Replication and Bayesian Hierarchical Modelling of Power-Distribution Equipment Adoption in Rwanda, 2000–2026

"background": "The adoption of power-distribution equipment is a critical infrastructural process in developing economies. Prior studies have modelled adoption rates using deterministic frameworks, which often fail to account for regional heterogeneity and uncertainty inherent in long-term infrastructure planning. ", "purpose and objectives": "This study aims to replicate and critically evaluate a previous national adoption model using a Bayesian hierarchical framework. The objective is to produce more robust probabilistic forecasts and quantify the uncertainty in the adoption rates of key equipment types across different regions. ", "methodology": "A replication study was conducted, applying Bayesian hierarchical modelling to historical adoption data. The core model is specified as y{ij \ (\ + \ xij, \²), with \ \ (\\, \²), where yij is the adoption rate in district i of region j, and x₈₉ represents a socio-economic covariate. Posterior distributions were estimated using Markov chain Monte Carlo methods. ", "findings": "The Bayesian model indicates a 95% credible interval for the national adoption rate of composite pylons of 0. 34, 0. 41 by the forecast horizon, which is notably wider and more conservative than the point estimate of 0. 42 from the original study. The hierarchical structure revealed significant regional variation, with the posterior probability of the adoption rate in the Eastern Province exceeding the national average being 0. 87. ", "conclusion": "The replication confirms the general trajectory of equipment adoption but demonstrates that failing to account for regional random effects and forecast uncertainty can lead to overconfident and potentially misleading planning assumptions. ", "recommendations": "Infrastructure planning models should incorporate hierarchical structures to capture sub-national variation and report probabilistic forecasts. Utilities are advised to use credible intervals for risk-aware resource allocation. ", "key words": "Bayesian hierarchical model, infrastructure planning, replication study, power distribution, adoption forecasting, uncertainty