Am J Crim Just (2019) 44:146–165
https://doi.org/10.1007/s12103-018-9438-6
The Kids Aren’t Alright: School Attachment, Depressive
Symptoms, and Gun Carrying at School
Stephen J. Watts 1 & Karli Province 1 & Kayla Toohy 1
Received: 10 January 2018 / Accepted: 21 March 2018 /
Published online: 5 April 2018
# Southern Criminal Justice Association 2018
Abstract Social science has frequently examined the relationships between school
environment and delinquency, mental health and delinquency, and school environment
and mental health. However, little to no research to date has examined the interrelationship between these variables simultaneously, especially at it relates specifically to
delinquent acts committed at school. The current study uses data from the National
Longitudinal Study of Adolescent to Adult Health (Add Health) to look at the
interrelationship between these variables. What is found in this data is that the
relationship between negative mental health states and delinquency at school, specifically measured as depressive symptoms and gun carrying at school, respectively, is
possibly a spurious one, wherein both of these variables are partly shaped by school
attachment, which accounts for their correlation. Implications for theory and policy are
discussed.
Keywords School violence . Guns . Depression . School attachment . Add health
Introduction
School violence, including bullying, fights, assaults, and mass shootings, is a major
concern in America. This is despite the fact that our schools are relatively safe (Kann
et al., 2016), with fear of school violence, in particular fear of school shootings, largely
* Stephen J. Watts
sjwatts@memphis.edu
Karli Province
klprvnce@memphis.edu
Kayla Toohy
krtoohy@memphis.edu
1
Department of Criminology and Criminal Justice, University of Memphis, 311 McCord Hall,
Memphis, TN 38152, USA
Am J Crim Just (2019) 44:146–165
147
being driven by sensationalized coverage of a few events (Cornell, 2006). But any level
of school violence is problematic because it can lead to, beyond the obvious negatives
like injury and death, a negative learning environment for students. The research that
has been done to understand school violence has largely focused on factors such as the
parent-child relationship, substance use, having a history of violence, and bullying
(Brockenbrough, Cornell, & Loper, 2002; Cornell, 2006; Lipsey & Derzon, 1998;
Resnick, Ireland, & Borowsky, 2004).
Children and adolescents carry guns to school for any number of reasons,
including fear of victimization and peer pressure (see Beardslee, Docherty,
Mulvey, Schubert, & Pardini, 2017; Esselmont, 2014; May, 1999; Melde,
Esbensen, & Taylor, 2009; van Geel, Vedder, & Tanilon, 2014; Wilcox, May, &
Roberts, 2006), but when school shootings in particular have been studied, they
have generally been looked at from a mental health standpoint (Borum, Cornell,
Modzeleski, & Jimerson, 2010). When high-profile school shootings specifically are
examined, there is often at the center an offender with some sort of mental health issue,
whether that issue is short-term or long-term. But perhaps this excessive focus on
mental health has left the full picture of why school shootings occur unclear. There
could be factors related specifically to the school environment that increase negative
emotionality and risky behaviors in school. The current study will focus on one such
risky behavior that can serve as a precursor to school violence or directly precipitate
school violence, carrying a gun at school. Specifically, we examine the relationship
between school attachment, depressive symptoms, and gun carrying at school in a
nationally representative, panel study of American adolescents to see if an adolescent’s
positive or negative identification with their school can partly shape depressive symptoms and gun carrying at school, and whether these associations can potentially reveal
the relationship between depressive symptoms and gun carrying to be spurious. By
spurious, we mean that the presumed causal relationship between depressive
symptoms and gun carrying is actually a mere correlation, a correlation driven
by the two variables jointly having a relationship with and partly being shaped by
school attachment. In the following section we will review the literature on mental
health and school violence, and identify how school environment, and in particular
school attachment, fits into this relationship.
Literature Review
Mental Health and School Violence
School violence has many sources, as do risky behaviors that can precede school
violence, like gun carrying. The literature on school violence and risky behaviors has
focused on a myriad of factors as potential causes, including substance use, a history of
violence, poor socioeconomic background, bullying, and truancy (Bailey, Flewelling,
& Rosenbaum, 1997; Brockenbrough et al., 2002; Cornell, 2006; DuRant, Krowchuk,
Kreiter, Sinal, & Woods, 1999; Esselmont, 2014; Muschert, 2007; Shetgiri, Boots, Lin,
& Cheng, 2016; Simon, Crosby, & Dahlberg, 1999; van Geel et al., 2014; Webster,
Gainer, & Champion, 1993). But much of the research on school violence and risky
behaviors that can precede school violence, gun carrying in particular, has additionally
focused on a broader topic of concern within violence in general and gun violence in
148
Am J Crim Just (2019) 44:146–165
particular, mental health. Much of this literature has pointed to depression, anger,
anxiety, and fear as risk factors for both school violence and risky behaviors like gun
carrying (DuRant et al., 1999; Juan & Hemenway, 2017; Melde et al., 2009; Muschert,
2007; Shetgiri et al., 2016; Simon et al., 1999; Webster et al., 1993; Wilcox et al.,
2006), and the literature on general strain theory in particular has focused on stress and
its link to school violence (Agnew, Brezina, Wright, & Cullen, 2002; Hinduja &
Patchin, 2007; Patchin & Hinduja, 2011).1
The research that has focused specifically on those school shootings that have
received widespread media coverage is very clear in its focus on mental health
(Harter, Low, & Whitesell, 2003; Juan & Hemenway, 2017; Meloy, Hempel,
Mohandie, Shiva, & Gray, 2001; Muschert, 2007). With the widespread availability
of guns in the United States considered a constant risk factor, there has been a strong
focus on mental health issues as an important individual-level predictor of perpetrating
a school shooting. Even when other factors, like romantic rejection (Klein, 2005),
bullying victimization (Burgess, Garbarino, & Carlson, 2006; Harter et al., 2003; Leary,
Kowalski, Smith, & Phillips, 2003; Meloy et al., 2001), and a troubled home life
(Webber, 2003), are considered, these factors usually end up mattering because they
cause the shooter mental distress, usually in the form of depression and/or anger.
If we step back from school shootings in particular, and focus instead on school
violence and risky behaviors in general, and consider negative emotions as an important cause of these issues, then that leads us to question, where do these negative
emotions come from that directly relate to school violence and risky behaviors? A
logical answer seems to be that negative emotions that directly relate to school violence
and risky behaviors at school should trace back in part to stressors related to school.
And indeed, this can be seen in the prior research on school violence and its focus on
peer rejection and bullying victimization and, at the macro-level, school-level factors
like the quality of faculty-student relationships (Moore, Petrie, Bragaand, &
McLaughlin, 2003) and the ability of administrators to enforce rules and respond to
threats (Fox & Harding, 2005). All of these factors taken together should shape an
individual’s level of attachment to and positive identification with their school, their
peers, their teachers, and school rules. In the next section, we turn to a discussion of
school attachment, and its potential role in generating both negative emotions and
school violence and risky behaviors like gun carrying among adolescents.
School Attachment, Mental Health, and School Violence and Risky Behaviors
Strong social bonds to important social institutions, like family and school, have long
been seen as important deterrents to delinquency and crime, among adolescents and
adults, in the criminological literature (Akers & Sellers, 2016). Most of this is due to the
development of social bonding theory by Travis Hirschi (1969) and its subsequent
popularity. Social bonding theory has been frequently tested and widely endorsed in the
criminological literature (Akers & Sellers, 2016). Hirschi’s theory argues that most
1
While this study will invoke many of the concepts found in general strain theory, we do not use general strain
theory as a theoretical guide, because rather than focusing on a lack of school attachment as a strain that
increases negative emotions which then result in deviance (i.e., the mediating relationship general strain theory
proposes), we instead focus on and find that the negative emotions-deviance relationship is a spurious one,
with school attachment as the key confounding variable.
Am J Crim Just (2019) 44:146–165
149
individuals do not offend because of the social controls placed on us as individuals in
the form of social bonds. A social bond is a force within one’s environment that
connects them to society’s moral constraints (Winfree Jr & Abadinsky, 2009).
Through these social bonds, we are attached to conventional individuals, committed
to societal norms, involved in conventional activities, and we come to believe in the
moral force behind society’s norms. Social bonding theory thus explains crime and
delinquency as the result of an individual’s social bonds being weak or nonexistent.
Based on this theory, it then makes sense that school attachment, the social bond to
one’s school, would be important in the etiology of crime and delinquency, especially
crime and delinquency at school. And indeed, this is exactly what the literature shows,
with measures of school attachment relating to various types of youth offending across
a wealth of studies in the criminology and education literatures (e.g., Bond et al., 2007;
Cernkovich & Giordano, 1992; Chapple, Tyler, & Bersani, 2005; Dornbusch, Erickson,
Laird, & Wong, 2001; Henrich, Brookmeyer, & Shahar, 2005; Henry & Slater, 2007;
Herrenkohl, Huang, Tajima, & Whitney, 2003; Hirschfield & Gasper, 2011; Longshore,
Chang, & Messina, 2005; Payne, 2009; Resnick, Harris, & Blum, 1993; Watts, 2017;
Wilson, 2004; Zhang & Messner, 1996). As an example, Henry and Slater (2007)
found that overall levels of school attachment within a school decrease individual-level
adolescent alcohol use behaviors.
Most studies in the criminological literature focus on school attachment and
delinquent behavior in general (e.g., Cernkovich & Giordano, 1992; Longshore et al.,
2005). Less often seen in this literature are studies that look specifically at the link
between school attachment and criminal and delinquent behaviors committed in the
school setting. Exceptions include a study by Jenkins (1997), who found in a sample of
7th and 8th graders that school attachment decreased school crime, misconduct, and
nonattendance, and a study by Hirschfield and Gasper (2011), who found in a sample
of inner-city Chicago elementary school students that elements of school engagement
decreased delinquency in school.
So the research shows a clear link between school attachment and crime and
delinquency among adolescents, even if the research on crime and delinquency at
school specifically is somewhat lacking. This leaves the issue of the potential
relationship between school attachment and mental health. To date, a number of studies
in the education literature have examined the relationship between school attachment,
referred to in this literature as Bschool connectedness,^ and mental health among
adolescents. These studies have consistently found that school attachment has a
strong relationship with mental health among adolescents, with low levels of
school attachment serving as a risk factor for negative mental health states, such
as experiencing depression and anxiety (Bond et al., 2007; Lester, Waters, &
Cross, 2013; Romo & Nadeem, 2007; Shochet, Dadds, Ham, & Montague,
2006). For example, Shochet and colleagues (2006) found in a sample of
Australian youths that school attachment (Bconnectedness^) predicted depressive
symptoms a year later for boys and girls, even when controlling for prior
depressive symptoms. Of note, they found that the reverse was not true, with
depressive symptoms not predicting school attachment a year later when
controlling for prior school attachment. Bond et al. (2007) found that low school
attachment, even in the presence of good overall social connectedness, increased
risks for depressive symptoms and anxiety in a 2–4 year follow up.
150
Am J Crim Just (2019) 44:146–165
So prior literature has tied school attachment to both crime and delinquency and
mental health among adolescents. However, these issues have largely been examined
separately. With few exceptions, research to date has not considered the interrelationship between school attachment, mental health, and crime and delinquency among
adolescents. One recent study, conducted by Juan and Hemenway (2017), considered
all of these factors together, with the authors taking the view that negative mental health
states, specifically depression, have primacy in how these variables connect. They
argued that depression has its effects on crime and delinquency, specifically measured
as carrying a gun at school, partly through depression’s effect on Bsocial
connectedness,^ a measure that included aspects of the relationship with one’s peers.
They argue for a mediating model whereby depression undermines social connectedness, which can then lead to gun carrying, and thus, social connectedness can act as a
Bbuffer^ between depression and acts of deviance at school like gun carrying.
However, the prior research cited above would argue just as strongly that school
attachment would hold primacy, especially when the focus is on crime and delinquency
in the school setting. Putting together the prior literature on school attachment, mental
health, and youth crime and delinquency in the school setting, an argument can be
made that the relationship between negative mental health states and offending in
schools could potentially be spurious. It may be that school attachment precedes and
partly shapes both mental health and deviance in school, and that the relationship
between them is driven by their joint relationship with school attachment. This is the
possibility that the current study will seek to assess.
The Current Study
This paper seeks to build on the existing literature on school attachment, mental health,
and crime and delinquency among adolescents. This will be done by examining the
relationship between school attachment, depressive symptoms, and a specific act of
deviance at school, carrying a gun, in a longitudinal probability sample of adolescents.
In so doing, we will expand on the literature in several important ways. First, this is one
of few studies to consider school attachment and focus specifically on delinquency at
school. Second, this is one of very few studies to consider school attachment, mental
health, and crime and delinquency simultaneously. Lastly, this will be the first study to
assess whether the relationship between negative mental health states and offending
among adolescents is in part driven by school attachment. This possibility has been
hinted at in the various literatures cited, but has to date not been examined empirically
with adequate data.
This study draws on prior literature, particularly expectations from social bonding
theory and findings concerning school connectedness and depression in the education
literature, to derive four hypotheses concerning the relationship between school attachment, depressive symptoms, and gun carrying at school. Hypothesis 1 predicts that
school attachment will be associated with depressive symptoms such that higher levels
of school attachment will mean lower levels of depression. Hypothesis 2 predicts that
without considering school attachment, depressive symptoms will be associated with
gun carrying at school such that more depressive symptoms will correspond with a
higher likelihood of carrying a gun to school. Hypothesis 3 predicts that school
attachment will be associated with gun carrying at school such that higher levels of
Am J Crim Just (2019) 44:146–165
151
school attachment will correspond with a lower likelihood of carrying a gun to school.
Finally, Hypothesis 4 predicts that controlling for school attachment will reduce the size
and significance of the depressive symptoms-gun carrying at school relationship. Thus,
the depressive symptoms-gun carrying relationship will ultimately be spurious, a mere
correlation partly driven by their joint relationship with school attachment. Fig. 1
presents the proposed empirical model. With this proposed empirical model, the
authors are not trying to suggest that school attachment is the only explanation for
depressive symptoms and/or gun carrying among adolescents, but rather, school
attachment can partly explain both, and in so doing, explain the correlation between
depressive symptoms and gun carrying at school.
Methods
Data
The National Longitudinal Study of Adolescent to Adult Health (Add Health) is a
longitudinal panel study of a nationally representative, school-based probability sample
of youth who were in grades 7–12 in the United States beginning in 1994–95 (Udry,
2003). Add Health employed cluster and systematic sampling methods to select 80
public and private high schools and 52 middle schools to be representative of U.S.
regions, urban composition, school size and type, and ethnicity (Harris et al., 2003).
Add Health first collected data with an in-school survey, then, for a subsample,
followed up with a series of in-home interviews conducted in 1994–95 (Wave I),
1996 (Wave II), 2001–02 (Wave III), and 2007–08 (Wave IV). The current study
utilized data from the restricted data set, which is identical to the public-use data set
but contains the full sample of respondents. Data from respondents who completed
both the Wave I & II in-home surveys, and for whom sampling weights were available,
was utilized (N = 13,568). Table 1 provides descriptive statistics for the study variables.
Wave II
(past 30 days)
Wave I
(past 7 days)
Depressive
Symptoms
Gun Carrying at
School
School Attachment
Wave I
(current/most recent school year)
Fig. 1 Empirical model
152
Am J Crim Just (2019) 44:146–165
Table 1 Descriptive statistics
Variables
Range
Full Sample
(N = 13,568)
Mean (SE)
Dependent variable
Gun Carrying at School Wave II
0/1
.01 (.00)
Depressive Symptoms Wave I
0–51
10.46 (.06)
School Attachment Wave I
6–38
27.64 (.04)
Independent variables
Controls
Male
0/1
.49 (.00)
White
0/1
.52 (.00)
Black
0/1
.20 (.00)
Native American
0/1
.02 (.00)
Asian
0/1
.07 (.00)
Other
0/1
.02 (.00)
Hispanic
0/1
.17 (.00)
Age Wave II
11–23
16.23 (.01)
Public Assistance Wave I
0/1
.10 (.00)
Maternal Warmth
1–5
4.37 (.01)
Maternal Supervision
3–15
5.83 (.02)
Out of School Suspension Wave II
0/1
.10 (.00)
English Grave Wave II
1–4
2.16 (.01)
Gang Member Wave II
0/1
.05 (.00)
Peer Delinquency Wave I
0–9
2.42 (.02)
Thoutfully Reflective Decision Making
4–20
8.81 (.02)
Self-control
1–5
2.99 (.01)
Fighting in School Wave II
0/1
.06 (.00)
Gun easily accessible in home Wave II
0/1
.16 (.00)
Because these statistics are weighted and adjusted for survey design, standard errors are produced rather than
standard deviations. 86 respondents reported that a gun was the weapon they carried to school most often in
the 30 days prior to the WII interview
Measures
Dependent Variable: Gun Carrying at School Wave II
The dependent variable is a single dichotomous measure from Wave II that measures
gun carrying at school. While school shootings were often mentioned in the literature
review, school shootings are an exceedingly rare phenomenon, and Add Health does
not contain measures on using a gun at school. Additionally, Add Health contains
limited measures about frequencies of school-specific violent or delinquent behavior,
with most measures of delinquent and criminal behaviors in the dataset not specifying
Am J Crim Just (2019) 44:146–165
153
the domain (i.e., at home, at school, in the neighborhood) in which the behavior
occurred.
Respondents were first asked if they had carried a weapon at school since the Wave I
interview. For those respondents that answered affirmatively, a follow-up question asked
what one kind of weapon they carried to school most often in the past 30 days. Possible
answers included a handgun, some other type of gun (rifle, shotgun, etc.), a club or bat, a
knife or razor, or something else. We used this question to create a dichotomous measure
of whether someone reported that the weapon they carried most often to school in the
past 30 days was a gun of any kind (1 = yes). While this measure would be better if it
focused on gun carrying in general rather than as a preferred weapon, the time frame
covered does add confidence in the causal ordering of the proposed empirical model.
Eighty-six respondents reported that a gun was the weapon they carried to school most
often in the 30 days prior to the Wave II interview.2
Independent Variables
Depressive Symptoms Wave I To assess depressive symptomatology we drew on 18
questions from Wave I based on the Centers for Epidemiological Study-Depression
Scale (CES-D) (Radloff, 1977, 1991), that we summed in an index where higher scores
denote higher levels of depressive symptomatology (range = 0 to 51, alpha = .86).
These items asked about feelings of loneliness, fear, sadness, fearfulness, etc. in the
7 days prior to the Wave I interview, with responses ranging from never or rarely (0) to
most of the time or all of the time (3). The purpose of this construct in the current study
is not to assess depressive symptoms based on criteria for diagnosis of major depression, but rather, to assess the level of depressive symptomatology among Add Health
respondents along a spectrum. Past research has similarly measured depressive symptomatology in Add Health (Goodman, Slap, & Huang, 2003; Hallfors et al., 2004;
Hallfors, Waller, Bauer, Ford, & Halpern, 2005; Iratzoqui, 2015; Kaufman, 2009;
Steuber & Danner, 2006; Watts & McNulty, 2013). The full list of items that make
up the depressive symptoms measure can be found in the appendix.
In focusing on depressive symptoms in the relationship between mental and gun
carrying, we are not seeking to discount the importance of other emotions, such as
anger, hostility, and/or fear. We are simply limited by what measures are available in the
Add Health, and there are not valid and complete measures of these other emotions in
the early waves of data collection.3
School Attachment Wave I The social bond to and positive identification with one’s
school is represented by an 8 item additive index of questions drawn from Wave I that
2
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 Bsvyset^ 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 Bsvyset^
command, which is not advised by Add Health, and the Bfirthlogit^ command is used to run a type of rare
events logit estimation, results that are substantively identical to those presented in this paper are produced.
These results are available upon request.
3
For example, while later waves of data have better questions concerning anger, the only one available in
Waves I and II comes from the parent interview in Wave I, where the target respondent’s parent was asked,
BDoes your child have a bad temper?^ Responses were simply recorded as Byes^ or Bno.^
154
Am J Crim Just (2019) 44:146–165
measure a respondent’s sense of belonging to their school where higher scores denote
higher levels of school attachment (range = 6 to 38, alpha = .74). Respondents were
asked about whether during the current or most recent school year they had had trouble
getting along with teachers or fellow students, if they felt close to people at school, felt
like a part of their school, believed their fellow students were prejudiced, were happy to
be at their school, whether teachers treated students fairly, and if they felt safe at their
school. Taken together, these items assess how attached and committed a respondent
was to their school, and how much they felt they belonged to and were safe at their
school during the current or most recent school year at Wave I. Past research has
similarly measured the concept of school attachment in Add Health (Johnson, Crosnoe,
& Thaden, 2006; McNeely & Falci, 2004; McNeely, Nonnemaker, & Blum, 2002;
Pearson, Muller, & Wilkinson, 2007; Watts, 2017). The full list of items that make up
the school attachment measure can be found in the appendix.
Controls
Numerous controls were considered to check the robustness of any observed relationships of central concern, and to help avoid model misspecification and omitted variable
bias (Mustard, 2003). General controls included Male, dummy variables for Hispanic,
non-Hispanic Black, Native American, Asian, and Other (with non-Hispanic White as
the reference category), Age (Wave II),4 and one measure of socioeconomic status,
Public Assistance (Wave I).5 The measure of public assistance indicates whether the
respondent’s family was receiving any sort of public assistance, such as welfare or food
stamps, at Wave I (1 = yes).
In addition to the more general controls, a number of controls were included because
they may relate simultaneously to school attachment, negative mental health states, and
general delinquent behavior. Controlling for these variables is an important test of the
robustness of any relationships uncovered between school attachment, depressive
symptoms, and gun carrying in schools. For this reason, we included two measures
of parenting, Maternal Warmth and Maternal Supervision. Maternal Warmth is a single
item that asked respondents at Wave I if their mother was warm and loving towards
them most of the time, with responses ranging from 1 (strongly agree) to 5
(strongly disagree). Maternal Supervision consists of three items from Wave I
that asked respondents how often their mother was at home when they left for
school, returned from school, and went to bed, with higher scores denoting less
supervision.
We also included two school-related controls. Out of School Suspension is a single
item from Wave II that asked if during the current or most recent school year the
respondent had received an out of school suspension (1 = yes). English Grade also
4
While there is a large age range in Add Health, the vast majority (88.6%) of respondents were between 14
and 18 years old at Wave II.
5
Regression models run with additional SES controls for parental education, parental employment, and total
family income produce identical results for the key variables of interest. With these SES measures infrequently
showing statistical significance and having little impact on the results, a more parsimonious model with only
the public assistance variable is presented. Results with all of these SES measures included are available upon
request.
Am J Crim Just (2019) 44:146–165
155
comes from Wave II, consisting of one item where respondents were asked what their
most recent grade was in English/language arts, with responses ranging from 1 (A) to 4
(D or lower). We also included two measures of delinquent peer affiliations. Gang
Member comes from Wave II, and consists of one item that asked if respondents had
ever been initiated into a named gang (1 = yes). Peer Delinquency comes from Wave I,
and consists of three items that asked respondents how many of their three closest
friends smoke at least one cigarette per day, drink alcohol at least once a month, and
smoke marijuana at least once a month. This measure of peer delinquency has been
shown in past studies using Add Health data to have predictive validity (Beaver,
Wright, & DeLisi, 2008; Watts & McNulty, 2015).
We additionally controlled for two conceptually similar variables related to impulse
control and the use of reason. Thoughtfully Reflective Decision Making consists of four
items from Wave I that recorded responses to four statements: When you have a problem
to solve, one of the first things you do is get as many facts about the problem as possible;
When you are attempting to find a solution to a problem, you usually try to think of as
many different approaches to the problem as possible; When making decisions, you
generally use a systematic method for judging and comparing alternatives; After
carrying out a solution to a problem, you usually try to analyze what went right and
what went wrong. Possible responses ranged from 1 (strongly agree) to 5 (strongly
disasgree). Higher scores denote less reflective decision making (Paternoster &
Pogarsky, 2009). Self-Control is measured in consideration of Hirschi’s (2004) redefinition of the concept as the tendency to examine the full range of outcomes from one’s
behavior. As such, this measure consists of one item from Wave I wherein respondents
were asked whether when making decisions they usually go with their Bgut feeling^
without thinking too much about the consequences of each alternative, with responses
ranging from 1 (strongly agree) to 5 (strongly disagree) (Paternoster & Pogarsky, 2009).
Finally, we included measures concerning violence and gun access. The variable
Fighting in School measured at Wave II whether the most recent fight a respondent got
into occurred at school (1 = yes) or somewhere else. We included this measure because
violent behaviors like fighting tend to correlate highly with risky behaviors like carrying
a weapon to school (Forrest, Zychowski, Stuhldreher, & Ryan, 2000; Lowry, Powell,
Kann, Collins, & Kolbe, 1998; Pickett et al., 2005; Resnick et al., 1997; Resnick et al.,
2004). The variable Gun Easily Accessible in Home utilizes a measure from Wave II
where respondents were asked if a gun is easily available to them in their home (1 = yes).
Studies have linked access to guns to youth violence (Blumstein & Cork, 1996;
Kennedy, Piehl, & Braga, 1996; Schubiner, Scott, & Tzelepis, 1993).6
Analytic Strategy
We tested our hypotheses utilizing ordinary least squares (OLS) regression (depression)
and logistic regression (gun carrying at school).7 The models presented test whether
school attachment at Wave I (current/most recent school year) significantly associates
6
While fear of victimization has also consistently been tied to gun carrying among adolescents (see Melde
et al., 2009; Wilcox et al., 2006), measures assessing fear of victimization are not available in Add Health.
7
For these OLS models we computed the ln + 1 of the depression variable to smooth out the distribution. For
this logged variable, skewness = −.81 and kurtosis = .77.
156
Am J Crim Just (2019) 44:146–165
with depressive symptoms at Wave I (past 7 days) (Hypothesis 1), whether depressive
symptoms significantly associate with gun carrying at school at Wave II when school
attachment is not considered (Hypothesis 2), whether school attachment at Wave I
significantly associates with gun carrying at school at Wave II (Hypothesis 3), and
whether controlling for school attachment reduces to insignificance the depressive
symptoms-gun carrying at school relationship (Hypothesis 4). The appropriate weight,
cluster, and strata variables were utilized in all analyses to account for the complex
sampling design of Add Health.8 Tests utilizing Variance Inflation Factors (VIFs) show
that multicollinearity is not a concern in regards to the independent variables in the
OLS or logistic regression models.
Results
Table 2 presents the results of OLS models with depressive symptoms at Wave I
regressed on school attachment at Wave I and the controls.9 To smooth out the
distribution of the depressive symptoms variable, we computed the ln + 1 of this
variable (skewness = −.81) for use in these models. Model 1 in Table 2 shows school
attachment at Wave I has a highly significant, negative association with depressive
symptoms at Wave I, supporting Hypothesis 1. In model 2, the controls are included to
check the robustness of this effect, and the coefficient for school attachment is still
highly significant, and only modestly reduced. Among the controls, males report less
depressive symptoms than females, black, Native American, and Hispanic respondents
report more depressive symptoms than non-Hispanic whites, older respondents report
more depressive symptoms, and those whose families were receiving public assistance
report more depressive symptoms. Also, less maternal warmth means more depressive
symptoms, those who had been suspended from school and those who had worse
grades report more depressive symptoms, more peer delinquency means more depressive symptoms, and less reflective decision making and less self-control mean more
depressive symptoms.
Table 3 presents the results of logistic regression models with odds ratios with gun
carrying at school at Wave II regressed on depressive symptoms at Wave I, school
attachment at Wave I, and the controls.10 To ease the interpretation of the log odds, we
z-score transformed the depressive symptoms and school attachment measures for use
in these models. Model 1 of Table 3 includes only depressive symptoms at Wave I,
which has a significant, positive association with gun carrying at school at Wave II,
8
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
Bsvyset^ command.
9
Stata 13.1 will not produce standardized (beta) coefficients for regression models that have been Bsvyset.^
This is for statistical reasons. Stata 13.1 and other programs are only willing to consider variance decomposition when data are understood to be independent and identically distributed (IID). Using the Bsvyset^
command is specifically meant to correct for the data not being IID, therefore Stata 13.1 is unwilling to
consider ratios of variance in these models.
10
When data are Bsvyset^ 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 Bfitstat^ command after running a logistic
regression model.
Am J Crim Just (2019) 44:146–165
157
Table 2 Depressive symptoms Wave I (ln + 1) regressed on attachment Wave I and controls (N = 13,568)
Variable
Model 1
Model 2
Coef. (SE)
Coef. (SE)
−.05 (.00)***
−.04 (.00)***
Independent variable
School attachment Wave I
Controls
Male
−.18 (.02)***
Black
.15 (.03)***
Native American
.09 (.05)
Asian
.33 (.04)***
Other
.06 (.07)
Hispanic
.20 (.03)***
Age Wave I
.03 (.01)***
Public Assistance Wave I
.07 (.03)*
Maternal Warmth
−.12 (.01)***
Maternal Supervision
.00 (.00)
Out of School Suspension Wave II
.06 (.03)*
English Grave Wave II
.05 (.01)***
Gang Member Wave II
.05 (.04)
Peer Delinquency Wave I
.03 (.00)***
Thoutfully Reflective Decision Making
.01 (.00)**
Self-control
−.07 (.01)***
Fighting in School Wave II
.03 (.03)
Gun easily accessible in home Wave II
−.01 (.03)
Constant
3.66 (.05)***
3.30 (.12)***
R-Sq.
.13
.23
This table includes unstandardized coefficients (linearized SEs) from OLS regression models. Non-Hispanic
White is the reference category for all race/ethnicity variables. *p < .05, **p < .01, ***p < .001
supporting Hypothesis 2. With the depressive symptoms index having been zscore transformed, the odds ratio of 1.42 means that for every standard deviation (one unit) increase in depression, the odds of carrying a gun at school
increase by approximately 42%.
Model 2 seeks to check the robustness of the depressive symptoms effect in the
presence of the control variables. And indeed, even with the controls included,
depressive symptoms still have a robust, significant association with gun carrying at
school, with the log odds having been marginally affected (decreased from 1.42 to
1.35). With all the controls considered, Hypothesis 2 is still supported, and this is with
several of the controls having large effects on gun carrying at school. Among the
controls, males are more likely than females to carry a gun at school, those who identify
as black or other are more likely than non-Hispanic whites to carry a gun at school,
current and/or former gang members are more likely to carry a gun at school, more peer
delinquency means a higher likelihood of carrying a gun at school, and having a gun
easily accessible in one’s home means a higher likelihood of carrying a gun at school.
158
Am J Crim Just (2019) 44:146–165
Table 3 Gun carrying at school wave II regressed on depressive symptoms wave I (z-scored), school
attachment wave I (z-scored), and controls
Variables
Model 1
Model 2
Model 3
Model 4
1.05 (.14)
1.18 (.18)
Independent variables
Depressive symptoms wave I
1.42 (.15)** 1.35 (.20)*
School attachment wave I
.47 (.07)*** .62 (.09)**
Controls
Male
23.95 (14.97)***
Black
2.72 (1.10)*
22.64 (14.32)***
3.10 (1.31)**
Native American
.57 (.47)
.43 (.39)
Asian
1.48 (.95)
1.82 (1.18)
Other
18.05 (9.06)***
15.53 (7.58)***
Hispanic
2.21 (1.09)
2.62 (1.26)*
Age Wave II
.88 (.11)
.87 (.10)
Public Assistance Wave I
1.45 (.61)
1.43 (.64)
Maternal Warmth
.79 (.14)
.84 (.15)
Maternal Supervision
1.03 (.07)
1.04 (.08)
Out of School Suspension Wave II
2.04 (.79)
2.06 (.77)
English Grave Wave II
.84 (.18)
.80 (.18)
Gang Member Wave II
8.14 (3.09)***
6.80 (2.56)***
Peer Delinquency Wave I
1.17 (.09)*
1.14 (.09)
Thoutfully Reflective Decision Making
.97 (.07)
.95 (.06)
Self-control
.80 (.13)
.82 (.13)
Fighting in School Wave II
.90 (.50)
.88 (.46)
Gun easily accessible in home wave II
3.03 (.99)**
2.91 (.99)**
Constant
.01 (.00)*** .00 (.01)*
.01 (.00)*** .00 (.01)*
Mckelvey and Zavoina’s R-Sq.
.21
.56
.90
.90
This table includes odds ratios (linearized SEs) from logistics regression models. Non-Hispanic White is the
reference category for all race/ethnicity variables. *p < .05, **p < .01, ***p < .001
Model 3 seeks to directly test Hypothesis 3 and Hypothesis 4, the two of which
together are the final checks whether school attachment reveals the depressive
symptoms-gun carrying relationship to be spurious. To that end, model 3 of Table 3
includes only depressive symptoms and school attachment. As can be seen in model 3
of Table 3, school attachment at Wave I has a highly significant, negative association
with gun carrying at school at Wave II, supporting Hypothesis 3. Meanwhile, the
inclusion of school attachment alone reduces the odds ratio for depressive symptoms
to insignificance (p-value = 0.693), supporting Hypothesis 4. With the school attachment variable having been z-score transformed, the odds ratio of .47 means that a
standard deviation (one unit) increase in school attachment lowers the odds of carrying
a gun at school by approximately 53%. The combination of results found in Tables 2
and 3 support Hypothesis 3 and Hypothesis 4, and the positive relationship
between depressive symptoms and gun carrying at school in the current data
Am J Crim Just (2019) 44:146–165
159
appears to be a positive correlation driven by these variables being mutually
associated with school attachment.
As a final check of the robustness of the school attachment association with gun
carrying at school, model 4 includes all of the study variables. In model 4, depressive
symptoms is once again insignificant, while school attachment is still highly significant,
and has an association that is only modestly reduced (odds ratio reduced to .62 from
.47) from model 3. The controls show similar effects in model 4 as they did in model 2,
with the exceptions that in model 4 the Hispanic coefficient is now significant and the
peer delinquency coefficient is no longer significant.11
Discussion and Conclusion
This paper examined the relationship between school attachment, depressive symptoms, and gun carrying at school among adolescents in a nationally representative
probability sample. Four hypotheses about the nature of the relationships between these
variables were proposed, all of which were supported. School attachment had a
significant association with depressive symptoms, depressive symptoms and school
attachment both had significant associations with gun carrying at school, and accounting for school attachment reduced the depressive symptoms-gun carrying relationship
to insignificance, suggesting they are only related in so far as they are both partially
caused by school attachment.12
These results build upon past research in important ways. This is one of only a few
studies to date to consider school attachment, mental health, and crime and delinquency
at school simultaneously and within the same longitudinal probability sample. In so
doing, this allowed us to think about prior theory and research and propose an empirical
model for how these variables relate. This led us to propose a model, informed by
previous theory and research, whereby school attachment shapes both depression and
gun carrying at school, with the results supporting this model.
The results of the current study have important implications for theory and policy. In
terms of theory, the results reaffirm the utility and importance of Hirschi’s social
bonding theory. School attachment is clearly an important social bond that deters gun
carrying at school among adolescents. Additionally, these results, and those of past
studies, show that school attachment is not only a social bond of importance in regards
to crime and delinquency among adolescents, but also adolescent mental health.
Among school-aged children, it may very well be that as much as relationships with
11
Traditional tests of model fit improvement, such as the likelihood ratio test or Wald test, will not run in
STATA when data is Bsvyset.^ However, if the models are run without the Bsvyset^ command, which is not
recommended in the case of these data, the likelihood ratio test produces a significant chi-square statistic
suggesting model fit improvement when school attachment is considered between models 1 and 3 (chi-sq. =
65.81, prob. > chi-sq. = 0.0000), and models 2 and 4 (chi-sq. = 26, prob. > chi-sq. = 0.0000).
12
Several alternative moderator and mediator models were considered, none of which fit the data as well as
the presented models. School attachment and depressive symptoms did not significantly interact in their
association with gun carrying at school, at Waves I or II. Also, mediating models were not supported. When
controlling for school attachment at Wave I, depressive symptoms, at Waves I and II, do not significantly relate
to gun carrying. Also, controlling for depressive symptoms does nothing to change the magnitude of the
relationship of school attachment, at Waves I or II, with gun carrying. Results are available upon request.
160
Am J Crim Just (2019) 44:146–165
parents and peers, positive or negative identification with one’s school is a driver of
feelings of happiness, sadness, inclusion, exclusion, anger, anxiety, etc.
In terms of policy, the results of the current study make for the possibility of novel
programming to simultaneously deal with multiple issues faced by adolescents.
Obviously, on its face, reducing the availability of guns would likely result in less
gun carrying in schools. But given the controversy surrounding attempts to curb access
to guns, we would rather focus on programming related to a concept at the heart of this
study, school attachment.
Much of our public rhetoric concerning youth violence, gun violence, and school
violence has centered on issues related to mental health. However, this focus may miss
out on an important factor, which is where negative mental health states come from.
The present results suggest that programmatic efforts to strengthen adolescent investment in and positive identification with school could help deal with issues related to
both negative mental health states and school-based deviance and violence. As such,
we would echo some of what prior researchers have called for, namely efforts to
improve classroom management, increase involvement in extracurricular activities,
the creation of fair and tolerant disciplinary practices, smaller class sizes, strong
teacher support for academic goals, positive and respectful adult-student relationships, and emotionally supportive environments for students with minority
statuses (McNeely et al., 2002; Monahan, Oesterle, & Hawkins, 2010). These
efforts to increase school attachment may produce improved mental health,
decreased delinquency, and improved academic performance among adolescents,
thus providing a lot of bang for their buck.
While this study does make important contributions, limitations should be noted. An
important limitation to note concerns the nature of the dependent variable. Very little
gun carrying at school exists in this sample, and this largely relates to how the questions
were asked concerning weapon carrying. In order to identify those that had carried a
gun to school in the past 30 days, they had to have carried a gun more often than
another type of weapon. A more ideal measure would have covered a larger period of
time and asked about gun carrying in general rather than as a preferred weapon. This
would have provided more variability in the dependent variable as well as picked up all
of the gun carriers in the sample. With the way the question is worded now, there may
be individuals in the sample who carried guns to school who were not picked up
because they did not carry a gun most often, instead favoring a knife, etc. Additionally,
future research could look at a similar model that focuses on weapon carrying at
7school more generally, rather than only focusing on gun carrying, and the
operationalization of mental health could be expanded to include emotions like anger,
hostility, and fear.
Another limitation concerns the time overlap in the independent variables. While the
school attachment measures do cover a much larger amount of time than the depression
measures (current/most recent school year vs. past 7 days), more separation between
these measures would provide further confidence in the causal ordering of the school
attachment, depressive symptoms, and gun carrying relationship.
In conclusion, the nature of the relationship between school environment, mental
health, and crime and delinquency among adolescents begs to be studied more. Other
sources of data should be utilized that contain similar and alternate measures of the key
variables in the current study to seek replication of the present findings. Longitudinal
Am J Crim Just (2019) 44:146–165
161
data that covers an earlier part of the life course would be especially desirable for
supporting the causal relationships proposed in this study. With the implications of
these results holding so much potential in terms of public and social policy, the field of
criminology should seek to further assess if these findings hold up with alternate data
sources and other modeling strategies.
Appendix: Items for scaled measures
Depression Wave I
How often was each of the following things true in the past week?
1. You were bothered by things that usually don’t bother you.
2. You didn’t feel like eating, your appetite was poor.
3. You felt that you could not shake off the blues, even with help from your family
and your friends.
4. You felt you were just as good as other people. (reverse coded)
5. You felt depressed.
6. You felt that you were too tired to do things.
7. You felt hopeful about the future. (reverse coded)
8. You thought your life had been a failure.
9. You felt fearful.
10. You were happy. (reverse coded)
11. You talked less than usual.
12. You felt lonely.
13. People were unfriendly to you.
14. You enjoyed life. (reverse coded)
15. You felt sad.
16. You felt that people disliked you.
17. It was hard to get started doing things.
18. You felt life was not worth living.
School Attachment Wave I
During the current/most recent school year, how often have you had trouble…
1.
2.
3.
4.
5.
6.
7.
8.
9.
getting along with your teachers?
getting along with other students?
How much do you agree or disagree with the following statements?
You feel close to people at your school.
You feel like a part of your school.
Students at your school are prejudiced.
You are happy to be at your school.
The teachers at your school treat students fairly.
You feel safe in your school.
162
Am J Crim Just (2019) 44:146–165
References
Agnew, R., Brezina, T., Wright, J. P., & Cullen, F. T. (2002). Strain, personality traits, and delinquency:
Extending general strain theory. Criminology, 40(1), 43–72.
Akers, R. L., & Sellers, C. S. (2016). Criminological theories : Introduction, evaluation, and application (7th
ed.). New York: Oxford University Press.
Bailey, S. L., Flewelling, R. L., & Rosenbaum, D. P. (1997). Characteristics of students who bring weapons to
school. Journal of Adolescent Health, 20(4), 261–270.
Beardslee, J., Docherty, M., Mulvey, E., Schubert, C., & Pardini, D. (2017). Childhood risk factors associated
with adolescent gun carrying among black and white males: An examination of self-protection, social
influence, and antisocial propensity explanations. Law and Human Behavior. https://doi.org/10.1037
/lhb0000270.
Beaver, K. M., Wright, J. P., & DeLisi, M. (2008). Delinquent peer group formation: Evidence of a gene X
environment correlation. The Journal of Genetic Psychology, 169(3), 227–244.
Blumstein, A., & Cork, D. (1996). Linking gun availability to youth gun violence. Law and Contemporary
Problems, 59(1), 5–24.
Bond, L., Butler, H., Thomas, L., Carlin, J., Glover, S., Bowes, G., & Patton, G. (2007). Social and school
connectedness in early secondary school as predictors of late teenage substance use, mental health, and
academic outcomes. Journal of Adolescent Health, 40(4), 357–3e9.
Borum, R., Cornell, D. G., Modzeleski, W., & Jimerson, S. R. (2010). What can be done about school
shootings? A review of the evidence. Educational Researcher, 39(1), 27–37.
Brockenbrough, K. K., Cornell, D. G., & Loper, A. B. (2002). Aggressive attitudes among victims of violence
at school. Education and Treatment of Children, 25(3), 273–287.
Burgess, A. W., Garbarino, C., & Carlson, M. I. (2006). Pathological teasing and bullying turned deadly:
Shooters and suicide. Veictims and Offenders, 1(1), 1–14.
Cernkovich, S. A., & Giordano, P. C. (1992). School bonding, race, and delinquency. Criminology, 30(2),
261–291.
Chapple, C. L., Tyler, K. A., & Bersani, B. E. (2005). Child neglect and adolescent violence: Examining the
effects of self-control and peer rejection. Violence and Victims, 20(1), 39–53.
Cornell, D. (2006). School violence: Fears vs facts. Mahway, NJ: Lawrence Erlbaum.
Dornbusch, S. M., Erickson, K. G., Laird, J., & Wong, C. A. (2001). The relation of family and school
attachment to adolescent deviance in diverse groups and communities. Journal of Adolescent Research,
16(4), 396–422.
DuRant, R. H., Krowchuk, D. P., Kreiter, S., Sinal, S. H., & Woods, C. R. (1999). Weapon carrying on school
property among middle school students. Archives of Pediatrics & Adolescent Medicine, 153(1), 21–26.
Esselmont, C. (2014). Carrying a weapon to school: The roles of bullying victimization and perceived safety.
Deviant behavior, 35(3), 215–232.
Forrest, K. Y., 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(3), 133.
Fox, C., & Harding, D. J. (2005). School shootings as organizational deviance. Sociology of Education, 78(1),
69–97.
van Geel, M., Vedder, P., & Tanilon, J. (2014). Bullying and weapon carrying: A meta-analysis. JAMA
Pediatrics, 168(8), 714–720.
Goodman, E., Slap, G. B., & Huang, B. (2003). The public health impact of socioeconomic status on
adolescent depression and obesity. American Journal of Public Health, 93(11), 1844–1850.
Hallfors, D. D., Waller, M. W., Ford, C. A., Halpern, C. T., Brodish, P. H., & Iritani, B. (2004). Adolescent
depression and suicide risk: Association with sex and drug behavior. American Journal of Preventive
Medicine, 27(3), 224–231.
Hallfors, D. D., Waller, M. W., Bauer, D., Ford, C. A., & Halpern, C. T. (2005). Which comes first in
adolescence—Sex and drugs or depression? American Journal of Preventive Medicine, 29(3), 163–170.
Harris, K. M., Florey, T., Tabor, J., Bearman, P. S., Jones, J., & Udry, J. R. (2003). The National Longitudinal
Study of Adolescent Health: Research design. Retrieved from http://www.cpc.unc.
edu/projects/addhealth/design.
Harter, S., Low, S. M., & Whitesell, N. R. (2003). What have we learned from columbine: The impact of the
self-system on suicidal and violent ideation among adolescents. Journal of School Violence, 2(3), 3–26.
Henrich, C. C., Brookmeyer, K. A., & Shahar, G. (2005). Weapon violence in adolescence: Parent and school
connectedness as protective factors. Journal of Adolescent Health, 37(4), 306–312.
Am J Crim Just (2019) 44:146–165
163
Henry, K. L., & Slater, M. D. (2007). The contextual effect of school attachment on young adolescents’
alcohol use. Journal of School Health, 77(2), 67–74.
Herrenkohl, T. I., Huang, B., Tajima, E. A., & Whitney, S. D. (2003). Examining the link between child abuse
and youth violence: An analysis of mediating mechanisms. Journal of Interpersonal Violence, 18(10),
1189–1208.
Hinduja, S., & Patchin, J. W. (2007). Offline consequences of online victimization: School violence and
delinquency. Journal of School Violence, 6(3), 89–112.
Hirschfield, P. J., & Gasper, J. (2011). The relationship between school engagement and delinquency in late
childhood and early adolescence. Journal of Youth and Adolescence, 40(1), 3–22.
Hirschi, T. (1969). Causes of Delinquency. Berkeley, CA: University of California Press.
Hirschi, T. (2004). Self-control and crime. In R. F. Baumeister & K. D. Vohs (Eds.), Handbook of selfregulation: Research, theory, and applications (pp. 537–552). New York, NY, US: Guilford Press.
Iratzoqui, A. (2015). Strain and opportunity: A theory of repeat victimization. Journal of Interpersonal
Violence, 33, 1–22. https://doi.org/10.1177/0886260515615146.
Jenkins, P. H. (1997). School delinquency and the school social bond. Journal of Research in Crime and
Delinquency, 34(3), 337–367.
Johnson, M. K., Crosnoe, R., & Thaden, L. L. (2006). Gendered patterns in adolescents’ school attachment.
Social Psychology Quarterly, 69(3), 284–295.
Juan, S. C., & Hemenway, D. (2017). From depression to youth school gun carrying in America: Social
connectedness may help break the link. Cogent Social Sciences, 3(1), 1314877.
Kann, L. et al. (2016). Youth risk behavior surveillance—United States, 2015. MMWR. Surveillance
Summaries, 65.
Kaufman, J. M. (2009). Gendered responses to serious strain: The argument for a general strain theory of
deviance. Justice Quarterly, 26(3), 410–444.
Kennedy, D. M., Piehl, A. M., & Braga, A. A. (1996). Youth violence in Boston: Gun markets, serious youth
offenders, and a use-reduction strategy. Law and Contemporary Problems, 59(1), 147–196.
Klein, J. (2005). Teaching her a lesson: Media misses boys’ rage relating to girls in school shootings. Crime,
Media, Culture, 1(1), 90–97.
Leary, M. R., Kowalski, R. M., Smith, L., & Phillips, S. (2003). Teasing, rejection, and violence: Case studies
of the school shootings. Aggressive Behavior, 29(3), 202–214.
Lester, L., Waters, S., & Cross, D. (2013). The relationship between school connectedness and mental health
during the transition to secondary school: A path analysis. Journal of Psychologists and Counsellors in
Schools, 23(2), 157–171.
Lipsey, M. W., & Derzon, J. H. (1998). Predictors of violent and serious delinquency in adolescence and early
adulthood: A synthesis of longitudinal research. In R. Loeber & D. P. Farrington (Eds.), Serious and
violent juvenile offenders: Risk factors and successful interventions (pp. 86–105). Thousand Oaks, CA:
Sage Publications.
Longshore, D., Chang, E., & Messina, N. (2005). Self-control and social bonds: A combined control
perspective on juvenile offending. Journal of Quantitative Criminology, 21(4), 419–437.
Lowry, R., Powell, K. E., Kann, L., Collins, J. L., & Kolbe, L. J. (1998). Weapon-carrying, physical fighting,
and fight-related injury among US adolescents. American Journal of Preventive Medicine, 14(2), 122–
129.
May, D. C. (1999). Scared kids, unattached kids, or peer pressure: Why do students carry firearms to school?
Youth & Society, 31(1), 100–127.
McNeely, C., & Falci, C. (2004). School connectedness and the transition into and out of health-risk behavior
among adolescents: A comparison of social belonging and teacher support. Journal of School Health,
74(7), 284–292.
McNeely, C. A., Nonnemaker, J. M., & Blum, R. W. (2002). Promoting school connectedness: Evidence from
the national longitudinal study of adolescent health. Journal of School Health, 72(4), 138–146.
Melde, C., Esbensen, F. A., & Taylor, T. J. (2009). ‘May piece be with you’: A typological examination of the
fear and victimization hypothesis of adolescent weapon carrying. Justice Quarterly, 26(2), 348–376.
Meloy, J. R., Hempel, A. G., Mohandie, K., Shiva, A. A., & Gray, B. T. (2001). Offender and offense
characteristics of a nonrandom sample of adolescent mass murderers. Journal of the American Academy
of Child & Adolescent Psychiatry, 40(6), 719–728.
Monahan, K. C., Oesterle, S., & Hawkins, J. D. (2010). Predictors and consequences of school connectedness.
The Prevention Researcher, 17(3), 3–6.
Moore, M. H., Petrie, C. V., Bragaand, A. A., & McLaughlin, B. L. (2003). Deadly lessons: Understanding
lethal school violence. Washington, DC: National Academies Press.
Muschert, G. W. (2007). Research in school shootings. Sociology Compass, 1(1), 60–80.
164
Am J Crim Just (2019) 44:146–165
Mustard, D. B. (2003). Reexamining criminal behavior: The importance of omitted variable bias. The Review
of Economics and Statistics, 85(1), 205–211.
Patchin, J. W., & Hinduja, S. (2011). Traditional and nontraditional bullying among youth: A test of general
strain theory. Youth Society, 43(2), 727–751.
Paternoster, R., & Pogarsky, G. (2009). Rational choice, agency and thoughtfully reflective decision making:
The short and long-term consequences of making good choices. Journal of Quantitative Criminology,
25(2), 103–127.
Payne, A. A. (2009). Girls, boys, and schools: Gender differences in the relationships between school-related
factors and student deviance. Criminology, 47(4), 1167–1200.
Pearson, J., Muller, C., & Wilkinson, L. (2007). Adolescent same-sex attraction and academic outcomes: The
role of school attachment and engagement. Social Problems, 54(4), 523–542.
Pickett, W., Craig, W., Harel, Y., Cunningham, J., Simpson, K., Molcho, M., et al. (2005). Cross-national study
of fighting and weapon carrying as determinants of adolescent injury. Pediatrics, 116(6), e855–e863.
Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for research in the general population.
Applied Psychological Measurement, 1(3), 385–401.
Radloff, L. S. (1991). The use of the Center for Epidemiologic Studies Depression Scale in adolescents and
young adults. Journal of Youth and Adolescence, 20(2), 149–166.
Resnick, M. D., Harris, L. J., & Blum, R. W. (1993). The impact of caring and connectedness on adolescent
health and well-being. Journal of Paediatrics and Child Health, 29(s1), S3–S9.
Resnick, M. D., Bearman, P. S., Blum, R. W., Bauman, K. E., Harris, K. M., Jones, J., et al. (1997). Protecting
adolescents from harm: Findings from the National Longitudinal Study on adolescent health. JAMA,
278(10), 823–832.
Resnick, M. D., Ireland, M., & Borowsky, I. (2004). Youth violence perpetration: What protects? What
predicts? Findings from the National Longitudinal Study of adolescent health. Journal of Adolescent
Health, 35(5), 424–4e1.
Romo, L. F., & Nadeem, E. (2007). School connectedness, mental health, and well-being of adolescent
mothers. Theory into Practice, 46(2), 130–137.
Schubiner, H., Scott, R., & Tzelepis, A. (1993). Exposure to violence among inner-city youth. Journal of
Adolescent Health, 14(3), 214–219.
Shetgiri, R., Boots, D. P., Lin, H., & Cheng, T. L. (2016). Predictors of weapon-related behaviors among
African American, Latino, and white youth. The Journal of Pediatrics, 171, 277–282.
Shochet, I. M., Dadds, M. R., Ham, D., & Montague, R. (2006). School connectedness is an underemphasized
parameter in adolescent mental health: Results of a community prediction study. Journal of Clinical Child
& Adolescent Psychology, 35(2), 170–179.
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(5), 340–348.
Steuber, T. L., & Danner, F. (2006). Adolescent smoking and depression: Which comes first? Addictive
Behaviors, 31(1), 133–136.
Udry, J. R. (2003). The National Longitudinal Study of adolescent health (add health). Chapel Hill, NC:
Carolina Population Center, University of North Carolina at Chapel Hill.
Watts, S. J. (2017). The link between child abuse and neglect and delinquency: Examining the mediating role
of social bonds. Victims & Offenders, 12(5), 700–717.
Watts, S. J., & McNulty, T. L. (2013). Childhood abuse and criminal behavior: Testing a general strain theory
model. Journal of Interpersonal Violence, 28(15), 3023–3040.
Watts, S. J., & McNulty, T. L. (2015). Delinquent peers and offending: Integrating social learning and
biosocial theory. Youth Violence and Juvenile Justice, 13(2), 190–206.
Webber, J. A. (2003). Failure to hold: The politics of school violence. New York: Rowman & Littlefield.
Webster, D. W., Gainer, P. S., & Champion, H. R. (1993). Weapon carrying among inner-city junior high
school students: Defensive behavior vs aggressive delinquency. American Journal of Public Health,
83(11), 1604–1608.
Wilcox, P., May, D. C., & Roberts, S. D. (2006). Student weapon possession and the Bfear and victimization
hypothesis^: Unraveling the temporal order. Justice Quarterly, 23(4), 502–529.
Wilson, D. (2004). The interface of school climate and school connectedness and relationships with aggression
and victimization. Journal of School Health, 74(7), 293–299.
Winfree Jr, L. T., & Abadinsky, H. (2009). Understanding Crime: Essentials of Criminological Theory (Third
Edition ed.). Boston, MA: Cengage.
Zhang, L., & Messner, S. F. (1996). School attachment and official delinquency status in the People’s Republic
of China. Sociological Forum, 11(2), 285–303.
Am J Crim Just (2019) 44:146–165
165
Stephen J. Watts is an Assistant Professor in the Department of Criminology and Criminal Justice at the
University of Memphis. His research focuses on the victimization-offending overlap, the testing of criminological theories, and the integration of these theories into a biosocial framework for explaining antisocial
behavior.
Karli Province is a master’s student and graduate research assistant at the University of Memphis in the
Department of Criminology and Criminal Justice. Currently, her research focuses on GxE interactions,
homicide, and GIS.
Kayla Toohy is a master’s student and graduate research assistant at the University of Memphis in the
Department of Criminology and Criminal Justice. Currently, her research focuses on racial and ethic
heterogeneity and its relationship to homicide. Her other research interest is the utilization of GIS to study
crime patterns.
American Journal of Criminal Justice is a copyright of Springer, 2019. All Rights Reserved.
Am J Crim Just (2019) 44:146–165
https://doi.org/10.1007/s12103-018-9438-6
The Kids Aren’t Alright: School Attachment, Depressive
Symptoms, and Gun Carrying at School
Stephen J. Watts 1 & Karli Province 1 & Kayla Toohy 1
Received: 10 January 2018 / Accepted: 21 March 2018 /
Published online: 5 April 2018
# Southern Criminal Justice Association 2018
Abstract Social science has frequently examined the relationships between school
environment and delinquency, mental health and delinquency, and school environment
and mental health. However, little to no research to date has examined the interrelationship between these variables simultaneously, especially at it relates specifically to
delinquent acts committed at school. The current study uses data from the National
Longitudinal Study of Adolescent to Adult Health (Add Health) to look at the
interrelationship between these variables. What is found in this data is that the
relationship between negative mental health states and delinquency at school, specifically measured as depressive symptoms and gun carrying at school, respectively, is
possibly a spurious one, wherein both of these variables are partly shaped by school
attachment, which accounts for their correlation. Implications for theory and policy are
discussed.
Keywords School violence . Guns . Depression . School attachment . Add health
Introduction
School violence, including bullying, fights, assaults, and mass shootings, is a major
concern in America. This is despite the fact that our schools are relatively safe (Kann
et al., 2016), with fear of school violence, in particular fear of school shootings, largely
* Stephen J. Watts
sjwatts@memphis.edu
Karli Province
klprvnce@memphis.edu
Kayla Toohy
krtoohy@memphis.edu
1
Department of Criminology and Criminal Justice, University of Memphis, 311 McCord Hall,
Memphis, TN 38152, USA
Am J Crim Just (2019) 44:146–165
147
being driven by sensationalized coverage of a few events (Cornell, 2006). But any level
of school violence is problematic because it can lead to, beyond the obvious negatives
like injury and death, a negative learning environment for students. The research that
has been done to understand school violence has largely focused on factors such as the
parent-child relationship, substance use, having a history of violence, and bullying
(Brockenbrough, Cornell, & Loper, 2002; Cornell, 2006; Lipsey & Derzon, 1998;
Resnick, Ireland, & Borowsky, 2004).
Children and adolescents carry guns to school for any number of reasons,
including fear of victimization and peer pressure (see Beardslee, Docherty,
Mulvey, Schubert, & Pardini, 2017; Esselmont, 2014; May, 1999; Melde,
Esbensen, & Taylor, 2009; van Geel, Vedder, & Tanilon, 2014; Wilcox, May, &
Roberts, 2006), but when school shootings in particular have been studied, they
have generally been looked at from a mental health standpoint (Borum, Cornell,
Modzeleski, & Jimerson, 2010). When high-profile school shootings specifically are
examined, there is often at the center an offender with some sort of mental health issue,
whether that issue is short-term or long-term. But perhaps this excessive focus on
mental health has left the full picture of why school shootings occur unclear. There
could be factors related specifically to the school environment that increase negative
emotionality and risky behaviors in school. The current study will focus on one such
risky behavior that can serve as a precursor to school violence or directly precipitate
school violence, carrying a gun at school. Specifically, we examine the relationship
between school attachment, depressive symptoms, and gun carrying at school in a
nationally representative, panel study of American adolescents to see if an adolescent’s
positive or negative identification with their school can partly shape depressive symptoms and gun carrying at school, and whether these associations can potentially reveal
the relationship between depressive symptoms and gun carrying to be spurious. By
spurious, we mean that the presumed causal relationship between depressive
symptoms and gun carrying is actually a mere correlation, a correlation driven
by the two variables jointly having a relationship with and partly being shaped by
school attachment. In the following section we will review the literature on mental
health and school violence, and identify how school environment, and in particular
school attachment, fits into this relationship.
Literature Review
Mental Health and School Violence
School violence has many sources, as do risky behaviors that can precede school
violence, like gun carrying. The literature on school violence and risky behaviors has
focused on a myriad of factors as potential causes, including substance use, a history of
violence, poor socioeconomic background, bullying, and truancy (Bailey, Flewelling,
& Rosenbaum, 1997; Brockenbrough et al., 2002; Cornell, 2006; DuRant, Krowchuk,
Kreiter, Sinal, & Woods, 1999; Esselmont, 2014; Muschert, 2007; Shetgiri, Boots, Lin,
& Cheng, 2016; Simon, Crosby, & Dahlberg, 1999; van Geel et al., 2014; Webster,
Gainer, & Champion, 1993). But much of the research on school violence and risky
behaviors that can precede school violence, gun carrying in particular, has additionally
focused on a broader topic of concern within violence in general and gun violence in
148
Am J Crim Just (2019) 44:146–165
particular, mental health. Much of this literature has pointed to depression, anger,
anxiety, and fear as risk factors for both school violence and risky behaviors like gun
carrying (DuRant et al., 1999; Juan & Hemenway, 2017; Melde et al., 2009; Muschert,
2007; Shetgiri et al., 2016; Simon et al., 1999; Webster et al., 1993; Wilcox et al.,
2006), and the literature on general strain theory in particular has focused on stress and
its link to school violence (Agnew, Brezina, Wright, & Cullen, 2002; Hinduja &
Patchin, 2007; Patchin & Hinduja, 2011).1
The research that has focused specifically on those school shootings that have
received widespread media coverage is very clear in its focus on mental health
(Harter, Low, & Whitesell, 2003; Juan & Hemenway, 2017; Meloy, Hempel,
Mohandie, Shiva, & Gray, 2001; Muschert, 2007). With the widespread availability
of guns in the United States considered a constant risk factor, there has been a strong
focus on mental health issues as an important individual-level predictor of perpetrating
a school shooting. Even when other factors, like romantic rejection (Klein, 2005),
bullying victimization (Burgess, Garbarino, & Carlson, 2006; Harter et al., 2003; Leary,
Kowalski, Smith, & Phillips, 2003; Meloy et al., 2001), and a troubled home life
(Webber, 2003), are considered, these factors usually end up mattering because they
cause the shooter mental distress, usually in the form of depression and/or anger.
If we step back from school shootings in particular, and focus instead on school
violence and risky behaviors in general, and consider negative emotions as an important cause of these issues, then that leads us to question, where do these negative
emotions come from that directly relate to school violence and risky behaviors? A
logical answer seems to be that negative emotions that directly relate to school violence
and risky behaviors at school should trace back in part to stressors related to school.
And indeed, this can be seen in the prior research on school violence and its focus on
peer rejection and bullying victimization and, at the macro-level, school-level factors
like the quality of faculty-student relationships (Moore, Petrie, Bragaand, &
McLaughlin, 2003) and the ability of administrators to enforce rules and respond to
threats (Fox & Harding, 2005). All of these factors taken together should shape an
individual’s level of attachment to and positive identification with their school, their
peers, their teachers, and school rules. In the next section, we turn to a discussion of
school attachment, and its potential role in generating both negative emotions and
school violence and risky behaviors like gun carrying among adolescents.
School Attachment, Mental Health, and School Violence and Risky Behaviors
Strong social bonds to important social institutions, like family and school, have long
been seen as important deterrents to delinquency and crime, among adolescents and
adults, in the criminological literature (Akers & Sellers, 2016). Most of this is due to the
development of social bonding theory by Travis Hirschi (1969) and its subsequent
popularity. Social bonding theory has been frequently tested and widely endorsed in the
criminological literature (Akers & Sellers, 2016). Hirschi’s theory argues that most
1
While this study will invoke many of the concepts found in general strain theory, we do not use general strain
theory as a theoretical guide, because rather than focusing on a lack of school attachment as a strain that
increases negative emotions which then result in deviance (i.e., the mediating relationship general strain theory
proposes), we instead focus on and find that the negative emotions-deviance relationship is a spurious one,
with school attachment as the key confounding variable.
Am J Crim Just (2019) 44:146–165
149
individuals do not offend because of the social controls placed on us as individuals in
the form of social bonds. A social bond is a force within one’s environment that
connects them to society’s moral constraints (Winfree Jr & Abadinsky, 2009).
Through these social bonds, we are attached to conventional individuals, committed
to societal norms, involved in conventional activities, and we come to believe in the
moral force behind society’s norms. Social bonding theory thus explains crime and
delinquency as the result of an individual’s social bonds being weak or nonexistent.
Based on this theory, it then makes sense that school attachment, the social bond to
one’s school, would be important in the etiology of crime and delinquency, especially
crime and delinquency at school. And indeed, this is exactly what the literature shows,
with measures of school attachment relating to various types of youth offending across
a wealth of studies in the criminology and education literatures (e.g., Bond et al., 2007;
Cernkovich & Giordano, 1992; Chapple, Tyler, & Bersani, 2005; Dornbusch, Erickson,
Laird, & Wong, 2001; Henrich, Brookmeyer, & Shahar, 2005; Henry & Slater, 2007;
Herrenkohl, Huang, Tajima, & Whitney, 2003; Hirschfield & Gasper, 2011; Longshore,
Chang, & Messina, 2005; Payne, 2009; Resnick, Harris, & Blum, 1993; Watts, 2017;
Wilson, 2004; Zhang & Messner, 1996). As an example, Henry and Slater (2007)
found that overall levels of school attachment within a school decrease individual-level
adolescent alcohol use behaviors.
Most studies in the criminological literature focus on school attachment and
delinquent behavior in general (e.g., Cernkovich & Giordano, 1992; Longshore et al.,
2005). Less often seen in this literature are studies that look specifically at the link
between school attachment and criminal and delinquent behaviors committed in the
school setting. Exceptions include a study by Jenkins (1997), who found in a sample of
7th and 8th graders that school attachment decreased school crime, misconduct, and
nonattendance, and a study by Hirschfield and Gasper (2011), who found in a sample
of inner-city Chicago elementary school students that elements of school engagement
decreased delinquency in school.
So the research shows a clear link between school attachment and crime and
delinquency among adolescents, even if the research on crime and delinquency at
school specifically is somewhat lacking. This leaves the issue of the potential
relationship between school attachment and mental health. To date, a number of studies
in the education literature have examined the relationship between school attachment,
referred to in this literature as Bschool connectedness,^ and mental health among
adolescents. These studies have consistently found that school attachment has a
strong relationship with mental health among adolescents, with low levels of
school attachment serving as a risk factor for negative mental health states, such
as experiencing depression and anxiety (Bond et al., 2007; Lester, Waters, &
Cross, 2013; Romo & Nadeem, 2007; Shochet, Dadds, Ham, & Montague,
2006). For example, Shochet and colleagues (2006) found in a sample of
Australian youths that school attachment (Bconnectedness^) predicted depressive
symptoms a year later for boys and girls, even when controlling for prior
depressive symptoms. Of note, they found that the reverse was not true, with
depressive symptoms not predicting school attachment a year later when
controlling for prior school attachment. Bond et al. (2007) found that low school
attachment, even in the presence of good overall social connectedness, increased
risks for depressive symptoms and anxiety in a 2–4 year follow up.
150
Am J Crim Just (2019) 44:146–165
So prior literature has tied school attachment to both crime and delinquency and
mental health among adolescents. However, these issues have largely been examined
separately. With few exceptions, research to date has not considered the interrelationship between school attachment, mental health, and crime and delinquency among
adolescents. One recent study, conducted by Juan and Hemenway (2017), considered
all of these factors together, with the authors taking the view that negative mental health
states, specifically depression, have primacy in how these variables connect. They
argued that depression has its effects on crime and delinquency, specifically measured
as carrying a gun at school, partly through depression’s effect on Bsocial
connectedness,^ a measure that included aspects of the relationship with one’s peers.
They argue for a mediating model whereby depression undermines social connectedness, which can then lead to gun carrying, and thus, social connectedness can act as a
Bbuffer^ between depression and acts of deviance at school like gun carrying.
However, the prior research cited above would argue just as strongly that school
attachment would hold primacy, especially when the focus is on crime and delinquency
in the school setting. Putting together the prior literature on school attachment, mental
health, and youth crime and delinquency in the school setting, an argument can be
made that the relationship between negative mental health states and offending in
schools could potentially be spurious. It may be that school attachment precedes and
partly shapes both mental health and deviance in school, and that the relationship
between them is driven by their joint relationship with school attachment. This is the
possibility that the current study will seek to assess.
The Current Study
This paper seeks to build on the existing literature on school attachment, mental health,
and crime and delinquency among adolescents. This will be done by examining the
relationship between school attachment, depressive symptoms, and a specific act of
deviance at school, carrying a gun, in a longitudinal probability sample of adolescents.
In so doing, we will expand on the literature in several important ways. First, this is one
of few studies to consider school attachment and focus specifically on delinquency at
school. Second, this is one of very few studies to consider school attachment, mental
health, and crime and delinquency simultaneously. Lastly, this will be the first study to
assess whether the relationship between negative mental health states and offending
among adolescents is in part driven by school attachment. This possibility has been
hinted at in the various literatures cited, but has to date not been examined empirically
with adequate data.
This study draws on prior literature, particularly expectations from social bonding
theory and findings concerning school connectedness and depression in the education
literature, to derive four hypotheses concerning the relationship between school attachment, depressive symptoms, and gun carrying at school. Hypothesis 1 predicts that
school attachment will be associated with depressive symptoms such that higher levels
of school attachment will mean lower levels of depression. Hypothesis 2 predicts that
without considering school attachment, depressive symptoms will be associated with
gun carrying at school such that more depressive symptoms will correspond with a
higher likelihood of carrying a gun to school. Hypothesis 3 predicts that school
attachment will be associated with gun carrying at school such that higher levels of
Am J Crim Just (2019) 44:146–165
151
school attachment will correspond with a lower likelihood of carrying a gun to school.
Finally, Hypothesis 4 predicts that controlling for school attachment will reduce the size
and significance of the depressive symptoms-gun carrying at school relationship. Thus,
the depressive symptoms-gun carrying relationship will ultimately be spurious, a mere
correlation partly driven by their joint relationship with school attachment. Fig. 1
presents the proposed empirical model. With this proposed empirical model, the
authors are not trying to suggest that school attachment is the only explanation for
depressive symptoms and/or gun carrying among adolescents, but rather, school
attachment can partly explain both, and in so doing, explain the correlation between
depressive symptoms and gun carrying at school.
Methods
Data
The National Longitudinal Study of Adolescent to Adult Health (Add Health) is a
longitudinal panel study of a nationally representative, school-based probability sample
of youth who were in grades 7–12 in the United States beginning in 1994–95 (Udry,
2003). Add Health employed cluster and systematic sampling methods to select 80
public and private high schools and 52 middle schools to be representative of U.S.
regions, urban composition, school size and type, and ethnicity (Harris et al., 2003).
Add Health first collected data with an in-school survey, then, for a subsample,
followed up with a series of in-home interviews conducted in 1994–95 (Wave I),
1996 (Wave II), 2001–02 (Wave III), and 2007–08 (Wave IV). The current study
utilized data from the restricted data set, which is identical to the public-use data set
but contains the full sample of respondents. Data from respondents who completed
both the Wave I & II in-home surveys, and for whom sampling weights were available,
was utilized (N = 13,568). Table 1 provides descriptive statistics for the study variables.
Wave II
(past 30 days)
Wave I
(past 7 days)
Depressive
Symptoms
Gun Carrying at
School
School Attachment
Wave I
(current/most recent school year)
Fig. 1 Empirical model
152
Am J Crim Just (2019) 44:146–165
Table 1 Descriptive statistics
Variables
Range
Full Sample
(N = 13,568)
Mean (SE)
Dependent variable
Gun Carrying at School Wave II
0/1
.01 (.00)
Depressive Symptoms Wave I
0–51
10.46 (.06)
School Attachment Wave I
6–38
27.64 (.04)
Independent variables
Controls
Male
0/1
.49 (.00)
White
0/1
.52 (.00)
Black
0/1
.20 (.00)
Native American
0/1
.02 (.00)
Asian
0/1
.07 (.00)
Other
0/1
.02 (.00)
Hispanic
0/1
.17 (.00)
Age Wave II
11–23
16.23 (.01)
Public Assistance Wave I
0/1
.10 (.00)
Maternal Warmth
1–5
4.37 (.01)
Maternal Supervision
3–15
5.83 (.02)
Out of School Suspension Wave II
0/1
.10 (.00)
English Grave Wave II
1–4
2.16 (.01)
Gang Member Wave II
0/1
.05 (.00)
Peer Delinquency Wave I
0–9
2.42 (.02)
Thoutfully Reflective Decision Making
4–20
8.81 (.02)
Self-control
1–5
2.99 (.01)
Fighting in School Wave II
0/1
.06 (.00)
Gun easily accessible in home Wave II
0/1
.16 (.00)
Because these statistics are weighted and adjusted for survey design, standard errors are produced rather than
standard deviations. 86 respondents reported that a gun was the weapon they carried to school most often in
the 30 days prior to the WII interview
Measures
Dependent Variable: Gun Carrying at School Wave II
The dependent variable is a single dichotomous measure from Wave II that measures
gun carrying at school. While school shootings were often mentioned in the literature
review, school shootings are an exceedingly rare phenomenon, and Add Health does
not contain measures on using a gun at school. Additionally, Add Health contains
limited measures about frequencies of school-specific violent or delinquent behavior,
with most measures of delinquent and criminal behaviors in the dataset not specifying
Am J Crim Just (2019) 44:146–165
153
the domain (i.e., at home, at school, in the neighborhood) in which the behavior
occurred.
Respondents were first asked if they had carried a weapon at school since the Wave I
interview. For those respondents that answered affirmatively, a follow-up question asked
what one kind of weapon they carried to school most often in the past 30 days. Possible
answers included a handgun, some other type of gun (rifle, shotgun, etc.), a club or bat, a
knife or razor, or something else. We used this question to create a dichotomous measure
of whether someone reported that the weapon they carried most often to school in the
past 30 days was a gun of any kind (1 = yes). While this measure would be better if it
focused on gun carrying in general rather than as a preferred weapon, the time frame
covered does add confidence in the causal ordering of the proposed empirical model.
Eighty-six respondents reported that a gun was the weapon they carried to school most
often in the 30 days prior to the Wave II interview.2
Independent Variables
Depressive Symptoms Wave I To assess depressive symptomatology we drew on 18
questions from Wave I based on the Centers for Epidemiological Study-Depression
Scale (CES-D) (Radloff, 1977, 1991), that we summed in an index where higher scores
denote higher levels of depressive symptomatology (range = 0 to 51, alpha = .86).
These items asked about feelings of loneliness, fear, sadness, fearfulness, etc. in the
7 days prior to the Wave I interview, with responses ranging from never or rarely (0) to
most of the time or all of the time (3). The purpose of this construct in the current study
is not to assess depressive symptoms based on criteria for diagnosis of major depression, but rather, to assess the level of depressive symptomatology among Add Health
respondents along a spectrum. Past research has similarly measured depressive symptomatology in Add Health (Goodman, Slap, & Huang, 2003; Hallfors et al., 2004;
Hallfors, Waller, Bauer, Ford, & Halpern, 2005; Iratzoqui, 2015; Kaufman, 2009;
Steuber & Danner, 2006; Watts & McNulty, 2013). The full list of items that make
up the depressive symptoms measure can be found in the appendix.
In focusing on depressive symptoms in the relationship between mental and gun
carrying, we are not seeking to discount the importance of other emotions, such as
anger, hostility, and/or fear. We are simply limited by what measures are available in the
Add Health, and there are not valid and complete measures of these other emotions in
the early waves of data collection.3
School Attachment Wave I The social bond to and positive identification with one’s
school is represented by an 8 item additive index of questions drawn from Wave I that
2
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 Bsvyset^ 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 Bsvyset^
command, which is not advised by Add Health, and the Bfirthlogit^ command is used to run a type of rare
events logit estimation, results that are substantively identical to those presented in this paper are produced.
These results are available upon request.
3
For example, while later waves of data have better questions concerning anger, the only one available in
Waves I and II comes from the parent interview in Wave I, where the target respondent’s parent was asked,
BDoes your child have a bad temper?^ Responses were simply recorded as Byes^ or Bno.^
154
Am J Crim Just (2019) 44:146–165
measure a respondent’s sense of belonging to their school where higher scores denote
higher levels of school attachment (range = 6 to 38, alpha = .74). Respondents were
asked about whether during the current or most recent school year they had had trouble
getting along with teachers or fellow students, if they felt close to people at school, felt
like a part of their school, believed their fellow students were prejudiced, were happy to
be at their school, whether teachers treated students fairly, and if they felt safe at their
school. Taken together, these items assess how attached and committed a respondent
was to their school, and how much they felt they belonged to and were safe at their
school during the current or most recent school year at Wave I. Past research has
similarly measured the concept of school attachment in Add Health (Johnson, Crosnoe,
& Thaden, 2006; McNeely & Falci, 2004; McNeely, Nonnemaker, & Blum, 2002;
Pearson, Muller, & Wilkinson, 2007; Watts, 2017). The full list of items that make up
the school attachment measure can be found in the appendix.
Controls
Numerous controls were considered to check the robustness of any observed relationships of central concern, and to help avoid model misspecification and omitted variable
bias (Mustard, 2003). General controls included Male, dummy variables for Hispanic,
non-Hispanic Black, Native American, Asian, and Other (with non-Hispanic White as
the reference category), Age (Wave II),4 and one measure of socioeconomic status,
Public Assistance (Wave I).5 The measure of public assistance indicates whether the
respondent’s family was receiving any sort of public assistance, such as welfare or food
stamps, at Wave I (1 = yes).
In addition to the more general controls, a number of controls were included because
they may relate simultaneously to school attachment, negative mental health states, and
general delinquent behavior. Controlling for these variables is an important test of the
robustness of any relationships uncovered between school attachment, depressive
symptoms, and gun carrying in schools. For this reason, we included two measures
of parenting, Maternal Warmth and Maternal Supervision. Maternal Warmth is a single
item that asked respondents at Wave I if their mother was warm and loving towards
them most of the time, with responses ranging from 1 (strongly agree) to 5
(strongly disagree). Maternal Supervision consists of three items from Wave I
that asked respondents how often their mother was at home when they left for
school, returned from school, and went to bed, with higher scores denoting less
supervision.
We also included two school-related controls. Out of School Suspension is a single
item from Wave II that asked if during the current or most recent school year the
respondent had received an out of school suspension (1 = yes). English Grade also
4
While there is a large age range in Add Health, the vast majority (88.6%) of respondents were between 14
and 18 years old at Wave II.
5
Regression models run with additional SES controls for parental education, parental employment, and total
family income produce identical results for the key variables of interest. With these SES measures infrequently
showing statistical significance and having little impact on the results, a more parsimonious model with only
the public assistance variable is presented. Results with all of these SES measures included are available upon
request.
Am J Crim Just (2019) 44:146–165
155
comes from Wave II, consisting of one item where respondents were asked what their
most recent grade was in English/language arts, with responses ranging from 1 (A) to 4
(D or lower). We also included two measures of delinquent peer affiliations. Gang
Member comes from Wave II, and consists of one item that asked if respondents had
ever been initiated into a named gang (1 = yes). Peer Delinquency comes from Wave I,
and consists of three items that asked respondents how many of their three closest
friends smoke at least one cigarette per day, drink alcohol at least once a month, and
smoke marijuana at least once a month. This measure of peer delinquency has been
shown in past studies using Add Health data to have predictive validity (Beaver,
Wright, & DeLisi, 2008; Watts & McNulty, 2015).
We additionally controlled for two conceptually similar variables related to impulse
control and the use of reason. Thoughtfully Reflective Decision Making consists of four
items from Wave I that recorded responses to four statements: When you have a problem
to solve, one of the first things you do is get as many facts about the problem as possible;
When you are attempting to find a solution to a problem, you usually try to think of as
many different approaches to the problem as possible; When making decisions, you
generally use a systematic method for judging and comparing alternatives; After
carrying out a solution to a problem, you usually try to analyze what went right and
what went wrong. Possible responses ranged from 1 (strongly agree) to 5 (strongly
disasgree). Higher scores denote less reflective decision making (Paternoster &
Pogarsky, 2009). Self-Control is measured in consideration of Hirschi’s (2004) redefinition of the concept as the tendency to examine the full range of outcomes from one’s
behavior. As such, this measure consists of one item from Wave I wherein respondents
were asked whether when making decisions they usually go with their Bgut feeling^
without thinking too much about the consequences of each alternative, with responses
ranging from 1 (strongly agree) to 5 (strongly disagree) (Paternoster & Pogarsky, 2009).
Finally, we included measures concerning violence and gun access. The variable
Fighting in School measured at Wave II whether the most recent fight a respondent got
into occurred at school (1 = yes) or somewhere else. We included this measure because
violent behaviors like fighting tend to correlate highly with risky behaviors like carrying
a weapon to school (Forrest, Zychowski, Stuhldreher, & Ryan, 2000; Lowry, Powell,
Kann, Collins, & Kolbe, 1998; Pickett et al., 2005; Resnick et al., 1997; Resnick et al.,
2004). The variable Gun Easily Accessible in Home utilizes a measure from Wave II
where respondents were asked if a gun is easily available to them in their home (1 = yes).
Studies have linked access to guns to youth violence (Blumstein & Cork, 1996;
Kennedy, Piehl, & Braga, 1996; Schubiner, Scott, & Tzelepis, 1993).6
Analytic Strategy
We tested our hypotheses utilizing ordinary least squares (OLS) regression (depression)
and logistic regression (gun carrying at school).7 The models presented test whether
school attachment at Wave I (current/most recent school year) significantly associates
6
While fear of victimization has also consistently been tied to gun carrying among adolescents (see Melde
et al., 2009; Wilcox et al., 2006), measures assessing fear of victimization are not available in Add Health.
7
For these OLS models we computed the ln + 1 of the depression variable to smooth out the distribution. For
this logged variable, skewness = −.81 and kurtosis = .77.
156
Am J Crim Just (2019) 44:146–165
with depressive symptoms at Wave I (past 7 days) (Hypothesis 1), whether depressive
symptoms significantly associate with gun carrying at school at Wave II when school
attachment is not considered (Hypothesis 2), whether school attachment at Wave I
significantly associates with gun carrying at school at Wave II (Hypothesis 3), and
whether controlling for school attachment reduces to insignificance the depressive
symptoms-gun carrying at school relationship (Hypothesis 4). The appropriate weight,
cluster, and strata variables were utilized in all analyses to account for the complex
sampling design of Add Health.8 Tests utilizing Variance Inflation Factors (VIFs) show
that multicollinearity is not a concern in regards to the independent variables in the
OLS or logistic regression models.
Results
Table 2 presents the results of OLS models with depressive symptoms at Wave I
regressed on school attachment at Wave I and the controls.9 To smooth out the
distribution of the depressive symptoms variable, we computed the ln + 1 of this
variable (skewness = −.81) for use in these models. Model 1 in Table 2 shows school
attachment at Wave I has a highly significant, negative association with depressive
symptoms at Wave I, supporting Hypothesis 1. In model 2, the controls are included to
check the robustness of this effect, and the coefficient for school attachment is still
highly significant, and only modestly reduced. Among the controls, males report less
depressive symptoms than females, black, Native American, and Hispanic respondents
report more depressive symptoms than non-Hispanic whites, older respondents report
more depressive symptoms, and those whose families were receiving public assistance
report more depressive symptoms. Also, less maternal warmth means more depressive
symptoms, those who had been suspended from school and those who had worse
grades report more depressive symptoms, more peer delinquency means more depressive symptoms, and less reflective decision making and less self-control mean more
depressive symptoms.
Table 3 presents the results of logistic regression models with odds ratios with gun
carrying at school at Wave II regressed on depressive symptoms at Wave I, school
attachment at Wave I, and the controls.10 To ease the interpretation of the log odds, we
z-score transformed the depressive symptoms and school attachment measures for use
in these models. Model 1 of Table 3 includes only depressive symptoms at Wave I,
which has a significant, positive association with gun carrying at school at Wave II,
8
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
Bsvyset^ command.
9
Stata 13.1 will not produce standardized (beta) coefficients for regression models that have been Bsvyset.^
This is for statistical reasons. Stata 13.1 and other programs are only willing to consider variance decomposition when data are understood to be independent and identically distributed (IID). Using the Bsvyset^
command is specifically meant to correct for the data not being IID, therefore Stata 13.1 is unwilling to
consider ratios of variance in these models.
10
When data are Bsvyset^ in Stata 13.1,…
Purchase answer to see full
attachment
Why Choose Us
- 100% non-plagiarized Papers
- 24/7 /365 Service Available
- Affordable Prices
- Any Paper, Urgency, and Subject
- Will complete your papers in 6 hours
- On-time Delivery
- Money-back and Privacy guarantees
- Unlimited Amendments upon request
- Satisfaction guarantee
How it Works
- Click on the “Place Order” tab at the top menu or “Order Now” icon at the bottom and a new page will appear with an order form to be filled.
- Fill in your paper’s requirements in the "PAPER DETAILS" section.
- Fill in your paper’s academic level, deadline, and the required number of pages from the drop-down menus.
- Click “CREATE ACCOUNT & SIGN IN” to enter your registration details and get an account with us for record-keeping and then, click on “PROCEED TO CHECKOUT” at the bottom of the page.
- From there, the payment sections will show, follow the guided payment process and your order will be available for our writing team to work on it.