"background": "The adoption of engineered water treatment systems in sub-Saharan contexts is critical for public health, yet robust longitudinal evidence on adoption drivers remains sparse. Previous evaluations of such interventions have often relied on cross-sectional data or lacked rigorous counterfactual analysis, limiting causal inference. ", "purpose and objectives": "This study replicates and extends a prior quasi-experimental evaluation of community-scale water treatment adoption. Its objectives are to verify the original study's effect estimates using an expanded longitudinal dataset, to conduct a comprehensive diagnostic of the quasi-experimental design's robustness, and to test a refined model incorporating maintenance cost variables. ", "methodology": "A difference-in-differences framework is employed, leveraging phased programme rollout across villages. The core statistical model is Adoption{it = \0 + \1 (Treatedit) + \2Xit + \ + \ +, where Xit includes household and village covariates. Robust standard errors are clustered at the village level. Diagnostics include placebo tests, balance checks, and an assessment of parallel trends. ", "findings": "The replication confirms a positive adoption effect but of a smaller magnitude than originally reported. The adjusted programme effect is a 12-percentage-point increase in sustained household adoption (95% CI: 8 to 16). Diagnostic tests indicate the parallel trends assumption holds only after controlling for baseline water source type. A key new finding is that adoption elasticity to maintenance costs is significant (\ = -0. 21, p < 0. 01). ", "conclusion": "The original study's central finding is robust to replication, though the effect size is more modest. The quasi-experimental design is valid only when accounting for pre-intervention heterogeneity in water infrastructure. Long-term adoption is highly sensitive to ongoing operational costs, a previously underemphasised factor in engineering implementation models. ", "recommendations": "Future engineering interventions must integrate life-cycle cost analysis into adoption models from
Diagne et al. (2019) studied this question.