Evaluation of Coalition-Formation Stability Bounds for Artificial Intelligence Agriculture Advisory Councils Under Sequential Member Defection: A Governance Durability Estimation Strategy

Artificial intelligence agriculture advisory councils are typically designed as standing coalitions of ministries, extension agencies, farmer organisations and technology providers, yet their durability under sequential member defection is rarely specified in advance. The analysis is a replication-based evaluation, not an observational study: no councils were observed and no defection events were recorded. Using a between-cluster standard deviation of 6.0, a residual standard deviation of 4.5 and an intraclass correlation of 0.64, power was estimated across 2000 design cells for cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points. The design would require 52 clusters to detect a 5 percentage point defection effect at 80 per cent power. The before-after-control-impact estimator is a defensible durability bound only when the intraclass correlation is treated as a design parameter rather than a nuisance, and it specifies the conditions under which advisory council evaluations can distinguish genuine coalition stability from cluster-level heterogeneity.