Crime Measures and Housing Prices: an Analysis Using Quantile Regression and Spatial Autocorrelation
研究犯罪指标与房价的关系，使用西雅图2008-2020年数据，通过分位数回归和空间自相关分析发现，不考虑空间自相关时犯罪率上升1%使房价下降0.55%，考虑后则使房价上升0.80%，且距犯罪热点距离与房价负相关。
Abstract Crime is a disamenity, so buyers should be willing to pay more for a house (all else equal) in a low crime area, suggesting that high crime rates depress housing prices. Conversely, it is plausible that criminals prefer wealthier areas because of the higher expected returns from their transgressions. This study examines the link between measures of crime and prices of residential housing. Our data begin in 2008 and end in 2020 for Seattle, Washington, including all reported felonies (756 , 304); 911 calls (1 , 528 , 303); all recorded residential real estate transactions (61,902 after filtering), as well as the corresponding property characteristics; demographic data and the associated changes. The inherent endogeneity between crime rates and housing prices forces us to find an instrument for crime rates. After rejecting several plausible choices, based on the Wu-Hausman test, the key variable in our empirical analysis is the number of 911 calls (contemporaneous and lagged) in a given beat. Our results present somewhat mixed evidence on the impact of crime on housing prices. Without adjustment for spatial autocorrelation, a 1 percentage point increase in crime rates (instrumented by the number of 911 calls) leads to approximately a 0.55% decrease in house prices. However adjusting for spatial autocorrelation changes this figure to a 0.80% increase . We further show that distance to a crime hotspot is significantly negatively related to housing prices, suggesting that criminals choose to operate in wealthier areas, which is consistent with our findings incorporating spatial autocorrelation.