Testing the income convergence hypothesis under new spatial weight matrices: The European example

This study proposes innovative approaches to spatial econometric methods to examine income convergence in Europe. The primary objective of the research is to shape the spatial weight matrix, constructed based on neighborhood relationships, in a way that accounts for the unequal effects of each neighbor, thereby achieving more accurate predictions. Using the proposed spatial weight matrices, applications have been carried out within the framework of the "convergence hypothesis" for European countries during the period from 2000 to 2021. In this context, absolute and conditional convergence hypotheses have been tested through "fixed effects" and "random effects" panel data models established with "queen," "threshold," "two-threshold," and "neighbor-asymmetric" weight matrices, as well as through "spatial lag model," "spatial error model," "spatial Durbin model," and "general spatial model." When examining the models based on the proposed weight matrices, it is observed that models constructed with the "neighbor" weight matrix generally exhibit the best performance. The performance of models established with threshold and two-threshold spatial weight matrices cannot surpass that of models constructed with conventional weight matrices. Therefore, findings indicate that only the neighboring asymmetric weight matrix out of the proposed threshold, two-threshold and neighboring asymmetric weight matrices gives better results than the conventional weight matrix. According to the analysis conducted with the neighbor-asymmetric weight matrix, income divergence exists in Europe. The identification of income divergence suggests that, rather than developing countries catching up with developed countries, the economic development gap between them is widening. Countries with low initial income levels have not converged toward those with high initial income levels, as posited by the neoclassical model; instead, they have exhibited lower growth rates. This situation indicates that the existing inequality in Europe persists or has worsened. For policymakers, considering local economic conditions and social structures emerges as a critical factor in developing strategies to reduce income inequalities. As a result, in this study, it has been shown that the spatial effect of neighboring observation values higher or lower than the value of any unit may be different from other neighboring units. Therefore, to achieve better estimates by obtaining higher log-likelihood values, it would be beneficial to utilize the proposed neighbor-asymmetric spatial weight matrix in future spatial econometric studies.