Structural Covariates of Crime Rates: Studies of Invariance

长期以来，社会科学家一直试图解释犯罪行为和个人受害行为在不同个人之间的差异，同样重要的是，解释不同时间和空间的差异。尽管在过去几十年里发表了许多关于不同国家以下单位犯罪率的结构协变量的研究，但关于在某一分析水平和某一特定时间点的一项研究中发现的经验关系在多大程度上推广到其他时间点的其他分析水平，我们知之甚少。到目前为止，大多数关注犯罪率变化的研究都利用了与特定样本相关的相当独特的关联模式。在这项研究中，兰德博士和科恩博士研究了社会结构(经济、人口、社会)决定犯罪率的因素，包括总体水平和时间。他们将使用作为之前NSF项目的一部分创建的数据集，该项目包含1960年、1970年和1980年的七项FBI指数犯罪和许多结构协变量的信息，以及三个聚合级别(城市、SMSA、州)的信息。将使用几种统计技术来检验社会指标和犯罪率之间的经验规律性。在对每个协变量的状态进行评估的基础上，将进一步考虑可能导致不同水平差异的聚合效应。这项研究应该会对美国这一时期的犯罪率进行最全面的研究。此外，将解决的问题是犯罪学研究的理论建设和方法论的基础。作为对理论的贡献，这项研究将对几种不同因果取向的支持者提供的关系的一致性进行测试。作为对方法论的贡献，它将展示以回归为基础的犯罪率解释的内在局限性，以及使用高度相关的衡量标准来理解犯罪率差异的问题。此外，这项研究应该通过检查面板和横截面模型之间的关系来增加研究实践。

Social scientists have long sought to explain both the variation of criminality and personal victimization across individuals and, as importantly, across time and space. Despite the fact that numerous studies of structural covariates of crime rates for various subnational units have been published over the past several decades, little is known about the extent to which empirical relationships found in one study with one level of analysis and at one particular point in time generalize to other levels of analysis at other time points. To date, most studies focusing on crime rate variation have capitalized on rather unique patterns of association related to particular samples. In this research, Drs. Land and Cohen examine the social structural (economic, demographic, social) determinants of crime rates across levels of aggregation and time. They will use a data set created as part of a previous NSF project which contains information on seven FBI Index crimes and numerous structural covariates for 1960, 1970, and 1980 and at three levels of aggregation (cities, SMSA, states). Several statistical techniques will be used to test the empirical regularities between social indicators and crime rates. On the basis of an assessment of the status of each covariate, additional thought and work will go into consideration of possible aggregation effects that might account for discrepancies across the levels. The research should result in the most comprehensive study of crime rates in the U.S. over this period. In addition, the questions that will be addressed are fundamental to theory construction and methodology in criminological research. As a contribution to theory, the study will provide tests of the constancy in relationships offered by proponents of several different causal orientations. As a contribution to methodology, it will demonstrate the inherent limits to regression-based explanations of crime rates and the problem of using highly correlated measures to understand differences in crime rates. Also, the study should add to research practice by inspecting relationships between panel and cross-sectional models.