A1436
Title: Predicting income inequality: A non-stationary spatial covariance approach
Authors: Ya-Mei Chang - National Chung Cheng University (Taiwan) [presenting]
Yi-Chen Lin - Tamkang University (Taiwan)
Pei-Hsin Huang - Tamkang University (Taiwan)
Truc Thuy Thi Le Nguyen - Tamkang University (Taiwan)
Abstract: Inequality predictions are vital for effective cohesion policy. A spatial model for the Gini index is proposed to account for the heteroskedasticity and non-stationary spatial dependence found in European regional income inequality data. Model residuals are represented as a sum of multiple basis functions, stationary processes, and a white-noise process. To enhance computational efficiency, the least absolute shrinkage and selection operator (lasso) is employed to simultaneously select relevant processes and estimate parameters. An analysis of data from 29 European countries over the period 2009-2017 reveals that the Baltic and Nordic regions exhibit high variance and strong spatial correlation. In contrast, western and southern europe show weak spatial dependence. By accommodating non-stationary spatial covariance, the proposed spatial model reduces the mean squared error by approximately 20\% compared to standard non-spatial models that assume independent and identically distributed (i.i.d.) residuals.
