The Relationship Between Bullying Victimization and School Avoidance: An Examination of Direct Associations, Protective Influences, and Aggravating Factors

ABSTRACT

This study examines the impact of bullying victimization on school avoidance by proposing the following hypotheses: (1) Net of other factors, students who have experienced bullying victimization are more likely to engage in school avoidance behaviors; (2) There are protective factors that will decrease this relationship between bullying victimization and avoidance behaviors; and (3) There are aggravating factors that will increase this relationship between bullying victimization and avoidance behaviors. Binary logistic regression models are estimated, using data from the 2011 National Crime Victimization Survey: School Crime Supplement, with a nationally representative sample of 6,547 middle and high school students. Findings provide strong support for the effect of bullying victimization on school avoidance, such that students who have experienced bullying are more likely to avoid places in school. Additionally, results suggest that participation in school activities and fear can influence the relationship between bullying victimization and school avoidance.

KEYWORDS:

• Bullying
• peer victimization
• school advoidance
• fear of crime

Acknowledgments

The authors would like to thank Drs. David B. Wilson, David Weisburd, and Devon Johnson for their help on previous work contained herein.

Notes

1. The NCVS provides three classes of weights, which include household weight, person weight, and incident weight. These weights are designed to produce national estimates of counts of households, people, and crime incidents from the NCVS and SCS sample cases (U.S. Department of Justice, Bureau of Justice Statistics, 2013). If a study aims to describe current rates, incidents, or population characteristics that are nationally representative, weights are appropriate to use, and must be selected correctly from among the weight variables available in the public use file (U.S. Department of Justice, Bureau of Justice Statistics, 2013). Because the current research is concerned primarily with the relationship between bullying victimization and avoidance behaviors, and not with the production of national counts or rates, the use of weights is not necessarily appropriate. For modeling purposes, sampling weights are used so that characteristics in the sample are more like characteristics in the population. This would have an impact if the sample data departed greatly from the national estimate representations of the relationships in question. In past analyses of NCVS data, researchers have conducted analyses primarily without sample weights. For instance, work appearing in distinguished academic journals has conducted analyses without weights, and noted that coefficients were unlikely to change when unit weights were turned on (for example, see Bachman, 1998; Lauritsen, Gatewood, Planty, Rand, & Truman, 2012).

2. Minority students included 804 Black students (12.3%), 309 Asian/Native American/Pacific Islander students (4.7%), and 203 categorized as “other” (3.1%).

3. There are many different ways to create a scale of these items. For instance, it is possible to collapse responses into one dichotomous variable that represents whether students were bullied or not bullied (“yes” or “no”) or to collapse responses into one variable that represents direct bullying victimization, indirect bullying victimization, and no bullying victimization (e.g., pushed you, shoved you, tripped you, or spit on you vs. made fun of you, called you names, or insulted you, in a hurtful way vs. did not experience bullying victimization). It is also possible to use a principal component analysis for creating the bullying victimization scale; however, after further investigation it appears that the current data does not lend itself to factor analysis even though this method has the ability to capture latent characteristics that are central to bullying victimization experiences. This is evident when examining the nature of the measures, as the bullying victimization items might not be highly correlated with one another even if an individual has experienced a high level of bullying. Factor analysis creates factors based on a high degree of correlation among items and assumes that when a correlation exists that there is a common underlying latent construct. The items on this scale are not necessarily mutually exclusive and responding “yes” to one does not necessarily increase the likelihood of responding “yes” to another. As such, one variable representing whether students were bullied or not bullied (“yes” or “no”) was created for the current research (Table 1). This scale is of interest as it is a straightforward measure, capturing whether students self-reported bullying victimization or not.

4. Again, it is possible to use a principal component analysis to create the avoidance scale; however, after further investigation it appears that the current data does not lend itself to factor analysis even though this method has the ability to capture latent characteristics that are central to avoidance behaviors. This is evident when examining the nature of the measures, as the avoidance items might not be highly correlated with one another even if an individual has engaged in a high level of avoidance behaviors. Factor analysis creates factors based on a high degree of correlation among items and assumes that when a correlation exists that there is a common underlying latent construct. The items on this scale are not necessarily mutually exclusive and responding “yes” to one does not necessarily increase the likelihood of responding “yes” to another. As such, these items are collapsed into one dichotomous variable that represents whether students did or did not avoid places in school or school altogether (“yes” or “no”, Table 2). This scale is of interest due to the lack of variation in the data and the dichotomous responses for each avoidance item.