Modelagem espacial do crescimento do PIB em cenários distintos de desempenho agrícola do estado do RS

Abstract

Spatial variability of the economy growth highlights local features to be taken into account when policy makers plan on regional development. In this sense, this work sought to understand the spatial association between the growth of the Gross Domestic Product (cPIB) and the growth of Value of Agricultural Production (cVPA) over the state of Rio Grande do Sul (RS) throughout the years from 2004 to 2015. In the first chapter spatial exploratory analyses were carried out with different kernels to depicted local averages, standard deviations, z-scores and correlations. Following, a Monte Carlo test was applied to assess whether the relationship between cPIB and cVPA shows to be non-stationary in the RS space. It was found the force of association varies along the western and eastern regions of the state. The western region showed the strongest positive associations between cPIB and cVPA while the eastern region showed the weakest negative association. In the second chapter results from geographically weighted regression models (GWR) were compared against those from global models. GWR parameterization was carried out by setting the Boxcar kernel with the optimum bandwidth defined through a cross-validation approach. The GWR models overtook global regression over all criterion used to evaluate model performances. GWR models presented smaller global error, higher overall determination coefficient and less spatially autocorrelated residuals. It was possible to verify that the northwest region has the highest conversion of cVPA into cPIB while the performance of agriculture has little effect on the economy of the metropolitan region. The relationship between cPIB and cVPA oscillates in time and space in non-stationary process that has modulated the economic and agricultural development over the RS state.