Preferences for Crop Insurance Programs among Smallholder Cocoa Farmers in Ghana

本研究旨在通过分析农民对各种经济作物保险计划，特别是传统保险计划的偏好，为农作物保险计划的文献做出贡献。鉴于天气的不确定性和不完善的市场的重要性，我们开发了一个理论框架，以显示如何生产风险和流动性约束影响农民作物保险的选择。 在某种程度上，作物保险购买决策的数据在加纳是不可用的，我们打算进行一个选择实验调查，以了解农民的偏好，在该国的阿散蒂和西部地区的各种作物保险计划。然后将分两个阶段调查作物保险计划的偏好。研究的第一阶段将考虑决定参加或不参加，使用离散选择模型，然后选择替代作物保险计划，如果农民表示愿意参加，使用多项logit模型。第二阶段的分析将使用选择实验数据来检查农民的偏好和支付意愿（WTP）的作物保险属性。我们将采用Hess和Stathopoulos（2013）提出的混合潜变量模型框架，其中调查参与度指标的数据与受访者对所述选择问题的回答联合建模。这种方法明显地解释了由受访者之间潜在的不同参与程度所造成的规模异质性，同时也克服了内隐偏差和测量误差的问题。 最后，我们将使用Mariel et al.（2015）开发的混合潜在类方法来研究农民之间的偏好异质性和异质性来源。

The study aims at contributing to the literature on crop insurance programs by analyzing farmers preferences for various cash crop insurance programs, particularly the traditional insurance programs. Given the significance of weather uncertainty and imperfect markets, we develop a theoretical framework to show how production risks and liquidity constraints affect farmers crop insurance choices. To the extent that data on crop insurance purchase decsions are not available in Ghana, we intend to conduct a choice experiment survey to elicit farmers preferences for various crop insurance programs in the Ashanti and Western regions in the country. Preferences for crop insurance programs will then be investigated in a two-stage process. The first-stage of the study will consider the decision to participate or not to participate, using a discrete choice model and then the choice among alternate crop insurance programs, given that the farmer has expressed the willingness to participate in it, using a multinomial logit model. The second-stage of the analysis will use the choice experiment data to examine farmers preferences and willingness-to-pay (WTP) for crop insurance attributes. We will employ the hybrid latent variable model framework proposed by Hess and Stathopoulos (2013), in which data on indicators of survey engagement are jointly modeled with respondents answers to the stated choice questions. This approach distinctly accounts for scale heterogeneity caused by potentially different levels of engagement among the respondents, while also overcoming the problems of endogeneity bias and measurement errors. Finally, we will examine preference heterogeneity and the sources of heterogeneity among the farmers, using the hybrid latent class approach developed by Mariel et al. (2015).