Journal Design Engineering Masthead
African Civil Engineering Journal | 23 April 2001

Bayesian Hierarchical Modelling of Municipal Infrastructure Asset Adoption Rates in Senegal

A Methodological Evaluation Dataset (2000–2026)
M, a, m, a, d, o, u, N, d, i, a, y, e, ,, A, ï, s, s, a, t, o, u, D, i, a, g, n, e
Bayesian modellingMunicipal infrastructureAsset managementSub-Saharan Africa
Bayesian hierarchical logistic growth model estimates municipal-level adoption rates.
Dataset integrates national surveys, municipal records, and engineering project inventories.
Method quantifies uncertainty in projections, with credible intervals widening over time.
Framework captures significant spatial clustering in infrastructure adoption trajectories.

Abstract

{ "background": "Municipal infrastructure asset management in developing nations often relies on aggregated national statistics, which obscure sub-national variation and hinder localised planning. A methodological gap exists for robust, probabilistic models that can estimate adoption rates of diverse infrastructure systems at the municipal level while accounting for spatial and temporal heterogeneity.", "purpose and objectives": "This Data Descriptor presents a structured dataset and methodological framework for evaluating municipal infrastructure adoption. The primary objective is to provide a reproducible, Bayesian hierarchical modelling approach to estimate and project adoption rates for water, sanitation, and transport assets, enabling uncertainty quantification for engineering decision-making.", "methodology": "A longitudinal dataset was constructed from national surveys, municipal records, and engineering project inventories. The core statistical model is a Bayesian hierarchical logistic growth model: $\\text{logit}(p{ijt}) = \\alphaj + \\beta{1i} X{ijt} + \\epsilon{ijt}$, where $p{ijt}$ is the adoption rate for infrastructure type $i$ in municipality $j$ at time $t$, $\\alphaj \\sim \\text{Normal}(\\mu\\alpha, \\sigma\\alpha)$ are municipality-level random effects, and $\\epsilon{ijt}$ captures observation error. Hamiltonian Monte Carlo sampling was used for inference.", "findings": "The methodological evaluation demonstrates that the model successfully captures significant spatial clustering in adoption trajectories, with posterior credible intervals for municipal-level growth rates providing actionable uncertainty bounds. A key illustrative result is that the model estimates a central 90% credible interval for the adoption rate of piped water connections in a typical municipality to widen by approximately 15 percentage points when projecting five years forward, highlighting the critical importance of probabilistic forecasting.", "conclusion": "The dataset and modelling framework offer a substantial improvement over deterministic projection methods for infrastructure planning. The Bayesian hierarchical approach formally integrates uncertainty and spatial dependence, providing a more reliable evidence base for municipal engineers and policymakers.", "recommendations": "Adoption of this probabilistic modelling framework