非寿险未决赔款准备金稳健评估方法研究

发布时间：2018-03-05 14:13

本文选题：未决赔款准备金　切入点：案均赔款法　出处：《山东科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：未决赔款准备金是非寿险准备金中最重要的一部分,因此对未决赔款准备金的评估显得尤为重要。对于非寿险公司来说,由于某些准备金评估周期较长的特征,通常某事故年的最后一进展年的索赔额是未知的。因此,在公司对未决赔款准备金进行评估时,要考虑很多的影响因素,针对不同的数据特征也要选取不同的评估方法。事实上,在实际的保险案例中,非寿险公司在录入真实索赔数据时,可能会出现某些人为的输入错误,譬如小数点前移或者后移一位、少输入或多输入一位数等等,这都会导致最终索赔额的巨大变化。另外,还有一些非人为行为(如自然灾害、信用欺诈、经济危机等)所导致的索赔数据突然变大或者变小也是时有发生的,这些也会很大的影响准备金评估的结果,因此对未决赔款准备金评估方法的稳健性进行研究很有意义。本文首先介绍了非寿险实务中未决赔款准备金评估的常用确定性模型与方法,从这些模型的结构、假设条件和数据基础入手,全面分析了模型的特点。其次,以确定性方法中的案均赔款法为例,引入离群值,运用残差箱线图法和两点法对相关索赔数据进行离群值检验,然后针对离群值提出了一种稳健的案均赔款法,对进展因子和结案率数据的选取方式加以修正,并通过实证证明,稳健的案均赔款法可以对离群值进行有效的识别和调整,最终索赔额估计结果也比传统方法更加平稳。在此基础上由未决赔款准备金评估的确定性方法过渡到随机性方法中,首先介绍实务中常用的一些随机性模型,并分析各方法的优缺点。其次,通过在准备金评估的双广义线性模型中引入离群值,考虑离群值对未决赔款准备金评估的影响,基于拟似然估计方法,提出了一种稳健的估计值方程,来对离群值进行识别和调整,使得模型在处理存在离群值的数据时,可以得到更加稳健的评估结果。进而,更深层次的对双广义线性模型进行研究,调整赔付损失的进展模式,改善双广义线性模型的线性预估量,将不易量化的内部因素和外部因素引入到模型中,以此拓宽模型的适用范围,并改善模型的预测效果。最后,利用R软件对所提出的模型给出算例分析,得到相关结论。
[Abstract]:The outstanding claim reserve is the most important part of the non-life insurance reserve, so the evaluation of the pending claim reserve is particularly important. Usually the amount claimed in the last year of an accident year is unknown. Therefore, a number of factors need to be taken into account when companies assess outstanding claims reserves, In fact, in actual insurance cases, when non-life insurance companies enter true claim data, there may be some artificial input errors. If the decimal point moves forward or backward by one decimal place, by entering less or more digits, and so on, this can lead to a huge change in the amount of the final claim. In addition, there are also some non-human acts (such as natural disasters, credit fraud, etc.). The sudden increase or decrease in claim data caused by the economic crisis, etc., also happens from time to time, which can also have a significant impact on the outcome of reserve assessment, Therefore, it is very meaningful to study the robustness of the evaluation methods of pending claim reserve. Firstly, this paper introduces the commonly used deterministic models and methods in non-life insurance practice, from the structure of these models, Based on the hypothesis and data basis, the characteristics of the model are analyzed. Secondly, taking the case average compensation method in the deterministic method as an example, the outlier value is introduced, and the outlier values of the related claim data are tested by the residual box diagram method and the two-point method. Then, a robust average compensation method is proposed for outliers, which modifies the selection of progress factors and closure rate data, and proves that the robust average compensation method can effectively identify and adjust outliers' outliers. The result of final claim estimation is also more stable than the traditional method. On this basis, the deterministic method of evaluation of pending claim reserve is transferred to randomness method. Firstly, some stochastic models commonly used in practice are introduced. The advantages and disadvantages of each method are analyzed. Secondly, by introducing outliers into the bilinear model of reserve evaluation, considering the influence of outliers on the evaluation of pending compensation reserve, the quasi-likelihood estimation method is used. In this paper, a robust estimation equation is proposed to identify and adjust outliers, so that the model can obtain more robust evaluation results when dealing with data with outliers. The further research on the bilinear model, the adjustment of the progress model of compensation loss, the improvement of the linear predictor of the bilevel linear model, the introduction of internal and external factors which are not easy to quantify into the model. In order to widen the scope of application of the model and improve the prediction results of the model. Finally, an example analysis of the proposed model is given by using R software, and the relevant conclusions are obtained.
【学位授予单位】：山东科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F224;F840

【参考文献】