ABSTRACT
A scant literature has identified gun carrying as a potential risk factor for victimization at the individual level. To date, however, research has generally focused on high-risk individuals rather than samples drawn from the general population. Additionally, prior studies have not often enough included controls robust enough to feel strongly that the relationship between gun carrying and victimization, gun victimization in particular, is not simply the spurious outcome of factors that influence both variables. The current study uses data from Add Health participants (N = 13,568) to look at the effect of gun carrying on gun victimization among adolescents. Results suggest that even when robust controls are considered, a measure of gun carrying significantly and positively correlates with gun victimization. The results support a model of the gun carrying-gun victimization relationship wherein gun carrying increases risks for gun victimization independent of factors that may influence both risky behaviors and victimization. Implications for theory and policy are discussed.
Disclosure statement
No potential conflict of interest was reported by the author.
Notes
1. This setup of looking at Wave I effects on Wave II gun victimization is preferable to looking at the independent and dependent variables within the same wave because it allows for a separation in the time periods covered, which is important for causal ordering and because the two variables cover different amounts of time (30 days vs. 12 months).
2. While fear of victimization has also consistently been tied to gun carrying among adolescents (see Melde et al., Citation2009; Wilcox et al., Citation2006), measures assessing fear of victimization are not available in Add Health.
3. Given the rarity of the dependent variable, rare events logit estimation was considered. However, this is not an option in Stata 13.1 when the “svyset” command is used to correct for sampling design to unbias coefficients, and thus standard logit estimation was utilized. If the models are run without the “svyset” command, which is not advised by Add Health, and the “firthlogit” command is used to run a type of rare events logit estimation, results that are substantively identical to those presented in this article are produced. These results are available upon request.
4. These variables, and guidance on how to appropriately use them, are provided by Add Health. Specifically, Add Health respondents have an individual weighting variable, a cluster variable based on their school, and a stratification variable based on their census region. These variables are identified in Stata 13.1 by using the “svyset” command.
5. When data are “svyset” in Stata 13.1, the M & Z pseudo r-sq. value is the only one of the numerous possible pseudo r-sq. values that is presented when using the “fitstat” command after running a logistic regression model.
6. To follow up on the results in model 2 of , interactions between the independent variable and each control variable were examined in a separate model. None of these interaction terms reached statistical significance. Results are available upon request.
7. Traditional tests of model fit improvement, such as the likelihood ratio test or Wald test, will not run in STATA when data is “svyset.” However, if the models are run without the “svyset” command, which is not recommended in the case of these data, the likelihood ratio test produces a significant chi-square statistic (chi-sq. = 327, prob. > chi-sq. = 0.0000), suggesting the addition of the further control variables in model 2 results in a statistically significant improvement in model fit.
8. Add Health contains no measures, at Waves I or II, about general gun carrying behavior.