{ "background": "The governance of water treatment infrastructure in South Africa faces significant challenges, including variable adoption rates of advanced systems and a lack of robust, data-driven evaluation frameworks for policy intervention. Existing assessments often rely on aggregate statistics that mask critical regional and technical heterogeneities.", "purpose and objectives": "This policy analysis aims to develop and demonstrate a novel Bayesian hierarchical modelling framework to evaluate the determinants of water treatment system adoption. The objective is to provide a methodological tool for infrastructure governance that quantifies adoption drivers and their uncertainties, informing targeted policy.", "methodology": "A Bayesian hierarchical logistic model is constructed, formalised as $y{ij} \\sim \\text{Bernoulli}(p{ij}), \\; \\text{logit}(p{ij}) = \\alpha{j[i]} + \\beta X{ij}$, where $y{ij}$ is the adoption status for facility $i$ in municipality $j$, $\\alpha_j$ are municipality-level random effects, and $\\beta$ are coefficients for facility-level covariates $X$. The model integrates multi-level data on technical, financial, and institutional factors.", "findings": "The analysis reveals substantial regional variation, with municipality-level random effects showing a posterior credible interval of [-2.1, 1.8] on the log-odds scale. A key concrete finding is that operational budget allocation is a stronger predictor of adoption than initial capital investment, with a 10% increase in operational budget share associated with a 15% higher probability of adopting advanced treatment systems.", "conclusion": "The Bayesian hierarchical model provides a superior, evidence-based framework for diagnosing adoption barriers in water treatment infrastructure, capturing both systemic and localised factors essential for effective governance.", "recommendations": "Policy should shift focus towards securing sustained operational expenditure alongside capital projects. Infrastructure governance bodies should adopt probabilistic, multi-level modelling to prioritise interventions in underperforming regions and allocate resources based on quantified drivers of adoption.", "key words": "Infrastructure governance, Bayesian statistics