Social registries, defined as information systems that support intake, registration, and eligibility assessment processes for social protection programs, are being developed and expanded in sub-Saharan Africa. In environments where data is scarce and informality is high, collecting and centralizing information on households or individuals offer significant potential for programming purposes. Social registries facilitate the adoption of common processes for program targeting and delivery, which can bring countries on a potentially more efficient path toward poverty reduction. For instance, during the COVID-19 pandemic, countries that had an existing social registry responded faster and reached more people than others.
In Liberia, where household-level surveys are infrequent, establishing a social registry through door-to-door data collection offers a unique opportunity that goes beyond administration of social protection programs. It could serve as a platform for interested government agencies, donors or development partner to inform service design, planning and delivery, in various sectors. However, using this door-to-door data collection approach is no small feat, as it requires enumerating wide geographical areas with often difficult terrains, potentially as dense as Lagos or as sparsely populated as the Sahel.
An illustrative case is the development of the Liberia Household Social Registry (LHSR), as part of the Liberia Social Safety Net Project (LSSNP), which ran from 2016 to 2023. The Ministry of Gender, Children, and Social Protection led the project with the support from the World Bank. More than 250,000 households, which is about 20% of the population, are now enrolled into the LHSR, but while this meets the corresponding LSSNP target, substantial hurdles had to be overcome.
Challenging terrains in Liberia often complicate and increase implementation costs, including data collection efforts for building the Liberia Household Social Registry (LHSR). Photo Credit: Liberia Ministry of Gender, Children and Social Protection (MoGCSP).
Initial delays were caused by a political transition and challenges in securing a suitable data collection firm. The onset of COVID-19 further halted the project just as data collection was set to begin. Fieldwork eventually commenced three years after the project’s approval, concluding in April 2021 with data from 200,000 households. However, this data had significant quality issues, raising important concerns about its reliability and utility.
To correct the errors, the World Bank and the LSSNP Project Implementing Unit decided to revisit the households that were incorrectly captured. Yet, further analysis revealed misclassifications related to building identifiers, complicating efforts to isolate the concerned buildings for the revisits.
Lines connect households that live under the same roof, according to the data. Credit: Liberia Ministry of Gender, Children and Social Protection (MoGCSP).
To address this issue, the team developed an ad-hoc deep-learning model to identify misclassified households. The model, named “Automated Correction of Misclassified Household Structure Data for Liberia,” identified errors by clustering nearby households that also had a high probability of being in the same building based on their GPS coordinates and photos.
A wrong assignment caught by the model, same structure numbers but different buildings. Credit: Credit: Liberia Ministry of Gender, Children and Social Protection (MoGCSP).
Although the cleaning processes took several months, they proved successful in identifying and revisiting the most problematic structures, ultimately resulting in a significantly cleaner and more usable social registry. The following insights gained from this and other experiences within the project can be valuable for those interested in establishing a national social registry.
First, prioritize quality and standards: look for clues and take nothing for granted. Most challenges faced during the project pertained data quality and productivity, often common in data collection. However, the amount and intended use of the data being collected demanded establishing detailed protocols, requiring weekly (or more frequent) discussions, which involved unpacking every piece of information to find defects and improve quality.
Second, adopt the project as a platform for data-driven or operational innovations. The LSSNP became a vehicle and an opportunity for introducing innovative advancements in the field of social protection in Liberia. These included a machine learnings application as earlier described, the use of building footprints from a public source (OpenBuildings) for planning field activities, and the introduction of scannable LHSR-ID Cards for identifying households and buildings in an ID-scarce environment.
Third, seeking opportunities for early uptake is key to ensure sustainability. In January 2020, the Liberian Cabinet endorsed the LHSR as the primary tool for targeting social protection programs in the country. In September 2022, the project team demonstrated the LHSR's capabilities at a National Social Protection Steering Committee meeting. The LSSNP also developed a Management Information System (MIS), which supports beneficiary recruitment and enables electronic payments, among other functionalities. The LHSR and the MIS are being used by three government-led social protection programs, with more expected as systems strengthening continues under the REALISE Project.
Fourth, foster a team with foresight, and focus on effective and transparent communication. Throughout the project, challenges encountered ranged from minor hiccups to major roadblocks. Strategizing and communicating, often informally, were opportunities to think constructively. Further, they served as a space for strengthening the team’s ever-evolving purpose, ultimately helping the government achieve its worthy goal: establishing an essential building block for social protection programming in Liberia.
Join the Conversation