Econometria espacialdiscutindo medidas para a matriz de ponderação espacial

Resumo

Considering the increasing use of methods, theories and techniques from urban and regional sciences, spatial economics and econometrics at overlapping areas of economics and public policy, it becomes relevant to the field of public policy analysis to understand and comprehend such set of knowledge. The main objective of this dissertation is to critically discuss aspects of the spatial econometrics literature and its applicability, with a special focus on the comparison of different methods and spatial weight matrices. This study tries to analyze different alternative measures to construct spatial weight matrices, as well as test by experimental simulation and empirical application the robustness of some of these alternative models. Given the indications at the literature and evidences show at this study, the more reliable strategy for a spatial specification search seems to be the use of model’s estimates in conjunction with a few specification tests. Moran’s I is a good test to identify the presence of spatial correlation, while the LM test on the dependent variable is a good test to identify this specific kind of spatial correlation. It is possible to notice that, in general terms, there is not so much difference in terms of expectation on non-spatial coefficients, exception for the constant term, when spatial estimates are compared with traditional ones. From this, the indication that can be made is that the use of spatial econometrics techniques should take place when the estimation of the spatial correlation parameters are the analysis target. Comparing the use of the contiguity matrix in place of a distance based matrix, it is possible to verify that the contiguity matrix is not, in general terms, a good substitute. When used on spatial regressions, the spatial coefficients are partially estimated and the error autocorrelation, when non-existent, is wrongly identified.

Ficha do documento

Tipo

Dissertação

Ano

2006

Instituição

Fundação Getulio Vargas

Fonte

Repositório da FGV

Idioma

Português

Acesso

Não informado

Identificador

oai:repositorio.fgv.br:10438/2451

Temas

Políticas PúblicasFinanças PúblicasDesenvolvimento RegionalAdministração Pública

Palavras-chave

Finanças públicasSimulação de Monte CarloMatriz de ponderação espacialEconometria espacialSpatial econometricsSpatial weight matrixMonte Carlo simulationAdministração públicaEconometriaMétodo de Monte Carlo