Experimental Design
Study Area
Central Mozambique is a diverse region when it comes to the production of maize. The drought risk faced by households varies significantly over relatively small regions and is influenced by altitude, soil differences, and proximity to rivers. At the same time, access to roads and distance from urban centers affects the availability of inputs and the development of markets for agricultural products, further complicating the production and marketing decisions of farm households. To be suitable for the questions considered in the study, potential sites thus needed be maize producing regions susceptible to drought and have minimal penetration by improved maize varieties.
Three districts which exemplified the above characteristics were chosen for the study: Machaze district in Manica province, Tambara districtd in Manica province; and Morrumbala district in Zambezia province. Working with department of agriculture extension agents and local partners, we then identified a set of feasible communities. For the purpose of this study feasible meant that a community was located in a maize growing area of the district, the community was accessible (though perhaps with difficulty) even during the rainy season, and the community was as insulated as possible from other study communities in the district.
Sample Selection and Randomization
In total a set of 72 communities were chosen (24 in each districts). All communities are in maize growing regions. The feasible set of communities was chosen to maximize the distance between each potential candidates subject to locating the required number of communities in the district. This selection ensures coverage over the various maize growing zones within each district and greatly reduces the chance of spillovers.
During the collection of baseline data, the political and security situation in the northern two districts (Tambara and Morrumbala) deteriorated and these sites became infeasible for research. This resulted in the loss of 48 study communities. A replacement district (Nhamatanda district in Sofala province) was chosen and baseline data collection resumed. Thirty-nine replacement communities were identified in Nhamatanda district, resulting in a total of sixty-three communities in the sample.
The study involves two treatment arms and a control group. In the first treatment arm, the research team and seed company partners undertook community-level marketing efforts for drought-tolerant maize (DTM) seed. The second treatment arm proceeded similarly, though in this set of communities the drought-tolerant seed was bundled with a satellite-based index insurance product. The bundle was available to all community members at a market rate. Control communities received no marketing intervention, only a yearly survey. Drought-tolerant maize seed is available commercially, though without insurance, and control community members could purchase it if they wanted to. Due to the isolated nature of study communities and extremely thin input supply markets, very few households outside of the treatment communities are expected to do so.
Given spatial variation in soils, weather, and access to inputs, our design is stratified geographically. Triads of proximate communities were constructed from the full set of feasible study villages. Randomization of treatments then took place within each triad—with one community being randomly assigned to control, one community to the DTM-only treatment and one being assigned to the DTM and Index-insurance treatment. Treatment was clustered, so all households in a community received the same treatment.
At the individual level, the research team worked with extension agents from the department of agriculture and local community leaders to obtain rosters of maize growing households in each study community. After household rosters were collected and verified for each community, households were assigned a number from a random draw and ordered accordingly. Twenty-one households in each community were then invited to take part in the study. These households would participate in a yearly survey and receive personal invitations to the community meetings (described below) at which the technologies being studied were introduced and promoted.
Description of Intervention
The intervention is best described as a community-level marketing intervention. Study participants, however, also received additional incentives to induce their attendance and participation in marketing activities and encourage experimentation with drought tolerant seed. Both community level activities and individual encouragements are detailed below.
In 2015 the research team and partner seed companies facilitated educational meetings in each treatment community. At these meetings, the characteristics and benefits of drought-tolerant maize seed were presented, as well as best practices and recommended levels of chemical fertilizer. Local extension agents from the department of agriculture were also present and answered questions about fertilizer use and other maize-related issues.
In the first year (2015), educational activities were uniform across both treatment arms. Communities received information on drought-tolerant seed only (without insurance). Study participants received personal invitations to these meetings, though all community members were informed about the meeting and invited to attend. Study participants also received 50 gram test packets of the seed varieties being promoted and a 1 kg test pack of one variety (which was chosen by lottery). All additional participants received only a 50 gram test packs. To put these quantities in perspective, the recommended seed use rate for maize is 25 kilograms per hectare. Additional kilogram-sized packs were available for sale to community members not participating in the study or study participants who wished to test the variety they did not receive through the lottery.
The two treatment arms diverged in 2016, which was the first year in which insured drought-tolerant maize seed was made available for purchase. As in the previous year, community meetings were organized through the department of agriculture and local leaders. At these meetings, two drought-tolerant maize seed varieties were presented, after which they were made available for purchase. All community members had been made aware, in advance, of the date and time of the meeting and were encouraged to attend. Households participating in the study also received personal invitations and were told that they would have the opportunity to participate in a lottery for discounts on the price of drought-tolerant seed purchased after the community meeting. Those communities in the bundled DTM satellite-based index insurance treatment arm also received educational sessions on the insurance product and the nature of index insurance. The seed purchased in these communities came bundled with the insurance product, which is described briefly below.
In 2017, the research team and seed company partners once again ran educational meetings and community seed sales. Households participating in the survey received personal invitations and random discounts as in the previous year.
Insurance Details
In one-third of communities, DTM was sold bundled with a satellite-based index insurance product. These technologies were not sold separately, though DTM seed does exist as a stand-alone product in the market. The inclusion of insurance raised the cost of the seed by about 20% and covered the replacement of the seed itself should one of the indices on which the contract was based be triggered. Two indices underlaid the insurance product. The first index is rainfall in the forty days following planting. If rainfall levels fall below a designated threshold over the first forty days after planting began in the community, the insurance will replace the insured seed the following year. The second index was a full-season prediction of maize yields, calibrated to historic production levels within the community. If full season yields were estimated to fall below a designated threshold, the insurance would replace the insured seed the following year.
The insurance was offered by Hollard Insurance, located in Maputo, Mozambique. Partner seed companies collected lists of farmers purchasing insured seed and provided these farmers with vouchers indicating their use of insured seed. The partner seed companies then remitted the insurance premiums to Hollard Insurance. In the event of an insurance payment, the seed company would replace farmers’ seed and collect the vouchers, which they would submit to Hollard Insurance to receive payment for the replaced seed.
Due to the index insurance design of the product. It was not necessary for farmers to report drought losses nor for the insurance company to send a claims adjuster to verify losses. The indices were monitored by the research team, which provided them to the insurance company. Results of the indices were communicated to farmers via SMS messages, both directly and through community leaders.
Complier definition
The marketing intervention described above took place at the community level. Its random assignment will allow us to make statements about the impacts of the availability of the technologies in question on various agricultural behaviors and household welfare measures. We will also be interested in the impact of these technologies on those households which choose to adopt them. We will refer to study households that adopt the technology promoted in their treatment arm when it is made available as compliers.