Spatial patterns and determinants of residential burglary in Gauteng for 2011/2012

Abstract

South Africa ranks amongst the highest in terms of reported crimes in the world, yet there are not many published research papers that study the spatial dynamics of crime. This report used statistical and spatial methods to explore determinants and patterns of burglary in Gauteng, South Africa. The variables considered for modelling were informed by the Routine Activities (RA) theory and other local studies. The methods of analysis used were ordinary least squares (OLS) regressions and geographically weighted regressions (GWR) - an extension of the OLS regression used when there is spatial heterogeneity in the data. The findings show that the measure of social cohesion and physical neighbourhood characteristics such as roads density, percentage of the population unemployed together with the percentage of the population that moved in the last 5 years were significant predictors of residential burglary. The findings provide stronger support for the social disorganisation (SOD) than for the RA theory in the South African context. The spatial patterns suggested that there were hot spots, colds spots clusters in certain parts of the study region. Evidence of heteroscedasticity of the influential factors of burglary was found, although marginally. Although heteroscedasticity at 5 percent significance level was not found in the model variables, the results of the GWR improved the OLS model at 10 percent significance level. The GWR results implied that different strategies of crime prevention could be implemented in different parts of Gauteng.