Municipal infrastructure asset management in Rwanda has historically relied on static, condition-based assessments, limiting the ability to forecast long-term efficiency and allocate resources proactively. This creates a significant policy challenge for sustaining gains in service delivery and capital planning. This policy analysis develops and evaluates a novel time-series forecasting model to quantify efficiency gains within municipal infrastructure asset management systems. The objective is to provide a predictive tool for evidence-based policy formulation and investment prioritisation. A vector autoregression (VAR) model, Yₜ = A₁Yₓ-₁ +. . . + AₚYₓ- + ₜ, was specified using national panel data on asset condition, maintenance expenditure, and service output indicators. Model parameters were estimated using feasible generalised least squares, with robust standard errors to account for heteroskedasticity. The model forecasts a 22% cumulative improvement in systemic asset management efficiency by the end of the forecast horizon, contingent on sustained current policy settings. Forecast error variance decomposition indicates that maintenance expenditure shocks account for approximately 60% of the variation in efficiency gains. The forecasting model demonstrates that targeted policy interventions in maintenance planning can yield substantial, measurable efficiency improvements in municipal asset management, moving beyond reactive approaches. Policy should institutionalise the use of dynamic forecasting models within municipal planning cycles. A dedicated national fund should be established, with allocations informed by model projections to maximise long-term efficiency gains. asset management, infrastructure, time-series forecasting, policy analysis, efficiency, municipalities This article provides the first application of a multivariate time-series model to forecast efficiency trends in a national municipal infrastructure system, offering a replicable methodological framework for evidence-based infrastructure policy.
Niyonzima et al. (Sat,) studied this question.