Ensaios utilizando econometria espacial: aplicações em modelos de decisão de famílias em áreas de risco e precificação e demanda por gasolina em postos de combustíveis de Fortaleza -CE

Abstract

The thesis presents a set of 3 articles that seek to demonstrate, from spatial analysis, whether the agents' decisions are influenced by the actions of their neighbors, that is, whether there is a spatial pattern in decision making. The first chapter analyzes the decision of the families that were removed from the banks of the Cocó River in Fortaleza (Rio Cocó Project), and who had to decide between receiving compensation for the improvements made to the property they lived in or being relocated in housing projects of the Minha Casa Minha Vida Program. We sought to investigate whether individual decisions were spatially correlated with neighbors' decisions. The data used for the analysis are from the monitoring of the project, making it possible to observe characteristics of the families, the residence and also the georeferencing of the houses. Thus, a spatial probit model was estimated (WILHELM, DE MATOS; 2013), finding evidence favorable to the spatial effect of interaction between neighbors. The second chapter of the thesis aims to test and model the existence of neighborhood effect and spatial autocorrelation in models of demand for gasoline in Fortaleza, using microinformation of quantity and price of gasoline at the level of gas stations in Fortaleza. For this purpose, information from electronic invoices from the Treasury Department of the State of Ceará (SEFAZ) is used, and spatial panel data models are estimated using the Generalized Moments Method (GMM) (MILLO, et al, 2012; SANTOS E FARIA, 2012). The main results found reveal that neglecting the spatial correlation in demand models implies bias in the estimation of gasoline price elasticity. The third chapter of the thesis analyzes whether the process of fuel price adjustments after a cost shock presents a spatial pattern among gas stations in Fortaleza. More specifically, it aims to analyze the existence of a spatial pattern in the increase in prices charged, making the gas stations closest to each other present similar price variations. For the analysis, matrices of spatial weights were created at different distances and the value of Moran's I was observed for the observation and spatial autocorrelation test. The main results found show that there is a spatial pattern of the percentage change in prices and that as the distance between stations increases, the strength of this spatial pattern becomes smaller.