More Places than Crimes: Implications for Evaluating the Law of Crime Concentration at Place
More Places than Crimes: Implications for Evaluating the Law of Crime Concentration at Place
复制标题
DOI：
10.1007/s10940-016-9324-7
复制
发表时间：
2017-09-01
影响因子：
3.6
通讯作者：
Steenbeek, Wouter
中科院分区：
文献类型：
--
作者：
Bernasco, Wim;Steenbeek, Wouter
犯罪和场所文献缺乏衡量和报告犯罪集中度的标准方法。我们建议用洛伦兹曲线报告犯罪集中度，并用基尼系数进行总结，并提出洛伦兹曲线和基尼系数的广义版本，以纠正犯罪数据稀疏（即犯罪数量少于地方）时的偏差。所提出的概括基于这样的原则：观察到的犯罪集中度不应与完全平等进行比较，而应与给定数据的最大平等进行比较。当犯罪数量接近空间单位数量时，概括逐渐逼近原始洛伦兹曲线和原始基尼系数。利用海牙市两类犯罪的地理编码犯罪数据，我们展示了原始洛伦兹曲线和基尼系数与广义版本之间的差异。我们证明，在犯罪数据稀疏的情况下，概括可以更好地表示犯罪集中度，并且如果犯罪数据稀疏，它们可以改善犯罪集中度的比较。建议研究人员在报告和总结地方犯罪集中度时使用洛伦兹曲线和基尼系数的广义版本。当地点数量超过犯罪数量时，广义版本比原始版本更好地代表了犯罪集中的潜在过程。广义洛伦兹曲线、基尼系数及其方差很容易计算。
The crime and place literature lacks a standard methodology for measuring and reporting crime concentration. We suggest that crime concentration be reported with the Lorenz curve and summarized with the Gini coefficient, and we propose generalized versions of the Lorenz curve and the Gini coefficient to correct for bias when crime data are sparse (i.e., fewer crimes than places).The proposed generalizations are based on the principle that the observed crime concentration should not be compared with perfect equality, but with maximal equality given the data. The generalizations asymptotically approach the original Lorenz curve and the original Gini coefficient as the number of crimes approaches the number of spatial units.Using geocoded crime data on two types of crime in the city of The Hague, we show the differences between the original Lorenz curve and Gini coefficient and the generalized versions. We demonstrate that the generalizations provide a better representation of crime concentration in situations of sparse crime data, and that they improve comparisons of crime concentration if they are sparse.Researchers are advised to use the generalized versions of the Lorenz curve and the Gini coefficient when reporting and summarizing crime concentration at places. When places outnumber crimes, the generalized versions better represent the underlying processes of crime concentration than the original versions. The generalized Lorenz curve, the Gini coefficient and its variance are easy to compute.