BMS. It gives prominence to analyze the asymmetric information problems (i.e. dynamic moral hazard and hunger for bonus) the policyholder may display under the BMS setting, furthermore, the governance strategies are put forward correspondingly. According to the study plan, the main conclusions drawn from the chapters are as follows:Chapter 1 sums up the domestic and abroad literatures on BMS, points out the limitation of BMS study from the angle of actuarial science purely, that is, ignoring the policyholder’s behavioral preference and arousing change risks to BMS pricing. Therefore, it suggests to use behavioral game to coordinate the transaction relationship between the insured and the insurer, the problems to be tackled are brought forward subsequently.Chapter 2 concludes the practical experience for the two layers of the premium classification mechanism, evaluates the practice of premium classification in Chinese motor insurance market. It shows that ex ante classification is lack of efficiency. BMS has deficiencies in certain aspects such as the span between bonus and malus, premium differentiation etc., although it is of the ability to converge to stationary probability distribution quickly.Chapter 3 assesses experience rating scheme systematically relying on economics theories. According to static welfare analysis, BMS can’t improve the market efficiency for statutory motor insurance with fixed coverage, while it plays a role to optimize the resource-allocating efficiency of insurance market for commercial motor insurance with variable coverage. Otherwise, BMS can restrain the opportunism behavior that policyholders may reveal under the circumstances of asymmetric information. In this case, it is of practical significance to introduce BMS into motor insurance business.Chapter 4 investigates the problem how to integrate the two layers of the premium classification mechanism. Firstly, it indicates that there exists substitutive and repulsive relationship between the two layers in theory. Secondly, it explains the limitation that the current actuarial model of the optimal BMS has, that is, segmenting the inherent linkage between the two layers, which results in the possibility that the policyholder confronts dual punishment or reward. Then, an insurance pricing model integrated ex ant classification with BMS is set up, which follows the idea that the degree of BMS differentiating base premium is only aimed to the residual risk heterogeneity after ex ant classification. Finally, a comparison between the former model and the integrated one is carried out on the basis of actual data, which validate the hypothetic effects on integration.Chapter 5 discusses the two representations of dynamic moral hazard under the circumstances of BMS, i.e. occurrence dependence effect and contract-time one, and governance strategies on dynamic moral hazard are provided. When premium just accounts for a tiny proportion to the policyholder’s income, occurrence dependence effect comes into existence. That is, if moral hazard does exist, along with the increasing of the marginal cost for an accident, the policyholder’s driving behavior becomes more cautious. It finds that there exists contract-time effect in Chinese motor insurance market by non-parameter hypothesis testing. That is, the policyholder’s behavioral preference isn’t stationary under dynamic incentive situations. The more near to the adjusting time of insurance contract, the more cautious the policyholder’s driving behavior shows.Chapter 6 studies the origin of hunger for bonus and the corresponding governance strategies. Hunger for bonus is a behavioral tendency that the policyholder bears small losses to obtain favorable renewal premium. It is obvious that hunger for bonus truncate the policyholder’s true loss distribution, which increases the pricing risks the insurer faces and gets BMS into the puzzledom of financial disbalance. Under the framework of BMS, the policyholder’s claim strategy can be regarded as choosing a preferable means of loss financing by comparing financing costs between retention and claim. Therefore, the foundation of hunger for bonus is that the financing cost corresponding to claim is greater than that corresponding to retention. As setting proper contractual deductible and optimizing the bonus-malus scales are feasible routes to reduce the financing cost corresponding to claim under the framework of BMS, they both exert restrainable effects on hunger for bonus.Chapter 7 summarizes the whole research of the paper. Focusing on the practical status of Chinese motor insurance, it come out some revelations or suggestions, the deficiencies and further research directions being put forward together. The main innovations of the paper are as follows:1. Through modifying the econometric model established by Chiappori & Salanie (2000), it constructs a multivariable Probit model to evaluate the efficiency of ex ant classification in motor insurance, attached with an empirical analysis using Chinese policy data. On the one hand, multivariable Probit model can be used to test the differentiating effect of a classifying variable, so it has certain directions to screen proper ex ant classifying variable and define its value space. On the other hand, relying on judging whether there exists residual adverse selection or not among motor insurance business, the overall evaluation on the efficiency of ex ant classification obtained by multivariable Probit model seems to be more reliable, which is verified by the empirical analysis.2. To rectify the defect involved in the current actuarial model of the optimal BMS, it brings forward a BMS actuarial model integrated with ex ant classification. Theoretical analysis shows that the relationship between ex ant classification and BMS should be a substitutive and repulsive one. The current actuarial model of the optimal BMS is based on the risk heterogeneity among the portfolio to calculate bonus-malus coefficients rather than the residual risk heterogeneity after ex ant classification, so it is inevitable to implement dual punishment or reward to the policyholders. However, the BMS actuarial model adjusts the base premium just on the residual heterogeneity. Empirical analysis proves that the new model does improve the relationship between the two layers.3. Summarizing the two representations, i.e. occurrence dependence effect and contract-time one, it also provides corresponding governance strategies. Through constructing a game model, it discusses the relationship between the premium variation and the cautiousness the policyholder exerts, and reflects that there exists occurrence dependence under the framework of BMS, that is, there is negative relationship between occurrence probability and claim cost. Ascending deductible scheme is suggested to consolidate occurrence dependence effect, for purpose of stabilizing the policyholder’s risk attribute. By designing a non-parameter hypothesis testing, it discloses that there is contract-time effect in Chinese motor insurance market. Furthermore, staggering the coverage period and the recording one is a credible way to alleviate contract-time effect.4. Combined with analyzing the origin of hunger for bonus, it concludes its governance means, i.e. setting proper contractual deductible and optimizing the bonus-malus scales. The two governance means share the same idea that controlling the loss financing cost corresponding to claim, and make the relative cost corresponding to retention is greater than that corresponding to claim, in order toweaken the foundation of hunger for bonus. Properly speaking, only when contractual deductible(d) and the policyholder’s retentive deductible z(d) satisfy z’(d) = -1,contractual deductible reaches optimization. As for optimizing the bonus-malus scales, it should rely on exponential loss function to get bonus-malus coefficients.更多还原