This paper reports the results of a field experiment to assess the collaborative effects of community participation in the Ugandan oil and gas sector. Our research design assesses collaborative impacts as relational between community members and different decision-makers in the sector and measures these impacts from the point of view of local people. Local people often face power imbalances in collaborative governance. Decision-makers are increasingly attempting to mitigate such imbalances to improve outcomes for communities, but little experimental evidence exists showing the impact of such efforts. Using multilevel ordered logit models, we estimate positive treatment effects, finding that encouraging the equitable participation of communities improves collaboration with other actors. Next, we use machine-learning techniques to demonstrate a method for targeting communities most likely to benefit from the intervention. We estimate that purposefully targeting communities that would benefit most yields a treatment effect about twice as large, relative to pure random assignment. Our results provide evidence that interventions mindful of community needs can improve collaborative governance and shows how such communities can be most effectively targeted. The experiment took place across 107 villages (53 treatment and 54 control) and the unit of statistical analysis is the household, where we report outcomes measured from 6,062 household surveys (approximately half at baseline and half at endline).

中文翻译：

---

社区如何从协作治理中受益：乌干达石油和天然气的实验证据

本文报告了一项实地实验的结果，以评估社区参与乌干达石油和天然气部门的协作效果。我们的研究设计将协作影响评估为社区成员与该部门不同决策者之间的关系，并从当地人的角度衡量这些影响。地方人民在协同治理中经常面临权力失衡。决策者越来越多地试图减轻这种不平衡以改善社区的结果，但很少有实验证据表明这种努力的影响。使用多级有序 logit 模型，我们估计了积极的治疗效果，发现鼓励社区的公平参与可以改善与其他参与者的合作。下一个，我们使用机器学习技术来展示一种针对最有可能从干预中受益的社区的方法。我们估计，相对于纯随机分配，有目的地瞄准最受益的社区会产生大约两倍的治疗效果。我们的结果提供的证据表明，考虑到社区需求的干预措施可以改善协作治理，并展示了如何最有效地针对此类社区。实验在 107 个村庄（53 个处理组和 54 个控制组）进行，统计分析的单位是家庭，我们报告了 6,062 个家庭调查的测量结果（大约一半在基线，一半在末线）。我们估计，相对于纯随机分配，有目的地瞄准最受益的社区会产生大约两倍的治疗效果。我们的结果提供的证据表明，考虑到社区需求的干预措施可以改善协作治理，并展示了如何最有效地针对此类社区。实验在 107 个村庄（53 个处理组和 54 个控制组）进行，统计分析的单位是家庭，我们报告了 6,062 个家庭调查的测量结果（大约一半在基线和一半在末线）。我们估计，相对于纯随机分配，有目的地瞄准最受益的社区会产生大约两倍的治疗效果。我们的结果提供的证据表明，考虑到社区需求的干预措施可以改善协作治理，并展示了如何最有效地针对此类社区。实验在 107 个村庄（53 个处理组和 54 个控制组）进行，统计分析的单位是家庭，我们报告了 6,062 个家庭调查的测量结果（大约一半在基线和一半在末线）。