Can Integrated Agriculture-Nutrition Programmes Change Gender Norms on Land and Asset Ownership? Evidence from Burkina Faso

Abstract: This article uses a mixed-methods approach to analyse the impact of an integrated agriculture and nutrition programme in Burkina Faso on women’s and men’s assets, and norms regarding ownership, use and control of assets. We use a cluster-randomised controlled trial to determine whether productive asset transfers and increased income-generating opportunities for women increase women’s assets over time. Qualitative work on gender norms finds that although men still own and control most assets, women have greater decision-making power and control over home gardens and their produce, and attitudes towards women owning property have become more favourable in treatment areas. The goal of HKI’s E-HFP programme in Burkina Faso was to improve maternal and child health and nutrition outcomes through a nutrition-sensitive agriculture programme targeted to women that includes a small transfer of agriculture and animal assets, training in optimal agriculture and animal-raising practices and optimal health and nutrition practices delivered through a behaviour change communication (BCC) strategy. These impacts are expected to come through three primary programme impact pathways: The programme may also improve maternal and child health and nutrition outcomes through increasing women’s access to and control over resources (Ruel & Alderman, 2013). These include resources such as additional income from the sale of products from home/village production activities, improved knowledge, skills and self-confidence in agriculture, health and nutrition gained through trainings, or increases in bargaining power through the programme’s transfer of productive assets. Changes in gender norms around asset control/ownership may also influence the programme’s potential impact on agriculture, health and nutrition outcomes. In Burkina Faso, two key components of the programme sought to directly increase women’s access to and control over agriculture-related assets. First, HKI worked with land owners in communities to identify and obtain rights to land that could be used for a Village Model Farm (VMF), run by women Village Farm Leaders (VFLs) who were beneficiaries of the programme. 5 The VMFs served as training sites for women to learn about homestead food production and small-animal rearing, to enable them to start their own home production activities, and, in some cases, as a place to work and harvest produce. 6 HKI supplied the VMFs with the agriculture and animal inputs (primarily chicks) at the beginning of the programme. In addition, HKI provided beneficiary women seeds, saplings, small gardening tools and chicks for their own home production activities. 7 HKI partnered with the International Food Policy Research Institute (IFPRI) to evaluate the programme’s impacts using a randomised control design coupled with two rounds of qualitative research. The quantitative longitudinal impact evaluation assessed the programme’s impact on outcomes including production, consumption, asset ownership, food security, health and nutrition-related knowledge and practices, and maternal and child health and nutrition outcomes, among others. The two rounds of qualitative research were used to understand how and why the programme did or did not have the expected impacts, as well as how the programme influenced women’s control over and ownership of productive assets, and related social norms toward this. Villages were selected according to a three-step process. First, Gourma Province in eastern Burkina Faso (Figure 1) was selected because HKI had experience implementing nutrition and health programmes in this area. Within this region, four departments were selected where HKI and other non-governmental organisations did not have much prior activity, to avoid biasing results due to participation in other (possibly similar) programmes. Second, within these four departments, villages that had access to water in the dry season and were therefore capable of carrying out a gardening project were identified (N = 55). A list of households with children under 12 months of age was then compiled for each of these 55 villages. Third, after stratifying the villages by department and village size to maintain a balanced distribution of geographic locations and village sizes between treatment and control villages, villages were randomly selected into 25 control villages and 30 treatment villages. Treatment villages received gardening and small animal raising interventions and health and nutritional counselling through HKI’s BCC strategy. 8 All households in these villages that had children 3–12 months of age at baseline (2010) were invited to participate in the programme and associated study. The baseline survey was conducted in 2010 and the endline in 2012. Map of study site in Gourma Province. Households for the qualitative research were randomly selected from each of the 30 treatment villages, and from 15 (of the 25) control villages that participated in the baseline survey. For the first round of qualitative research (2011), semi-structured interviews (SSIs) were conducted with five randomly selected households in each village included in the qualitative research. Two of the five households in each village were selected to complete a longer SSI that collected more in-depth information (Table 1). The same households participated in the first and second rounds of qualitative research to the extent possible. If a household from the first round of was not available to participate in the second round, a replacement household was randomly selected from the list of households that participated in the baseline survey. Overview of methods and participants from intervention villages and control villages Source: Compiled by authors. Note: aOne village dropped out of the programme and study before the first round of qualitative research, resulting in a total of 14 villages for the first and second rounds of qualitative research and for the endline survey for the impact evaluation. The reason for the village to drop out was a lack of social cohesion due to conflict. Gender-specific questions Source: Compiled by the authors. The units of analysis for the impact evaluation were the household as well as individuals within the household. The household head was asked to answer questions about the different household members, their health, education and dwellings. Both male and female respondents were then interviewed separately about issues including asset and animal ownership (Dillon, Moreira, Olney, Pedehombga, & Quinones, 2012). For the qualitative research, SSIs were carried out with households and key informants in treatment and control villages. In the first round, SSIs with beneficiary women covered issues related to implementation and utilisation of the programme components, perceptions of the quality of these components and barriers and enablers to inform their optimal implementation and utilisation. SSIs were also carried out with control households to establish a counterfactual. The second round of qualitative research (2012) used SSIs to understand men’s and women’s views about acquisition, use and ownership of assets and agricultural decision-making in treatment and control villages. Women were interviewed in each household, and their husbands were interviewed for selected modules. The baseline survey was conducted between February and April 2010 and the endline between February and May 2012 with the same households. Impacts were estimated for specific outcomes comparing results from the treatment villages to those from control villages, using a difference in differences specification controlling for baseline characteristics. The pooled specification was estimated with the following regression: where ΔYEndline − YBaseline is the change in programme indicator variable between the endline and baseline survey, which could be either a household-level or individual-specific indicator. Treated indicates whether the household or individual received the E-HFP programme or not (1 = treated, 0 = not treated). The specification also included baseline characteristics of the household or child depending on the programme indicator variable chosen. Balancing tests of several baseline characteristics are presented in Table A1 in the Online Appendix, and demonstrate that household characteristics such as household size and education of the head of household, as well as men’s and women’s asset values were balanced at baseline. No characteristics can be rejected at the 5 per cent level of statistical significance. The specifications still include baseline characteristics to increase the precision of estimates. The regressions were also estimated with corrections for clustering at the village level, the unit at which intervention was assigned, and for attrition using inverse probability weights (Wooldridge, 2002); see Table A2 in the Online Appendix for the attrition specification used to correct for observable attrition bias in our estimates. The first round of qualitative data collection was carried out in May and June of 2011 and the second round in May and June of 2012. Qualitative data were manually coded by grouping similar responses together and looking for common themes among the respondents. Both the impact evaluation and the qualitative research addressed questions that aimed to understand the gender-related impacts of the programme, presented in Table 2.