Crime and divorce. Can one lead to the other? Using Multilevel Mixed Models

Cross-sectional and time-series studies of the influence of divorce on crime and the reverse are few in developing and developed countries. Questions arise as to whether divorce causes crime, the reverse, or both effects exist in Jordan. The objectives are to investigate the relationship between divorce and crime, determining whether the clustering in divorce and in crime within governorates exist and whether divorce and crime increase or decrease over time. The study design was a cross-sectional time-series analysis. Several Jordanian statistical yearbooks and surveys issued by the Jordanian Statistics Department provided the data of 12 governorates over 14 years (2000–2013). After calculating the divorce rate (DR) and crime rate (CR), multilevel mixed-effects linear regression was performed, estimating three models each for divorce and crime. Comparison between these models was explained in intraclass correlation, the proportional change in the variance of the response variable, and the deviation. The statistical and social epidemiological concepts of contextual phenomena confirm that the rates of divorce and crime in the same governorate are more similar to each other than to those from different governorates. Using the CR as a predictor for the DR reduced the within-governorate variance more than four times the between-governorates variance. Using the DR as a predictor for the CR reduced the within-governorate variance and inflated the between-governorates variance. Using time as a predictor for the DR reduced the within-governorate variance dramatically higher than the between-governorates variance and as a predictor for the CR reduced the within-governorate variance but inflated the between-governorates variance a small amount. Thus, both divorce and crime lead to the other.

在发展中国家与发达国家中，探讨离婚对犯罪的影响及二者反向因果关系的横断面研究与时间序列研究均较为匮乏。在约旦，学界尚存疑问：究竟是离婚引发犯罪，还是犯罪导致离婚，抑或是二者互为因果？本研究旨在探讨离婚与犯罪之间的关联，验证约旦各省份内是否存在离婚率与犯罪率的集聚现象，并分析二者随时间推移的变化趋势。本研究采用横断面时间序列分析的研究设计。本研究的数据来源于约旦统计局（Jordanian Statistics Department）发布的多份约旦统计年鉴与调查数据，涵盖2000年至2013年共14年间约旦12个省份的相关信息。在计算离婚率（divorce rate, DR）与犯罪率（crime rate, CR）后，本研究采用多层混合效应线性回归模型，分别针对离婚与犯罪两类因变量构建三种回归模型。本研究通过组内相关系数、响应变量方差的比例变化量以及偏差值对上述模型进行比较分析。情境现象的统计学与社会流行病学概念证实：同一省份内的离婚率与犯罪率相较于不同省份的对应指标，彼此间具有更高的相似性。以犯罪率（CR）作为离婚率（DR）的预测变量时，省份内方差的缩减幅度较省份间方差的缩减幅度高出四倍有余。以离婚率（DR）作为犯罪率（CR）的预测变量时，省份内方差出现缩减，而省份间方差则有所增大。以时间作为离婚率（DR）的预测变量时，省份内方差出现了远大于省份间方差的显著缩减；而以时间作为犯罪率（CR）的预测变量时，省份内方差虽出现缩减，但省份间方差仅出现小幅增大。综上可见，离婚与犯罪二者互为因果关系。