One Acre Fund Tree Program
Summary
- What is the program? The agricultural sector accounts for a large share of employment in low-income countries. One Acre Fund runs a tree program where they provide seedlings and training to small-scale farmers in sub-Saharan Africa. This program aims to generate economic returns for farmers who plant, prune, and harvest fast-growing and maturing tree species over an extended period.
- What is its evidence of effectiveness? There is evidence from one randomized controlled trial in Kenya that One Acre Fund’s tree program increases the number of trees planted after one year. There is no direct causal evidence that additional trees increase farmers’ profits and household income over the long term. To generate our best guess on the effect of the tree program on farmers’ incomes, we rely on projections from One Acre Fund, which extrapolate increases in the number of trees planted to long-term effect on income. Given the lack of high-quality direct evidence on farmers’ incomes, we are highly uncertain about the evidence for effectiveness. However, One Acre Fund has shared survey evidence suggesting that farmers commonly use and sell wood products, making it appear plausible that increasing the number of trees grown could generate additional income.
- How cost-effective is it? Our best guess is that the cost-effectiveness of One Acre Fund’s tree program in most contexts is slightly below the level of cost-effectiveness of programs to which we generally expect to direct marginal donations.1 However, it is close enough to our threshold that our bottom line could change if we receive positive updates to some assumptions used. The higher estimated cost-effectiveness in some contexts is driven by low baseline household consumption levels, low programmatic costs, and a small income stream that we model as persisting for 10 to 15 years. Our cost-effectiveness estimate of this intervention relies on several parameters about which we are highly uncertain, including the costs of the program at scale, the quantity of wood products used over time and income received by farmers for those products, and the number of surviving trees per farmer over time.
- Does it have room for more funding? Additional funding would likely be used to maintain One Acre Fund’s programming in the geographic areas where it currently focuses, as well as to expand the program to new regions and countries.
- Bottom line: We plan to continue investigating this opportunity and may consider funding an evaluation or data gathering of the program to help improve our understanding of the outcomes of One Acre Fund’s tree program.
Published: December 2022
What is the problem?
In low-income countries, employment is concentrated in the agricultural sector,2 and the poorest households are more likely than others to work in agriculture, according to data compiled and reported by the World Bank.3 One survey conducted across five countries in Africa indicated that a majority of small-scale farmers do not have trees on their land.4 Of farmers that do grow trees for productive purposes, trees used for timber appear to be less prevalent than other types of trees.5 Timber trees may generate income for households who can sell timber products.6 Tree planting may therefore have the potential to raise incomes among poor households.
What is the program?
One Acre Fund’s tree program aims to encourage farmers to grow more trees by providing seedlings and training on growing practices.7 The goal of their tree program is to generate economic returns for farmers who plant, prune, and harvest trees over an extended period.8
One Acre Fund currently operates in nine countries (Burundi, Ethiopia, Kenya, Malawi, Nigeria, Rwanda, Tanzania, Uganda, and Zambia) and reaches more than one million farmers.9
The bundle of services provided varies from country to country.10 One Acre Fund has introduced the tree program in all nine countries, often as an addition to their standard bundle of agricultural inputs (although in some contexts it now acts as a standalone program).11 Beginning in 2022, One Acre Fund’s preferred model is to provide seedlings (rather than seeds) to farmers, who will then receive training to transplant them into the ground.12
Does the program have strong evidence of effectiveness?
We are moderately confident that providing seedlings and training to farmers increases the number of trees grown by farmers in the short-term. This is based on one randomized controlled trial (RCT) that has estimated the impact of One Acre Fund’s tree program on the number of trees grown by adopter farmers, as well as internal monitoring data and modeling shared with us by One Acre Fund.
We do not have direct evidence of the effect of additional trees grown on household income or consumption, which is the primary outcome we model in our cost-effectiveness analysis. To generate our best guess on the effect of the tree program on farmers’ incomes, we rely on projections from One Acre Fund.
Do tree kits cause farmers to have more trees?
We identified one RCT (Laterite 2021) that directly measures the impact of One Acre Fund’s tree program on the number of trees planted. The study was conducted in two counties of Kenya with a sample of 1,852 farmers.13 Farmers in the treatment group paid an estimated 555 KES (around $5 USD)14 for grevillea tree seeds and other inputs, plus training for planting and maintaining the trees.15 Laterite assessed the effects of the program on the number of trees (both grevillea and other types) planted using baseline and midline data.16 This trial found that treatment caused an increase in the number of grevillea trees planted in the last 12 months of about 7.5 trees.17
This is broadly similar to findings from One Acre Fund’s monitoring data. One Acre Fund routinely tries to estimate the number of trees that are attributable to their program through internal monitoring surveys. They estimate the number of new trees planted that are attributable to their program by running annual “adoption and tree planting surveys.” They also use “survival surveys” to measure the number of seedlings that are still alive one year after planting.18 The number of (planted and surviving) seedlings grown by One Acre Fund farmers is then compared to those of a comparator group of farmers.19 Across five different country programs (including both seed kit and seedlings distribution), One Acre Fund reports that the number of incremental trees surviving per adopter ranges from 4.7 to 21.6.20
We guess these estimates are at least partially over-estimated because they may not fully capture the possibility that farmers may decide to plant fewer other types of trees. Laterite 2021 found evidence that is suggestive of some substitution away from other trees within the timber category toward the One Acre Fund-supported species.21 This implies that the program’s impact on the overall number of timber trees planted could be lower than the species-specific estimate suggests.
Do additional trees lead to increases in long-term income?
We are not aware of any evidence that directly measures the income or consumption effects of One Acre Fund’s tree program. However, we think it is plausible that additional trees would lead to a long-term income stream, given survey evidence showing farmers’ expected use and sale of wood products. To estimate income and consumption effects, we rely on projections provided by One Acre Fund. To estimate income effects, One Acre Fund estimates the following parameters:
- Revenue generated by trees, which takes as inputs:
- The number of trees grown attributable to the program (“incremental seedlings”)
- The market value of trees and wood products at different ages or for different uses (e.g., firewood or bean poles)
- Farmer use of trees (i.e., what percent of trees are pruned or harvested for various purposes).
- Costs of the tree program,22
which takes as inputs:
- Input costs (i.e., farmer contribution to seedlings or seed kits)
- Farmer time costs and land opportunity costs
- Time horizons (the number of years that benefits accrue)
One Acre Fund estimates these parameters using the results from several surveys. As noted above, “survival surveys” are used to estimate the number of incremental seedlings that were grown and remain alive after one year. One Acre Fund models revenue from trees with “ongoing use surveys” of farmers to understand how they use tree products and measure earnings from any sales.23 They use “market checks”24 and “tree trader surveys”25 to establish the price of tree products. They then estimate the number of tree products produced by each tree each year.26 Input costs are generally zero, aside from some specific scenarios.
Key uncertainties
The modeling approach shared by One Acre Fund appears logical to us, but relies on a number of important assumptions. We think that the following areas are the most important to understand with further investigation or data gathering.
- Defining the treatment effect: We are uncertain about the likelihood that differences in farmer characteristics (for example, secure land rights), rather than the program, could be driving differences in treatment effect. One Acre Fund reports that they have taken steps to mitigate these concerns.27
- The treatment effects for adopters from the Laterite 2021 RCT are larger than One Acre Fund’s internal monitoring indicates, which is likely partially due to context-specific characteristics.28 Thus, we aren’t able to use the RCT results to benchmark against One Acre Fund’s internal monitoring.
- Substitution effects with tree planting: Different types of trees have different uses and thus different potential impacts on income. As discussed above, there is some evidence that One Acre Fund’s program causes substitution from other timber species to the One Acre Fund-supported grevillea tree. This could imply that using the grevillea-specific treatment effects overestimates income effects, particularly for “whole tree” uses (like furniture and construction), which we guess are only carried out with timber trees. On the other hand, there is also some evidence that the program causes positive spillovers to other tree planting.29 Some of the uses modeled (e.g., bean poles and firewood/charcoal) may be accomplished by non-timber trees. If this is true, using the grevillea-specific treatment effect may underestimate the benefits of the program from these uses.
- Opportunity costs for farmers: It is possible that farmers could also substitute away from other productive activities due to their participation in the tree program in ways that are not captured by the modeled farmer labor and land opportunity costs. One Acre Fund has told us in conversation that this is unlikely because they believe that farmers generally have excess capacity.30 Laterite 2021 found a decrease in the percentage of farmers planting vegetables, which could reduce consumption or income.31 But Laterite 2021 reports no evidence of an effect of the seedlings program on farmers planting other plants, having grazing land, or having livestock.32
- Survival of seedlings: One Acre Fund and Laterite 2021 both implicitly assume that incremental seedlings that survive at least one year after planting will continue to grow so that farmers can make money from them in future years.33 One Acre Fund has noted that most tree deaths occur within 18 months of planting and that they would expect subsequent tree death to be equal between treatment and control groups.34 However, if the rate of tree death is equal between treatment and control groups on an ongoing basis, it will reduce the treatment effect over time.35 If the rate of tree death is higher in control groups (which could be the case if One Acre Fund’s training for farmers helps them keep seedlings alive), it could increase the treatment effect over time.36
- Value of revenue from ongoing use: Revenue from before trees are harvested (such as that from pruning) constitutes a substantial portion of the modeled benefits.37 There is a wide range in the estimated average revenue from these commodities across countries, and we are uncertain about how confident we should be in these estimates.
Other evidence from the literature
We are not aware of additional RCTs on the effect of agroforestry interventions on economic outcomes. We did a shallow review of observational evidence but do not put weight on this evidence (see footnote for details).38
How cost-effective is the program?
We conducted a preliminary cost-effectiveness analysis of One Acre Fund’s tree program. As of November 2022, we estimate that in most contexts this program is slightly below the level of cost-effectiveness of programs we expect to direct marginal donations to.39 However, it is close enough to our threshold that our bottom line could change if we get more information.40
Note that our cost-effectiveness analyses are simplified models that do not take into account a number of factors. There are limitations to this kind of cost-effectiveness analysis, and we believe that cost-effectiveness estimates such as these should not be taken literally due to the significant uncertainty around them. We provide these estimates (a) for comparative purposes and (b) because working on them helps us ensure that we are thinking through as many of the relevant issues as possible.
This cost-effectiveness analysis is at an early stage, and we think it’s likely that our bottom line cost-effectiveness estimate will change with further review.
Providing farmers with seedlings and training is a low-cost intervention that may lead to long-term increases in agricultural income for low-income households. This leads to high cost-effectiveness in some contexts.
A sketch of the cost-effectiveness model (based on estimates for Rwanda) is below:
- Cost per adopter: One Acre Fund thinks the program would cost $0.12 per tree (in labor and land opportunity costs) for farmers that adopt seedlings. The program would cost One Acre Fund around $1.83 per adopter.41
- Increase in number of trees per adopter: One Acre Fund estimates that each farmer that adopts will plant and cultivate six additional trees after accounting for substitution.42
- Income per tree: One Acre Fund estimates that the trees would yield an average annual profit of $0.76 across 15 years of potential benefits.43 We use a discount rate of 6.5%.44
- Downward adjustments: We apply a downward adjustment of 48% due to concerns like relying on projections rather than direct evidence.45
- Baseline consumption: We estimate that baseline household consumption is $1,100, and households have around four members at the beginning of the program (which we have assumed will decline over the 15-year period).46
- Cost-effectiveness: Our best guess is that the cost-effectiveness of One Acre Fund’s tree program in Rwanda is around 10 times greater than that of unconditional cash transfers.
However, we have high uncertainty about the costs of the program at scale, the quantity of wood products sold and income received by farmers for those products, the number of surviving trees (including One Acre Fund-supported species and trees overall) per program adopter, and whether seedlings induced by One Acre Fund's programs survive at different rates than those planted by farmers without the program.
- Costs. We have moderate uncertainty about the cost estimates that we use. We used estimates provided to us by One Acre Fund, which differ substantially across contexts, and which may include or exclude some organizational costs that are inconsistent with how we typically model programs. We believe they are the most informed costs at our disposal at this stage of our investigation, and expect that we will refine these if we investigate a specific opportunity.
- Number of incremental trees. We are highly uncertain about the number of One Acre Fund-supported tree species and the number of total trees that are likely to survive per program adopter. The estimated effect size among adopters in Laterite 2021 was higher than the effect size indicated in One Acre Fund’s monitoring.47 However, we are unsure how representative the population studied in the Laterite 2021 was of One Acre Fund’s typical programs, so do not feel confident using its findings to benchmark One Acre Fund’s monitoring against. In addition, the estimated number of incremental surviving trees caused by One Acre Fund’s tree program differs significantly between countries.48 We do not fully understand the drivers of these differences, nor whether the wide range of values is plausible.
- Tree survival rate for program adopters. We are also highly uncertain about the long-term tree survival of program adopters compared with counterfactual farmers. We have assumed in our model that incremental trees that survive at least one year after planting will go on to survive for around 15 years. Although we have applied a downward adjustment to account for the possibility that some of these trees would later die, this is highly subjective. We are uncertain whether the survival rate of One Acre Fund program participants’ trees may be higher than that of the counterfactual farmer, and how this could differ by tree species.
- Quantity of wood product used and income received. We also have high uncertainty about how many ongoing wood products farmers are likely to sell, as well as the income that farmers are likely to receive from sales of different products. The income derived from ongoing use of trees (such as firewood and bean poles) accounts for a substantial portion of the modeled benefits.49 However, we are unsure about whether these estimates are reasonable, particularly because there is a wide range of estimated values across countries. We would work to refine this key input if we investigate a specific grant opportunity.
Is there room for more funding?
In conversations with One Acre Fund, they have reported that they will have around $20 million in annual room for more funding by 2025, and $50 million in annual room for more funding by 2027. This is because around 80% of their current tree program funding is from a one-time campaign that won’t renew past 2024.50 We have not vetted this claim, and intend to investigate this further in any grant investigation.
Of the countries that we believe to be the most cost-effective (Burundi, Rwanda and Kenya), One Acre Fund’s projections indicate that there will be around $25 million in total room for more funding over the next three years, and $64 million in total room for more funding over the next six years in a less certain scenario.51
We expect that the cost-effective room for more funding of One Acre Fund’s tree program could increase if (a) an additional trial or data gathering allowed us to have more confidence in their internal monitoring efforts, thereby increasing our confidence in the cost-effectiveness of new contexts; (b) One Acre Fund were interested in testing different model variations; or (c) our best guess of farmer costs was updated.
Key questions for further investigation
The primary focus of further investigation on One Acre Fund’s tree program will be to more precisely understand the impact of providing seedlings to farmers on the number of mature trees available for income generation, and the income derived from said trees.
Questions we will aim to answer by speaking with One Acre Fund include:
- Can we recommend funding any data-collection activities (such as surveys or studies) that would provide external validation to One Acre Fund’s monitoring data? This could give us more confidence in several key model parameters, including the estimates of the number of incremental trees grown, the presence (or lack thereof) of a substitution effect, the value of wood products, farmer use of timber trees, and the opportunity cost for farmers.
- How do modeling assumptions or methods differ across countries, and is this driving some of the bottom-line results?
- Do seedlings of One Acre Fund participants survive (after the one-year mark) at a higher rate than non-One Acre Fund participants’ trees?
Other questions we plan to ask as part of further investigation include the following:
- Is household income likely to grow at a higher or lower rate than wood commodity prices?
- What is the likelihood that trees survive around 15 years, given that they survive at least one year after planting?
- What is the probability of positive or negative spillover effects if a program was scaled up?
Our process
To date, our investigation process has consisted of:
- Several conversations with One Acre Fund.52
- Reviewing the Laterite 2021 RCT on One Acre Fund’s tree program in Kenya.
- Conducting a shallow search for RCTs on other interventions related to agroforestry.
- Reviewing documents that One Acre Fund sent us after our conversations and in response to our queries.
Sources
- 1
As of the writing of this page (November 2022), GiveWell expects to fund opportunities that are ten or more times as cost-effective as unconditional cash transfer programs, such as that of GiveDirectly.
- 2
In 2019, 59% of employment in low-income countries was in agriculture, based on International Labor Organization estimates. World Bank, "Employment in agriculture as percentage of total employment in low-income countries, International Labour Organization, ILOSTAT database," 2021.
- 3
See Castañeda et al. 2016 (working paper), p. 13, figure 6. We have not vetted the data presented in this figure or in the analysis as a whole.
- 4
“The data for this chapter pertain to five of the six countries originally covered under the Living Standards Measurement Study–Integrated Surveys on Agriculture (LSMS-ISA) initiative. The countries are Ethiopia, Malawi, Nigeria, Tanzania, and Uganda. The first challenge is to define what plants to consider as trees. Based on an initial classification using biological definitions, the study carried out focus group interviews with experts to refine and validate the definition (box 13.1). This resulted in a crop classification that combined the biological description of each crop/tree and its economic role on the farm. The five categories are fruit trees; cash crop trees; timber and fuelwood trees; plants, herbs, grasses, and roots; and those not identified. The study focused on the first three. . . . Across the study countries, about one-third of African farms report growing trees, often in the proximity of existing forests. Nonetheless, the stock of trees on farms varies substantially by country and category of tree (table 13.1). On the one hand, there are the cases of Nigeria and Malawi, where the prevalence of trees on farms is relatively low. In these countries, only 16 and 23 percent (respectively) of landholders report having trees on their farmland. On the other hand, in Tanzania, Ethiopia, and Uganda, trees on farms are considerably more prevalent (54, 38, and 30 percent of landholders, respectively).” Miller, Munoz-Mora and Christiaensen 2018, ch. 13 in Christiaensen and Demery 2018, pp. 116-17.
- 5
“The prevalence of trees also varies across countries by tree type. In Tanzania, for example, fruit trees are especially widespread, with 45 percent of the landholders having at least one plot with fruit trees. Trees for timber and fuelwood were reported among 18 percent of smallholders in Tanzania, but very little elsewhere. In contrast, in Ethiopia, less than 4 percent of smallholders report having trees for timber and fuelwood (likely an underestimate, given that eucalyptus trees were not properly captured in the questionnaires). But Ethiopia has the highest proportion of farms with tree cash crops (32 percent), especially coffee.” Miller, Munoz-Mora and Christiaensen 2018, ch. 13 in Christiaensen and Demery 2018, pp. 117-18.
- 6
“Trees can enable economic gains through sale and consumption of timber and non-timber products, as well as the environmental services they provide." Miller et al. 2020, p. 2.
- 7
- “Tree-planting programs have been central to our impact strategy, and in 2021 we expanded our investment in them dramatically by supporting over 2 million farmers to plant more than 40 million trees, of which we project 20 million will reach maturity." One Acre Fund, Cultivating New Frontiers: 2021 Annual Report, p. 10.
- "One Acre Fund’s team of field officers provides expert, tailored training directly to farmers throughout the year. This includes training on . . . tree planting." One Acre Fund, "Training"
- 8
“Our research found timber trees in particular provide both a strong financial return to farmers (driven by low input and labor costs; low susceptibility to disease and theft; and high market demand for tree products) and strong environmental benefits (e.g., erosion control, soil enhancement, and carbon sequestration)." One Acre Fund, Summary of 2019-2021 tree seedling randomized trial, p. 2.
- 9
“After first starting in Kenya in 2006, we've grown to now serve over 1.4 million hardworking smallholder farmers across nine countries in Eastern and Southern Africa.” One Acre Fund, "Countries we serve". The web page also lists the countries where One Acre Fund works.
- 10
“OAF works in six countries and provides a unique bundle of services in each country.” GiveWell's non-verbatim summary of a conversation with One Acre Fund, March 13, 2018, p. 1. (Note that the number of countries where One Acre Fund works has grown since this conversation.)
- 11
- “In our full-service program, we work directly with farmers throughout the year, which generally includes:
- Providing quality farm products on credit, which farmers repay over the full growing season.
- Delivering these products within walking distance of farmers’ homes, in time for planting.
- Training farmers on effective agricultural practices and how to sell any harvest surplus."
One Acre Fund, "Our model".
- “In 2010, 1AF introduced a Tree Program to their core loan package with the aim of increasing the client farmers’ tree assets as a source of extra income from their sale. The Tree Program contains grevillea seeds, planting fertilizer, large planting the seeds, smaller planting bags for raising individual seedlings once they are big enough, and a set of trainings specifically on tree planting and maintenance.” One Acre Fund, Summary of 2019-2021 tree seedling randomized trial.
- One Acre Fund, comments on a draft of this page, September 23, 2022 (unpublished).
- “In our full-service program, we work directly with farmers throughout the year, which generally includes:
- 12
- “Additionally, in focus group discussions, both farmers and 1AF field officers preferred a program model that offered tree seedlings (rather than tree seeds), as the former are easier to plant and more likely to survive. After considering all facets (farmer demand, impact, cost, etc.), 1AF has decided to shift efforts in Kenya towards a model in which 1AF grows (or contracts local entrepreneurs to grow) tree seedlings, and delivers those seedlings directly to farmers, which should increase both the adoption rate among farmers and overall number of trees planted per adopter, albeit with higher program cost.”
- "Finally, most farmers reported attending only one or two Tree Program trainings from four delivered by Field Officers; improving training attendance is another opportunity for 1AF."
One Acre Fund, Summary of 2019-2021 tree seedling randomized trial
- 13
- “At baseline, 226 training groups in 37 1AF training sites were assigned to the treatment and control arms and 1,852 farmers from these groups were sampled and interviewed.” Laterite, 2019-2021 Tree Seedling RCT: endline report, 2021, p. 8.
- "Laterite was contracted by 1AF in 2018 to undertake a randomized control trial (RCT) to evaluate the impact of the Tree Program in the Kericho and Uasin Gishu counties of Kenya, where 1AF expanded their program in the 2019 long rains season." Laterite, 2019-2021 Tree Seedling RCT: endline report, 2021, p. 1
- 14
“Farmers in the trial districts, i.e., treatment farmers in our sample, were offered the tree kit for a total of KES 505 [sic]. This price consisted of a bag with 10g of grevillea seeds, two bags for planting, 100 planting sockets, and tree fertilizer (priced together at KES 200), training for the tree kit (KES 200), loan interest (KES 130), transportation and quality assurance (KES 25).” Laterite, 2019-2021 Tree Seedling RCT: endline report, 2021, p. 38. We believe that “505” was a typo because the listed costs add up to 555 KES (200 + 200 + 130 + 25).
- 15
“The treatment farmers each received a tree kit, which contained 10 grams of grevillea tree seeds, a planting bag, sockets for seedlings, tree fertilizer, and a set of trainings specifically on planting and maintaining trees.” One Acre Fund, Summary of 2019-2021 tree seedling randomized trial, p. 3
- 16
Baseline (June 2019) and midline (September 2020) reports were provided to GiveWell. See Laterite, Randomized control trial of the Tree Program in 2019 expansion areas of Kenya: Baseline findings report, 2019 and Laterite, One Acre Fund Tree Program: Midline evaluation report, 2020.
- 17
"One-year after program participation, 1AF farmers were growing 7.5 additional grevillea trees (relative to control farmers). This number increases to 31.2 additional grevillea trees when restricting the analysis to ‘adopters’ (the subset of farmers who successfully germinated seeds and transplanted seedlings into the ground)." One Acre Fund, Summary of 2019-2021 tree seedling randomized trial, p. 1.
- 18
“Tree Quantity
We want to know how many trees a participating farmer has planted due to the program. Survey teams count the number of trees that 1AF and non1AF (comparison) plant. In order to mitigate the challenge of seasonality (see “Important Definitions”), 1AF enumerators follow-up after the season for tree planting has closed but before the dry season has started to count the number of trees that tree planters from both groups have planted. Several months later, after a dry season has completed and the rains have begun, enumerators re-visit these farms and count the total number of seedlings alive. The total number of incremental trees is then calculated by removing the number of surviving seedlings that comparison farmers have from the total surviving seedlings that 1AF farmers have. . . .
Adoption and tree planting survey. This is a field-based survey in which we observe how many trees were planted that season shortly after trees were distributed among 1AF and non1AF farmers. This allows us to estimate the number of trees planted, and the percent of farmers who were given trees who actually adopted them (i.e. went home and planted seedlings), and the incremental trees planted for 1AF and comparison farmers.
Survival survey. This survey is conducted after 12 months or more after the tree distribution. The primary purpose here is to get a tree survival rate after tree planting and a final incremental tree count. We generally find that many trees do not survive the first year after planting (e.g. due to flooding, drought, pests, a cow may eat it etc…), and so it is important to account for this loss. We estimate survival rates of around 35%-65% depending on the country, season and tree species.” One Acre Fund, Tree impact estimation methodology, 2022, pp. 3-4. - 19
The exact definition of this comparator group varies with context and can include neighboring farmers who can’t participate in the trees program because they’re not participants in One Acre Fund core staple crop programs, cluster RCTs or matching studies. One Acre Fund reports that the “neighboring non-participating farmer” methods “produces outcomes that are highly similar to the RCTs.”
“Counterfactual Selection
In our tree impact assessments, we compare the tree planting behavior of 1AF tree adopters to counterfactual farmers so that we only take credit for the impact of incremental tree planting. The methodology on how the counterfactual farmers are selected depends on how the tree program is spread out on the ground and who is eligible to participate in the program.- Neighboring non-participating farmers: In instances, where the tree program is only open to clients of the 1AF Core staple crop program, we select neighboring non-participating farmers to serve as the counterfactuals. To select the counterfactual farmers, we typically ask the 1AF farmers to recommend their neighbors who will be interested in the program in the next season. Another factor mitigating the selection bias is that the 1AF farmers have enrolled for an agricultural program and not the tree program. They receive the trees just as a top up to their agricultural package. Here, the neighboring farmers face the same agroecological and tree growing conditions as the 1AF tree farmers, but do not have access to the tree program because they are not clients of the 1AF Core agricultural program. In comparisons between 1AF internal and independent external evaluations of our core program, we have generally found that this quasi-experimental method produces outcomes that are highly similar to the RCTs.
Clustered studies: In several countries, the 1AF tree program is open to any farmer residing in the areas where 1AF is operating. As a result, the ‘neighboring non-participating farmers’ approach described above is no longer valid since those non participating farmers are eligible to participate in the tree program. In such cases, we attempt to run clustered randomized studies or cluster-matching studies (e.g. we have done this in Ethiopia and Malawi). Running such clustered studies are much more resource intensive (due to sample size requirements) but they mitigate several concerns about selection bias since the counterfactual farmers have not been exposed to the 1AF tree program in any way.” One Acre Fund, Tree impact estimation methodology, 2022, p. 5.
- 20
- Burundi (seedlings): 13.1, Ethiopia (seedlings): 21.6, Kenya (seedkit): 18.3 , Malawi (seedlings): 4.7, Rwanda (seedlings) 6.8. One Acre Fund, comments on a draft of our cost-effectiveness analysis of One Acre Fund's tree program, October 7, 2022.
- According to One Acre Fund, this is largely driven by differences in distribution amounts and tree survival rates: “This range is primarily driven by the number of trees distributed in each country program which varies by context, and secondarily because of different survival rates in different countries.” One Acre Fund, comments on a draft of this page, September 23, 2022 (unpublished).
- 21
Laterite 2021 found that treatment caused an increase in the number of grevillea trees planted in the last 12 months of about 7.5 trees. The intention-to-treat (ITT) effect for the number of all trees planted was 13.8 trees, which is larger than the grevillea-specific estimate. This suggests that One Acre Fund’s tree program has positive spillovers, encouraging farmers to plant other trees. However, the ITT for timber trees in particular is 3.8 trees, which is lower than the grevillea-specific estimate. See Laterite, One Acre Fund Tree Program: Midline evaluation report, 2020, p. 23, Table 10, and p. 26, Table 13.
- 22
See our cost-effectiveness analysis of One Acre Fund's tree program, “Summary Tables” sheet.
- 23
"Ongoing use survey. These surveys are administered 3-4 times per year with randomly selected farmers with existing mature trees of the 1AF provided species to better understand their tree uses (e.g. for timber, firewood, fence poles etc…). We conduct these surveys multiple times throughout the year in order to capture any seasonal effects. We also ask farmers to let us know the price per unit they received if they sold any tree products." One Acre Fund, Tree impact estimation methodology, 2022, p. 4.
- 24
"Market checks: We conduct market checks to get the price of tree products when they are sold in a market. This helps us understand the opportunity cost of the tree products when farmers use their own trees instead of having to buy them in the markets." One Acre Fund, Tree impact estimation methodology, 2022, p. 4.
- 25
"Tree trader surveys. These surveys are administered to tree traders in the areas of program operation. In these surveys we ask traders what they pay for various tree species at various ages." One Acre Fund, Tree impact estimation methodology, 2022, p. 4.
- 26
"One Acre Fund surveying teams collect data on the different ways that farmers use the trees, the ages of the trees for different uses, and the value that farmers derive from the different uses. For example, we can assign a value to firewood and fencing poles and then estimate which years farmers will start using what % of their trees for those values. This is combined with cost data, which includes input, labor, and land opportunity costs." One Acre Fund, Tree impact estimation methodology, 2022, p. 1.
- 27“Counterfactual Selection
In our tree impact assessments, we compare the tree planting behavior of 1AF tree adopters to counterfactual farmers so that we only take credit for the impact of incremental tree planting. The methodology on how the counterfactual farmers are selected depends on how the tree program is spread out on the ground and who is eligible to participate in the program.- Neighboring non-participating farmers: In instances, where the tree program is only open to clients of the 1AF Core staple crop program, we select neighboring non-participating farmers to serve as the counterfactuals. To select the counterfactual farmers, we typically ask the 1AF farmers to recommend their neighbors who will be interested in the program in the next season. Another factor mitigating the selection bias is that the 1AF farmers have enrolled for an agricultural program and not the tree program. They receive the trees just as a top up to their agricultural package. Here, the neighboring farmers face the same agroecological and tree growing conditions as the 1AF tree farmers, but do not have access to the tree program because they are not clients of the 1AF Core agricultural program. In comparisons between 1AF internal and independent external evaluations of our core program, we have generally found that this quasi-experimental method produces outcomes that are highly similar to the RCTs.
- Clustered studies: In several countries, the 1AF tree program is open to any farmer residing in the areas where 1AF is operating. As a result, the ‘neighboring non-participating farmers’ approach described above is no longer valid since those non participating farmers are eligible to participate in the tree program. In such cases, we attempt to run clustered randomized studies or cluster-matching studies (e.g. we have done this in Ethiopia and Malawi). Running such clustered studies are much more resource intensive (due to sample size requirements) but they mitigate several concerns about selection bias since the counterfactual farmers have not been exposed to the 1AF tree program in any way.” One Acre Fund, Tree impact estimation methodology, 2022, p. 5.
- 28
- One Acre Fund's internal monitoring for the full Kenya program indicates 20.3 trees per adopter before accounting for substitution (Table on p. 8), whereas Laterite’s indicates 31 trees per adopter (Table on p. 9). See One Acre Fund, One Acre Fund Tree Program: Compendium of materials for GiveWell, 2022.
- "These findings [from the RCT] are substantially greater than 1AF’s internal findings for tree-planting in Kenya in 2019. 1AF believes the two districts in which the Laterite study took place are overall a more hospitable area for tree-planting (e.g., local market prices per tree are substantially higher) compared to the 40 districts in which 1AF’s internal measurement took place." One Acre Fund, Summary of 2019-2021 tree seedling randomized trial, p. 1.
- 29
Laterite 2021 found that treatment caused an increase in the number of grevillea trees planted in the last 12 months of about 7.5 trees. The intention-to-treat (ITT) effect for the number of all trees planted was 13.8 trees, which is larger than the grevillea-specific estimate. This suggests that One Acre Fund’s tree program has positive spillovers, encouraging farmers to plant other trees. However, the ITT for timber trees in particular is 3.8 trees, which is lower than the grevillea-specific estimate. See Laterite, One Acre Fund Tree Program: Midline evaluation report, 2020, p. 23, Table 10, and p. 26, Table 13.
- 30
Conversation with One Acre Fund, June 14, 2022 (unpublished).
- 31
“The results show that the program decreased the share of farmers who planted vegetables in the past 12 months by 4.8 percentage points, holding everything else equal, using a difference-in-difference model. This effect is significant at the 5% level. However, the effect is not significant on the number of different vegetable types a farmer planted.” Laterite, 2019-2021 Tree Seedling RCT: endline report, 2021, pp. 37-38.
- 32
“We find no statistically significant treatment effect on whether a farmer had planted any type of plant, had grazing land, or had livestock on their farm (Figure 10), except for vegetables. For other plants and livestock, we also do not find a statistically significant treatment effect on the number of types of plants and Tropical Livestock Units that a farmer grew or owned. The results show that the program decreased the share of farmers who planted vegetables in the past 12 months by 4.8 percentage points, holding everything else equal, using a difference-in-difference model. This effect is significant at the 5% level. However, the effect is not significant on the number of different vegetable types a farmer planted.
Overall, we conclude that the farmers do not incur a major opportunity cost in terms of land-use because of the Tree Program. The treatment effect on a higher share of farmers planting grevillea and more grevillea trees being planted on the treatment farmers’ land does not displace other types of plants or livestock, except possibly vegetables, and does not affect whether farmers have grazing land.
While the treatment effect might affect the area planted with each type of plant, qualitative insights suggest that this is not the case. Farmers and key informants argue that grevillea trees are excellent trees for intercropping and explain that grevillea trees are usually planted with other crops, including coffee, and provide shade to avoid drying.” Laterite, 2019-2021 Tree Seedling RCT: endline report, 2021, pp. 37-38. - 33
“Farmers were surveyed at baseline and 1 year after tree distribution. (A limited number were also surveyed and participated in qualitative follow-up 2 years after tree distribution). The study was originally intended to measure impact over a two-year time period; however, quantitative data collection had to be curtailed due to COVID, and the full data collection effort occurred only 1 year after tree distribution.” One Acre Fund, Summary of 2019-2021 tree seedling randomized trial, p. 3.
- 34
- “We believe these results are still highly valid, especially as the literature suggests that after 1 year, tree survival is fairly well assured (with most tree deaths taking place in the first dry season after planting) and any increases in tree death beyond the first year are likely to be roughly equal between the treatment and control groups. Put another way, the incremental number of trees per treatment farmer (relative to control farmer) was unlikely to have changed between year 1 and year 2. ” One Acre Fund, Summary of 2019-2021 tree seedling randomized trial, p. 3.
- One Acre Fund later told us that while tree survival is fairly well assured after 12 months have passed in regular rainfall markets, in drier markets it may not be assured until 18 months after planting. One Acre Fund, comments on a draft of this page, September 23, 2022 (unpublished).
- 35
For calculations, see GiveWell, One Acre Fund seedlings BOTEC, "Seedling survival adjustment" tab.
- 36The Laterite 2021 RCT found that farmers assigned to treatment were significantly more likely to know the correct soil for planting seedlings than farmers assigned to control, but it found no significant differences in knowledge of germination, watering, survival or pruning practices: “We find an intention-to-treat effect for the knowledge on using the correct soil for planting tree seeds. There was a 13.2 percentage point increase in the share of farmers that knew that they should gather soil from close to a healthy tree, use topsoil only, use a mix of topsoil and sand, or specifically not using clay soil at midline in the treatment group compared to the control group. In aggregate, the treatment also has a positive and statistically significant effect on the number of knowledge questions that the farmers answer correctly. There was no significant change in the knowledge of other best practices asked at baseline. These questions were about the time farmers should wait for a seed to germinate (two months or more), how often they should water the seeds if it does not rain (at least once a day), what they can do to maximize germination rates, and correct pruning (25% of the branches or less). The lack of effect these other practices might be explained by the low attendance rate to the tree trainings. These trainings might not have been offered at all in some places according to the One Acre Fund.”Laterite, 2019-2021 Tree Seedling RCT: endline report, 2021, p. 4.
- 37
See the average annual revenue from the sale of firewood and bean poles in comparison to the average annual revenue from whole tree sales here in GiveWell, One Acre Fund seedlings BOTEC, 2022.
- 38
We are aware of some observational evidence on similar programs. Castle et al. 2021 conducted a meta-analysis of seven quasi-experimental studies that measured the effects of agroforestry programs on income in low- and middle-income countries: "We identified eight studies that measured the effects of agroforestry on income. We conducted a meta-analysis based on seven of these included studies that measured the impacts of agroforestry on income." Castle et al. 2021, p. 22. They found a positive but relatively small effect, with a moderately high level of heterogeneity between studies (see Figure 7, p. 23). Castle et al. 2021 reports some evidence that technical assistance programs promoting fallow, intercropping, or fertilizer trees also appear to improve yields and increase income. One Acre Fund also conducted a literature review on evaluations of programs introducing trees to small-scale farmers and identified four additional studies conducted in East Africa. These studies (Nabunya 2017, Thorlakson and Neufeldt 2012, Hughes et al. 2020, and Kiyani et al. 2017) all found positive impacts of agroforestry programs on income. We have not vetted any of these non-experimental studies thoroughly.
- 39
The values provided by One Acre Fund that are used in the cost-effectiveness analysis are current as at October 2022, but are continuously being updated as One Acre Fund conducts monitoring and data collection.
- 40
- For an example of the cost-effectiveness of our recommendations, see this page. As of November 2022, we estimate that we are likely to direct funding to opportunities at least 10 times as cost-effective as unconditional cash transfers.
- Our cost-effectiveness analysis estimates that this program will range from 0.1 to 10.0 times as cost-effective as unconditional cash transfers, depending on the country modeled (and associated costs).
- See our cost-effectiveness analysis of One Acre Fund’s tree program, CEA tab, "Cost-effectiveness in multiples of cash transfers, after all adjustments" row.
- 41
See here in GiveWell, One Acre Fund seedlings BOTEC, 2022, for land and opportunity costs and here for information about costs for One Acre Fund.
- 42
- 43
See our One Acre Fund seedlings BOTEC, Rwanda tab, "Average over years 0-15." The calculated average includes 15 years of tree growth plus "year 0" in which the tree was planted.
- 44
We use a larger discount rate than is typical for GiveWell models (6.5% rather than 4%). This is to account for the greater temporal uncertainty associated with wood prices compared to long-term cognitive benefits.
- 45
See the adjustment here (100% - 48% downward adjustment = 52%).
- 46
See here in GiveWell, One Acre Fund seedlings BOTEC, 2022, for household consumption, and here for household size.
- 47
One Acre Fund's internal monitoring for the full Kenya program indicates 20.3 trees per adopter before accounting for substitution (Table on p. 8), whereas Laterite’s indicates 31 trees per adopter (Table on p. 9). See One Acre Fund, One Acre Fund Tree Program: Compendium of materials for GiveWell, 2022.
- 48
- The estimated number of incremental surviving trees caused by One Acre Fund's tree program by country: Burundi (seedlings): 13.1, Ethiopia (seedlings): 21.6, Kenya (seedkit): 18.3 , Malawi (seedlings): 4.7, Rwanda (seedlings) 6.8. One Acre Fund, comments on a draft of our cost-effectiveness analysis of One Acre Fund's tree program, October 7, 2022 (unpublished).
- According to One Acre Fund, this difference between countries is largely driven by differences in distribution amounts and tree survival rates: “This range is primarily driven by the number of trees distributed in each country program which varies by context, and secondarily because of different survival rates in different countries.” One Acre Fund, comments on a draft of this page, September 23, 2022 (unpublished).
- 49
See the average annual revenue from the sale of firewood and bean poles in comparison to the average annual revenue from whole tree sales here in GiveWell, One Acre Fund seedlings BOTEC, 2022.
- 50
One Acre Fund, One Acre Fund Tree Program: Compendium of materials for GiveWell, 2022, p. 14.
- 51
See our cost-effectiveness analysis of One Acre Fund's tree program, “Program Costs and RFMF” tab.
- 52
For example, see GiveWell's non-verbatim summary of a conversation with One Acre Fund, June 14, 2022.