ABSTRACT
For youth between the 8th and the 12th grades, parents and adult caretakers play a critical role in shaping their behaviors. The parent-youth relationship has several important dimensions, and each plays a unique role in preventing youth delinquency. The current study seeks to explore the heterogeneity among youth in Arizona regarding parental neglect, parental monitoring, and exposure to family violence. We further investigated how these family characteristics correlated with self-reported delinquency. We analyzed survey data collected from a sample of youth from Arizona (n = 38,945) and used latent class analysis (LCA) to capture these family properties. We identified a five-group model suggesting that these dimensions did not necessarily overlap. We also found that group membership was correlated with self-reported arrests as well as multiple delinquent behaviors. Youth with low exposure to family violence and from families with effective monitoring had the lowest probability of engaging in delinquent behaviors. Insufficient parental monitoring and exposure to family violence had slightly different correlates for different types of delinquency.
Acknowledgments
We thank Jacob Forston and Cheyenne Weaver for their research assistance.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1. Participants could choose more than one racial and ethnic groups to identify with. Therefore, the total percentages of racial and ethnic groups exceeded 100%. For ease of presentation and interpretation, we included all non-White racial and ethnic groups as dummy variables in our regression models, in which case the actual reference group was White youth with non other racial and ethnic identity. We also conducted a sensitivity test by creating a multi-racial category (therefore making each youth belonging to one and only one category), and found no notable difference in the findings.
2. We are aware of the long debates on the acceptability of treating Likert scales as continuous scores. In addition to the analysis presented here, we also attempted to estimate LCA models treating the eight scale indicators as either nominal or ordinal variables. Both sets of models ran into convergence problems when the number of classes became large, presumably due to the level of complexity. A recent simulation-based study suggested that even with a Likert scale containing a small number of points (like a 4-point one used in the present study), treating the responses continuously would not gravely distort the empirical findings (H. Wu & Leung, 2017). Therefore, we present the sensitivity test to support the robustness of our findings.