빈집과 절도･폭력범죄의 관계에 대한 공간가중회귀모형의 적용

With the application of a geographically weighted regression(GWR) model, this study examined the relationship between the abandoned house and theft･violent crime in S city, South Korea. This study found that the observed variables within census output areas (geographic units of analysis) were spatially dependent. Spatial regression models assuming the existence of spatial dependence among variables might be suitable than traditional OLS models; however, the spatial regression models fail to consider the existence of spatial non-stationarity at the local level, whose existence is common in reality. Thus, this study applied a GWR model in examination of the relationship between the abandoned house and crime by controlling spatial dependence and non-stationarity. This study found that a 1% increase in abandoned houses in an output area is estimated to increase 0.128 cases of theft. On the other hand, this study did not find any statistically significant effect of the abandoned house on violent crime. Moreover, increases in alcohol outlet and street light density are estimated to positively impact the number of theft and violent crime. Based on the finding that the relationship between structural variables and crimes varied in its direction and strength at the local level, this study suggested that customized crime prevention activities should be implemented within in localized areas.