单产损失相关性是农业自然灾害的重要特征，对农业保险的赔付风险会存在重大影响，是影响农业风险分散效率的关键因素。个体损失间的相关性被认为是传统农业保险的主要缺点之一。由于指数保险的赔付主要依赖单一性的指数确定，研究中也认为指数保险自身也面临着严重的系统性风险。农户单产相关性对于传统保险和指数保险风险分散效率的有何影响？以风险分散效率为基准，是否存在临界农户单产相关系数，可以指导产品模式选择？开展相应的研究工作，可为区域保险模式的选取和大灾风险的转移提供科学参考。 本文选取区域产量指数保险（Area yield protection, AYP）和农户产量保险（Yield protection, YP）分别代表指数农业保险以及传统农业保险，利用农业保险大灾风险和参保农户福利变化作为衡量风险分散效率的指标，通过理论推导、蒙特卡罗仿真对不同农户单产相关系数下两类保险产品的优劣进行了对比分析。在此基础上，利用湖南省常德市9县区的794个入户调研数据对上述结果进行了验证，并对中国农业保险的模式选择等问题进行了讨论，以期对农业保险发展模式的选取，以及农业大灾风险分散机制的设计提供科学依据。论文得出的主要结论如下： 1）当农户单产相关系数较大，保障水平较低时，区域产量指数保险相较于农户产量保险拥有较低的大灾风险。任何假设情景下，区域产量指数保险保额损失率的期望值和最大可能损失均小于农户产量保险；只有当农户单产相关系数 ，且保障水平 时，区域产量指数保险保额损失率的标准差和大灾附加才会大于农户产量保险；在任何假设情景下，区域产量指数保险保额损失率的偏度大于农户产量保险。 2）只采用纯费率作为保险费率时，当农户单产相关系数较低时，农户产量保险对农户福利的提升要优于区域产量指数保险，反之，则区域产量指数保险更优。在以纯费率作为保险费率的前提下，无论是保成本还是保收益，单独购买农户产量保险或者区域产量指数保险，都会使得农户福利提升；当农户单产相关系数 时，农户产量保险对农户福利的提升要优于区域产量指数保险，当农户单产相关系数 时，则部分农户购买区域产量指数保险的福利提升效果更优。随着农户单产相关系数的增加，两种保险产品对农户福利提升的程度差距越来越小，农户单产相关系数越大，两者对农户的福利提升效果最为接近。 3）采用风险费率作为保险费率时，农户产量保险与区域产量指数保险各有优势。采用风险费率作为保险费率时，农户产量保险和区域产量指数保险对农户福利的提升效果低于以纯费率作为保险费率时的情况。当农户单产相关系数 或 时，农户产量保险对农户福利的提升效果优于区域产量指数保险；当农户单产相关系数 且保障水平较低时，部分农户购买区域产量指数保险对其福利的提升效果优于农户产量保险，保障水平较高时，农户购买区域产量指数保险会使得自身福利提升效果降低的可能性大于农户产量保险，此时农户产量保险更具有优势。 4）利用湖南省常德市9县区的794个入户调研数据较好的验证了仿真计算的结果。对于大灾风险的对比：9个县区的对比结果中有8个县的区域产量指数保险保额损失率的期望小于农户产量保险；44%的结果显示区域产量指数保险保额损失率的大灾附加小于农户产量保险；9个县区的对比结果中有7个县的区域产量指数保险保额损失率的最大可能损失小于农户产量保险。从分县结果来看，除安乡县因调研数据量较少，结果误差较大外，其余各县实际的参数对比指标与相似仿真情景下的参数对比指标的差值基本在0~0.46之间。上述结果与利用仿真数据计算的结果类似。 实际调研的农户单产相关系数 取值范围在0.2-0.48之间，对于两种保险的各种假设情景下，以纯费率作为保险费率时，农户产量保险对农户福利的提升效果要优于区域产量指数保险。以风险费率作为保险费率时，当农户单产相关系数 ，保障水平 =0.9，72%的计算结果显示农户产量保险对农户福利的提升效果更优。从分县结果来看，在各假设情景下，每个县实际的福利提升效果与相似仿真情景下的福利提升效果的差值基本在0.28~15.46之间。上述结果与利用仿真数据计算的结果类似。 上述研究结果表明：在中高农户单产相关系数水平下，区域产量指数保险的大灾风险更有可能超过农户产量保险，而此时区域产量指数保险在农户福利提升的效果上优于农户产量保险。考虑到大田作物农户间单产相关水平恰好处于中等水平，因此区域产量指数保险从风险分散效率的角度而言具有可行性。但特定区域内的保险模式选择，取决于当地的农户数量、农户单产相关系数以及保障水平的设置，需要经过实证分析和计算后才能确定。本研究通过一般性模型假定和数值仿真获得的临界相关性水平参数，可以在更广泛的场景中判定保险模式选择时得到应用。本文的研究框架和结果对保险模式的选取和大灾风险分散机制的建立均具有重要意义。

The correlation of yield loss is an important feature of agricultural natural disasters, and it will have a significant impact on the payment of agricultural insurance, which is a key factor affecting the efficiency of agricultural risk diversification. The correlation between individual losses is considered to be one of the main shortcomings of traditional agricultural insurance. However, since the payment of the index insurance mainly depends on the uniform index, the previous research also believes that index insurance also faces serious systemic risks. What is the impact of farmer yield correlation on the efficiency of traditional insurance and index insurance risk diversification? Based on the efficiency of risk diversification, is there a critical farmer yield correlation coefficient that can guide product model selection? Carrying out corresponding research can provide a scientific reference for the selection of regional insurance models and the catastrophic risk diversification. This paper selects Area Yield Protection (AYP) and Yield Protection (YP) as representatives of index agricultural insurance and traditional agricultural insurance. Using the catastrophic risk of agricultural insurance and the changes of welfare of insured farmers as indicators to measure the efficiency of risk diversification, through theoretical derivation, Monte Carlo simulation compares the advantages and disadvantages of two types of insurance products under the different coefficient of yield correlation. On this basis, the results were verified by using 794 survey data of the farmer level from 9 counties in Changde City, Hunan Province, and discuss the model selection of China's agricultural insurance from various angles, with a view to selecting the development model of agricultural insurance, and the design of the agricultural disaster catastrophic risk diversification mechanism provides a scientific basis. The main conclusions of the paper are as follows: 1) When the correlation coefficient between farmers' yields is larger and the coverage level is lower , the area yield protection has lower catastrophic risk than the yield protection. In any situation, the expectation of loss cost ratio and the possible maximum loss-cost ratio of the area yield protection are less than the yield protection; only when the farmer's yield correlation coefficient , the coverage level , the standard deviation and the catastrophic loading of loss cost ratio of the area yield protection are more than yield protection; in any situation, the skewness of loss cost ratio of the area yield protection is greater than the yield protection. 2) When only the actuarially fair premium rates are used as the premium rates, when the farmer's yield correlation coefficient is lower, the yield protection will improve the farmer's welfare better than the area yield protection. Otherwise, the area yield protection will be better. Under the actuarially fair premium rates as the premium rates, whether it is guaranteed cost or income, the purchase of the yield protection or the area yield protection separately will increase the farmer's welfare; when the farmer's yield correlation coefficient , the yield protection will improve the farmer's welfare better than the area yield protection. When the farmer's yield correlation coefficient , the welfare improvement effect of some farmers purchasing area yield protection is better. With the increase of the correlation coefficient of farmers' yields, the gap between the two insurance products to improve the welfare of farmers is getting smaller and smaller, and the correlation coefficient of farmers' yields is larger, and the effect of the two insurance products on the welfare improvement of farmers is the closest. 3) When using the risk rates as the premium rates, the yield protection and the area yield protection have their own advantages. When the risk rates are used as the premium rates, the yield protection and the area yield protection on the improvement of farmer welfare is lower than when the the actuarially fair premium rates are used as the premium rates. When the correlation coefficient of farmer's yield or , the improvement of the yield protection on the welfare of farmers is better than the area yield protection; when the farmer's yield coefficient and the coverage level is low , the improvement of the area yield protection on the welfare of farmers is better than the yield protection for some farmer; when the coverage level is high , purchasing the area yield protection makes the possibility of welfare reduction is greater than the yield protection. At this time, the yield protection has advantages. 4) The results of simulation calculations were verified by using 794 household survey data from 9 counties in Changde City, Hunan Province. Among the comparison results of 9 counties, the expectation of the loss cost ratio of the area yield protection are less than the yield protection in 8 counties; 44% of the results show that the catastrophic loading of loss cost ratio of the area yield protection is less than the yield protection; Among the comparison results of 9 counties, the possible maximum loss-cost ratio of the area yield protection are less than the yield protection in 7 counties; From the results of each county, except for the small amount of research data, the error of the result is large in Anxiang County, the difference between the actual parameter comparison index and the parameter comparison index under similar simulation scenarios is basically 0~0.46 in other counties. The above results are similar to those calculated using simulation data. The actual research on the yield coefficient of farmer ranges from 0.2 to 0.48, for different scenarios of two types insurance, when the actuarially fair premium rates is used as the premium rates, the yield protection is better than the area yield protection for the improvement of farmer welfare. When the risk rate is used as the premium rate, the farmer's yield correlation coefficient , the coverage level = 0.9, 72% of the calculation results show that the yield protection has a better effect on the improvement of farmers' welfare. From the results of each county, in each scenario, the difference between the actual welfare improvement of each county and the welfare improvement under similar simulation scenarios is basically 0.28~15.46. The above results are similar to those calculated using simulation data. The above research results show that: ① the yield coefficient of farmer ranges from 0.2 to 0.8, the catastrophic risk of the area yield protection is more likely to exceed the yield protection. ②the yield coefficient of farmer ranges from 0.2 to 0.8, the area yield protection is superior to the yield protection in terms of the improvement of farmers' welfare. For the selection of insurance models in specific regions, the selection should be based on the number of farmers, the yield coefficient of farmers and coverage level. This study uses model assumptions and numerical simulations, not for a specific region, and strives for a universal conclusion, in order to be applied in a wider range of scenarios. The research framework and results of this paper are of great significance for the selection of insurance models and the establishment of catastrophic risk diversification system.