Autor: Quintaba Pablo Aníbal*, Herrera Gómez Marcos**

Institución: *UNCuyo, **Universidad Nacional de Rio Cuarto

Año: 2023

JEL: C21, C23


The spatial weighting matrix plays a pivotal role in spatial econometrics and remains an active area of research. In this study, we apply recent advancements in machine learning for estimating the spatial weights matrix in econometric models. By employing LASSO strategies and incorporating geographical restrictions, we directly derive the spatial weighting matrix from the available data. This approach removes the necessity for arbitrary criteria set by researchers. As an empirical example, we explore the relationship among the salary of registered salary workers of Argentine provinces. Using monthly information between 2014 and 2022, we identify breakpoints in the wage time series and determine whether the breaks occur due to the movements within each province or due to neighboring provinces.