살인범죄의 원인에 대한 거시적 분석: GIS를 활용한 공간회귀모델링

Abstract

Every single social phenomenon including crime has a spatial aspect and it is spatially dependent, in general. The spatial dependence causes spatial autocorrelation in macro-level crime analysis, which in turn makes the standard errors of regression coefficients inflated and biased. Eventually, the property of best linear unbiased estimate (BLUE) of the ordinary least squares estimates does not hold any more. To solve this problem, spatial regression modeling is required. In Korean criminal justice area, however, no spatial regression analysis has been attempted yet. Thus, this study tried the spatial regression modeling of homicide rates in 241 districts within the past three years. The result of Global Moran's I showed that homicide rates are spatially dependent. The spatial lag model discovered that divorce rate is the most important homicide indicator within Korean communities, which implied that greater social bond (collective commitment to community well-being) as well as the presence of an extended support system could be responsible for lower homicide rates. More spatial analyses for diverse neighborhood sizes and characteristics, various crime types, and cross-level interaction effects are to be made for effective crime prevention policies that fit each neighborhood's unique feature.