Qualitative interview resulting from semi-structured household interviews and focus group discussions that aimed to assess the impact of development activities that are intended to benefit poor men, women and children; and how their income and food security is changing. The study took place in four rural village sites: Masumbankhunda and Karonga areas in Malawi and Tigray and Oromia areas in Ethiopia.How can the impact of development activities intended to benefit poor men, women and children caught up in complex processes of rural transformation best be assessed? The research set out to develop and evaluate a protocol for impact assessment based on self-reported attribution without the use of comparison groups as an alternative to experimental or quasi-experimental designs based on statistically inferred attribution. The three year project, starting in September 2012, was led by James Copestake at the University of Bath, and conducted in collaboration with three NGOs - Self Help Africa, Farm Africa, and Evidence for Development. It was jointly funded by the UK Economic and Social Research Council (ESRC) and Department for International Development DFID. The research has three strands that applied to four projects: two in Ethiopia and two in Malawi Strand 1 comprises a baseline and two rounds of annual monitoring of food security and income at the household level by NGO staff. Strand 2 comprises two rounds of annual in-depth interviewing to elicit self-reported attribution from intended project beneficiaries. Strand 3 comprises two rounds of qualitative evaluation of what Strand 2 added to the understanding of project stakeholders.

The full data collection methodology is outlined in the attached documentation files. The approach used is Qualitative Impact Assessment Protocol (QUIP). This approach aims to collect information on changes in people's lives over the same period of time as the development intervention that is being assessed by the implementing NGO. Semi structured questionnaires were used with selected beneficiary households that were sampled using a stratified randomised sub-sample of the households in the quantitative monitoring survey, using a combination of open-ended and closed questions. Field researchers were recruited and trained, but given no information about the development project being evaluated to avoid confirmation bias. This is called 'blinding' in the methodology. The dataset consists of Excel files that contain all anonymised interview transcripts for individual households and focus groups. This dataset is linked to a related set of quantitative monitoring data for the same projects.