Conservation Agriculture 2009-2011 - Malawi

Abstract
---------------------------

The randomized control trial impact evaluation tests different strategies for communicating information about agricultural technologies to smallholder maize farmers in 8 districts in Malawi. The objective is to provide information to the Ministry of Agriculture and Food Security as to how best to use its limited resources to increase rates of adoption of new technologies. There are four primary dimensions to the evaluation: agricultural technologies, communication methods, incentives and gender.

Geographic coverage
---------------------------

Regional Coverage

Analysis unit
---------------------------

Households

Kind of data
---------------------------

Sample survey data [ssd]

Sampling procedure
---------------------------

District Selection:
Out of the 12 districts scheduled to be included in the ADP-SP in 2009-10, 8 were chosen as evaluation sites. Four are dry districts where pit planting is relevant: Balaka, Chikwawa, Neno, and Rumphi. Composting was promoted in the other four districts: Dedza, Mchinji, Mzimba, and Zomba. Together, these districts cover the major agro-ecological zones of Malawi and are spread through the South, Central, and Northern regions. District selection was not random; rather, it was based on the schedule for ADP-SP and the relevance of the technologies we are interested in.

Selection of Sections and Villages:
From a list of all the sections in the 8 districts staffed by an extension worker, 60 sections were randomly selected from the 4 districts assigned to conservation farming, and 60 sections from the 4 districts assigned to nutrient management. Because there are more districts staffed by AEDOs in the districts assigned to nutrient management, the probabilities of selection are not equal. For the CF districts, we chose 60 out of 176 possible districts. For the NM districts, 60 were chosen out of a possible 281. For each of the 120 selected treatment sections, one village was randomly selected from a list of all villages provided by DAES will provide a list of all the villages in the selected sections. The selection of the villages was weighted by the number of farm families per village.

Randomized Assignment of Evaluation Components:
To evaluate each of the four components of the project, certain subsets of the village were randomly selected for each component. Thus, there are four overlapping dimensions:
- Incentives: To address selection bias, sections were allocated to various treatment groups randomly. Of the 120 sections, 60 were randomly assigned to an "incentive" condition. Those selected for the incentive will be offered (but will not necessarily receive) a performance-based incentive.
- Communication Strategies: Next, the type of communication strategy for the section was randomly assigned. 25 are randomly assigned to "extension worker" (AEDO) status, 50 to Lead Farmer (LF) status, and the final 45 to "Peer Farmer" (PF) status. Note that while extension workers continued to be used in all areas (in some cases communicating directly to farmers and in others communicating through peer or lead farmers), the evaluation focused on different communicators (AEDO or LF or PF) in different areas.
- Gender: For the 50 LF villages, the gender of the lead farmer was randomly assigned. 25 LF villages were assigned to male lead farmers (LF-M), and 25 others were assigned to female lead farmers (LF-F). Out of the 45 PF villages, 22 were randomly assigned to have majority men among the set of peer farmers (PF-M), and the other 23 were randomly assigned to have majority women (PF-F). In other words, we encouraged these villages to choose more peer farmers from the assigned gender rather than the other gender.

摘要
---------------------------
本随机对照试验影响评估旨在测试向马拉维8个地区的自耕农玉米种植者传达农业技术信息的不同策略。其目标是向农业与粮食安全部提供信息，以指导其如何最佳利用有限的资源，提高新技术的采用率。评估主要包含四个维度：农业技术、传播方法、激励措施和性别。
地理覆盖范围
---------------------------
区域覆盖
分析单元
---------------------------
家庭
数据类型
---------------------------
样本调查数据 [ssd]
抽样程序
---------------------------
地区选择:
在2009-10年度计划纳入ADP-SP的12个地区中，选定了8个地区作为评估地点。其中4个为干旱地区，坑种技术在其中适用：巴拉卡、奇库瓦、内诺和鲁姆菲。其他4个地区推广了堆肥技术：德德扎、奇钦吉、姆齐马和宗巴。这些地区共同涵盖了马拉维的主要农业生态区，并分布在南部、中部和北部地区。地区选择并非随机，而是基于ADP-SP的时间表以及我们感兴趣技术的相关性。
区域和村庄的选择:
从8个地区中所有由推广员负责的区域列表中，从分配给保育农业的4个地区中随机选择了60个区域，从分配给养分管理的4个地区中随机选择了60个区域。由于养分管理地区的地区中有更多的AEDO工作人员，选择概率并不相等。对于保育农业地区，我们从176个可能的地区中选择了60个。对于养分管理地区，从281个可能的地区中选择了60个。对于每个选定的120个处理区域，从DAES提供的所有村庄列表中随机选择了一个村庄。村庄的选择根据每个村庄的农场家庭数量进行加权。
评估组件的随机分配:
为了评估项目的四个组成部分，对村庄的某些子集进行了每个组件的随机选择。因此，存在四个重叠的维度：
- 激励措施：为了解决选择偏差，区域被随机分配到不同的处理组。在120个区域中，有60个被随机分配到“激励”条件。被选中的激励措施将提供（但不一定收到）基于绩效的激励。
- 传播策略：接下来，随机分配了区域的使用通信策略。25个被随机分配到“推广员”（AEDO）状态，50个分配到“主要种植者”（LF）状态，最后的45个分配到“同行种植者”（PF）状态。请注意，尽管推广员在所有地区都继续使用（在某些情况下直接向农民传达，在其他情况下通过同行或主要种植者传达），但评估重点集中在不同地区的不同传播者（AEDO或LF或PF）。
- 性别：对于50个LF村庄，随机分配了主要种植者的性别。25个LF村庄被分配给男性主要种植者（LF-M），另外25个分配给女性主要种植者（LF-F）。在45个PF村庄中，22个被随机分配到主要由男性同行种植者组成（PF-M），另外23个被随机分配到主要由女性组成（PF-F）。换句话说，我们鼓励这些村庄从指定的性别中而不是其他性别中选择更多的同行种植者。