Abstract
The reliability of water treatment systems is a critical infrastructure challenge. Previous national assessments have often relied on cross-sectional data, limiting the ability to analyse temporal trends and control for unobserved heterogeneity across facilities. This study replicates and extends a prior national assessment using a longitudinal framework. Its primary objective is to apply panel-data methods to estimate system reliability, isolating time-invariant facility-specific effects from operational performance trends. We construct a facility-level panel dataset from operational performance records. Reliability is modelled as a function of maintenance expenditure, operator training, and asset age using a two-way fixed effects estimator: $Reliability{it} = \alphai + \lambdat + \beta X{it} + \epsilon{it}$, where $\alphai$ and $\lambda_t$ are facility and year fixed effects. Inference is based on robust standard errors clustered at the municipal level. A 10% increase in preventative maintenance expenditure is associated with a 2.3 percentage point increase in system reliability (95% CI: 1.7 to 2.9). The facility fixed effects account for over 40% of the variation in the outcome, underscoring the importance of unobserved, time-invariant factors like design and location. Panel-data estimation provides a more rigorous methodological foundation for evaluating water treatment system performance by controlling for persistent, unobserved differences between facilities. National monitoring programmes should adopt longitudinal data collection protocols. Infrastructure investment decisions must account for significant baseline differences between facilities, which are revealed by the panel approach. infrastructure reliability, panel data, fixed effects, water treatment, maintenance, South Africa This study provides the first application of a two-way fixed effects panel model to national water treatment reliability data, demonstrating that a substantial portion of performance variation is attributable to time-invariant facility characteristics.