This paper uses data from Registration of Real Estate Actual Transaction Prices, and position data of housing theft, motor vehicle theft, motorcycle theft, bike theft, robbery, and forceful taking in Taipei to investigate whether housing prices are lower in the regions with high crime concentrations, and higher in the regions with low crime concentrations as expectation. We use the traditional hedonic pricing method with the concept of the spatial autoregressive model to analyze. The empirical result shows that when we control the six criminal variables simultaneously, these types of crime will lower housing prices significantly except bike theft.