Sex Differences in the Health Risk Behavior Outcomes of Childhood Bullying Victimization

ABSTRACTShort- and long-term health consequences of bullying victimization are well documented and include physical and mental health issues as well as increased involvement in risky behavior, but research exploring sex differences in victimization outcomes is still limited. This study uses data from the National Longitudinal Survey of Youth—1997 to examine the consequences of victimization by sex and, more specifically, relationships between bullying victimization and later health risk behaviors—including risky sexual activity, smoking, alcohol use, and drug use. Multivariate analyses identified sex differences for specific health risk indicators, and a substantial difference was evident for overall risk.KEYWORDS: biological sexbullyinghealth risk behaviorsvictimization Notes1. With the exception of age and household income, demographic and control variables as well as the bullying victimization variable were taken from Wave 1. Descriptive statistics presented here represent the full sample as available at Wave 1 (n = 8,984). Descriptive statistics for the health risk behavior outcome variables represent the follow-up sample at Wave 6 (n = 7,898).2. The initial sampling strategy for the NLSY97 involved a multistage, stratified random sample as well as oversampling of specific groups. Sampling weights were created at each wave to account for this. While the NLSY specifies that using sampling weights in regression analysis is not advised (see "Sample Weights & Design Effects" at http://www.nlsinfo.org/nlsy97/nlsdocs/nlsy97/use97data/weights.html), all multivariate models presented here were reanalyzed with the Wave 6 sampling weights and robust standard errors in Stata 12. Results were substantively similar to those presented here and are available on request.3. With the exception of household income, missing data was minimal, and list-wise deletion was used. Including the household income measure results in the loss of nearly 1,500 cases (17% of the sample) in the multivariate models. There are significant differences on a number of measures between those missing cases and those retained, with the important exception of the bullying victimization measure. The models were estimated again leaving out the household income measure, and results were substantively similar, so we have elected to retain the household income measure since income does play a significant role in predicting health risk behaviors.4. The coefficient comparison test as described by Paternoster et al. (Citation1998) is not appropriate with logistic regression, so this comparison is only conducted with the summative risk index.