Securing Schools? Examining the Longitudinal Relationships Between Student Behavior and Engagement With Security Measures in Urban School Systems

Abstract

The use of school security measures has increased over the last two decades. Yet prior research suggests school security measures have a deterrent effect on student misbehavior. Existing studies often focus on school-level comparisons in security as opposed to examining how students within a given school differ in their interaction with security measures (i.e., within-school differences). To address this gap in the literature, the current study estimates the association between individual students’ engagement with security and multiple forms of maladaptive student behavior in school. In particular, this study is guided by two research questions: 1) What is the relationship between students’ engagement with school security measures and their engagement in problem behaviors; and, 2) To what extent do the relationships between engagement with security and student behavior problems differ by student race and ethnicity? Longitudinal data were collected from students at two separate time points in one academic year (N=359) across eight schools in one urban school district. Using a series of models to examine how students’ engagement with school security measures is related to their perpetration of student behavior, findings highlight negative associations between engagement with school security and non-serious violent and weapons-related crime. While the school security change score and students’ engagement in problem behaviors was no different for Black students than it was for students who were non-Black or non-Hispanic, the negative association between engagement with security and behavior indicated a stronger deterrent effect for Hispanic students. Findings suggest that engagement with school security should be examined at the within-school level and with consideration that racial and ethnic differences might vary from student to student within any given school. Moreover, long-term programming goals should be established when developing process for securing schools with emphasis on how security measures might influence individual students differently within the school setting.

Keywords

school safety behavior school security student engagement survey research Longitudinal

Introduction

Although school violence and student misbehavior has generally trended downward over the last two decades, there has been a notable increase in the use of school security measures across the United States (Musu et al., 2019). One might infer that the decline in frequency and severity of problematic student behavior in school is attributed to an increase in the deployment of security measures; however, prior research suggests that school security measures and student behavioral outcomes are largely unrelated (Reingle Gonzalez et al., 2016; Tanner-Smith et al., 2018; Turanovic et al., 2019). Moreover, while the increased implementation of school security measures is well documented (Musu et al., 2019), there is little research that has examined within-school differences in students’ engagement with security measures and how this engagement influences student behavior (Cuellar & Coyle, 2021).

This study examines the associations between engagement with school security measures and various forms of student misbehavior through two competing theoretical lenses: (a) opportunity theories of crime that suggest higher engagement with security should predict less violence and misbehavior, and (b) a school criminalization perspective that suggests greater engagement with security should be unrelated to or even increase school violence and student misbehavior. Using longitudinal data from students across eight high schools in one inner-city school district, this study explores the relationship between student engagement with school security and self-reported indicators of violent behavior, non-violent behavior, substance use, and property crime in school.

Literature Review

Prevalence of Violence and Other Problem Behaviors in School

According to national-level statistics, schools in the United States are experiencing the lowest rates of student problem behaviors witnessed in decades. During the 2015–2016 school year, 79% of public schools recorded that one or more incidents of violence, theft, or other crimes had taken place, amounting to 1.4 million crimes. Although these figures are large and represent a meaningful social problem, they are lower than they have been since at least the year 2000. Similar downward trends exist for various indicators of maladaptive student behavior, including physical fighting (33% in 2001 to 24% in 2017), gang presence (20% in 2001 to 9% in 2017), bullying (29% in 1999–2000 to 12% in 2015–2016), and hate crimes (12% in 2001 to 6% in 2017). This decline in school violence parallels broader trends of steady reductions in rates of youth crime and violence and could be partially attributed to restorative justice initiatives and increased student awareness of misbehavior in school (Musu et al., 2019). Engagement in and exposure to school violence can have negative effects on student outcomes. Students who perpetrate violence in school are at increased likelihood to drop out of high school, particularly among disadvantaged male youths (Staff & Kreager, 2008). In general, adolescent exposure to violence and peer delinquency is associated with greater health risk behaviors, poor mental health outcomes and lower high school performance and retention rates (Boynton-Jarrett et al., 2012; Cuellar, Coyle & Weinreb, 2021). Moreover, exposure to violence and maladaptive student behavior in school has been linked to aggression, particularly among secondary school youth (O’Keefe, 1997). Additional research indicates that school-based violence and victimization not only has a negative psychological effect on youth in schools, but can also affect the educational environment in which it occurs, compromising students’ feelings of safety and connectedness and impeding on the nurturing environment schools traditionally aim to establish (Eisenbraun, 2007).

School Security in the United States

One common approach to limit school violence and address day-to-day problem behavior in school is to employ various school security measures. These measures can include physical security measures (e.g., security cameras) and policies aimed at enhancing security (e.g., required visitor sign-ins). The use of these security measures in schools has increased dramatically since 2000 due to various factors, such as increased media attention to school violence and a culture of violence in North America (Addington, 2009). In fact, the percentage of students who reported observing the use of security cameras rose from 39% in 2001 to 84% in 2017, and those who reported observing police officers in their school rose from 39% in 2001 to 84% in 2017 (Musu et al., 2019). This growth is generally reflected in school-level statistics. The percentage of schools reporting use of security cameras has increased from 19% in 1999 to 81% in 2016, while the percentage of schools reporting the use of school police has increased from 54% in 2001 to 70% in 2014 (Musu-Gillette et al., 2017; Musu et al., 2019). Moreover, urban schools tend to see higher rates of violence and greater reliance on security measures that tend to focus on reducing criminal opportunity, which results in higher arrest rates for non-violent crimes (Servoss & Finn, 2014; Theriot & Cuellar, 2016).

Research suggests that student exposure to security measures differs by race at both the school and student level (Cuellar & Coyle, 2021). At the school level, researchers have consistently demonstrated a relationship between schools that serve a larger proportion of minority youth and use of “target-hardening” practices that aim to promote safety through structural and legal avenues (Kupchik & Ward, 2014; Mowen & Parker, 2017; Steinka-Fry et al., 2016). Moreover, nationally representative data suggest that schools that serve predominantly African American students tend to rely heavily on authoritarian practices (Cuellar, 2016; Servoss & Finn, 2014). At the student level, research suggests there are significant differences in student engagement with security measures by race (Shedd, 2015). For example, Cuellar and Coyle (2021) used data from a large urban school district and found that within-school variation in security measure engagement differed by race. More specifically, African American students were found to have a significantly higher likelihood of reporting being searched randomly in school for contraband when compared to their White and Hispanic counterparts, reflecting racialized patterns of security and surveillance enacted upon students.

Competing Theoretical Perspectives Regarding School Security

The potential effects of security measures on school violence can be framed in terms of competing theoretical perspectives. Opportunity theories of crime suggest that students engage in offending behaviors when they perceive low risk with high reward. Thus, limiting opportunity to engage in offending behavior will generally reduce the occurrence of offending behaviors within an particular setting (Hannon, 2002). Applied within the context of school security, opportunity theories of crime suggests that increasing school security should reduce students’ engagement in crime, violence, and other disorderly behaviors by way of limiting student opportunity to offend while increasing the risk of getting caught. One common application is routine activity theory. Routine activity theory suggests that offenders will offend when there is a confluence in time and space of a motivated offender, suitable target, and lack of capable guardianship (Cohen & Felson, 1979). Security measures would serve as the guardianship expected to reduce crime, and more interaction would further reduce the occurrence of maladaptive behavior in school. School resource officers are a good example of how this theory might be applied in school settings; student risk of getting caught engaging in violence or other forms of misbehavior should increase with the presence of security personnel in school, thus causing students to rethink engagement with increased risk and potentially low reward.

The school criminalization perspective contends that increasing security is ineffective at reducing students’ problem behaviors and may have unintended negative consequences, such as increased arrest rates, higher levels of absenteeism, poor academic performance and lower graduation rates, and a school climate in which developing connection to students and school is more challenging (Hirschfield, 2008; Kupchik & Monahan, 2006; Shedd, 2015). One argument focuses on negative expectancy effects; as students witness more security measures, they internalize the notion that their school is a dangerous place and consequently behave in ways that align with this perceived norm (Tanner-Smith & Fisher, 2016). This would suggest that increased engagement with security measures may in fact increase students’ problem behaviors. Another contention central to the school criminalization perspective is that racial and ethnic inequities are reproduced as schools exert more surveillance and control over students (Kupchik & Ellis, 2008; Kupchik & Ward, 2014). Prior research has shown that both within and between schools, non-White students experience more intensive and invasive surveillance than their White peers (Cuellar & Coyle, 2021; Kupchik & Ward, 2014; Shedd, 2015; Steinka-Fry et al., 2016). However, few studies have examined the extent to which security measures operate differently across different racial/ethnic groups in shaping outcomes related to school violence (Fisher et al., 2018). Researchers must focus on understanding differences by race and ethnicity to promote diversity and equity for students in school. This focus represents an extension of this literature into a new, policy-relevant area.

Empirical Literature Linking Security Measures and Student Behavior

Most research linking security measures to violence and other maladaptive behaviors in school has relied on school-level measures of security without attending to within-school differences in students’ engagement with security (Gawley et al., 2021). In other words, researchers have traditionally focused on outcomes that reflect the frequency of behavior reports for a given school as a whole as opposed to how these reports might differ across students within a given school. With this in mind, several studies have examined between-school differences in the use of a variety of security measures. One approach that has received attention is the use of metal detectors. In a systematic review, Hankin et al. (2009) analyzed seven studies that utilized self-report surveys on the use of metal detectors. Only one of these provided limited support that metal detectors reduce the amount of weapons brought to school. Furthermore, research suggests that metal detectors compromise students’ feelings of safety in United States schools (Gastic, 2011; Gastic & Johnson, 2014), and are used disproportionally in United States schools characterized by more violence and a large percentage of minority students enrolled (Toldson, 2012).

Another common approach to school security is policing. School policing has seen exponential growth over the last several years. However, there is little empirical evidence regarding the outcomes of law enforcement in schools, and the existing research suggests no reliable benefits to incorporating police in schools (Livingston et al., 2019). For example, in a meta-analysis by Fisher & Hennessy (2016), the researchers found that discipline rates increased following the implementation of school resource officers (Fisher & Hennessy, 2016). This supports previous research which suggests that school policing is associated with increased arrest rates for non-serious crimes in school (Na & Gottfredson, 2013). Early data suggested policing in schools is a promising practice that might decrease arrest rates while promoting a sense of safety among students by mitigating student arrests (Brown, 2006; Brown & Benedict, 2005; Johnson, 1999). However, research in the last decade paints a different picture. Using nationally representative data, Na and Gottfredson (2013) found that schools employing school resource officers (SROs) tend to report more arrests for weapons and drug charges as well as higher rates of arrests for minor offenses. Related research has linked police presence in schools to higher arrest rates (Homer & Fisher, 2020; Owens, 2017; Theriot & Cuellar, 2016) and more behavior problems (Gottfredson et al., 2020). Additionally, the interaction with SROs appears to have a detrimental effect on school climate (Devine, 1996; Nolan, 2011). For example, students who report negative interactions with SROs have been found to have poorer academic outcomes and more behavioral problems in school (Theriot & Cuellar, 2016; Wolf, 2014). It can also result in racially disparate outcomes pertaining to the use of security measures in school (Author, forthcoming; Shedd, 2015).

Additional literature has examined the relationship between schools’ use of multiple security measures and student behavior. One national study found mostly null effects between security measures and exposure to school violence, suggesting no practical relationship to suggest the implementation of security measures reduces the risk or occurrence of student misbehavior (Tanner-Smith et al., 2018). Fisher et al., (2018) conducted an analysis of the Educational Longitudinal Study 2002 and found that adolescents in schools with more security measures report higher odds of being threatened with harm, while no difference existed in odds of participating in a physical altercation or having something stolen (Fisher et al., 2018). Cuellar (2016) used the School Survey on Crime and Safety to examine the relationships between various types of security measures and violence in schools, identifying a significant positive association between physical security measures (such as metal detectors, locked gates, and security cameras) and the occurrence of physical attacks and fights, while legal security measures (such as school policing practices) were positively associated with weapons reported in school.

Finally, two systematic reviews are important to note. First, a systematic review found that, among 32 different studies, an increased use of security measures was not generally associated with a reduction in school violence (Reingle Gonzalez et al., 2016). Second, a more recent meta-analysis provides further evidence to suggest security measures do not prevent school violence (Turanovic et al., 2019) This body of\ research consistently highlights the notion that although security measures are a popular and increasingly utilized approach to addressing school violence, they do not appear to be effective in reducing school violence in a way that warrants their potential cost to students and the school environment.

The Current Study

The current study examines the longitudinal association between individual students’ engagement with security and multiple forms of violence perpetration in school. In particular, this study is guided by two research questions: 1) What is the relationship between students’ engagement with security measures and their engagement in problem behaviors (i.e., serious violent crime, nonserious violent crime, substance-related crime, weapon-related crime, property crime, and disorder); and, 2) To what extent do the relationships between engagement with security and student behavior problems differ by student race and ethnicity? To answer these questions, we examine how students’ engagement with security measures is related to changes in a variety of behavior problems at school, and how these relationships differ by race/ethnicity.

Method

Student Participants

The sample consisted of 9^th grade students (N = 359) across one large urban school district. The sampling frame (ninth grade students in eight participating schools) was represented by 1603 students (22.39% response rate). Among participants within the eight participating schools, 41.2% reported as male and 58.4% reported as female. In regard to race and ethnicity, 24.2% reported as White, 25.0% reported as Black or African American, 2.2% reported as Asian American, and 44.5% reported as other/mixed. Approximately 61.8% reported as Hispanic. In regard to socioeconomic status, 42.7% reported living with a single parent or guardian and 84.9% of the sample reported receiving free or discounted lunch during the school day. Across the district, 48.2% reported as female and 51.8% reported as male. Approximately 41.9% of students reported as Black or African American, 49.3% as Hispanic, and 7.6% reported as White. Approximately 75.0% of students were reported as being socioeconomically disadvantaged across the district. When compared to statistics provided by the state’s progress reports that provide demographic information for all students in the district, females, White students, and Hispanic students are overrepresented, while African American students are underrepresented in the final sample.

Additional Information Regarding Participating Schools

The urban area from which data were collected is located in the greater New York City area. The city in which the school is located has a current population of 311,549 with a density of approximately 12,800 people per square mile. The school district that participated in this study serves over 35,000 students enrolled in over 60 schools. The students that were targeted for this study were all ninth graders enrolled in the 14 schools that were formal public high schools (not charter schools) within the participating school district. The district serves a diverse student body. Across the district, approximately 41.9% of students reported as Black or African American, 49.3% as Hispanic, and 7.6% reported as White in the 2018–2019 school year.

According to the district, suspensions are down over 35% from last year, due to restorative justice initiatives, yet reported incidences of violence, vandalism, illegal substances, and weapons have decreased over the past three years, although many students still report feeling unsafe at school. Chronic absenteeism and student retention remains a major challenge for the district; approximately 50% of students in the public high schools sampled were identified as chronically absent (missed more than 18 days) for the 2014-2015 school year.

Across the participating schools, there were 1603 ninth graders eligible to participate in the study. The mean student enrollment was 983, with the median student enrollment being 614 students. Schools in the sample tended to be racially and ethnically heterogeneous in that the district-wide racial/ethnic composition was generally reflected in all school-level indicators of race and ethnicity among the Black (M = 48.63; SD = 29.94), Hispanic (M = 41.63; SD = 22.03), and White (M = 8.23; SD = 9.76) student populations. These school-level percentages were similar in all but one school, which had a larger percentage of White students in comparison to Black students, with the Hispanic student population making up the majority of students enrolled.

Student Recruitment Process: Both parents and students were required to opt-in to the research protocol. Parental consents were sent home two weeks prior to the administration of surveys. The researchers appointed a liaison at each school to direct the dissemination and collection of parental consents (i.e., opt-in forms). All ninth-grade students were invited to participate in the research. After parental consents were returned, the liaison contacted the researcher and the student surveys were scheduled. Student consents were read at the time the survey was distributed to students whose parent(s) or legal guardian(s) returned their consents. Liaisons at each school were appointed by their school’s administration and oversaw that all student consents were completed and matched the students that were asked to engage in the research. Students completed surveys in classrooms, cafeteria settings, libraries, and/or auditorium settings depending on the school’s facilities, and the researchers were available throughout the surveys in the case of student questions regarding the survey content. Liaisons were also present during all surveys to answer student questions and help students return to their classrooms. No study incentives were provided for participants.

Survey Development

The survey examined students’ engagement with school security, student perpetration of and victimization by school violence, and school culture and climate. Indicators of engagement with security and school violence were based off of data collected in the School Survey on Crime and Safety. Indicators of school culture and climate were operationalized using the Maryland Safe and Supportive Schools (MDS3) Survey (Bradshaw et al., 2014). Demographic characteristics were requested from participants. The final survey contained 119 items and took most students 20–30 minutes to complete. Surveys were completed across eight different schools throughout the participating district across two time points: once at the end of the Fall 2018 and once at the end of the Spring 2019 term (May 2019).

Variables

Engagement with School Security Measures

Student engagement with school security measures was the focal predictor in this study. To measure this construct, we used item response theory (IRT) to create a single latent composite score (θ) that represented students’ engagement with school security measures. Although IRT is perhaps most frequently applied in the field of educational testing, it has proven to be beneficial for modeling a variety of other latent constructs (Yang & Kao, 2014), including schools’ approaches to security (Fisher et al., 2018) and school climate (Johnson, 1999). θ is typically scaled so that it has a mean of 0 and each unit change represents a change in one standard deviation. In this study, a θ value of 1.0, for example, represents a level of engagement with security measures that is one standard deviation greater than the mean. Similarly, the coefficients presented in this study that link θ (i.e., students’ engagement with school security measures) to outcome variables can be interpreted in terms of standard deviation unit changes in θ.

IRT is particularly useful for measuring engagement with school security measures for multiple reasons. First, because most schools use more than one security measure in tandem (Steinka-Fry et al., 2016), IRT is useful from a data reduction perspective; rather than measuring engagement with each individual security measure, θ accounts for students’ engagement with each of the (possibly many) school security measures. Other measurement strategies such as additive scales or indices also have this advantage, but they typically force each item to be weighted identically.

Thus, the second major advantage of IRT is that the information contributed by each individual item to the latent measure of engagement with security measures differs according to the “difficulty” and “discrimination” parameters associated with each item (de Ayala, 2009). Each item contributes differentially to θ based on the item’s difficulty and discrimination parameters. The difficulty parameter in IRT is typically thought of as the difficulty of answering a test question correctly—in the current application, it can be interpreted as the difficulty of engaging with a given school security measure. Some school security measures—like wearing a school uniform—are relatively easy to engage with because students in schools with a uniform policy encounter the policy every day; others—like drug sniffing dogs—are more difficult because they are used infrequently. The discrimination parameter measures the extent to which an item corresponds with change in the value of θ; for example, more frequent engagement with a security measure with a large discrimination value would lead to a greater increase in theta than the same amount of engagement with a security measure with a smaller discrimination value. IRT considers these differing levels of difficulty and discrimination when calculating θ, which better reflects real-world engagement with security measures than other measurement approaches.

In the current study, 11 items were used in the IRT model, including student responses to the following: “How often do you…”: (1) check in to the front desk at school; (2) travel through locked/monitored gates to get to school; (3) pass through a metal detector; (4) use a structured anonymous threat reporting system; (5) interact with school security officer; 6) searched using a metal detector at school (e.g., wand); (7) checked by a drug-sniffing dog at school; (8) randomly searched for contraband at school; (9) required to take a drug test at school; (10) within eyesight of a security camera at your school; (11) required to wear/present a form of identification. Response options to all items were ordinal (0 = None of the Time; 1 = Some of the Time; 2 = Most of the Time; and 3 = All of the Time).

Student Problem Behavior

Items representing student behavior were operationalized based on the types of incidents outlined in the SSOCS. The survey asked students how often they engaged in various behaviors. The prompt read “Thinking of your time at school during a regular school day, how often have you engaged in the following behaviors?,” and included a list of 14 items. In line with prior research using the SSOCS data (Devlin & Gottfredson, 2018; Fisher, Higgins, & Homer, 2018; Na & Gottfredson, 2013), we grouped behaviors into categories. Serious violent crime included (a) physical attack/fight with a weapon, (b) robbery, and (c) sexual assault. Nonserious violent crime included (a) physical attack/fighting without a weapon, and (b) threatening physical attack. Substance-related crime included (a) distributing illegal drugs or alcohol, and (b) using illegal drugs or alcohol. Weapon-related crime included (a) possessing a firearm, and (b) possessing a knife or weapon other than a firearm. Property crime included (a) theft, and (b) vandalizing school grounds. Finally, disorderly behavior included (a) hating on a population different than yourself, (b) gang activity, and (c) bullying. Response options to all items were ordinal (0 = None of the Time; 1 = Some of the Time; 2 = Most of the Time; and 3 = All of the Time). Within each of these groups, we calculated mean scores that can be interpreted on this 0-3 scale.

Student Race/Ethnicity

Students self-reported their race/ethnicity as part of the survey. They responded to the following questions: (a) What is your race? (White, African American, Asian American, Other [open response]), and (b) Are you of Hispanic/Latino origin (No, Yes). For the current analysis, we created a set of dummy variables were used to measure student race/ethnicity. The first category, Black, included students who identified as African American but not Hispanic/Latino. The second category, Hispanic, included students who identified as Hispanic/Latino regardless of race. The reference group, then, included all students who were non-Black, non-Hispanic.

Although we have operationalized race/ethnicity in this way, we consider this student-level measure a rough proxy for individuals’ place within U.S. society’s system of racial stratification. As such, we do not interpret racial and ethnic differences as resulting from individual-level characteristics, but as resulting from a broader social system that has marginalized people of color for centuries. Additionally, we recognize that racialized groups are not monolithic; the experiences of individuals in the same racial/ethnic group are not all the same. Although this study is unable to disentangle these within-group differences, the findings presented here should not be understood as applicable to all members of racial/ethnic groups. These principles broadly align with a critical quantitative perspective (Gillborn et al., 2018) that is informative for conducting quantitative analyses of race and racism.

Control Variables

School climate variables measuring students’ perceptions of school climate at Wave 1 were used as control variables in the models. These measures were based on the MDS3 survey (Bradshaw et al., 2014), and included scale scores for the domains of (a) safety, (b) engagement, and (c) environment. Perceived safety is measured by items representing perceived safety, bullying and aggression, and general drug use. Engagement is measured by items representing Environment is measured by items operationalizing rules and consequences, physical comfort, support, and disorder. For additional information on how these scales were created, please refer to Bradshaw and colleagues (2014).

Data Analysis

All analyses used path analysis to estimate the relationship between students’ engagement with school security and the various measures of problem behaviors. Path analysis is a particularly useful approach here primarily because it readily permits the use of full-information maximum likelihood, a state-of-the art method for handling missing data (described further below). The data analysis proceeded in four steps. First, the IRT score for engagement with school security at Wave 1 was used to predict each Wave 2 dependent variable in turn while controlling for Wave 1 measures of safety, engagement, and environment. Because the school security items here were measured on an ordinal Likert-type scale, a unidimensional graded response model was used. This approach models both the discrimination and difficulty parameters while accommodating ordinal measures. Moreover, a clustered sandwich estimator was used to account for the clustering of students within schools. Using this model, we calculated item response scores for each observation which were used in subsequent models; these scores represent the engagement with school security measures of individual students. Table 1 displays the discrimination and difficulty value for each item included in the graded response model. Although some applications of IRT modeling suggest dropping items with particularly large discrimination or difficulty values (de Ayala, 2009), we retained all items regardless of their discrimination or difficulty values. We made this choice because we were not interested in creating the most parsimonious measurement model, but to capture as complete a picture of students’ engagement with school security as possible. Second, we calculated change scores in the measure of engagement with school security and used that variable to predict each Wave 2 dependent variable, still controlling for safety, engagement, and environment. Third, we added lagged dependent variables to each model using Wave 1 measures. Fourth, to examine differences by race/ethnicity, we calculated multiplicative interaction terms between the measure of students’ engagement with school security and indicators for Black and Hispanic. All significant interaction terms were plotted and probed (Preacher et al., 2006).

Table 1.

Difficulty and Discrimination Parameters for IRT Model of School Security.

	Wave 1			Wave 2
Variable	Discrimination	Difficulty		Discrimination	Difficulty
Pass through metal detectors	−0.74	>=1	5.67	1.66	>=1	−2.92*
		>=2	5.42		>=2	-2.74*
		3	4.31		3	-2.43*
Limited access to social networking sites	−0.25	>=1	3.92	0.31	>=1	-2.44
		>=2	1.63		>=2	−0.19
		3	0.39		3	1.70
Travel through locked, controlled, monitored gates	−0.12	>=1	−1.46	0.46*	>=1	−0.97
		>=2	−2.13		>=2	−0.46
		3	−3.76		3	0.16
Required to wear uniform	−0.06	>=1	43.74	0.30	>=1	−5.82
		>=2	20.11		>=2	−1.64
		3	1.14		3	1.07
Interact with school security officer	0.25	>=1	−3.21	0.43**	>=1	0.95*
		>=2	4.62		>=2	4.70**
		3	10.48		3	7.35**
Within eyesight of security camera	0.29	>=1	−7.34	1.08	>=1	−1.83
		>=2	−3.45		>=2	−0.96
		3	2.94		3	0.98**
Prohibited from using cell phones	0.30	>=1	−7.67	0.57**	>=1	−3.35*
		>=2	0.19		>=2	1.21*
		3	4.02		3	4.03
Use structured anonymous reporting system	0.50	>=1	4.98*	0.91***	>=1	2.39***
		>=2	6.19*		>=2	3.68***
		3	9.03*		3	4.69***
Searched by metal detector wand	0.60**	>=1	−1.50	0.86*	>=1	−1.36
		>=2	0.35		>=2	0.30*
		3	0.75		3	0.68**
Randomly searched for contraband at school	0.77	>=1	2.73	1.05*	>=1	2.54*
		>=2	3.78		>=2	3.48*
		3	4.07		3	3.92*
Required to wear clear bookbag during school	0.91	>=1	3.94	0.14	>=1	18.18
		>=2	4.53		>=2	21.87
		3	4.99		3	31.84
Check in at front desk to get to school	0.94	>=1	1.27	0.30	>=1	3.05
		>=2	2.13		>=2	5.33
		3	2.43		3	6.36
Required to wear ID	1.16	>=1	-0.02	0.71	>=1	−1.56
		>=2	1.56		>=2	1.80
		3	1.99		3	4.01
Required to take drug test at school	1.31	>=1	3.19	0.23	>=1	15.26
Required to take drug test at school		2	4.30		2	21.86
Searched using drug-sniffing dog	1.75	>=1	2.69	2.87	>=1	1.95*
		>=2	3.00		>=2	2.27*
		3	3.14		3	2.35**

Note. *p < .05; **p < .01; ***p < .001. Variables are sorted in ascending order of discrimination values at Wave 1. Both models used a clustered sandwich estimator to account for school-level clustering.

Missing Data

Many of the variables used in these analyses had a small amount of missing data, ranging from 0% to 21.45%. Missingness was primarily the result of students who were unexpectedly absent on the day surveys were administered. While efforts were made to visit each school multiple times, absenteeism was unpredictable and uncontrollable due to the design. Dropping cases with missing data can lead to biased estimates and is thus frequently an inappropriate technique for handling missing data (unless the data are missing completely at random, a very strict assumption; Allison, 2001). Here, we used full information maximum likelihood (FIML) to handle the missing data (Allison, 2001). This technique allows for the generation of model estimates using the data from all observations without missing data on a given variable. FIML avoids deleting observations with missing data and provides estimates without imputing any data.

Results

Descriptive statistics for the key study variables are presented in Table 2. As shown, no behavior problems were particularly common among the students in this sample. At the end of the school year in which data were collected, the most frequent behavior problems were nonserious violent crime (M = 0.21, SD = 0.52), disorder (M = 0.16, SD = 0.42), and property crime (M = 0.11, SD = 0.41). Less frequent behavior problems were drug crime (M = 0.08, SD = 0.44), serious violent crime (M = 0.06, SD = 0.25), and weapon crime (M = 0.03, SD = 0.24). The first measure of engagement with school security measures had a mean of 0.00 (SD = 0.61) and the school security change score had a mean of 0.00 (SD = 0.82).

Table 2.

Descriptive Statistics for Key Study Variables.

	M	SD	Min	Max	% missing
Dependent Variables
Serious Violent Crime (W2)	0.06	0.25	0.00	2.67	13.93
Nonserious Violent Crime (W2)	0.22	0.52	0.00	4.00	13.93
Substance-related Crime (W2)	0.08	0.44	0.00	4.00	14.21
Weapon-related Crime (W2)	0.03	0.24	0.00	3.50	13.93
Property Crime (W2)	0.11	0.41	0.00	4.00	13.93
Disorder (W2)	0.16	0.42	0.00	4.00	13.93
Independent Variables
Engagement with Security (W1)	0.00	0.61	-1.34	2.33	0.00
Engagement with Security Change Score	0.00	0.82	-3.31	3.58	0.00
Safety (W1)	1.14	0.64	0.11	2.56	16.43
Engagement (W1)	1.98	0.41	1.00	3.00	17.83
Environment (W1)	1.60	0.33	0.65	3.00	21.45
Serious Violent Crime (W1)	0.04	0.18	0.00	1.67	14.48
Nonserious Violent Crime (W1)	0.19	0.53	0.00	4.00	14.48
Substance-related Crime (W1)	0.04	0.19	0.00	1.50	15.88
Weapon-related Crime (W1)	0.01	0.09	0.00	1.00	15.32
Property Crime (W1)	0.08	0.34	0.00	2.50	15.32
Disorder (W1)	0.13	0.30	0.00	2.00	14.76
Black	0.19	0.39	0.00	1.00	0.00
Hispanic	0.62	0.49	0.00	1.00	11.70

The first set of models assessed the extent to which students’ engagement with school security measures at Wave 1 was associated with their subsequent engagement in behavior problems at Wave 2, controlling for their Wave 1 perceptions of school climate (i.e., school safety, engagement, and environment). These results are presented in Panel A of Table 3. As shown, greater engagement with security in the middle of the year was associated with more frequent involvement in problematic behaviors at the end of the year, including nonserious violent crime (b = 0.17, SE = 0.05, p = .002), weapon-related crime (b = 0.07, SE = 0.04, p = .049), and property crime (b = 0.11, SE = 0.07, p = .036). These effects are small in magnitude; the regression coefficients indicate that a one unit increase in engagement with security is equal to increases of approximately 32%, 29%, and 27% of a standard deviation in the three outcome variables, respectively.

Table 3.

Regression Models predicting Frequency of Engagement in Behavior Problems.

	Serious Violence				Nonserious Violence			Substances			Weapons			Property			Disorder
	b	SE	p	b	SE	p	b	SE	p	b	SE	p	b	SE	p	b	SE	p
Panel A: Security at wave 1
Security	0.07	0.04	.051	0.17	0.05	.002	0.08	0.05	.092	0.07	0.04	.049	0.11	0.05	.036	0.12	0.07	.062
Safety	0.03	0.04	.542	0.12	0.07	.080	0.09	0.09	.321	−0.01	0.03	.750	0.03	0.07	.718	0.08	0.07	.233
Engagement	−0.12	0.07	.083	−0.20	0.08	.007	0.07	0.15	.658	−0.15	0.10	.126	−0.09	0.14	.500	−0.07	0.11	.518
Environment	−0.02	0.10	.862	−0.05	0.14	.715	-0.19	0.16	.243	−0.05	0.15	.750	−0.12	0.17	.464	−0.11	0.21	.598
Panel B: Using school security change scores
Security change	−0.05	0.03	.095	−0.07	0.03	.034	-0.03	0.03	.328	−0.05	0.02	.049	−0.04	0.03	.135	−0.05	0.03	.052
Safety	0.03	0.04	.358	0.13	0.07	.041	0.10	0.09	.256	0.00	0.03	.997	0.03	0.06	.580	0.09	0.06	.168
Engagement	−0.10	0.07	.139	−0.17	0.09	.059	0.09	0.15	.548	−0.12	0.10	.204	−0.07	0.14	.594	−0.06	0.12	.603
Environment	−0.02	0.10	.840	−0.06	0.17	.724	-0.20	0.17	.232	−0.06	0.16	.684	−0.13	0.18	.477	−0.11	0.23	.641
Panel C: Adding lagged dependent variables
Security change	−0.05	0.03	.064	−0.05	0.02	.004	−0.03	0.03	.309	−0.05	0.02	.046	−0.04	0.03	.169	−0.05	0.03	.102
Safety	0.03	0.04	.433	0.12	0.06	.037	0.10	0.09	.264	0.00	0.03	.984	0.03	0.06	.577	0.10	0.06	.099
Engagement	−0.10	0.07	.175	−0.12	0.06	.034	0.10	0.14	.468	−0.12	0.10	.208	−0.07	0.12	.590	0.00	0.07	.982
Environment	−0.03	0.10	.780	−0.07	0.15	.624	−0.22	0.16	.177	−0.06	0.16	.688	−0.13	0.18	.441	−0.14	0.19	.468
Wave 1 Behavior	0.20	0.09	.027	0.26	0.14	.063	0.18	0.20	.365	0.05	0.10	.600	0.04	0.08	.648	0.25	0.19	.189
Panel D: Adding race/ethnicity moderators
Security change	−0.03	0.02	.099	0.05	0.06	.369	−0.033	0.04	.363	0.01	0.02	.538	0.02	0.03	.441	0.08	0.04	.036
Black	−0.02	0.04	.587	−0.01	0.11	.930	−0.049	0.05	.323	0.01	0.01	.314	−0.06	0.08	.425	−0.06	0.05	.275
Security x Black	0.02	0.03	.586	−0.16	0.11	.135	0.048	0.05	.358	−0.01	0.03	.752	−0.02	0.05	.679	−0.07	0.07	.319
Hispanic	0.03	0.03	.405	−0.03	0.05	.494	0.007	0.02	.711	0.05	0.03	.125	0.00	0.04	.950	0.00	0.04	.955
Security x Hispanic	−0.03	0.02	.179	−0.11	0.04	.012	−0.013	0.02	.595	−0.08	0.05	.110	−0.09	0.03	.002	−0.17	0.04	.000
Safety	0.04	0.04	.402	0.11	0.05	.026	0.107	0.09	.244	0.00	0.03	.853	0.04	0.06	.553	0.10	0.06	.112
Engagement	−0.09	0.08	.217	−0.13	0.06	.032	0.109	0.14	.448	−0.13	0.10	.190	−0.06	0.12	.610	0.00	0.07	.975
Environment	−0.03	0.10	.802	−0.07	0.14	.600	−0.221	0.18	.214	−0.06	0.15	.682	−0.14	0.17	.428	−0.15	0.17	.387
Wave 1 Behavior	0.18	0.10	.080	0.27	0.14	.062	0.180	0.20	.362	0.10	0.17	.548	0.04	0.08	.625	0.25	0.20	.216

However, these models only show a static picture of students’ engagement with school security measures at one time point, and are unable to assess the extent to which students’ engagement with security increased or decreased across the school year. To address this limitation, the next set of models uses change scores to examine how change over time in students’ engagement with school security measures is associated with behavior problems at the end of the school year. These results are presented in Panel B of Table 3. As shown, increases in engagement with security over the year was associated with less frequent involvement in problematic behaviors at the end of the year, including nonserious violent crime (b = −0.07, SE = 0.03, p = .034) and weapon-related crime (b = −0.05, SE = 0.02, p = .049). Notably, these relationships are in the opposite direction of the previous set of models, which indicates that increased engagement with security is associated with less frequent nonserious violent crime and weapon-related crime. These effects are smaller than in the first set of models; the regression coefficients indicate that a one unit increase in the change score is equal to decreases of approximately 14% of a standard deviation for nonserious violent crime and 21% of a standard deviation for weapon-related crime.

One limitation of these models is that they are unable to account for students’ prior behavior. To address this, the next set of models added a lagged measure of each dependent variable to control for students’ baseline levels of engagement in each of the behaviors. These results are presented in Panel C of Table 3. As shown, controlling for Wave 1 behavior yielded results that were substantively similar to those without this additional control variable. Increases in engagement with security across the year was associated with less frequent end-of-year involvement in nonserious violent crime (b = −0.05, SE = 0.02, p = .004) and weapon-related crime (b = −0.05, SE = 0.02, p = .046). This represents decreases of approximately 10% of a standard deviation for nonserious violent crime and 21% of a standard deviation for weapon-related crime.

The final set of models investigated differences by race and ethnicity, focusing on the experiences of both Black and Hispanic students. As noted above, we understand these analyses to point to systems of racial stratification, not essential features of various racial and ethnic groups. These models incorporated multiplicative interaction terms between the change score for engagement with school security measures and indicators of whether a student was Black or Hispanic. These results are presented in Panel D of Table 3. As shown, there were no significant interactions between the school security change score and the indicator of being Black. This suggests that the relationship between the school security change score and students’ engagement in problem behaviors was no different for Black students than it was for students who were non-Black, non-Hispanic. However, three models revealed significant interactions between the measures of school security change scores and the indicator of whether a student was Hispanic, showing that increased engagement with security measures was related to decreased involvement in behavior problems. This included models predicting nonserious violent crime (b = −0.11, SE = 0.04, p = .012), property crime (b = −0.09, SE = 0.03, p = .002), and disorder (b = −0.17, SE = 0.04, p < .001).

The interaction for the model predicting nonserious violent crime is displayed in Panel A of Figure 1. As shown, at one standard deviation below the mean school security change score, the point estimates showed that Hispanic students had a slightly higher frequency of nonserious violent crime. However, at one standard deviation above the mean, Hispanic students had a lower frequency of nonserious violent crime. This indicates that increasing students’ engagement with school security measures was associated with declines, which might be particularly true for Hispanic students’ frequency of nonserious violent crime. A similar pattern emerged for the model predicting property crime. At one standard deviation below the mean school security change score, Hispanic and White students were nearly identical in their frequency of property crime. However, at one standard deviation above the mean, Hispanic students engaged in less frequent property crime whereas White students engaged in more property crime.

Figure 1.

Interactions Between the School Security Change Score and Hispanic

Finally, the model predicting disorder was again quite similar. At one standard deviation below the mean school security change score, Hispanic and non-Black, non-Hispanic students were nearly identical in their frequency of disorder. However, at one standard deviation above the mean, Hispanic students engaged in less frequent disorder whereas non-Black, non-Hispanic students engaged in more frequent disorder. Together, these models presented in Panel D of Figure 3 suggest that increasing students’ engagement with school security measures was associated with small improvements in some behaviors, and that these effects might be particularly relevant to Hispanic students in urban settings.

Discussion

The current study explores the association between individual students’ engagement with school security and multiple forms of school violence perpetration. Using an IRT approach to measure students’ engagement with school security measures allowed us to operationalize engagement with all types of security measures as part of a single latent measure of engagement with school security. Key findings highlight small negative associations between engagement with school security and non-serious violent crime and weapons-related crime. While the school security change score and students’ engagement in problem behaviors was no different for Black students than it was for students who were non-Black, non-Hispanic, the negative association between engagement with security and behavior differed for Hispanic students relative to non-Black, non-Hispanic students. These findings make a unique contribution to the literature on school safety in that the constructs explored are specific to within-school differences in an urban school district, suggesting even when a school uses the same security measures, students engage with them differently and each student is likely influenced by security measures in a different way.

This study’s findings represent an extension of the school safety literature into a new, policy-relevant area. While researchers have examined between-school variation in engagement with school security measures, there has been limited empirical examination in how school security influences student behaviors differently within a school. The direction of the associations between engagement with security measures and student behavior supports the school criminalization perspective over opportunity theories of crime. This is particularly relevant for the relationships between non-serious and weapons offenses, while no relationship between serious violent crime existed. However, differences by race/ethnicity yield important information that should be considered. Specifically, when examining Hispanic students, some evidence of a deterrent effect emerged, as it was found that interaction with school security measures overall was associated with a small reduction in Hispanic students’ engagement in non-serious crime, disorder, and property crime, while the same effect did not exist for Black or White students. This provides some support for opportunity theories of crime, particularly among Hispanic students.

Although deterring problem behaviors might appear to be a desirable outcome for the Hispanic group within the context of routine activity theory, practical concerns of the school and current political environment must be considered. In particular, the current political environment has led to unique challenges for families of immigrants in the United States, particularly Hispanic families. It is estimated that 2654 immigrant children have been separated from their parents or caregivers as a result of Trump administration policies (ACLU, 2020). As a result, we have seen a more xenophobic climate emerge that has resulted in what might be perceived as a less welcoming country. This may explain why increased surveillance resulted in decreased non-violent behavior for the Hispanic students but not the Black or White students. That is, to the extent that some students are undocumented, increased surveillance may lead to changed behavior due to students fearing the consequences for them and their families. However, further research will need to explore the extent to which feeling surveilled in school might influence behavior, and how this relationship might vary for different students. Further research including both quantitative and qualitative methodologies would be beneficial. Quantitative research could focus on collecting data from large, representative samples to examine the within-school and within-student variation that might potentially drive how students access and engage with school security measures. Qualitative research can provide more in-depth context concerning the day-to-day nuances of how student engagement with security is perceived within the school context and how different racial and socioeconomic backgrounds might drive these perceptions. Qualitative discussion of these findings with students, parents, school administrators, and community stakeholders can help researchers highlight additional gaps in the knowledgebase concerning how security measures can be optimized to promote student connectedness alongside securing schools.

Another finding that is important to note is that there were higher reports of property crime amongst the White sample when compared to the African American and Hispanic subsamples. The differences in reporting property crime between these groups might be explained by differential selection hypothesis (Chambliss, 1994, 1995), which assumes that systematic and potentially oppressive factors influence the reporting of specific maladaptive and undesirable behaviors in school. In other words, differential deployment of security measures within a given school might implicitly work against racial minorities if they are received and interacted with differently, which therefore functions as the primary factor underlying Whites’ underrepresentation in more serious crime statistics (Piquero & Brame, 2008). If systematic oppression is attributable in any way to the disparate outcomes identified in this study, then it is fair to conclude that White students will have higher reports of non-serious behavior whereas racial or ethnic minority groups might have higher reports of more violent or undesirable behaviors, thus highlighting bias in how security and discipline is executed in schools. Future research should consider how implicit racial bias and systemic oppression factors into the accuracy of school records when comparing student-level outcomes concerning discipline and behavior across race and ethnicity.

The positive association between engagement with school security at Wave 1 and behavior at Wave 2, which was different in direction for Black students, might suggest racial disparities in the criminalization of student behavior (Hirschfield, 2008). While there are small effect sizes identified in the models, associations provide evidence to suggest that the criminalization of student behavior might be particularly problematic for minority students, builds on previous research that suggests schools with students of color are more likely to rely on school security measures (Servoss & Finn, 2014), and interaction with these measures tends to be associated with increased engagement in non-serious and non-violent behavior (Schreck & Miller, 2003). In sum, these findings suggest African American students are especially susceptible to the negative ramifications of security measures in urban schools, supported by student- and school-level data that have highlighted similar relationships (Homer & Fisher, 2020; Weisburst, 2019). However, policymakers and stakeholders should use caution when interpreting these findings due to the small effect sizes associated with the relationships identified.

Implications for Research

Research on the influence of security measures on student behavior has commonly used school-level data to examine and characterize differences. However, recent research focusing on student-level within-school differences in school security suggests that it would behoove scholars to consider the individualized effects that school security measures might have on students in school. As evidenced in this study, within-school differences in student-level outcomes (Cuellar & Coyle, 2021; Fisher et al., 2018), and differential engagement with school security may help researchers better understand the influence of school security measures on student behavior.

This study found that rather than treating school security measures as binary, using an ordinal scale was useful for capturing within-school differences in the extent to which students engage with security (as in Cuellar & Coyle, 2021). Researchers should consider operationalizing school security based on student engagement with security measures in their schools and explore the most effective ways of doing so. This approach can yield important information consistent with a person-in-environment perspective concerning school security, and can help delineate both student- and school-level trends in a variety of outcomes attributable to school security measures that are increasingly being employed.

Finally, future research should consider contextual factors that might moderate the relationships between security measures and student behavior. Current literature and the findings from the present study suggest that other contextual factors at the school-level (e.g., racial/ethnic composition; other school policies beyond security) might moderate the experiences of Black, Hispanic, and White students with security and their behavior. This is an important area of research that should be explored in future school-based studies.

Policy Implications

Results of the present study provide some evidence that school security measures might be ineffective in addressing maladaptive student behavior alone. However, in a time of heightened racial tensions in the United States, the primary recommendation from these data is to include culturally appropriate methods for assessing security needs and implementing programs to secure schools with appreciation of diversity and equity in mind. This appears particularly relevant to Hispanic students in our sample. While there were desirable reductions in student behavior as a result of interaction with school security among Hispanic students, this outcome may be offset by the fear and trauma they encounter with increased surveillance, particularly if immigration issues are a concern. Culturally-sensitive crime prevention initiatives and programs must be considered when administrators and educators discuss programs to be implemented in their school. Planning should be done with a multicultural approach that is inclusive to the specific needs of the student body in each school. When planning programs to secure their school, administrators and educators must connect with students, parents, and community stakeholders to consult how approaches to security might be received and how they might affect students over time. Community-based participatory frameworks might be helpful in guiding an inclusive approach to assessing security needs and implementing effective programs aimed at keeping youth safe. More generally, crime prevention efforts should focus on supporting and engaging students rather than increasing formal surveillance and treating them as potential criminals. Schools are likely to benefit from including the voices of multiple populations when making decisions about their crime prevention strategies. Finding ways to meaningfully solicit and listen to perspectives from marginalized groups may be particularly useful. This may include students, parents, community members and organizations, and a variety of other stakeholders. Democratizing the decision-making process around school violence prevention is likely to improve buy-in from a wide variety of individuals and also increase the anticipation of unintended negative consequences that have followed schools’ use of security measures in particular. In a national climate where police-free schools are appearing nationwide, the time may be right for a broader reconsideration of how school security measures and other school violence prevention initiatives are used.

Limitations

Existing measures such as the MDS3 and the SSOCS were used to operationalize school climate, student behavior, and school security. However, these constructs have not been assessed for construct validity and it is unclear as to the convergence and divergence of items operationalizing the various constructs on the survey, with the exception of the MDS3, which has been piloted and used extensively across the literature. In regard to external validity, there is a significant lack of generalizability as a result of the nonprobability sampling method and the use of a single school district. Moreover, demographics suggest differences in the sample when compared to the sampling frame.

Additionally, as noted above, although this study examined differences across racial and ethnic groups in an effort to operationalize broad social systems of racial stratification, it relied on limited information that risks oversimplifying social structures. Racial stratification in the United States is deep-seated and complex, and students’ position in this system cannot be adequately measured using the data available to us in the survey. Similarly, these data were unable to capture the variability within racial and ethnic groups in relation to their experiences with both school security measures and behavior problems in school. Finally, this study was focused on quantitative survey data. Such a deductive approach fails to yield detailed accounts pertaining to the lived experiences of students in schools surveyed. More in-depth qualitative information might yield a different picture and provide information that we don’t know regarding how various school security measures were implemented and used.

Finally, policymakers and stakeholders must use caution in interpreting the practical significance of the associations identified due to many of the small effect sizes referenced. While this study was exploratory, these effect sizes suggest that some of the differences that emerged might not carry practical significance. While the associated p-values in the analyses suggest that the differences are likely not by chance alone, the magnitude of the difference should be understood and carefully interpreted when applying these findings to program development, implementation, and assessment.

Conclusion

Although prior research has addressed the link between the presence of school security measures and student behavior, research has not yet examined the extent to which the frequency of students’ engagement with school security measures relates to outcomes related to student behavior as it might differ from student-to-student within the same school. This study focused on these student-level differences in behavioral outcomes attributable to student engagement with school security measures. Findings suggest that engagement with school security varies at the student-level among students who attend the same school. More specifically, findings suggest there might be racial and ethnic differences in the relationship between school security and non-serious violence and weapon-related crime. With this in mind, long-term programming goals should be established when developing process for securing schools with emphasis on how security measures might influence individual students differently within the school setting. Future research should consider findings of the current study when assessing the effectiveness of school security measures in addressing student problem behaviors in school.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Matthew J. Cuellar

Author Biographies

Matthew J. Cuellar, PhD, is an Assistant Professor in the School of Social Work at University of Alaska Anchorage. His research focuses on school security and student connectedness and performance in urban school systems, with a particular focus on addressing racial and ethnic disparities in student-level outcomes attributable to school security.

Benjamin W. Fisher, PhD, is an Assistant Professor in the Department of Criminal Justice at Florida State University. His research interests center on school criminalization, with a particular focus on issues of equity. His research has addressed school resource officers, school security, exclusionary discipline, perceptions of safety at school, school climate, and inequality in these areas.

References

ACLU (2020). Immigrants’ rights and detention. Retreived from https://www.aclu.org/issues/immigrants-rights/immigrants-rights-and-detention

Addington

(2009). Cops and cameras: Public school security as a policy response to columbine. American Behavioral Scientist, 52(10), 1426–1446. https://doi.org/10.1177/0002764209332556

Allison

P. D.

(2001) Missing data (Vol. 136). Sage publications.

Boynton-Jarrett

Hair

Zuckerman

(2013). Turbulent times: Effects of turbulence and violence exposure in adolescence on high school completion, health risk behavior, and mental health in young adulthood. Social Science & Medicine, 95, 77–86. https://doi.org/10.1016/j.socscimed.2012.09.007

Bradshaw

Waasdorp

Debnam

Johnson

S. L.

(2014). Measuring School Climate in High Schools: A Focus on Safety, Engagement, and the Environment. Journal of School Health, 84(9), 593–604. https://doi.org/10.1111/josh.12186

Brown

(2006). Understanding and assessing school police officers: A conceptual and methodological comment. Journal of Criminal Justice, 34(6), 591–604. https://doi.org/10.1016/j.jcrimjus.2006.09.013

Brown

Benedict

W. R.

(2005). Classroom cops, what do the studentsthink? A case study of student perceptions of school police and security officers conducted in an Hispanic community. International Journal of Police Science & Management, 7(4), 264–285. https://doi.org/10.1350/ijps.2005.7.4.264

Chambliss

W. J.

(1994). Policing the ghetto underclass: The politics of law and law enforcement. Social Problems, 41(2), 177–194. https://doi.org/10.1525/sp.1994.41.2.03x0433q

Chambliss

W. J.

(1995). The institutionalization of racism through law. In Hawkins

(Ed.), Race, ethnicity, and crime (pp. 235–258). State University of New York Press.

10.

Cohen

L. E.

Felson

(1979). Social change and crime rate trends: A routine activity approach. American sociological review, 44(4), 588–608. https://doi.org/10.2307/2094589

11.

Cuellar

M. J.

(2016). School safety strategies and their effects on the occurrence of school-based violence in United States high schools: An exploratory study. Journal of School Violence, 29(2), 1–18. https://doi.org/10.1080/15388220.2016.1193742

12.

Cuellar

M. J.

Coyle

(2021). Assessing disparities in school safety: implications for promoting equality in current efforts to keep kids safe. Security Journal, 34, 658–684. https://doi.org/10.1057/s41284-020-00254-2

13.

Cuellar

Coyle

Weinreb

(2021). Dealing with the day-to-day: harnessing school climate to address the effects of student victimization on academic performance. Psychology in the Schools , 58(9), 1799–1815. https://doi.org/10.1002/pits.22560

14.

de Ayala

R. J.

(2009). The theory and practice of item response theory. Guilford.

15.

Devine

(1996). Maximum security: The culture of violence in inner-city schools. University of Chicago Press.

16.

Devlin

D. N.

Gottfredson

D. C.

(2018). The roles of police officers in schools: Effects on the recording and reporting of crime. Youth Violence and Juvenile Justice, 16(2), 208–223. https://doi.org/10.1177/1541204016680405

17.

Eisenbraun

(2007). Violence in schools: Prevalence, prediction, and prevention. Aggression and Violent Behavior, 12(4), 459–469. https://doi.org/10.1016/j.avb.2006.09.008

18.

Fisher

B. W.

Hennessy

E. A.

(2016). School resource officers and exclusionary discipline in U.S. high schools: a systematic review and meta-analysis. Adolescent Research Review, 1(3), 217–233. https://doi.org/10.1007s40894-015-0006-8

19.

Fisher

B. W.

Mowen

T. J.

Boman

J. H.

(2018). School security measures and longitudinal trends in adolescents’ experiences of victimization. Journal of Youth and Adolescence, 47(6), 1221–1237. https://doi.org/10.1007/s10964-018-0818-5

20.

Gastic

(2011). Metal detectors and feeling safe at school. Education and Urban Society, 43(4), 486–498. https://doi.org/10.1177/0013124510380717

21.

Gastic

Johnson

(2014). Disproportionality in daily metal detector student searches in U.S. public schools. Journal of School Violence, 14(3), 299–315. https://doi.org/10.1080/15388220.2014.924074

22.

Gawley

Cuellar

M. J.

Coyle

(2021). A theoretical and empirical assessment of authoritarianism’s effects on behavior, attendance, and performance in urban school systems. Contemporary Justice Review, 24(2), 197–217. https://doi.org/10.1080/10282580.2021.1881894

23.

Gillborn

Warmington

Demack

(2018). QuantCrit: education, policy, ‘Big Data’and principles for a critical race theory of statistics. Race Ethnicity and Education, 21(2), 158–179. https://doi.org/10.1080/13613324.2017.1377417

24.

Gottfredson

D. C.

Crosse

Tang

Bauer

E. L.

Harmon

M. A.

Hagen

C. A.

Greene

A. D.

(2020). Effects of school resource officers on school crime and responses to school crime. Criminology & Public Policy, 19(3), 905–940. https://doi.org/10.1111/1745-9133.12512

25.

Hankin

Hertz

Simon

(2009). Impacts of metal detector use in schools: Insights from 15 years of research. Journal of School Health, 81(2), 100–106. https://doi.org/10.1111/j.1746-1561.2010.00566.x

26.

Hannon

(2002). Criminal opportunity theory and the relationship between poverty and property crime. Sociological Spectrum, 22(3), 363–381. https://doi.org/10.1080/02732170290062676

27.

Hirschfield

P. J.

(2008). Preparing for prison? The criminalization of school discipline in the USA. Theoretical Criminology, 12(1), 79–101. https://doi.org/10.1177/1362480607085795

28.

Homer

E. M.

Fisher

B. W.

(2020). Police in schools and student arrest rates across the United States: Examining differences by race, ethnicity, and gender. Journal of School Violence, 19(2), 192–204. https://doi.org/10.1080/15388220.2019.1604377

29.

Johnson

I. M.

(1999). School violence: The effectiveness of a school resource officer program in a southern city. Journal of Criminal Justice, 27(2), 173–192. https://doi.org/10.1016/s0047-2352(98)00049-x

30.

Kupchik

Ellis

(2008). School discipline and security: Fair for all students?. Youth & Society, 39(4), 549–574. https://doi.org/10.1177/0044118x07301956

31.

Kupchik

Monahan

(2006). The new American school: Preparation for post‐industrial discipline. British Journal of Sociology of Education, 27(5), 617–631. https://doi.org/10.1080/01425690600958816

32.

Kupchik

Ward

(2014). Race, poverty, and exclusionary school security: An empirical analysis of US elementary, middle, and high schools. Youth Violence and Juvenile Justice, 12(4), 332–354. https://doi.org/10.1177/1541204013503890

33.

Livingston

Rossheim

Stidham Hall

(2019). A descriptive analysis of school and school shooter characteristics and the severity of school shootings in the United States, 1999-2018. Journal of Adolescent Health, 64(6), 797–799. https://doi.org/10.1016/j.jadohealth.2018.12.006

34.

Mowen

T. J.

Parker

K. F.

(2017). Minority threat and school security: Assessing the impact of Black and Hispanic student representation on school security measures. Security Journal, 30(2), 504–522. https://doi.org/10.1057/sj.2014.42

35.

Musu-Gillette

Zhang

Wang

Zhang

Oudekerk

B. A.

(2017). Indicators of school crime and safety: 2016. National Center for Education Statistics, US Department of Education, and Bureau of Justice Statistics, Office of Justice Programs, US Department of Justice.

36.

Musu

Zhang

Wang

Zhang

Oudekerk

B.A.

(2019). Indicators of school crime and safety: 2018 (NCES 2019-047/NCJ 252571). National Center for Education Statistics, U.S. Department of Education, and Bureau of Justice Statistics, Office of Justice Programs, U.S. Department of Justice.

37.

Gottfredson

D. C.

(2013). Police officers in schools: effects on school crime and the processing of offending behaviors. Justice Quarterly, 30(4), 619–650. https://doi.org/10.1080/07418825.2011.615754

38.

Nolan

(2011). Police in the hallways: Discipline in an urban high school. U of Minnesota Press.

39.

O’Keefe

(1997). Adolescents’ exposure to community and school violence: Prevalence and behavioral correlates. Journal of Adolescent Health, 20(5), 368–376. https://doi.org/10.1016/S1054-139X(97)80131-0

40.

Owens

E. G.

(2017). Testing the school‐to‐prison pipeline. Journal of Policy Analysis and Management, 36(1), 11–37. https://doi.org/10.1002/pam.21954

41.

Piquero

A. R.

Brame

R. W.

(2008). Assessing the race-crime and ethnicity-crime relationship in a sample of serious adolescent delinquents. Crime and Delinquency, 54(3), 390–422. https://doi.org/10.1177/0011128707307219

42.

Preacher

K. J.

Curran

P. J.

Bauer

D. J.

(2006). Computational tools for probing interaction effects in multiple linear regression, multilevel modeling, and latent curve analysis. Journal of Educational and Behavioral Statistics, 31(4), 437–448. https://doi.org/10.3102/10769986031004437

43.

Reingle Gonzalez

Jetaline

Jennings

(2016). Structural school safety measures, SROs, and school-related delinquent behavior and perceptions of safety. Policing: An International Journal of Police Strategies and Management, 39(3), 438–454. https://doi.org/10.1108/pijpsm-05-2016-0065

44.

Schreck

Miller

(2003). Sources of Fear of Crime at School. Journal of School Violence, 2(4), 57-79. https://doi.org/10.1300/J202v02n04_04

45.

Servoss

Finn

(2014). School security: For whom and with what results?. Leadership and Policy in Schools, 13(1), 61–92. https://doi.org/10.1080/15700763.2014.890734

46.

Shedd

(2015). Unequal city: Race, schools, and perceptions of injustice. Russell Sage Foundation.

47.

Staff

Kreager

(2008). Too Cool for School? Violence, Peer Status and High School Dropout. Social Forces, 87(1), 445–471. https://doi.org/10.1353/sof.0.0068

48.

Steinka-Fry

K. T.

Fisher

B. W.

Tanner-Smith

E. E.

(2016). Visible school security measures across diverse middle and high school settings: Typologies and predictors. Journal of Applied Security Research, 11(4), 422–436. https://doi.org/10.1080/19361610.2016.1210482

49.

Tanner-Smith

E. E.

Fisher

B. W.

(2016). Visible school security measures and student academic performance, attendance, and postsecondary aspirations. Journal of Youth and Adolescence, 45(1), 195–210. https://doi.org/10.1007/s10964-015-0265-5

50.

Tanner-Smith

E. E.

Fisher

B. W.

Addington

L. A.

Gardella

J. H.

(2018). Adding security, but subtracting safety? Exploring schools’ use of multiple visible security measures. American Journal of Criminal Justice, 43(1), 102–119. https://doi.org/10.1007/s12103-017-9409-3

51.

Theriot

M. T.

Cuellar

M. J.

(2016). School resource officers and students’ rights. Contemporary Justice Review, 19(3), 363–379. https://doi.org/10.1080/10282580.2016.1181978

52.

Toldson

I. A.

(2012). Editor’s Comment: Insecurity at Black Schools: When Metal Detectors Do More Harm Than Good. The Journal of Negro Education, 81(4), 303–306. https://doi.org/10.7709/jnegroeducation.81.4.0303

53.

Turanovic

Pratt

Kulig

Cullen

(2019). Individual, institutional, and community sources of school violence: A meta-analysis. NCJRS. https://www.ncjrs.gov/pdffiles1/nij/grants/253934.pdf

54.

Weisburst

E. K.

(2019). Patrolling public schools: The impact of funding for school police on student discipline and long‐term education outcomes. Journal of Policy Analysis and Management, 38(2), 338–365. https://doi.org/10.1002/pam.22116

55.

Wolf

(2014). Arrest decision making by school resource officers. Youth Violence & Juvenile Justice, 12(2), 137–151. https://doi.org/10.1177/1541204013491294

56.

Yang

F. M.

Kao

S. T.

(2014). Item response theory for measurement validity. Shanghai Archives of Psychiatry, 26(3), 171–177. https://doi.org/10.3969/j.issn.1002-0829.2014.03.010