Unstructured Socializing with Peers and Delinquent Behavior: A Genetically Informed Analysis

Abstract

A large body of research finds that unstructured socializing with peers is positively associated with delinquency during adolescence. Yet, existing research has not ruled out the potential for confounding due to genetic factors and factors that can be traced to environments shared between siblings. To fill this void, the current study examines whether the association between unstructured socializing with peers and delinquent behavior remains when accounting for genetic factors, shared environmental influences, and a variety of non-shared environmental covariates. We do so by using data from the twin subsample of the National Longitudinal Study of Adolescent to Adult Health (n = 1200 at wave 1 and 1103 at wave 2; 51% male; mean age at wave 1 = 15.63). Results from both cross-sectional and lagged models indicate the association between unstructured socializing with peers and delinquent behavior remains when controlling for both genetic and environmental influences. Supplementary analyses examining the association under different specifications offer additional, albeit qualified, evidence supportive of this finding. The study concludes with a discussion highlighting the importance of limiting free time with friends in the absence of authority figures as a strategy for reducing delinquency during adolescence.

Appendices

Appendix 1. Information on DF Model Estimation

The DF model takes the following form:

$${X_{1j}} = {\beta _0} + {\beta _1}{\ddot X_{2j}} + {\beta _2}{\ddot X_{2j}}*{R_j} + {\varepsilon _{1j}}$$

where: X _1j is the score on unstructured socializing for twin 1 from twin pair j; β ₀ is the intercept and is typically not interpreted substantively; ${\beta _1}{\ddot X_{2j}}$ provides an estimate of the shared environment by estimating the impact of twin 2’s mean-centered score of X on his/her co-twin’s score; ${\beta _2}{\ddot X_{2j}}*{R_j}$ provides an estimate of heritability by calculating the degree to which the mean-centered version of X ₂ is a better predictor of X ₁ for siblings that share more genetic overlap (i.e., when R = 1 for MZ twins relative to R = 0.50 for DZ twins); and ε _1j is the error term that captures all sources of variance in X _1j that are not attributable to heritability or shared environments. Note that we estimated the DF model using double-entry, which required us to adjust the standard errors to account for the clustering of twins in pairs.

Appendix 2. Information on Fixed Effects Model Estimation

The fixed effects model takes the following form:

$${Y_{ij}} = {\varphi _0} + {\varphi _1}{X_{ij}} + {\varphi _2}{D_{1j}} + \mathop {\sum}\limits_{j = 1}^J {{\varphi _j}}$$

where: Y _ij is the outcome of focus; φ ₀ is an intercept term revealing the average level of Y for respondents who score a zero on all the right-hand side variables; φ ₁ X _ij captures the impact of unstructured socializing on Y; φ ₂ D _1j is a fixed effect that captures any random differences that may exist between twins labeled twin 1 and those labeled twin 2; and $\mathop {\sum}\nolimits_{j = 1}^J {{\varphi _{j\,}}}$ captures the collective influence of all the twin-level fixed effects for all twin pairs J.