Better Futures GuideDonate

Raising The Village: technical report

This report describes our evaluation of Raising The Village in detail. A shorter version is available on our Raising The Village page.

Published: November 2026

Summary

  • Our recommendation. We recommend Raising The Village as an effective giving opportunity. It runs a two-year poverty graduation program for whole villages in Uganda, Rwanda, Tanzania and the Democratic Republic of the Congo.
  • Our headline result. We estimate that Raising The Village (RTV) doubles one person’s consumption for one year for $27. This is about 6x as cost-effective as cash transfers provided by GiveDirectly.
  • Raising The Village’s program is a form of poverty graduation, an approach that has been tested in 40 randomized trials, over nearly two decades. Poverty graduation programs provide a bundle of resources and support for a period of roughly one or two years, on the premise that households are most likely to exit extreme poverty in a durable way when several of their constraints are addressed at once. In a recent meta-analysis of 40 randomized trials in 26 countries, graduation programs raised consumption by about 11%, income by 15% and assets by 40% on average.
  • Gains from poverty graduation programs tend to last for years after the program ends. Across the trials in the meta-analysis, consumption gains were on average as large or larger at later measurements, up to four years after the program ended. Most trials that followed households for longer found gains still present seven or more years later, although sometimes at a lower level. In some trials, gains narrowed because comparison households’ living standards rose.
  • However, programs differ, and so do their impact and cost-effectiveness. Some programs cost much less than others, and results can vary widely from one graduation program to the next.
  • Raising The Village is an unusually low-cost version of poverty graduation. We estimate that it will cost about $161 per household in Uganda in 2027–28, as compared with a median of about $520 among the graduation programs started since 2016 in the meta-analysis cited above.
  • Yet, a randomized trial found Raising The Village’s effects on par with higher-cost programs. In a trial covering 335 villages, approximately three years after the program began, households in program villages had:
    • 10% higher consumption just after harvest, consumption here meaning the value of what a household eats and uses, including food it grows
    • 39% higher consumption in the lean season, the months leading up to harvest (measured nine months later). In that season, 4% of households in program villages had a child go to bed hungry in the past month, against 25% in comparison villages
    • 26% higher income and 31% more wealth in livestock, savings and other assets
    • Wider benefits including improved health, with adults losing about one day less of work a month to illness, and children 6 percentage points less likely to miss school through illness; less depression among adults; greater trust within the village
  • Room for more funding. Raising The Village reports a funding gap of between $8 million and tens of millions of dollars a year in the growth scenarios it has shared with us, and we think it can use additional donations at the cost-effectiveness we have modeled.
  • In sum, we recommend Raising The Village. During our evaluation, we found it to be a well-run, transparent organization. The intervention it delivers, poverty graduation, is among the best-evidenced ways to reduce extreme poverty. RTV delivers it at unusually low cost, with effects on par with higher-cost programs.

Introduction

This report sets out our evaluation of Raising The Village, conducted in July to November 2026. We first define poverty graduation and review the evidence base for it, including how it compares with cash transfers. We then describe Raising The Village’s model and the randomized trial of its program, including why we find its notable results plausible. We then present the findings of our cost-effectiveness analysis. There, we discuss how we model the persistence of effects using weighted scenarios drawn from long-run trials, our adjustments for changes to the program since the trial and for government co-funding, a comparison with GiveDirectly’s cash transfers, and a Monte Carlo uncertainty analysis with 10,000 draws. We close with Raising The Village’s room for more funding and its organizational capacity.

Many people and organizations gave generously of their time throughout our work on this evaluation, offering expertise, input on our process, feedback on our analyses, and comments on drafts of this report. We are grateful to Samantha Carter, Joshua Deutschmann, Kelsey Jack, Mahreen Mahmud, Heidi McAnnally-Linz, Andrés Parrado, Emma Riley, Patrick Stadler, as well as staff at Ambitious Impact (Morgan Fairless, Filip Murár, Steve Thompson), the BRAC Institute of Governance and Development, the Center for Global Development, Dovetail Impact Foundation, End Ultrapoverty: A Global Collaborative, Founders Pledge, GiveWell, IDinsight, Innovations for Poverty Action, the International Initiative for Impact Evaluation (3ie), J-PAL, The Life You Can Save, the Livelihood Impact Fund, Northwestern University’s Global Poverty Research Lab, and Prevail Fund. All views and any errors are our own. We thank Raising The Village for its openness throughout this evaluation: its team answered detailed questions, shared internal cost and monitoring data, and reviewed a draft of this report for factual accuracy. The conclusions are ours alone.

We also thank Aqsa Zaidi, who contributed throughout the research and evaluation process; Arshia Haque, who contributed research on poverty graduation and our methodology; Matt Lee, who built early cost-effectiveness estimates of nominated organizations; and Sylia Hanana, who led our initial mapping of poverty graduation organizations.

Dollar figures are in January 2026 US dollars unless stated otherwise. Figures marked PPP are adjusted for purchasing power parity, and give their base year. This report reflects information available to us as of November 16, 2026. We used AI tools to help with literature searches, checking claims against sources, model checks, drafting and editing. Our team reviewed all outputs; the research design, judgments and conclusions are our own. Better Futures Guide has received funding from the Livelihood Impact Fund (LIF), which is Raising The Village’s largest funder. We conduct our work independently from LIF, although its staff has provided us with valuable input and feedback.

The approach: Poverty Graduation

More than 800 million people live in extreme poverty, on less than $3 a day.1 Poverty graduation is a type of program built on the premise that individuals and families remain in extreme poverty because they face several constraints at once. For instance, a family may have no money to invest, little access to training in farming or business skills, and no savings to absorb a setback, such as a failed harvest or an illness.

Graduation programs therefore address several constraints at the same time. For a period of typically one to two years, they provide a combination of resources and support: for example, livestock or a cash grant to start or expand a small business or farm, training in how to run it more effectively, regular coaching visits to help apply that training and overcome obstacles, and access to a savings group or account. An aim is a lasting exit from extreme poverty, in which households sustain higher living standards after the program ends. The approach was pioneered in Bangladesh by BRAC in 2002 and has been adapted and tested continuously since.2

There is no single definition of poverty graduation. The common core is the bundle of support, usually delivered over one to two years, that provides households living in extreme poverty both the means and the skills to build a lasting livelihood. A 2026 meta-analysis by Carter, Jee, Karlan and Premand, covering 40 randomized trials in 26 countries, identifies five core components. It considers a program to be poverty graduation if it includes at least three of them:3

ComponentDescriptionPurpose
Asset or cash transferA substantial transfer of productive goods, equipment or livestock, or a lump sum of cashProvides the means to start or expand an income-earning activity
TrainingStructured instruction with a set curriculum, for example in farming, livestock or business skillsBuilds the skills needed to produce and earn more from that activity
Consumption supportSmaller transfers of cash or food to meet short-term consumption needsCovers basic needs while the activity becomes established, reducing the pressure to sell or use up what the household has received
CoachingRecurring contact with a coach or mentor, through one-on-one or group meetingsHelps participants apply the training, plan, stay motivated and address problems as they arise
SavingsHelp to save regularly, often through a savings group, access to an account, or savings targetsBuilds a reserve for emergencies and, in savings groups, gives members access to credit

Components and definitions: Carter et al. (2026), Appendix Table A.1, with our examples. Purposes: our summary.

In the meta-analysis, programs with all five components tended to have larger effects on economic outcomes, as did programs with larger transfers. Larger effects did not mean greater cost-effectiveness, however: lower-cost programs tended to be more cost-effective. Additionally, the data did not single out any component as either essential or dispensable; for example, a large transfer was not a necessary requirement for large effects.4

Evidence from randomized trials

Graduation is among the most well-evidenced approaches to reducing extreme poverty. The 40 trials in Carter et al. span 26 countries, about a quarter of them in fragile or post-conflict settings. On average, programs raised household consumption by about 11%, income by 15% and assets by 40%. Savings, food security, mental health and women’s say in household decisions improved as well.5

The table below presents trials that speak to the persistence of gains: those that surveyed households more than once, or several years after the program began.

Persistence of consumption gains in trials with longer-term follow-up6

Program and countryWhen effects measured (time since program start)Effect on consumption, relative to comparison householdsChange in effect over time
BRAC, Bangladesh2, 4 and 7 years; 17 years (separate study)+5%, then +10%, similar at 7 years; +14% at 17 yearsGrew, then held. Households below an asset threshold fell back
Bandhan, India18 months, 3, 7 and 10 years+16%, +11%, +29%, +21%Grew to year 7; narrowed to year 10 as comparison households caught up. Still significant at year 10.
REST, Ethiopia2, 3 and 7 years+16%, +18%, then +9%Held for three years, then halved by year 7 as living standards also rose among comparison households7
Six-country trial (Ethiopia, Ghana, Honduras, India, Pakistan, Peru; includes the India and Ethiopia trials above)2 and 3 yearsPositive in four countries at both roundsStable. In Honduras, gains fell back after many of the chickens given to participants died of disease
Afghanistan2 and 5 years+30%, then about +16%Narrowed during drought, conflict and COVID-19 but remained positive
YemenAbout 4 yearsNo significant effect on consumptionDelivery was disrupted by conflict: coaching visits were cancelled and consumption support was suspended for over a year
Zambia (government-run)About 2.5 and 4.5 years+0.24, then +0.44 SDGrew between the two surveys
Niger (government-run)6 and 18 monthsUp to +0.25 SDStable or grew

The evidence shows that graduation programs have the potential to deliver lasting gains. Most trials in Carter et al. measured effects between six months and four years after the program ended, and across them consumption gains were on average as large or larger at later measurements. A few trials followed households for longer. In India, gains grew for seven years and were still significant after ten. In Ethiopia, consumption was still 9% higher after seven years; although that is a lower effect size than found after three years, this decrease is mainly due to rising living standards among comparison households.8 In Bangladesh, one study found meaningful impact 17 years after the intervention, including intergenerational effects.9

However, some trials have shown no or only short-term impact.10 Several programs in the meta-analysis had small and statistically insignificant effects on consumption. In several of these cases, the program was not carried out as planned, because of conflict, instability or other operational problems. In Burkina Faso, cash grants were delayed, and field staff contracts ended around the time the grants arrived, so participants received little coaching afterwards. In Yemen, unrest led to cancelled coaching visits, and consumption support was suspended for over a year. In South Sudan, the implementing organization’s local office closed about six months after the asset transfers, while coaching was still under way. In Honduras, most participants chose chickens as their asset, and many of the chickens later died of disease. Fragile settings do not rule out strong results, however. Programs in Somalia and Afghanistan were carried out as planned and had some of the largest effects in the meta-analysis.

Meta-analysis: graduation programs raise consumption, income and assets.

Average effects in a meta-analysis of 40 randomized trials of graduation programs. The comparison bar is the median comparison-group level across studies, so bar totals are illustrative. Source: Carter et al. 2026, Appendix Table A.3.

Comparison with unconditional cash transfers

Graduation programs and unconditional cash transfers both aim to reduce extreme poverty. Whereas cash gives households money and the choice of how to use it, graduation programs typically give a smaller transfer, together with training, coaching and help with saving, so that households can build lasting sources of income and consumption. While both approaches are supported by evidence, the overall evidence suggests that graduation programs that are well run and low-cost can produce longer-lasting gains than cash, at a lower cost per household.

Graduation rests on the idea of a ‘big push’: support on several fronts at once, substantial enough for households to raise their living standards durably (that is, to ‘graduate’ out of extreme poverty). That is why programs typically combine the components described above. A transfer of goods or cash provides the means to start or expand an income-earning activity. Training builds the skills to earn more from it. Support for basic needs reduces the pressure to sell or use up what the household has received. Coaching helps participants apply the training, plan, stay motivated and overcome obstacles, and help with saving builds a reserve for emergencies. In graduation programs started since 2016, the median program cost is about $520 per household.11

In comparison, cash typically involves a larger transfer with less emphasis on the other components. In GiveWell’s model of GiveDirectly’s Cash for Poverty Relief program, the transfer is about $865 to $1,100 per household (2024 US dollars), depending on the country. Households consume about 60% of the transfer in the first year and invest the rest. As a result, about 42 to 45% of the consumption gain to recipients falls in the first year, and 68 to 70% within three years.12 Some studies have identified longer-term impacts, including in Kenya (unpublished) and Uganda.13

The two approaches appear to be converging. Graduation programs are becoming lower-cost and more similar to cash transfers, while GiveDirectly is experimenting with incorporating aspects of the graduation approach. In graduation programs, the main transfer is now most often a lump sum of cash rather than a physical asset such as livestock,14 and costs have fallen as implementers sought lower-cost designs and governments took programs on.15 GiveDirectly, for its part, now pairs its cash transfers with light-touch coaching and savings support.16 It is also testing whether adding training and behavioral support to large transfers can match graduation programs at lower cost.17

Overall, the literature suggests that recent graduation programs can be more cost-effective than cash transfers, provided the costs are low. Carter et al. (2026) compared the two under three assumptions about how long gains last.18 In each scenario, graduation outperformed cash when looking at graduation programs started since 2016, while cash was more cost-effective in two out of three scenarios for programs started before 2016.19

These results are consistent with two earlier meta-analyses. Kondylis and Loeser (2021), covering mostly older programs, found that on average, graduation programs raised consumption more than cash per dollar transferred, but were less cost-effective at the time horizons most trials measured. The gap was smaller for low-cost graduation programs, and results varied more across graduation programs than across cash transfers.20 Leight et al. (2024), a meta-analysis of cash and ‘cash-plus’ programs, found that adding training or similar support to cash transfers raised consumption effects per dollar transferred by about half, but that such additions are cost-effective only under specific circumstances.21

Taken together, these findings suggest that graduation programs, if well-designed, well-run and low cost, have the potential to be substantially more cost-effective than cash transfers.

Poverty Graduation Sector Recommendation: Raising The Village

Raising The Village runs a 24-month program in remote, underserved villages, which it selects in partnership with local governments.22 Every household in the village takes part, not only the poorest. Raising The Village costs much less per household than a classic graduation program.23 Notably, its trial version had the lowest program cost in a large 2026 meta-analysis of graduation programs. For 2027–28, we estimate that it will cost about $161 per household in Uganda, its largest country, against Carter et al.’s median of about $520 for programs started since 2016.24

We consider RTV’s costs remarkably low, given that Raising The Village’s impact per household, as measured by a randomized trial (Mahmud and Riley 2026), is comparable to other graduation programs.

Raising The Village’s model

Raising The Village operates in Uganda, Tanzania, Rwanda, and the Democratic Republic of the Congo. Its program works with a village in phases: planning, six intensive months of training and inputs, and a transition to community ownership. After two years, the program ends, but RTV continues to carry out monitoring that includes annual surveys for future program refinement. More specifically:

  • Pre-planning. Raising The Village meets with local government to select last-mile villages: places with limited services, such as schools, clinics, or piped water, where families are reliant on farming for both food and income. Staff then plan the program with the village. They map and survey households, ask the village to list its needs, and select seed packages for later distribution based on local climate data and meetings with households.
  • Core implementation (months 0–6). Households receive seeds and seedlings selected for their village and context: for example, high-yield bean, groundnut, and maize seed, seed potatoes, and vegetable and fruit seeds. The village receives tools that households can borrow, such as hoes, forks, and spades, as well as shared equipment. For instance, trial villages received a milling machine to grind grain, spray pumps for pest control, and water drums and watering cans for gardens. Households attend one-day training sessions timed to the crop cycle, covering planting, organic fertilizer, pests, perennial crops such as coffee, post-harvest storage, and selling produce collectively, plus sessions on savings groups, budgeting, business, gender roles, and other topics. The village elects volunteer committees (which Raising The Village calls champions), including for agriculture, livestock, and sanitation, who receive extra training, mentor their peers, and look after shared assets.25
  • Transition (months 7–12). Each quarter, independent data collectors survey a sample of households to check whether at least 90% are using the practices they were trained in.26 Villages below that level receive extra training and support. Committee members visit every household on a fixed schedule to provide their peers with support.
  • Community ownership (months 12–24). Volunteer committees, supported by government frontline workers who are involved from the core implementation phase, take over more of the day-to-day work, such as mentoring neighbors and managing shared tools. Staff shift from delivery to mentoring and handover, with refresher trainings where data show a need. At month 24, the village graduates. Village clusters short of their income and production targets can stay in this phase for six more months.
  • Follow-up (months 25–60). Program activity ends. In order to inform future program design, Raising The Village surveys a sample of households each year until five years after the start of the program to track metrics including income, production, assets, and food security.

Raising The Village has incorporated the core components of the graduation approach as follows.

ComponentIn Raising The Village’s program
Asset or cash transferA small package of goods: seeds and seedlings for each household, and shared tools and equipment for the village; no cash.
TrainingGroup training in farming, livestock care, hygiene, financial literacy, business, and gender roles, with demonstration plots in the village.
Consumption supportNot incorporated. The closest analog is the introduction of kitchen gardens at the beginning of the program, where households grow vegetables for their own meals. This supports food consumption.
CoachingCoaching visits by staff, and elected village committees that mentor households and check that at least 90% adopt each practice. At under 90%, villages receive additional support.
SavingsAt least two savings and loan groups per village, usually saving for livestock, household assets or a small business.

In addition to these components, Raising The Village’s program also covers water and health. Villages receive new or rehabilitated water sources where needed, households are trained and coached on water, sanitation and hygiene, and health outreach sessions are held in the village.27

Whole-village approach

Most graduation programs target the poorest households in a village. Raising The Village works with every household instead, and actively involves community volunteers in its programming. According to its theory of change, this provides multiple benefits simultaneously:

  • New practices spread. New practices – e.g., for increasing yields or saving money – spread through the community via shared learning and peer accountability.
  • Downsides of household targeting are bypassed. It avoids selecting households within a community in which even better-off households tend to be poor. Selecting households can be costly, error-prone, and perceived as unfair.
  • Costs are lower. Training can be given to groups, volunteers can mentor neighbors, and no money is spent on deciding which households are eligible.

As mentioned above, a core part of the village model is the formation of committees for topics such as agriculture, livestock and sanitation. Committee members are elected by the village and receive an additional day of training in their area. Beyond skill mentorship, committee members are responsible for ensuring that at least 90% of households attend each training and adopt the practices taught. In this way, they support sustained skill adoption within the village. Raising The Village’s staff check at quarterly visits that committee members are carrying out their duties and meeting their 90% targets.28

We are not aware of any other graduation program that works with whole villages. Raising The Village also differs from most graduation programs in its use of village committees and in the few assets it provides upfront. We are nonetheless confident that it is a graduation program and that the evidence on graduation applies to it. Definitions vary. Carter et al. (2026) count a program as graduation if it has at least three of the five core components. Raising The Village has four, and its trial is part of their meta-analysis. BRAC’s Graduation Essentials are stricter and include basic needs support, which Raising The Village does not provide; however, the author of BRAC’s Graduation Essentials, Heidi McAnnally-Linz, told us she considers Raising The Village’s kitchen gardens a form of consumption support. In addition, one of the largest effects in the meta-analysis comes from a program without consumption support, and Carter et al. cite Raising The Village’s trial as evidence that a large transfer is not necessary for large effects.29

Of the differences, we consider the small asset transfer the most relevant to impact, but not for whether it counts as graduation or not, as the literature does not define a minimum value. Universal coverage and village committees concern how the program is delivered, not what households or villages receive. They may affect cost-effectiveness (see below), but not whether the program counts as graduation.

Scale and growth plans

Raising The Village is a scaling organization. Over the past decade, it has grown from reaching approximately 24,000 people in Uganda in 2016 to 618,000 in 2025 (647,000 across its three countries). It now works in four countries: in 2024, it began work in Rwanda and the eastern Democratic Republic of the Congo (DRC); in 2026, it began work in Tanzania. Across its countries, the number of people reached grew by 55% in 2025 and grew by 55% again in 2026, to about one million (1,002,415).30

Raising The Village is pursuing a strategy of rapid growth. It aims to reach approximately 2 million people in 2027 and between 2.5 and 4 million in 2028. Between these two years, it projects approximately 61% of its spending to be for programs in Uganda, 18% for Tanzania, 12% in the DRC and 9% in Rwanda.31

Growth increasingly runs through governments. In Uganda, Raising The Village told us it expects government staff seconded to its program to support delivery in about 75% of its coverage area by 2027, while it continues to manage the program. In Rwanda, government frontline workers already deliver the program and the government subsidizes seed. In the DRC, Raising The Village works with local partners who deliver its program in conflict-affected areas.32 Collaboration with governments and local NGOs shifts part of the cost of delivery to them, but also adds new challenges. Across graduation programs, government involvement is associated with lower costs but also smaller effects, although Carter et al. find no difference in cost-effectiveness.33 We discuss this further under ‘Changes to the program since the trial & external validity adjustments’ and ‘Organizational capacity’.

Evidence of impact

Trial results

Raising The Village’s impact has been independently investigated through a randomized controlled trial (Mahmud and Riley 2026). The researchers randomly assigned 114 clusters of villages, 335 villages in all, to receive the program or not. They surveyed 3,852 households in 2020, before the program reached their villages. They surveyed the same households again in August to October 2023, just after a harvest. A phone survey in June 2024 measured consumption and hunger in the lean season.34

While the study has not yet been published in a peer-reviewed journal, we consider the working paper’s results credible because the study meets common quality standards, including: pre-registration, an analysis plan lodged before the endline data collection was complete, balanced groups, low attrition, and main results that remain significant after adjusting for the number of outcomes tested.35 Additionally, its impact findings are in line with the broader literature on graduation.36

The results of a follow-up study by the same authors, measuring the program’s impact five years after the start, are expected next year. Two new randomized trials are planned: one in Eastern Uganda, starting with a 2027 cohort, and one in the DRC, planned for 2028 and still at the concept stage. The follow-up study and the Uganda trial are funded by GiveWell.37

In the trial’s endline survey, households in program villages consumed 10% more per month than households in comparison villages ($59 more, on a base of $583, in 2023 PPP dollars). The gap was wider in the lean season, the months before harvest when stored food runs low: then, program households consumed 39% more than comparison households ($150 more, on a base of $387, in 2023 PPP dollars). The same survey shows the difference in hunger directly: In the lean season, 38% of comparison households reported an adult going to bed hungry in the past month, and 25% reported a child doing so. In program villages, the figures were reduced to 6% and 4%. The endline survey also found higher income and wealth in program villages. Income was 26% higher. Wealth other than land (assets, livestock, savings and net lending) was 31% higher. The share of households below the $2.15-a-day poverty line (2017 PPP) was 7 percentage points lower at the endline (28% versus 35% in comparison villages) and 26 percentage points lower in the lean season (39% versus 65%).38

About half of the income gain came from selling more crops, and a third from more households running small businesses.39 In the first growing season, harvests of the main crops rose: beans by 15%, maize by 14%, and groundnuts by 33%, partly because households planted about 15% more land. Households also grew more types of crops.40 Enumerators observed that the farming practices taught in training had been adopted. For example, 44% of program households had a compost pit, against 20% of comparison households. Households also reported using more than twice as much fertilizer.41

The poorest quarter of households gained about as much in dollars as wealthier ones, and therefore much more in proportion. Their income was 41% higher than that of the poorest quarter in comparison villages, and their wealth nearly doubled. By the endline, their income was statistically indistinguishable from that of the second-wealthiest quarter of comparison households. This addresses a potential concern about universal programs: that better-off households capture the benefits.42

Beyond the headline figures, the study shows how household economies changed. Non-economic effects are discussed under ‘Spillovers and non-consumption effects’ below.

  • In the lean season, program households ate 53% more food from their own farms, by value, than comparison households.43
  • Households saved more. Savings held in village groups were 27% higher, and households bought chickens and goats beyond the pig some received.44
  • Households owned more assets, for example beds and mattresses, solar panels, jerry cans, and pots and pans.45
  • In both growing seasons, more program households hired farm labor themselves (4 to 6 percentage points more).46

The trial took place during the COVID-19 pandemic. The baseline survey was carried out in 2020; the first round was done by phone during Uganda’s lockdown, when incomes were temporarily depressed. The authors therefore describe households using the comparison group’s endline statistics rather than its baseline. Their percentage effects are also calculated against endline means, and we do the same in our cost-effectiveness analysis. The results also do not meaningfully change when baseline controls are excluded.47 Even so, the program was tested under unusual conditions, and we treat this as a reason for extra caution when interpreting its results (see ‘Plausibility’, directly below).

Plausibility

Mahmud and Riley’s trial shows Raising The Village achieving effects similar to other graduation programs, at what we calculate to be about a third of their cost.48 Most of the difference in costs can be traced to the program design:

  • Classic graduation programs give each selected household a relatively large asset or cash grant. Raising The Village gives every household relatively inexpensive inputs (worth about $66 PPP on average in the trial cohort, for seeds, tools and livestock), plus a share of village-level equipment such as milling machines.
  • Raising The Village provides no consumption stipend.
  • Raising The Village spends nothing on selecting households within villages.49
  • Because every household in the village takes part, training can be delivered in groups, and ongoing mentoring is done by elected volunteer committees, with periodic staff visits, rather than by paid coaches.50

Because of the exceptional cost-effectiveness Mahmud and Riley find – and because the literature describes the difficulty of streamlining graduation programs without affecting their effectiveness51 – we have taken extra care to verify the plausibility of its impact. Conversations with Raising The Village’s leadership, consultation with experts, a comparison with the poverty graduation literature, and an analysis of impact mechanisms and data have led us to believe that Mahmud and Riley’s results are plausible.

Most importantly, Raising The Village appears to be an exceptionally data-driven organization, with fast feedback loops between its M&E and program activities. During the course of our evaluation, we noted several instances of Raising The Village changing its program in response to its own data. For example, in 2022 it stopped providing livestock in new villages and kept only livestock training, after its household data showed that livestock transfers did not recover their cost by the time households completed the program. Its household data also showed higher returns from perennial crops, so it added training in growing them.52

Raising The Village tracks its results mainly through an Annual Household Survey, run after the June to July harvest. In each village it samples households at random and follows them from baseline to month 60. Its 2025 Impact Report draws on about 42,000 surveyed households in 1,575 villages. It runs the same survey in comparable peer villages where it does not work and compares how the two groups change over time.53 In addition, leadership reviews program data quarterly and decides on changes for the next quarter.54 Field staff know what targets they are optimizing toward and are empowered to adjust their work based on information that is rapidly shared back with them.

In addition, RTV’s work is highly adapted to a village’s local context and resources. The seeds and seedlings it provides, for crops grown both to eat and to sell, are chosen from village climate data (altitude, soil type, temperature, rainfall) and design sessions with households.55 It teaches households cost-effective methods to improve yields, such as making organic fertilizer in household compost pits. Village committees and coaching visits then follow up on the training to survey and increase uptake.56

Carter et al.’s (2026) meta-analysis57 and Raising The Village’s own M&E data – which it makes available publicly in its yearly Impact Reports – support Mahmud and Riley’s findings. Carter et al. find that graduation programs typically increase consumption by a similar share as Raising The Village (see below). Raising The Village’s own monitoring, which compares changes in its partner villages with those in similar villages it does not serve, finds larger gains in household income and production than the trial did.58 This monitoring is not randomized, so we do not directly use it in our cost-effectiveness analysis, but it points in the same direction as the trial.

We have approached experts to further confirm the plausibility of Mahmud and Riley’s results, with a particular emphasis on its lean season effect. Among others, Samantha Carter (co-author of Carter et al. (2026)’s meta-analysis, currently Livelihood Impact Fund) and Joshua Deutschmann (University of Chicago’s Development Innovation Lab; co-author of Deutschmann et al. (2025) on One Acre Fund in Kenya) confirmed they saw no reason to assume that the results overstate Raising The Village’s impact. Heidi McAnnally-Linz (Livelihood Impact Fund, Raising The Village’s largest funder; formerly BRAC and IPA) remarked that “even if the results [were] half as good, it’s still one of if not the best deal in poverty reduction.” Kelsey Jack (UC Berkeley; co-author of Fink, Jack and Masiye (2020) and Augenblick et al. (2026) on consumption smoothing in the lean season) considers the lean season premium plausible and expects it to fade more slowly than the post-harvest effect, because consumption variance tends to fall faster than mean consumption rises as households become better off. In her view, multiple practices Raising The Village teaches, such as storage, perennials and crop diversification, kitchen gardens and savings groups, are all particularly useful for addressing consumption seasonality.

Overall, we think Raising The Village’s data-driven practices and culture help explain how it has achieved a level of costs far below those of other graduation programs at a similar level of impact. (See the sections on effect size and external validity below on the adjustments we apply in our cost-effectiveness model to account for the remaining uncertainty about RTV’s impact.)

Cost-effectiveness

In this section we discuss our cost-effectiveness estimate for Raising The Village. Our cost-effectiveness analysis (CEA) combines Raising The Village’s projected costs for 2027–28 with the effects Mahmud and Riley measured in their randomized trial, and expresses the result as the cost of doubling one person’s consumption for one year.

Our CEA consists of several parts, each discussed in more detail below. It begins with cost: Raising The Village’s projected cost per participant in 2027–28 in each of its four countries of operation. It then takes the trial’s effect on household consumption, measured separately for the months after harvest and for the lean season, when food runs low and the program’s effect is nearly four times larger in percentage terms. Because the trial followed households for three years, we model seven scenarios for how long the effect lasts beyond that, weighted by our judgment of how well each fits Raising The Village’s impact mechanism, its own monitoring data and long-run studies of other graduation programs. We then make a series of adjustments, including for how the trial’s findings compare with the wider graduation literature. Finally, we apply and adjust the RCT evidence to each of the four countries, subtract the value that government co-funding would have had elsewhere, and divide by cost. Our estimate uses our best guess for each input. The uncertainty range around it comes from 10,000 simulations across the main uncertain inputs.

Metric: consumption doublings

We measure impact through consumption: the value of what a household eats, buys and uses in a month. This covers food, whether bought, grown at home or received as a gift. It also covers everyday items such as fuel, soap, clothing, transport, medical care and school fees.

For households living in extreme poverty, consumption reflects living standards more representatively than income. The reason is that these households grow much of their own food. Take a family that harvests more beans and eats them. The family is better off, but it has earned no extra money, so a survey of income records shows no change. Consumption is also one of the main outcomes in the graduation literature: using it lets us compare programs on more similar terms.

Consumption is an imperfect measure. It comes from surveys in which families estimate what they have eaten and spent, and studies do not all measure it in the same way. Even so, it remains the most direct measure available of how households are living.

Our headline result is in consumption doublings: doubling one person’s consumption for one year. We use proportional rather than dollar gains on the principle that an extra dollar matters more to a poorer person than to a richer one. We discount future doublings 4% per year. This reflects the risk that projected benefits do not arrive, for example because AI reshapes the economies these households depend on; because doublings already value gains relative to income, we add no discount for rising incomes.59

Graduation programs affect more than consumption. For instance, in the Raising The Village trial, households in program villages earned about $40 (2023 PPP) more in cash income a month than comparison households just after harvest, but spent only about $19 (2023 PPP) more.60 Households appear to have saved or invested part of the difference. Although these gains are not counted in our headline estimate of consumption doublings, they inform how we model the persistence of the consumption effect, because investments in assets can keep producing income and savings are able to be spent on later consumption needs.

Our findings

As discussed above, graduation programs are among the better-evidenced tools against extreme poverty. Two elements drive our results for Raising The Village: its program costs much less per household than other graduation programs, and the trial found effects of similar size to theirs. Our CEA estimates that:

  • Raising The Village doubles one person’s consumption for one year for $27. This excludes potential spillovers to other village clusters and health effects, which we did not model.
  • In 80% of 10,000 simulations that vary the main uncertain inputs in our model, the cost per doubling falls between about $16 and $62.
  • Raising The Village is approximately 6x as cost-effective as cash transfers from GiveDirectly (80% of simulations: 2.4x to 9.6x). Even when all of GiveDirectly’s spillovers and health effects are counted, Raising The Village remains 3.5x as cost-effective (80% of simulations: 1.5x to 6.1x). It beats cash in more than 99% of our simulations, and in more than 95% when all of cash’s spillovers and health effects are counted. These shares reflect the uncertainty we model, not all the ways our model could be wrong.
  • Raising The Village is 11x GiveWell’s benchmark. GiveWell, a leading charity evaluator, measures programs as multiples of a fixed cash-transfer benchmark.61 This is a separate yardstick from the GiveDirectly comparison above.

Schematic overview

The following interactive tool shows a simplified version of each step of the CEA. Values that involve judgment can be changed to see how they affect the endline result.

Below, we highlight what we consider the most important findings and judgment calls in our CEA. For further details and reasoning please consult the full CEA.

Effect size and internal validity

Mahmud and Riley’s main finding comes from an in-person survey just after the harvest: they found treated households consumed 10% more than control households ($59 (2023 PPP) per month). A separate phone survey in the lean season, when food is scarce, found a much larger gap: 39% ($150 (2023 PPP) per month).62 In our CEA, we calculate consumption doublings for each season separately. We then combine them, weighted by the number of lean months in a year. This way, we account for consumption smoothing: the fact that a family profits not only from additional consumption, but also from being able to eat more when food is scarcest.63

As discussed above, we have made extra effort to verify the effect sizes Mahmud and Riley found because of Raising The Village’s unique level of cost-effectiveness. That has led us to make the following adjustments.

  • We apply a downward adjustment of 10% to the post-harvest effect. Carter et al.’s (2026) meta-analysis pools consumption results from 52 graduation program arms. Its model pulls each study’s result toward the average of all studies, more so for noisy or unusual results. Carter et al.’s model shrinks Mahmud and Riley’s post-harvest effect by about 2%.64 Because this is the first randomized trial of Raising The Village and it has not yet been published in a peer-reviewed journal, we apply a larger, more conservative discount.
  • We discount the lean season’s extra gain by 30%. We are not aware of other trials of graduation programs that measure a lean season effect. Mahmud and Riley measured it using a phone survey, as opposed to an in-person survey for the endline effect. The phone survey was of high quality: It reached 97% of households, used the same consumption questions as the in-person survey, and interviewers did not know which villages had the program. Audio audits of 5% of calls, transcribed by an independent enumerator, found no tip-off mentions of Raising The Village.65 However, to account for uncertainty around this effect size and its persistence, we count only 70% of its excess over the post-harvest gain.
  • We see no reason to further adjust Mahmud and Riley’s results, for instance because of potential bias. The trial villages were selected through RTV’s usual processes, and the program was no more expensive in the trial than in routine cohorts, so we see no sign that the trial villages received special treatment.66 The authors registered the trial in 2020 and lodged an analysis plan before the final data collection ended. Attrition was low and balanced across groups, the groups were similar at baseline, and the main results survive correction for multiple testing.67

One alternative we considered with regard to internal validity was to anchor our adjustments to Raising The Village’s level of cost-effectiveness instead of its effect size. That would have led to steeper discounts, as the former is more unusual than the latter. Because we think there are plausible explanations for Raising The Village’s effect size given its lower costs (see above), we do not believe such an extra discount is warranted.

The cost-effectiveness analysis itself includes further explanations for our adjustments.

Persistence of effects

Mahmud and Riley measured effects nearly three years after the program started (the endline). As a consequence, how long the effects last beyond that is a main driver of uncertainty in our model.

Because there is no randomized evidence yet on how long Raising The Village’s effects last, we have modeled multiple potential scenarios. Some scenarios follow persistence trajectories of other graduation programs in the literature; others follow from Raising The Village’s own monitoring data and its impact mechanism. Based on a comparison between Raising The Village’s impact mechanism and the literature, we assign a probability weight to each scenario.

The scenarios and their weights are informed by the following considerations. The first two point to effects that persist; the second two point to effects that fade:

  • Most of the consumption gain is food (77%), and most of that is home-grown (86%).68 This rests on practices RTV teaches, such as composting and post-harvest drying. There is little reason to expect that farmers would abandon these practices after observing them for a few years. Elsewhere in Uganda, farmers kept using such practices three years after program support ended (Fishman et al. 2022). In Mozambique, gains from a one-time input subsidy persisted in later years without it (Carter, Laajaj and Yang 2021).
  • Raising The Village’s own monitoring shows the gap in daily income and production with peer households still widening two years after the trial ended, though more slowly.69
  • However, about 30% of the trial’s income gain comes from more small businesses.70 Businesses like these close at relatively high rates71 and are less likely to be founded after the intervention has ended.
  • Households without the program are likely to improve over time, which narrows the gap even if treated households don’t lose ground. Poverty is slowly falling in the region. In Raising The Village’s own data, peer households’ poverty probability fell from 31% to 24% over five years.72

In their meta-analysis of graduation programs, Carter et al. (2026) find larger consumption effects at later follow-ups across studies, although the evidence varies. However, few trials follow households beyond four years. We therefore use this finding cautiously, and we conservatively model an upward trend in only one scenario, with a 5% weight.

The table below describes the persistence scenarios we model and the weights we give each one. We cap impact at 15 years after endline (i.e., 18 years after the start of the intervention) to avoid becoming speculative. Additionally, because we discount future years by 4% per year, later impact counts less: e.g., a consumption doubling in year 18 is worth 51% of a consumption doubling in the first year. GiveWell’s 2024 back-of-the-envelope estimate for Raising The Village counts about three years of effects (range two to eight).73 We model longer persistence because of the mechanisms above and the long-run evidence on the persistence of graduation. Our conversations with Samantha Carter, Joshua Deutschmann, and Kelsey Jack have informed our judgments on persistence (see Plausibility above); all scenarios and judgments are our own.

Scenario, from the end of the trialWeightReasoning
Gradual decline
Decline to zero after 8 years20%Effects fall linearly by 12.5% a year, as in Carter et al.’s “conservative” linear-decay scenario. It is based on an Ethiopian program, although it is more conservative than that: consumption there was still 9% higher after seven years, against 18% after three, and the decrease was mainly due to rising living standards among comparison households.74
Decline to zero after 15 years25%A decline at about 7% per year. This is roughly consistent with agricultural outcomes remaining stable, but business outcomes gradually fading and/or the control group gradually catching up.
Decline to 50% over 15 years, then stops20%A more optimistic version of the previous scenario, roughly consistent with agricultural outcomes that improve further (e.g., because of further investments in livestock), business outcomes fading less, and/or the control group catching up more slowly.
Stable first
Stable for two years after endline, then no effect5%This could occur in case of an abrupt loss, for example after a shared shock such as drought or livestock disease.
Stable for 5 years, then fades over 1020%RTV’s own monitoring shows impact still growing, slowly, between years four and five, about two years after endline. This scenario extends that plateau a few years, then assumes a gradual decline.
Stable for 15 years, then stops5%Carter et al.’s permanent persistence scenario, cut off at 15 years. Given the slow but steady decline of poverty levels, and because RTV’s effects are arguably more transferable between villages than more traditional graduation programs which provide expensive assets, we think it’s plausible that poverty levels would slowly decrease over time without Raising The Village.
Increase first
Grows 5 years, then fades over 105%Based on Carter et al.’s estimated slope (+$48 (2020 PPP) a year; the 5-year and 10-year periods are our judgment). Although the organization’s M&E data shows modest growth between the fourth and fifth year (i.e., up to two years after endline), we have not seen evidence of its impact growing further. See previous scenario for why we think it’s plausible that poverty levels would slowly decrease over time without Raising The Village.

Cumulative doublings per scenario

Each line adds up one person’s consumption doublings under one scenario for how long gains last after the trial. The black line shows the weighted average of the scenarios, using the weights in the table above.

Changes to the program since the trial & external validity adjustments

Raising The Village has informed us of the following changes to the program since the trial that could affect cost-effectiveness: more delivery by seconded government staff in Uganda, the program no longer giving livestock, an added emphasis on perennial crops, and support for collective marketing.75 In this section, we discuss these changes in more detail and our adjustments for them. The figures below are from Raising The Village’s monitoring data. Because our CEA is built on the trial, they enter it only through our judgment on adjustments.

Raising The Village’s monitoring data give no sign that changes since the trial have reduced the program’s cost-effectiveness. In Uganda, impact has not dropped across cohorts, and in the DRC, the first cohort shows relative gains at 12 months about as large as those of Uganda cohorts at the same stage. Rwanda’s first cohort shows smaller gains at 24 months (see table below).76

Government staff increasingly deliver part of the program in Uganda. In half of its Ugandan partner districts, government extension workers are seconded to the program and spend three to four days a week on it alongside Raising The Village’s staff. Raising The Village expects this to cover about 75% of its operations by 2027, including all new districts, and is testing delivery with seconded government staff against delivery by its own staff.77 This shifts part of the staff cost to the government. Although delivery continues to be supervised by RTV, it risks a lower level of implementation fidelity. Please see the table below for our specific adjustments.

The program no longer gives livestock. Raising The Village’s own analysis found that whereas animals raised household value at graduation, this increase was not enough to offset their cost. Specifically, their data showed households realized an average gain of $20.49 in livestock income and $22.44 in livestock income plus consumption, but at an average cost of $27.59 per household, the ROI was negative. Consequently, we think it is plausible that dropping livestock has raised cost-effectiveness, at least over the program period. Households continue to build up livestock without the transfer. In the trial, households also bought goats and chickens, beyond the pigs some of them received, using higher incomes and savings groups.78 In Raising The Village’s 2023 cohort, which started after it began phasing out livestock distribution, the share of households owning livestock rose from 60% at baseline to 78% at graduation, against 58% to 60% in peer villages.79 However, to be conservative, we apply a modest downward applicability adjustment in our CEA as a consequence of this change.80

The program has added training on perennial crops. Data from the 2022 and 2023 cohorts showed little income from perennials. However, Raising The Village saw perennials as a promising lasting income source after graduation: perennials keep producing for years. It added perennial crop training in 2024. In the first cohort to receive it, perennial income per household rose from $23 at baseline to $102 at year 1, and the value of perennial crops from $55 to $163; by year 1, perennials made up 43% of agricultural income.81 If such gains last, they support persistence. We made no upward adjustment for this change.

Raising The Village has introduced support for collective marketing. Previously, higher agricultural production did not always raise income, because households sold individually and often received lower prices due to limited bargaining power. In 2024, collective marketing was introduced as a formal part of the program. Households now pool produce and negotiate through bulk sales. Raising The Village describes this as a move from focusing predominantly on production outcomes to addressing broader market-access and value-generating pathways. It reports that households taking part earned about $174 more than those that did not, and that potatoes sold collectively fetched $49.58 a bag against $26.84 sold individually.82 We make no upward adjustment for this change.

Other considerations

Our CEA models the expected cost-effectiveness of a donation in 2027 and 2028. We have separately modeled the expected impact in the four countries Raising The Village plans to operate in during this period. The rationale behind their external validity adjustments is described in the table below.

Please note: The heavier adjustments for Rwanda, the DRC and Tanzania should be read in context. Raising The Village calls Rwanda and the DRC its Learning Hubs: deliberately small programs where it tests how its model performs under different delivery arrangements, including government-led delivery in Rwanda and local-partner delivery in the DRC’s fragile setting. Tanzania is its newest program. These are the kinds of tests it will need to undertake if it is to grow from about one million participants a year to several million, as it intends. Early cohorts in a new country cost more and are less certain to deliver. We model that, as the figures show. However, a scaling organization that did not run such tests would be the greater risk, and projected 2027–28 spending outside Uganda remains a minority (39%) of the total.

CountryShare of spendingApplicability of trial to this context (External Validity)Justification
Uganda61%85%Same country as the trial, which already tested routine delivery at scale. Later cohorts did at least as well as the trial cohorts on Raising The Village’s monitoring. We discount for the end of livestock provision (see above), fast growth into new districts, and the growing share delivered by seconded government staff (which makes implementation fidelity less certain; however, government extension workers already supported delivery in the trial, and Raising The Village supervises delivery83).
Tanzania18%70%The program started in 2026, so 2027–28 spending falls in its second and third years. As a rough anchor: in Uganda, Raising The Village’s 2020 cohort, launched during the COVID-19 pandemic, gained about 70 to 80% as much as its 2021–23 cohorts by month 24.84 We use the lower estimate to account for the country transfer.
DRC12%50%Ongoing conflict and displacement threaten both delivery and persistence.85 Because armed groups target high-value assets, the program has been adapted, e.g., with emphasis on more transportable assets like guinea pigs and shorter-horizon crops.86 Delivery runs through local partners. However, Raising The Village’s monitoring at 12 months shows relative gains about as large as those of Uganda cohorts at the same stage, well above what our 50% adjustment implies.87
Rwanda9%70%Stable operating environment. Raising The Village’s monitoring at 24 months shows gains over peer households, but smaller income and production gains and lower returns than Uganda cohorts at the same stage.88 However, government extension workers deliver the program, which makes implementation fidelity less certain. Farms are also smaller, leaving less room to plant more land.89

The shares of spending, included above, are based on Raising The Village’s projections, based on historical cost metrics by country (see below).

Uganda, which gets most of the money, is the most cost-effective country.

Dots: $ per consumption doubling at central inputs, as in the table above. Bars: the middle half of 10,000 simulations (25th to 75th percentile); bar thickness shows each country’s share of planned spending. Lower is better. The dots sit near the low end of the bars because several uncertain inputs have more room to fall than to rise.

Spillovers and non-consumption effects

Our model does not include spillovers beyond a trial cluster or effects other than consumption.

We think positive spillovers are plausible, but we are unsure of their size. The program serves every household in a cluster of villages, so spillovers within village clusters are already part of the trial’s estimate.90 Between clusters of villages, Mahmud and Riley find suggestive evidence of positive spillovers. They suggest local demand as the cause: for example, traders are more likely to travel to an area when enough is for sale there.91 We think spillovers are likely to be lower than those for cash – which GiveWell credits with spillovers to neighbors worth about 60% of the benefits to recipients92 – because cash spillovers come mainly from recipients spending a large transfer locally (about $1,000 per household in Kenya, over 15% of local GDP).93 Raising The Village’s transfer is a small package of seeds and seedlings, plus tools shared at village level.94

We think negative economic spillovers or meaningful general equilibrium effects are unlikely. The trial found no sign of negative spillovers: no effect on households in nearby comparison villages, on local prices, or on wages paid.95 We think this could change if programs like this reach most villages in an area, because more supply could then lower crop prices. We believe that is unlikely soon, because most crops already leave the area through outside traders.96

Mahmud and Riley identified positive health and mental health effects in the trial.97 We also think it is plausible that some health effects found for cash transfers, such as lower child mortality,98 could apply to Raising The Village, although the trial did not measure mortality. Because we have not defined moral weights to translate these effects into our main metric, we have not included them in our model. However, the trial’s findings include:

  • Respondents in program villages (the household head or spouse) lost about one day less of work to illness in the past month (3.8 days, against 4.7 in comparison villages).99
  • Fewer school-age children missed school for illness in the past month: 37%, against 43% in comparison villages, with a larger drop for girls.100
  • Respondents in program villages were about 6 percentage points less likely to show symptoms of depression, and reported less stress and higher life satisfaction and happiness.101

Cost per person

For each country, we use the cost per person that Raising The Village projects for 2027 and 2028.102 These figures cover all of Raising The Village’s own spending, including overhead and the start-up costs of its newer country programs.

Raising The Village notes that these figures should be treated as high-level projections for planning purposes only. In regions where program operations are less mature (Tanzania) or where it has a smaller operating presence (Rwanda and DRC), actual costs may vary as strategic plans and annual budgeting cycles evolve, with higher start-up costs included.103 We therefore raise its projections in our CEA: by 5% in Uganda, 20% in the DRC and Tanzania, and 30% in Rwanda. This is not a judgment on Raising The Village’s budgeting. Across organizations, costs more often end up above projections than below them, and projections for newer programs are less certain. We also vary each country’s cost in our uncertainty analysis.

CountryCost per person (USD 2026)Uncertainty adjustment applied
Uganda$325%
DRC$4020%
Rwanda$4430%
Tanzania$4720%

Some of the program costs are paid for by governments or partner organizations. These costs, or in-kind support, are not part of the cost figures.104 In the CEA, we instead subtract what those contributions would likely have achieved if spent elsewhere, using GiveWell’s estimate of the value of government spending.

In Uganda, $1,000 buys about 45 consumption doublings.

Uganda at central inputs. Our overall estimate, $27, is higher because it includes the other three countries.

Comparing with cash

We compare our cost-effectiveness estimate with GiveWell’s estimate of cash transfers from GiveDirectly, using their 2024 analysis of its Cash for Poverty Relief program.105 To make the best possible comparison, we use an estimate for cash that counts gains to recipients plus gains to their close neighbors. This is because Mahmud and Riley compared whole clusters of villages, so their estimate already includes any gains that pass between neighbors. Based on Egger et al. (2022), we count 37% of GiveWell’s modeled gains to non-recipients. This is a rough estimate based on the share we think would arise within an area the size of a trial cluster of two to four villages.106 We average GiveWell’s five countries using its own weights and convert to January 2026 dollars.

Counted this way, cash transfers cost $151 per consumption doubling. Counting recipients’ gains alone, the figure is $185; counting all spillovers, it is $114. Raising The Village is about 6 times as cost-effective as cash on our like-for-like comparison. It is 3.5 times as cost-effective even when all of cash’s spillovers and health effects are counted ($95 per doubling) and none of its own beyond the cluster.

Please note that our comparison with cash is different from GiveWell’s benchmark, which was based on its pre-2024 estimate of GiveDirectly’s cost-effectiveness.

Uncertainty analysis

Inputs to a cost-effectiveness model are ultimately estimates. Trial results such as Mahmud and Riley’s (2026) carry sampling error, and several of our inputs are judgment calls. We therefore report a range alongside our best estimate, and show which inputs affect it most. We do this in three ways.

  • We run the model 10,000 times. Each run draws a value for each uncertain input from its plausible range: the trial’s effects, the cost in each country, and some of our own judgment calls (a Monte Carlo analysis). Our headline estimate uses our best guess for each input; the range around it comes from these runs, and their average is also $27. Details are described in the CEA.
  • We model persistence as seven scenarios, not one. How long the effect lasts is among our largest uncertainties. Each scenario is weighted by the evidence for it and by how well it fits the program (see above).
  • We separately report the sensitivity of individual parameters in the tornado graph below.

Some inputs move our cost per doubling much more than others. The four that move it most are how long the effect lasts, how well the trial’s results apply in Uganda, the size of the lean season effect, and the cost per person in Uganda.

Uncertainty analysis

10,000 simulated results, each drawing every uncertain input from its range. Our estimate, $27, uses our best guess for each input; the curve shows the 10,000 simulations, 80% of which fall between about $16 and $62 (median $30). Cash: GiveDirectly, like-for-like (recipients plus spillovers within a village cluster, January 2026 dollars). No results lie beyond $300.

Elements that move the result the most

Each bar moves one input from its 25th to its 75th percentile while the others stay at their best-guess values. Bars are centered on our estimate ($26.97), which uses the best guess for every input. Lower is better. Grey: the discount rate is a value choice, not an uncertainty, shown at 3% and 6% instead of 4%.

Room for more funding

We are confident Raising The Village can use substantial additional funding in 2027–28 at the cost-effectiveness we have modeled. Its 2026 budget is about $33 million, and it aims to grow substantially over the next few years.107 In every growth scenario it has shared with us, it reports a funding gap ranging from about $8 million to tens of millions of dollars per year. Raising The Village will finalize its 2027 plan and budget at the end of November 2026, pending decisions by its board in consultation with stakeholders.108

Raising The Village itself names financial sustainability, including avoiding reliance on a few large funders, as an important constraint it faces.109 Its funding is relatively concentrated, with the Livelihood Impact Fund (LIF) being its largest funder by a wide margin. LIF has indicated that it is open to investing more, conditional on both increasing funding from other donors and demonstrated continued capacity by RTV to deliver with quality.110 As a result, funding from new sources could potentially also unlock additional funding from current funders.

Value of additional donations

We think Raising The Village can deploy additional donations at the level of cost-effectiveness we have modeled. That estimate is an average across its four countries, because it is currently difficult to predict in which country an unrestricted donation will be spent. Some of RTV’s funders restrict their grants to particular countries, and Raising The Village can move its unrestricted money between countries in response.111

Consequently, our model weights the four countries by their share of Raising The Village’s projected spending in 2027 and 2028, averaged over its base and accelerated scenarios (see above). This is our best estimate of where additional donations will be spent, though the actual split may differ.

Organizational capacity

Our evaluation found Raising The Village to be a well-run, transparent, and unusually data-driven organization. In reviewing organizational capacity, we identified no concerns to change our recommendation.

Overall, Raising The Village is pursuing a strategic path to becoming a meaningfully larger, multi-country organization that increasingly delivers through governments and partners. While that poses risks, including to senior team bandwidth, we believe it is adequately investing in the organizational structures needed to make that a success. We discuss this in more detail below.

Management, delivery capacity and growth

As described above, we think the quality of Raising The Village’s operations and its fast data-driven learning cycles are key to its impact and cost-effectiveness. Its ability to keep operating at this level depends on how the organization adapts in order to maintain these practices at a much larger scale.

Since its inception, Raising The Village has been led by its founder and CEO. Running an organization several times its current size asks more of a senior team and requires an institutionalized management approach. Raising The Village recognizes this. In our conversations, it named leadership and organizational capacity for rapid growth as its most pressing constraint. It is actively targeting this: it is recruiting into senior roles and has invested heavily in reshaping its senior team. Additionally, it is working with external experts to optimize the allocation of decision-making capacities.112 In sum, and after consulting experts familiar with its plans, we think Raising The Village recognizes the leadership risk in its current operating model and is addressing it appropriately.

Raising The Village increasingly uses governments and local partners to deliver its program. In Uganda, government extension workers are seconded in half of partner districts, three to four days a week, and Raising The Village expects seconded government staff to support delivery across about three-quarters of its operational footprint in 2027. In Rwanda, government frontline workers deliver the program, and the government subsidizes seed and covers the workers’ per diems. In Tanzania, the government seconds extension workers and covers fuel and part of the cost of seed and seedlings. In the DRC, local organizations deliver the program, supported mostly remotely by Raising The Village’s Uganda team, and conflict has caused intermittent pauses.113

Delivering through partners can pose risks for implementation quality and fidelity. However, this approach builds on the trial, where government vets, extension workers and health workers already helped deliver the program. Their time, priced at Raising The Village’s own salary rates, was worth about as much as the time of its own field staff.114 Additionally, Raising The Village is currently actively testing delivery with seconded government staff against delivery by its own staff, which will show whether quality holds up and allow it to mitigate quality risks.115

We have found no issues regarding Raising The Village’s finances or governance. It publishes audited accounts going back to 2016, and its auditors have given a clean (unmodified) audit opinion in each of the last five years.116 It is also an unusually transparent organization: its annual Impact Report sets out its methods, cohort data and cost-effectiveness figures.117

Conclusion

Raising The Village runs a low-cost, whole-village version of poverty graduation. We estimate that it will cost about $161 per household in Uganda in 2027–28, against a median of about $520 for recent graduation programs, and in its randomized trial it raised consumption about as much as far costlier programs. The largest gains came in the lean season, when families are hungriest. We estimate that it doubles one person’s consumption for one year for $27 (about $16 to $62 in 80% of our simulations), about 6 times as cost-effective as cash transfers from GiveDirectly. Given its low costs, strong evidence, practice of changing its program in response to its own data, and a funding gap of at least $8 million a year, we are pleased to recommend Raising The Village as an effective giving opportunity.

Notes

  1. World Bank Poverty and Inequality Platform, September 2026 release. The line is $3.00 per person per day in 2021 purchasing power parity (PPP) dollars. In plain terms, this is not what $3 buys after being exchanged into local currency. It is the amount of goods and services that $3 would have bought in the United States in 2021, covering all of a person’s needs for the day, including the value of food a household grows for itself. ↩︎
  2. J-PAL 2023; Banerjee et al. 2015; Carter et al. 2026, section 4.3. ↩︎
  3. Carter et al. 2026, sections 2.1 and 4.4.1. ↩︎
  4. J-PAL 2023, which reviews trials that removed livestock, savings support or coaching; Bossuroy et al. 2022 (Niger) for a cheaper version that stayed cost-effective. Carter et al. 2026, sections 4.2 and 4.4.2, for the pattern across programs. ↩︎
  5. Carter et al. 2026, sections 2.2 and 4.1, Table A.3. ↩︎
  6. In table order: Bandiera et al. 2017, Balboni et al. 2022, Ahmed et al. 2026; Banerjee, Duflo and Sharma 2021; Barker et al. 2024; Banerjee et al. 2015; Bedoya et al. 2019, 2023; Brune et al. 2022; Botea et al. 2025; Bossuroy et al. 2022. ↩︎
  7. The year-seven effect is significant at the 5% level, but not after the authors’ adjustment for testing multiple outcomes (Barker et al. 2024). At year seven, the effect on assets was still significant; the effects on income and food security were not (Online Appendix Table A4). ↩︎
  8. Barker et al. 2024. The year-seven effect is significant at the 5% level, but not after the authors’ adjustment for testing multiple outcomes. At year seven, the effect on assets was still significant; the effects on income and food security were not (Online Appendix Table A4). ↩︎
  9. Ahmed et al. 2026, initial evidence on the intergenerational effects of the Ultra-Poor Graduation program. ↩︎
  10. Carter et al. 2026, page 13 and Figure 2; Banerjee et al. 2015 (Honduras). Some households fall back even when average gains last: Balboni et al. 2022. ↩︎
  11. Carter et al. 2026, Data Appendix Tables 1 and 3; our calculation: median $415 for 31 arms started since 2016, in 2020 dollars (× 1.25 for January 2026). ↩︎
  12. GiveWell 2024, GiveDirectly cost-effectiveness analysis; ranges across its five countries, discounted at 4%. ↩︎
  13. Kenya: early results, GiveDirectly 2026, long-term cash research; GiveWell 2024, section 4.2. Uganda: one-time cash grants of about $400 per person to young adults (2008). After nine years, income and consumption had converged, but treated participants kept more assets and skilled work (Blattman, Fiala and Martinez 2020). After twelve years, during COVID-19 lockdowns, men’s income and employment were again higher (Fiala et al. 2025). ↩︎
  14. BRAC International 2025. ↩︎
  15. Carter et al. 2026, sections 1, 2.4 and 4.3. ↩︎
  16. GiveDirectly 2025, large lump-sum cash transfers and graduation. ↩︎
  17. GiveDirectly 2025, What giving people money doesn’t fix; GiveDirectly 2025, large lump-sum cash transfers and graduation; GiveDirectly 2026, plus cash. ↩︎
  18. Carter et al. 2026, Figure 9 and Table A.12; consumption gains only; cash figures from Crosta et al. 2026; cash cost is the transfer plus 24% for delivery. ↩︎
  19. Carter et al. 2026, section 4.4.2 and Table A.12. The 2016-onward group includes the Raising The Village trial. Without it, by our approximate calculation, graduation still beats cash in all three scenarios (e.g., 4.6 against 1.9 under estimated persistence; Data Appendix Table 4). ↩︎
  20. Kondylis and Loeser 2021, Table 1 and pages 6, 38 and 42 to 44. ↩︎
  21. Leight, Hirvonen and Zafar 2024, Table 3, pages 18 and 24, and Table A6. ↩︎
  22. Raising The Village Impact Report 2025, pages 7, 30, 82 and 125 to 128; Mahmud and Riley 2026, sections 1 to 3; Raising The Village, answers to our questions, October 2026. These are the sources for this whole section except where another note is given. ↩︎
  23. $288 against $3,612 per household, both in 2014 PPP dollars. Mahmud and Riley 2026, Online Supplement Table B2. ↩︎
  24. Carter et al. 2026, Data Appendix Tables 1 and 3: lowest of the 48 program arms with cost data; the median is our calculation; 2020 dollars × 1.25 for January 2026. ↩︎
  25. Raising The Village Impact Report 2025, pages 65 and 127. ↩︎
  26. Raising The Village Impact Report 2025, page 128. ↩︎
  27. Mahmud and Riley 2026, section 2.2; Raising The Village Impact Report 2025, pages 61 and 127. ↩︎
  28. Mahmud and Riley 2026, section 2.2. ↩︎
  29. Carter et al. 2026, abstract, sections 4.1.1 and 4.2, and Data Appendix Tables 1 and 2; BRAC International 2025. ↩︎
  30. Raising The Village Impact Report 2025, page 22; answers to our questions, October 2026. ↩︎
  31. Rounded; average of the base and accelerated scenarios in Raising The Village’s expansion budget model. Raising The Village notes these are high-level projections that may change as its regional programs mature and scale and as cost efficiencies are realized. ↩︎
  32. Raising The Village, answers to our questions, October 2026. ↩︎
  33. Raising The Village, answers to our questions, October 2026; Carter et al. 2026, section 4.2. ↩︎
  34. Mahmud and Riley 2026, sections 2.3 and 4, and footnote 40. ↩︎
  35. AEA RCT Registry, AEARCTR-0006415 (registered September 2020); Mahmud and Riley 2026, title page, sections 4.2, 4.4 and 4.5, and Table 2. ↩︎
  36. Mahmud and Riley 2026, sections 5 and 7. ↩︎
  37. Five-year follow-up: GiveWell 2025. New trial: GiveWell grants database, entries one and two. ↩︎
  38. Mahmud and Riley 2026, Table 2, Table A26, and Online Supplement Tables OS26 and OS28. ↩︎
  39. Mahmud and Riley 2026, Table A9. ↩︎
  40. Mahmud and Riley 2026, Table A13 panel A and Online Supplement Table OS13. ↩︎
  41. Mahmud and Riley 2026, Online Supplement Table OS14 and Table A16. ↩︎
  42. Mahmud and Riley 2026, Table A6 and Figure 1. ↩︎
  43. Mahmud and Riley 2026, Online Supplement Table OS27. ↩︎
  44. Mahmud and Riley 2026, section 5.2, Tables A21 and A23. ↩︎
  45. Mahmud and Riley 2026, Online Supplement Tables OS17–OS18. ↩︎
  46. Mahmud and Riley 2026, Table A17. ↩︎
  47. Mahmud and Riley 2026, footnotes 19, 22 and 28, and Online Supplement Table OS3. ↩︎
  48. Raising The Village’s projected cost in Uganda, about $32 per person, is about $161 per household of 4.98 people, against a median of about $520 for graduation programs started since 2016 (January 2026 dollars; Carter et al.’s 2020 dollars × 1.25). Carter et al. 2026, Data Appendix Tables 1 and 3; Raising The Village, answers to our questions, October 2026; our calculation. ↩︎
  49. Mahmud and Riley 2026, section 1, Table A2 and Online Supplement Table B2. ↩︎
  50. Mahmud and Riley 2026, footnote 7 and section 2.2. ↩︎
  51. In Ethiopia, a lighter program without intensive coaching had limited effects (Leight et al. 2026); in Ghana, the asset alone did not produce similar gains, and savings support alone gave only short-lived gains (Banerjee et al. 2022). ↩︎
  52. Raising The Village, answers to our questions, October 2026. ↩︎
  53. Raising The Village Impact Report 2025, pages 11 and 127 to 130; answers to our questions, October 2026. ↩︎
  54. Conversations with Raising The Village, October 2026. ↩︎
  55. Mahmud and Riley 2026, section 2.2. ↩︎
  56. Mahmud and Riley 2026, sections 2.2 and 5.1. ↩︎
  57. Carter et al. 2026, section 4.1.1: +11% on average, against +10% after the harvest in the trial (Mahmud and Riley 2026, Table A25). ↩︎
  58. Raising The Village Impact Report 2025, pages 33 and 43. ↩︎
  59. GiveWell uses the same metric for economic impact and applies a 4% time discount rate as well. ↩︎
  60. Mahmud and Riley 2026, Table 2, Table A20 and Online Supplement Table OS16; pages 26 and 27. Income here covers crop and livestock sales, business profits, wages and rent. The value of food a household grows and eats counts only in consumption. ↩︎
  61. GiveWell 2025, cost-effectiveness analyses, footnote 1. ↩︎
  62. Mahmud and Riley 2026, Table 2 and Online Supplement Table OS26. ↩︎
  63. Morduch 1995. ↩︎
  64. Carter et al. 2026, Figure 2, with exact values provided by the authors (November 2026): the trial’s post-harvest effect shrinks from $49.95 to $49.03 (2020 PPP) a month, a reduction of about 1.8%. ↩︎
  65. Mahmud and Riley 2026, section 4.4 and footnotes 39 and 40. ↩︎
  66. Mahmud and Riley 2026, section 2.1 and Table A2; Raising The Village Impact Report 2025, page 29 (our comparison after currency conversion). ↩︎
  67. Mahmud and Riley 2026, page 1, pages 17 to 20 and Tables A3, A5 and 2; AEA RCT Registry, AEARCTR-0006415. ↩︎
  68. Mahmud and Riley 2026, Table 2 and Online Supplement Table OS21. ↩︎
  69. Raising The Village Impact Report 2025, page 33 (2020 cohort). ↩︎
  70. Mahmud and Riley 2026, section 5.1 and Table A9. ↩︎
  71. McKenzie and Paffhausen 2019: across more than 14,000 firms in twelve developing countries, small firms close at about 8% a year. ↩︎
  72. Raising The Village Impact Report 2025, page 15: poverty probability (Grameen PPI) of peer households in the 2020 cohort, baseline to month 60. ↩︎
  73. GiveWell 2024, back-of-the-envelope cost-effectiveness estimate for Raising The Village. ↩︎
  74. Barker et al. 2024; Carter et al. 2026, section 4.4.2. The year-seven consumption effect in Ethiopia is significant at the 5% level, but not after the authors’ adjustment for testing multiple outcomes. The effect on assets was still significant; the effects on income and food security were not (Barker et al. 2024, Online Appendix Table A4). ↩︎
  75. Raising The Village, answers to our questions, October 2026. ↩︎
  76. Raising The Village Impact Report 2025, page 27; Raising The Village, answers to our questions, October 2026. ↩︎
  77. Raising The Village, answers to our questions, October 2026. ↩︎
  78. Mahmud and Riley 2026, section 5.2 and Table A23. ↩︎
  79. Raising The Village Impact Report 2025, page 116. ↩︎
  80. Raising The Village, answers to our questions, October 2026. ↩︎
  81. Raising The Village, answers to our questions, October 2026. ↩︎
  82. Raising The Village, answers to our questions, October 2026. ↩︎
  83. Mahmud and Riley 2026, Table A2 and Online Supplement B.1, and call with Raising The Village, October 2026; answers to our questions, November 2026. ↩︎
  84. Raising The Village Impact Report 2025, page 27: annual household value (income, production and livestock assets) above peer villages at month 24, $327 for the 2020 cohort vs. $402 to $467 for 2021–23. ↩︎
  85. Raising The Village, answers to our questions, October 2026. ↩︎
  86. Raising The Village, email to us, November 2026. ↩︎
  87. Raising The Village, answers to our questions, October 2026; our comparison with Uganda cohorts at month 12. ↩︎
  88. Raising The Village, answers to our questions, October 2026 (first Rwanda cohort at 24 months: income and production $288 above peer households, cumulatively; total program value $365 against a cost of $251, a return of 1.45x); Raising The Village Impact Report 2025, pages 29 and 43 (Uganda cohorts 2020 to 2023 at 24 months: returns of 5x to 7x). ↩︎
  89. Farming households in Rwanda’s Western Province, where Raising The Village works, cultivate about 0.3 hectares on average (all parcels operated, owned or rented), against about 0.45 hectares planted per season by households in the trial’s comparison villages (National Institute of Statistics of Rwanda, Agricultural Household Survey 2024, Table 14; Raising The Village Impact Report 2025, page 24; Mahmud and Riley 2026, Table A16). That may leave less room to expand the area planted, which treated households in the trial increased by 15 to 18% per season (same table). Raising The Village also expects government work programs and household land size to affect results in Rwanda (answers to our questions, October 2026). ↩︎
  90. Mahmud and Riley 2026, sections 2.3 and 6.2. ↩︎
  91. Mahmud and Riley 2026, section 6.2 and Table A31. The total effect including spillovers is significant (consumption at 5%, income at 1%); the spillover part is not tested separately. The authors call it suggestive: the trial was not designed to test spillovers. ↩︎
  92. GiveWell 2024, GiveDirectly cost-effectiveness analysis: 59% to 68% of recipient benefits by country. ↩︎
  93. GiveWell 2024; Egger et al. 2022. ↩︎
  94. Mahmud and Riley 2026, section 2.2 and Table A2. ↩︎
  95. Mahmud and Riley 2026, sections 6.1 and 6.2, Tables A30 and A31, and Online Supplement Tables OS38 and OS41. ↩︎
  96. Mahmud and Riley 2026, section 6.1. ↩︎
  97. Mahmud and Riley 2026, section 5.3. ↩︎
  98. GiveWell credits cash with lower child mortality through more food, medicine and health-related home improvements (GiveWell 2024). For Raising The Village, mortality reduction due to less hunger is plausible (Mahmud and Riley 2026, Online Supplement Table OS28). ↩︎
  99. Mahmud and Riley 2026, Online Supplement Table OS29. ↩︎
  100. Mahmud and Riley 2026, Online Supplement Table OS24. ↩︎
  101. Mahmud and Riley 2026, Table A27. ↩︎
  102. Raising The Village expansion budget, October 2026: average of the 2027 and 2028 cost per participant (see CEA, Cost tab). Raising The Village marks the 2027 figures as placeholders, to be updated at the end of November 2026. ↩︎
  103. Raising The Village, answers to our questions, October 2026, and expansion budget, October 2026. ↩︎
  104. Raising The Village, answers to our questions, October and November 2026; see CEA, Government investment tab. ↩︎
  105. GiveWell 2024, GiveDirectly cost-effectiveness analysis, November 2024. ↩︎
  106. Egger et al. 2022, footnote 2 and page 2611; Mahmud and Riley 2026, sections 2.1 and 2.3. See CEA, tab “Cash comparison”. ↩︎
  107. Raising The Village expansion budget, October 2026; answers to our questions, October 2026. ↩︎
  108. Raising The Village, answers to our questions, October 2026. ↩︎
  109. Raising The Village expansion budget, October 2026; answers to our questions, October 2026. ↩︎
  110. Livelihood Impact Fund, email to us, November 2026. ↩︎
  111. Raising The Village, answers to our questions, November 2026; expansion budget, October 2026. ↩︎
  112. Conversations with Raising The Village; Raising The Village, answers to our questions, October 2026, and email to us, November 2026. ↩︎
  113. Raising The Village, answers to our questions, October 2026. ↩︎
  114. Mahmud and Riley 2026, Table A2 and Appendix B.1. ↩︎
  115. Raising The Village, answers to our questions, October 2026. In our CEA, we subtract the value the government’s contribution would have had elsewhere, and reflect less certain implementation fidelity in our applicability estimates. ↩︎
  116. Raising The Village Inc., audited financial statements, 2021 to 2025 (CAD). All audited statements since 2016: raisingthevillage.org/reports. ↩︎
  117. Raising The Village Impact Report 2025, pages 25 to 30 and 129 to 130. ↩︎

Sources

Back to top ↑