Policy Research Working Paper 11084 Impacts and Spillovers of a Low-Cost Multifaceted Economic Inclusion Program in Chad Patrick Premand Pascale Schnitzer Development Impact Group & A verified reproducibility package for this paper is Social Protection and Labor Global Department available at http://reproducibility.worldbank.org, March 2025 click here for direct access. Policy Research Working Paper 11084 Abstract This study analyzes the direct effects and local spillovers of a and off-farm micro-enterprises. The intervention improved low-cost multifaceted economic inclusion program through women’s empowerment and some dimensions of social a randomized controlled trial in Chad. The intervention well-being. The findings show evidence of positive local included group savings promotion, micro-entrepreneur- spillovers, with improvements in food consumption and ship training, and a lump-sum cash grant delivered to poor economic activities among households that were not female beneficiaries of a regular cash transfer program. It assigned to the economic inclusion program in targeted was designed to address multiple constraints to productivity villages. The results are consistent with the intervention and livelihoods, but at a much lower cost (approximately broadly improving saving, sharing, and financial support $104 per household) than most stand-alone nongovern- mechanisms, as well as potential demand-side effects in the mental organization graduation pilots and government-led labor and product markets. Once spillovers are accounted economic inclusion programs. The results show substantial for, the intervention becomes cost-effective without assum- impacts on food consumption 18 months after the inter- ing that any impact persists past the follow-up survey at vention. A reallocation of labor between economic activities 18 months. is observed, along with higher revenues from agriculture This paper is a product of the Development Impact Group, Development Economics and the Social Protection and Labor Global Department. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at ppremand@worldbank.org and pschnitzer@ worldbank.org. A verified reproducibility package for this paper is available at http://reproducibility.worldbank.org, click here for direct access. RESEA CY LI R CH PO TRANSPARENT ANALYSIS S W R R E O KI P NG PA The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Impacts and Spillovers of a Low-Cost Multifaceted Economic Inclusion Program in Chad 1 Patrick Premand (DIME, World Bank) Pascale Schnitzer (World Bank) Keywords: poverty, social protection, graduation, economic inclusion, livelihoods, cash grant, social protection, spillovers, field experiment. JEL Codes: D13, D14, O12, O13, O17, I38. 1 This paper is based on a collaboration between the Chad Safety Nets Unit (CFS), which implemented the Social Safety Nets Pilot Project (PPFS) under the supervision of the Ministry of Economy and Development Planning, the World Bank West Africa Social Protection program, and DIME. We thank Dr Japhet Doudou Beindjila, Derry Gotobé Florence and all CFS staff, as well as Rony Djekombe, Surat Nsour, Mahamane Maliki, Moukenet Azoukalne, Mbaipeur Nenodji, Pantaleo Creti, and the World Bank Sahel Adaptive Social Protection Program team for a fruitful collaboration. CFS led the intervention implementation together with IHDL (Initiative Humanitaire pour le Développement Local) and AIDER (Association d’Appui aux Initiatives de Développement Rural), and with technical assistance from the World Bank (coordinated by Pascale Schnitzer) and Trickle Up (coordinated by Yéréfolo Malle). Simplex Consulting collected the baseline and follow-up survey data, with technical support from Karim Paré. Marine Colon de Franciosi and Vincent Mermet Bijon provided excellent research assistance for data analysis. Vincent Mermet Bijon prepared the replication package. We are grateful to Thomas Bossuroy, Eeshani Kandpal and Kelsey Wright for contributions at various parts of the study, as well as to Manuela Angelucci, Harounan Kazianga and a DIME reviewer for comments. The study was pre-registered in the AEA RCT registry: https://www.socialscienceregistry.org/trials/6898. The study received approval from the Innovations for Poverty Action Institutional Review Board #00006083. The computational reproducibility of the results has been verified by DIME analytics. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the World Bank and its affiliated organizations, those of the Executive Directors of the World Bank, or the governments they represent. 1. Introduction Social protection and safety net programs have spread rapidly around the world over the past 20 years. Their expansion has been informed by extensive evidence showing that cash transfers improve welfare, human capital and productive capacities (Alderman and Yemtsov, 2012, Kondylis and Loeser, 2021, Crosta et al., 2024). Policy makers are increasingly looking to complement regular cash transfers with additional interventions to further boost livelihoods and income-generating activities. For instance, there has been a recent surge in government social protection systems incorporating multifaceted economic inclusion programs (Beegle et al., 2018; Andrews et al., 2021; Arévalo-Sánchez et al., 2024). This evolution has been inspired by encouraging evidence on “cash plus” (Daidone et al., 2019; Leight et al., 2024) and graduation programs (e.g. Banerjee et al. 2015; Bandiera et al., 2017; Bedoya et al., 2019; Banerjee et al, 2021; Balboni et al., 2021). NGOs have played a key role in developing and producing early evidence on the effectiveness of multifaceted graduation programs. Government policy makers around the world then grappled with several questions as they considered how to integrate similar models into their national systems (Andrews et al., 2021; Arévalo-Sánchez et al., 2024). For instance, policy makers wondered how to optimize multifaceted interventions so that they would not be overly complex to implement. They looked for ways to lower costs to ensure scale-up would be possible without prohibitive budgets. And they asked about the aggregate effects of these policies on the local economy and beyond. These questions have fueled operational innovations and a new wave of research, such as this study. This paper makes two main contributions to the literature. First, it presents experimental evidence on the effectiveness of a light, low-cost multifaceted economic inclusion intervention added to a government cash transfer program in Chad. Second, it tests whether the multifaceted economic inclusion intervention induces local spillovers among non-participants in targeted villages. Specifically, we designed an RCT of a multifaceted economic inclusion intervention that included group savings promotion, micro-entrepreneurship training and a lump-sum cash grant. The RCT was implemented in a sample of 88 villages targeted by a regular cash transfer program for extremely poor households (Daye et al, 2024). The design involved a two-step randomization. In the first step, 57 villages were assigned to the economic inclusion intervention, and 31 villages were assigned to control. In the second step, within treatment villages a subset of cash transfer recipient households was assigned to participate in the economic inclusion intervention. Based on this design, we identify the impact of the 2 economic inclusion intervention across treatment and control villages. We also identify local spillovers from the economic inclusion intervention within treatment villages on both cash transfer recipient and non-recipient households. The economic inclusion intervention was implemented in 2019, and we analyze results based on a follow-up survey fielded in 2021, 18 months after its completion. Results show that the economic inclusion intervention has substantial direct welfare effects on participant households. It increases household food consumption by 13% and a composite household welfare index (covering food security and consumption) by 0.17 standard deviations. Consistent with these welfare improvements, we find changes in income-generating activities for the women eligible for the economic inclusion intervention, including some labor reallocation along with higher revenues from agriculture and off-farm micro-enterprises. Participation in savings groups and amounts saved also remain higher 18 months after the end of the intervention. These economic impacts are associated with a broader enhancement in well-being, in particular greater women’s empowerment and control over household resources, as well as improvements in dimensions of social well-being, although not of psychological well-being. Importantly, results reveal that the economic inclusion intervention has positive local spillovers on non- participant households in treated villages. The patterns of the spillovers mirror those of the direct impacts. Food consumption and a standardized welfare index improve for non-participant households. A reallocation of labor and higher revenues from agriculture and micro-enterprises are also observed. This again is associated with broader improvements in well-being, in particular women’s empowerment and dimensions of social well-being. Lastly, we compare impacts to costs. When we only consider direct effects on food consumption, we find that the economic inclusion intervention is cost-effective under weak assumptions, for instance if impacts after the follow-up survey dissipate at a rate of 75% per year or less. This places the estimates from the Chad pilot program at the top of the distribution of impacts per unit of cost for the graduation programs reviewed by Kondylis and Loeser (2021), and slightly below the benefit-cost ratios from Niger in Bossuroy et al. (2022). Once the spillovers are accounted for, the intervention becomes cost-effective based on impacts already observed by the 18-months follow-up survey, without assuming any persistence. In this case, the internal rate of return of the economic inclusion intervention is 73%. Economic spillovers of much smaller magnitude than those we estimate would be sufficient for the intervention to be cost-effective without assuming that any impact persists after the follow-up survey. 3 Graduation or economic inclusion programs are based on the premise that extremely poor households face multiple constraints to improve their income-generating capacities and well-being, so that multifaceted interventions are needed for sustained poverty reduction. Graduation pilots achieved strong results in several countries (e.g. Banerjee et al. 2015; Bandiera et al., 2017; Bedoya et al., 2019; Banerjee et al, 2021; Balboni et al., 2021). Research subsequently unpacked program components to disentangle the contributions from capital, training, coaching, or psychosocial interventions (Sedlmayr et al. 2020; Banerjee et al. 2022; Bossuroy et al., 2022; Botea et al., 2023). In parallel, governments looked for ways to integrate multifaceted economic inclusion approaches in their national social protection systems. The Sahel has been at the forefront of these efforts: government teams working on adaptive social protection across the region shared an objective to boost households’ livelihoods and resilience. This occurred in a broader political economy context where coverage of regular safety nets remained limited and questions about the productive impacts and cost- effectiveness of social protection interventions were recurrent, creating a demand to strengthen and document impacts on investments and livelihoods. Starting in 2016, government safety net teams across the Sahel collaborated to design a multifaceted economic inclusion model that could be delivered on top of national cash transfer programs (Archibald et al., 2021). Niger, Senegal, Mauritania and Burkina Faso implemented a package that included group savings promotion, micro-entrepreneurship training, psychosocial interventions and a lump-sum cash grant. The model was designed for scaling, already at a cost much lower than graduation programs in the literature. It achieved strong impacts on economic and psychosocial outcomes in Niger (Bossuroy et al. 2022), solid results in Mauritania and urban Senegal (Bossuroy et al., 2025a, 2025b), with more muted effects in a challenging context due to COVID and insecurity in Burkina Faso (Bossuroy et al., 2024). Chad decided to implement a simplified version of the Sahel economic inclusion package as part of a pilot program embedded in the inception phase of its social protection system. It kept the objective of addressing multiple constraints to boost the livelihoods and productivity of very poor households, but simplified the content so that it could be implemented in a shorter time frame (3-4 months instead of 12-18 months). It streamlined some components, such as the group savings modality, and removed others, such as the psychosocial interventions. The amount of the one-off cash grant was also lowered (to FCFA 45,000) compared to other Sahel countries (where it ranged from FCFA 80,000 in Niger to FCFA 150,000 in Senegal). This paper thus documents the effectiveness of an economic inclusion package that 4 was light and had a very low cost ($104 in nominal terms, or $252 in 2016 PPP). Results show that it has substantial impacts both on participants and non-participant households in target communities. Broader questions about the nature of potential spillovers from policies to support livelihoods, including through graduation and economic inclusion programs, often arise in the policy dialogue and in the academic literature. At times, these questions are formulated as concerns that targeted interventions may worsen outcomes for non-participants in treated areas. This partly stems from debates on safety net programs, for instance about the social implications of targeting (e.g. Della Guardia et al., 2022), the potential negative spillovers on non-participants’ psychological well-being (Haushofer et al., 2019, Baird et al., 2013), or increases in the price of non-tradable goods in the presence of market frictions (Filmer et al, 2023; Cunha et al., 2019). In addition, there is a possibility that livelihood programs promote specific economic activities or businesses that could crowd out opportunities for non-participants, create competitive pressures or induce market-stealing (Bloom et al., 2007; Drexler et al, 2014; Rotemberg, 2019; McKenzie and Puerto, 2019). Studies that measure spillovers from multifaceted graduation or economic inclusion programs remain rare.2 Published studies to date have rarely found significant negative or positive spillovers (e.g. Banerjee et al. (2015), Bandiera et al. (2017), Blattman et al., (2016), Sedlmayr et al. (2020), Baird et al., 2024)). Bandiera et al (2017) find signs of positive spillovers on wages and assets, but Sedlmayr et al. (2020) find weak signs of negative spillovers. Botea et al. (2023) find evidence of negative spillovers on mental health, echoing some negative point estimates found by Banerjee et al. (2015) on subjective well-being. The limited spillovers found to date might be surprising given there is a broader literature highlighting spillovers from specific interventions that comprise the bundle of economic inclusion or graduation programs. There are several pathways through which positive spillovers might be expected. First, savings groups can replicate and spread over time (e.g. Beaman et al., 2014), as such expanding opportunities for households that were not initially targeted to mobilize financial resources for investments (e.g. Marguerie and Premand, 2023).3 Second, cash transfers or grants can increase loans or private transfers between households (Angelucci and De Giorgi, 2009; Ribas, 2020; Aguila et al., 2 A series of ongoing studies are in the process of documenting spillovers of economic inclusion interventions, but do not yet have published results on those spillovers (e.g. Abate et al. 2024; Brune et al., 2024; Coville et al. 2022; Fernandez et al. 2024). 3 Spillovers can also arise through the expansion of access to bank accounts or micro finance services (Demont, 2016; Dupas et al., 2019). 5 2022; Carraro and Ferrone, 2024). Cash transfers or grants can also induce substantial multiplier effects in the local economy, for instance generating externalities by boosting local demand for products (e.g. Egger et al., 2022; Papineni et al, 2024). This may arise due to slack or underutilization of factors of production in the presence of indivisibilities, creating the possibility of large increases in micro- enterprise productivity following cash injections, particularly in poor rural economies (Walker et al., 2024). Third, spillovers may stem from more frequent social relations, for instance through interactions with leaders (Macours and Vakis, 2014), peers (Berge et al., 2021), or participation in savings groups (Marguerie and Premand, 2023). Fourth, beyond the financial or product markets, spillovers may arise through the labor markets, for instance when the demand for labor increases in households’ income- generating activities or social protection programs increase equilibrium wages (Ambler et al., 2018; Berg et al., 2018; Muralidharan et al., 2023; Imbert and Papp (2015); Franklin et al. (2024)). Lastly, training interventions can also induce knowledge spillovers.4 Our RCT is not designed to experimentally isolate a single or specific source of spillovers, but we find evidence consistent with several mechanisms. Over time, non-beneficiaries become more likely to participate in a savings group, suggesting improved access to capital (in line with Beaman et al., 2014, Marguerie and Premand, 2024). There is more financial support and sharing within communities, including higher food consumption from gifts among cash transfer recipient households, and a stronger ability to mobilize financial resources in case of shocks. Spillover effects on some economic activities are observed, including for cash transfer non-recipient households, possibly because regular cash transfers already spurred activities for cash transfer recipients (Daye et. al, 2024). We find some evidence pointing to an increase in the use of paid labor in agriculture, consistent with some spillovers occurring through the labor market. Results are also compatible with demand-side effects from cash injection on the product market (as in Egger et al., 2022). Taken together, these results are suggestive of the intervention improving saving, sharing, financial support mechanisms, as well as potential demand-side effects in the labor and product markets. More generally, the findings also show that targeting livelihood interventions to poor households can induce broader benefits in the local economy, which partly mitigates concerns about targeting social programs within small geographical areas when budgets are insufficient to achieve universal coverage. 4 In addition, a study focusing on children’s outcomes finds evidence of spillovers from a graduation program on nutrition outcomes through transmission of information and nutrition practices (Raza et al. (2018)). 6 The paper is structured as follows. Section 2 provides additional details on the Chad cash transfer program and economic inclusion intervention. Section 3 details the RCT design, data, and estimation strategy. Section 4 presents results on direct impacts, indirect impacts, and cost-effectiveness. Section 5 concludes. The main figures and tables are in the appendix, with additional tables in a supplementary appendix. 2. Context and Interventions The Chad Safety Net Chad is a fragile country where poverty affects 42% of the population and is particularly widespread in rural areas (Savadogo and Sanoh, 2021). Located in the Sahel, Chad suffers from civil conflict and is one of the most climate-vulnerable countries globally, facing recurrent food crises exacerbated by droughts and floods, which further deepen food insecurity. Gender inequality in Chad is also among the highest in the world. In 2016, the Government of Chad, in partnership with the World Bank, introduced a social assistance pilot as an initial step in establishing a social safety net system. The pilot supported poor and vulnerable households through cash transfers combined with an economic inclusion program in two regions, and with a cash-for-work intervention in N’Djamena. The focus of this study is on the economic inclusion pilot program. In the two regions where the intervention took place, Logone Occidental and Barh-el-Ghazal, the 14 poorest cantons were prioritized, and within these, 88 villages were selected through a lottery process.5 Within selected villages, the poorest households were identified using a Proxy Means Test (PMT) followed by community validation (Della Guardia et al. 2022). A total of 6,200 households—4,650 in Logone Occidental and 1,550 in Barh-el-Ghazal—benefited from cash transfers of 45,000 CFA francs (US$75) provided every three months for two years, starting in December 2017. The transfers were given to women recipients, in most cases the wife of the household head.6 The transfers represented approximately 25% of monthly consumption for recipient households. An economic inclusion 5 In this paper, we focus on the 88 villages in which the economic inclusion intervention was randomized, which had 20 cash transfer recipient households or more. Twenty-two additional villages were targeted by the cash transfer program but had fewer than 20 beneficiaries. 6 A few households did not have an adult woman, in which case the recipient was a man. 7 intervention was then provided to a subset of cash transfer beneficiaries, starting in July 2019, approximately 18 months since the first cash transfer was made. Two earlier studies analyzed cash transfers before the introduction of the economic inclusion intervention. Twelve months after the first transfer, Daye et al. (2024) find significant positive effects on household consumption and women’s business activities. Women who receive the transfers are twice as likely to start a new business compared to non-recipients, with a 53% increase in their profits. Additionally, women’s self-efficacy strongly increases, and there is an 18% reduction in the likelihood of depression. A qualitative study by Della Guardia et al. (2022) highlights that non-recipient households in cash transfer villages also experience significant positive economic impacts, but that the program may have led to social cleavages and divisions by not being able to cover all poor households, with unclear overall effects for non-recipients. Economic inclusion interventions A simplified economic inclusion program was introduced for 1,798 cash transfer recipients. Its goal was to further boost revenues from income-generating activities. Compared to other economic inclusion interventions in the Sahel and in the literature, the intervention was streamlined. The cost was lower, at $104 in nominal terms ($252 PPP) per participant. It was implemented over a shorter duration, just 3-4 months, as opposed to 12-24 months for most other interventions. The economic inclusion intervention was delivered by the Government Safety Net Unit, with the support from two local NGOs. The intervention included savings facilitation, coaching, micro-entrepreneurship training, and lump-sum cash grants, as detailed below. Savings facilitation. The intervention facilitated the creation of savings groups consisting of 10 to 25 members, with an average of 20 members per group. These groups operated a “Table Banking” model, which includes regular meetings when each member decides whether and how much to save or to borrow from the pooled group savings. Inspired by other savings group models, such as Village Savings and Loan association (VLSAs), the approach has relatively short cycles and allows participants to borrow funds immediately after the savings contributions are pooled into a common pot. A total of 120 savings groups (88 in Logone Occidental and 22 in Barh-el-Ghazal) were initially established. Coaching. Coaching is a process to support participants in their economic activities. NGO coaches provide individual coaching focused on helping participants develop their income-generating activities 8 and achieve their economic goals. Coaches also act as facilitators and guide the functioning of the savings groups. Micro-entrepreneurship Training. A 5-day micro-entrepreneurship training was delivered to the groups of participants by staff from the two local NGOs. The curriculum was adapted from the ILO’s Manage Your Business Better (GERME) Level 1 training, and specifically tailored for illiterate participants. It covers essential skills in micro-entrepreneurship, such as basic accounting, management principles, market research, planning, saving, and investment. Additionally, the training includes guidance on selecting economic activities, with a focus on analyzing risks and opportunities at the local level. However, the content does not cover technical training in specific income-generating activities. Lump-Sum Cash Grant. A one-time lump-sum cash grant of 45,000 CFA francs was provided on top of the payment of the last regular cash transfer (also of 45,000 CFA). The goal of the additional transfer was specifically to promote investments in income-generating activities. The total transfer of 90,000 CFA was distributed through the microfinance agency which also provided the regular cash transfers. The study can be considered a proof of concept for a light economic inclusion program. The pilot was small-scale and introduced in a nascent social protection system by layering on a regular cash transfer program implemented by the government. The economic inclusion interventions were managed and supervised by the government safety net unit but implemented in partnership with two local NGOs. We do not have fine-grained data on the quality of implementation, but it was considered generally satisfactory, on par with other countries in the Sahel.7 3. Experimental design and data 3.1 Experimental design The RCT aims to estimate both the direct impacts of the economic inclusion intervention on participant households and the indirect impacts on non-participant households within targeted villages. The economic inclusion intervention is layered on top of the regular cash transfer program. Out of the 88 villages (with more than 20 beneficiaries) participating in the regular cash transfer program, 57 were 7 Specifically, we would not consider this pilot an outlier with extraordinarily high implementation quality. The research team was not as involved supporting program monitoring in Chad relative to other Sahel countries. The international NGO that provided advice and technical assistance to the government (Trickle Up) for the implementation of the productive measures was the same that played a similar role in other Sahel countries. 9 randomly assigned to receive the economic inclusion intervention. Within these villages, 1,798 cash transfer recipient households were randomly assigned to participate in the economic inclusion intervention, and 2,207 continued receiving the regular cash transfers only. The remaining 31 villages serve as a control group, with 1,861 households receiving regular cash transfers without additional support. By comparing the outcomes of households assigned to the economic inclusion program with those of households only receiving cash transfers in control villages, this design identifies the direct impact and cost-effectiveness of the economic inclusion intervention (Figure 1). Additionally, by comparing outcomes for households not assigned to the economic inclusion intervention between treated and control villages, this design also identifies local spillover effects, that is, the indirect effects of the economic inclusion intervention on non-participant households in targeted villages. These spillover effects can be analyzed separately for cash transfer recipient and non-recipient households. 3.2 Timeline, sampling and data The cash transfer program began in December 2017. The RCT of the economic inclusion program took place between 2019 and 2021. A baseline survey was conducted prior to the implementation of the economic inclusion intervention in November and December 2018. The simplified economic inclusion intervention was then implemented between July and October 2019. The follow-up survey was collected in March and April 2021, approximately 18 months after the cash grants were distributed. This timing aligns with RCTs in other Sahel countries (e.g. Bossuroy et al., 2022). Although the program was rolled out before the COVID-19 crisis, the follow-up survey recall period (often 12 months) covers a period affected by the pandemic. The baseline and follow-up survey sample includes the 88 villages benefiting from the cash transfer program.8 At baseline, the sample included 1,438 households, with 1,033 households from treatment villages (of which 665 received cash transfers) and 405 from control villages (of which 202 received cash transfers).9 The sample was increased at follow-up to a total of 4,030 households across the 88 8 Twenty-two other villages with fewer than 20 cash transfer household beneficiaries did not enter the sample for the experiment because the economic inclusion was not randomized. 9 The survey was a follow-up for a study on the impact of cash transfers (Daye et al. (2024), which we use as a baseline for the economic inclusion intervention. In the 57 villages randomly assigned to the economic inclusion intervention, the baseline survey aimed to sample 21 households per village, including 14 cash transfer recipient 10 villages.10 This was done to increase statistical power, including for the spillover analysis.11 Data was successfully collected in 88 percent of the planned sample (3,532 households). Figure 1 provides the break-down of the effective sample: data was collected for 2,093 households in treatment villages (of which 882 received cash transfers and productive inclusion interventions and 619 cash transfers only) and 1,439 households in control villages (of which 884 received cash transfers). Attrition during the follow-up survey was balanced across groups (see Table 1, panel D). The baseline and follow-up surveys gathered data on household demographics, consumption, food insecurity, income-generating activities, and psychosocial well-being. Some sections focused on household-level information, such as for consumption or food insecurity, while others collected individual-level data, including on income-generating activities, savings, debt, psychosocial well-being or women’s empowerment. The survey was conducted with the individuals eligible for the transfers or productive measures (almost always women) and heads of households. 3.3 Estimation strategy and balance As mentioned above, the economic inclusion intervention was randomized in two steps, first at the village then at the household level. The analysis relies on two main specifications to estimate direct impacts and local spillovers 18 months after the intervention. The specification to estimate direct impacts writes: () = + + . + + with ℎ the outcome of interest for observation h at follow-up, ℎ is the baseline outcome, ℎ is a treatment indicator that takes a value 1 for cash transfer recipient households randomly assigned to the households and 7 non-recipients. In the 31 control villages that were not selected to receive the economic inclusion measures, 14 households were sampled per village (split equally among cash transfer recipients and non- recipients). In smaller villages, the planned sample did not always reach these targets. 10 In the 57 villages randomly assigned to the economic inclusion intervention, the follow-up survey sample included 42 households per village, including 18 cash transfer recipients receiving the economic inclusion intervention, 12 recipients of cash transfers alone, and 12 non-recipients. In the 31 control villages, the sample included 54 households per village (including 33 cash transfer recipients and 21 non-recipient households). In smaller villages, the planned sample did not always reach these targets. 11 This sample allows detecting minimum direct effects in the range of 0.21-0.22 standard deviations for most outcomes of interest, except for food consumption per capita and the food insecurity experience scale, for which the minimum detectable effect is above 0.3 due to higher intra-cluster correlations. The spillover sample allows detecting effects of 0.12-0.15 standard deviations for most outcomes, though again a bit higher for the food insecurity experience scale (0.215) or per capita food consumption (0.295). 11 economic inclusion intervention in treatment villages, and a value of 0 for cash transfer recipient households in control villages, represents randomization strata fixed effects12 and ℎ is the error term. The (pooled) specification to estimate local spillovers writes: () = + + . + + with ℎ takes the value of 1 for cash transfer recipient households randomly assigned not to receive the economic inclusion intervention or cash transfer non-recipient households in treatment villages, and a value of 0 for cash transfer recipient and non-recipient households in control village.13 The spillovers can also be estimated solely on cash transfer recipients, i.e. those randomly assigned not to participate in the economic inclusion intervention in treatment villages, and cash transfer recipient households in control villages: () = + + . + + ; ℎ recipient Similarly, the spillovers can also be estimated only on cash transfer non-recipient households across treatment and control villages: () = + + . + + ; ℎ − In all specifications, standard errors are clustered at the village level. Note that each specification includes lag baseline outcomes as control, similar to an ANCOVA specification. Some households that were not in the baseline sample were added at follow-up, as mentioned in Section 3.2. For those observations, we impute the baseline mean of their group within village (cash transfer and economic inclusion recipients, cash transfer recipients, or cash transfer non-recipients) in each specification above. Table 1 presents baseline balance tests for the treatment and control groups used to estimate direct effects based on specification 1 (columns 1 and 2), and for the pooled spillover specification 2 (columns 3 and 4), respectively at the level of the household (Panel A), eligible individual (woman) (Panel B), and (male) heads of household (Panel C). (Table A.1 in the appendix provides similar balance tests for 12 Specifically, there are 16 strata based on 4 dimensions: region, land, population and wealth. 13 When estimating the pooled spillover specification, we include weights to ensure that the share of cash transfer recipient and non-recipient households is the same across treatment and control. Indeed, this share is lower in treated villages (619/(592+619)=0.51) than in control villages (884/(555+884)=0.61) since some cash transfer recipients are assigned to also participate in the production inclusion interventions. Results are robust when estimated without the weights. 12 specifications (3) and (4), which unpack spillovers for cash transfer recipient, respectively non-recipient households.) Treated households have an average of 5.5 members. The household head is the cash transfer program recipient in 21% of the households, meaning in most cases that these are female-headed. Among the households, 17% are polygamous. The eligible individuals are almost all female (98%), hence we use the shorthand ‘eligible women’ for that group.14 They are 25.6 years old and have 1.5 years of education on average. When presenting results for household heads, we focus on those who are not also the eligible individual, hence in almost all those cases (98%) the household heads are male, and we use the shorthand ‘male household heads’ for this group. Panel A reveals some imbalances for the direct specification at the household level. Specifically, there is a notable imbalance in non-food consumption, which is higher in the treatment group at baseline, as well as the (reversed) food insecurity experience scale, which indicates lower food security in the treatment group at baseline. Treated households tend to be slightly smaller, too. Given the imbalance in non-food consumption, the discussion will emphasize more the results for food consumption, which is balanced at baseline.15 In addition, we build an index of the various food security and consumption measures. This aggregate index is balanced across the treatment and control groups. Lastly, to further mitigate potential concerns about baseline imbalances, we include lag baseline outcomes as control in each specification, as mentioned above. We also report p values from permutation tests for the main results. Importantly, imbalances are mostly observed in the household-level variables. Panels B and C show that at the individual level the direct specification is well-balanced both for the eligible individuals (women) and (male) household heads, with the area cultivated by the household head the only exception (household heads in treated households cultivate a marginally smaller area). 14 The other 2% of households did not have an adult woman when the individual recipients were identified. 15 Total daily consumption per adult equivalent is the sum of daily household food and non-food consumption, divided by the number of adult equivalents per household. The number of adult equivalents is calculated using the OECD equivalence scale which assigns a value of 1 to the first adult in the household, 0.7 to each additional adult, and 0.5 to each child. For food consumption, we ask about household consumption of a variety of food products in the last week, including food consumed from own production, purchased or gifted. To get food expenditure, we multiply amounts consumed by prices. For food products that are both purchased and consumed, we use reported purchase prices. For food products that are consumed but not purchased, we use median purchase prices. We winsorize consumed values within each food product and we divide weekly household totals by 7. For non-food consumption, we ask about monthly (and yearly) expenses on a wide variety of goods and services that are typically consumed on a monthly (or yearly) basis. For each of these expenditures, we winsorize values at the finest level possible and re-scale to daily values. 13 Similar balance patterns are found for the pooled spillover specification. We note that, in this case, food consumption is also imbalanced, but again that the aggregate normalized welfare index across all consumption and food security outcomes is balanced. The individual samples remain well-balanced, except for the eligible women being a bit younger and cultivating slightly more land, and household heads being a bit younger and operating slightly fewer businesses in treatment villages. 4. Results 4.1 Direct impacts Economic welfare We start by presenting results for direct impacts based on specification (1). Table 2 contains household economic welfare outcomes, including consumption and food security indicators. The economic inclusion intervention has substantial direct welfare effects on participant households 18 months after its completion. We find a significant increase in (food) consumption. Treated households’ consumption is higher by 54.2 FCFA per adult equivalent per day, or 14% relative to control. Food consumption represents 76% of total consumption among control households, and the increase in household consumption is driven by an increase in food consumption by 37.8 FCFA per adult equivalent per day, 13% or 0.16 standard deviations relative to control (Table A2).16 The point estimate for non-food consumption is positive and not negligible (14.2 FCFA per day), but not statistically significant.17 Results are more muted for the food security proxies. Point estimates for measures of dietary diversity tend to be positive (by 2.1 points or 6% relative to control for the Food Consumption Score, and by 0.12 points or 2% relative to control for the Dietary Diversity Score), but not for a (reverse-coded) index of the frequency of food insecurity experiences, and with no statistically significant effect for any of the food security indicators. These patterns suggest that households primarily increase consumption of food 16 The top panel of Table A10 shows results for the components of food consumption. Impacts on food consumption are in part explained by higher consumption from own production (+8.9 FCFA per adult equivalent per day, significant at the 10 percent level). The point estimates for food purchased (+13.4) and gifted (+15.5) are also positive but not statistically significant (possibly due to higher noise). 17 Since food consumption represents the largest share of consumption, and there is an imbalance in non-food consumption but not food consumption at baseline (as mentioned above in Section 3.3), we consider the food consumption estimate as our main estimate of the intervention’s welfare effects, including for the cost- effectiveness analysis in Section 4.3. We also emphasize results for the aggregate welfare index, which is balanced at baseline. 14 products they already frequently consume, such as cereals, without having a more diverse diet.18 Overall, given the effects on (food) consumption and positive point estimates for 2 out of 3 food security indicators, the intervention increases the composite household welfare index capturing both food security and consumption by 0.17 standard deviations (see standardized effects on components in Table A2). We thus conclude that the economic inclusion intervention improves household economic welfare. Section 4.3 below discusses the magnitude of these effects relative to costs and the broader literature. Livelihoods The primary goal of the economic inclusion intervention was to strengthen households’ productive capacities and livelihoods. To understand the potential drivers of improvements in household welfare, we thus analyze impacts on a range of livelihood outcomes. For each type of livelihood activity, we first present impacts for the eligible women, followed by the (male) household heads. Table 3 presents direct effects on agriculture. Agriculture is one of the main income-generating activities in the study sample, with almost three-quarters of eligible women and male household heads involved in agriculture in the control group at follow-up. The intervention tends to decrease the time that eligible women spend working in agriculture on their own plots by 5.7 days (from an average of 27 days, column 3) as well as on other household members’ plots (by 7.5 days from an average of 30 days, column 4). This is equivalent to a reduction of time spent working in agriculture by about 1 day per month over the year. However, the value of agricultural production on eligible women’s own plots increases substantially, by 6,891 FCFA, a 34% increase. This is consistent with an observed increase in food consumption from own production (Table A10, top panel). Similar patterns are observed for the (male) household heads: they decrease their time spent working on their own plots (by 9.3 days, from an average of 64 days), but the increase in the value of agricultural production is not statistically significant for them. Results suggest a re-organization of household labor for agricultural production, with both eligible women and male household heads decreasing their labor inputs, and other household members also decreasing their time spent on the plots managed by the eligible women and male household heads (column 5). While the use of inputs such as fertilizers remains unchanged, the likelihood of using paid agricultural labor increases substantially (by 7.8pp, from a base of 21% among control households, Table 18 The more muted effects on food security may also be due to the timing of the follow-up survey, which took place before the lean season. Indeed, impacts of livelihood interventions on food security have been shown to vary during the year due to seasonality (e.g. Premand et al., 2024). 15 A3).19 This suggests that households in part substitute their own labor with paid labor, and this labor reallocation is associated with a higher value of production. Table 4 presents results for off-farm income-generating (business) activities. Of eligible women in the control group, 60% are engaged in off-farm micro business activities, and the treatment does not increase the likelihood of operating an off-farm business or the number of off-farm businesses operated. Point estimates suggest that eligible women spend 1.2 days more working in off-farm businesses per month, which is marginally not statistically significant (with a p value of 0.105), but consistent with the observed decrease in time worked in agriculture. There is also suggestive evidence that the number of months when businesses operate during the year increases (by 0.7 months from a mean of 6.04 months, a 12% increase), although this is marginally not statistically significant (p=0.14). Results point to an increase in revenues from off-farm businesses (by 31,077 FCFA per year, or 27% relative to control), significant at the 10% level. A smaller relative increase in profit is observed (7,075 FCFA per year, or 18% relative to control), but it is not statistically significant. This might be due to the measure being noisy, or to increases in the costs of operating the businesses. We do not find evidence of significant additional investments in business assets.20 Taken together, these results suggest that eligible women allocate more of their labor to off-farm business activities, and in part adjust the ways these activities operate, leading to higher revenues though not necessarily higher profits. Note that cash transfers alone had already produced a significant positive impact on off-farm business activities before the economic inclusion intervention was introduced, which may have reduced the margins for additional improvements (Daye et al, 2024). Table A4 shows that there is limited impact on salaried employment, which remains rare in the study sample, with 7% of eligible women in control working (informally) for someone else. An increase in the number of days that eligible women work for someone else is observed, but from a low base, and without a corresponding increase in earnings. Similarly, a small increase in the number of days worked in agriculture by male household heads is noted. This could be consistent with local labor markets being more active, given the higher use of external labor in agricultural activities. 19 Unfortunately, we do not have detailed measures of the number of days worked by agricultural laborers from outside the household, so that we cannot quantify the increase in agricultural labor and its costs. 20 We note that our measure of business assets may not capture all forms of investment in businesses that households make, although both Bossuroy et al. (2022) and Marguerie and Premand (2024) find evidence of investments based on a similar variable. 16 Table 5 documents impacts on livestock (columns 1-4). The likelihood of holding livestock, the time spent raising livestock, and an overall index of the number of animals owned all remain stable both for eligible women and household heads. While the revenue from livestock sales is stable for eligible women, it declines for male household heads, suggesting that they become less likely to sell livestock, possibly in part due to the observed changes in agricultural activities and off-farm businesses. Finally, Table 5 provides evidence of lasting effects on participation in informal savings group (columns 5-6). Of the eligible women, 53% remain active member of savings groups 18 months after the end of the program, a 44 percent increase relative to control. This shows that the savings groups that were established sustained operation well after the program concluded, consistent with results in other West African countries (e.g. Bossuroy et al., 2022, Marguerie and Premand, 2023). Also consistent with continued engagement, eligible women contributed 2,253 FCFA more to a savings group relative to control, again a large 64% increase.21 Empowerment and Psychosocial Wellbeing The effects on economic welfare and livelihoods are associated with changes in broader dimensions of well-being for eligible women. Results on livelihoods point to shifts within the households, in particular labor reallocations, as well as increases in revenues from agriculture and off-farm businesses for eligible women. Importantly, the share of eligible women’s revenues in total household revenue is 46% in the control group but increases by 7.3pp in the treatment group (Table 6, column 1). In other terms, the intervention increases the relative share of household revenues generated by women so that it reaches more than 50%. Improvements in women’s empowerment are also noted (Table 6, column 2), particularly an increase in women’s control over household resources. This is driven by additional decision-making power for their own health care, children’s education, as well as greater influence over the use of their partner’s earnings (see first row of Table A14 for components of the index). In contrast, no change in other dimensions of empowerment, such as the quality of intra-household dynamics (Table 6, column 3, with components in Table A15), is observed. Turning to psychosocial well-being, the results highlight improvements in various dimensions of social wellbeing, including financial support, social standing and collective action (Table 6, columns 4-7). 21 Participation in a savings group was only measured for women, so that we do not have comparable measures for (male) household heads. 17 Households report a higher ability to leverage financial support in case of need. Specifically, they are more likely to report being able to mobilize money from others when exposed to a shock and also self- report a higher degree of financial support in their communities (for components of the index, see Table A16, top panel).22 This occurs without participants having more people they seek or provide advice, though they do report having more role models (for components of the index, see Table A17, top panel). Eligible women also report a higher social standing: they feel more respected and that their opinion is more considered in the community (for components of the index, see Table A18, top panel). Furthermore, an index of collective action improves, driven by participants being more likely to be members of associations and having responsibilities in these associations (Table A19, top panel). These effects might be driven by the sustained participation in saving groups, which created new leadership positions. Despite this, there is no improvement in broader aspects of social cohesion, though participants report that they are slightly less likely to have enemies (Table A20, top panel). The results thus show improvements in economic welfare, economic activities, women’s empowerment and dimensions of social well-being. Still, we do not observe significant improvements in psychological well-being, self-efficacy or expectations for the future (Table 6, columns 9-11, with components of the indices shown in Tables A21-A23, top panel). This contrasts with results obtained in other Sahel countries, particularly Niger, where substantial effects on psychological well-being were found. This may in part be due to the intervention being shorter, lighter, and not including psychosocial interventions, as well as from the economic effects being more muted (e.g. standardized welfare effect of 0.16 standard deviation in Chad, compared to 0.25 standard deviations in Niger). While the light, low-cost multifaceted intervention induced broad impacts on economic outcomes in Chad, it may not have been sufficient to induce mental well-being benefits. We further discuss the magnitude of effects in Section 4.3. 4.2. Local spillover effects We now discuss local spillover effects on households that were not assigned to the economic inclusion intervention in treated villages, based on specification (2). In treated villages, there are two types of non-participant households, some who receive the regular cash transfer program (specification (3)), and others who do not (specification (4)). We present the main results for the pooled sample of non- 22 This is in line with the positive (but noisy) point estimate on food consumption from gifts (Table A10, top panel). 18 participants (specification (2)). Tables in appendix show disaggregated effects (specifications (3) and (4)), which we discuss in the text when there are signs of differences between results for the two groups. Note that the study was not powered to detect spillovers separately for each subgroup, hence this analysis is more exploratory and suggestive. Economic welfare Table 7 provides evidence of spillovers on non-participant households’ economic welfare. Non- participant households’ consumption is higher by 35.9 FCFA per adult equivalent per day, or 9%, relative to similar households in control villages. This is driven by food consumption, which represents 77% of total consumption and is higher by 26.3 FCFA per adult equivalent per day (or 9% relative to control).23,24 These spillovers have a smaller magnitude than the direct welfare effect, but not by much, representing 0.11 standard deviations for food consumption (Table A5) compared to direct effects of 0.16 standard deviations (Table A2). While these results point to substantial spillovers, evoking the large effects found in Egger et al. (2022), we interpret the size of the estimated effect on consumption with caution in light of some of the imbalances noted in Section 3.3. There is also evidence of spillovers on food security. The two indicators capturing dietary diversity have positive coefficients, with a significant increase in the food consumption score (by 3.6 points) and in the household dietary diversity score (by 0.15 point). There is no significant effect in the (reverse-coded) index capturing the frequency of experiences of food insecurity. Still, jointly considering food security and consumption indicators show a positive effect on the composite welfare index by 0.17 standard deviations. This is noteworthy as this indicator is balanced at baseline. Spillovers on food consumption appear driven by cash transfer recipient households (Table A8). Among cash transfer recipients, a significant increase in consumption from gifts is observed (Table A10, third panel), which is suggestive of more sharing taking place for this group. Point estimates in Table A8 are positive for cash transfer non-recipients, but not statistically significant. Since the study is not powered to detect spillovers for each subgroup, we interpret these results with caution given the relatively large 23 The point estimate for non-food consumption is positive (10.4 FCFA per day), but not statistically significant. 24 This is in part explained by higher consumption from own production (+6.5 FCFA per adult equivalent per day) and food gifted (+21.5 FCFA). The point estimate for food purchased (+1.3 FCFA) is close to zero. None of the coefficients from the decomposition of food consumption is statistically significant (possibly due to higher noise) (Table A10, second panel). 19 confidence intervals, especially as both groups display similar improvements in dietary diversity (such as the food consumption score) and in the composite welfare index. Livelihoods The welfare spillovers are indicative that the economic inclusion intervention has positive effects on the local economy in participating villages. Thus, we next consider how the economic activities of non- participant households are affected. We find evidence of improvements that are consistent with the patterns observed for the direct effects, with spillovers arising through channels such as improvements in saving, sharing and financial support mechanisms, as well as potential demand-side effects in the labor and product markets. Table 8 document spillovers on agricultural activities. As for direct effects, we find evidence that individuals in non-participant households decrease their time worked in agriculture (by 6 days per year for eligible women on their own plots). Other household members also provide less labor to the plots managed by eligible women and male household heads. Male household heads cultivate slightly less land (by 0.4 hectares, though there was some imbalance in this indicator at baseline) and tend to work fewer days on their plots (by 6 days, but not statistically significant). Taken together, these results are indicative of a labor reallocation and re-organization of agricultural activities, following which the value of agricultural revenues either remains stable (for eligible women) or increases (for male household heads). This is accompanied by an increase in the share of households hiring labor (by 27%) and a decrease in the purchase of seeds (by 24%) (Table A6). Increases in revenues tend to be driven by the plots managed by male heads from households who do not receive regular cash transfers (Table A11). We do not observe significant spillover effects on the overall likelihood that eligible women in non- participant households hold wage jobs or on their wage earnings, but we do see a small increase in the number of days worked in wage jobs (Table A7). We also note a small increase in days worked by male household heads in agricultural wage jobs. This is in line with a more active labor market and the increase in hired labor noted above. Although the magnitude is limited, this suggests that some spillovers may have occurred through labor markets. In parallel, Table 9 reveals an increase in time worked in off-farm income-generating activities, particularly for women who spend 2.4 additional days per month working in those activities. Importantly, the number of months women-led business activities operate during the year is also higher 20 by nearly 1 month on average, which contributes to increasing total revenues as well as profits. This may partly explain why women reduce time worked in agriculture. The increase in the number of months when businesses operate is consistent with demand-side effects, such as those documented for cash grants by Egger et al. (2022), who find substantial multiplier effects in the local economy. Walker et al. (2024) discuss that such effects can be large in the presence of slack in micro-enterprises, with scope to increase revenues and profits without increases in costs. Our study context is precisely one where such demand-side effects may be positive and large in light of the parameters discussed by Walker et al. (2024). Indeed, the communities are remote and isolated, with both consumers and producers strongly relying on the local markets, hence with a high propensity to spend locally. For instance, households are located 78 minutes from the nearest market on average. Micro businesses mostly purchase inputs from and sell products to the community. For instance, 54% purchase inputs from the community, with 30% from outside the community (and the rest not purchasing inputs during the survey recall period). Similarly, 63% sell output within the community, with 20% outside the community (and the rest not selling during the survey recall period). Table 10 reveals mixed results on spillovers for livestock (columns 1-4). On the one hand, eligible women in non-participant households spend a bit more time raising livestock, but neither an overall livestock index nor revenues change for them. Male household heads spend the same time raising livestock, and do not have more livestock, but they still report a decrease in the value of sales. This again may be related to the increases in revenues from other types of activities mentioned above. Furthermore, the decrease in livestock sales appear mostly driven by cash transfer recipient households (Table A13), who may have other means to deal with shocks or lower need for precautionary savings than through livestock. For instance, these are also the households that report higher food consumption from gifts (Table A10, third panel). Lastly, Table 10 (columns 5-6) sheds light on another pathway that may contribute to the spillover effects. Over time, eligible women from non-participant households start joining savings groups, with the share of eligible women from non-participant households engaged in savings group higher by 9.8pp, or 33% relative to control at endline. This effect tends to be larger among households not receiving regular cash transfers, who are less likely to participate in savings groups in the first place (Table A13). These results suggest that the expansion of savings groups contributes to increase access to financial 21 resources among program non-participants, which may help them also finance their income-generating activities.25 Empowerment and Psychosocial Wellbeing Table 11 documents spillovers on women’s empowerment and psychosocial well-being. Direct effects indicated a shift in the role of women and in their bargaining power in the household. We also find evidence of spillovers on women’s empowerment (Table 11). As for participants, the share of revenues generated by eligible women in non-participant households increases to become higher than 50% (from 47.3% in control to 53%) (Table 11, column 1). This is again associated with an increase in women’s decision-making power in their households (Table 11, column 2), including their influence on decisions about daily spending, large purchase, family planning, their own health care or children’s education (see index components in Table A14, second panel).26 These improvements in women’s empowerment are not associated with changes in intra-household dynamics or the relationship with their partner, however (Table A15). Finally, as for participants, Table 11 shows improvements in some dimensions of social well-being, such as financial support (the ability to mobilize financial help in case of shocks) and collective action (participation in associations and leadership responsibilities), but without broader changes in social cohesion, psychological well-being (mental health) or future expectations. Importantly, there is clearly no negative effect on the subjective well-being of non-participants in this context, as the results in Botea et al. (2023) suggest in a context where economic spillovers were not observed. Taken together, these results suggest that the intervention induced broader changes in saving, sharing and financial support mechanisms within communities, as well as potential demand-side effects in the labor and product markets. The findings also show that targeting livelihood interventions to poor households induces broader benefits in the local economy, which partly mitigates concerns about 25 Besides these enhancements in savings mechanisms, we do not find evidence from changes in the amounts borrowed or in financial transfers between households. 26 There is also suggestive evidence that attitudes towards women’s work improve, with a reduction in the small share of women who say they are not free to work (not shown), which might be indicative of shifts in social norms related to women’s work. This may either be driven by the intervention or contribute to explaining some of its spillover effects. 22 targeting social programs within small geographical areas when budgets are insufficient to achieve universal coverage. 4.3 Cost-effectiveness A key aspect that motivated the RCT of the Chad multifaceted economic inclusion package is that, although it was designed to address multiple constraints to productivity and livelihoods, it had a much lower cost (FCFA 62,000 per household, approximately $104 in 2019 nominal terms, or $252 in 2016 PPP terms) than most stand-alone NGO graduation programs and government-led economic inclusion programs in the literature. For instance, the initial graduation pilots were stand-alone programs which proved effective but had relatively high costs. Using 2016 PPP conversion factors, the cost was $1,475 PPP in India, $4,215 PPP in Ethiopia, $5,483 PPP in Ghana, $6,044 PPP in Pakistan (Banerjee et al., 2015b), and $6,183 PPP in Afghanistan (Bedoya et al. 2019). In the Sahel, the economic inclusion package developed for delivery through adaptive social protection systems was designed for scaling and at a much lower cost. For instance, the cost was $584 PPP in Niger, $672 PPP in Burkina Faso, $963 PPP in Senegal, and $1,646 PPP in Mauritania. The Chad economic inclusion program is particularly low-cost.27 Table 12 (panel 1, column 1) shows that the program’s (nominal) costs included FCFA 48,000 (or 79%) for the cash grant (and its payment fees) and FCFA 13,000 (or 21%) for the costs of activities by field agents to facilitate savings groups and deliver the micro-entrepreneurship training. Table 12 (panel 2) summarizes the welfare effects. The welfare effects in column 1 are based on the direct impacts on food consumption from Table 2, which is equivalent to an increase of FCFA 50,129 per year and per household. We then calculate the net present values of welfare gains under several assumptions, including impacts dissipating at 75% per year, 50% per year, 25% per year, or persisting in perpetuity. For each scenario, panel 3 then displays the related benefit-cost ratios, and panel 4 the internal rates of returns. When focusing on direct effects (column 1), costs are larger than benefits by the time of the follow-up survey, assuming full dissipation of impacts thereafter. If impacts dissipate by 75% per year 27 It represents half the cost of the full package implemented in Niger, but is also slightly below the least expensive psychosocial package. This also holds when looking at the review by Andrews et al. (2021), who find that “The total cost of economic inclusion programs is between $41 and $2,253 (in 2011 PPP) per beneficiary over the duration (3.6 years on average) of each program.” 23 after the follow-up survey, the benefit-cost is just around the break-even point (1.01). If impacts dissipate by 50% each year, the intervention is clearly cost-effective, with benefits 1.45 times larger than costs and an IRR of 32%. In their review of graduation programs, Kondylis and Loeser (2021) find an impact of 0.52 dollar per unit of transfer, or an impact per unit of cost in the range of 0.1-0.2. Yet the programs they consider tend to be the higher-costs NGO-led programs. The impact on consumption per unit of cost (0.78) from Chad is in the top of the range of their estimates, in part driven by the low cost. We also note that, although costs are lower (by a factor of 2) than in Niger (Bossuroy et al., 2022), the magnitude of welfare effects (in PPP terms) is also lower (by approximately 3 times). As a result, the cost-effectiveness ratios are lower than those obtained from Niger. Importantly, estimates in column 1 do not factor in spillovers. Accounting for spillovers can have large implications for cost-effectiveness.28 Column 2 adds the spillover welfare effects to our cost- effectiveness calculations. We consider the share of the population that are cash transfer recipients but not assigned to the economic inclusion intervention,29 and factor in the estimated spillovers for that group from Table A8 (column 1). We focus on this subgroup since it is the one for whom spillover effects on food consumption are significant. We find that the economic inclusion intervention becomes cost- effective at the time of the follow-up survey, without assuming any sustained impact thereafter, once the spillovers are accounted for (IRR=73%). If impacts dissipate at a rate of 75% per annum, accounting for spillovers also pushes the intervention well above the break-even point (IRR=98%). While we estimate relatively large spillovers and cannot fully rule out than their magnitude might be overestimated given some of the observed baseline imbalances, we note that even much smaller positive spillovers would be sufficient for the intervention to pass the break-even point without assuming that any impact is sustained after the follow-up survey. Under a medium scenario where impacts dissipate by 50% each year, the internal rate of return of the economic inclusion intervention is 123%, with benefits 3.1 times larger than costs. These results also highlight that the cost-effectiveness of other economic inclusion programs in the Sahel and the broader literature might be higher than previously documented in case there are also positive spillovers in those contexts. 28 For another example, see Janzen et al. (2021), who study a livelihood intervention that explicitly encourages households to pass benefits to others. 29 Specifically, the coverage rate among cash transfer recipients is 2207/4405=55.1%. 24 5. Conclusion This paper presents the results of an RCT assessing the direct and indirect impacts of a low-cost, multifaceted economic inclusion program in Chad. The intervention, implemented alongside a regular cash transfer program targeting extremely poor households, included group savings promotion, micro- entrepreneurship training, and a lump-sum cash grant. The findings highlight that, despite being more streamlined and less expensive than comparable programs, the light economic inclusion intervention produced welfare gains for both participant and non-participant households in targeted villages 18 months after its completion. In particular, the study finds significant increases in food consumption and in an overall household welfare index. These improvements are linked to a reallocation of labor across income-generating activities, with higher revenues from agriculture and off-farm micro-enterprises. Furthermore, the intervention fosters women's empowerment and enhances some dimensions of social well-being. The magnitude of impacts and cost-effectiveness tends to be smaller than other government-led programs in the Sahel, such as in Niger (Bossuroy et al., 2022). Still, they place at the top of the distribution of impact per unit of cost compared to other graduation program in the literature, particularly the more expensive NGO-led programs (see review in Kondylis and Loeser (2020)). In addition, the study reveals significant positive local spillovers on non-participant households in treated villages, a feature that has remained largely undocumented in previous RCTs of graduation programs. Accounting for these spillovers makes the Chad program cost-effective under weak assumptions. Spillover effects largely mirror direct impacts, with improvements in food consumption, economic activities, and women's empowerment for both cash transfer recipient and non-recipient households. The analysis suggests that these spillovers arise through multiple channels, with the program inducing broader changes in saving, sharing and financial support mechanisms, as well as potential demand-side effects in the labor and possibly product markets. Given some baseline imbalances in consumption for the spillover sample, it is possible that the magnitude of spillover impacts may be overestimated. Nevertheless, we find evidence of positive spillover effects across multiple channels, including for an overall welfare index and other livelihood outcomes that are well-balanced at baseline. Importantly, welfare spillovers of a much smaller magnitude would be sufficient for the intervention to be cost-effective after 18 months without making any assumption that impacts are sustained past endline. 25 The study contributes to the growing literature on economic inclusion programs by providing evidence on the effectiveness of a light, low-cost intervention in a challenging context. Various countries around the world are currently testing simplified or streamlined versions of the multifaceted graduation model, and it remains to be seen if lighter packages can be cost-effective and impactful across settings. The study also sheds light on the potential for economic inclusion programs to generate positive spillovers through several channels, showing that these interventions targeted to poor households can induce broader benefits in the local economy. The cost-effectiveness of other economic inclusion programs in the Sahel and beyond might be higher than previously estimated in case they induce similar local spillovers. Furthermore, this study offers insights into the trade-offs between delivering programs with universal coverage in limited geographical areas or targeted programs in broader regions. Notably, targeted household interventions can generate positive externalities that benefit non-participant households within villages as well. References Abate, G., D. Egger, J. Loeser and M. Vinez. 2024. "The General Equilibrium Effects of Graduation Programs: Experimental Evidence from Ethiopia." AEA RCT Registry. Savadogo, A. and A. Sanoh. 2021. “Chad Poverty Assessment: Investing in Rural Income Growth, Human Capital, and Resilience to Support Sustainable Poverty Reduction.” Washington, DC: World Bank. Aguila, E., A. Kapteyn and E. Meijer, 2022. “Effects of Permanent Income Increases on Neighbors: Evidence from an Experiment.” Mimeo. Alderman, H. and Ruslan Yemtsov. 2014. “How Can Safety Nets Contribute to Economic Growth?” World Bank Economic Review 28 (1): 1–20. Ambler, K., A. de Brauw and S. Godlonton, 2018. “Rural labor market responses to large lumpy cash transfers: Evidence from Malawi.” Mimeo. Andrews, C., A. de Montesquiou, I. Arévalo Sánchez, P. V. Dutta, B. V. Paul, S. Samaranayake, J. Heisey, T. Clay, and S. Chaudhary. 2021. The Potential to Scale: Economic Inclusion for the Poorest. The State of Economic Inclusion Report. Washington, DC: World Bank. Angelucci, M. and G. De Giorgi. 2009. “Indirect Effects of an Aid Program: How Do Cash Transfers Affect Ineligibles' Consumption?” American Economic Review 99 (1): 486-508. Archibald, E., T. Bossuroy and P. Premand. 2020. “Productive Inclusion Measures and Adaptive Social Protection in the Sahel: A Case Study.” Washington, DC: The World Bank. 26 Arévalo-Sánchez, I., J. Heisey, S. Chaudhary, T. Clay, V. Strokova, P. V. Dutta and C. Andrews. 2024. The State of Economic Inclusion Report 2024: Pathways to Scale. Washington, DC: World Bank. Baird, S., J. de Hoop and B. Ozler. 2013. “Income shocks and adolescent mental health.” The Journal of Human Resources, Vol, 48(2). Baird, S., C. McIntosh, B. Özler and U. Pape, 2024 “Asset transfers and anti-poverty programs: Experimental evidence from Tanzania.” Journal of Development Economics, Volume 166. Banerjee, A., E. Duflo, N. Goldberg, D. Karlan, R. Osei, W. Parienté, J. Shapiro, B. Thuysbaert, and C. Udry. 2015. “A Multifaceted Program Causes Lasting Progress for the Very Poor: Evidence from Six Countries.” Science 348 (6236). Balboni, C., O. Bandiera, R. Burgess, M. Ghatak, and A. Heil. 2021. “Why Do People Stay Poor?” NBER Working Paper No. 29340. Bandiera, O., R. Burgess, N. Das, S. Gulesci, I. Rasul, and M. Sulaiman. 2017. “Labor Markets and Poverty in Village Economies.” The Quarterly Journal of Economics 132 (2): 811–70. Banerjee, A., E. Duflo, and G. Sharma. 2021. “Long-Term Effects of the Targeting the Ultra Poor Program.” American Economic Review: Insights, Volume 3(4), 471-486. Banerjee, A., D. Karlan, R. Osei, H. Trachtman, and C. Udry. 2022. “Unpacking a Multifaceted Program to Build Sustainable Income for the Very Poor.” Journal of Development Economics, Vol. 155. Beaman, L., D. Karlan and B. Thuysbaert. 2014: “Saving for a (not so) Rainy Day: A Randomized Evaluation of Savings Groups in Mali,” NBER Working Paper No. 20600. Bedoya, G., A. Coville, J. Haushofer, M. Isaqzadeh, and J. P. Shapiro. 2019. “No Household Left Behind: Afghanistan Targeting the Ultra Poor Impact Evaluation,” NBER Working Papers, No. 25981. Beegle, K., A. Coudouel and E. Monsalve. 2018. Realizing the Full Potential of Social Safety Nets in Africa. Washington, D.C.: The World Bank. Berg, E., S. Bhattacharya, D. Rajasekhar and R. Manjula. 2018. “Can Public Works Increase Equilibrium Wages? Evidence from India’s National Rural Employment Guarantee.” World Development 103: 239– 54. Berge, L. I. O. and A. J. Garcia Pires. 2021. “Measuring Spillover Effects from an Entrepreneurship Programme: Evidence from a Field Experiment in Tanzania.” The Journal of Development Studies, 57(10), 1755–1775. Blattman, C., E. P. Green, J. Jamison, M. C. Lehmann and J. Annan. 2016. "The Returns to Microenterprise Support among the Ultrapoor: A Field Experiment in Postwar Uganda." American Economic Journal: Applied Economics 8 (2): 35–64. 27 Bloom, N., M. Schankerman, and J. Van Reenen. “Identifying Technology Spillovers and Product Market Rivalry” Econometrica 81, no. 4 (2013): 1347–93. Bossuroy, T., M. Goldstein, B. Karimou, D. Karlan, H. Kazianga, W. Pariente, P. Premand, C. Thomas, C. Udry, J. Vaillant, and K. Wright. 2022. “Tackling psychosocial and capital constraints to alleviate poverty,” Nature, 605, 1–7. Bossuroy, T., D. Karlan, W. Pariente, P. Premand, C. Udry, J. Vaillant, and K. Wright. 2024. “Impact des mesures d’inclusion productive du programme Burkin Naong Sa Ya au Burkina Faso.” Washington DC : Banque Mondiale. Bossuroy, T., D. Karlan, W. Pariente, P. Premand, C. Udry, J. Vaillant, and K. Wright. 2025a. “Impact des mesures d’inclusion productive du programme Yom Kom Kom au Sénégal.” Washington DC : World Bank. Bossuroy, T., D. Karlan, W. Pariente, P. Premand, C. Udry, J. Vaillant, and K. Wright. 2025b. “Impact des mesures d’inclusion productive du programme Tekavoul en Mauritanie.” Washington DC : World Bank. Botea, I., A. Brudevold-Newman, M. Goldstein, C. Low and G. Roberts. 2023. “Supporting Women’s Livelihoods at Scale: Evidence from a Nationwide Multifaceted Program.” NBER Working Paper No. 31625. Brune, L., N. Goldberg, D. Karlan, D. Parkerson, and C. Udry. 2024. "Randomized Impact Evaluation of a Graduation Program on Livelihoods in Refugee and Host Communities in Uganda." AEA RCT Registry. Carraro, A. and L. Ferrone. 2024. “Feed Thy Neighbour: How Social Ties Shape Spillover Effects of Cash Transfers on Food Security and Nutrition.” Journal of African Economies, Vol. 33(2). Coville, A., D. Jeong, and P. Premand 2022. "Promoting economic inclusion at scale through self and wage employment support." AEA RCT Registry. Crosta, T., D. Karlan, F. Ong, J. Rüschenpöhler and C. Udry. 2024. “Unconditional Cash Transfers: A Bayesian Meta-Analysis of Randomized Evaluations in Low and Middle Income Countries.” NBER Working Paper No. 32779. Cunha, J. M, G. De Giorgi, and S. Jayachandran. 2019. “The Price Effects of Cash Versus In-Kind Transfers.” The Review of Economic Studies 86 (1): 240-281. Daidone, S., B. Davis and S. Handa. 2019. “The Household and Individual-Level Productive Impacts of Cash Transfer Programs in Sub-Saharan Africa.“ American Journal of Agricultural Economics, Vol 101(5) Daye, M, E. Kandpal and P. Schnitzer. 2024. “The effect of cash transfers on women's productive activities and subjective well-being: evidence from an ultra-poor setting.” Mimeo. Della Guardia, Anne & Lake, Milli & Schnitzer, Pascale, 2022. "Selective inclusion in cash transfer programs: Unintended consequences for social cohesion." World Development, Vol. 157(C). 28 Demont, T., 2016, “Microfinance spillovers: A model of competition in informal credit markets with an application to Indian villages.” European Economic Review, Vol. 8: 21-41, Drexler, A., G. Fischer and A. Schoar. 2014. "Keeping It Simple: Financial Literacy and Rules of Thumb." American Economic Journal: Applied Economics 6 (2): 1–31. Dupas, P., A. Keats and J. Robinson, 2019. “The Effect of Savings Accounts on Interpersonal Financial Relationships: Evidence from a Field Experiment in Rural Kenya.” The Economic Journal, Vol 129(617): 273–310. Egger, D., J. Haushofer, E. Miguel, P. Niehaus and M. Walker (2022): “General equilibrium effects of cash transfers: experimental evidence from Kenya.” Econometrica, 90. Fernandez, C., S. Ferrah, A. Guariso, M. Holmlund, T. Mitchell and C. Newman, 2024. "Entrepreneurship support for host and displaced communities." Washington DC : World Bank. Filmer, D., J. Friedman, E. Kandpal and J. Onishi. 2023. “Cash Transfers, Food Prices, and Nutrition Impacts on Ineligible Children.” The Review of Economic and Statistics, Vol 105(2). Franklin, S., C. Imbert, G. Abebe and C. Mejia-Mantilla. 2024. “Urban Public Works in Spatial Equilibrium: Experimental Evidence from Ethiopia.” American Economic Review 114(5): 1382–1414. Haushofer, J., J. Reisinger and J. Shapiro. 2019. “Is Your Gain My Pain? Effects of Relative Income and Inequality on Psychological Well-being.” Mimeo. Imbert, C. and J. Papp. 2015.“Labor Market Effects of Social Programs: Evidence from India’s Employment Guarantee.” American Economic Journal: Applied Economics 7(2): 233–63. Janzen, S., N. Magnan, S. Sharma and T.M. William. 2021. “Pay it Forward: A Mechanism for Achieving Scale in Anti-Poverty Programs.” Mimeo. Kondylis, F. and J. Loeser. 2021. “Intervention Size and Persistence.” Policy Research Working Paper, No. 9769. World Bank, Washington, DC. Leight, J., K. Hirvonen and S. Zafar. 2024. “The effectiveness of cash and cash plus interventions on livelihoods outcomes: Evidence from a systematic review and meta-analysis.” IFPRI Discussion Paper 2262. Washington, DC: International Food Policy Research Institute. Macours, K. and R. Vakis. 2014. “Changing Households’ Investment Behaviour through Social Interactions with Local Leaders: Evidence from a Randomised Transfer Programme?” The Economic Journal, 124 (576), 607-633. Marguerie, A. and P. Premand. 2023. “Savings Facilitation or Capital Injection? Impacts and Spillovers of Livelihood Interventions in Post-Conflict Côte d’Ivoire.” Policy Research Working Papers; 10563. World Bank, Washington, DC. 29 McKenzie, D. and S. Puerto. 2021. "Growing Markets through Business Training for Female Entrepreneurs: A Market-Level Randomized Experiment in Kenya." American Economic Journal: Applied Economics 13 (2): 297–332. Muralidharan, K., P. Niehaus and S. Sukhtankar. 2023. “General Equilibrium Effects of (Improving) Public Employment Programs: Experimental Evidence from India.” Econometrica 91(4): 1261–95. Papineni, S., P. Gonzalez, M. Goldstein and J. Friedman. 2024. “Cash Is Queen: Local Economic Effects of Unconditional Cash Transfers to Women in West Africa.” Mimeo. Premand, P., P. Christian, C. Fernandez, J. Heirman, E. Jospe, J. Waidler, P. Bajpai, O. Eltoukhi, H. Rajaram, H. 2024. “Impact Evaluation of an Integrated Resilience Programme in Niger.” World Food Programme Office of Evaluation. Raza, W. A., E. Van de Poel, and T. Van Ourti. 2018 “Impact and spill-over effects of an asset transfer program on child undernutrition: Evidence from a randomized control trial in Bangladesh”, Journal of Health Economics, Vol. 62: pp 105-120. Ribas, R. P. 2020. Liquidity constraints, spillovers, and entrepreneurship: evidence from a cash transfer program. Small Business Economics 55:4, 1131-1158. Rotemberg, M. 2019. “Equilibrium Effects of Firm Subsidies.” American Economic Review 109 (10): 3475–3513 Sedlmayr, R., A. Shah and M. Sulaiman. 2020. “Cash-plus: Poverty Impacts of Alternative Transfer- Based Approaches.” Journal of Development Economics 144: 102418. Walker, M. W., N. Shah, E. Miguel, D. Egger, F. S. Soliman, and T. Graff. 2024. “Slack and Economic Development” NBER Working Paper No. 33055. Figure 1. Study Design and Follow-up Sample Figure 2 – Timeline Table 1: Balance table - direct impacts and spillovers (1) (2) (1)-(2) (3) (4) (3)-(4) Control Treatment Pairwise T-test Control Pooled Spillover Pairwise T-test Variable Mean/(SE) Mean/(SE) Asymp/[Perm] P-values Mean/(SE) Mean/(SE) Asymp/[Perm] P-values Panel A: Household No. of Household Members 5.921 5.476 0.065* 6.153 5.780 0.059* (0.194) (0.142) [0.088*] (0.154) (0.121) [0.026**] No. of Adult Equivalent 3.750 3.492 0.059* 3.915 3.683 0.041** (0.110) (0.078) [0.115] (0.089) (0.069) [0.016**] Polygamous Household 0.168 0.175 0.874 0.146 0.141 0.867 (0.032) (0.024) [0.941] (0.024) (0.013) [0.875] HH head is also Eligible Individual 0.149 0.208 0.110 0.222 0.239 0.493 (0.023) (0.029) [0.434] (0.016) (0.018) [0.601] Welfare and Food security Z-score -0.000 0.079 0.512 -0.183 -0.055 0.187 (0.084) (0.086) [0.873] (0.069) (0.069) [0.235] Food Security 2.465 1.825 0.042** 2.074 1.592 0.042** (0.255) (0.181) [0.154] (0.189) (0.139) [0.024**] Food Consumption Score 55.611 54.596 0.521 55.433 56.014 0.709 (1.112) (1.130) [0.304] (1.270) (0.914) [0.861] Dietary Diversity Score 7.045 7.154 0.358 6.973 7.016 0.651 (0.082) (0.086) [0.576] (0.069) (0.066) [0.498] Daily Food Consumption (FCFA, ad. equiv.) 361.613 377.728 0.607 318.708 366.863 0.073* 32 (24.081) (20.118) [0.801] (18.997) (18.742) [0.011**] Daily Non-Food Consumption (FCFA, ad. equiv.) 95.687 125.643 0.003*** 85.767 107.827 0.008*** (6.347) (7.724) [0.002***] (5.468) (6.003) [0.006***] Total Daily Consumption (FCFA, ad. equiv.) 458.309 502.142 0.197 404.903 474.574 0.018** (25.336) (22.575) [0.368] (19.717) (21.530) [0.002***] N 202 332 405 701 Panel B: Eligible Individual (woman) Female 0.970 0.982 0.451 0.970 0.970 0.984 (0.013) (0.008) [0.399] (0.010) (0.006) [0.788] Age 24.743 25.605 0.144 32.681 31.015 0.018** (0.407) (0.425) [1.000] (0.529) (0.449) [0.002***] Years of Education 1.782 1.524 0.404 1.279 1.268 0.964 (0.253) (0.179) [0.953] (0.178) (0.163) [0.631] Mental Health Z-Index 0.292 0.204 0.432 0.019 0.005 0.875 (0.088) (0.070) [0.944] (0.072) (0.055) [0.949] Self Efficacy Z-Index -0.015 -0.041 0.737 -0.019 0.011 0.642 (0.061) (0.050) [0.790] (0.043) (0.048) [0.546] Social Standing Z-Index 0.059 -0.028 0.454 -0.034 0.001 0.636 (0.081) (0.083) [0.507] (0.058) (0.046) [0.293] Head Works in Agriculture 0.886 0.916 0.410 0.862 0.901 0.250 (0.030) (0.020) [0.601] (0.029) (0.018) [0.219] No. of Days Worked in Agriculture (per rainy season) 58.545 58.202 0.943 56.879 57.850 0.811 Continued on next page Table 1: Balance table - direct impacts and spillovers – continued from previous page (1) (2) (1)-(2) (3) (4) (3)-(4) Control Treatment Pairwise T-test Control Pooled Spillover Pairwise T-test Variable Mean/(SE) Mean/(SE) Asymp/[Perm] P-values Mean/(SE) Mean/(SE) Asymp/[Perm] P-values (3.920) (2.731) [0.724] (3.300) (2.404) [0.436] Harvest Value (1000 FCFA) 18.234 16.634 0.677 15.414 16.487 0.688 (3.320) (1.975) [0.403] (2.237) (1.482) [0.587] No. of Plots Owned/Managed 0.723 0.723 0.999 0.704 0.795 0.364 (0.087) (0.061) [0.809] (0.080) (0.062) [0.108] Area of Plots Owned/Cultivated (ha) 0.816 0.758 0.714 0.733 0.890 0.240 (0.130) (0.094) [0.773] (0.106) (0.082) [0.050*] Works in Non-Ag Business 0.426 0.431 0.927 0.368 0.364 0.923 (0.043) (0.033) [0.914] (0.033) (0.029) [0.479] No. of Days Worked in Non-Ag Business (per month) 5.163 5.398 0.803 4.620 5.110 0.462 (0.731) (0.595) [0.527] (0.468) (0.477) [0.138] Profits (yearly, 1000 FCFA) 34.929 39.181 0.619 29.380 32.724 0.601 (6.545) (5.545) [0.349] (4.609) (4.470) [0.324] No. of Businesses Owned/Managed. 0.797 0.831 0.733 0.701 0.750 0.548 (0.081) (0.059) [0.561] (0.056) (0.059) [0.133] Business Assets (1000 FCFA) 4.951 5.634 0.509 4.645 5.149 0.575 (0.759) (0.707) [0.192] (0.635) (0.641) [0.301] N 202 332 405 701 Panel C: Household Head (man) - Eligible Individual Excluded 33 Female 0.006 0.004 0.772 0.003 0.002 0.633 (0.006) (0.004) [1.000] (0.003) (0.002) [0.763] Age 29.017 29.848 0.444 37.790 34.618 0.009*** (0.809) (0.725) [1.000] (1.061) (0.564) [0.001***] Years of Education 4.262 3.662 0.316 3.683 3.435 0.641 (0.481) (0.357) [0.350] (0.434) (0.309) [0.430] Works in Agriculture 0.855 0.844 0.797 0.825 0.808 0.656 (0.032) (0.026) [0.576] (0.031) (0.025) [0.204] Harvest Value (1000 FCFA) 75.922 90.411 0.191 70.159 72.213 0.818 (7.128) (8.428) [0.336] (6.381) (6.264) [0.895] No. of Plots Owned/Managed 2.192 2.141 0.820 2.140 2.134 0.981 (0.186) (0.128) [0.889] (0.180) (0.132) [0.811] Area of Plots Owned/Cultivated(ha) 2.905 2.603 0.431 2.701 2.668 0.927 (0.308) (0.229) [0.099*] (0.297) (0.209) [0.493] Works in Non-Ag Business 0.355 0.369 0.767 0.321 0.269 0.164 (0.036) (0.032) [0.953] (0.029) (0.023) [0.063*] No. of Days Worked in Non-Ag Business (per month) 2.337 3.255 0.194 2.606 2.682 0.893 (0.443) (0.548) [0.145] (0.478) (0.305) [0.906] Business Profits (yearly, 1000 FCFA) 28.309 23.802 0.588 26.077 27.184 0.853 (6.333) (5.431) [0.680] (4.793) (3.614) [0.679] No. of Businesses Owned/Managed 0.337 0.357 0.792 0.289 0.385 0.074* (0.047) (0.061) [0.740] (0.037) (0.038) [0.116] Business Assets (1000 FCFA) 4196.221 5759.506 0.422 4339.206 3324.824 0.391 Continued on next page Table 1: Balance table - direct impacts and spillovers – continued from previous page (1) (2) (1)-(2) (3) (4) (3)-(4) Control Treatment Pairwise T-test Control Pooled Spillover Pairwise T-test Variable Mean/(SE) Mean/(SE) Asymp/[Perm] P-values Mean/(SE) Mean/(SE) Asymp/[Perm] P-values (1271.756) (1475.233) [0.377] (968.518) (682.766) [0.356] N 172 263 315 525 Panel D: Response Rate Response Rate at Follow-Up 0.873 0.883 0.705 0.865 0.890 0.294 (0.021) (0.017) [0.586] (0.018) (0.016) [0.298] Response Rate Baseline to Follow-Up 0.906 0.901 0.860 0.899 0.905 0.821 (0.023) (0.019) [0.947] (0.021) (0.017) [0.773] Notes : Standard errors for all tests are clustered at the village level. Fixed effects using randomization strata are included in all estimation regressions. Both asymptotic and permutation p-values are computed for the pairwise tests. The randomization inference tests are performed with 1,000 permutations, preserving both the randomization strata and the cluster structure. *** p < 0.01, ** p < 0.05, * p < 0.1. 34 Table 2: Direct impacts on household welfare Consumption Food Security (1) (2) (3) (4) (5) (6) (7) Total Daily Daily Food Daily Non-Food Food Dietary Food Composite Consumption Consumption Consumption Security Diversity Consumption Z-Score (FCFA, ad. equiv.) (FCFA, ad. equiv.) (FCFA, ad. equiv.) (reversed FIES) Score (HDDS) Score (FCS) Index Treatment 54.193∗∗∗ 37.805∗∗∗ 14.211 -0.184 0.122 2.098 0.168∗ (19.209) (12.956) (9.482) (0.231) (0.093) (1.898) (0.085) Observations 1741 1741 1766 1766 1766 1766 1766 Control mean 394.38 299.70 94.69 2.09 6.43 34.36 0.00 Permutation T-test 3.18∗∗∗ 3.01∗∗∗ 2.55∗∗ -0.71 1.36 1.01 2.06∗ Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. The randomization inference tests are performed with 1,000 permutations, preserving both the randomization strata and the cluster structure. *** p < 0.01, ** p < 0.05, * p < 0.1. All indices are standardized with respect to the control group in that survey round. Components in columns (2) to (6) are used to construct the Composite Z-Score Index (7). 35 Table 3: Direct impacts on agriculture (1) (2) (3) (4) (5) (6) Works in Area of Work Days Work Days Work Days from Harvest Value Agriculture own Plots on on other other Household Member (rainy season, (0,1) cultivated (ha) Own Plots Household Plots on Own Plots 1000 FCFA) Panel A: Eligible Individual (woman) Treatment -0.006 -0.109 -5.697∗∗ -7.470∗∗ -9.299∗ 6.891∗∗ (0.033) (0.486) (2.815) (3.080) (4.916) (3.179) Observations 1766 1766 1766 1766 1766 1713 Control mean 0.73 1.49 26.95 30.38 33.40 20.28 Permutation T-test -0.13 -0.24 -2.11∗ -2.43∗∗ -2.33∗∗ 1.93∗ Panel B: Household Head (man) Treatment -0.041 0.061 -9.338∗∗ -0.027 -29.039∗∗∗ 9.161 (0.030) (0.328) (3.996) (0.666) (10.339) (7.773) Observations 1487 1487 1487 1487 1487 1441 Control mean 0.72 2.87 63.97 1.66 96.81 87.05 Permutation T-test -1.39 0.10 -2.37∗∗ -0.04 -3.09∗∗∗ 1.73 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline 36 outcomes. Robust standard errors are shown in parentheses, clustered at the village level. The randomization inference tests are performed with 1,000 permutations, preserving both the randomization strata and the cluster structure. *** p < 0.01, ** p < 0.05, * p < 0.1. Table 4: Direct impacts on off-farm businesses (1) (2) (3) (4) (5) (6) (7) (8) (9) Works in No. of Days in Non-Ag Business Business No. of Months Business Business Bought Non-Ag Non-Ag Business Revenues (yearly, Profits (yearly, Business in Assets Closure Inputs from Business (0,1) Businesses (monthly) 1000 FCFA) 1000 FCFA) Operation (1000 FCFA) (0,1) Regional Market Panel A: Eligible Individual (woman) Treatment 0.024 0.073 1.230 31.077∗ 7.075 0.700 0.879 0.000 0.067∗∗ (0.044) (0.058) (0.750) (18.500) (6.123) (0.474) (0.714) (0.023) (0.031) Observations 1766 1766 1766 1766 1766 1766 1766 1766 1766 Control mean 0.60 0.94 7.05 113.07 40.15 6.04 4.59 0.14 0.17 Permutation T-test 0.54 1.28 1.61 1.78 1.17 1.63 1.31 0.02 2.14∗ Panel B: Household Head (man) Treatment 0.039 0.015 0.435 14.727 9.057 0.276 159.374 0.014∗∗ 0.026 (0.038) (0.039) (0.464) (12.748) (5.997) (0.308) (522.353) (0.007) (0.017) Observations 1487 1487 1487 1487 1487 1487 1487 1487 1487 Control mean 0.32 0.23 2.15 48.69 18.36 1.80 2081.44 0.01 0.05 Permutation T-test 1.08 0.41 1.04 1.39 1.48 0.90 0.41 2.07∗∗ 1.59 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. The randomization inference tests are performed with 1,000 permutations, preserving 37 both the randomization strata and the cluster structure. *** p < 0.01, ** p < 0.05, * p < 0.1. Table 5: Direct impacts on livestock and savings Livestock Savings (1) (2) (3) (4) (5) (6) Works in Days Spent Livestock Livestock Member of Total Savings Livestock Raising Livestock Sale Revenue Count Savings Group (yearly, (0,1) (monthly) (1000 FCFA) (TLU) (0,1) 1000 FCFA) Panel A: Eligible Individual (woman) Treatment 0.008 1.061 -0.140 -0.002 0.161∗∗∗ 2.253∗ (0.045) (0.821) (0.808) (0.043) (0.041) (1.160) Observations 1766 1766 1766 1766 1766 1766 Control mean 0.52 5.42 3.99 0.20 0.37 3.53 Permutation T-test 0.24 1.29 -0.17 -0.04 3.89∗∗∗ 1.55 Panel B: Household Head (man) Treatment -0.005 0.086 -6.849∗∗ 0.037 (0.037) (0.897) (3.056) (0.049) Observations 1487 1487 1487 1487 Control mean 0.52 10.97 17.39 0.30 38 Permutation T-test -0.13 0.10 -2.24∗∗ 0.77 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. The random- ization inference tests are performed with 1,000 permutations, preserving both the randomization strata and the cluster structure. *** p < 0.01, ** p < 0.05, * p < 0.1. TLU represents Tropical Livestock Units. Table 6: Direct impacts of women’s empowerment and psychosocial well-being (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Beneficiary Control over Intra-Household Financial Social Social Collective Social Mental Self Future Share of Total HH Resources Dynamics Support Support Standing Action Cohesion Health Efficacy Expectations Revenues (%) Z-Index Index Index Index Index Index Index Index Index Index Treatment 7.302∗∗ 0.115∗ 0.145 0.256∗∗∗ 0.083 0.171∗ 0.226∗∗∗ -0.036 0.016 0.050 0.056 (3.051) (0.066) (0.088) (0.064) (0.067) (0.099) (0.074) (0.085) (0.124) (0.058) (0.088) Observations 1766 1740 1766 1766 1766 1766 1766 1766 1766 1766 1373 Control mean 46.34 0.00 -0.07 -0.04 -0.05 -0.07 0.02 -0.03 0.10 0.07 0.09 Permutation T-test 2.88∗∗∗ 1.83 1.66 4.00∗∗∗ 1.24 1.59 3.06∗∗ -0.42 0.04 0.87 0.65 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. The randomization inference tests are performed with 1,000 permutations, preserving both the randomization strata and the cluster structure. *** p < 0.01, ** p < 0.05, * p < 0.1. All indices are standardized with respect to the control group in that survey round. Results from components of each index are provided in Table A14 (control over household resources index components), Table A15 (intra-household dynamics index components), Table A16 (financial support index components), Table A17 (social support index components), Table A18 (social standing index components), Table A19 (collective action index components), Table A20 (social cohesion and community closeness index components), Table A21 (mental health index components), Table A22 (self-efficacy index components), and Table A23 (future expectations index components). 39 Table 7: Spillovers on household welfare Consumption Food Security (1) (2) (3) (4) (5) (6) (7) Total Daily Daily Food Daily Non-Food Food Dietary Food Composite Consumption Consumption Consumption Security Diversity Consumption Z-Score (FCFA, ad. equiv.) (FCFA, ad. equiv.) (FCFA, ad. equiv.) (reversed FIES) Score (HDDS) Score (FCS) Index Pooled Spillover 35.903∗ 26.302∗ 10.357 -0.033 0.152∗ 3.631∗∗ 0.174∗∗ (18.845) (14.255) (8.635) (0.223) (0.082) (1.733) (0.077) Observations 2603 2603 2650 2650 2650 2650 2650 Control mean 381.59 295.53 85.99 1.95 6.34 33.15 -0.10 Permutation T-test 2.43∗∗ 2.16∗∗ 1.79∗ -0.29 1.92∗ 2.10∗ 2.43∗∗ Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. The randomization inference tests are performed with 1,000 permutations, preserving both the randomization strata and the cluster structure. *** p < 0.01, ** p < 0.05, * p < 0.1. All indices are standardized with respect to the control group in that survey round. Components in columns (2) to (6) are used to construct the Composite Z-Score Index (7). 40 Table 8: Spillovers impacts on agriculture (1) (2) (3) (4) (5) (6) Works in Area of Work Days Work Days Work Days from Harvest Value Agriculture own Plots on on other other Household Member (rainy season, (0,1) cultivated (ha) Own Plots Household Plots on Own Plots 1000 FCFA) Panel A: Eligible Individual (woman) Pooled Spillover 0.017 -0.280 -6.035∗∗ -4.031 -9.796∗ 1.555 (0.032) (0.437) (2.821) (2.468) (4.949) (2.677) Observations 2650 2650 2650 2650 2650 2570 Control mean 0.70 1.42 27.17 26.16 35.49 20.25 Permutation T-test 0.62 -0.61 -2.16∗ -1.63 -2.02∗ 0.56 Panel B: Household Head (man) Pooled Spillover -0.023 -0.394∗ -5.955 0.024 -24.460∗∗∗ 14.269∗∗ (0.028) (0.224) (3.583) (0.592) (8.678) (6.943) Observations 2095 2095 2095 2095 2095 2028 Control mean 0.70 2.85 58.97 1.75 99.66 79.70 Permutation T-test -0.79 -1.69 -1.71 0.04 -2.98∗∗∗ 2.17∗ Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline 41 outcomes. Robust standard errors are shown in parentheses, clustered at the village level. The randomization inference tests are performed with 1,000 permutations, preserving both the randomization strata and the cluster structure. *** p < 0.01, ** p < 0.05, * p < 0.1. Table 9: Spillovers impacts on off-farm businesses (1) (2) (3) (4) (5) (6) (7) (8) (9) Works in No. of Days in Non-Ag Business Business No. of Months Business Business Bought Non-Ag Non-Ag Business Revenues (yearly, Profits (yearly, Business in Assets Closure Inputs from Business (0,1) Businesses (monthly) 1000 FCFA) 1000 FCFA) Operation (1000 FCFA) (0,1) Regional Market Panel A: Eligible Individual (woman) Pooled Spillover 0.009 0.065 2.433∗∗∗ 42.420∗∗ 13.677∗∗ 0.938∗ 0.832 -0.029 0.034 (0.044) (0.065) (0.801) (16.519) (5.861) (0.508) (0.703) (0.019) (0.030) Observations 2650 2650 2650 2650 2650 2650 2650 2650 2650 Control mean 0.55 0.86 6.51 97.66 34.98 5.44 4.26 0.14 0.15 Permutation T-test 0.19 1.14 3.08∗∗∗ 2.64∗∗∗ 2.35∗∗ 1.99∗ 1.27 -1.47 1.13 Panel B: Household Head (man) Pooled Spillover 0.037 -0.020 0.478 11.581 3.815 0.093 173.503 -0.005 0.034∗∗ (0.034) (0.035) (0.373) (10.853) (4.064) (0.294) (511.970) (0.004) (0.015) Observations 2095 2095 2095 2095 2095 2095 2095 2095 2095 Control mean 0.30 0.22 2.02 44.36 17.46 1.70 1853.82 0.01 0.05 Permutation T-test 1.06 -0.40 1.21 1.34 0.93 0.35 0.15 -1.19 2.23∗∗ Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. The randomization inference tests are performed with 1,000 permutations, preserving 42 both the randomization strata and the cluster structure. *** p < 0.01, ** p < 0.05, * p < 0.1. Table 10: Spillovers impacts on livestock and savings Livestock Savings (1) (2) (3) (4) (5) (6) Works in Days Spent Livestock Livestock Member of Total Savings Livestock Raising Livestock Sale Revenue Count Savings Group (yearly, (0,1) (monthly) (1000 FCFA) (TLU) (0,1) 1000 FCFA) Panel A: Eligible Individual (woman) Pooled Spillover 0.034 1.653∗∗ 0.751 -0.005 0.098∗∗∗ 2.065∗∗ (0.040) (0.791) (0.883) (0.049) (0.034) (0.907) Observations 2650 2650 2650 2650 2650 2650 Control mean 0.46 5.18 3.88 0.19 0.30 3.19 Permutation T-test 0.76 2.09∗ 0.85 -0.09 2.86∗∗ 2.35∗∗ Panel B: Household Head (man) Pooled Spillover 0.030 0.688 -6.321∗∗ 0.079 (0.034) (0.782) (2.467) (0.064) Observations 2095 2095 2095 2095 Control mean 0.46 9.28 15.94 0.29 43 Permutation T-test 0.78 0.88 -2.56∗∗ 1.24 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. The random- ization inference tests are performed with 1,000 permutations, preserving both the randomization strata and the cluster structure. *** p < 0.01, ** p < 0.05, * p < 0.1. TLU represents Tropical Livestock Units. Table 11: Spillovers on women’s empowerment and psychosocial well-being (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Beneficiary Control over Intra-Household Financial Social Social Collective Social Mental Self Future Share of Total HH Resources Dynamics Support Support Standing Action Cohesion Health Efficacy Expectations Revenues (%) Z-Index Index Index Index Index Index Index Index Index Index Pooled Spillover 5.731∗ 0.115∗∗ 0.083 0.199∗∗∗ 0.061 0.146 0.120∗ 0.024 0.004 0.002 0.110 (2.954) (0.051) (0.084) (0.071) (0.073) (0.103) (0.066) (0.085) (0.115) (0.065) (0.079) Observations 2650 2587 2650 2650 2650 2650 2650 2650 2650 2650 1914 Control mean 47.28 0.04 -0.04 -0.08 -0.01 -0.05 -0.03 -0.02 0.01 0.02 -0.04 Permutation T-test 2.20∗∗ 2.44∗∗ 0.99 2.81∗∗ 0.84 1.42 1.83∗ 0.28 -0.04 0.01 1.49 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. The randomization inference tests are performed with 1,000 permutations, preserving both the randomization strata and the cluster structure. *** p < 0.01, ** p < 0.05, * p < 0.1. All indices are standardized with respect to the control group in that survey round. Results from components of each index are provided in Table A14 (control over household resources index components), Table A15 (intra-household dynamics index components), Table A16 (financial support index components), Table A17 (social support index components), Table A18 (social standing index components), Table A19 (collective action index components), Table A20 (social cohesion and community closeness index components), Table A21 (mental health index components), Table A22 (self-efficacy index components), and Table A23 (future expectations index components). 44 Table 12: Cost-Benefit Analysis Treatment Effect Only Treatment and Spillover Panel 1: Program costs per beneficiary, 1000 FCFA Coordination 10 10 Field Operations 4 4 Cash Grant Transfer fees 3 3 Cash Grant 45 45 Total costs, calculated as if all incurred immediately at beginning of year 0 61 61 (1) Total costs, inflated to year 1 at 5% annual discount rate 64 64 Panel 2: Benefits per household, 1000 FCFA, all values inflated to year 1 at 5% annual social discount rate (2) Year 1 food consumption treatment effect 50 106 (3) B1: Year 2 onward food consumption treatment effect, assuming dissipation of 75% 15 32 B2: Year 2 onward food consumption treatment effect, assuming dissipation of 50% 43 92 B3: Year 2 onward food consumption treatment effect, assuming dissipation of 25% 119 253 (4) C: Year 2 onward food consumption treatment effect, assuming year 1 gains persist in perpetuity 904 1914 (5) A: Total benefits: (2) = (5), 5% discount rate, no impact after year 1 50 106 (6) B1: Total benefits: (2) + (3) = (6), 5% discount rate, 75% annual dissipation 65 138 B2: Total benefits, 5% discount rate, 50% annual dissipation 94 198 B3: Total benefits, 5% discount rate, 25% annual dissipation 169 359 (7) C: Total benefits: (2) + (4) = (7), 5% discount rate, assuming year 1 gains persist in perpetuity 954 2020 Panel 3: Benefit/cost ratios (8) A: Benefit/cost ratio: (5) / (1) = (8), 5% discount rate 78% 165% A: Benefit/cost ratio, 7% discount rate 76% 162% A: Benefit/cost ratio, 10% discount rate 74% 157% (9) B1: Benefit/cost ratio: (6) / (1) = (9), 5% discount rate, 75% annual dissipation 101% 214% B2: Benefit/cost ratio, 5% discount rate, 50% annual dissipation 145% 307% B3: Benefit/cost ratio, 5% discount rate, 25% annual dissipation 263% 557% (10) C: Benefit/cost ratio: (7) / (1) = (10), 5% discount rate, assuming year 1 gains persist in perpetuity 1480% 3135% Panel 4: Real internal rate of return (IRR) A: Assuming dissipation of 100% after year 2, at 5% discount rate -18% 73% B1: Assuming annual dissipation of 75% 7% 98% B2: Assuming annual dissipation of 50% 32% 123% B3: Assuming annual dissipation of 25% 57% 148% C: Assuming effects are sustained in perpetuity 82% 173% 45 Appendix 46 Table A1: Balance table - decomposition of spillovers (1) (2) (1)-(2) (3) (4) (3)-(4) Control CT Spillover Pairwise T-test Control Non-CT Spillover Pairwise T-test Variable Mean/(SE) Mean/(SE) Asymp/[Perm] P-values Mean/(SE) Mean/(SE) Asymp/[Perm] P-values Panel A: Household No. of Household Members 5.921 5.532 0.091* 6.384 6.122 0.334 (0.194) (0.123) [0.115] (0.184) (0.199) [0.223] No. of Adult Equivalent 3.750 3.517 0.077* 4.080 3.912 0.280 (0.110) (0.070) [0.091*] (0.107) (0.113) [0.164] Polygamous Household 0.168 0.129 0.286 0.123 0.158 0.271 (0.032) (0.018) [0.405] (0.024) (0.020) [0.440] HH head is also Eligible Individual 0.149 0.189 0.223 0.296 0.307 0.759 (0.023) (0.024) [0.385] (0.030) (0.023) [0.561] Welfare and Food security Z-score -0.000 0.040 0.732 -0.365 -0.185 0.081* (0.084) (0.083) [0.835] (0.078) (0.067) [0.090*] Food Security 2.465 1.708 0.015** 1.685 1.433 0.296 (0.255) (0.173) [0.009***] (0.195) (0.141) [0.294] Food Consumption Score 55.611 55.042 0.707 55.256 57.355 0.303 (1.112) (1.034) [0.509] (1.640) (1.214) [0.399] Dietary Diversity Score 7.045 7.120 0.519 6.901 6.872 0.806 (0.082) (0.084) [0.355] (0.085) (0.084) [0.911] Daily Food Consumption (FCFA, ad. equiv.) 361.613 391.588 0.349 276.229 332.047 0.038** 47 (24.081) (21.148) [0.361] (17.908) (19.655) [0.006***] Daily Non-Food Consumption (FCFA, ad. equiv.) 95.687 115.525 0.037** 75.895 97.217 0.018** (6.347) (6.967) [0.017**] (6.545) (6.069) [0.012**] Total Daily Consumption (FCFA, ad. equiv.) 458.309 507.176 0.169 352.029 428.667 0.011** (25.336) (24.803) [0.148] (20.294) (21.710) [0.002***] N 202 333 203 368 Panel B: Eligible Individual (woman) Female 0.970 0.982 0.431 0.970 0.954 0.336 (0.013) (0.007) [0.331] (0.013) (0.011) [0.326] Age 24.743 25.105 0.508 40.581 39.160 0.235 (0.407) (0.368) [0.590] (0.933) (0.751) [0.220] Years of Education 1.782 1.474 0.329 0.778 0.984 0.372 (0.253) (0.190) [0.378] (0.165) (0.161) [0.207] Mental Health Z-Index 0.292 0.070 0.044** -0.252 -0.084 0.115 (0.088) (0.065) [0.082*] (0.088) (0.060) [0.151] Self Efficacy Z-Index -0.015 0.008 0.773 -0.024 0.014 0.706 (0.061) (0.053) [0.642] (0.065) (0.077) [0.592] Social Standing Z-Index 0.059 -0.072 0.175 -0.126 0.102 0.015** (0.081) (0.051) [0.445] (0.068) (0.062) [0.006***] Head Works in Agriculture 0.886 0.922 0.311 0.837 0.872 0.380 (0.030) (0.019) [0.241] (0.033) (0.022) [0.522] No. of Days Worked in Agriculture (per rainy season) 58.545 59.562 0.824 55.222 55.492 0.956 Continued on next page Table A1: Balance table - decomposition of spillovers – continued from previous page (1) (2) (1)-(2) (3) (4) (3)-(4) Control CT Spillover Pairwise T-test Control Non-CT Spillover Pairwise T-test Variable Mean/(SE) Mean/(SE) Asymp/[Perm] P-values Mean/(SE) Mean/(SE) Asymp/[Perm] P-values (3.920) (2.395) [0.604] (3.699) (3.303) [0.566] Harvest Value (1000 FCFA) 18.234 14.450 0.322 12.578 19.312 0.044** (3.320) (1.904) [0.258] (2.435) (2.252) [0.014**] No. of Plots Owned/Managed 0.723 0.775 0.646 0.685 0.823 0.226 (0.087) (0.073) [0.430] (0.090) (0.070) [0.055*] Area of Plots Owned/Cultivated (ha) 0.816 0.863 0.778 0.650 0.927 0.053* (0.130) (0.108) [0.475] (0.107) (0.094) [0.009***] Works in Non-Ag Business 0.426 0.426 0.990 0.310 0.277 0.465 (0.043) (0.037) [0.545] (0.035) (0.029) [0.678] No. of Days Worked in Non-Ag Business (per month) 5.163 5.664 0.591 4.079 4.348 0.728 (0.731) (0.583) [0.292] (0.570) (0.525) [0.503] Profits (yearly, 1000 FCFA) 34.929 34.850 0.992 23.858 29.794 0.426 (6.545) (5.065) [0.576] (4.711) (5.790) [0.377] No. of Businesses Owned/Managed. 0.797 0.829 0.767 0.606 0.641 0.649 (0.081) (0.070) [0.328] (0.051) (0.059) [0.409] Business Assets (1000 FCFA) 4.951 4.898 0.960 4.340 5.495 0.316 (0.759) (0.761) [0.784] (0.766) (0.859) [0.164] N 202 333 203 368 Panel C: Household Head (man) - Eligible Individual Excluded 48 Female 0.006 0.000 0.317 0.000 0.004 0.319 (0.006) (0.000) [0.086*] (0.000) (0.004) [0.365] Age 29.017 29.044 0.978 48.343 43.608 0.005*** (0.809) (0.544) [0.562] (1.404) (0.845) [0.001***] Years of Education 4.262 3.622 0.293 2.986 3.133 0.801 (0.481) (0.372) [0.216] (0.463) (0.360) [0.945] Works in Agriculture 0.855 0.815 0.360 0.790 0.796 0.909 (0.032) (0.029) [0.200] (0.042) (0.030) [0.359] Harvest Value (1000 FCFA) 75.922 76.154 0.983 63.070 65.757 0.799 (7.128) (8.405) [0.866] (8.199) (6.682) [0.554] No. of Plots Owned/Managed 2.192 2.174 0.939 2.077 2.071 0.980 (0.186) (0.144) [0.930] (0.207) (0.150) [0.598] Area of Plots Owned/Cultivated(ha) 2.905 2.642 0.478 2.457 2.711 0.543 (0.308) (0.210) [0.186] (0.327) (0.263) [0.463] Works in Non-Ag Business 0.355 0.304 0.263 0.280 0.212 0.141 (0.036) (0.028) [0.223] (0.037) (0.027) [0.034**] No. of Days Worked in Non-Ag Business (per month) 2.337 2.896 0.343 2.930 2.337 0.483 (0.443) (0.390) [0.336] (0.738) (0.419) [0.367] Business Profits (yearly, 1000 FCFA) 28.309 29.660 0.864 23.394 23.191 0.980 (6.333) (4.800) [0.652] (6.472) (4.957) [0.875] No. of Businesses Owned/Managed 0.337 0.433 0.137 0.231 0.306 0.201 (0.047) (0.044) [0.187] (0.038) (0.045) [0.353] Business Assets (1000 FCFA) 4196.221 3184.711 0.525 4511.189 3550.816 0.549 Continued on next page Table A1: Balance table - decomposition of spillovers – continued from previous page (1) (2) (1)-(2) (3) (4) (3)-(4) Control CT Spillover Pairwise T-test Control Non-CT Spillover Pairwise T-test Variable Mean/(SE) Mean/(SE) Asymp/[Perm] P-values Mean/(SE) Mean/(SE) Asymp/[Perm] P-values (1271.756) (960.503) [0.592] (1408.772) (773.512) [0.389] N 172 270 143 255 Panel D: Response Rate Response Rate at Follow-Up 0.873 0.906 0.229 0.853 0.865 0.619 (0.021) (0.018) [0.230] (0.019) (0.018) [0.628] Response Rate Baseline to Follow-Up 0.906 0.913 0.832 0.892 0.894 0.943 (0.023) (0.023) [0.885] (0.028) (0.019) [0.791] Notes : Standard errors for all tests are clustered at the village level. Fixed effects using randomization strata are included in all estimation regressions. Both asymptotic and permutation p-values are computed for the pairwise tests. The randomization inference tests are performed with 1,000 permutations, preserving both the randomization strata and the cluster structure. *** p < 0.01, ** p < 0.05, * p < 0.1. 49 Table A2: Direct impacts on standardized welfare indices Consumption Food Security Indicators (1) (2) (3) (4) (5) (6) Daily Food Daily Non-Food Food Security Consumption Dietary Composite Consumption Consumption (reversed FIES) Score (FCS) Diversity Score Z-Score Z-Index Z-Index Z-Index Z-Index (HDDS) Z-Index Index Treatment 0.162∗∗∗ 0.138 -0.077 0.118 0.100 0.168∗ (0.056) (0.092) (0.097) (0.098) (0.076) (0.085) Observations 1741 1766 1766 1766 1766 1766 Control mean 0.00 0.00 0.00 0.00 0.00 0.00 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. All indices are standardized with respect to the control group in that survey round. Components in columns (2) to (6) are used to construct the Composite Z-Score Index (7). 50 Table A3: Direct impacts on agricultural inputs (1) (2) (3) (4) Purchased Used Used Used seeds chemical phytosanitary Paid (0,1) fertilizer (0,1) products (0,1) Labor (0,1) Treatment -0.098∗∗∗ -0.006 0.002 0.078∗∗ (0.029) (0.010) (0.008) (0.037) Observations 1766 1766 1766 1766 Control mean 0.37 0.03 0.02 0.21 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. 51 Table A4: Direct impacts on salaried employment All Salaried Employment Agricultural Employment (1) (2) (3) (4) (5) (6) Works in Days Worked Wage Works in Days Worked Wage Salaried in Salaried Earnings (yearly, Salaried in Salaried Earnings (yearly, Employment (0,1) Employment 1000 FCFA) employment Employment 1000 FCFA) Panel A: Eligible Individual (woman) Treatment 0.019 0.195∗∗ 0.022 0.016 0.092 0.507 (0.018) (0.085) (1.156) (0.016) (0.069) (0.548) Observations 1766 1766 1766 1766 1766 1766 Control mean 0.07 0.24 3.20 0.04 0.13 1.02 Panel B: Household Head (man) Treatment -0.000 0.051 -0.679 0.013 0.140∗ 0.699 (0.019) (0.161) (3.431) (0.012) (0.082) (0.659) Observations 1487 1487 1487 1487 1487 1487 Control mean 0.11 0.75 13.66 0.03 0.15 1.47 52 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. Table A5: Spillovers impacts on standardized welfare indices Consumption Food Security Indicators (1) (2) (3) (4) (5) (6) Daily Food Daily Non-Food Food Security Consumption Dietary Composite Consumption Consumption (reversed FIES) Score (FCS) Diversity Score Z-Score Z-Index Z-Index Z-Index Z-Index (HDDS) Z-Index Index Pooled Spillover 0.113∗ 0.100 -0.014 0.199∗∗ 0.124∗ 0.174∗∗ (0.061) (0.084) (0.094) (0.088) (0.067) (0.077) Observations 2603 2650 2650 2650 2650 2650 Control mean -0.02 -0.08 -0.06 -0.06 -0.08 -0.10 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. All indices are standardized with respect to the control group in that survey round. Components in columns (2) to (6) are used to construct the Composite Z-Score Index (7). 53 Table A6: Spillovers impacts on agricultural inputs (1) (2) (3) (4) Purchased Used Used Used seeds chemical phytosanitary Paid (0,1) fertilizer (0,1) products (0,1) Labor (0,1) Pooled Spillover -0.087∗∗∗ -0.005 -0.000 0.049∗ (0.026) (0.009) (0.008) (0.026) Observations 2650 2650 2650 2650 Control mean 0.36 0.03 0.02 0.18 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. 54 Table A7: Spillovers impacts on salaried employment All Salaried Employment Agricultural Employment (1) (2) (3) (4) (5) (6) Works in Days Worked Wage Works in Days Worked Wage Salaried in Salaried Earnings (yearly, Salaried in Salaried Earnings (yearly, Employment (0,1) Employment 1000 FCFA) employment Employment 1000 FCFA) Panel A: Eligible Individual (woman) Pooled Spillover 0.015 0.174∗ 0.998 0.010 0.069 0.702 (0.016) (0.097) (1.307) (0.013) (0.070) (0.541) Observations 2650 2650 2650 2650 2650 2650 Control mean 0.06 0.24 3.03 0.03 0.14 0.97 Panel B: Household Head (man) Pooled Spillover 0.018 0.132 4.925 0.015 0.135∗ 1.441 (0.019) (0.160) (4.422) (0.012) (0.079) (0.888) Observations 2095 2095 2095 2095 2095 2095 Control mean 0.10 0.68 11.97 0.03 0.13 1.35 55 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. Table A8: Decomposition of spillovers on household welfare Consumption Food Security (1) (2) (3) (4) (5) (6) (7) Total Daily Daily Food Daily Non-Food Food Dietary Food Composite Consumption Consumption Consumption Security Diversity Consumption Z-Score (FCFA, ad. equiv.) (FCFA, ad. equiv.) (FCFA, ad. equiv.) (reversed FIES) Score (HDDS) Score (FCS) Index CT Spillover 45.358∗∗ 34.181∗∗ 10.182 -0.016 0.134 3.355∗ 0.188∗∗ (20.522) (14.632) (9.877) (0.232) (0.088) (1.862) (0.084) Observations 1488 1488 1503 1503 1503 1503 1503 Control mean 394.38 299.70 94.69 2.09 6.43 34.36 0.00 Non-CT Spillover 21.118 14.022 12.203 -0.102 0.184∗ 4.219∗∗ 0.162∗ (20.989) (17.129) (8.718) (0.251) (0.096) (1.895) (0.084) Observations 1115 1115 1147 1147 1147 1147 1147 Control mean 360.74 288.73 72.14 1.72 6.19 31.21 -0.25 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. All indices are standardized with respect to the control group in that survey round. Components in columns (2) to (6) are used to construct the Composite Z-Score Index (7). 56 Table A9: Decomposition of spillovers on standardized welfare indices Consumption Food Security Indicators (1) (2) (3) (4) (5) (6) Daily Food Daily Non-Food Food Security Consumption Dietary Composite Consumption Consumption (reversed FIES) Score (FCS) Diversity Score Z-Score Z-Index Z-Index Z-Index Z-Index (HDDS) Z-Index Index CT Spillover 0.146∗∗ 0.099 -0.007 0.184∗ 0.109 0.188∗∗ (0.063) (0.096) (0.097) (0.095) (0.072) (0.084) Observations 1488 1503 1503 1503 1503 1503 Control mean 0.00 0.00 0.00 0.00 0.00 0.00 Non-CT Spillover 0.060 0.118 -0.043 0.232∗∗ 0.150∗ 0.162∗ (0.073) (0.084) (0.105) (0.096) (0.079) (0.084) Observations 1115 1147 1147 1147 1147 1147 Control mean -0.05 -0.22 -0.16 -0.17 -0.19 -0.25 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. All indices are 57 standardized with respect to the control group in that survey round. Components in columns (2) to (6) are used to construct the Composite Z-Score Index (7). Table A10: Food Consumption Components (1) (2) (3) Daily Self-Produced Daily Purchased Daily Gifted Food Consumption Food Consumption Food Consumption (FCFA, ad. equiv.) (FCFA, ad. equiv.) (FCFA, ad. equiv.) Treatment 8.901∗ 13.424 15.541 (4.727) (13.606) (11.555) Observations 1745 1745 1745 Control mean 37.79 229.80 32.12 Pooled Spillover 6.488 1.321 21.514 (4.827) (18.239) (13.164) Observations 2606 2606 2606 Control mean 35.69 227.86 31.56 CT Spillover 5.461 3.824 26.915∗ (5.514) (18.784) (14.986) Observations 1488 1488 1488 Control mean 37.79 229.80 32.12 Non-CT Spillover 7.897 -1.222 12.911 (5.746) (20.502) (12.430) Observations 1118 1118 1118 Control mean 32.29 224.71 30.66 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. 58 Table A11: Decomposition of spillovers impacts on agriculture (1) (2) (3) (4) (5) (6) Works in Area of Work Days Work Days Work Days from Harvest Value Agriculture own Plots on on other other Household Member (rainy season, (0,1) cultivated (ha) Own Plots Household Plots on Own Plots 1000 FCFA) Panel A: Eligible Individual (woman) CT Spillover 0.031 -0.143 -4.868 -6.129∗∗ -8.398 1.381 (0.033) (0.560) (3.320) (2.898) (5.407) (3.006) Observations 1503 1503 1503 1503 1503 1463 Control mean 0.73 1.49 26.95 30.38 33.40 20.28 Non-CT Spillover -0.004 -0.501 -8.068∗∗∗ -0.553 -12.527∗∗ 0.465 (0.041) (0.304) (2.703) (2.687) (5.938) (3.151) Observations 1147 1147 1147 1147 1147 1107 Control mean 0.66 1.30 27.52 19.44 38.83 20.21 Panel B: Household Head CT Spillover -0.052 -0.311 -8.457∗∗ 0.301 -32.929∗∗∗ 8.236 (0.034) (0.238) (3.779) (0.724) (9.087) (8.228) 59 Observations 1293 1293 1293 1293 1293 1257 Control mean 0.72 2.87 63.97 1.66 96.81 87.05 Non-CT Spillover 0.034 -0.583 -1.599 -0.514 -10.862 26.640∗∗∗ (0.035) (0.396) (4.443) (0.888) (12.615) (8.124) Observations 802 802 802 802 802 771 Control mean 0.67 2.81 49.31 1.91 105.16 65.12 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. Table A12: Decomposition of spillovers impacts on off-farm businesses (1) (2) (3) (4) (5) (6) (7) (8) (9) Works in No. of Days in Non-Ag Business Business No. of Months Business Business Bought Non-Ag Non-Ag Business Revenues (yearly, Profits (yearly, Business in Assets Closure Inputs from Business (0,1) Businesses (monthly) 1000 FCFA) 1000 FCFA) Operation (1000 FCFA) (0,1) Regional Market Panel A: Eligible Individual (woman) CT Spillover -0.011 0.047 2.815∗∗∗ 35.392∗ 12.451∗ 0.660 0.898 -0.025 0.027 (0.048) (0.068) (0.948) (20.148) (7.152) (0.556) (0.827) (0.023) (0.036) Observations 1503 1503 1503 1503 1503 1503 1503 1503 1503 Control mean 0.60 0.94 7.05 113.07 40.15 6.04 4.59 0.14 0.17 Non-CT Spillover 0.038 0.098 1.832∗∗ 54.184∗∗∗ 15.320∗∗∗ 1.390∗∗∗ 0.797 -0.036 0.045 (0.046) (0.072) (0.825) (14.854) (5.331) (0.508) (0.731) (0.025) (0.031) Observations 1147 1147 1147 1147 1147 1147 1147 1147 1147 Control mean 0.48 0.72 5.66 73.12 26.75 4.47 3.72 0.14 0.13 Panel B: Household Head CT Spillover 0.025 -0.015 0.330 9.602 3.237 0.063 -34.040 0.001 0.039∗∗ (0.040) (0.041) (0.464) (13.107) (4.719) (0.323) (704.788) (0.006) (0.019) Observations 1293 1293 1293 1293 1293 1293 1293 1293 1293 Control mean 0.32 0.23 2.15 48.69 18.36 1.80 2081.44 0.01 0.05 60 Non-CT Spillover 0.064∗ -0.026 0.838∗ 16.353 5.656 0.193 667.431 -0.015∗∗ 0.027 (0.038) (0.041) (0.490) (15.105) (5.846) (0.378) (586.075) (0.007) (0.019) Observations 802 802 802 802 802 802 802 802 802 Control mean 0.24 0.19 1.78 36.00 15.71 1.49 1414.43 0.02 0.04 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. Table A13: Decomposition of spillovers on livestock and savings Livestock Savings (1) (2) (3) (4) (5) (6) Works in Days Spent Livestock Livestock Member of Total Savings Livestock Raising Livestock Sale Revenue Count Savings Group (yearly, (0,1) (monthly) (1000 FCFA) (TLU) (0,1) 1000 FCFA) Panel A: Eligible Individual (woman) CT Spillover 0.004 2.061∗∗ 0.194 -0.039 0.073∗ 2.522 (0.046) (1.009) (1.068) (0.049) (0.040) (1.827) Observations 1503 1503 1503 1503 1503 1503 Control mean 0.52 5.42 3.99 0.20 0.37 3.53 Non-CT Spillover 0.082∗ 1.019 1.594 0.050 0.136∗∗∗ 4.635∗∗∗ (0.043) (0.741) (1.212) (0.062) (0.036) (1.195) 61 Observations 1147 1147 1147 1147 1147 1147 Control mean 0.37 4.79 3.70 0.19 0.19 2.65 Panel B: Household Head CT Spillover -0.004 -0.315 -9.047∗∗∗ 0.036 (0.038) (0.840) (2.965) (0.064) Observations 1293 1293 1293 1293 Control mean 0.52 10.97 17.39 0.30 Non-CT Spillover 0.093∗∗ 2.596∗∗∗ -0.854 0.166 (0.043) (0.967) (4.236) (0.100) Observations 802 802 802 802 Control mean 0.35 6.03 13.16 0.26 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. TLU represents Tropical Livestock Units. Table A14: Control over household resources index components (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Controls HH Daily Large Family Own Partner’s Child Daily Large Purchases Family Planning Own Healthcare Resources Spending Purchases Planning Healthcare Earnings Education Spending Unilateral Unilateral Unilateral Unilateral Index Influence (1-3) Influence (1-3) Influence (1-3) Influence (1-3) Influence (1-3) Influence (1-3) Power (1-3) power (1-3) power (1-3) power (1-3) Treatment 0.115∗ 0.060 0.067 0.075 0.123∗∗∗ 0.106∗ 0.140∗∗∗ -0.033 -0.006 0.055 0.005 (0.066) (0.043) (0.050) (0.065) (0.039) (0.055) (0.047) (0.038) (0.046) (0.060) (0.041) Observations 1740 1702 1704 1559 1708 1534 1635 1727 1729 1578 1737 Control mean 0.00 2.44 2.40 2.09 2.58 1.84 2.46 2.36 2.29 1.93 2.52 Pooled Spillover 0.115∗∗ 0.103∗∗∗ 0.082∗∗ 0.092∗ 0.115∗∗∗ 0.075 0.130∗∗∗ 0.039 0.017 0.039 0.002 (0.051) (0.033) (0.038) (0.054) (0.032) (0.049) (0.034) (0.034) (0.039) (0.050) (0.038) Observations 2587 2502 2498 2114 2525 2131 2371 2549 2547 2150 2579 Control mean 0.04 2.45 2.41 2.10 2.59 1.83 2.46 2.37 2.32 1.95 2.54 CT Spillover 0.105 0.084∗∗ 0.069 0.075 0.114∗∗∗ 0.069 0.109∗∗∗ 0.022 0.017 0.047 0.002 (0.070) (0.038) (0.050) (0.063) (0.037) (0.062) (0.041) (0.040) (0.049) (0.063) (0.047) Observations 1483 1453 1450 1342 1448 1309 1407 1478 1477 1365 1481 Control mean 0.00 2.44 2.40 2.09 2.58 1.84 2.46 2.36 2.29 1.93 2.52 Non-CT Spillover 0.123∗∗ 0.133∗∗∗ 0.101∗∗ 0.129∗∗ 0.116∗∗∗ 0.086 0.168∗∗∗ 0.065 0.008 0.011 -0.003 (0.057) (0.040) (0.040) (0.057) (0.038) (0.055) (0.041) (0.043) (0.043) (0.056) (0.044) Observations 1104 1049 1048 772 1077 822 964 1071 1070 785 1098 Control mean 0.11 2.47 2.43 2.11 2.62 1.82 2.47 2.38 2.36 2.01 2.57 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. 62 Table A15: Intra-household dynamics index components (1) (2) (3) (4) (5) (6) (7) (8) (9) Intra-Household Partner Disagrees Trusts Partner Household Household Household Household Dynamics Dynamics with Partner Partner Inclusiveness Dynamics Allows Family Tensions Inclusiveness Index Index (1-4) (1-4) (1-4) Index Visits (0, 1) Infrequent (1-4) (1-4) Treatment 0.145 0.060 0.013 0.065 0.053 0.149 0.023 0.077 0.073 (0.088) (0.071) (0.047) (0.054) (0.059) (0.093) (0.017) (0.058) (0.069) Observations 1766 1580 1551 1555 1566 1766 1766 1766 1766 Control mean -0.07 -0.00 3.07 3.24 3.48 -0.08 0.90 3.49 2.75 Pooled Spillover 0.083 0.001 -0.031 0.049 0.010 0.101 0.023∗ 0.021 0.063 (0.084) (0.075) (0.048) (0.055) (0.054) (0.086) (0.012) (0.058) (0.069) Observations 2650 2249 2186 2192 2226 2650 2650 2650 2650 Control mean -0.04 -0.00 3.07 3.23 3.49 -0.05 0.91 3.50 2.76 CT Spillover 0.088 0.012 -0.044 0.039 0.035 0.097 0.011 0.010 0.104 (0.093) (0.077) (0.055) (0.057) (0.059) (0.100) (0.016) (0.064) (0.078) Observations 1503 1368 1341 1344 1353 1503 1503 1503 1503 Control mean -0.07 -0.00 3.07 3.24 3.48 -0.08 0.90 3.49 2.75 63 Non-CT Spillover 0.078 -0.012 0.000 0.076 -0.036 0.111 0.042∗∗∗ 0.041 -0.003 (0.091) (0.109) (0.071) (0.075) (0.075) (0.082) (0.013) (0.059) (0.071) Observations 1147 881 845 848 873 1147 1147 1147 1147 Control mean 0.00 -0.01 3.08 3.23 3.51 0.01 0.93 3.52 2.76 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. Table A16: Financial support index components (1) (2) (3) (4) (5) (6) (7) Financial Able to raise No. of Financial No. of Financial No. of Financial No. of Financial Village Support funds in case Supporters Supporters (Other Supporters Supporters Financial Support Index of shocks (1-4) (Siblings) Family Members) (Friends) (Others) (1-4) Treatment 0.256∗∗∗ 0.160∗∗∗ 2.855 12.889 4.738 12.089 0.133∗∗∗ (0.064) (0.060) (23.838) (12.915) (5.009) (11.154) (0.047) Observations 1766 1766 1766 1766 1766 1766 1766 Control mean -0.04 1.60 39.06 0.43 2.70 0.19 2.72 Pooled Spillover 0.199∗∗∗ 0.125∗∗ 13.244 3.489 7.629 10.042 0.102∗ (0.071) (0.058) (22.018) (9.961) (5.625) (9.629) (0.052) Observations 2650 2650 2650 2650 2650 2650 2650 Control mean -0.08 1.55 34.10 7.74 1.78 0.18 2.72 CT Spillover 0.218∗∗ 0.132∗ 21.551 3.431 6.430 4.969 0.116∗∗ (0.083) (0.068) (32.154) (3.227) (7.958) (4.851) (0.051) Observations 1503 1503 1503 1503 1503 1503 1503 Control mean -0.04 1.60 39.06 0.43 2.70 0.19 2.72 Non-CT Spillover 0.170∗∗ 0.114∗ -0.468 3.748 9.453 18.696 0.081 64 (0.073) (0.063) (19.107) (26.655) (7.504) (17.660) (0.064) Observations 1147 1147 1147 1147 1147 1147 1147 Control mean -0.14 1.47 26.20 19.39 0.32 0.17 2.73 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. Table A17: Social support index components (1) (2) (3) (4) (5) (6) (7) Social No. of No. of people No. of people No. of people No. of people No. of Support Role to ask advice who seek advice to ask advice who seek advice Market Index Models on activities on activities on disputes on disputes Intermediaries Treatment 0.083 0.872∗∗ 0.174 0.128 0.096 -0.055 0.011 (0.067) (0.351) (0.121) (0.125) (0.111) (0.119) (0.086) Observations 1766 1766 1766 1766 1766 1766 1766 Control mean -0.05 2.75 1.94 1.23 1.70 1.23 0.85 Pooled Spillover 0.061 0.353 0.163 0.016 0.059 0.019 0.074 (0.073) (0.329) (0.138) (0.134) (0.128) (0.137) (0.091) Observations 2650 2650 2650 2650 2650 2650 2650 Control mean -0.01 3.02 1.87 1.40 1.72 1.40 0.83 CT Spillover 0.113 0.755∗∗ 0.196 0.123 0.165 0.041 0.075 (0.078) (0.311) (0.155) (0.121) (0.148) (0.144) (0.103) Observations 1503 1503 1503 1503 1503 1503 1503 Control mean -0.05 2.75 1.94 1.23 1.70 1.23 0.85 Non-CT Spillover -0.022 -0.267 0.102 -0.158 -0.114 -0.003 0.070 (0.088) (0.527) (0.153) (0.207) (0.144) (0.196) (0.111) Observations 1147 1147 1147 1147 1147 1147 1147 Control mean 0.06 3.44 1.77 1.66 1.75 1.67 0.79 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. 65 Table A18: Social standing index components (1) (2) (3) (4) (5) Social Good Respected Opinion Social Standing Person Person Followed Position Index (0,1) (0,1) (0,1) (0,1) Treatment 0.171∗ 0.192 0.329∗ 0.366∗∗ 0.104 (0.099) (0.193) (0.166) (0.154) (0.154) Observations 1766 1762 1766 1766 1766 Control mean -0.07 6.01 5.30 4.95 4.55 Pooled Spillover 0.146 0.240 0.257 0.225 0.159 (0.103) (0.206) (0.168) (0.157) (0.158) Observations 2650 2635 2650 2650 2650 Control mean -0.05 5.96 5.41 5.12 4.46 CT Spillover 0.187∗ 0.212 0.352∗∗ 0.334∗ 0.229 (0.107) (0.208) (0.173) (0.170) (0.177) Observations 1503 1495 1503 1503 1503 Control mean -0.07 6.01 5.30 4.95 4.55 Non-CT Spillover 0.074 0.287 0.103 0.052 -0.003 (0.113) (0.227) (0.194) (0.183) (0.174) Observations 1147 1140 1147 1147 1147 Control mean -0.02 5.89 5.58 5.38 4.31 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. 66 Table A19: Collective action index components (1) (2) (3) (4) (5) (6) Collective No. of No. of Community No. of Works Action Associations Association Project Volunteering with Index where Member Responsibilities Donations (FCFA) Days Community (1-4) Treatment 0.226∗∗∗ 0.385∗∗∗ 0.080∗∗∗ 6.756 0.194 0.057 (0.074) (0.098) (0.027) (34.838) (0.209) (0.034) Observations 1766 1766 1766 1766 1766 1766 Control mean 0.02 0.58 0.08 213.57 1.07 2.96 Pooled Spillover 0.120∗ 0.162∗∗ 0.049∗∗ 11.020 0.002 0.055 (0.066) (0.075) (0.021) (29.622) (0.166) (0.035) Observations 2650 2650 2650 2650 2650 2650 Control mean -0.03 0.52 0.09 180.38 1.00 2.93 CT Spillover 0.079 0.130 0.051∗ -19.155 -0.008 0.029 (0.076) (0.084) (0.028) (37.427) (0.190) (0.037) Observations 1503 1503 1503 1503 1503 1503 67 Control mean 0.02 0.58 0.08 213.57 1.07 2.96 Non-CT Spillover 0.183∗∗ 0.208∗∗ 0.045 58.473∗ 0.016 0.094∗∗ (0.078) (0.095) (0.029) (33.016) (0.193) (0.046) Observations 1147 1147 1147 1147 1147 1147 Control mean -0.12 0.42 0.10 127.50 0.89 2.88 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. Table A20: Social cohesion and community closeness index components (1) (2) (3) (4) (5) (6) Social Cohesion Trusts No. of Don’t Community Community and Community Village Trusted Villagers Have Tensions Inclusiveness Closeness Index Women (1-4) (1-10) Enemies (1-4) Infrequent (1-4) (1-4) Treatment -0.036 0.048 -0.149 -0.143∗∗ -0.018 0.059 (0.085) (0.048) (0.178) (0.061) (0.046) (0.051) Observations 1766 1766 1766 1766 1766 1766 Control mean -0.03 2.90 4.97 3.25 3.29 2.33 Pooled Spillover 0.024 0.072 0.025 -0.051 -0.058 0.047 (0.085) (0.044) (0.174) (0.055) (0.054) (0.052) Observations 2650 2650 2650 2650 2650 2650 Control mean -0.02 2.89 4.93 3.26 3.30 2.35 CT Spillover 0.052 0.084∗ -0.005 -0.048 -0.041 0.080 (0.086) (0.049) (0.179) (0.061) (0.061) (0.059) Observations 1503 1503 1503 1503 1503 1503 68 Control mean -0.03 2.90 4.97 3.25 3.29 2.33 Non-CT Spillover -0.019 0.055 0.068 -0.054 -0.084 -0.006 (0.095) (0.047) (0.196) (0.068) (0.062) (0.056) Observations 1147 1147 1147 1147 1147 1147 Control mean -0.01 2.86 4.88 3.28 3.33 2.38 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. Table A21: Mental health index components (1) (2) (3) (4) (5) (6) Mental Less Less Life Inner Self-Reported Health Depression Disability Satisfaction Peace Mental Index (0-70) (0-28) (1-10) (1-10) Health Treatment 0.016 -0.057 -0.254 0.133 0.228 -0.103 (0.124) (1.232) (0.540) (0.169) (0.225) (0.104) Observations 1766 1766 1766 1766 1766 1766 Control mean 0.10 48.44 22.30 4.93 5.77 0.11 Pooled Spillover 0.004 0.082 -0.058 0.076 0.097 -0.090 (0.115) (1.051) (0.502) (0.171) (0.217) (0.100) Observations 2650 2650 2650 2650 2650 2650 Control mean 0.01 47.57 21.91 4.80 5.77 0.03 CT Spillover 0.026 0.203 -0.048 0.138 0.146 -0.100 (0.120) (1.124) (0.527) (0.187) (0.228) (0.105) Observations 1503 1503 1503 1503 1503 1503 Control mean 0.10 48.44 22.30 4.93 5.77 0.11 Non-CT Spillover -0.052 -0.434 -0.134 -0.033 0.032 -0.072 (0.123) (1.119) (0.547) (0.177) (0.232) (0.105) Observations 1147 1147 1147 1147 1147 1147 Control mean -0.12 46.19 21.30 4.61 5.77 -0.11 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. 69 Table A22: Self-efficacy index components (1) (2) (3) (4) (5) (6) (7) (8) Self Put effort Stay on Cope Adapt Find Find Do as Efficacy to solve your plan and with contigencies and handle multiple usually well as Index problems (1-4) achieve goals (1-4) (1-4) difficulties (1-4) solutions (1-4) solutions (1-4) others (1-4) Treatment 0.050 -0.013 0.005 0.020 0.041 0.016 0.033 0.068∗ (0.058) (0.033) (0.033) (0.035) (0.036) (0.037) (0.035) (0.040) Observations 1766 1766 1766 1766 1766 1766 1766 1766 Control mean 0.07 3.06 3.03 2.92 3.01 2.93 2.96 2.96 Pooled Spillover 0.002 -0.028 -0.006 -0.036 0.022 0.002 0.010 0.039 (0.065) (0.035) (0.037) (0.039) (0.036) (0.035) (0.036) (0.040) Observations 2650 2650 2650 2650 2650 2650 2650 2650 Control mean 0.02 3.01 2.98 2.90 3.01 2.92 2.95 2.92 CT Spillover -0.038 -0.070∗ -0.043 -0.061 0.008 -0.003 0.013 0.031 (0.065) (0.038) (0.043) (0.042) (0.035) (0.038) (0.039) (0.040) Observations 1503 1503 1503 1503 1503 1503 1503 1503 Control mean 0.07 3.06 3.03 2.92 3.01 2.93 2.96 2.96 Non-CT Spillover 0.066 0.041 0.051 0.001 0.042 0.018 0.003 0.054 70 (0.091) (0.049) (0.051) (0.057) (0.049) (0.048) (0.051) (0.055) Observations 1147 1147 1147 1147 1147 1147 1147 1147 Control mean -0.07 2.94 2.91 2.86 3.02 2.90 2.94 2.86 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. Table A23: Future expectations index components (1) (2) (3) Future Expected Expected Expectations Social Status Life Satisfaction Index (0-10) (1-10) Treatment 0.056 0.055 0.074 (0.088) (0.066) (0.067) Observations 1373 1263 1249 Control mean 0.09 2.60 2.63 Pooled Spillover 0.110 0.092 0.084 (0.079) (0.066) (0.056) Observations 1914 1748 1734 Control mean -0.04 2.51 2.55 CT Spillover 0.072 0.058 0.049 (0.086) (0.068) (0.066) Observations 1141 1046 1028 Control mean 0.09 2.60 2.63 Non-CT Spillover 0.212∗∗ 0.168∗∗ 0.170∗∗ (0.092) (0.083) (0.065) Observations 773 702 706 Control mean -0.26 2.33 2.41 Notes: Results presented are OLS estimates that include controls for region fixed effects and, where possible, baseline outcomes. Robust standard errors are shown in parentheses, clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1. 71