Exploring the Genetic Origins of Adolescent Victimization in a Longitudinal Sample of Adoptees

Abstract

Behavioral genetic research has consistently revealed that genetic factors explain a significant proportion of variance in antisocial phenotypes. Despite some initial evidence, the extent to which genetic factors influence adolescent victimization remains largely undetermined. The current study partially addresses this gap in the literature by employing an adoption-based research design to estimate genetic influences on victimization. Participants consisted of adoptees drawn from the National Longitudinal Study of Adolescent Health (Add Health). Multivariate analyses revealed that adoptees who had a biological criminal father or a biological criminal mother were at increased risk for being victimized in adolescence and also to be victimized repeatedly. This is the first study to use an adoption-based research design to show an association between genetic factors and the risk for adolescent victimization.

Keywords:

• Add Health
• adoption
• biosocial
• genetics
• victimization

Acknowledgments

This research uses data from Add Health, a program project designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and funded by a grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 17 other agencies. Special acknowledgment is due to Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining data files from Add Health should contact Add Health, Carolina Population Center, 123 West Franklin Street, Chapel Hill, NC 27516 ([email protected]). No direct support was received from grant P01--HD31921 for this analysis.

Notes

1. We should point out that we refer to victimization as an antisocial phenotype to capture the fact that victimization is typically viewed as a negative outcome. Likewise, we also chose to characterize victimization as an antisocial phenotype in the current study because victimization is partially the result of antisocial traits and behaviors (Schreck, 1999; Wolfgang, 1957). As a result, the logic linking genes to victimization hinges on the antisocial origins to victimization. This is not to say that random events or even prosocial processes play no role in the etiology of victimization; rather, we are emphasizing the very real possibility that genes might relate to victimization via the effects that genes have on antisocial phenotypes.

2. We realize that the item dealing with whether the adoptee had been in a physical fight could be tapping both victimization and offending. As a result, we reestimated the models after removing this item from the wave 1 victimization scales. The substantive results remained the same so we opted to retain this item in the analyses. In addition, we note that the wave 2 victimization scale did not include this item, but the results were again substantively the same. As a result, we do not see this item as biasing the findings in any particular direction.

3. A reviewer pointed out that the statistical models may be more correctly specified by including a measure which controls for offending/delinquency. We opted not to include such a measure for two reasons. First, a measure of offending/delinquency would not render any genetic effect spurious since offending/delinquency would not be causing genetic risk (i.e., genetic risk would be antecedent to offending/delinquency). Second, and relatedly, although offending/delinquency might mediate the association between genetic risk and victimization, the Add Health data do not allow us to disentangle the temporal ordering between victimization and offending. Although these data are longitudinal, victimization and offending are highly interrelated and a simple longitudinal model would not be able to capture the dynamic processes that unfold to produce the victim-offender overlap. Victimization, for instance, could precede offending, offending could precede victimization, or there could be feedback effects. Given the limitations with the data and the inability to distinguish among these different interpretations, we opted to calculate all of the models without controlling for offending/delinquency. However, we agree that this is an important avenue of research for future studies to explore.

4. We realize that setting the covariates at their means when plotting the predicted probabilities produces predicted probabilities that do not apply to any specific person. The reason for this is because we are employing means for dichotomous variables, such as gender and race, wherein people are either male/female or white/nonwhite. The alternative to this approach would be to arbitrarily select a profile (e.g., a white female) and calculate the predicted probabilities for that profile. We opted to calculate the predicted probabilities when the covariates were set to their means for two main reasons. First, if we selected to focus on one profile, then we would arbitrarily be selecting one profile and leaving out all others. This could give the false impression that the findings would be substantially different if we selected another profile. Second, we are really interested in the absolute difference in slopes between the two groups (i.e., those who possess genetic risk and those who do not). Using different profiles would not affect the difference between slopes and thus would not bias the interpretation of the predicted probabilities that are graphed in the figures. In other words, utilizing the mean scores when calculating predicted probabilities does not change the substantive results and can be thought of as the overall average predicted probability for all respondents in the sample.