Abstract
A wealth of scholarship generally finds that marriage protects against crime, but there is less consistent evidence for cohabitation. In this article, we contribute to scholarship on marriage and put forward new evidence about cohabitation by examining marital and cohabiting partnerships as transitions with distinct stages of entry, stability, and dissolution. We use within-person change models with contemporary data from the National Longitudinal Survey of Youth 1997 to analyze these stages for the full sample and separately for men and women. The findings show differential protective associations of marriage and cohabitation depending on the stage of the partnership. Both recently formed cohabiting partnerships and stable cohabiting partnerships are associated with reductions in the level of offending, although to a lesser degree than marital relationships. Cohabiting partnerships that are stable, in that they have lasted at least a year, are associated with larger decreases in offending, particularly among women.
Acknowledgements
The authors would like to thank Jonathan Brauer, and Justice Quarterly Editors and Anonymous Reviewers for thoughtful comments on earlier versions of this manuscript. We would also like to thank Germán Rodriguez for statistical advice and Dawn Koffman for programming advice.
Notes
1 In addition to partnership stage, the quality of the relationship is likely to be an important factor, where higher- quality partnerships are more protective compared to lower-quality partnerships (Skardhamar et al., Citation2015). Relationship quality and partnership stage, although related, are not necessarily analogous. We return to this point later in the paper.
2 A few studies have measured aspects of these stages separately. Bersani and Doherty (Citation2013) and Larson et al. (Citation2016) have focused on dissolution, while Lyngstad and Skardhamar (Citation2013), McGloin et al. (Citation2011), and Laub et al. (Citation1998) have explored the courtship period.
3 We do not include survey weights in the descriptive statistics and regression models for several reasons. First, we use longitudinal data from 1998 to 2011 and the use of weights with longitudinal data is not straightforward (NLSY Citationn.d.b). Second, as we describe and later test in the paper, NLSY97 changed the universe of people asked the criminal offending questions in 2004. Because of this issue, combined with the longitudinal nature of the sample, sampling weights are not appropriate for descriptive statistics. Third, we refrain from using weights in the multivariate regression models, following Winship and Radbill (Citation1994).
4 When the partner ID variable was missing and the respondent was partnered in consecutive years (about 7% of cases in which people were partnered in consecutive years), we assumed that the respondent was partnered with the same person.
5 Although prior research has examined relationship duration (e.g. Blokland & De Schipper, Citation2016), we do not directly address duration for several reasons. First, the inclusion of a count variable of duration means that the protective associations of relationships are highest just prior to dissolution; whereas, we suggest that the time period around dissolution is the least protective for offending. Second, we conducted supplemental analyses, and we find no association between duration and offending among stable partnerships. In these models (the equivalent of Table ), we distinguish among partnership stages of entry and dissolution, and we include controls for duration of stable married and stable cohabiting partnerships.
6 We refrain from examining relationship quality because of two limitations with the NLSY97 data. First, quality questions are asked for years 2000 to 2008 only (or, a subset of the analytic sample 1998–2011). Second, and more importantly, quality questions are asked for current partnerships only; therefore, information on quality for recently dissolved partnerships (e.g. cohabiting to single or married to divorced) would have to come from the previous wave. This is problematic, since the same quality information refers to stable or recently entered into partnerships (for the year prior to dissolution) and recently dissolved partnerships, for partnerships that eventually experience dissolution. Despite these limitations, we conducted two sets of supplemental analyses to examine whether distinctions by partnership stage are driven by relationship quality. These analyses are based on a quality measure that is the average (from 0 to 10) of responses to three questions that capture respondent perceptions about the degree to which their partners care for them, how close they feel to their partner, and the amount of conflict (reverse coded) in the relationship. If partnership stage is conflated with quality, we would expect to find that: (1) levels of quality align with stage, where stable partnerships have the highest mean quality; (2) high quality partnerships are similarly associated with offending (regardless of the partnership stage); and (3) low quality partnerships are similarly associated with offending (regardless of partnership stage). Descriptive analyses demonstrated that, within marriages and cohabiting partnerships, quality is actually highest at the entry stage rather than stability stage, suggesting that partnership stages are not explained fully by relationship quality. In regression models (the equivalent of Table ), we did not find evidence of points 2 and 3; instead, although high and low quality distinctions sometimes mattered within stage, these differences did not override distinctions across partnership stages of entry, stability, and dissolution.
7 To further address this clustering, we also include robust standard errors.
8 In the logistic regression model with fixed effects, we are unable to account for the fact that some of the respondents are siblings, since the model is not multilevel and since we cannot cluster the standard errors. As a result, we checked the robustness of these results by employing a multilevel logistic regression model, adopting the same approach we did for the negative binomial model. The results from this approach were substantially similar to the findings estimated by the fixed effects model.
9 When the coefficient is greater than 0, the incident rate ratio is calculated as: exp(coefficient) – 1. When the coefficient is less than 0, the incident rate ratio is calculated as: 1 – exp(coefficient).
10 This was determined by making cohabitation the reference group.
11 This was determined by changing the reference group in the analyses to stable cohabitation.
12 We also conducted additional analyses to examine whether associations by partnership stage differed depending on the presence or absence of children (as opposed to simply including a control for children in the models). Prior work indicates that the combined presence of marriage and children (full family package) has a greater protective association than marriage alone, particularly among men (Zoutewelle-Terovan, van der Geest, Liefbroer, & Bijleveld, Citation2014). In our models we found little consistent evidence of a family package effect across stages. In the logistic regression specification for women, we find suggestive evidence that divorce is more negatively associated with offending when children are present (b = –.660, p-value < .05 with children compared to b = –.507, p-value = n.s. without children), although these coefficients are not different from one another and we found no evidence of similar potential differences in the negative binomial specification.
Additional information
Notes on contributors
Aaron Gottlieb
Aaron Gottlieb is an Assistant Professor at the Jane Addams College of Social Work, University of Illinois at Chicago. His research explores the causes and consequences of mass incarceration and punitive criminal justice practices in the United States.
Naomi F. Sugie
Naomi F. Sugie is an Assistant Professor in the Department of Criminology, Law and Society at the University of California, Irvine. Her research examines the consequences of incarceration and other forms of criminal justice contact for individuals and their partners. Her recent work focuses on reentry, employment, and criminal record stigma, and she is particularly interested in the use of new technologies to study hard-to-reach groups.