随着经济发展、人口增长和城市化进程加快，我国灾害发生的频率和损失程度均在不断上升，由此引发的巨灾损失亦不断地增加，巨灾则不断地挑战保险公司的偿付能力。如果积累的巨灾准备金不能满足经营需要，很有可能导致保险公司的破产。目前，我国关于财产保险巨灾风险准备金测算的研究还相对较少，实践亟需相关研究的指导。本文从保险公司整体经营角度出发，从财务报表的角度，研究其面临的总体巨灾风险状况，并计算风险准备金的合理计提方法。

本文利用2003年至2015年广西某财产保险公司的保费收入与保险赔案数据，开展了巨灾风险准备金测算的研究。首先分别针对现金流的进项保费收入的时空特征、现金流的出项总赔付金额的时空特征以及引起总赔付金额极具跳跃的巨灾赔付的时空特征分析，在此基础上，运用时间序列的自回归移动平均自回归平均模型建模对保费收入、普通赔案损失进行拟合及预测，运用嵌入POT模型（Peak Over Thresholds，门限极值模型）的GPD分布（General Pareto Distribution，广义帕累托分布）对单次巨灾发生的损失进行拟合与预测，利用泊松过程的指数分布对巨灾发生的频率预测，并运用蒙特卡洛方法仿真未来10年120期的保费收入、普通赔案损失、巨灾损失，随机产生5000次仿真结果。最后在一定的假设前提下，运用现金流分析方法，建立研究区的准备金盈余动态变化模型，并采用敏感度分析的方式，测算研究区的巨灾风险准备金提取规模。经过上述研究，得到以下结论：

1）保费收入与保险赔付均随时间呈现明显的增长趋势，巨灾的发生频率、赔款金额均存在较大的波动。研究区的巨灾赔付主要由雪灾、台风、暴雨、冰雹、其他原因五类致灾因子引起，其中占比最高的是雪灾，有超过80%的巨灾赔付来自2008年雪灾的影响；其次是台风，占比16%。在空间格局上，巨灾赔付较大的几个城市分别为南宁、来宾、柳州、钦州、防城港、百色、玉林、北海。

2）对保费收入、普通赔案的赔付、巨灾赔付时间序列的拟合和仿真显示，保费收入、普通赔案在未来10年有一定的趋势增长，月度保费收入有92.71%的概率在10亿元以内，月度普通赔案的赔付较现金流入项的变化程度平缓，月度普通赔案的赔付有88.97%的概率在10亿元以内；巨灾发生具有偶然性，有65.8%的概率在1亿以内，24.2%的概率巨灾损失在[1,10]亿内，并且保费收入、普通赔款、巨灾赔付两两之间的无相关性。

3）对保险公司经营的稳定性与财务可持续性的情景模拟结果显示，随着巨灾风险准备金计提比例的提高，准备金被击穿的概率随之降低，准备金首次击穿的时间推后；其中，计提比例每上升一个百分点，击穿概率下降约0.22个百分点，首次击穿时间推后约8期。保险公司经营的稳定性随之增强。但同时保险公司保费利润率也随之降低，计提比例每上升一个百分点，保费利润率在20%及以上的概率降低0.92个百分点。依据敏感性测试结果，建议该财险公司选取保费收入的[6%, 12%]作为巨灾风险准备金的储备，以稳定保险公司经营。

本论文的结果立足于广西某财险公司的现实经营环境进行测算，从保险公司财务报表的思路，以月度为统计标准，打破常见的单一巨灾风险准备金的定式提取模式，将研究的巨灾风险的种类扩展为所在区域的全部自然灾害，并拓宽险种的产品维度，将研究的巨灾风险准备金扩展为保险公司经营的所有保险险种的巨灾风险准备金评估思路，运用现金流量分析、敏感度分析方法测算出研究区的巨灾风险准备金提取规模。相关测算方法可以为财险公司测算巨灾风险准备金提供支撑，研究结果也可为广西当地公司的巨灾风险准备金科学提取提供依据。

With the acceleration of economic development, population growth and urbanization, the frequency and loss of disasters in our country are constantly increasing, and the resulting catastrophic losses are also increasing. The catastrophic disasters constantly challenge the ability of insurance companies to repay. If the accumulated catastrophe reserves can not meet the operational needs, it is very likely to lead to the bankruptcy of the insurance company. At present, the study of catastrophic risk at home and abroad is mainly used at the pricing level, and the calculation of catastrophic risk reserve is less. This article mainly studies the overall catastrophic risk reserve from the perspective of the overall business of the insurance company and from the perspective of the financial statements. The reasonable reserve and calculation method of specific risk reserve are put forward.

Based on the data of premium income and insurance losses of a property insurance company in Guangxi from 2003 to 2015, this paper studies the calculation of catastrophe risk reserve. Firstly, aiming at the spatio-temporal characteristics of the input premium income of cash flow, the spatio-temporal characteristics of the total payout amount of cash flow and the spatio-temporal characteristics of catastrophe payout, which causes the total payout amount to jump greatly, respectively, on this basis, the seasonal ARIMA model of time series is used to fit and predict the premium income and the loss of ordinary payout cases, and the embedded POT model, The GPD distribution is used to fit and predict the losses caused by a single catastrophe, the exponential distribution of Poisson process is used to predict the frequency of catastrophe, and Monte Carlo method is used to simulate the premium income, ordinary compensation losses and catastrophe losses for 120 periods in the next 10 years, with 5,000 simulation results generated randomly. Finally, under certain assumptions, the cash flow analysis method is used to establish the dynamic change model of reserve surplus in the study area, and the sensitivity analysis method is used to measure the extraction scale of catastrophe risk reserve in the study area. After the above research, the following conclusions are obtained:

1) There is a certain trend between insurance compensation and monthly time. With the passage of time, insurance compensation increases at a certain speed. However, due to the occurrence of catastrophe, there are years in which compensation occasionally jumps sharply, such as the snowstorm in 2008. Catastrophes in the study area are mainly caused by five types of disaster-causing factors, including snow disaster, typhoon, rainstorm, hail and other causes. Among them, snow disaster accounts for the highest proportion of disaster-causing compensation for natural disasters, followed by typhoon, and over 80% of disaster-causing compensation comes from the impact of snow disaster in 2008. The second is the impact of typhoons, accounting for 16%. The cities with large compensation for the catastrophe are Nanning, Laibin, Liuzhou, Qinzhou, Fangcheng Port, Baise, Yulin and Beihai. Most of the losses were caused by the snowstorm in 2008, while Fangcheng Port, Qinzhou and Beihai are located near the sea and vulnerable to typhoons. According to the characteristics of premium income, compensation in ordinary compensation cases and catastrophe compensation, seasonal ARIMA model and POT model based on GPD distribution can be used for breathtaking prediction.

2) The stability and sustainability of insurance company's operation are related to the scale of catastrophe risk reserve provision. The author set up eight scenarios to conduct sensitivity tests. When the scale of catastrophe risk reserve increases, the probability of reserve breakdown decreases, and the time for first breakdown of reserve may be postponed, thus enhancing the stability of insurance company's operation. However, there is also a decrease in the operating efficiency of insurance companies. Through comparative analysis, it is suggested that a property insurance company in Guangxi select [6%,12%] of the premium income as the reserve for catastrophe risk to stabilize the operation of insurance companies.

The results of this paper are based on the actual operating environment of a property insurance company in Guangxi. Although there are some deficiencies, it can provide practical reference for scientific extraction of catastrophe risk reserve for the company.