Interventions to promote social cohesion in sub-Saharan Africa

Abstract

This article presents a synthetic review of impact evaluations examining the effectiveness of community-driven development (CDD) and curriculum interventions in improving social cohesion in sub-Saharan Africa. The review found weakly positive impacts of CDD and curriculum interventions on social cohesion outcomes, although only two findings were replicated across studies: one positive and one negative. Causal chain analysis of data on implementation and contextual factors relating to the CDD interventions found that broad and substantive participation was often lacking, suggesting the interventions have often not been carried out in accordance with the theory of CDD.

Keywords:

• social cohesion
• social capital
• evaluation
• Africa
• community-driven development
• participation

Acknowledgements

The authors sincerely thank the study authors for their interesting and important work. They are also grateful to research assistants Travis Coulter and Elizabeth Edelstein, to the numerous researchers who replied to their listserv calls, to Hugh Waddington and Howard White from 3ie, and to two anonymous reviewers. The authors also gratefully acknowledge 3ie/GDN for supporting this review and Columbia University's Institute for Social and Economic Research and Policy for administrative assistance.

Notes

1. The MSMS was designed by Sherman et al., based on Cook and Campbell's (1979) seminal work, as a tool to evaluate methodologies for their review of over 500 crime-prevention interventions for the US Congress (for a summary, see Sherman et al. 1998). The scale has since been considered a key resource for appraising quantitative studies (for instance, Government Social Research undated). The MSMS is a five-point scale ranging from weakest (one) to strongest (five) in terms of the robustness of the causal claims, or internal validity. In short, only studies with both a strong comparison group and pre-test and post-test data provide solid evidence of impact.

2. We could have imposed our own solution to the problem by extracting a single outcome score by factorising the separate indicators. However, without access to the raw data from each study, covariances between various indicators within studies could not be estimated and proper factorisation was not possible. As such, we worked with collections of estimates for each study, appreciating that this makes it difficult to determine whether the findings are globally significant.

3. Tables with details of the data collected from the studies for the causal chain analysis are available in the appendix of King et al. 2010.

4. Neither of these projects involved sub-project selection by the community so this is perhaps not surprising.