Tackling psychosocial and capital constraints to alleviate poverty

The research protocol was approved by Innovations for Poverty Action Institutional Review Board and preregistered in the AEA RCT Registry (study 0002544). The pre-analysis plan is registered at https://www.socialscienceregistry.org/versions/52534/docs/version/document. All survey participants completed informed consent. They were not compensated for their time as they were all part of the national cash transfer programme.

Interventions

Niger context and cash transfer programme

Niger, one of the poorest countries in the world, has a rural poverty rate50 of 51.4% and ranks last in human development indicators51. Landlocked in the Sahel, its population is highly exposed to climatic shocks and food insecurity. More than 90% of Nigerien households have a member engaged in agriculture, but agricultural production is dominated by low-productivity subsistence farming with little market access. Only about 25% of farmers commercialize any crop and only 10% of villages have a permanent market. Non-agricultural activities are scarce as a primary occupation in rural areas (less than 10%) but are a secondary occupation for about a third. They mostly consist of agricultural transformation and trading. The wage sector only employs 4% of the workforce, mostly in public sector jobs concentrated in the capital. More than a third of Nigerien women do not participate in the labour force, overwhelmingly because of the burden of housework52.

After repeated humanitarian interventions in response to shocks and food insecurity, the Government of Niger set up a social protection system. Its cornerstone was a national cash transfer programme that provided monthly payments of 10,000 XOF for two years (US$15.95, US$38.95 purchasing power parity-adjusted (PPP)), which represented approximately 11% of yearly household consumption for targeted poor rural households. The programme was rolled out in three main phases and reached 100,000 beneficiary households between 2012 and 2019. We studied the 3rd phase of the programme, implemented from 2016 to 2019, which reached approximately 22,500 households. The cash transfers were unconditional but were delivered with child development promotion activities for all households.

The national cash transfer programme applied geographical targeting before using household-level poverty targeting. The programme selected the communes with highest poverty rates in all eight regions in the country. In practice, most selected communes were rural. Within communes, all villages were eligible and public lotteries were organized to select beneficiary villages. Poverty-targeting methods were applied to determine the beneficiary households. Within selected households, a woman over 20 was the recipient of the cash transfers.

Multi-faceted interventions

To address constraints to participation in income-generating activities and economic diversification, the multi-faceted programme combined three main sets of interventions and was delivered on top of the regular cash transfer programme53. The core components promoted financial inclusion, basic micro-entrepreneurship skills and market access. A second component addressed capital constraints by providing a lump-sum cash grant intended for productive purposes. A third component provided psychosocial interventions that aimed to strengthen aspirations and interpersonal and intrapersonal skills, as well as to address gender and social norms. Supplementary Appendix 1 describes how the intervention was delivered through the government-led national cash transfer programme.

Core components included in all three treatment variants

1. Coaching. The coaching component facilitated the delivery and coordination of the various interventions. Beneficiaries formed groups of 15 to 25 members and selected a coach to mentor them throughout the programme. Coaches were men or women from the village, generally selected for their capacity to advise on income-generating activities and to represent the group for service providers and market agents. Coaches facilitated the implementation of group-based programme activities, including promoting the attendance of beneficiaries at meetings and coordinating with service providers. They led group-level coaching sessions, during which challenges and opportunities for income-generating activities were discussed. The group-level coaching sessions occurred during weekly savings group meetings, as described below. Coaches also provided some individualized follow-up to beneficiaries.

2. Saving groups. The groups of beneficiaries formed a village savings and loans association (VSLA), with initial training from the coach. The group received a VSLA kit, elected members to leadership positions and determined the rules governing the association. Key decisions included the cost of a saving ‘share’, maximum loan size, interest rate and duration of a savings cycle. Group members also defined other parameters, such as a mandatory contribution to an emergency fund and penalties. At weekly meetings, members purchased between one and five shares in the savings fund, contributed a fixed amount to the emergency fund, and could take out a short-term loan from the savings fund. A full savings cycle lasted between 9 and 12 months, at which point the accumulated savings, interest, and penalty fees were shared among members in proportion to the number of savings shares owned by each member.

3. Micro-entrepreneurship training. A week-long micro-entrepreneurship training was delivered to the groups. The curriculum was adapted from the International Labour Organization’s Start and Improve Your Business (SIYB) level 1 training, which is tailored to non-literate participants. The curriculum covered fundamental micro-entrepreneurship skills, including basic accounting and management principles, market research, planning and scheduling, saving, and investing. In addition, the training focused on the choice of livelihood activities and the preparation of a basic business plan.

4. Access to markets. Coaches were trained to deliver information sessions on market access. Depending on production cycle timing, they held group sessions to discuss where to buy inputs for agricultural activities, how to choose suppliers, or where to sell products.

The capital component

A lump-sum cash grant of 80,000 XOF (US$127 (US$311 in 2016 PPP)) was provided to promote investments in income generating activities. Payments were not conditional on participation in other programme activities.

The psychosocial components

The psychosocial components included community-level programming, which consisted of community sensitization on social norms and aspirations, and individual-level programming, which consisted of life skills training for the beneficiaries. While they were relatively light, they aimed to trigger three main mechanisms: (1) to build personal psychological assets, including self-efficacy, self-worth, aspirations, and optimism about the future, while developing behavioural skills related to interpersonal communication, problem-solving, leadership, and goal setting; (2) to promote social empowerment, including social standing in the community, community support and solidarity, and supportive social norms around women’s income generating activities; and (3) to foster positive intra-household dynamics, including interpersonal trust, closeness, and conflict resolution, as well as women’s decision-making power and control over resources. We also expected several of these mechanisms to improve mental health. Supplementary Appendix 2 provides a detailed description of the psychosocial components.

1. Community sensitization on aspirations and social norms. The full community, including elders, economic and traditional leaders, and programme beneficiaries and their husbands (or other family members), were invited to attend a video screening and community discussion. Programme staff projected a short video in local languages that depicted the story of a couple that overcomes household and personal constraints and develops economic activities, with support from family and their community. As a result, they become more economically resilient. After the screening, trained facilitators guided a public discussion on social norms, aspirations, and community values. The sensitization integrated multiple approaches to social and behaviour change. These include role models in the video, peer effects in the audience construction, goal setting and social consensus techniques in the discussion, and values alignment in both the video and discussion.

2. Life skills training. A week-long life-skills training was organized for groups of beneficiaries. Grounded in participatory, problem-centred learning, the training included role plays, games, and case studies. The nine modules of the curriculum focused on building skills for effective decision-making, problem-solving, goal setting, interpersonal communication, and women’s leadership, while simultaneously building self-worth, self-efficacy, and aspirations. In addition, discussions prompted participants to relate their economic goals to broader values and to spousal, gender, and generational roles. The training was delivered by private trainers contracted by the government through small firms.

Randomized controlled trial design and data

Experimental design

In total, approximately 100,000 households have participated in the Niger cash transfer programme since 2012. This study focused on the 3rd wave of the programme, which reached 22,507 beneficiary households in 329 villages in 17 communes of the 5 most populous of Niger’s eight regions (Dosso, Maradi, Tahoua, Tillaberi and Zinder; see Supplementary Fig. 1 for a map of study communes). All of the villages that received cash transfers in the 17 communes were included in our sample. After grouping small neighbouring villages that have less than 8 beneficiaries for ease of programme operations, 322 villages entered the randomization.

The study is a cluster-randomized controlled trial in which villages with existing cash transfer beneficiaries were randomly allocated to one of the four arms (Table 1): one control group (81 villages), and three treatment arms with variants of the intervention components (80 villages in Capital, 78 villages in Psychosocial and 83 villages in Full). Within each village there was no additional randomization across households, and thus all eligible households within each village received the same treatment.

Randomization of the villages was stratified by the 17 communes and the targeting method used to select cash transfer beneficiaries in each village (which is part of a complementary study54) and took place in public lotteries. To promote the transparency gained from public lotteries while maintaining balance across targeting methods, we proceeded in two stages. First, for each commune we randomly assigned villages into four lists stratified by targeting method. The strata were based on a categorical variable with four values, one for each of three randomized targeting methods and a fourth for not being part of the targeting study. This stage did not assign the experimental arm label to each list. Second, we organized a public lottery in each of the 17 communes to randomly assign each list to one of the four experimental arms. The lottery was organized by the cash transfer programme government team and held in the capital of the commune in the presence of village chiefs or elders.

One limitation of this design is that we could not include a fourth treatment arm with core components only. While we can therefore test the importance of including capital on top of the core and psychosocial components (by comparing the Full arm to the Psychosocial arm), if the psychosocial components change the marginal value of the capital, then we would not estimate the effect of providing capital as part of a programme without those psychosocial components. Likewise, we test the importance of including psychosocial components on top of a design that includes the lump-sum capital transfer (by comparing the Full arm to the Capital arm). Note that earlier work on the Niger national cash transfer programme has shown that cash transfers (either alone or combined with group savings facilitation as in the core component) increased savings and livestock accumulation, but had little average effects on earnings from income-generating activities or economic diversification55,56.

Sampling, timeline and data

Out of the 22,507 cash transfer beneficiaries that were assigned to the 4 treatment variants, 4,712 households were drawn into a sample for data collection (1,206 households in control, 1,191 households in capital, 1,112 households in psychosocial and 1,203 households in full). Before the study, we conducted power calculations assuming an intracluster correlation of 0.10 (based on data from Ghana6 and a Niger national household survey) and equal sized arms. To maximize power, we sampled all villages in this phase. Sampling 15 households per village allowed for minimum detectable sizes of 0.057s.d. between arms, before adjusting for baseline outcomes or strata.

Extended Data Figure 1 summarizes the study timeline. Baseline data collection took place between April and June 2017. The public lotteries took place after data collection in June 2017. The intervention was delivered between September 2017 and January 2019. Two follow-up surveys were collected. The midpoint occurred in February and March 2019, a median of 6 months (3 to 9 months) post-intervention (that is, after the delivery of the lump-sum grant in treatment arms with the capital component). The endpoint survey occurred a year later in February and March 2020, a median of 18 months post-intervention (after the delivery of the cash grants in treatment arms with the capital component). Survey teams, blind to treatment status, were assigned to villages; but the participant could reveal treatment status in the last module of the midpoint survey. During the fieldwork, a remote team checked and updated the field plan for treatment balance across teams and survey weeks.

Supplementary Table 1 reports descriptive baseline statistics and balance tests across the experimental arms for a set of pre-specified variables. The sample was extremely poor. Fewer than 8% of beneficiaries were literate and they had, on average, less than 0.5 year of schooling. Beneficiaries were 38 years old on average, and 99% were female. They took about 70 min to get to the nearest market. On the whole, the random assignment created well-balanced experimental arms.

At the midpoint and endpoint, 95.0% and 91.3% of baseline households were successfully interviewed, respectively. Attrition was balanced across the treatment arms (Supplementary Table 1, bottom panel).

Supplementary Table 2 documents compliance with treatment assignment based on administrative data. Across all treatment arms, the participation rate in VSLA meetings was 92%, and the attendance rate in the micro-entrepreneurship training was 95%. By design, there was more variation in the delivery of individual coaching visits, with on average 52% of beneficiaries receiving coaching visits each month. Across the Psychosocial and Full treatment groups, 94% of beneficiaries attended life skills training and 89% attended the community sensitizations. Across the Capital and Full treatment groups, 99.9% of beneficiaries received the cash grants.

Estimation strategy

We estimate separate intent-to-treat treatment effects for each (treatment) arm for pre-specified outcomes based on the following specification:

$${Y}_{i,t}={\beta }_{p,t}{T}_{{\rm{P}}{\rm{s}}{\rm{y}}{\rm{c}}{\rm{h}}{\rm{o}}{\rm{s}}{\rm{o}}{\rm{c}}{\rm{i}}{\rm{a}}{\rm{l}}}+{\beta }_{c,t}{T}_{{\rm{C}}{\rm{a}}{\rm{p}}{\rm{i}}{\rm{t}}{\rm{a}}{\rm{l}}}+{\beta }_{f,t}{T}_{{\rm{F}}{\rm{u}}{\rm{l}}{\rm{l}}}+\delta {Y}_{i,0}+{\boldsymbol{\gamma }}\,+{{\rm{\varepsilon }}}_{i,t}$$

(1)

where Yi,t is the outcome of interest for household or individual i at midpoint or endpoint (t = 1 or t =2); TPsychosocial, TCapital and TFull are indicators for village assignment to the Psychosocial, Capital, or Full treatment arm; γ is a vector of randomization strata fixed effects. We estimate this specification separately for each follow-up. Standard errors are clustered at the village level, the unit of randomization. To increase precision, we include a control for the outcome at baseline (Yi,0) when available. When not available for a subset of households, we set the baseline control to the mean outcome in the randomization strata and include a dummy for a missing measurement at baseline. βp,t, βc,t and βf,t are the main parameters of interest. They capture the impact of each treatment arm for regular cash transfer beneficiary households.

To estimate the added value of the cash grant and psychosocial components (or gross marginal effects), we report three additional tests for each data collection round:

First (H1), we test the added value (or gross marginal effect) of the cash grant with H0: βfβp = 0.

Second (H2), we test the added value (or gross marginal effect) of the psychosocial interventions (the community sensitization intervention and life skills training) with H0: βfβc = 0.

Third (H3), we test for equality of treatment effects between the Capital and Psychosocial arms, which is the same as testing equality of gross marginal effects of the cash grants and psychosocial interventions, with: H0: βcβp = 0.

Note that gross marginal effects are inclusive of complementarities with the core components.

Finally, we test for equality of treatment effects between data collection rounds to uncover any temporal effects (for each treatment arm separately).

We conduct our analysis in accordance with a pre-analysis plan. We pre-specified in our pre-analysis plan two primary economic outcomes: consumption per adult equivalent and the (reverse of) FAO’s Food Insecurity Experience Scale (FIES57,58). Although it was pre-specified as a secondary outcome, we also report the Food Consumption Score (FCS59 in the main outcome Extended Data Table 1, since it provides another measure of food security that captures the beneficiary women’s dietary diversity. Other notable deviations include slight changes of the grouping of outcome variables for expository clarity, and the presentation of standardized effect sizes for key outcomes. Supplementary Appendix 4 summarizes deviations from the pre-analysis plan.

We pre-specified a range of intermediary outcomes to capture the pathways through which the interventions were expected to affect the primary economic outcomes, as well as a range of psychosocial well-being measures (see Supplementary Appendix 3 for more information on psychosocial outcomes). We discuss key intermediary outcomes in the results section, with additional results in the annex. Supplementary Tables 3, 4 provide more details on variable construction.

To account for multiple hypotheses, we calculate P-values adjusted within each treatment arm within predetermined families of variables, and report corrections in Supplementary Table 5. Following our pre-analysis plan, we also calculate P-values controlling for both the false discovery rate (FDR) and the family-wise error rate (FWER). The FWER is our preferred correction and is displayed in the extended data tables.

Cost–benefit calculations

The intervention was designed as low-cost to ensure it could be scaled-up through government systems. Extended Data Table 9 details programme costs obtained from administrative data, per beneficiary of each intervention arm. In 2016 PPP US$, total costs were US$263 for the Psychosocial arm, US$482 for the Capital arm and US$584 for the Full arm. We do not account for cash transfer programme costs (including targeting or payment) since these were incurred for the control group as well. The programme costs were substantially lower than similar graduation programmes implemented in other contexts: US$1,475 PPP in India, US$4,215 PPP in Ethiopia, US$5,483 PPP in Ghana, US$6,044 PPP in Pakistan6 and US$6,183 PPP in Afghanistan25.

We perform a conservative calculation of estimated benefits that only considers impacts on consumption (obtained from the specification in equation (1)), without accounting for impacts on assets or psychosocial well-being. Cumulated consumption impacts are calculated as half the impacts on yearly consumption at midpoint plus impacts on yearly consumption at endpoint. We consider various scenarios regarding the sustainability of impacts after endpoint. First, we consider zero impacts after endpoint (scenario A). We then consider various yearly rates of dissipation of impacts, including 75% (scenario B1), 50% (scenario B2) and 25% (scenario B3). Lastly, we assume impacts are sustained in perpetuity (scenario C), as in the benchmark case used by some other studies6. We use a 5% discount rate when calculating benefit-cost ratios.

We also perform cost-effectiveness calculations of benefits to psychological well-being. For each treatment arm, we compute the cost per 0.1s.d. increase in life satisfaction, as assessed by the Cantril ladder at endpoint. We choose a benchmark of 0.1s.d. given it is approximately the meta-analytic effect of economic interventions on psychological well-being60. We additionally compute the cost per case of depression averted within each arm, using the CESD-10 self-report measure of depression at both follow-ups.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Related Posts

Leave a Reply

Your email address will not be published.

%d bloggers like this: