Rebuilding the Social Cohesion of Returnees in Post-Conflict Mozambique

The intervention consists of a one-time, in-person community dialogue held with small groups of local residents. The dialogue is structured and moderated, and focuses on everyday community life, experiences related to displacement and return, and peaceful coexistence after conflict.
Participants are randomly assigned to one of two dialogue formats. In one format, the discussion is facilitated by a respected religious leader from the local area. In the other format, the discussion is facilitated by a government official. The content and structure of the dialogue are the same across formats; only the identity of the facilitator differs.
A third group of participants does not take part in any dialogue and serves as a comparison group.

The intervention consists of a structured, facilitated community dialogue implemented among adult residents of Palma, Cabo Delgado, who were displaced by the armed conflict and have since returned to the town. The study uses an individual-level randomized design with three groups: (i) a control group with no dialogue intervention, (ii) a dialogue facilitated by a religious leader, and (iii) a dialogue facilitated by a government official.
Participants assigned to one of the two treatment groups take part in a small-group, in-person dialogue session with other residents from the same neighborhood. Dialogues follow a standardized protocol developed by the research team to ensure consistency across sessions and facilitators. The content focuses on shared community experiences rather than political positions, including: (i) everyday life in the neighborhood after return, (ii) experiences of displacement and loss, (iii) perceptions and stereotypes within the community, and (iv) aspirations for peaceful coexistence and rebuilding social ties. Discussions are designed to promote respectful listening, mutual recognition, and constructive engagement, without advocating specific policy positions or outcomes.
The two treatment arms differ only in the identity of the facilitator. In one arm, the dialogue is led by a locally respected religious leader (e.g., imam or pastor). In the other arm, the dialogue is led by a government official with responsibilities at the local or district level. Facilitators receive the same training, use the same discussion guide, and are instructed to follow identical moderation rules, including neutrality, inclusive participation, and non-confrontational facilitation. Participants are not informed in advance about which type of facilitator they will interact with.
The control group does not participate in any dialogue session during the study period but is surveyed on the same schedule as treated participants. All participants complete a baseline survey prior to the intervention and a follow-up survey after the dialogue. Surveys measure social cohesion, trust, cooperation, perceptions of community members, attitudes toward leaders and institutions, and views related to conflict, peace, and coexistence. Additional measurement tools include vignettes and indirect questioning techniques to capture sensitive attitudes.
The intervention is designed to isolate whether and how the messenger or institutional identity of the dialogue facilitator affects the ability of structured community conversations to rebuild trust and social cohesion in a fragile, post-conflict setting.

2026-02-13

2026-03-28

The primary outcomes of interest measure social cohesion, trust, and attitudes toward community members, leaders, and institutions in a post-conflict setting. These outcomes are designed to capture both explicit and implicit dimensions of beliefs and preferences, as well as behavioral intentions related to cooperation and peaceful coexistence.

The main primary outcomes are:
1. Social cohesion and willingness to cooperate
- Willingness to cooperate with neighbors and other community members.
- Acceptance of shared decision-making and participation in community activities.
- Willingness to interact, work with, or accept leadership from individuals with whom respondents may have had conflicts in the past.
2. Trust
- Trust in other community members.
- Trust in religious leaders.
- Trust in government officials and public institutions.
- Trust in people associated (directly or indirectly) with the conflict.
3. Attitudes toward conflict, violence, and coexistence
- Acceptance or rejection of violence as a means to address grievances.
- Attitudes toward insurgents or former combatants, including forgiveness, punishment, or reintegration.
- Perceived prevalence of violent or “malfeitor” behavior within the community.
4. Perceptions of legitimacy and moral authority
- Perceived fairness, integrity, and appropriateness of leaders (religious or governmental) as community representatives.
- Beliefs about whether leaders act in the interest of the community or selectively favor certain groups.
5. Implicit and sensitive attitudes
- Implicit associations between authorities, insurgents, and positive or negative attributes.
- Sensitive attitudes related to violence, trust, and social norms measured through indirect elicitation methods.
6. Emotional responses to community narratives
Emotional reactions expressed during open-ended responses about the community, the state, and the conflict, measured through multiple modalities.

Several primary outcomes are constructed indices or latent measures that combine information from multiple survey instruments and measurement modes. These outcomes are constructed as follows:
1. Social Cohesion Index
This index aggregates standardized responses from multiple survey items capturing:
Willingness to cooperate with others in the neighborhood.
Acceptance of shared community decision-making.
Comfort interacting with individuals linked to past conflict.
Beliefs about whether community members generally cooperate or exploit one another.
Items are standardized and combined into an index using either simple averaging or inverse-covariance weighting, depending on robustness checks.
2. Trust Indices
Separate trust indices are constructed for:
Interpersonal trust (neighbors, community members).
Religious authority trust (religious leaders).
State trust (local, provincial, and national government officials).
Each index combines Likert-scale responses measuring confidence, perceived honesty, and perceived fairness of the relevant group.
3. Attitudes Toward Violence and Reintegration
This outcome combines:
Explicit survey questions on acceptance of violence.
Preferences regarding punishment versus forgiveness of insurgents.
Willingness to accept former insurgents back into the community.
To reduce social desirability bias, this outcome is complemented by indirect measures (see below).
4. Implicit Association Test (IAT) Outcome
Implicit attitudes are measured using an Implicit Association Test that captures the strength of automatic associations between:
Authorities vs. insurgents, and
Positive vs. negative emotional attributes.
The primary IAT outcome is the standardized D-score, computed using response latencies and error rates following established IAT scoring procedures. Higher values indicate stronger implicit associations favoring authorities over insurgents.
5. List Experiment and Graphical List Experiment Outcomes
Sensitive attitudes related to violence, trust, and social norms are measured using:
A standard list experiment with neutral control items and a sensitive item.
A graphical list experiment using images instead of text to reduce literacy constraints.
The primary outcome is the estimated prevalence of agreement with the sensitive item, inferred from differences in mean item counts between treatment and control lists.
6. Vignette-Based Evaluation Outcomes
Respondents evaluate fictional characters presented in vignettes that vary by role (e.g. authority figure or insurgent) and behavior. Outcomes include:
Trust in the character.
Perceived moral character and fairness.
Willingness to interact with or follow the character.
These measures capture normative judgments and moral evaluations in a controlled narrative setting.
7. Emotional Response Measures (Multimodal)
For a subset of open-ended questions, responses are captured using:
Video (facial expressions),
Audio (voice and speech patterns), and/or
Text (verbatim responses), depending on participant consent.
Primary emotional outcomes are constructed by extracting indicators of emotional valence and intensity from:
Facial emotion analysis (e.g. positive vs. negative affect),
Audio-based emotion and sentiment features,
Text-based sentiment and thematic analysis.
These measures are used to complement self-reported attitudes and capture affective responses that may not be fully expressed in survey scales.
8. Composite Primary Outcome
For some analyses, a pre-specified composite index combines standardized measures of:
Social cohesion,
Trust,
Attitudes toward violence and coexistence,
Implicit and indirect attitude measures.
This composite outcome is designed to capture overall changes in social relations and conflict-related attitudes resulting from the intervention.

In addition to the primary social cohesion / trust / conflict-attitudes endpoints, we will analyze the following secondary outcomes:
1. Economic well-being and expectations
- Current income bracket and expected income in 5 years.
- Occupational expectations for self and aspirations for children/youth in the household.
2. Perceived safety and fear
- Self-reported neighborhood and town security.
- Avoidance behavior (places the respondent fears going).
- Sleep disruption due to fear and ability to “rest mentally” at night.
3. Rumors, information environment, and perceived uncertainty
- Frequency of hearing security-related rumors.
- Behavioral responses to rumors (e.g., avoiding places due to rumors).
- Beliefs about whether rumors are usually true, and perceived ability to identify false information.

4. Community governance and dispute resolution
- Perceptions of how decisions are made in the neighborhood (leader-driven vs deliberative).
- Perceived participation of men/women/youth in decisions.
- Perceived fairness and transparency of dispute resolution.

5. Treatment experience and perceived quality of dialogue
- For participants who attended a dialogue session: perceptions of facilitator neutrality, respect, balance, and whether the dialogue increased willingness to cooperate.

The study uses a randomized controlled design with three groups. Eligible participants are randomly assigned at the individual level to one of the following conditions: (i) a control group that does not participate in any dialogue activity, (ii) a group invited to participate in a community dialogue facilitated by a religious leader, or (iii) a group invited to participate in a community dialogue facilitated by a government official.
Random assignment is conducted prior to the intervention. Participants in all groups are surveyed using the same instruments before and after the intervention. The comparison across groups allows the study to estimate the causal effects of participating in a community dialogue, as well as whether the type of facilitator influences outcomes.
The study is implemented in the town of Palma, Cabo Delgado, Mozambique.

The study is implemented as an individual-level randomized controlled trial among adult residents of the town of Palma, Cabo Delgado, Mozambique, who were previously displaced by the armed conflict and have since returned. Eligible participants are identified through household listing and screening procedures conducted prior to baseline data collection.
Following the baseline survey, participants are randomly assigned to one of three experimental groups: (i) a control group that does not participate in any dialogue activity during the study period, (ii) a treatment group invited to participate in a structured community dialogue facilitated by a religious leader, and (iii) a treatment group invited to participate in an identical dialogue facilitated by a government official. Randomization is conducted by the research team using a computerized procedure and is stratified by neighborhood to ensure balance across local areas.
The two treatment arms differ only in the institutional identity of the facilitator. All dialogue sessions follow the same written protocol, cover the same discussion themes, and are subject to the same moderation rules. Facilitators receive identical training and are instructed to remain neutral, encourage inclusive participation, and avoid advocacy or political messaging. Participants are informed that the activity is part of an academic research project but are not told that different facilitator types are being compared.
Dialogue sessions are conducted in small groups drawn from the same neighborhood. Participants assigned to treatment are invited to attend a single dialogue session during the intervention period. The control group does not receive any alternative activity during this time.
Data collection consists of a baseline survey conducted prior to randomization and a follow-up survey conducted after the intervention. Surveys are administered in person by trained enumerators using tablets and are available in Portuguese and relevant local languages. The survey instruments include standard survey questions, vignettes, indirect questioning techniques (list and graphical list experiments), and an Implicit Association Test. For a subset of questions, respondents provide open-ended answers that are recorded as text, audio, and/or video, subject to informed consent.
The primary analysis estimates intention-to-treat effects by comparing outcomes across the three randomized groups. All participants are analyzed according to their original assignment, regardless of attendance. Pre-specified secondary analyses explore heterogeneity by baseline characteristics, including prior exposure to conflict, displacement experiences, and baseline trust in leaders and institutions.
The experimental design is intended to identify both the overall effect of structured community dialogue and the differential impact of facilitator identity on social cohesion, trust, and conflict-related attitudes in a fragile, post-conflict setting.

public lottery: individuals randomly pick a number from a bag. that number will correspond to a cohort, and that cohort can either be control, religious leader meeting, or government official meeting.

cluster: cohort level (group of ~9 people). each cohort will be randomly allocated to control, religious leader treatment or government official treatment.

Yes

Religious leader treatment: 30 cohorts
Government official treatment: 30 cohorts
control group: 36 cohorts
total 96 clusters

864

30 clusters religious leader treatment
30 clusters government official treatment
36 clusters control.