Leveraging the Power of Household Surveys in Agricultural Research for Development

Abstract

The number of smallholders globally is estimated to lie between 300 and 580 million. These farmers are important for global food production, but bear the brunt of the world’s social, economic, environmental, and health challenges. Agricultural Research for Development (AR4D) seeks to: 1) develop interventions that can improve smallholder productivity and welfare; 2) inform strategies for deploying these interventions that target those who need them. To achieve this, AR4D needs a coherent body of evidence about the heterogeneity of smallholders, their practices, and their environment in relation to development outcomes.

I present a suite of tools which could help researchers use the data they are already collecting to better evaluate and target development interventions. A systematic review was conducted to assess whether current research practices in AR4D would enable the research community to build a body of evidence on smallholder heterogeneity in relation to development outcomes. Farm surveys were the main source of data in this type of research. A lack of coordination, openness, and transferability prevents researchers from using disparate farm surveys for the synthesis of knowledge. In collaboration with experts in the design and implementation of household surveys, I identified the necessary characteristics of a community-maintained tool, that could address these issues. I designed software that exhibits these characteristics using principles of sustainable research software engineering. The initial prototype facilitates the collection, processing, and sharing of interoperable farm survey data. In collaboration with members of the AR4D community, I assess the utility of this tool, and discuss procedures for ongoing contribution and maintenance. A concrete example of what the analysis of harmonised household survey data can provide is given. I outline a procedure for combining unilateral survey efforts, linking to spatial data sources, and modelling spatial patterns of development outcomes. This analysis shows the importance of local-level heterogeneity in development outcomes.

The software and analysis procedures are imperfect, but this thesis has shown that they can adapt to meet the needs of the AR4D community. Increased funding, changes in research culture, and institutional incentives will be needed to accelerate the widespread uptake and impact of the work presented in this thesis.

Date of Award	18 Jun 2024
Original language	English
Awarding Institution	University of Bristol
Sponsors	The Alan Turing Institute
Supervisor	Andrew Dowsey (Supervisor), Jim Hammond (Supervisor), Christopher J Woods (Supervisor), William J Browne (Supervisor) & Mark C Eisler (Supervisor)