我国气象灾害频发、农业抗灾基础薄弱，农业生产面临各种风险。现代农业生产中，农业保险是分散农业风险、稳定农业生产的重要手段。我国现行的农业保险制度以政策性保险为主体，但政策性农业保险存在财政补贴制度不健全、道德风险和逆选择突出、交易成本较高等问题。天气指数保险是一种较新的农业保险类型，它以第三方气象部门提供的实际气象观测数据作为理赔依据，能够有效解决政策性农业保险面临的问题。 面对各种农业风险，农户的风险偏好是影响其农业生产决策和生产活动的一个重要因素。本文以微观经济学和行为经济学中的前景理论为基础，运用经济学实验方法，定量分析农户的风险偏好，并将风险偏好进一步细化为风险厌恶、损失厌恶和决策权重三个变量；结合调研数据，对以上三个风险偏好变量进行了描述分析，运用OLS回归模型从个人特征和家庭特征两方面分析了影响农户风险偏好的因素；基于Ajzen的计划行为理论，构建了研究农户购买天气指数保险决策的理论模型，以此为基础运用二项Logistic回归模型进行实证分析，旨在找出农户风险偏好的三个变量对购买农业保险决策的影响。 基于以上实证分析得出如下结论：（1）面对农业生产风险，农户进行生产决策时会受到风险偏好的影响；多数农户的风险偏好类型都表现为风险厌恶型，具有厌恶损失和高估小概率事件并低估大概率事件的决策倾向。（2）农户的风险偏好受到一定社会经济特征的影响。年龄越大，厌恶风险的程度越低；年龄越大、教育水平越低、耕作年限越短，厌恶损失的程度越高。（3）农户的风险偏好会影响其购买农业保险的决策。其中，损失厌恶越强烈，购买天气指数保险的可能越小；主要社会经济特征中，男性、年龄较小、耕作年限较长的农户，购买天气指数保险的意愿更高。针对以上结论，本文提出相关建议：增加农民收入和小麦种植天气指数保险的保费补贴，加强小麦种植天气指数保险的宣传工作，深化农户对农业风险和农业保险的认知。

Frequent meteorological disasters and weak disaster-resistant foundation bring various risks on agricultural productions in China. Agricultural insurance is an important means for risk diversification and steady production. The current agricultural insurance system in China takes policy-related insurances as the main body, which faces problems such as defective fiscal subsidies, moral hazards, adverse selections, and high transaction costs. On the other side, weather index insurances, which provide compensation on the basis of meteorological observation data from meteorological department rather than actual crop loss, could effectively solve these problems of traditional agricultural insurances.In the face of various agricultural risks, farmers’ risk preferences have important effects on their decision-making of agricultural activities. Firstly, based on the microeconomics and the Prospect Theory in behavioral economics, farmers’ risk preferences are refined into three variables quantitatively with methods of economic experiments: risk aversion, loss aversion and decision weighting. Secondly, descriptive statistics and an OLS regression are used to identify personal and household factors associated with farmers’ risk preferences. Thirdly, on the basis of Theory of Planned Behavior, theoretical model is established to analyse how farmers’ risk preferences affect their insurance purchasing decision, and is empirically evidenced by a binary Logistic regression. All the research results are as follows: (1) In the face of agricultural risks, farmers are under the influences of risk preferences, a majority of them are risk averse, and show psychological tendencies of loss aversion and a reverse S-shaped decision weighting function. (2) Farmers’ risk preferences are affected by a certain social and economic characteristics. The older farmers are less likely to be risk averse, while the older, less educated and less cultivating-experienced farmers are more likely to be loss aversion. (3) Farmers’ risk preferences affect their decision whether to cover weather index insurance. The less loss-averse ones are less possible to be covered, and the male, younger and more cultivating-experienced ones show higher willingness to be covered. Finally, three advices are brought out for the development of agricultural insurance: increase farmers’ income and premium subsidy, strengthen the publicity of weather index insurance, and promote farmers’ understanding of agricultural risks and insurance products.