Fortifying food systems over the coming decades will be critical in sub-Saharan Africa, where crop yields could decline by 5 to 17 percent by 2050, despite a rapidly growing population. In the face of climate change, it’s crucial that every investment made to improve and adapt agriculture returns results. But how do you evaluate in a sustainable and efficient way whether diverse agricultural programs—from soil testing in Kenya or land rights in Namibia—led to benefits, especially when every dollar spent on evaluation is a dollar less spent on implementation?
This was the question we tackled at a recent workshop with ReNAPRI, the Regional Network of Agricultural Policy Research Institutes, on the power of using geospatial data and tools for impact evaluations.
Impact evaluations are critical to understanding how development programs impact local communities and if they are having the intended results. However, traditional approaches like Randomized Control Trials (RCTs), where a treatment (the program) is randomly assigned, can be difficult to carry out. RCTs need to be planned far in advance, require multiple rounds of data collection, and can only be implemented in circumstances where randomization of treatment is ethical. To overcome these cost and time barriers, AidData often employs Geospatial Impact Evaluations (GIEs) so we can use Earth Observation (EO) data—information we get from satellites—and cutting-edge statistical methods to help us mimic the conditions of an RCT, while also allowing researchers to conduct the evaluation retrospectively if needed.
However, the barriers are quite high to confidently undertaking a GIE. Multiple studies of evaluations of development lending and spending find that less than 10% are impact evaluations, and of those, less than 40% used relevant data or had high analytical validity. To do it right, evaluators need access to the right information; programs need to have outcomes that can be measured using geospatial data; and advanced statistical methods need to be accurately used. Given these constraints, there has been slow uptake among researchers in sub-Saharan Africa to using GIE methods.
As part of our work with the Hewlett Foundation, we are keenly interested in building the capacity of African researchers to integrate the GIE methodology and apply it in their own work. In November 2023, we teamed up with the Regional Network of Agricultural Policy Research Institutes (ReNAPRI) to conduct a GIE training session during their yearly research conference. ReNAPRI is a Pan-African network of 16 agricultural policy research institutions located in 15 countries. Their purpose is to provide a way for like-minded research centers to collaborate around mutual topics of interest and increase the technical capacity of their staff to undertake agricultural research.
The ReNAPRI network hosted their 10th Annual Stakeholder Conference in Victoria Falls, Zimbabwe, where research directors from each of the institutes along with other interested stakeholders gathered to share insights around the theme of “Fortifying Africa’s food system through increased productivity, climate resilience and adaptation.” Prior to the main event, we held a one-day GIE training for the research directors which included a discussion on the benefits of GIEs, program and data requirements, and the statistical methods required for the analysis.
After spending the first half of the day discussing methods for conducting GIEs, we spent the second half workshopping potential GIE projects from the training participants. This approach allowed participants to share their own thoughts about the project, help assess what further information was needed for a successful GIE, and determine which statistical methods would be best to use. We also discussed available training guides for participants, including AidData’s Toolkit for Agricultural Geospatial Impact Evaluations.
The projects spanned a diversity of topics, from soil testing in Kenya to land rights in Namibia. Each researcher had 5 minutes to present their project followed by Q&A time with us and the other ReNAPRI researchers. We also provided suggestions on everything from how to build a research design around the project and how to add geospatial data to an existing impact evaluation, to where to find available and relevant data. As each additional researcher presented their potential GIE project, the other training participants felt more comfortable contributing to the workshopping of the GIE design.
We look forward to continuing the conversation with ReNAPRI research directors and their staff around promising GIE opportunities. Due to the complex nature of GIEs, we’ve found that learning these methods is best done when applied to a real-world project, and we’ll supply sustaining support for ReNAPRI researchers as they further explore geospatial research.
