Ending Extreme Poverty through Social Protection? Experimental Evidence from 10-year Nationwide Randomised Cash Transfers in Rwanda and Malawi
Lead Research Organisation:
UNIVERSITY OF OXFORD
Department Name: Economics
Abstract
650 million people, or 8.5% of the world's population, still live below the international poverty line of $2.15 per day. Despite concerted efforts from the international community, the UN's first Sustainable Development Goal (SDG) of ending extreme poverty by 2030 therefore remains highly ambitious. Which interventions and policies can make this a reality at the scale and speed required?
A robust body of research has shown that unconditional cash transfers (UCTs) - cash delivered to people living in poverty, no strings attached - are effective tools for improving people's livelihoods (Bastagli et al. 2016). GiveWell (a trusted evaluator of philanthropic initiatives) ranks cash transfers among the most cost-effective development interventions, higher than any other non-health intervention. Based on these findings, the non-governmental organisation GiveDirectly (GD) - in collaboration with the governments of Rwanda and Malawi - is embarking on an ambitious programme aiming to 'accelerate the eradication of extreme poverty' in both countries. It will provide $550 to every person living below USD 2.15 per day - a total of over 10 million recipients. The programme amounts to USD 6 billion (a quarter of GDP) over 10 years.
Although many small-scale randomized controlled trials (RCTs) have demonstrated that poor households use cash transfers effectively to improve their livelihoods at the micro level, it is not clear what the aggregate implications of such programmes are. Beyond directly benefitting recipients, a nationwide programme may affect non-recipients through changes in market and social interactions as well as political processes. Will increased spending by recipients on locally produced goods and services generate higher incomes for non-recipients? Will an influx of cash lead to deteriorating social relationships or inequality? Are there political repercussions? Does the periodicity, mode of delivery, speed of roll-out, or targeting within the household matter? How does supply in different sectors and geographies adjust? Will there be inflation? Can cash alone provide a big push to accelerate the structural transformation historically associated with the development process (Rosenstein-Rodan (1943))?
The programme is uniquely ambitious in scale and presents an unprecedented opportunity to generate rigorous evidence on how cash transfers affect entire national economies. Too often, big ideas are not evaluated carefully and independently, leaving us little better off in terms of knowing what works (see e.g. the Millennium Villages). Methodologically, studying phenomena at this scale is challenging: The 2019 Nobel Prize in Economics awarded to Banerjee, Duflo, and Kremer credits randomized controlled trials (RCTs) as providing 'reliable answers about the best ways to fight global poverty'. Yet, critics have argued that RCTs are often too small-scale and context-specific to study 'big' questions (Bardhan (2005); Easterly (2006); Deaton (2010)). However, such concerns may be unfounded. Rigour and relevance need not necessarily be compromised when randomization at-scale is feasible. But, no RCT has been able to evaluate interventions at this scale.
To do this, a new approach bridging micro- and macroeconomics that combines randomisation across large geographical units with careful spatial modelling is required. The Oxford Centre for the Macro-Experimental Study of Cash Transfers will design and implement - in collaboration with GD and the Malawian and Rwandan governments - a two-country nationwide RCT to independently evaluate impacts of the programme - the largest cash transfer RCT to date. In the process, it will develop the methodological tools for a novel macro-experimental approach to economics, allowing researchers to rigorously study national development interventions, bringing cutting edge research to implementation at scale, and impacting the lives of millions.
A robust body of research has shown that unconditional cash transfers (UCTs) - cash delivered to people living in poverty, no strings attached - are effective tools for improving people's livelihoods (Bastagli et al. 2016). GiveWell (a trusted evaluator of philanthropic initiatives) ranks cash transfers among the most cost-effective development interventions, higher than any other non-health intervention. Based on these findings, the non-governmental organisation GiveDirectly (GD) - in collaboration with the governments of Rwanda and Malawi - is embarking on an ambitious programme aiming to 'accelerate the eradication of extreme poverty' in both countries. It will provide $550 to every person living below USD 2.15 per day - a total of over 10 million recipients. The programme amounts to USD 6 billion (a quarter of GDP) over 10 years.
Although many small-scale randomized controlled trials (RCTs) have demonstrated that poor households use cash transfers effectively to improve their livelihoods at the micro level, it is not clear what the aggregate implications of such programmes are. Beyond directly benefitting recipients, a nationwide programme may affect non-recipients through changes in market and social interactions as well as political processes. Will increased spending by recipients on locally produced goods and services generate higher incomes for non-recipients? Will an influx of cash lead to deteriorating social relationships or inequality? Are there political repercussions? Does the periodicity, mode of delivery, speed of roll-out, or targeting within the household matter? How does supply in different sectors and geographies adjust? Will there be inflation? Can cash alone provide a big push to accelerate the structural transformation historically associated with the development process (Rosenstein-Rodan (1943))?
The programme is uniquely ambitious in scale and presents an unprecedented opportunity to generate rigorous evidence on how cash transfers affect entire national economies. Too often, big ideas are not evaluated carefully and independently, leaving us little better off in terms of knowing what works (see e.g. the Millennium Villages). Methodologically, studying phenomena at this scale is challenging: The 2019 Nobel Prize in Economics awarded to Banerjee, Duflo, and Kremer credits randomized controlled trials (RCTs) as providing 'reliable answers about the best ways to fight global poverty'. Yet, critics have argued that RCTs are often too small-scale and context-specific to study 'big' questions (Bardhan (2005); Easterly (2006); Deaton (2010)). However, such concerns may be unfounded. Rigour and relevance need not necessarily be compromised when randomization at-scale is feasible. But, no RCT has been able to evaluate interventions at this scale.
To do this, a new approach bridging micro- and macroeconomics that combines randomisation across large geographical units with careful spatial modelling is required. The Oxford Centre for the Macro-Experimental Study of Cash Transfers will design and implement - in collaboration with GD and the Malawian and Rwandan governments - a two-country nationwide RCT to independently evaluate impacts of the programme - the largest cash transfer RCT to date. In the process, it will develop the methodological tools for a novel macro-experimental approach to economics, allowing researchers to rigorously study national development interventions, bringing cutting edge research to implementation at scale, and impacting the lives of millions.
Organisations
- UNIVERSITY OF OXFORD (Lead Research Organisation)
- Government of Malawi (Collaboration)
- University of California, San Diego (UCSD) (Collaboration)
- Government of Rwanda (Collaboration)
- Centre for Economic Policy Research (Collaboration)
- University of California, Berkeley (Collaboration)
- GiveDirectly (Project Partner)
| Description | The UKRI Future Leaders Fellowship has enabled the setting up of the Oxford Centre for Macro-Experimental Development. This is a research group stuying how development interventions can be scaled up to achieve maximum impact, and how research designs employing experimentation at-scale together with structural modeling can be brought together to improve such a scale-up. As a first case-study, we evaluate the planned nationwide rollout of lump sum unconditional cash transfers (delivered through mobile money) amounting to approximately one year of income to all adults within two entire countries by the NGO GiveDirectly (GD). To date, we have set up the research center and research group, consisting of a post-doctoral fellow, several doctoral students (currently 2-3), two pre-doctoral fellows, and a research manager. We have also begun setting up the data and research infrastructure in one of the countries -- Malawi -- as GD has started enrolling over 100,000 recipients in the district in a multi-level clustered randomized controlled trial that will study the impact of transfers at high-frequency on households, local markets and firms, prices, health, and political processes. We have completed baseline data collection and cash transfer enrolment is currently underway. A key innovation of the program is methodological: We hope to build a research platform that is as collaborative and open-access as possible, allowing other researchers (particularly from low- and middle-income countries and local universities) to contribute to the research, propose survey questions and modules, access the data, etc. to maximize joint learning. Moreover, together with the STEG group at CEPR, there is also an effort underway to use this experiment as litmus test of economic models by pre-registering their predictions ex-ante and testing them against the evidence. |
| Exploitation Route | At this stage, no findings/outcomes are available yet, but they will be soon. The findings will impact social protection policy in Malawi and Rwanda (the two focus countries of the initiative) and beyond. Unconditional cash transfers -- by virtue of their simplicity, scalability, and increasingly documented positive impacts -- are the most prevalent and fastest growing social protection policy. This initiative will help understand its impact not only on recipients, but local markets and national economies at large, and help improve their targeting and design. Methodologically, the Oxford Macro-Experimental Development group will help understand how markets in developing countries work, and how to combine experimentation at-scale in combination with macroeconomic modeling to better predict and understand the impact of development interventions at-scale. |
| Sectors | Communities and Social Services/Policy Education Healthcare Government Democracy and Justice Other |
| Description | While early in the process, the project has been received with great attention both within academia and by policymakers. Within academia, the idea that macroeconomic models should be brought into the process ex-ante, and that model predictions, measures of success, and tests for these models should be pre-specified (akin to how primary outcomes and statistical tests are increasingly pre-specified in reduced-form RCT analyses in the social sciences) has begun to be discussed more widely, including at two dedicated workshops organized by the initiative. Among policy makers, while we do not have any findings yet (the program we study has not yet been implemented), the need for evidence of such policies at-scale to inform future nationwide rollouts and reforms of social protection system has garnered support, and an understanding of the research agenda and design has been developed. |
| Sector | Communities and Social Services/Policy,Other |
| Impact Types | Societal Economic Policy & public services |
| Description | GiveDirectly Operations |
| Geographic Reach | National |
| Policy Influence Type | Implementation circular/rapid advice/letter to e.g. Ministry of Health |
| Impact | see above. |
| Description | Rwanda Social Protection |
| Geographic Reach | National |
| Policy Influence Type | Participation in a guidance/advisory committee |
| Impact | Through this collaboration we've established a strong need for evidence-based policy-making, and have built capacity for understanding rigorous evidence and how this can inform policymaking at various levels. We have also shifted policymakers beliefs in favor of trialing cash-based graduation (which has been shown to be highly effective in other contexts, but has not yet been adopted in Rwanda to date). |
| Description | GD Phase I Contribution |
| Amount | $350,000 (USD) |
| Organisation | GiveDirectly |
| Sector | Charity/Non Profit |
| Country | United States |
| Start | 01/2025 |
| End | 06/2026 |
| Title | Pre-specification in Macro/Structural Modeling |
| Description | This is (to our knowledge) the first large-scale randomized controlled trial that brings in macro- and structural economists (in collaboration with STEG) EX ANTE to write models, influence research design, propose and suggest changes to questionnaire design into the research process. Typically, macroeconomic models are disciplined ex-post, using out-of-sample moments to target, or other goodness-of-fit based tests. For the first time, this collaboration will pre-specify the models developed and their predictions based on baseline data only. They will pre-specify a) a target / success metric of their model, and b) a test for validating the model against that target BEFORE the experiment has taken place. This has been widely adopted in the social sciences for reduced-form outcomes in randomized controlled trials, but has not yet been adopted by more structural economists or by macroeconomists. |
| Type Of Material | Improvements to research infrastructure |
| Year Produced | 2026 |
| Provided To Others? | Yes |
| Impact | We hope that through this initiative, we will develop a minimum protocol for pre-specifying structural work and model validation. After the end of the collaboration, we hope to write a summary article on what we have learned (in addition to the individual papers), and propose this minimum protocol for the profession to adopt. We hope this will increase transparency and credibility of model validation using experimental data in the future. |
| Title | GiveDirectly STEP RCT - Chiradzulu Baseline Data |
| Description | We have shared (anonymized) baseline data from a) household census, b) household survey, c) enterprise census, d) enterprise survey, e) market price survey, with all researchers involved in the STEG initiative. These are fully representative datasets on the economy of Chiradzulu district, the first district within the GD STEP initiative in Malawi. |
| Type Of Material | Database/Collection of data |
| Year Produced | 2026 |
| Provided To Others? | Yes |
| Impact | The data will be used by STEG collaboration researchers to discipline/calibrate/estimate their models on baseline data, and to derive predictions about endline impacts of the program. |
| Description | Bringing Pre-specification to Macro-Economics |
| Organisation | Centre for Economic Policy Research |
| Country | United Kingdom |
| Sector | Charity/Non Profit |
| PI Contribution | Together with the Structural Transformation and Economig Growth Group (STEG) at CEPR and the NGO GiveDirectly, the Centre for Macro-Experimental Development at Oxford set up a joint initiative to study the general equilibrium effects of cash transfers. The aim of the collaboration is for macro-development researchers to bring state-of-the art structural economic modelling to bear on the question of what effects -- particularly macro-economic effects that go beyond what has been well studied at smaller scale -- we should expect from the nationwide cash transfer program rolled out by GiveDirectly (and governments/organisations elsewhere). Moreover, the initiative aims for participants and modellers to commit ex-ante to writing down and calibrating a model on baseline data only, and to a) set out ex-ante what 'success' would mean for the model (e.g. prediction accuracy), and b) pre-specify a statistical test of this measure of success. The timeline of the collaboration is as follows: Feb 2025: Call for proposals issued (36 submissions received, 18 shortlisted for full proposals and reviewed by two external reviewers, 6 funded) Jul 2025: Workshop in Malawi, field visits, presentation of proposals, etc. Jan 2026: Baseline data shared with participants May 2026: Carnegie/Rochester/NYU conference. Final pre-analysis papers presented and published in a special issue of the Journal of Monetary Economics Jan 2028: Endline data shared with researchers 2028/9: Final papers/tests completed, taking stock conference |
| Collaborator Contribution | STEG issued the call for proposals and organized its evaluation independently of Oxford CMED. They contributed 50% of funds for proposals and workshops, totalling GBP 100,000, and helped with workshop organization in Malawi. They also organized the Carnegie/Rochester/NYU conference, and took over editorship of the special issue of the JME where structural pre-analysis papers will be published. |
| Impact | - Call for propoals (Jan 2025) - 5-day conference workshop in Malawi with field visits, etc. (Jul 2026) - Cleaned and documented final baseline datasets shared for households, enterprises, markets, prices, etc. (Jan-Feb 2026) - 3-day conference in Rochester (May 2026) - Special issue of the Journal of Monetary Economics with pre-analysis papers (tbd, 2026) - Taking-stock conference (tbd) |
| Start Year | 2024 |
| Description | Malawi Data Hub |
| Organisation | Government of Malawi |
| Country | Malawi |
| Sector | Public |
| PI Contribution | The 'Malawi Date Hub' is a collaboration between the University of Malawi, Department of Economics (represented by Dr. Bertha Bangara Chikadza, Senior Lecturer in Macroeconomics, and Dr. Gowokani Chijerwe Chirwa, Head of the Department) the National Statistics Office (represented by Hector Kankuwe, Director of Agriculture and Economics Statistics), the Oxford Centre for Macro-Experimental Development (OCMD, represented by Dennis Egger and David Henning). The hub will promote evidence-based policymaking, support academic research and enhance coordination and collaboration across the National Statistical System (NSS). The Malawi Data Hub supports a) international collaborations between researchers, and b) helps boost institutional capacity building in Malawi. It provides research assistance, travel, and capacity building to clean and prepare as-yet unused administrative data on the Malawian economy at the national level (including CPI microdata, VAT data, firm registries, tax returns, economic censuses and surveys, etc.). Furthermore, to make this data publicly available, we will develop the necessary technology for matching individuals without a common identifier based on Personally Identifiable Information (PII), and the legal guardrails to facilitate this data access and protect the privacy of those whose data are stored at NSO. With regards to international collaboration, the hub supports: a) research assistance opportunities for Malawian students or early-career researchers hosted at the University of Malawi, b) seed funding for early career researchers and PhD students to conduct research in Malawi, including small research grants for up graduate students and post-docs (5 for Malawians, 5 for non-Malawian) working on macro-development topics in Malawi. Especially for the Malawian students, we will provide information sessions to program participants on how to apply for additional funding using existing links with the Centre for the Study of African Economies at Oxford. We will also enable match-making between Malawian and international researchers and joint projects will be prioritized.. Finally, the hub will provide a capacity-building program to assist the University of Malawi and the National Statistics Office of Malawi in generating insights from its growing data collection, with a focus on a topic of their choosing. We propose satellite and remote sensing data, but the exact topic will be decided on in collaboration with NSO and the University of Malawi. This will help form stronger links between the NSO and academia. Lastly, the Malawi Data Hub will support the ongoing collaboration between STEG, OCMD, and GiveDirectly to study the impacts of cash transfers at scale by combining the nationwide randomised rollout of cash transfers with the newly accessible administrative data. |
| Collaborator Contribution | The University of Malawi and the NSO are hosting the data hub and providing research assistant support as well as maintaining connections and relationships, and co-organize capacity building workshops. See details above. |
| Impact | - none as of yet, though the University of Malawi and the National Office of Statistics participated in the joint workshop in July. |
| Start Year | 2025 |
| Description | Malawi Social Protection Policy |
| Organisation | Government of Malawi |
| Country | Malawi |
| Sector | Public |
| PI Contribution | We have been in continued engagement with the Government of Malawi, both locally in the first treatment district (Chiradzulu) as well as at national scale, primarily through the Ministry of Gender which houses all social protection policy. Engagement included a policy-focused workshop in Malawi in July, where cash transfers and their impact/relationship with existing social protection policy and budgetary requirements in Malawi were discussed between policy-makers and international and local macroeconomists. We also maintain connections to the Chiradzulu District administration with respect to data collection and capacity building, and are beginning to launch a collaboration to strengthen public health monitoring systems which will also allow us to study in more detail the health and (child) mortality impacts of the GiveDirectly programme. We have also contributed to a report by the Oxford Policy Engagement Network and FCDO for a policy-focused report on the state and direction and political background of Malawi's current social protection system which is due to be published shortly. |
| Collaborator Contribution | Government of Malawi representatives attend joint workshops and occasional meetings, and help review proposals. They also are instrumental in building connections and supporting capacity building at the local level within the local legal framework. |
| Impact | Very multidisciplinary, involving political scientists, public administration, public health and education systems, as well as political representatives. |
| Start Year | 2025 |
| Description | Phone- and other admin data for cash transfer targeting and evaluation |
| Organisation | University of California, Berkeley |
| Country | United States |
| Sector | Academic/University |
| PI Contribution | The Oxford Centre for Macro-Experimental Development has established a collaboration with Joshua Blumenstock at UC Berkeley and Emily Aiken at UC San Diego for research in both Malawi and Rwanda on how to use mobile-phone administrative data (call logs, message logs, mobile money transactions, etc.) for both a) targeting, and b) impact evaluation of ongoing cash transfer rollout. From our side, we will share anonymized outcome data, linked to phone metadata to analyse accuracy of phone-based targeting relative to survey-based data, and to evaluate impacts of cash transfers in survey vs. phone based data. This will be crucial for future more cost-effective targeting protocols in data-poor environments, as well as for streamlining the evaluation process using cost-saving admininstrative data. |
| Collaborator Contribution | Joshua Blumenstock and Emily Aiken are leading computer scientists working on using algorithmic approaches to target social protection and anti-poverty programs in developing countries with poor data environments. They have established data sharing agreements with telecom providers in both countries, and will work on making such data usable and developing the algorithms used to target poverty. |
| Impact | yes, with computer scientists |
| Start Year | 2025 |
| Description | Phone- and other admin data for cash transfer targeting and evaluation |
| Organisation | University of California, San Diego (UCSD) |
| Country | United States |
| Sector | Academic/University |
| PI Contribution | The Oxford Centre for Macro-Experimental Development has established a collaboration with Joshua Blumenstock at UC Berkeley and Emily Aiken at UC San Diego for research in both Malawi and Rwanda on how to use mobile-phone administrative data (call logs, message logs, mobile money transactions, etc.) for both a) targeting, and b) impact evaluation of ongoing cash transfer rollout. From our side, we will share anonymized outcome data, linked to phone metadata to analyse accuracy of phone-based targeting relative to survey-based data, and to evaluate impacts of cash transfers in survey vs. phone based data. This will be crucial for future more cost-effective targeting protocols in data-poor environments, as well as for streamlining the evaluation process using cost-saving admininstrative data. |
| Collaborator Contribution | Joshua Blumenstock and Emily Aiken are leading computer scientists working on using algorithmic approaches to target social protection and anti-poverty programs in developing countries with poor data environments. They have established data sharing agreements with telecom providers in both countries, and will work on making such data usable and developing the algorithms used to target poverty. |
| Impact | yes, with computer scientists |
| Start Year | 2025 |
| Description | Rwanda Cash Transfers Path to Scale |
| Organisation | Government of Rwanda |
| Country | Rwanda |
| Sector | Public |
| PI Contribution | Over the last years, I have continuously engaged with various stakeholders within the social protection space in Rwanda, including MINALOC (responsible for the social protection targeting and the social registry), LODA (responsible for implementing public works and cash transfer programs across local authorities), the Ministry of Finance, the NGO BRAC (a core implementation and research partner of the government's poverty graduation strategy), the NGO GiveDirectly (or primary research implementation partner), and various other stakeholders. The aim of these engagements has been to define a concrete learning agenda, and a research design and rollout to address this over the next decade or so. Conversations have progressed to a point where learning agenda is mostly agreed, and a phased approach to the rollout of cash, and its comparison with existing social protection mechanisms with accompanying research has been agreed, fousing a) first, on the role of social-registry vs. community based targeting in accuracy and support of the program, b) the question of how to target, within locations, different program components to different people (what works best for whom?), and c) the evaluation of the scale-up of both graduation and cash-transfer based social protection programs and their impacts on local economies. |
| Collaborator Contribution | Partners have engaged with us through multiple meetings and workshops, and have provided access to existing research and data from the social registry as well as pilot phases of the newly launched national poverty graduation program. |
| Impact | Very multidiscipilinary, including public policy, agronomy, computer scientists (for targeting, etc.), and policymakers |
| Start Year | 2025 |