A Behavioural Economic Analysis of Agricultural Investment Decisions in Uganda

发展中国家的农民在极其危险的环境中作业。旨在提高这种环境下农业生产率的政策--农业贷款、研究和推广、农村基础设施、作物保险等政策--需要考虑到农民如何做出农业投资决策。然而，他们的决策习惯中与高风险选择相关的关键因素仍然没有得到充分的理解。在这里，我们建议使用行为经济学研究方法在一些重要方面推进我们的理解。我们在这里提出的研究是新的，对于理解危险环境中包容性农业增长的行为制约因素，从而做出有效的政策回应，具有潜在的深远影响。为了展示这种潜力，我们选择了一个案例研究国家乌干达，在那里(A)由于非凡的农民投资行为已经实现了这种增长，以及(B)尽管如此，生产率增长的潜力仍然是大多数农业部门的特征，与新技术采用率低、农村市场稀薄和孤立、农民合作社失败等许多因素有关。为了阐明农民投资行为的动机，我们将使用经济实验，在这些实验中，参与者对真正的货币激励做出反应。在发展经济学中越来越多地使用经济实验；两位申请者都精通在发展背景下使用经济实验，他们所在的机构(国际开发学院和东安格利亚大学行为和实验社会科学中心)也是如此。这里提出的实验包含了许多新颖的设计特征，为了确保参与者理解说明，将进行试点。在实验中，参与者被要求从彩票中进行选择；他们选择的彩票是通过使用随机装置(例如有彩色球的袋子)实现的，然后计算并支付收入。抽奖的概率和可能的结果是系统地不同的，以衡量某些风险在参与者的脑海中是否倾向于被夸大(或淡化)。将实施几个不同的版本。首先，与钱只能赚(不能输)的标准方法不同，我们会在实验前几周交出资金(可能存在受试者在比赛日之前无法进入的银行账户)，这样我们就可以观察到与受试者已被视为自己的钱损失有关的风险选择行为。其次，为了测试是否存在风险被忽略的概率阈值，我们将通过允许受试者(以一定的价格)操纵结果的概率，让受试者控制他们承担的风险量。第三，为了研究社会互动对风险选择行为的影响，我们组成了随机的参与者对，他们的收入被汇集在一起，参与者被要求确定总收益应该如何重新分配。我们还将调查农民在考虑投资时是否存在偏见。要做到这一点，我们考虑使用研究区现有的关于给定土壤类型、农业投入等情况下产量分布的数据，这些数据可以转换为易于理解的概率陈述(“每英亩产量将超过X，Z倍中的Y”)。我们会向受访者描述一些情景，并邀请他们提出这样的声明，因为他们知道他们的回报会更大，声明越准确。这使我们能够发现决策中的任何偏见，然后这些偏见可能与我们在实验后进行的社会经济调查中获得的实际投资行为数据有关。

Farmers in developing countries operate in extraordinarily hazardous environments. Policies that aim to increase agricultural productivity in such environments - policies on agricultural lending, research and extension, rural infrastructure, crop insurance, and so forth - need to take into account how farmers take agricultural investment decisions. However, crucial elements of their decision-making habits that relate to risky choices remain ill-understood. Here we propose to use behavioural economic research methods to advance our understanding in a number of important ways.The research we propose here is new, and has potentially far-reaching implications for understanding the behavioural constraints on inclusive agricultural growth in hazardous environments, and thus for an effective policy response. In order to demonstrate this potential we have selected a case study country, Uganda, where (a) such growth due to remarkable farmer investment behaviour has been achieved, and (b) despite this, unrealised potential for productivity growth still characterises most of the agricultural sector, related to low adoption rates of new technologies, 'thin' and isolated rural markets, the failure of farmer cooperatives, among many other factors.In order to elucidate the motivation of farmers' investment behaviour, we will use economic experiments, in which participants respond to real monetary incentives. The use of economic experiments is increasing in development economics; both applicants are well-versed in their use in a development context, as are the institutions at which they are based (the School of International Development and the Centre for the Behavioural and Experimental Social Sciences (CBESS) at the University of East Anglia). The experiments proposed here contain a number of novel design features, which will be piloted in order to ensure that participants understand the instructions.In the experiments, participants are asked to make choices from among lotteries; the ones they choose are implemented with the use of a random device (e.g. a bag with coloured balls), after which earnings are calculated and paid out. Probabilities and possible outcomes of the lotteries are systematically varied, in order to measure whether certain risks tend to be exaggerated (or alternatively, downplayed) in participants' minds. Several variants will be implemented. First, in contrast to the standard approach where money can only be earned (not lost), we would hand over money for use in the experiments several weeks before (possibly in a bank account that subjects cannot access before game day), so that we observe risky choice behaviour that relates to losses of money that subjects have come to regard as their own. Second, in order to test whether a probability threshold exists below which risks are ignored, we would offer subjects control over the amount of risk they take through allowing them (at a price) to manipulate the probabilities of outcomes. Third, to study the influence of social interaction on risky choice behaviour, random pairs of participants are formed whose earnings are pooled and participants are asked to determine how the total earnings should be redistributed.We will also investigate whether farmers are biased when they consider making an investment. To do so, we have in mind to use existing data for the research area about the distribution of yields given soil type, agricultural inputs, and so forth, which can be converted into easily understood probabilistic statements ("yield will exceed X per acre, Y out of Z times"). We would describe to respondents a number of scenarios, and invite them to come up with such a statement, knowing that their reward will be greater, the more accurate the statement is. This allows us to detect any bias in decision-making that can then be related to the data on actual investment behaviour obtained in a socio-economic survey that we conduct after the experiments.