Violent Victimization in the Schools
Ilhan Turkmen
The School of Information Technology
University of Cincinnati
Email: turkmein@mail.uc.edu
Introduction
This paper examines the number violent victimization of students (dependent variable) at schools based on substance use, respondent age, school safety, extracurricular activities, and how many days skipped class, how often gangs in fights/violence at school and number of times in a fight (independent variables). These independent variables are chosen consistently with the victimization literature and previous victimization studies.
Our research question is “how can we decrease the likelihood of students being a victim at schools”. This research question is of vital importance because victims in our study are adolescents and they are affected worse by violent crimes. Our hypotheses are as follows;
H_{0}: There is no significant relationship between any of the above independent variables and the number of violent victimization.
H_{1}: There is a positive significant relationship between substance use and the number of violent victimization.
H_{2}: There is a negative significant relationship between school safety and the number of violent victimization.
H_{3}: There is a positive relationship between number of times in a fight and the number of violent victimization.
H_{4}: There is a positive significant relationship between the number of days skipped class and the number of violent victimization.
H_{5}: There is a positive significant relationship between gang fights/violence at school and the number of violent victimization.
H_{6}: There is a significant negative relationship between respondent age and violent victimization.
Literature Review
There are a couple of victimization theories in the literature. One of the victimization theories is lifestyle exposure theory which explains the likelihood of being a victim of a crime by lifestyle differences. A person’s lifestyle has impact on getting encountered to the high risk places, people, time etc. (Meier and Miethe, 1993).
Sullivan et al (2011) state that Ousey et al.’s (2008) prior examination of Rural Substance Abuse and Violent Project (RSVP) found out that delinquency, being member of a gang and having access to illicit drugs are positively related with victimization (p. 88). They also argue that school based victimization is related to delinquent behavior, illicit opportunities etc. They analyzed victimization by summing 6 separate illicit opportunities together, such as alcohol, drugs, and cigarettes etc.
Wilcox et al (2009) mention about Garofalo et al’s (1987) findings concerning guardianship and social control of student. They found that the students are more likely to be victims during the way to or from the school where school staff has little or no supervision. Supervision decreases the likelihood of being a victim of a violent crime. Again, Wilcox et al (2009) argue that “Even though structured before and afterschool activities are supervised by adults, the relative amount of supervision available can be much less than during normal school hours” (p.248). They also state that “The routine activity research has generally supported the notion that adolescent involvement in structured afterschool activities (as opposed to unstructured leisure activities) serves to protect against delinquent offending” (p.247).
Discussion and Results
The data that analyzed in this paper belong to ICPSR National Crime Victimization Survey: School Crime Supplement, 2007 (23041) study. The data were collected by United States Department of Commerce Bureau of the Census and the study was funded by United States Department of Justice Bureau of Justice Statistics. Cross sectional study and surveys were administered to the households between 12 and 18 years old in the US. These population also correspond 6^{th} graders and 12^{th} graders in school. Again, telephone and computer assisted interviews were conducted.
For our analysis purposes, some variables were added together and summed. We created new variables. Those variables were initially coded as 1=yes, 2= No. They were recoded as 1 and 0 for our analysis here. Variables that are added (and summed) for analysis are as follows;
For School Safety: VS035+VS036+VS037+VS038+VS039+VS040+VS041+VS042+ VS043 in which VS035 is security guards, VS036 is staff/adults in hallway, VS037 is metal detectors, VS038 is locked doors, VS039 is visitors sign in, VS040 is locker checks, VS041 is badges, VS042 is security cameras, and VS043 is code of conduct.
For variety of Substance Use: VS056 + VS057 + VS058 + VS059 + VS060 + VS061+ VS062 + VS063 + VS064 + VS064A + VS065 in which VS056 is possibility to get alcohol, VS057 is for marijuana, VS058 is for crack, VS059 is for cocaine, VS060 is for uppers, VS061 is for downers, VS062 is for LSD, VS063 is for PCP, VS064 is for heroine, VS065 is for other illegal drugs.
For Times Victimized: VS071+VS072+VS073+VS074+VS075+VS076+VS077 in which VS071 is for made fun of/called names, VS072 is for spread rumors, VS073 is for threatened you, VS074 is for pushed/shoved tripped etc., VS075 is for do things not wanted, VS076 is for excluded you, VS077 is for destroyed your property.
Table 1: Descriptive Statistics 
The table above illustrates the summary statistics of the data that are used in this paper. Respondents’ ages range from 12 to 18. As can be seen from this table, number of fights’ mean is 1.93 whereas the maximum number is 36. This is an indication of outliers in this data set. We could see it better via box plot or histogram table. It causes skewness to the right in histogram table. Again, number of days skipped class, substance use and times victimized have outliers.
We can also observe a missing data problem in this table. Number of times in fight and how many days skip class variables have low N number compared to others. That’s why, we have to address this as a limitation for our analysis. These two variables are widely used in violent victimization studies, thus these are important variables for us as well.
Figure1 illustrates the distribution of times of victimization. As can be seen in the table, the data are skewed to the right because most of the respondents were not victims of crimes. The data were not normally distributed.
Figure 1: Histogram Table for Times of Victimization

So as to create normality in the distribution, we made a log transformation. But again, the distribution is not normally distributed as can be seen in Figure 2. Furthermore, Table 2 illustrates that skewness value is 4 four times bigger than standard deviation which is an indicator of skewness. Therefore, we did not use the transformed log for our analysis. Normality could be better reached by setting victimization as a binary variable. In this case, all victimization responses would be added together. Thus 0’s and 1’s would be compared. But in this study, we are looking for further effects of independent variables. That’s why, we use this data as is. There are some other ways to (poisson) provide normality, but it is beyond this paper.
Figure 2: victimization log distribution  Table 2: Transformed log descriptive statistics 
Table 3 illustrates the correlation between variables. In the table, if the Pearson correlation is a positive value, then there is a positive relationship. It shows the strength and direction of linear relationship. If Sig. (2tailed) is lower than .05, then the relationship is significant (whether positive or negative).
The table indicates that school safety and number of fights have negative but insignificant relationship (.683). It implies that, the safer the school, the less fights will occur in that school. But again, the relationship is insignificant. On the other hand, substance use and number of fights have positive but again insignificant relationship (.277). Number of times of victimization and number of fights have positive and insignificant relationship as well (.063).
School safety and number of times of victimization have negative and significant relationship (with .005). The more schools have safe environment for students, the less the risk of being a victim. Again, the number of times of violent victimization and substance use have positive and significant relationship (with pvalue of .000). When the students have more access to illicit drugs and alcohol etc., they are more likely to be a victim.
Table 3: Correlations 
Surprisingly, substance use and school safety have positive significant relationship in this dataset (.000). More surprisingly, there is a negative but insignificant relationship between extracurricular activities and times of victimization (.788). Again unexpectedly, gang fights and times victimized have negative insignificant relationship (.719). On the other hand, there is a positive but insignificant relationship between number of days skipped class and times of victimization (.320).
So as to better analyze relationship and establish causality, we need to go further from our bivariate analysis. Thus, we conducted a regression analysis.
In this table, RSquare is the proportion of variance in the dependent variable. In another words, it illustrates how many per cent of the variance in dependent variable is explained by the independent variables. In this model, the dependent variable is “times victimized”, and our independent variables are substance use, respondent age, school safety, extracurricular activities (such as athletics etc.), how many days skipped class, how often gangs in fights/violence at school and number of times in a fight. R^{2} is .563 which means that 56 per cent of the variance in times victimized is explained by our regression model. The adjusted R^{2 }is also .181 in the model and it is fair enough for social sciences.
Table 4: Model Summary 
In this regression model none of the relationships are significant because Sig. values are above .05. In table 4, we observe that VIF value for extracurricular activities and the number of fights are bigger than 4.00 (which we accept as a threshold in our study). This indicates multicollinearity problem in our regression model. That also affects our models’ significance of relationships. So as to circumvent this problem, we are going to omit extracurricular activities and run the test again.
Table 5: Coefficients Table  

Table 6 illustrates that R^{2 }for the new model did not change considerably but the adjusted R^{2} increased from .181 to .261 which is the more strict measure for R^{2}.
Table 7 illustrates the new model without extracurricular activities. Now, the multicollinearity problem seems to be solved because the VIF values are around 1 and 2. But the relationships in the model are not still significant. All of the sig values (p values) are above .05 except for the number of fights. This independent variable has .027 pvalue which indicates that there is a significant relationship between number of fights and times victimized. And, as H_{3 }argues, the relationship is positive which implies that the number victimization will go up when the number of fights goes up as well. Thus, we reject the null hypothesis.
In this table, coefficient values illustrate the strength of the relationship. For example, as far as the relationship between the number of days skipped and the number of times victimized are concerned, we can interpret the table as “holding constant other variables, when there is a 1 unit increase in the number of days skipped, the number of violent victimization goes down by 66% because there is a negative relationship. Here, we have to note that, in our model, the relationship is not significant.
The number of victimization (times victimized – dependent variable) has negative and insignificant relationship with school safety. Although it is consistent with H_{2 }concerning the direction of the relationship, it is insignificant. That’s why, we fail to reject null hypothesis (for H_{2}).
The number of victimization and how many days skipped class have negative and insignificant relationship. Thus, we fail to reject null hypothesis for H_{4}.
The number of victimization and how often gangs in fights /violence at school have negative insignificant relationship. Again, we fail to reject null hypothesis for H_{5}.
There is a positive and insignificant relationship between times victimized and respondent age. We fail to reject the null hypothesis for H_{6}.
There is a positive relationship between the number of violent victimization and substance use. Thus, we fail to reject null hypothesis again for H_{1}.
Most of our findings are inconsistent with the literature. Wilcox et al. (2009) and Tillyer et al. (2011) found positive relationship between victimization and involvement in school activities. In our first regression we also found positive relationship but the relationship was insignificant. We omitted extracurricular activities variable so as to solve collinearity problem.
Again, Ousey et al. (2008) found positive relationship between victimization and access to illicit drugs. In our model, there is a positive relationship between victimization number and substance use. But again their relationship is insignificant.
Ousey et al. (2008) found positive relationship between victimization and selfcontrol. Accordingly, in our model, there is a positive relationship between number of fights and victimization. But again, the relationship is not significant.
Tillyer et al. (2011) found positive relationship between victimization and selfreported criminal behavior. Accordingly, in our model, there is a significant relationship between times victimized and substance use.
There are some limitations in our analysis as well. First, there are considerably missing data in variables that are widely used in the literature (number of times in a fight and how many days skipped). It probably affected our results.
Again, because the data were collected via cross sectional study, we have a causality problem. In cross sectional studies, causality may not be established well.
Conclusion and Implications
Based on the regression model that we adopted, our findings suggest that the school administrations should increase their schools’ safety level by increasing the allocation of personnel and equipment such as security cameras, locked doors and metal detectors.
Again, extracurricular activities seem to have positive relationship with the number of victimization although it is insignificant. The schools should be cautious about the ending times of these activities. They may end their extracurricular activities before the sunset so as to allow the students enough time to get their homes before it is dark. Furthermore, dismissal times should be better monitored by safety guards and other personnel.
Substance use seems to have positive relationship with the times victimized; therefore, supervision of the school inside and outside should be better addressed.
Finally, the number of engaging in a fight has positive significant relationship with victimization. Therefore, supervision of students during extracurricular and curricular activities is of vital importance in preventing fights.
A future study should address the missing data problem. It will affect the results considerably. Again, our descriptive statistics indicate that the data is not normally distributed.
References
Maier, R., F., Miethe, T., D. (1993). Understanding Theories of Criminal Victimization. Crime and Justice, 17, 459499
Ousey, G. C., Wilcox, P., & Brummel, S. (2008). Deja vu all over again: Investigating temporal continuity of adolescent victimization. Journal of Quantitative Criminology, 24, 307–355.
Sullivan, C. J., Wilcox, P., & Ousey, G. C. (2011). Trajectories of victimization from early to midadolescence. Criminal Justice and Behavior, 38, 85104.
Wilcox, P., Tillyer, M. S., & Fisher, B. S. (2009). Gendered opportunity? Schoolbased adolescent victimization. Journal of Research in Crime and Delinquency, 46(2), 245−269
Tillyer, M.S., Fisher, B. S., & Wilcox, P. (2011). The effect of school crime prevention on students’ violent victimization, risk perception, and fear of crime: A multilevel opportunity perspective. Justice Quarterly, 28(2), 249277.