Carrying weapons to school for protection: An analysis of the 2001 school crime supplement data

Abstract

This study focused on the reasons why adolescents bring guns to schools. It is argued that the etiology for carrying other weapons to school is not the same as that for carrying a gun to school for protection. Data from a nationally representative sample of school youth were analyzed with an appropriate analytic technique—multinomial logistic regression in dealing with nominal dependent variable. The results supported the hypotheses that both correlates and correlational strengths of carrying guns and other weapons to school for protection are different. Such factors as others' drug use, gangs at school, skipping school, perception, and age influence the probability of carrying other weapons to school, but are not related to carrying a gun to school. The three factors that relate to both carrying a gun and other weapons to schools are physical fights, peers' carrying guns, and gender. Their correlational strengths, however, are quite different.

Full text

Carrying weapons to school for protection:
An analysis of the 2001 school crime supplement data
Liqun Cao
a,
⁎, Yan Zhang
b
,NiHe
c
a
Department of Sociology, Anthropology, and Criminology, Eastern Michigan University, Ypsilanti, MI 48197, United States
b
College of Criminal Justice, Sam Houston State University, Box 2296 - S.H.S.U., Huntsville, TX 77341-2296, United States
c
College of Criminal Justice, Northeastern University, 360 Huntington Avenue, Boston, MA 02115, United States
Abstract
This study focused on the reasons why adolescents bring guns to schools. It is argued that the etiology for carrying other weapons to school is
not the same as that for carrying a gun to school for protection. Data from a nationally representative sample of school youth were analyzed with
an appropriate analytic technique—multinomial logistic regression in dealing with nominal dependent variable. The results supported the
hypotheses that both correlates and correlational strengths of carrying guns and other weapons to school for protection are different. Such factors
as others' drug use, gangs at school, skipping school, perception, and age influence the probability of carrying other weapons to school, but are not
related to carrying a gun to school. The three factors that relate to both carrying a gun and other weapons to schools are physical fights, peers'
carrying guns, and gender. Their correlational strengths, however, are quite different.
© 2008 Elsevier Ltd. All rights reserved.
Introduction
The fundamental purpose of government is to establish and
maintain order, a situation that is especially important for the
weaker groups in a society, such as the old and the young. The
study of carrying weapons to school is important because it
concerns the safety of one of the weak groups in society. When
gunshots are heard on school campuses, they arouse people's
strong emotions. The literature on why youth carry guns,
however, is clouded. The burgeoning literature that focuses on
weapon carrying aggregates less lethal weapon carrying and
gun carrying into the same category (Bailey, Flewelling, &
Rosenbaum, 1997; DuRant, Getts, Candenhead, & Woods,
1995; Forrest, Zychowski, Stuhldreher, & Ryan, 2000; Kulig,
Valentine, Griffith, & Ruthazer, 1998; Rountree, 2000; Simon,
Crosby, & Dahlberg, 1999; Wilcox & Clayton, 2001). Most
studies do not differentiate among purposes for carrying
weapons (Cook & Ludwig, 2004; DuRant et al., 1995; Forrest
et al., 2000; Kulig et al., 1998; Simon et al., 1999; Wilcox &
Clayton, 2001). The studies are local in nature (DuRant et al.,
1995; Kingery, Pruitt, & Heuberger, 1996; Kulig et al., 1998;
Lizotte, Krohn, Howell, Tobin, & Howard, 2000; Lizotte,
Tesoriero, Thornberry, & Krohn, 1994; Rountree, 2000; Sheley
& Brewer, 1995; Smith, 1996; Wilcox & Clayton, 2001), or
statewide (Bailey et al., 1997; Hayes & Hemenway, 1999;
Ruddell & Mays, 2003), or nonrandom (Braga & Kennedy,
2001; DuRant et al., 1995; Price, Desmond, & Smith, 1991;
Sheley & Wright, 1995). There is a conspicuous absence of
national studies on why guns are brought to school. This
study attempted to address some of the problems in the current
research on carrying guns and other weapons to school, and to
advance the understanding of the characteristics of those who
carry guns versus other weapons to schools.
Gun violence in the United States is both a public health
problem and a criminal justice issue (Fagan & Wilkinson, 1998;
Lizotte et al., 1994). Gun-related juvenile homicide peaked in
the early 1990s (Greenbaum, 1997). Despite a decline in recent
years, homicide rates in the U.S. remain high compared with
other industrialized countries (Zimring & Hawkins, 1997).
Firearms are still the weapons most frequently used for murder
(Federal Bureau of Investigation, 2002). The impact of gun
violence is even more pronounced among juveniles and young
adults. For persons between the ages of fifteen and twenty-four,
the homicide rate of 15.2 per 100,000 U.S. residents is higher
than the combined total homicide rate of all other industrialized
A
vailable online at www.sciencedirect.com
Journal of Criminal Justice 36 (2008) 154 –164
⁎Corresponding author. Tel.: +1 734 487 4246; fax: +1 734 487 7010.
E-mail address: lcao@emich.edu (L. Cao).
0047-2352/$ - see front matter © 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jcrimjus.2008.02.005

nations (Peters, Kochanek, & Murphy, 1998). From 1994
through 1999, about seven in ten murders in schools involved
some type of firearm, and approximately one in two murders
in schools involved a handgun (Perkins, 2003). Between 1994
and 1999, 172 students were murdered in school-associated
violence in the United States (Anderson et al., 2001).
Although gun violence in schools is rare, it has devastating
consequences when it occurs. It undermines the quality of
learning experiences, it reduces the positive activities of people
associated with the campuses, and it attracts negative media
coverage. Local, state, and national newspapers and television
swamp the schools to do stories on the incidents because gun
violence is significantly different from less lethal forms of
violence, such as fist fighting or even violence with a sharp
weapon, like a knife or a razor. While a low level of violence has
always been a part of the growing-up experience for teenagers,
the presence of a gun has changed the dynamics of interpersonal
conflict (DuRant et al., 1995). Whenever gunshots are heard on
school campuses, whether people live in a leafy suburb or in an
inner city, few can pretend that life is normal. When someone
brings a gun to school, campus safety becomes a major concern,
and the school is no longer perceived as a place with a special,
erudite environment protected from hazardous worldly happen-
ings. Due to the strong public reaction to the presence of guns in
schools, a better understanding of the etiology of such behavior
is warranted.
Literature review
Gun ownership is deeply rooted in the American culture. In
fact, America is described as a nation under gun (Cao, Cullen,
Barton, & Blevins, 2002; Wright, Rossi, & Daly, 1983). This
characterization reflects not only America's high rate of gun-
related violence, but also the widespread ownership of firearms.
There is a rich literature on why people own guns (Kleck, 1991;
Wright et al., 1983), and many theoretical models have been
developed to explain why Americans own guns (Cao, Cullen, &
Link, 1997). More recently, some of these models have been
extended to explain why adolescents carry weapons/guns
(Lizotte et al., 1994; Rountree, 2000; Wilcox & Clayton, 2001).
Despite the tremendous debate about the ownership of guns
among adults, there is a general consensus on the ownership of
guns by adolescents. Since the 1980s, states have passed laws
banning juvenile ownership of guns at an increasing frequency,
and in 1994, a federal law banned possession of handguns by
persons under eighteen years of age (Marvell, 2001). In addi-
tion, it is illegal everywhere in the U.S. to bring guns to schools
even in the states with laws permitting adults to carry concealed
guns (Kovandzic & Marvell, 2003). The current study focused
on the factors related to carrying guns/weapons to school among
adolescents.
A concentration of teenagers on a campus will inevitably re-
sult in conflicts. When violence-prone adolescents settle their
conflicts with guns, death is the likely result. Bringing a gun to
school, even without an incident, represents a major concern for
school administrators and for parents. Many views have been
proposed on why students bring weapons to school. Some of
these views were tested with data. According to the hypothesis
of fear and victimization (Wright et al., 1983), those who are
victims of crime and those who are more fearful of crime are
more likely to possess weapons. A related hypothesis, called the
“collective security hypothesis”(Cao et al., 1997; McDowall &
Loftin, 1983), suggests that people engage in self-help pro-
tective action, including arming themselves with guns, when
collective security is perceived to be low. Similarly, guns are
considered a symptom of “a culture of private justice”(Glaeser
& Glendon, 1998, p. 459). Those who feel vulnerable due to
previous victimization, are afraid of ineffective or reduced col-
lective security within the school, and believe in private retri-
bution are more likely to resort to self-help. In their statewide
sample of seventh and eighth graders in Illinois, however,
Bailey et al. (1997) found that victimization and fear for safety
were not significantly associated with weapon carrying for pro-
tection in their logistic multiple regression analysis. In a study
of African American adolescents living in and around public
housing, DuRant et al. (1995) found that frequency of weapon
carrying was positively associated with violent behaviors both
as an offender and as a victim among boys. Forrest et al. (2000)
confirmed in their national data that serious physical fighting
was associated with carrying weapons to school. They further
found that being stabbed before and feeling unsafe were both
associated with carrying weapons. Kingery et al. (1996) found
that carrying a handgun at school was associated with repeat-
ed victimization in a sample of rural Texas adolescents. In
Rountree's (2000) sample of Kentucky students, the effects of
fear and loathing were not significant for urban and western
counties, but experiencing threats increased the likelihood of
carrying weapons in eastern counties. In the further analysis of
the same data, however, Wilcox and Clayton (2001) found that
being threatened at school increased the likelihood of carrying
weapons to school while fear remained insignificant. Except for
Bailey's study, none of the studies specified the purpose of
bringing weapons to school. Their reasons for carrying weapons
to school for protection were only assumed.
The contextual factors are hypothesized to explain weapon
carrying to school. The presence of gang activities within
schools and juveniles involved with drugs are thought to
promote resorting to self-help in the form of carrying weapons to
school. Although involvement in crime and fear of victimization
often go hand in hand, the presence of gangs and drugs in school
can be regarded as contextual of weapon carrying to school
(Lizotte et al., 2000; Rountree, 2000). The presence of gang
activities are perceived by others as a threat, a situation that
stimulates the consciousness of self-protection. Past research
found that adolescents involved with drug use or sales were more
likely to carry weapons (Forrest et al., 2000; Kingery et al., 1996;
Kulig et al., 1998; Rountree, 2000; Sheley & Brewer, 1995;
Simon et al., 1999) and carrying guns (Lizotte et al., 2000).
It is proposed that weapon carrying to school may be in
response to normative influences in school. One such influence
is the effect of peers' weapon-carrying. This can also be re-
garded as part of pro-weapon socialization (Cao et al., 1997;
Rountree, 2000) or as a contagion effect (Loftin, 1986). Bailey
et al. (1997),Rountree (2000), and Wilcox and Clayton (2001)
155L. Cao et al. / Journal of Criminal Justice 36 (2008) 154–164

all found that indeed peer weapon carrying significantly
increased the likelihood of carrying a weapon to school. Lizotte
et al. (1994) found that peer gun ownership was related to
juvenile protective gun ownership, and Lizotte et al. (2000) dis-
covered that peer gun ownership for protection was associated
with carrying a hidden gun.
Another normative school-based influence is that a student's
social bond to school may affect his/her various deviant behav-
iors (Rountree, 2000), including weapon carrying. DuRant,
Kahn, Beckford, and Woods (1997) reported that poor academic
performance increased the likelihood of carrying weapons to
school. Kulig et al. (1998) found that skipping school was
correlated with weapon carrying among urban high school
students. Rountree's (2000) research, however, indicated that
school attachment, measured by six items of attitudes toward
teachers and school, decreased the likelihood of carrying weap-
ons in western counties, but the effect was not significant in
eastern counties and urban counties. Similarly, Wilcox and
Clayton (2001) found that strong school attachment reduced
the probability of carrying weapons to school. None of these
studies, however, separated weapon carrying to school from gun
carrying to school.
The current literature on carrying guns to school was limited
in several aspects. First, many studies tended to aggregate other
weapons with guns into one category (Bailey et al., 1997;
DuRant et al., 1995; Forrest et al., 2000; Kulig et al., 1998;
Rountree, 2000; Simon et al., 1999; Wilcox & Clayton, 2001).
While useful, such studies are misleading. There are qualitative
differences between other weapons (or even the sharp weapons,
such as a knife) and guns. While other weapons are potentially
deadly, they are more likely to cause injuries, and the process of
the violent act is slow and potentially preventable. By contrast,
the gun, the great equalizer, almost always provides a quick and
absolute solution. Once it is fired at someone, the result is
usually very devastating. It is more likely to cause death, or fatal
injury, and arouse fear among the public. In other words,
carrying other weapons may result in a low level of violence,
which has always been a part of school life, while carrying a gun
to school potentially leads to more instant and lethal violence.
Many otherwise law-abiding students may get involved with a
low level of violence sometimes in their lives, but never allow
themselves to be mixed up in more deadly violence. It is,
therefore, possible that those carrying guns to school and those
carrying other weapons to school are two different kinds of
students.
Second, most studies did not distinguish between the dif-
ferent purposes that students carry weapons to school (DuRant
et al., 1995; Forrest et al., 2000; Kulig et al., 1998; Simon et al.,
1999; Wilcox & Clayton, 2001). In studying why adolescents
carry handguns to school, Kingery et al. (1996) found that 48
percent of students said that it made them feel safer. Other
purposes for carrying guns to school included, in descending
order, attempting to kill (55 percent), bragging (20 percent),
bullying (19 percent), popularity (10 percent), and duress
(8 percent).
1
Other empirical studies on adolescents' carrying
weapons drew the conclusion not based on solid findings, but
on speculation. For example, without specifying the purpose of
carrying weapons to school, DuRant et al. (1995) speculated
that carrying weapons was indicative of and associated with
violent and aggressive behavior, and it was for purposeful
criminal activity, rather than self-protection. The purpose for
carrying weapons to school is a very important issue. In the
study of patterns of adolescent firearms ownership and use,
Lizotte et al. (1994) found that factors related to sports gun
ownership were largely different from those related to
protection gun ownership. It is, thus, possible that social cor-
relates that relate to carrying weapons may not necessarily be
related to carrying guns to school for protection and that when
they do share a correlation, the strength of that correlation may
differ.
Third, many studies focused on juvenile ownership of guns
(Bjerregaard & Lizotte, 1995; Lizotte et al., 1994; Price et al.,
1991) or carrying a hidden gun without specification of the
premises (Cook & Ludwig, 2004; DuRant et al., 1995; Kulig
et al., 1998; Lizotte et al., 2000). Many more adolescents might
own guns at home, but would not bring them to school. Owning
a gun and carrying a gun to school are two different events.
Although the two might be highly correlated, carrying a gun
before and after school and carrying a gun into the school
premises are also two different concepts.
Fourth, the data from previous studies were nonrandom,
local, or only statewide. Their generalizability was limited.
There were only three national studies on carrying guns/
weapons among adolescents. Sheley and Wright's (1995) data
were limited to inner-city high school males, Forrest et al.'s
(2000) national research was limited to carrying weapons
(guns, knives, razors, clubs, sticks, or pipes) to school without
specifying the purpose for carrying weapons, while Cook and
Ludwig's (2004) national study was limited to gun carrying
among all male adolescents aged fifteen to nineteen years old
regardless of whether they were enrolled in school or not, or
regardless of whether they carried their guns to school or not.
Statistics of the male teenage population under the age of
nineteen masked the fact that many of the adolescents,
especially the serious delinquents, had dropped out of the
school.
The current research contributed to filling the gaps in the
literature primarily through analysis of data in the 2001 School
Crime Supplement (SCS) as a part of the National Crime
Victimization Survey (NCVS). NCVS-SCS employs a repre-
sentative sample of all school populations in the U.S. and
separates carrying guns for protection from carrying other
weapons for protection to school. It attempts to explore whether
or not social correlates that relate to carrying other weapons for
protection and to carrying guns for protection are the same, and
that when they share a correlation, whether the strength of the
correlations is the same or not. In approaching the analysis, the
following hypotheses were tested:
Hypothesis 1. Personal experiences, such as victimization,
being bullied, fear, avoidance of attack, and being involved in
physical fights are related to carrying other weapons to school
for protection more strongly than to carrying guns to school for
protection.
156 L. Cao et al. / Journal of Criminal Justice 36 (2008) 154–164

Hypothesis 2. Contextual variables, such as gang presence,
drug availability, peer weapon carrying, and skipping school,
increase the likelihood of carrying a gun to school for protection
more than the likelihood of carrying other weapons to school for
protection.
Hypothesis 3. In addition to these two sets of hypotheses, an
auxiliary hypothesis was also tested: gun/weapon carrying to
school is related to students' perception of the school rules and
can be prevented by security measures.
While policy recommendations to improve the security of
school are many, the extant literature is almost blank regarding
the basic question of what works to stop weapons/gun carrying
to school. The SCS data happened to have measures of school
security and safety, and the perception of reinforcement of
school rules.
The above three sets of hypotheses were tested with two
groups of control variables—socio-demographic control vari-
ables and geographic distribution control variables. Among the
social demographic variables, the existing literature suggests
that older teens, males, African Americans/Hispanics, and
juveniles from lower-income families are more likely than
younger teens, females, Whites, and juveniles from mid- and
upper-income families to carry weapons to school (Bailey et al.,
1997; DuRant et al., 1995; Forrest et al., 2000; Hayes &
Hemenway, 1999; Kingery et al., 1996; Kulig et al., 1998;
Simon et al., 1999; Wilcox & Clayton, 2001). Since it was a
national study, geographic distribution impacts were controlled
for. The literature of guns in the general population indicated
that the South and West have a unique history of a subculture of
gun ownership (Cao et al., 2002; Cao et al., 1997; Cook &
Ludwig, 2004; Felson, Liska, South, & McNulty, 1994). As for
the relationship between residential areas and weapons carry-
ing, Bailey et al. (1997) did not find a significant relationship
between students of urban and suburban areas nor between
students of the urban and rural areas. Little empirical
information exists about the pattern of weapon carrying in
private schools. Almost all of the previous studies had been
conducted in public schools. The effects of the South and the
West, the effects of inner city versus suburbs and rural areas,
and the effects of private versus public schools were controlled
for. It is expected that students in the South, in rural areas, and in
public schools are more likely to bring weapons to schools than
students in the non-South, in urban areas, and in private schools.
Methods
The sample
The data from the 2001 School Crime Supplement (SCS, there-
after) were part of the National Crime Victimization Survey,
available from the National Archive of Criminal Justice Data
(NACJD) at the Inter-University Consortium for Political and
Social Research (ICPSR). The Supplement was an effort to pro-
vide information about school-related victimizations on a national
level so that policymakers as well as academic researchers and
practitioners can make informed decisions concerning this issue.
The data included those in households with adolescents between
the ages of twelve to eighteen who attended school at any time
during the six months preceding the interviews. The 2001 SCS
questionnaire was similar to the 1999 questionnaire, but added
many new questions (including several critical variables con-
cerning this study) and changed the wording of some existing
questions.
The 2001 SCS had data on 9,277 persons, excluding 374
persons (3.9 percent) without any values. Since the purpose of
the current research was to understand factors related to car-
rying weapons to school for protection, the analysis was re-
stricted to those who attended either public or private schools
during the last six months. Respondents who did not attend
school (450 or 4.7 percent) or who were home-schooled (273 or
2.8 percent) were excluded. Since none of the Asian American
and Native American respondents (432 or 4.5 percent) indicated
that they carried any weapons to school in the past six months,
this study was further restricted to the three dominant racial
and ethnic groups in the U.S.: non-Hispanic Whites, Hispanic
Whites, and African Americans. There were 9.1 percent missing
values in one or more variables. According to Malhotra (1987)
and Raymonds and Roberts (1987), missing data under 10
percent can generally be ignored except when the missing data
occur in a specific nonrandom fashion. To diagnose whether the
missing data were random or not, one dummy variable was
created, where cases with missing values were coded as 1 and
otherwise coded as 0. A couple of t-tests (if the independent
variable was categorical, then chi-square test was used) were
conducted based on the dummy variable and independent var-
iables. These tests did not exhibit any specific patterns. It was
concluded that the missing data were random. Therefore, list-
wise deletion of variables with missing data was applied for all
following analyses. The final number in the analysis was 7,391
school age respondents (see U.S. Department of Justice, Bureau
of Justice Statistics, 2001, for more detailed discussion).
Measures
Dependent variable
The dependent variable refers to carrying a specific category
of weapon (a gun, or other weapons) to school in the past six
months. This three-category variable was created by reorganiz-
ing three items in the survey so that those who brought a gun
to school were coded as 2, a knife and/or any other form of
weapons as 1, and those who never carried any weapons to
school served as the reference category in the analysis. The exact
wording of these items were: (1) “Some people bring guns,
knives, or objects that can be used as weapons to school for
protection. During the last six months, did you ever bring a gun
to school or onto school grounds?”(2) “During the last six
months, did you ever bring a knife to school or onto school
grounds? Include only knives brought as weapons?”and
(3) “During the last six months, did you ever bring some other
weapon to school or onto school grounds?”It was clear from
these items that the weapons/guns were for protective purposes
157L. Cao et al. / Journal of Criminal Justice 36 (2008) 154–164

and that the locations for carrying weapons were schools or
school grounds. There were twelve persons who carried two
types of weapon. The hierarchical rule of only counting the more
serious type of weapons (i.e., guns) was applied in the data
classification.
Independent variables
There were two groups of independent variables that were
supposedly related to guns/weapons carrying to school. Personal
experiences included victim experiences, being bullied, fear,
avoidance, physical fights, others' drug use, drug availability,
peers' carrying weapons, and gang presence. The variable vic-
timization was a binary variable with 1 indicating both violent-
crime victim and property-crime victim and “no”as the reference
group. Having been bullied at school was a binary variable where
those students who were bullied, felt rejected by other students,
and/or were called derogatory names were coded as 1 and those
without any such experiences were treated as the reference group.
Fear was a scale of three items measuring respondents' subjective
emotion of being afraid of being attacked (see Appendix A). The
Cronbach's alpha for “fear”was .76. Avoidance was an additive
scale that included the change of behavior because of fear.
The Cronbach's alpha for “avoidance”was .78. Both fear and
avoidance were coded in a way that the higher value meant being
more afraid of being attacked and more likely to change their
behavior because of fear (see Appendix A). Physical fight was a
dummy variable with those who were in one or more physical
fights at school coded as 1 and those who were not serving as the
reference group.
The social context included access to alcohol and drugs,
others' drug use, peers' carrying guns, gang presence, and
skipping classes. Access to alcohol and drugs was an additive
scale measuring the ease of securing alcohol and various drugs
on campus (see Appendix A). Lower value meant easier access
to alcohol and drugs. The Cronbach's alpha for “access to
alcohol and drugs”was .89. Others' drug use was a binary
variable with those who said that they knew for sure that other
students were on drugs or alcohol while at school as 1 and those
who did not know for sure as the reference group. Peers' carrying
guns to school was a dummy variable with those who knew other
students who brought a gun to school as 1 and those who did not
know as the reference group. Gang presence at school was a
three-category nominal variable. The “don't know”category
was retained because of its large proportion (13 percent). The
reference category was “no gang at school.”Skipping school
was a binary variable where those who admitted skipping school
in the past six months were coded as 1 and those who said they
did not were coded as the reference group.
2
In addition, measures concerning school security and safety
and concerning the perception of reinforcement of school rules
were included in the analysis.
3
The variable “school guards”
was a binary that asked whether the school had security guards
or had assigned police officers or not. The variable “perception
of school rule”was an index created out of five items (see
Appendix A). The higher scores indicated that the students did
not think the school rule was clear, fair, or strictly enforced. It
was hypothesized that police/school guards will reduce gun/
weapon carrying of students and that students who do not accept
school rules are more likely to bring a gun or other weapons to
school.
Control variables
There were two groups of control variables: demographic
variables and distributional variables. The demographic control
variables included age, gender, race/ethnicity, and parent's
education. Age was measured by the respondents' actual age at
the time of the survey. Gender was a binary variable with male as
1 and female as the reference group. Race/ethnicity was a series
of two dummy variables, where a Hispanic White was coded as 1
and African American was coded as 1 with non-Hispanic Whites
as the reference group. The data set had an item called “reference
person's education year.”Ninety-two percent of the reference
persons were either the respondent's father or mother. It was
coded as a continuous variable, ranging from 0 to 18. This
variable was used as the proxy for the students' social economic
status (Cao, Cao, & Zhao, 2004; Rountree, 2000).
Finally, there were three distributional control variables. The
first was the system control: those educated in a private school
were coded as 1 and those in a public school were coded as the
reference group. The second was the local geographic control
variable, which was a three-category nominal variable with the
central metropolitan area as the reference group. The third was
the national geographic control variable consisting of four
categories with the Northeast as the reference group. It was
expected that students attending a public school, living in the
inner city and living in the South, and West are more likely to
carry guns to school than students in a private school, in the
suburbs, and in the Northeast.
Statistical modeling
In the gun literature, multivariate research typically had
relied on binary logistic regression. This technique is proble-
matic because it restricts comparison to two categories when
in reality three or more categories are of theoretical interest. The
dependent variable, weapon carrying, is a three-category nom-
inal variable: carrying other weapons, carrying a gun, and not
carrying any weapons to school. Binary comparisons poten-
tially result in misspecified models (Lizotte et al., 1994), and
multinomial logistic regression is a more appropriate model for
such occasions (Cao et al., 1997; Greene, 1993; Long, 1997).
In this procedure, the probabilities of carrying any weapon
to school were estimated simultaneously with the probabilities
of not carrying any weapons and of carrying a gun to school.
The procedure used all the information simultaneously, thereby
producing the most efficient estimates with the integration of
a number of choice probabilities within an efficient closed
form. The interpretation of the coefficients, however, was less
straightforward. In analyzing the dependent variable with two
outcomes, it is acceptable to interpret the results directly from
regression coefficients or in odds ratios: with only one degree of
freedom left, the variation in the ratio of two choice probabilities
158 L. Cao et al. / Journal of Criminal Justice 36 (2008) 154–164

does reflect the changes in the probabilities per se.Applyingthe
same approach for the multinomial logit model, however, can be
misleading. With the presence of more than two alternatives, the
regression coefficients may not necessarily bear any relationship
with the changes in the probabilities (Cao et al., 1997; Greene,
1993). Since the probabilities for all choices always sum to unit,
the simultaneous structural variation in the proportional distribu-
tion of more than two alternatives is beyond what the changes in a
set of log odds can capture. Additional procedures are needed to
facilitate the understanding of the coefficients of the multinomial
logistic regression. The marginal effects of each independent
variable on each choice of the dependent variable are calculated.
These marginal effects, or discrete changes according to Long
(1997), shed new light on the logit coefficients because they not
only provide more reliable signs for the effects of each inde-
pendent variable on the dependent variable, but also reveal the
magnitude of each independent variable on the dependent var-
iable (Cao et al., 1997; Greene, 1993).
Results
Table 1 presents the descriptive sample characteristics. As
expected, carrying a weapon to school for protection is a rare
event. About 98 percent of students in the sample did not carry
any weapons to school for protection in the past six months.
Among those 2 percent who carried a weapon to school, the
majority of them (1.3 percent) carried other weapons to school,
and only .5 percent of them carried a gun.
4
The demographic
and geographic variables showed that the sample was very
representative of the U. S. school population.
Table 2 reports the results of the comprehensive model of
multinomial logistic regression. Multicollinearity tests were
conducted and results indicated no such threat to the data
analysis. The table presents the logit coefficients. It was found
that those who did not bring any weapon to school in the past six
months were contrasted with those who carried other weapons
to school for protection (Column 2), and with those who carried
guns to school for protection (Column 3). For Column 2 car-
rying of other weapons, the results showed that those who were
more afraid of being attacked, who were involved in physical
fights, who were aware of others' use of drugs, who knew
friends carrying a gun, who knew the presence of gangs at
school, and who skipped school were more likely to bring other
weapons to school. For school security measures, one's per-
ception of enforcement of school rules was related to carrying
other weapons to school for protection, but was not related to
carrying guns to school. Security guards were not related to
carrying any category of weapon to school. Other independent
variables in this column were not statistically significant. For
the control variables, age and gender were both related to car-
rying other weapons to school, while none of the geographic
distribution controls were statistically significant.
Finally, carrying a gun to school for protection (Column 3) was
statistically related only to physical fights and peers' carrying
guns to school: those who were involved in physical fights in the
past six months and those who had friends carrying guns to school
were more likely to bring guns to schools. Fear, having been
bullied, others' drug use, gang presence, and skipping school
were not related to carrying guns to school. The rest of the
independent variables were not statistically significant. Among
the control variables, only gender was related to gun carrying
to school. The model chi square was 254 and Nagelkerke R
2
was .184.
Table 3 presents the reduced model that reassessed all
significant predictors in the comprehensive model for weapon
carrying to school. Except fear of being attacked, all statistically
significant variables in Table 2 remained significant in the final
model. In addition, the marginal effects of each predictor were
calculated. These marginal effects revealed the change in the
predicted probability when the independent variable changed
one unit. If the independent variable was a categorical variable,
the marginal effects indicated the differences in the predicted
probability between two categories. Columns 2 and 3 of Table 3
Table 1
Descriptive sample characteristics (n = 7,391)
Characteristics Percentages or means
a
Cases
Dependent variable
Carrying weapons to school
Never carrying any weapons 98.2% 7,257
Carrying other weapons 1.3% 96
Carrying guns 0.5% 38
Independent variables
Victimization 8.2% 609
Being bullied (yes = 1) 21.6% 1,595
Fear of being attacked 3.65 (1.3)
Avoidance .17 (.77)
Physical fights 5.4% 402
Access to alcohol and drugs 46.21 (7.0)
Others' drug use 39.0% 2,884
Peer carrying gun 5.1% 377
Gangs at school
Yes 20.3% 1,498
Don't know 13.0% 958
No 66.7% 4,935
Skipping school (yes = 1) 8.9% 660
Security guards/police 65.6% 4,848
Perception of school rule 9.66 (2.2)
Control variables
Age 14.65 (1.8)
Gender (male = 1) 52.3% 3,862
Race/ethnicity
Hispanic White 16.2% 1,200
African American 13.9% 1,025
Non-Hispanic White 69.9% 5,166
Parent's education in years 13.12 (3.1)
Metropolitan statistical area (MSA)
Not MSA 15.5% 1,148
MSA not central 57.2% 4,228
MSA central 27.3% 2,015
Region
West 23.2% 1,707
Midwest 25.5% 1,886
South 34.2% 2,531
Northeast 17.1% 1,267
Public school 91.0% 6,726
a
For continuous variables, the means and standard deviations are presented.
159L. Cao et al. / Journal of Criminal Justice 36 (2008) 154–164

report the logit coefficients of all independent variables in the
reduced models, and the next three columns present the prob-
abilities of marginal effects.
From reading the sample estimates, it was learned that the
predicted probability for carrying other weapons to school was
.67 percent, the predicted probability of carrying guns to school
was .3 percent, and the predicted probability of not carrying any
weapons to school was 99.03 percent with all values of indepen-
dent variables fixed at the sample means. The results showed
clearly how small a predicted probability it was for students
with all characteristics in the sample to carry a gun to school.
The marginal effects showed that physical fights increased
the probabilities of bringing other weapons to school for pro-
tection by 1.8 percent and increased the probabilities of bringing
a gun to school for protection by .56 percent. Having a peer
carrying a gun increased the probabilities of bringing other
weapons to school by .9 percent, and the probabilities of
bringing a gun to school by 1.6 percent. Gang presence at
school increased the probabilities of bringing other weapons to
school by .5 percent. Having skipped school increased the
probability of bringing other weapons to school by .67 percent.
One score increased in the student's perception of reinforcement
of school rules increased the probabilities of carrying other
weapons to school by .12 percent. The other probabilities of the
marginal effects of perception were not statistically significant.
One year's increase in age increased the probability of carrying
other weapons to school by .1 percent. Being a male increased
the probabilities of bringing other weapons to school by .41
percent and of bringing a gun to school by .21 percent. Although
these effects were small, it was interesting to note that the effect
of physical fights was twice as strong on the probability of
carrying other weapons to school as to carrying a gun to school
(1.8 versus .6). The effect of peers' carrying guns was almost
twice as weak on the probability of carrying other weapons
for protection as to carrying a gun to school for protection
(.9 versus 1.6). The model chi square was 211 and Nagelkerke
R
2
was .155.
Discussion
By analyzing the 2001 School Crime Supplement data con-
sisting of a nationally representative sample of school adoles-
cents from ages twelve to eighteen, this study tested whether or
not the social correlates of carrying other weapons and for
carrying guns for protection to school are the same, and that
when they share a correlation, whether or not the strengths of
the correlation are different. Using an appropriate regression
technique of analysis—multinomial logistic regression in
dealing with the nominal dependent variable, the current
Table 3
Reduced model of multinomial logit of carrying weapon to school (coefficients
and probability reported)
Independent
variables
Estimated logit coefficient Effects on probability
Other weapons Guns p
1
p
2
p
3
b
1
b
2
Physical fights 1.383⁎⁎ 1.109⁎⁎ .0181 .0056 −.0237
Peer carrying gun .900⁎⁎ 1.958⁎⁎ .0090 .0160 −.0250
Other's drug use .763⁎⁎ .305 .0056 .0009 −.0065
Gangs at school
Yes .626⁎⁎ .755 .0050 .0029 −.0079
Don't know .363 .937⁎.0027 .0041 −.0068
Skipping school .732⁎⁎ .277 .0067 .0009 −.0076
Perception .165⁎⁎ .070 .0012 .0002 −.0014
Age .141⁎−.069 .0010 −.0002 −.0008
Gender (male = 1) .619⁎⁎ .719⁎.0041 .0021 −.0062
Intercepts −9.621⁎−6.410⁎
Sample estimates
a
.0067 .0030 .9903
Model chi-square 211 (d.f. = 18)
Nagelkerke pseudo
R square
.155
Note: p
1
= probability of carrying other weapons to school for protection;
p
2
= probability of carrying guns to school for protection;
p
3
= probability of not carrying any weapons to school.
a
Sample estimates refer to the estimated probabilities for those who carry
knives or any other weapons to school, who carry guns to school, and who
never carry any weapons to school, with the values of independent variables
fixed at the sample means.
⁎p≤.05.
⁎⁎ p≤0.01 (n = 7,391).
Table 2
Multinomial logit model of carrying weapons to school (logit coefficients and
standard errors reported)
Independent variables Other weapons Guns
b
1
b
2
Victimization .52 (.27) −.69 (.63)
Being bullied .40 (.24) .44 (.37)
Fear of being attacked .14⁎(.07) .02 (.12)
Avoidance −.24 (.14) −.43 (.30)
Physical fights 1.21⁎⁎ (.28) 1.08⁎(.43)
Others' drug use .60⁎(.28) .28 (.41)
Access to alcohol and drugs −.02 (.01) −.02 (.02)
Peer carrying gun .74⁎(.29) 1.85⁎⁎ (.39)
Gangs at school
Yes .61⁎(.26) .64 (.43)
Don't know .40 (.36) .87 (.48)
Skipping school .71⁎⁎ (.26) .28 (.47)
Security guards/police −.24 (.27) .07 (.45)
Perception of school rule .14⁎⁎ (.05) .08 (.08)
Age .65⁎⁎ (.23) .09 (.10)
Gender (male = 1) −.21 (.46) .76⁎(.37)
Race/ethnicity
Hispanic White −.66 (.40) −.81 (.59)
African American −.05 (.31) .09 (.43)
Parent's education .01 (.04) −.11 (.06)
Public school −.09 (.50) −.19 (.79)
Metropolitan statistical area (MSA)
Not MSA −.03 (.35) −.35 (.54)
MSA not central −.05 (.31) −.74 (.39)
MSA central
Region
West −.00 (.37) −.30 (.59)
Midwest −.00 (.36) −.28 (.57)
South .36 (.34) .19 (.51)
Intercepts −9.18⁎⁎ (1.69) −3.31 (2.47)
Model chi-square 254 (d.f. = 48)
Nagelkerke pseudo R square .184
*p
≤0.05.
** p ≤0.01 (n = 7,391).
160 L. Cao et al. / Journal of Criminal Justice 36 (2008) 154–164

model was capable of analyzing the data in a more refined way.
With the help of the marginal effects of probabilities, the results
showed that factors explaining other weapons carrying are not
necessarily able to explain gun carrying to school for protection.
It is, therefore, more informational to investigate these different
categories of weapon carryings separately.
The analyses revealed that both carrying a gun and other
weapons to school for protection in the past six months was a
rare event and the correlates of carrying a gun and of carrying
other weapons for protection were not the same. Both personal
experiences and social context were important social correlates
of carrying weapons to school for protection. Specifically,
carrying other weapons to school was significantly related to
eight factors, while carrying a gun to school for protection was
associated only with three factors. The variables correlated with
carrying other weapons to school were one's involvement in
physical fights, having peers' carrying guns, knowing others'
use of illegal drugs, skipping school, and gang presence at
school. Age and gender were also related to carrying other
weapons to school with older teenagers and boys being more
likely than girls. The results were quite consistent with the
findings in the previous studies (Bailey et al., 1997; DuRant
et al., 1995, 1997; Forrest et al., 2000; Lizotte et al., 2000, 1994;
Rountree, 2000; Simon et al., 1999; Wilcox & Clayton, 2001),
suggesting that weapon carrying is a partial response to the
situational imperatives that youth face in school (Lizotte et al.,
2000).
Carrying a gun to school for protection, however, was only
related to one's involvement in physical fights, having peers
bringing guns to school, and being male. These results were
unique in the sense that the existing literature (Bailey et al.,
1997; DuRant et al., 1995; Forrest et al., 2000; Rountree, 2000;
Wilcox & Clayton, 2001) failed to separate gun carrying from
other weapon carrying to school. Using an appropriate
regression technique of analysis—multinomial logistic regres-
sion in dealing with the nominal dependent variable, the results
revealed that factors explaining the weapons carrying to school
do not always explain carrying a gun to school. These results
have given credence to the claim that the correlates of carrying a
gun and other weapons to school are not the same. The low level
of violent inclination (including carrying other weapons to
school for protection) exhibited by some adolescents is not
necessarily transferable to the high level of violent tendency by
carrying a gun to school for protection.
More importantly, the strengths of correlations were different
as hypothesized. Physical fights were associated with carrying
other weapons to school for protection more strongly than with
carrying a gun to school for protection. Being a male was twice
as likely to carry other weapons to school as to carry a gun to
school. Taken together, these results indicated that more male
students are willing to settle their differences with the lower
level of violence than with the high level of violence, such as
guns. Peers' carrying guns, however, was related to carrying
guns to school for protection more strongly than to carrying
other weapons to school for protection. The heightened risk
stimulates the heightened need for self-protection. It was inter-
esting to point out that students' perception that their peers had
guns was not accurate. While only a .5 percent of students
carried guns to school, more than 5 percent of students thought
that their peers carried guns to school. Without the partitioning
of weapon carrying into the three separate categories, these
important differences might have been obscured in the larger
category of all weapon carrying. For the policy implications
from the current study, the results seem to suggest that policies
aimed at preventing general violence, such as physical fights
and peers' weapon carrying, will also reduce the probabilities of
carrying guns to school for protection, but policies aimed to
control gangs and drugs in schools only affect carrying other
weapons to school and may not have much impact on carrying
guns to school for protection.
Some insignificant effects in the comprehensive model are
worth discussing, too. None of the well-measured fear and
victimization variables was significant. This confirmed the
findings and arguments by Bailey et al. (1997),Rountree
(2000), and Simon et al. (1999) that weapon carrying to school
has a unique etiology that is different from adult weapon
carrying or adolescent weapon carrying outside school: it is
unrelated to victimization or fear. This interpretation is also
consistent with Webster, Gainer, and Champion's (1993)
argument that gun carrying among school students is part of a
larger set of highly aggressive delinquent behavior rather than a
defensive response to victimization. Neither the regional control
variables nor racial/ethnic control variables were significant.
Finally, there was no difference among students carrying any
weapons to school in terms of rural, suburban, and inner city.
These findings were inconsistent with the results reported by
Cook and Ludwig (2004), who found that the likelihood of
adolescent gun carrying increased with the prevalence of local
gun ownership. Their data, however, were limited to teenage
males between fifteen and nineteen years old, many of whom
might have dropped out of school. The premises of gun carrying
were not specified, nor were the purposes of carrying guns. In
contrast, the current data included both boys and girls aged
twelve to eighteen, the premises were specified as school
grounds, and the purpose of carrying gun/weapons was for
protection. Besides these differences in data, there are two
alternative interpretations. First, Sheley and Wright (1995)
argued that the community context of gun ownership and
availability was irrelevant to youth's gun involvement. There
may be higher rates of gun ownership in the South and in the
rural areas, but students in these areas do not necessarily bring
their guns to school more often. Similarly, the debate over adult
gun ownership has not affected the national consensus
regarding the restriction of gun ownership and gun carrying
for adolescents. The current results are consistent with these
arguments.
The second possibility is that the current data collected in
2001 were more recent than Cook and Ludwig's data collected
in 1995, and tough measures taken against weapons in school
may have shown their effects among students. According to the
1999 Annual Report on School Safety (U.S. Department of
Education and the U.S. Department of Justice, 1999), between
the 1997 and 1998 school years, the states and territories
reported that they had expelled an estimated 3,930 students for
161L. Cao et al. / Journal of Criminal Justice 36 (2008) 154–164

bringing firearms to school (a rather significant decline from
6,093 in the 1996–97 school year). These measures might have
driven the students with guns out of school and might have
prevented those who would otherwise carry a gun to school
from doing so. Data did not allow for much speculation in this
area.
Finally, the ancillary findings appear to indicate that various
security measures in schools are not significantly related to
carrying any category of weapons to school. The students' per-
ceptions of the strictness of rule enforcement in school is
weakly related to carrying other weapons to school with those
who regarded the rules as fair and strict being less likely to carry
other weapons to school. Although the cross-sectional data
cannot fully settle the debate of the effectiveness of various
security measures, these results seem to suggest cautiously that
schools' security in times of fiscal conservatism is unlikely to
improve much by purchasing more costly devices to stop
students from carrying weapons into school. It is more practical
for the school administrators to design a program that enhances
communication with students as well as with parents or guard-
ians of students and educate them about the potential danger of
guns and weapons in the hands of adolescents.
While the main conclusion that there is a difference in the
correlates and in the strength of correlates of carrying other
weapons and of carrying a gun to school for protection is solid,
there were several limitations to the current study. First, the small
number of cases in the category of carrying guns to school may
limit the generalizability of the study and may increase the effect
of possible measurement error. Second, the current analysis was
applicable only to non-Hispanic Whites, Hispanic Whites, and
African Americans. Other minority groups were excluded.
Third, although the sample was nationally representative and the
list of independent variables was comprehensive, the data were
from a victimization survey. As a result, many salient variables
that may be more predictive of carrying a gun to school were not
available. For example, to test the criminal-lifestyle hypothesis
fully, direct measures of drug sale/use and violent perpetrators
would have to be available, but none of these variables was
found in the victimization data. Finally, the proxy variable of
being involved in physical fights was the most consistent pre-
dictor of carrying weapons/guns to school, but it was unclear
whether these adolescents were largely perpetrators or largely
victims in these physical fights. This distinction might prove
to be very important (see Brown, 2004). The limitations of the
present study suggest possible avenues for future research.
Acknowledgements
We want to thank Professor Alethea Helbig for polishing the
language of the article. All viewpoints are those of the authors.
Appendix A. Items in the index measures
Fear (alpha = .758)
1. How often are you afraid that someone will attack you or
threaten to attack you at school?
2. How often are you afraid that someone will attack you or
threaten to attack you on the way to and from school?
3. Besides the times you are at school or going to or from
school, how often are you afraid that someone will attack
you or threaten to attack?
(1 = never, 2 = almost never, 3 = sometimes, 4 = most of the
time)
Avoidance (alpha = .781)
Did you stay away because you thought someone might at-
tack or threaten to attack you?
1. The shortest route to school
2. The entrance into school
3. Hallways/stairs in school
4. Parts of school cafeteria
5. Any school restrooms
6. Other places in school building
7. School parking lot
8. Other places on school grounds
9. Did you avoid any extracurricular activities at your school
because you thought someone might attack or threaten to
attack you?
10. Did you avoid any class at your school because you thought
someone might attack or threaten to attack you?
11. Did you stay home because you thought someone might
attack or threaten to attack you?
(1 = yes, 0 = no)
Access to alcohol and drugs (alpha = .885)
Would you say ______ to get at your school?
1. Alcoholic beverages
2. Marijuana
3. Crack
4. Other forms of cocaine
5. Uppers, such as ecstasy, crystal meth, or other illegal
stimulants
6. Downers such as GHB/sleep pill
7. LSD or acid
8. PCP or angel dust
9. Heroin or smack
10. Other illegal drugs
(1 = easy, 2 = fairly easy, 3 = fairly hard, 4 = hard, 5 = not
available at all)
Perception of school rule (alpha = .749)
1. Everyone knows what the school rules are
2. The school rules are fair
3. Punishment for breaking school rules is the same no matter
who you are
162 L. Cao et al. / Journal of Criminal Justice 36 (2008) 154–164

4. School rules are strictly enforced
5. If a school rule is broken, students know what kind of pun-
ishment will follow.
(1 = strongly agree, 2 = agree, 3 = disagree, 4 = strongly
disagree).
References
Anderson, M., Kaufman, J., Simon, T. R., Barrios, L., Paulozzi, L., Ryan, G., et al.
(2001). School-associated violent death in the United States, 1994–1999.
Journal of the American Medical Association,286(21), 2695−2702.
Bailey, S. L., Flewelling, R. L., & Rosenbaum, D. P. (1997). Characteristics of
students who bring weapons to school. Journal of Adolescent Health,20,
261−270.
Bjerregaard, B., & Lizotte, A. J. (1995). Gun ownership and gang membership.
Journal of Criminal Law and Criminology,86,37−58.
Braga, A., & Kennedy, D. M. (2001). The illicit acquisition of firearms by youth
and juveniles. Journal of Criminal Justice,29, 379−388.
Brown, B. (2004). Juveniles and weapons: Recent research, conceptual consid-
erations, and programmatic interventions. Youth Violence and Juvenile
Justice,2, 161−184.
Cao, L., Cao, J., & Zhao, J. (2004). Family, welfare and delinquency. Journal of
Criminal Justice,32(6), 565−576.
Cao, L., Cullen, F. T., Barton, S. M., & Blevins, K. R. (2002). Willingness to
shoot: Public attitudes toward defensive gun use. American Journal of
Criminal Justice,27,85−105.
Cao, L., Cullen, F. T., & Link, B. G. (1997). The social determinants of gun
ownership: Self-protection in an urban environment. Criminology,35(4),
629−657.
Cook, P. J., & Ludwig, J. (2004). Does gun prevalence affect teen gun carrying
after all? Criminology,42,27−54.
DuRant, R. H., Getts, A. G., Candenhead, C., & Woods, E. R. (1995). The asso-
ciation between weapon carrying and the use of violence among adolescents
living in and around public housing. Journal of Adolescent Health,17, 376−380.
DuRant, R. H., Kahn, J., Beckford, P. H., & Woods, E. R. (1997). The asso-
ciation between weapon carrying and fighting on school property and other
health risk and problem behaviors among high school students. Archives of
Pediatric Adolescent Medicine,151, 360−366.
Fagan, J., & Wilkinson, D. L. (1998). Guns, youth violence, and social identity
in inner cities. Crime and Justice,24, 105−188.
Federal Bureau of Investigation. (2002). Crime in the United States 2001.
Washington, DC: U.S. Department of Justice.
Felson, R. B., Liska, A. E., South, S. J., & McNulty, T. L. (1994). The subculture
of violence and delinquency: Individual versus school context effects. Social
Forces,73, 155−173.
Forrest, K. Y. Z., Zychowski, A. K., Stuhldreher, W. L., & Ryan, W. J. (2000).
Weapon-carrying in school: Prevalence and association with other violent
behaviors. American Journal of Health Studies,16, 133−140.
Glaeser, E. L., & Glendon, S. (1998). Who owns guns? Criminals, victims, and
the culture of violence. American Economic Review,88(2), 458−462.
Greenbaum, S. (1997). Kids and guns: From playgrounds to battlegrounds.
Juvenile Justice,3(2), 3−10.
Greene,W.H.(1993).Econometric analysis. Englewood Cliffs, NJ: Prentice Hall.
Hair, J. F., Jr., Black, W. C., Babin, B. J., Anderson, R. E., & Tatham, R. L.
(2006). Multivariate data analysis. Upper Saddle River, NJ: Prentice
Hall.
Hayes, D. N., & Hemenway, D. (1999). Age-within-school-class and adolescent
gun-carrying. Pediatrics,103(5), 64−76.
He, N., & Swatt, M. (2005). Do the right thing or does it matter? Analyzing
NCVS data with corrected standard errors. Paper presented at the 2005
annual meeting of Academy of Criminal Justice Sciences in Chicago, IL.
Kingery, P. M., Pruitt, B. E., & Heuberger, G. (1996). A profile of rural
Texas adolescents who carry handguns to school. Journal of School Health,
66,18−22.
Kleck, G. (1991). Point blank: Guns and violence in America. New York: Aldine
de Gruyer.
Kovandzic, T. V., & Marvell, T. B. (2003). Right-to-carry concealed handguns
and violent crime: Crime control through gun decontrol? Criminology and
Public Policy,2(3), 363−396.
Kulig, J., Valentine, J., Griffith, J., & Ruthazer, R. (1998). Predictive model of
weapon carrying among urban high school students: Results and validation.
Journal of Adolescent Health,22, 313−319.
Lizotte, A. J., Krohn, M. D., Howell, J. C., Tobin, K., & Howard, G. J. (2000).
Factors influencing gun carrying among young urban males over the
adolescent-young adult life course. Criminology,38,811−834.
Lizotte, A. J., Tesoriero, J. M., Thornberry, T. P., & Krohn, M. D. (1994).
Patterns of adolescent firearms ownership and use. Justice Quarterly,11,
51−73.
Loftin, C. (1986). Assaultive violence as a contagious social process. Bulletin of
the New York Academy of Medicine,62(5), 550−555.
Long, J. S. (1997). Regression models for categorical and limited dependent
variables. Thousand Oaks, CA: Sage.
Malhotra, N. K. (1987). Analyzing marketing research data with incomplete
information on the dependent variables. Journal of Marketing Research,24,
74−84.
Marvell, T. B. (2001). The impact of banning juvenile gun possession. Journal
of Law and Economics,44, 691−714.
McDowall, D., & Loftin, C. (1983). Collective security and demand for legal
handguns. American Journal of Sociology,88, 1146−1161.
Perkins, C. (2003). Weapon use and violent crime (NCJ 194820). Washington,
DC: U.S. Department of Justice.
Peters, K. D., Kochanek, M. A., & Murphy, S. (1998). Deaths: Final data for
1996 (National Vital Statistics Rep. 47-9). Atlanta, GA: Centers for Disease
Control and Prevention, Division of Vital Statistics.
Price, J. H., Desmond, S. M., & Smith, D. (1991). A preliminary investigation of
inner-city adolescents' perceptions of guns. Journal of School Health,61,
255−259.
Raymonds, M. R., & Roberts, D. M. (1987). A comparison of methods for
treating incomplete data in selection research. Education and Psychological
Measurement,47,13−26.
Rountree, P. W. (2000). Weapons at school: Are the predictors generalizable
across context? Sociological Spectrum,20, 291−324.
Notes
1. The total percentage exceeds 100 percent because the item asked
students to “mark all that apply”.
2. The data also included two other variables that were indicative of a
student's bond to school: students' academic performance measured by their
GPA and their expectations after high school graduation. Neither of them was
statistically significant. For the reason of parsimony, the results were not re-
ported in the comprehensive model.
3. There were seven other items concerning security measures of schools,
such as supervising the hallways, metal detectors, locking entrance and exit
doors during the day, requiring visitors to sign in, locker checks, wearing
student badges or picture IDs, and having security cameras. None of them,
singly or in combination, was significantly related to carrying weapons to
school. Their results were not reported in the analysis.
4. The percentage and the frequency of gun carrying teenagers in schools
were reported based on the filtered sample (i.e., n = 7,391). Although the
percentage would remain the same, the raw frequency counted, after sampling
weights were applied, would result in a much larger number of gun carrying
occurrences nationwide. He and Swatt (2005) suggested that coefficient
estimates and statistical conclusions in multivariate analysis do not vary
significantly when comparing weighted and unweighted NCVS data. For the
reason of parsimony, the unweighted data were used in the analysis. The
relatively small frequency and percentage of the gun-carrying category may be
a concern. To address this concern, a subsample of 500 non-weapon
respondents was drawn and analyzed (Hair, Black, Babin, Anderson, &
Tatham, 2006). The results were consistent with the results reported here.
163L. Cao et al. / Journal of Criminal Justice 36 (2008) 154–164

Ruddell, R., & Mays, G. L. (2003). Examining the arsenal of juvenile
gunslingers: Trends and policy implications. Crime and Delinquency,49,
231−252.
Sheley, J. F., & Brewer, V. E. (1995). Possession and carrying of firearms among
suburban youth. Public Health Reports,110 ,18−26.
Sheley, J. F., & Wright, J. D. (1995). In the line of fire: Youth, guns, and violence
in urban America. New York: Aldine de Gruyter.
Simon, T. R., Crosby, A. E., & Dahlberg, L. L. (1999). Students who carry
weapons to high school: Comparison with other weapon carriers. Journal of
Adolescent Health,24, 340−348.
Smith, M. D. (1996). Sources of firearm acquisition among a sample of inner-
city youths: Research results and policy implications. Journal of Criminal
Justice,24, 361−367.
U.S. Department of Education and the U.S. Department of Justice. (1999). 1999
annual report on school safety. Washington, DC: U.S. Government Printing
Office.
U.S. Department of Education and the U.S. Department of Justice. (2001).
National crime victimization survey: School crime supplement [Data file].
Washington, DC: Author.
Webster, D. W., Gainer, P. S., & Champion, H. R.(1993). Weapon carrying among
inner-city junior high school students: Defensive behavior vs. aggressive de-
linquency. American Journal of Public Health,83,1604−1608.
Wilcox, P., & Clayton, R. R. (2001). A multilevel analysis of school-based weap-
on possession. Justice Quarterly,18, 509−539.
Wright, J. D., Rossi, P., & Daly, K. (1983). Under the gun: Weapons, crime and
violence in America. New York: Aldine.
Zimring, F., & Hawkins, G. (1997). Crime is not a problem: Lethal violence in
America. New York: Oxford University Press.
164 L. Cao et al. / Journal of Criminal Justice 36 (2008) 154–164