Replication and Methodological Evaluation of a Difference-in-Differences Model for Transport Depot System Reliability in Nigeria, 2000–2026

{ "background": "Difference-in-differences (DiD) models are increasingly applied in engineering systems analysis to evaluate policy interventions, yet their methodological robustness in complex, real-world infrastructure contexts requires scrutiny. A seminal study applied a DiD framework to assess the impact of a centralised maintenance programme on transport depot reliability, presenting significant positive effects.", "purpose and objectives": "This study aims to replicate the original DiD analysis and conduct a rigorous methodological evaluation of its design and assumptions. The objective is to verify the reported findings and assess the model's sensitivity to key specification choices common in engineering applications.", "methodology": "We execute an exact replication using the original dataset and code, followed by a series of robustness checks. These include testing the parallel trends assumption, evaluating sensitivity to alternative fixed effects structures, and employing cluster-robust standard errors at the depot level. The core model is specified as Ydt = \β0 + \β1 + \β2 + \δ \(Treatd \· ) + \εdt.", "findings": "The exact replication confirms the original point estimate for the treatment effect (\δ = 0.18). However, methodological evaluation reveals the result is not robust: the effect becomes statistically insignificant at the 5% level when accounting for temporal autocorrelation, with the 95% confidence interval spanning from -0.04 to 0.40. The parallel trends assumption for pre-intervention periods is also violated for a key subsystem reliability metric.", "conclusion": "The reported positive effect of the centralised maintenance programme on depot reliability is not methodologically robust. This underscores the critical importance of rigorous diagnostic testing for DiD applications in infrastructure engineering studies, where system dynamics can violate core model assumptions.", "recommendations": "Future engineering studies employing DiD should pre-test the parallel trends assumption, explicitly report robustness checks for error structure,