Application of spatial regression models to income poverty ratios in middle delta contiguous counties in egypt

Abstract

Regression analysis depends on several assumptions that have to be satisfied. A major assumption that isnever satisfied when variables are from contiguous observations is the independence of error terms. Spatialanalysis treated the violation of that assumption by two derived models that put contiguity of observationsinto consideration. Data used are from Egypt's 2006 latest census, for 93 counties in middle delta sevenadjacent Governorates. The dependent variable used is the percent of individuals classified as poor (thosewho make less than 1$ daily). Predictors are some demographic indicators. Explanatory Spatial DataAnalysis (ESDA) is performed to examine the existence of spatial clustering and spatial autocorrelationbetween neighboring counties. The ESDA revealed spatial clusters and spatial correlation betweenlocations. Three statistical models are applied to the data, the Ordinary Least Square regression model(OLS), the Spatial Error Model (SEM) and the Spatial Lag Model (SLM).The Likelihood Ratio test andsome information criterions are used to compare SLM and SEM to OLS. The SEM model proved to bebetter than the SLM model. Recommendations are drawn regarding the two spatial models used.