Can Participatory Community-Based Targeting Improve Trust and Social Cohesion?

The key intervention in this project involves alternative variants of community-based targeting (CBT) of cash transfers. The CBT is conducted by six community leaders in each village, composed of the village (Kebele) leader, a community elder, religious leader, a teacher/development agent/ health extension worker, a women’s representative and youth representative. Community leaders were provided with pre-defined criteria (which we discuss in the experimental design section below) set by the research team, to identify and prioritize beneficiaries. We provide community leaders with a list of 20 randomly selected households in their village (community), and we then ask them to rank these households based on their need for social assistance, from the neediest to the least needy. We provide community leaders with either a hypothetical or real cash transfer of 20,000 Ethiopian Birr to distribute among those households deemed “eligible”. This mimics the CBT processes in many cases, including in the Ethiopian Productive Safety Net Programme (PSNP).

In this intervention, we exogenously vary whether villages are assigned to a hypothetical or actual (real) cash transfer, and whether the targeting process is participatory (involving community members in the process) or non-participatory (led only by community leaders). Thus, the project introduces two sources of variation and hence interventions:

(i) Hypothetical versus actual (real) cash transfer
Communities and hence community leaders are assigned either to a hypothetical or an actual (real) cash transfer of ETB 20,000 to be distributed to “eligible” households. In the hypothetical (control) group, community leaders were asked to rank community members and allocate a hypothetical cash transfer of ETB 20,000, while in the actual transfer group, community leaders were asked to rank and allocate a real transfer of ETB 20,000. This will allow us to compare whether and how targeting outcomes differ when targeting decisions are based on hypothetical and real cash transfers.
(ii) Participatory vs non-participatory targeting methods

The other variant of our intervention comes from the type of targeting method communities are assigned to: targeting process in the presence and absence of public disclosure and community members’ participation in the targeting process. In the non-participatory targeting group, only community leaders decide how households are ranked and how the transfer is allocated among the community members. In contrast, in the participatory approach, there is a public disclosure and requires community members to review the community leaders’ transfer allocation decisions, vote on the inclusion and exclusion, and potentially override leaders’ proposals through a structured and anonymous voting process. Community leaders assigned to the participatory targeting were informed that their decisions have to be consulted with potential beneficiary households before they can be implemented. These targeting dimensions allow us to compare how targeting outcomes differ in the presence (absence) of public disclosure and community members’ participation. More specifically, this intervention aims to test whether public disclosure and community members’ participation improve trust in local leadership and social cohesion among community members. The intervention is facilitated by trained enumerators in collaboration with community leaders, with all sessions closely monitored to ensure strict adherence to the experimental protocol and consistency across communities. Each community leader was given a token of appreciation of 350 Ethiopian Birr (about 2.5 USD). If they wish, community leaders were given the option to retain up to a maximum of 10 percent of the budget for covering “administrative” costs. The community-based targeting experiment started in November 2025 and the endline follow-up household survey is scheduled to start Mid-December 2025. Thus, the follow-up household survey will be conducted 2-4 weeks after the community-based experiment was conducted.

Research Questions
This study aims to assess the impact of community-based cash transfers as well as alternative CBT designs to improve trust in major actors involved in the delivery of social protection programs (including local leaders and NGOs) and social cohesion among community members. More specifically, we evaluate whether enforcing participation of community members (potential beneficiary households) in targeting processes can affect efficiency and legitimacy of targeting processes and hence trust in those involved in the delivery and targeting of these programs. As social protection programs serve a portion of potentially eligible populations, understanding what happens not just to beneficiary populations but to non-beneficiary populations remains an important agenda that has received limited attention. Our design allows us to compare the implication of alternative targeting of cash transfers on beneficiary and non-beneficiary households. Finally, we also explore potential mechanisms that may explain how different targeting methods affect trust and social cohesion. Thus, we raise the following research questions:
(i) Does public engagement in CBT enhance trust, social cohesion, efficiency and legitimacy of targeting?
(ii) Can participatory targeting reduce unintended negative effects on non-beneficiaries?
(iii) What are the mechanisms through which participatory community-based targeting affect the above-mentioned outcomes?

2025-11-13

2025-12-31

Primary outcomes
(i) Trust in local leaders
(ii) Trust in NGOs
(iii) Social cohesion among community members
(iv) Efficiency of targeting
(v) Legitimacy of targeting
(vi) Preference for future targeting

(i) Trust in local leaders is measured using incentive compatible investment game originally developed by Berge et al. (1995) as well as survey-based measures of trust. We focus on trust in village (kebele) leaders, who are involved in the targeting of the cash transfer. We administer these questions through Likert-scale questions of agreement and disagreement associated with respondents’ trust and confidence in these leaders. We aim to use both continuous and binary outcomes associated with these questions. In addition to the survey-based measures of trust, we will also measure trust using an incentive-compatible trust game. Trust in community leaders will be captured by the amount of money sent by the household to the leader. This is a continuous variable ranging from 0 to 100 Birr, in increments of 10 Birr (0, 10, 20, …, 100). We will also construct two binary indicators: (i) a positive-trust indicator equal to 1 if the household sends any positive amount (>0), and 0 otherwise; and (ii) a high-trust indicator equal to 1 if the household sends at least 50 Birr, and 0 otherwise. To measure the belief of households about community leaders’ trustworthiness, we will measure the expected return (continuous), defined as the amount the household expects the community leader to return.
(ii) Trust in NGOs is similarly measured using Likert-scale of agreement or disagreement.
(iii) Social cohesion: we aim to construct an index to capture social cohesion by combining several indicators measuring social interaction and participation in community-based organizations and activities (e.g., Nigus et al., 2026).
(iv) Efficiency of targeting is measured following Premand and Schnitzer (2021). We aim to capture efficiency by comparing the share of poor households included through each targeting method.
(v) Legitimacy of targeting is measured by eliciting respondents’ satisfaction and confidence in the targeting process and outcomes. Following Premand and Schnitzer (2021) we elicit whether respondents are very satisfied with the targeting method and whether they would like the same method to be used again in similar programs in the future
(vi) To elicit targeting preferences of respondents we directly ask them who should participate and be responsible for targeting similar cash transfer programs. Respondents can choose various actors, including community leaders as well as other actors.

(i) Interpersonal trust
(ii) Subjective well-being
(iii) Informal transfers

Experimental design
The intervention is carried out in 181 villages (enumeration areas) across several regions in Ethiopia, including Tigray, Afar, Amhara, Oromia, Sidama, South Ethiopia, Southwest Ethiopia, Central Ethiopia, Somali regions, as well as the Dire Dawa city administration. As described above, the experiment is implemented by six community leaders per village, each leader holding different customary or administrative roles in the community. This composition mimics targeting practices in Ethiopia, including those practices used in the PSNP. The intervention occurs at two levels: (i) community leaders and (ii) community members (households). In this study, we focus on the impact of alternative community-based targeting interventions on household-level outcomes. The same community leaders also participated in a similar targeting experiment conducted two years ago (Abay et al., 2024).
The intervention follows a community-level cluster randomization, where 181 villages (enumeration areas) are randomly assigned into a control group and two treatment arms. We stratified the random assignment of villages by region and previous participation in cash transfer programs. As shown in Figure 1 below, we randomly assigned about 34% of the villages (61 EAs) to the control group, while the remaining 36% (66 EAs) and 30% (54 EAs) were randomly assigned to an incentivized (non-participatory) and an incentivized and participatory targeting, respectively.


(1) Control: Hypothetical targeting and ranking for distributing a 20,000 Birr transfer (C):
Community leaders in this group are asked to rank 20 randomly selected households in each village based on their need for social assistance and allocate a hypothetical transfer of ETB 20,000 to those deemed “eligible”. We provided them five criteria to prioritize and rank households by asking them to prioritize households who (i) had difficulty satisfying their food needs; (ii) own no or little asset (e.g., livestock, land); (iii) have limited income-generating activities or capacity; (iv) have lost productive assets due to shocks (e.g., conflict, drought); and (v) have lost family members recently. Before ranking and community leaders first discuss these set of five pre-defined targeting criteria and report whether the criteria are exhaustive. If they wish, community leaders were given the option to retain up to a maximum of 10 percent of the budget for covering “administrative” costs. This is beyond the token of appreciation of 350 Ethiopian Birr (about 2.5 USD) for their participation. Before distributing the hypothetical transfer, each community leader was individually state how much of the budget (up to 10%) they would hypothetically allocate to cover their “administrative costs”. Community leaders then jointly agree on the amount of money to be retained. Then community leaders jointly rank all households from most to least needy and make a hypothetical allocation that does not exceed ETB 20,000.
(2) Incentivized targeting involving actual cash transfer (T1):
Community leaders in this group follow the same procedures as in the control group, except the leaders receive an actual budget of ETB 20,000. Community leaders collectively decide the share to be allocated to cover administrative costs, and then rank households using the same pre-defined criteria, and distribute real cash transfers to those community members deemed “eligible” accordingly. Community leaders record the amount allocated to each household and list the most important criterion used for the three top-ranked beneficiaries. Comparing Groups 1 and 2 allows us to examine the impact of real cash transfers on community leaders’ targeting decisions as well as ultimate impacts on trust and social cohesion among households.

(3) Incentivized and participatory targeting involving community members (T2):
Community leaders in this group also receive an actual budget of ETB 20,000 and rank and allocate the cash transfer to potential beneficiaries using the same selection criteria as in Group 2, including the decisions on the amount of administrative costs they keep for themselves. However, community leaders were asked to explicitly involve potential beneficiary households in the targeting processes. After community leaders propose the ranking and allocation of the cash transfer, a public deliberation session is held with at least two-thirds of the sample households. During this session, community leaders must publicly disclose the total budget allocated to the village (ETB 20,000), the amount of budget kept for themselves to cover their administrative cost, their proposed beneficiary ranking and allocations, and their justification for the selection of beneficiary households. Community members are then informed that they can anonymously vote to accept or override leaders’ inclusion and exclusion decisions. Beneficiary households must receive at least 50% support from participating households to be considered as eligible to receive the transfer. If not, these households may be replaced by community members-nominated households that secure at least 50% of the votes. If more nominees pass the threshold than positions available, those with the highest vote shares are selected. Community members cannot add new beneficiaries without replacing others, and including the administrative costs, total allocations remain within ETB 20,000. Importantly, households selected by community leaders who do not participate in the public deliberation or voting session are still eligible to receive the transfer; non-attendance does not exclude them from receiving the cash transfer. The final beneficiary list and allocations are also recorded after voting.

Overall, this experimental design allows us to identify (i) the impact of cash transfers by comparing hypothetical vs real cash transfers, and (ii) the impact of public disclosure and participatory targeting by comparing leaders-only targeting versus targeting where community members can overturn decisions.

Not available

The randomization was conducted through computer using Stata.

Village or community level

Yes

181 communities

About 2896 households

C (Control : hypothetical targeting): 61 villages
T1 (Incentivized targeting): 66 villages
T2 (Incentivized and participatory targeting): 54 villages

We compute the number of clusters and households needed to detect meaningful effect for the primary outcomes described above, assuming fixed number of households in each cluster (village). In the baseline (2019) sample there were an average of 20 households in each village, and we managed to trace 85 percent of them in the 2023 survey. In the current survey, we anticipate being able to trace about 80 percent of them, 16 households per village. Our power calculations aim to achieve the standard and widely adopted 80 percent power at a significance level of 5 percent. Given that we have several hypotheses and primary outcomes, we computed the number of clusters and associated sample size needed for each outcome separately. To evaluate the impact of participatory targeting on trust in various actors and social cohesion, we compute statistical power and sample size needed to detect a reasonable impact on the primary outcomes listed above: trust in local leaders, trust in NGOs, social cohesion, efficiency and legitimacy of targeting process, and preference for targeting of future social assistance programs. For each outcome, we assemble mean and standard deviations from previous round of our survey (see for example, Abay et al., 2024; Nigus et al., 2025; Abay et al., 2025). studies in similar settings. We extract impact estimates from previous studies evaluating similar interventions (e.g., Evans et al., 2019; Alessandro et al., 2021). Using these minimum detectable effects (MDEs), we compute the number of clusters and households needed for conducting pairwise comparisons of the different treatment arms. For example, as shown in Table 1, detecting a 12 percentage point improvement in trust in local leaders requires 53 clusters in each arm and a total of 159 clusters and 2544 households. Similarly, detecting an 8 percentage points improvement in trust in NGOs requires 46 communities in each arm and a total of 138 clusters and 2208 households. Considering these computations, we decided to keep a total of 181 communities, with about 33 percent of the 181 communities assigned to the control (hypothetical arm) with no actual transfer and divide the remaining communities into two groups. This yields a sample of 2896 households, which is larger than the sample size needed for most of the outcomes except for comparing efficiency in targeting using the share of poor households identified in each arm. We note that we only computed power calculations for those primary outcomes and those with baseline data. We stratify the random assignment of communities across regions and previous exposure to similar targeting exercise. To improve efficiency in our estimations, we will account for these stratifications in our estimations.