Rural development NGO accountability intermediaries and smallholder grievance escalation in Uganda: a difference-in-differences analysis of village agent mandates

Village agents are the front-line intermediaries through which rural development NGOs in Uganda receive and escalate smallholder grievances, and the question of whether a change in their mandate raises or lowers escalation is routinely evaluated with a before-after-control-impact design. the present analysis characterises the operating characteristics of that design under a declared variance structure rather than reporting field observations. It specifies a difference-in-differences estimator for a binary escalation outcome, fixes a between-cluster standard deviation of 6.0, a residual standard deviation of 4.5 and an intraclass correlation of 0.64, and estimates power across 2000 replications for each of the design cells formed by 8, 16, 24 and 32 clusters and effects of 3, 5 and 8 percentage points. Power is low at 8 and 16 clusters for every effect size, rises steeply at 24 clusters, and reaches conventional levels at 32 clusters for the 5 and 8 percentage point effects; 56 clusters are required for 80 per cent power at a 5 percentage point effect. The minimum detectable effect at 80 per cent power falls from a value above the largest tested effect at 8 clusters to a value below 5 percentage points at 32 clusters. The results imply that published null findings on village agent mandates in Uganda are frequently underpowered rather than evidence of no effect, and that the intraclass correlation, not the number of smallholders surveyed, is the binding constraint on the design.