Continuity and Change in Adolescent Victimization: Examining the Effects of Stable and Dynamic Influences

ABSTRACT

After decades of treatment as a fairly distinct topic, recent research on victimization has begun to draw on theoretical approaches previously directed at understanding criminal behavior. The current study expands this research by studying victimization and its relationship to key developmental influences with data from 3,976 adolescents. We first detail the longitudinal process that underlies continuity and change in victimization and then consider the impact of time-stable and time-varying covariates that reflect mechanisms within those explanations. Findings suggest that time-varying markers of risky lifestyle and attachment affect victimization, but also that victimization affects risky behaviors and prosocial ties.

KEYWORDS:

• adolescent victimization
• developmental
• school violence
• theories

Acknowledgments

The authors thank Richard R. Clayton, Pamela Wilcox, Scott A. Hunt, Michelle Campbell Augustine, Shayne Jones, Kimberly Reeder, Staci Roberts Smith, and Jon Paul Bryan for their contributions to the Rural Substance Abuse and Violence Project, which provides the data analyzed here. The conclusions and opinions of this study are the sole responsibility of the authors.

Funding

Data analyzed for this article were originally collected through funding from the National Institute on Drug Abuse (DA 11317, Richard R. Clayton, PI).

Notes

1. Protective factors can be defined as factors that reduce the likelihood of victimization as indicated by an inverse relationship (Goldbaum, Craig, Pepler, & Connolly, 2003; Jackson, Chou, & Browne, 2015) or events, opportunities, or experiences that buffer or diminish the likelihood of engaging in risky behaviors (Resnick, Harris, & Blum, 1993; Resnick, Ireland, & Borowsky, 2004). We focus on the former operationalization here.

2. Unlike data that are missing completely at random (Acock, 2005), the missing values in the current study may in fact be related to other variables in the model (i.e., “missing at random” or “not missing at random”). Based on the sample attrition, it seems that there are some instances where variable nonresponse is dependent on the distribution of that variable. This is most likely due to a general attrition problem that is relatively consistent across variables, but which could be associated with some items in the models (e.g., victimization). While not ideal when missingness at random cannot be assumed, a maximum likelihood approach to preserve all possible data values is still preferred to listwise deletion in such situations (Graham, 2009).

3. Given the study objectives, this approach has a few advantages over pooled time series regression (Hsiao, 2003). For example, this model allows for the estimation of both between- and within-individual covariate effects, which would not be possible in a fixed-effects regression analysis (Allison, 2009). Also, the latent growth curve portion of the model estimates the conditional trend over time and the variation around that trend and the autoregressive portion allows for testing freed direct effect coefficients at each stage. Last, and most importantly, use of this approach presents an opportunity to study linked, indirect structural relationships that would not be available in a pooled fixed or random effects model.

4. Full table results are not presented, but are available upon request.

5. Due to four potential outliers in the Wave 1 age variable (ages 17 to 19), an additional analysis was estimated without these cases to determine their influence on the estimates. The pattern of results was similar to that presented here. Therefore, the cases were included in the final analysis.

Additional information

Funding

Data analyzed for this article were originally collected through funding from the National Institute on Drug Abuse (DA 11317, Richard R. Clayton, PI).