Explaining and Predicting National Logistics Performance Across the Survey-Based Logistics Performance Index Era: Institutional Capacity, Digital Connectivity, and Explainable Decision Support

Abstract

While national logistics performance affects trade reliability and market access, as well as supply-chain resilience, it is unclear how much institutional capacity, digital connectivity, and trade integration contribute to this performance, both between countries and over time. This study analyzes explanatory patterns at all time points of the Logistics Performance Index (LPI) survey period (2007-2023), and predictive patterns through a seven-wave panel of 2007-2023. This is the primary sample of 975 observations from 160 economies. The principal inferential framework is a two-way fixed effects (2WFE) model and country-clustered standard errors, with Hausman and Mundlak diagnostics and specific robustness tests to complement, and Driscoll-Kraay sensitivity estimates applied. There is a positive relationship between Government Effectiveness, Regulatory Quality, Rule of Law, and LPI performance, which holds through various measurement, balanced-panel estimation, leave-one-wave-out, and 2023-specific tests.Government Effectiveness, Regulatory Quality and Rule of Law are positively related to LPI performance and are stable across various measurement and balanced-panel estimation, leave-one-wave out and 2023-specific tests. The model preferred is the fixed-effects model, in which no significant differences are found in the use of the Internet or in the level of openness of trade. The digital–institutional interaction is positive, but sensitive to covariance choice and other measures of digital. Following FDR correction, institutional capacity continues to be correlated with customs and infrastructure and logistics competency. The heterogeneity in income groups is only identified for the GDP-per-capita slope. On the held out 2023 wave, the R² values are around 0.77 for country-grouped machine learning and 0.808 for Random Forest. The top predictive features found by SHapley Additive exPlanations (SHAP) are Government Effectiveness, economic development, and fixed broadband. Comparing the explanatory signals across the other variables, the overall signal of institutional capacity is more stable than those of digital associations, which are only stable depending on how they are measured and specified.