Existing criminal geography research has always focused on the number of cases, but neglected their severity. Limited by the availability of data, cross-sectional analysis was more universal than longitudinal, which would cause endogeneity problems. Furthermore, spatial dependence of different independent variables has not been systematically examined. In order to fill these gaps, this study utilized criminal case records, points of interest, and road network data from 2012 to 2017 in Beijing to explore spatial pattern of crime number and its harm, and clarify the role of urban built environment in their forming process. In order to measure the extent of crime harm, criminal penalty by judicial authorities was used as the indicator. First, this research demonstrated that both crime number and crime harm showed geographical concentration and agglomeration. However, concentration extent of crime harm was higher than crime number, but agglomeration extent of crime harm was lower than crime number. The degree of harmspot's stability was weaker than crimespot, therefore geographical analysis of crime number cannot fully reveal the spatial pattern of social harms caused by criminal cases. Second, permeable space postulated by street eyes theory was unable to inhibit criminal activities, while high density land use, diversified urban functions, convenient transport branch networks, and close spatial proximity to crime-prone areas could lead to the increase of crime number and crime harm, which is consistent with the prediction of defensible space theory. Additionally, crime harm was more susceptible to conducive built environment factors. Third, most influences caused by various characteristics of the built environment showed spatial dependence. Specifically, density, diversity, design of road network, as well as commercial place had agglomeration spillover effect, which meant that both local and neighboring environmental elements had positive relationship with native criminal activities. In contrast, local educational institutions, parks, squares, hotels, bus stops, parking lots, and residential areas that were strictly supervised had no effect or negative effect on native criminal activities, but neighboring attractors had significant positive impact on native criminal patterns. Because of this spatial competition effect, security measures should not confine to local areas and attractors.