Estimations of pooled dynamic panel data model with time-space dependence of selected Sub-Saharan African urban agglomerations, 2000–2020
Estimations of pooled dynamic panel data model with time-space dependence of selected Sub-Saharan African urban agglomerations, 2000–2020

Summary/Abstract: The ongoing progress in the spatial econometrics literature includes an eternal debate on suitable methods to comprehensively estimate time-space dependence. Nevertheless, techniques for carrying out such studies are inherently confined to purely cross-sectional or time-series studies. Thus, this paper reviewed different estimation techniques previously applied in the time-space literature, particularly the general dynamic panel model and approaches integrating pooled OLS, FE, RE, GMM-difference (Diff), and GMM-systematic (Sys) in estimating the time-space dynamic panel model. To achieve this objective, The authors determined the predictive efficiency of each method by evaluating the root mean square error (RMSE) estimators of Monte Carlo simulations of urban agglomeration spanning 2000–2020 for 22 Sub-Saharan African countries. The results indicate that the estimates arrived at using the conventional methods (pooled OLS, FE, and RE) are not consistent if endogeneity problems plausibly exist. Satisfactorily, the estimation procedures based on GMM methods, including the assumption of sequential exogeneity of the independent variables, offer ideal options for overcoming problems of endogeneity, heterogeneity, and feedback effects. Explicitly, the GMM-Sys estimators are characterized by low bias and excellent efficiency and are hence ideal for modeling spatial concepts with time-space dependence.

• Details
• Contents