Online appendix B: Further analyses B.1 Further descriptive analysis

We then ran three moderator analyses on these results to detect whether the thethe relationship between interventions and social cohesion was affected by : 1) the risk of bias level of the included studies, 2) the study design,, or 3) the length of the exposure to the intervention. [...] We then ran three moderator analyses on these results to detect whether the relationshiprelationship between interventions and social cohesion was affected by: 1) the risk of bias level of the included studies, 2) the study design, or 3) the length of exposure to the intervention. [...] The authors posit that these tensions may have caused of reflect the needs of volunteers in others' the worsening of tolerance values among individuals in the control group, and the target communities, ie through that participation in the programme was successful in preventing these values community; (iii) collaborative contact, change from worsening, to a statistically significant degree. [...] (-) Government turnover blocked full implementation: One month following the training of facilitators, the local governor pressured the chairman of the organisation sponsoring the intervention (not the local NGO partner, but a national NGO affiliated with the World Bank) to shut the programme down. [...] (+) Group listening may facilitate impact: The evaluation design required that participants listen to the radio broadcast in groups, and the authors note that the effects are therefore measuring the combination of the drama + the effect of listening in group.