Phase I of the Capital Journey study examined whether BRAC’s program records can be linked into a longitudinal account of how women access and use capital. Drawing on key informant interviews, focus group discussions, and an assessment of roughly 145,000 datapoints across five branches, it found that inconsistent identifiers and late digitisation make cross-program linkage largely infeasible, with match rates below 4%. The study recommends a universal beneficiary identifier, a data-sharing protocol, and branch-level digitization as preconditions for longitudinal analysis.
Researchers: BRAC Central Data Unit and LightCastle Partners
Partners: BRAC; Gates Foundation
Timeline: 2024–2025
Status: Completed
Contact: Raisa Adiba; raisa.adiba@bracu.ac.bd
Context
Over more than two decades, BRAC has accumulated an extraordinary volume of beneficiary records across microfinance, ultra-poor graduation, health, education, and urban development. In principle, these records describe the same women moving between services over time, a ready-made longitudinal account of how poor households acquire, use, and lose access to capital. In practice, each program maintains its own identifiers, storage conventions, and digitization timelines, and much of the older record base remains on paper at branch level. Phase I of the Capital Journey study was commissioned to establish how much of that latent longitudinal value can actually be recovered from the data as it currently exists, and what would be required to unlock the rest.
Objectives
The study set out to determine the feasibility of using BRAC’s longitudinal programme data to reconstruct women’s capital journeys across services. It further sought to assess the quality, completeness, and accessibility of the records held at branch level, to identify the ecosystem factors that enable or obstruct women’s access to capital along those journeys, and to prescribe the data structures and governance arrangements BRAC would need to put in place for such analysis to become possible. The findings were also intended to inform the design of a subsequent phase of work.
Methodology
The study used a mixed-methods design across five branches selected for geographic and climatic contrast: Chowdhury Bazar (Habiganj), Seroil Colony (Rajshahi), Chilmari (Kurigram), Mongla (Bagerhat), and Gulshan (Dhaka). Fieldwork comprised 35 key informant interviews with program managers, branch managers, and field officers, five focus group discussions with microfinance beneficiaries, and on-site inspection of branch data repositories. On the quantitative side, roughly 145,000 datapoints across four programmes were assessed for completeness, consistency, and accessibility. Record-linkage exercises were then run between microfinance records and those of the ultra-poor graduation, health, education, urban development, and integrated development programmes, using national ID and contact number as candidate keys. Qualitative and quantitative evidence were read together to explain why observed linkage rates were so low.
Findings and Recommendations
Cross-programme linkage proved largely infeasible with current data. Fewer than 2% of ultra-poor graduation beneficiaries could be matched to microfinance records using national ID, and match rates across all programme pairs remained under 4%. Restricting to female beneficiaries with a recorded national ID raised the rate only to 3.31%, or 658 individuals. Microfinance, ultra-poor graduation, and urban development records were judged usable; health, education, and integrated development data only partially so; and no digitised climate change programme data was available. Digitisation at most branches began only in 2023, leaving earlier years on paper.
The study recommends five preconditions for longitudinal analysis: a data-sharing protocol aligned with GDPR, ISO, and FAIR principles; a universal beneficiary identifier with national ID cross-verification; a digital data topsheet tied to branch manager performance targets; a shift from vendor-led delivery to co-creation with BRAC’s central data teams; and revised officer targets to unblock progression from Dabi to Progoti. These findings shaped Phase II.
