Do Interactions With School Resource Officers Predict Students’ Likelihood of Being Disciplined and Feelings of Safety? Mixed-Methods Evidence From Two School Districts

Criminalization Model

Evidence on the relationship between students’ interactions with SROs and discipline outcomes is still evolving but has tended to find that the presence of SROs can result in larger numbers of students subject to school discipline (Fisher & Hennessy, 2016; Nance, 2016; Weisburst, 2019). For example, a recent meta-analysis by Fisher and Hennessy (2016) examined whether the mere presence of SROs shaped rates of school discipline in studies examining high schools, and found that, in one of their two models, the presence of SROs increased rates of school discipline (out-of-school suspensions, expulsions, arrests, and reported crimes; see also Nance, 2016; Theriot, 2016). A similar increase in out-of-school suspension rates among middle schoolers has also been found in work that leveraged plausibly exogenous variation in SRO placement that resulted from federal grants (Weisburst, 2019). Given that research consistently finds that students of color are considerably more likely than others to feel the brunt of school discipline, this criminalizing effect of SROs exacerbates racial inequality in school discipline and access to education (Jacobsen et al., 2019).

However, these studies’ findings must be understood in terms of two broad limitations. First, adding SROs to a school may increase the detection and punishment of student misbehavior, even if the overall behavior in the school improves or remains the same. Relying on official records of behaviors or punishments makes this threat to validity nearly impossible to disentangle. Second, studies comparing schools that have SROs to schools that do not may suffer from selection bias. It is likely that schools that decide to add SROs are systematically different from schools that do not, particularly in the degree to which they have perceived problems with crime or misbehavior. Thus, these studies’ findings may be shaped by characteristics of schools in addition to the presence of SROs. Research that examines within-school differences based on levels of exposure to SROs may help alleviate this methodological concern.

That said, qualitative work, particularly school-based ethnographies that critically examine SROs’ presence, help explain why SRO presence might increase disciplinary outcomes. Nolan (2011), for example, has shown how SROs can escalate situations with students, whereby SROs’ efforts to help patrol hallways and maintain school rules are interpreted as harassing by students, who respond in ways that sometimes result in school discipline or arrest (see also Mukherjee, 2007). Kupchik (2010) found the presence of SROs can subtly shift schools’ foci somewhat away from students’ social and emotional well-being and toward a law enforcement perspective, resulting in greater recognition of misbehavior as rule violations or illegal. Shedd’s (2015) analysis of policing in Chicago high schools considered what lessons youth who interact with police in schools learn from these experiences. She described how youth learn in schools to expect to be policed rigorously and often even harassed (or “told off”), resulting in the view that policing is procedurally unjust and coercive in a way that endangers, not protects, their security. Such work suggests SROs’ presence in schools creates a shift toward viewing children—particularly children of color—as potential threats and/or criminals (Fisher et al., 2020; Kupchik, 2010; Mukherjee, 2007; Nolan, 2011).

These studies align with work on the criminalization of youth (Hirschfield, 2008; Kupchik, 2010; Shedd, 2015) as well as the concepts of net-widening and, relatedly, net-deepening (Cohen, 1985; Irby, 2013, 2014; Van Dusen, 1981). Research on policing and youth of color outside of school (e.g., Durán, 2009; Gau & Brunson, 2015; Rios, 2011) finds frequent conflict between these two groups, with youth reporting feeling harassed and treated abusively. This may explain in part why, nationwide, schools with police have significantly higher arrest rates of Black students than White or Hispanic students (Homer & Fisher, 2020). Net-widening suggests that the presence of police in schools may result in more students coming into contact with law enforcement and thereby having their behaviors interpreted as a violation of a rule. Net-deepening suggests that the responses for such misconduct may, as a result of police presence, be made increasingly punitive (Irby, 2013, 2014). Addressing this phenomenon, Shedd (2015) writes, “My discussions with students suggested that the presence of police is not making them feel safer. Instead, it makes them feel imprisoned” (p. 93).

In line with this view, a parallel set of studies have directly considered the outcome of students’ feelings of safety at school and whether they are shaped by the presence of SROs. Several studies have considered how security strategies, broadly (but including SROs) shape student perceptions of safety. These studies have tended to find that more security predicts greater student fear (Bachman et al., 2011; Perumean-Chaney & Sutton, 2013). This result was confirmed by a review of 32 studies from both primary and secondary schools, showing that more safety measures are associated with poorer student perceptions of safety (Reingle Gonzalez et al., 2016), though, as explained in the next section, this relationship did not hold specifically for SROs.

A criminalization model, based on this critical literature, predicts negative outcomes to arise from police in schools, particularly for youth of color. This applies directly to our analysis by predicting the following results from our quantitative analysis: (a) interactions with SROs will relate to greater likelihood that a student is disciplined in school, (b) students with more interactions with SROs will be more fearful of their safety, and (c) these negative results will be found mostly among youth of color. In the qualitative analysis, the criminalization model suggests that we will find accounts of SROs engaging in school discipline and students internalizing fear or a sense of threat as a result of SRO interactions and that these narratives may be most pronounced among students of color.

Support Model

In contrast, the support model posits that SROs are an asset to schools and serve to reduce problem behaviors and threats of violence while increasing perceptions of safety. While the bulk of the evidence to date supports a criminalization model instead, it is important to note that prior research on outcomes of SROs is somewhat inconsistent and limited, with some research showing null or even positive and supportive effects of SROs (Johnson et al., 2018; Owens, 2017). For example, Owens (2017) found that SROs reduced school administrators’ reports of disruptive criminal incidents in school, and Johnson et al. (2018) found SROs’ presence to be related to greater perceptions of safety among students. Other recent research found that consequences of SROs’ presence (e.g., school crime rates, referrals to the police) vary based on how SROs interact with students, suggesting that SROs can be supportive if they focus more on mentoring roles as opposed to law enforcement (Devlin & Gottfredson 2018; Fisher & Devlin, 2020). Furthermore, research has found that the link between SROs and the criminalization of youth tends to occur primarily with low-income youth of color (e.g., Lynch et al., 2016; Nolan, 2011) who may arrive at school with very different experiences with and perceptions of police compared with other students. Although several studies have found negative outcomes of SROs, variability in student body demographics and in how SROs interact with students prevents us from generalizing to all SROs and all schools or from assessing whether these negative outcomes are produced by SROs’ interactions in schools as opposed to mirroring previously existing distrust and negative experiences in communities. As a result, whether SROs’ actions align with a support model or criminalization model may vary based on the school context and demographics of students served.

With regard to perceptions of safety, other work has found, in contrast to other security practices, positive relationships between the presence of SROs and perceptions of safety (Reingle Gonzalez et al., 2016) and that students who report better perceptions of SROs feel safer and have fewer disciplinary incidents (Penteck & Eisenberg, 2018; Theriot & Orme, 2016). It is possible that the net effect of SROs’ presence is high levels of support for students and protection from negative outcomes (e.g., school punishment, feeling unsafe), a scenario consistent with the description of SROs by advocates such as the National Association of School Resource Officers (Canady et al., 2012).

This may be particularly the case in certain contexts, such as the relatively affluent, suburban schools examined in this study. Indeed, Hirschfield (2010) theorized that schools’ use of security measures—including SROs—is likely to vary across contexts in the extent to which they serve more inclusionary or exclusionary functions. In particular, schools in more affluent areas are likely to be more inclusive in how school security measures are used, focusing on keeping out external threats rather than on monitoring students’ behaviors. Empirical research on SROs largely supports this hypothesis. One nationally representative study of schools found that SROs are more likely to be engaged in educational activities in schools with lower levels of disadvantage (Lynch et al., 2016). Another study examined how schools’ racial composition shapes SROs’ perceptions of the main threats to their school (Fisher et al., 2020). Drawing on interviews with SROs across two school districts, this study found SROs in schools with small proportions of non-White students viewed external threats (e.g., intruders) as the main threat to the school (Fisher et al., 2020). These studies suggest that context shapes how SROs do their job with suburban and rural SROs acting in ways that might be more indicative of the support model (Devlin & Gottfredson, 2018; Fisher et al., 2020). In this way, suburban and rural school SROs might be unlikely to engage in school discipline (Fisher et al., 2020; Lynch et al., 2016).

As we note later in our description of the study’s setting, the schools in this study expanded the presence of SROs as a direct response to the Sandy Hook shooting and for the primary purpose of protecting against external threats to the schools. As a part of their established relationship with the school systems, the SROs are prohibited by policy from engaging in school discipline. We describe elsewhere (Curran et al., 2019) that there is nuance in how this policy is interpreted and implemented, but this is a noteworthy factor given that nearly two thirds of schools with SROs nationwide lack written agreements with law enforcement agencies specifying the role of SROs in discipline (Correa & Diliberti, 2020). The setting of this study may thus provide a test of outcomes associated with SROs under a “best-case” scenario—one in which the stakeholders view SROs as a safety mechanism rather than disciplinarians, attempting to mitigate potential increases in student discipline through explicit policy.

A support model thus predicts positive outcomes to arise from police in schools and from students’ interactions with those SROs. Specifically, this model applies to our quantitative analysis by predicting that (a) interactions with SROs reduce or at least do not increase students’ risk of involvement in school discipline, and (b) students with more interactions with SROs will be less fearful in school. This model suggests that, in the qualitative data, we will find evidence of accounts of positive interactions between students and SROs, cases in which SROs deescalate disciplinary situations, and accounts of how SROs increase feelings of safety. A support model does not presume that SROs interact differently with different students, and thus offers no prediction for whether positive or negative results will be found mostly among youth of color or White youth.

Combining Frameworks

Our theoretical contribution is an application of the competing criminalization and support models within the context of a “dosage” framework. In other words, we draw on conceptual models that suggest that the degree to which the criminalizing or supportive functions of SROs are felt by students is a function of the frequency of interaction with, feelings of trust in, and comfort with SROs experienced by students. This conceptual framing aligns with program evaluation literature that focuses on “dose-response,” or the idea that outcomes are responsive to the amount of an intervention experienced (Rossi et al., 2019; Rowbotham et al., 2019). Similarly, economic theory uses the concepts of extensive and intensive margins to refer to whether individuals are working (extensive margin) and how much they work (intensive margin; Blundell et al., 2013). Extending to the case of SROs in schools, we argue that the effects of the criminalization or support models can be conceptualized as varying both as a function of whether an SRO is present (extensive margin) but also, relatively less studied, how often and in what ways SROs interact with students (intensive margin). Furthermore, this dosage impact may potentially vary based on the demographics of individual students, with a differential impact possible for students of color and other marginalized groups. This logic may also extend to differences across settings; SROs in schools with larger proportions of racial/ethnic minority students may more closely align with the criminalization model than SROs in schools with larger proportions of White students (Fisher et al., 2020; Hirschfield, 2010). In summary then, our theoretical framework posits the possibility of either positive or negative associations that are responsive to the dosage (both frequency and trust/comfort) of interactions with SROs and that may vary based on student characteristics such as race/ethnicity.

Contribution

Building on the existing body of literature, the current study makes three primary contributions. First, the study examines how differential frequency of interactions and trust/comfort with SROs relates to disciplinary and safety outcomes, moving beyond prior studies that have tended to focus only on the presence or lack thereof of an SRO in a school. A handful of recent exceptions to this include a study which found that students with poorer perceptions of SROs tend to experience more school discipline and feel less safe at school (Penteck & Eisenberg, 2018; Theriot, 2016). However, the bulk of existing research on the outcomes associated with SROs focuses on between-school differences, comparing outcomes across either (a) schools with and without SROs or (b) schools with SROs engaged in different roles. Much less is known about how SROs may differentially affect students within the same school and the extent to which students’ level of exposure to and comfort/trust in SROs might shape this.

Second, this study examines settings—including elementary schools and nonurban schools—that have been the subject of little prior research on SROs, thereby enhancing the scope of SRO research to settings that are experiencing the greatest increases in SRO presence. At a national level, both schools in suburban and rural settings report less use of serious disciplinary actions (36% and 30%, respectively) as compared with cities and towns (40% and 50%, respectively; Musu-Gillette et al., 2018) though racial disparities in use of discipline remain significant in both (Pearman et al., 2019). SRO presence, in particular, is associated with particularly high out-of-school suspension rates in high schools with a strong zero tolerance orientation to discipline coupled with a large proportion of non-White students (Fisher, 2016). Thus, further understanding the nature of SROs’ interactions with students and students’ trust and comfort in SROs may help illuminate reasons for the heterogeneous impacts of SROs on discipline and perceived safety (Fisher & Hennessy, 2016; Reingle Gonzalez et al., 2016). Considering that prior research finds that students of color can be more likely than White students to see police officers as threats to their safety (e.g., Rios, 2011; Shedd, 2015), it seems likely that such a better understanding of students’ perceptions of, trust, and comfort in SROs may be crucial for disentangling why prior research on consequences of SROs varies. Thus, the roles and impacts of SROs may be driven, at least in part, by their school’s setting. Using data from a more geographically diverse setting with SROs in elementary schools therefore has the potential to significantly expand our understanding of SROs in previously understudied settings.

Finally, by integrating both qualitative and quantitative data, the study provides unique insights into the mechanisms by which SROs may influence student outcomes, particularly the ways in which their presence may have both positive and negative consequences. It is not common in prior research on SROs to use a mixed-methods design. For example, prior literature has tended to understand student perceptions of safety as a linear, unidirectional outcome. Prior studies have hypothesized (but not tested) a more complex relationship: that the ubiquity of SROs socializes students to expect both tight security and threat of danger (Kupchik, 2010). This hypothesis complicates the presumed relationship between SROs’ presence, interactions with students, and students’ level of fear as it suggests that SROs may simultaneously act as a comfort (as they are expected to enhance safety) and an indicator of danger (as they are presumed to be needed in schools due to imminent threats). Work by Theriot (2016) found that the frequency of students’ interactions with SROs shaped their perceptions in complex ways; students who interacted more with SROs had more positive attitudes about SROs but were less connected to their schools. By analyzing quantitative and qualitative data, we are able to explore such complicated relationships in greater depth than prior studies and provide insights into why such relationships may exist.

Context

Our data came from a larger mixed-methods research study focusing on the implementation of SROs across all elementary schools in a single county in the southeastern United States (all data archived in the National Archive of Criminal Justice Data). This study also examines questions of school climate and safety as well as school disciplinary practices across all grade levels. The county includes two school districts with one school district covering the local county seat, Washington City Schools, and the other school district covering the remainder of the county area, Fairfield County Schools (pseudonyms). The two districts have student feeder patterns such that many students will attend both districts over the course of their K–12 career. As a result, the two school districts operate in tandem for many policy decisions, including that of placing SROs in schools.

We found this cooperation to be present with regard to school safety, with the safety coordinators of each district in regular contact with each other, regularly making many decisions about safety protocols in conjunction with each other. Beginning in the early 2000s, deputies from the county Sheriff’s Department and local police were placed as SROs in middle and high schools. After the shooting at Sandy Hook, SROs were expanded to all schools in the districts, and the Sheriff’s Department became the sole provider of SROs to the schools. Consequently, all SROs in both districts were trained, staffed, and supervised by the county Sheriff’s Department. Individual SROs were placed in specific schools by the law enforcement agency, with no input from school personnel including principals. All SRO personnel matters including school placement, transfers, and work schedules were determined solely by the law enforcement agency. The relationship between the Sheriff’s Department and the school districts was guided by memorandums of understanding (MOUs). These MOUs stated that SROs in these districts were prohibited from being involved in school discipline (although this term was not defined in the MOUs). Given that the Sandy Hook shooting was the primary motivator for expanding SROs to all schools in the districts, SROs generally perceived their top priority as preventing external threats from entering the school. Secondary to this goal, however, was a goal of building relationships with students and trust between students and police (Curran et al., 2020).

While the use of two districts may provide some increase in generalizability, the tight coupling of the two districts and the staffing of SROs by a single law enforcement agency means that we effectively treat the two districts as a single district in our analysis. While we show next how this setting mirrors many of those nationally that have experienced increases in SROs, we also note in our limitations that the study of a single county affects the generalizability of results.

Table 1 provides descriptive statistics on the districts. Fairfield County Schools cover a range of urbanicities from fringe rural to small city settings. On average, the schools in this study had 80% (all percentages rounded to nearest 10) White students with a range of 40% to 90% across all schools. The schools with the largest Hispanic populations were 40% Hispanic, and schools with the largest African American populations were 30% African American. Around 65% of schools in Washington City Schools qualified as Title 1 compared with 15% of county schools. In short, the schools in the sample represented a range of urbanicities, racial/ethnic makeup, and socioeconomic composition. Politically, the county is strongly conservative, with approximately 65% of voters (rounded for confidentiality) voting Republican in the most recent election (i.e., November 2020). The county’s demographics—generally suburban, more affluent, and relatively White—were similar to the setting of a number of high-profile school shootings, including Columbine and Sandy Hook. This is particularly relevant given that nonurban schools have seen some of the greatest expansion of SRO presence in the past two decades, making the setting of this study similar to locations that have both motivated the growth in SROs and seen the greatest increases in SRO presence.

OPEN IN VIEWER

Table 1

Descriptive Statistics on School Districts

Key independent variables	Fairfield County Schools	Washington City Schools
Number of schools	40 a	10 a
Enrollment	37,000 b	3,000 b
Percent African American	5% c	15% c
Percent Hispanic	5% c	25% c
Percent of students eligible for free or reduced price lunch	10% c	40% c
Percent of Title 1 eligible schools	15% c	65% c
Per-pupil expenditures	US$9,000 b	US$14,000 b

Source. The state’s Department of Education and Common Core of Data.

a

Rounded to the nearest ten. ^bRounded to the nearest thousand. ^cRounded to the nearest five.

Quantitative Data

Qualitative Data

In addition to surveys, we conducted group interviews with students across Washington City Schools for students in Grades 4 to 8. The group interviews were co-conducted by two facilitators, when possible, using a semi-structured protocol. The interview protocol asked questions about school climate, discipline, and the school’s SRO. A copy of the protocol is included as Supplemental Appendix A in the online version of the journal. As shown, questions were organized under a series of topical headings that were deemed relevant from existing literature and from the specific research questions of the study. Questions were designed to elicit conversation and dialogue about a topic, to be open-ended enough to allow emergent topics and themes to emerge, and included a number of probes to prompt respondents for specific examples or deeper explanation.

The semi-structured interview format was purposefully chosen to both elucidate discussion of predetermined topics while also leaving flexibility for discussion of emergent topics and themes. As such, group interview facilitators allowed conversation to flow among participants and followed up with additional probes and redirects to participant responses as appropriate. While facilitators strove to cover all questions/topics on the interview protocol in the course of the interview, the order in which topics were addressed and the time spent discussing particular topical areas was responsive to the salience of particular topics to the group and the flow of the conversation. All group interviews were conducted by members of the research team who were familiar with and helped develop the interview protocol.

At each elementary and middle school, the school’s administration selected the sample of students to include in the group interview. Each school reported that they selected a certain class, grade level, or after-school activity, and then passed out consent forms to all students in that group. We found no evidence that specific students were selected by the schools to participate although this could have taken place without the knowledge of the research team. The school administration reported that these groups were often selected based on availability to participate during noninstructional time. In elementary schools, we had only fourth and fifth grade students participate to maximize comprehension of the questions. The final sample of students included students who both turned in a parent consent form and were in school the day we conducted the group interview. While we do not have data on prior disciplinary histories of students who participated in interviews, it is likely, given the selection mechanism, that they were less likely to have prior disciplinary involvement than students on average. We conducted nine group interviews that each included between 2 and 10 participants. Although demographic data of group interview participants were not collected, participants generally represented the broader school context insofar as they were predominantly White and were fairly evenly split between males and females. All group interviews were audio recorded and transcribed for analysis.

Researcher Positionality

Before turning to our analytic strategy, we make mention of our positionality as researchers as it relates to the data collection and analysis. The lead author of this study is a White male who works as a professor studying education policy in an institution of higher education. He comes from a relatively affluent background and attended public K–12 schools in settings similar to that of the research site. As such, there was strong congruence between his socioeconomic background and race and that of many of the participants. Two of the authors are former public school teachers, and their work is often conducted through the perspective of an educator, a fact that was sometimes shared with participants.

Several of the group interviews had graduate assistants as facilitators. The racial/ethnic and socioeconomic background of these graduate assistants varied. Although most had previous experience working for schools, few had any established relationships with the county under study. While the general alignment between researchers and participants was generally seen as an asset for putting participants at ease and increasingly their likelihood of sharing candidly, we nevertheless recognize that certain participants, particularly those who did not share these characteristics of the dominant group, may have interacted differently as a result. Likewise, we recognize that our presentation of results and interpretation of their meaning are inherently filtered through our collective background and professional lenses.

Quantitative Analysis

This study used cross-sectional survey data to explore if the frequency of interactions with, trust in, and comfort with SROs is related to students’ disciplinary outcomes or feelings of safety. For the main set of results, we estimated fixed effects models using logistic regression with the dichotomized versions of the dependent variables. Our analytic strategy leveraged school and classroom fixed effects, depending on the model, as well as observable demographic controls of students to isolate relationships that were adjusted for a broad set of observable and unobservable characteristics of students and schools. In particular, the fixed effects approaches allowed us to base estimates off variation among students in the same school or classrooms, thereby implicitly controlling for factors such as school or grade level, the quality of school leadership, instructional practices, other safety and security measures in the school, demographic composition of the school, and the neighborhood contextual environment of the school. The final model took the following form:

P r ( Y i c s = 1 | G ) = 1 1 + e − G where G = β 0 + β 1 S R O _ L e s s M o n t h l y i c s + β 2 S R O _ M o n t h l y i c s + β 3 S R O _ W e e k l y i c s + β 4 S R O _ D a i l y i c s + ∑ n = 1 6 β n + 4 X i c s + γ c s + ε i c s

In our first model, the logistic regression included a series of indicators for frequency of student i in classroom c in school s interacting with the SRO (SRO_), where never interacting with the SRO served as the reference category. Following this series of frequency indicators, we included a vector of student characteristics (race/ethnicity and sex, X_ics). We then estimated models with school or classroom fixed effects (represented as γ c s ). The choice of fixed effects models was made to implicitly control for observable and unobservable characteristics that were constant within schools or classrooms, thereby mitigating selection bias. Further models of the same form were estimated in which the key independent variables of interest were indicators of comfort in talking to and trusting the SRO. Response options ranged from strongly disagree to strongly agree with strongly disagree serving as the reference category. Each of these models also included controls for the frequency of interactions with SROs, assessing the relationship between the trust/comfort measures net of the frequency of interactions.

In addition to these primary models, we also estimated models in which the frequency and trust/comfort measures were interacted with student race/ethnicity, school socioeconomic composition, as well as school grade level. For parsimony, we used the ordinal version of the primary independent variables in these models rather than the binary coded versions. For the racial subgroup analysis, we interacted these variables with an indicator of a student being White or the percentage of White students in a school due to small cell sizes among specific racial/ethnic groups. In other models, we found that the results did not differ when using further disaggregated racial/ethnic groups. For models examining variation by school socioeconomic composition, we interacted the primary variables with an indicator for being a Title 1 school. For the school-level analysis, we interacted the primary independent variables with binary indicators of whether a school was an elementary or middle school. These models allowed for an explicit consideration of whether relationships varied for students of color or by school level.

Although our analytic strategy accounted for many different plausible sources of endogeneity in the relationship between SRO interactions and student outcomes, our research design did not address all sources of endogeneity. In particular, we were limited to a small set of student demographic variables and could not address preexisting differences in student behavior, attitude, or academic performance. A student fixed effects approach would be a superior way to address this selection issue, but, with only one time point per student, we were unable to implement student fixed effects. Furthermore, given the cross-sectional nature of our data, we were unable to establish time order, meaning that we cannot confirm that SRO interactions preceded the measured outcomes. Without the benefit of an experimental design to test our research questions or survey responses from multiple time periods, we used a classroom and school fixed effects approach to account for endogenous differences between schools and between classrooms within schools and coupled this with qualitative analyses to provide further insight into the mechanisms and nuances of the relationships observed.

Qualitative Analysis

The SGI transcripts were loaded into NVivo and coded using the constant comparative method. All transcripts were coded by at least two team members. The purpose of the use of multiple coders was to reach “crystallization,” a qualitative concept that refers to the creation of a more complete view of a concept by engaging multiple perspectives in the coding of the data (Ellingson, 2008; Tracy, 2010, 2013). Coders kept coding memos in which theoretical notes, emergent themes, exemplar quotes, and negative cases were recorded. Coding continued until the research team felt they had reached saturation, defined in this study as the point at which no new themes emerged (Fusch & Ness, 2015; Saunders et al., 2018).

Through this process, the research team took a grounded-theory, iterative approach to identify anticipated as well as unanticipated themes in the data (Glaser & Strauss, 1967; Strauss & Corbin, 1990). The research team regularly met to discuss themes and clarify understanding of each theme as it emerged. The original development of themes was based on the group interview protocol as well as research team conversations on the content of the group interviews as experienced by facilitators.

For this study, we were interested in several specific themes related to the relationships between student perceptions of safety and disciplinary outcomes as well as student relationships and trust/comfort with SROs. The themes related to student disciplinary outcomes included SROs’ impact on student behavior and SROs’ role as a disciplinarian. In both codes, students were commenting directly on SROs’ involvement in the school’s disciplinary system. For the safety-related themes, we included codes for SROs’ impact on feelings of safety, how students perceived the purpose of having an SRO in their school, and students’ perceptions of safety in the school. We also included a code for SRO–student relationships that provided data on both discipline and safety. We include a copy of our full coding dictionary (both codes central to this analysis and others) as Supplemental Appendix B in the online version of the journal.

Two members of the research team then reviewed the resulting text from the relevant codes using the following guiding questions: (a) How do students describe SROs’ impacts on disciplinary situations and on their feelings of safety? (b) How do students describe SROs’ role in discipline? (c) From the student point of view, how do SROs communicate threats to safety and their role in maintaining safety? (d) When students describe their comfort and/or trust in the SRO, how do they relate these to their feelings of safety and experiences of discipline? Common responses and divergent views were then noted, and results were synthesized from the notes from these guiding questions and research team discussion.

Descriptive Statistics

As shown in Table 2, our survey sample characteristics generally mirrored those of the broader schools participating in the study. In particular, about half of participants were elementary students and around a quarter each were middle or high school students. Participants were predominantly White (79%), which reflected the overall student composition. Females were more likely than males to participate in the survey (57%–43%).

OPEN IN VIEWER

Table 2

Means and Standard Errors of Key Independent and Dependent Variables for the Full Sample and by Frequency/Quality of Interaction With SRO

Key variables	Full sample	Frequent SRO interactions	Infrequent SRO interactions	Agrees or strongly agrees comfortable talking to SRO	Neutral, disagrees, or strongly disagrees comfortable talking to SRO	Agrees or strongly agrees trusts SRO	Neutral, disagrees, or strongly disagrees trusts SRO	Sig. T-test Cols 2 and 3	Sig. T-test Cols 4 and 5	Sig. T-test Cols 6 and 7
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Often student interacts with SRO
Never	0.391	0	0.600	0.290	0.601	0.334	0.650		*	*
	(0.016)	(.)	(0.020)	(0.019)	(0.029)	(0.018)	(0.038)
Less than monthly	0.261	0	0.400	0.260	0.262	0.285	0.150			*
	(0.015)	(.)	(0.020)	(0.018)	(0.026)	(0.017)	(0.028)
Monthly	0.116	0.332	0	0.143	0.0594	0.119	0.100		*
	(0.011)	(0.027)	(.)	(0.014)	(0.014)	(0.012)	(0.024)
Weekly	0.149	0.427	0	0.191	0.0594	0.165	0.0750		*	*
	(0.012)	(0.028)	(.)	(0.016)	(0.014)	(0.014)	(0.021)
Daily	0.0839	0.241	0	0.116	0.0175	0.0970	0.0250		*	*
	(0.009)	(0.024)	(.)	(0.013)	(0.008)	(0.011)	(0.012)
Student feels comfortable talking to SRO
Strongly disagree	0.0295	0.0163	0.0365	0	0.0909	0.00831	0.125			*
	(0.006)	(0.007)	(0.008)	(.)	(0.017)	(0.003)	(0.026)
Disagree	0.0510	0.0130	0.0713	0	0.157	0.0263	0.163	*		*
	(0.007)	(0.006)	(0.011)	(.)	(0.022)	(0.006)	(0.029)
Neutral	0.244	0.0977	0.322	0	0.752	0.222	0.344	*		*
	(0.014)	(0.017)	(0.019)	(.)	(0.026)	(0.015)	(0.038)
Agree	0.385	0.404	0.376	0.570	0	0.402	0.313			*
	(0.016)	(0.028)	(0.020)	(0.020)	(.)	(0.018)	(0.037)
Strongly agree	0.290	0.469	0.195	0.430	0	0.342	0.0563	*		*
	(0.015)	(0.029)	(0.017)	(0.020)	(.)	(0.018)	(0.018)
Students trusts the SRO
Strongly disagree	0.0125	0.00326	0.0174	0.00336	0.0315	0	0.0688	*	*
	(0.004)	(0.003)	(0.005)	(0.002)	(0.010)	(.)	(0.020)
Disagree	0.0215	0.00651	0.0296	0.00839	0.0490	0	0.119	*	*
	(0.005)	(0.005)	(0.007)	(0.004)	(0.013)	(.)	(0.026)
Neutral	0.147	0.0945	0.176	0.0872	0.273	0	0.813	*	*
	(0.012)	(0.017)	(0.016)	(0.012)	(0.026)	(.)	(0.031)
Agree	0.358	0.306	0.386	0.337	0.402	0.438	0	*
	(0.016)	(0.026)	(0.020)	(0.019)	(0.029)	(0.018)	(.)
Strongly agree	0.460	0.590	0.391	0.564	0.245	0.562	0	*	*
	(0.017)	(0.028)	(0.020)	(0.020)	(0.025)	(0.018)	(.)
Frequency receiving consequence at school	0.442	0.450	0.438	0.450	0.427	0.454	0.388
	(0.017)	(0.028)	(0.021)	(0.020)	(0.029)	(0.019)	(0.039)
How often afraid of attack at school	0.205	0.163	0.228	0.161	0.297	0.183	0.306	*	*	*
	(0.014)	(0.021)	(0.018)	(0.015)	(0.027)	(0.014)	(0.037)
I feel safe at this school.	0.866	0.889	0.854	0.908	0.780	0.914	0.650		*	*
	(0.011)	(0.018)	(0.015)	(0.012)	(0.025)	(0.010)	(0.038)
The SRO makes me feel safer.	0.763	0.860	0.711	0.844	0.594	0.855	0.350	*	*	*
	(0.014)	(0.020)	(0.019)	(0.015)	(0.029)	(0.013)	(0.038)
Elementary school	0.533	0.674	0.457	0.552	0.493	0.572	0.356	*		*
	(0.017)	(0.027)	(0.021)	(0.020)	(0.030)	(0.018)	(0.038)
Middle school	0.257	0.173	0.303	0.262	0.248	0.262	0.237	*
	(0.015)	(0.022)	(0.019)	(0.018)	(0.026)	(0.016)	(0.034)
High school	0.210	0.153	0.240	0.186	0.259	0.166	0.406	*	*	*
	(0.014)	(0.021)	(0.018)	(0.016)	(0.026)	(0.014)	(0.039)
Student race
White	0.793	0.801	0.788	0.785	0.808	0.806	0.731			*
	(0.014)	(0.023)	(0.017)	(0.017)	(0.023)	(0.015)	(0.035)
Black	0.0488	0.0586	0.0435	0.0587	0.0280	0.0471	0.0563		*
	(0.007)	(0.013)	(0.009)	(0.010)	(0.010)	(0.008)	(0.018)
Asian	0.0692	0.0423	0.0835	0.0705	0.0664	0.0679	0.0750	*
	(0.009)	(0.012)	(0.012)	(0.010)	(0.015)	(0.009)	(0.021)
Hispanic	0.0351	0.0293	0.0383	0.0285	0.0490	0.0249	0.0813			*
	(0.006)	(0.010)	(0.008)	(0.007)	(0.013)	(0.006)	(0.022)
Other	0.0329	0.0391	0.0296	0.0369	0.0245	0.0332	0.0313
	(0.006)	(0.011)	(0.007)	(0.008)	(0.009)	(0.007)	(0.014)
Two or more races	0.0204	0.0293	0.0157	0.0185	0.0245	0.0194	0.0250
	(0.005)	(0.010)	(0.005)	(0.006)	(0.009)	(0.005)	(0.012)
Student is male	0.431	0.430	0.431	0.448	0.395	0.434	0.419
	(0.017)	(0.028)	(0.021)	(0.020)	(0.029)	(0.018)	(0.039)
Observations	882	307	575	596	286	722	160

Note. SRO = school resource officers.

When we compare the survey sample to the group interview sample, we do not have descriptive statistics on our group interview participants, but we can note some overall similarities and differences between the two samples. The key difference between the survey sample and the group interview sample was that only elementary and middle school students participated in the group interviews. The demographics of the two samples were very similar in that they were mostly White with a relatively equal representation of female and male students. The interviews only took place in Washington City Schools which is more diverse socioeconomically than Fairfield County Schools, making it possible that the interview participants were more likely to be economically disadvantaged than the survey participants. Washington City Schools is also classified as a small city while Fairfield County Schools have a range of urbanicities, making the interview sample a more homogeneously urban sample than the survey sample. Finally, it is possible that group interview participants were purposefully chosen by school staff, although we did not specifically hear from school administrators that they were purposefully selecting group interview participants. This purposeful selection was less likely to have occurred with the survey because every school distributed the survey to all students in a class/grade level (although there could have been purposeful selection of classes).

We note that these general descriptive characteristics are important for contextualizing our findings in view of the larger literature on school discipline. In particular, research has demonstrated that White students and females are less likely than their Black and Hispanic as well as male counterparts to experience exclusionary forms of school discipline as well as to experience arrests or other interactions with the justice system more broadly (Gregory et al., 2010; Hirschfield, 2008; Kupchik, 2016; Losen, 2014). Consequently, as we discuss later, our findings may be uniquely shaped by the context of our study in ways that limit their generalization to other settings, such as more urban environments. Yet, given the increasing presence of SROs in schools such as those in this study (Musu-Gillette et al., 2018), we argue that our results are important for understanding the expansion of SROs to broader settings. With regard to our key independent variables, we found that, overall, students generally reported comfort and trust with the SROs, despite the fact that many reported never interacting with the SRO. As shown in Table 2, 39% of respondents reported never interacting with the SRO, yet only around 8% reported disagreeing or strongly disagreeing that they were comfortable talking to the SRO and fewer than 4% reported disagreeing or strongly disagreeing with a statement that they trust the SRO. These findings suggest that the students in the sample generally believed in the goodness of law enforcement, and that they viewed them as legitimate authorities even without having personal interactions with their SROs. This mirrors the broader perceptions of law enforcement in the community, which had elected the same Sheriff to multiple consecutive terms and was generally supportive of the work done by the various law enforcement agencies. We note that this is a perspective that, based on prior research, may not be as ubiquitous in communities with more people of color or lower socioeconomic demographics.

We found that about 44% of the sample reported experiencing a disciplinary consequence and that such reports did not significantly differ by SRO interactions or trust/comfort in the SRO. Overall, students reported feeling safe in school with 87% of students agreeing or strongly agreeing that they feel safe and only 21% reporting fear of an attack. In general, students who reported frequent interactions with an SRO or trust talking to or comfort with the SRO reported a greater sense of safety and a lower likelihood of fearing an attack. Finally, 76% of students agreed or strongly agreed that the SRO in particular makes them feel safer. Such a response was significantly stronger among those that interacted frequently with the SRO or reported comfort and trust with the SRO. These descriptive findings, then, suggest relationships between measures of SRO interaction and student fear and safety but little relationship with school discipline. We turn next to results of fixed effects regression models and group interview qualitative data that further probe these descriptive findings.

SRO Interactions and Discipline

SROs were placed in elementary schools in these districts in direct response to concerns about an intruder-based school shooting, so we might not expect SROs to have an effect on student disciplinary outcomes. However, we explore the relationship between SROs and student discipline because of the prior evidence linking SROs with increased rates of punitive or exclusionary discipline (Fisher & Hennessy, 2016; Weisburst, 2019). In line with the descriptive findings, we found that, after accounting for school and classroom fixed effects as well as observable characteristics of students, the measures of interaction (frequency and trust/comfort) with an SRO were not significant predictors of the likelihood of reporting a school disciplinary consequence. Table 3 shows results from logistic regression models predicting the binary indicator of reporting at least one disciplinary consequence. Results from three separate regressions are shown. The top panel includes frequency of interaction as the key independent variable, whereas the middle and bottom panels include comfort talking to the SRO and trust in the SRO, respectively, as the key independent variables. Column 1 reports results from unadjusted models, Column 2 adds school fixed effects, and Column 3 includes classroom fixed effects along with student race and gender controls. As shown, while there were some significant relationships in the unadjusted models (Column 1), these relationships tended to become insignificant in both the school and classroom fixed effects models (Columns 2 and 3). Although not in line with the theoretical framing of support which hypothesized that SRO interactions would reduce disciplinary consequences, this finding was also not consistent with the criminalization perspective’s suggestion that greater interactions with SROs would predict more frequent disciplinary consequences. The lack of a significant finding may be reflective of the affluent suburban context, in which prior theory and research indicates SROs are likely to be primarily concerned with external threats rather than the misbehavior of students (Fisher et al., 2020; Hirschfield, 2010; Lynch et al., 2016).

OPEN IN VIEWER

Table 3

Coefficients and Standard Errors From Logistic Regression Models Predicting Disciplinary Consequences

Key independent variables	Consequence
	(1)	OR	(2)	OR	(3)	OR
Frequency of interaction
Less than monthly	0.589**	1.80	0.336 †	1.40	0.246	1.28
	(0.169)		(0.183)		(0.209)
Monthly	0.241	1.27	−0.126	0.88	−0.385	0.68
	(0.225)		(0.244)		(0.285)
Weekly	0.300	1.35	−0.145	0.87	−0.206	0.81
	(0.203)		(0.226)		(0.266)
Daily	0.336	1.40	0.110	1.12	0.00435	1.00
	(0.252)		(0.280)		(0.343)
n	920		920		790
Feels comfortable talking to SRO
Disagree	−0.0905	0.91	−0.153	0.86	−0.504	0.60
	(0.484)		(0.538)		(0.630)
Neutral	−0.506	0.60	−0.681	0.51	−0.960 †	0.38
	(0.409)		(0.464)		(0.553)
Agree	−0.273	0.76	−0.473	0.62	−0.993 †	0.37
	(0.401)		(0.459)		(0.554)
Strongly agree	−0.394	0.67	−0.636	0.53	−1.133*	0.32
	(0.411)		−0.469		(0.565)
n	913		913		784
Trusts the SRO
Disagree	1.603 †	4.97	1.871 †	6.49	1.045	2.84
	(0.896)		(0.973)		(1.119)
Neutral	1.242	3.46	1.217	3.38	1.003	2.73
	(0.789)		(0.844)		(0.961)
Agree	1.384 †	3.99	1.239	3.45	0.690	1.99
	(0.777)		(0.834)		(0.942)
Strongly agree	1.559*	4.75	1.132	3.10	0.544	1.72
	(0.777)		(0.831)		(0.934)
n	917		917		787
School fixed effects			Yes
Classroom fixed effects					Yes
Student race and gender controls					Yes

Note. Standard errors in parentheses. Omitted group is never interacted with SRO or strongly disagree. Models in middle and bottom panel include control for frequency of interactions with SRO. Sample sizes vary due to fixed effects groups with no variation within group as well as differential missing data patterns on primary variables. OR = odds ratio; SRO = school resource officers.

†

p < .1. *p < .05. **p < .01.

Indeed, such a view is supported by the responses of SGI participants. Although the students we talked to discussed awareness and interactions with SROs, they almost never linked these interactions to formal disciplinary situations. When asked whether their SRO ever administers discipline, the students in our group interviews generally responded with a resounding “no” though some did report that SROs assisted administrators with disciplinary investigations and could be involved in very serious situations. For example, when asked if the “[SRO had] ever gotten any students in trouble” students in one group stated “No. Not at all. Not ever ever ever ever ever” (SGI participant). In another group interview, students noted that SROs might be involved for very serious incidents like a weapon: “unless it were a really bad break. Like if you brought in like a knife” (SGI participant). In another, they noted that SROs might assist administrators with the use of security cameras or report behavior to teachers, but otherwise they reported that SROs had little formal disciplinary involvement.

Notably, in no cases in the SGIs did students give specific examples of SRO involvement in formal discipline such as administering consequences like suspension. Although this likely reflects in some part the selection effects of students participating in our group interviews, it also suggests that SROs’ roles in administering formal consequences were limited. Although we noted some evidence of more informal involvement in discipline—such as verbally reprimanding students or reporting incidents to teachers—SROs were more likely to be seen as counselors or alternatives to formal discipline such as explained by one student who described what an SRO would say to a student who was in trouble: “How did you get in trouble? What happened and like I’m not going to tell anyone. And then like, he’ll like just like make you feel—he’ll make people that are sitting at the front feel like better” (SGI participant). Whereas we discuss nuance to SROs’ roles in discipline in work elsewhere (Curran et al., 2019), our finding that the frequency or even quality of interactions with SROs was not predictive of differential reports of disciplinary consequences is consistent with the participating students’ descriptions of formal discipline not being a salient part of SROs’ activities. It seems then, in this context, the SROs generally avoided adopting a criminalization approach, and, in some cases, acted in ways that were consistent with the support model—though not so much so as to be predictive of lower disciplinary consequences.

SRO Interactions, Fear, and Safety

While SROs were not placed in these districts because of specific or concurrent safety concerns, the reasons SROs were in the schools are more proximally related to feelings of fear or safety than to disciplinary involvement. We found that the frequency of interactions with SROs did not significantly predict the outcome of fearing a physical attack at school or feeling safe at school. However, feeling comfortable talking to the SRO and trusting the SRO were associated with students’ feelings of safety. Results of logistic regression models predicting fear of a physical attack are shown in Table 4 and results of models predicting feelings of safety are shown in Table 5. As before, the tables are divided into three panels each representing a separate logistic regression model. As shown, students who reported feeling comfortable talking to the SRO or trusting the SRO were much more likely to report agreeing or strongly agreeing that they feel safe at school, even after accounting for observable characteristics and school or classroom fixed effects. Likewise, students who reported comfort talking to the SRO were less likely to report a fear of being physically attacked at school.

OPEN IN VIEWER

Table 4

Coefficients and Standard Errors From Logistic Regression Models Predicting Fears of an Attack at School

Key independent variables	Attack
	(1)	OR	(2)	OR	(3)	OR
Frequency of interaction
Less than monthly	0.0326	1.03	−0.0259	0.97	0.0354	1.04
	(0.199)		(0.209)		(0.236)
Monthly	−0.152	0.86	−0.169	0.84	0.0102	1.01
	(0.275)		(0.288)		(0.323)
Weekly	−0.523 †	0.59	−0.637*	0.53	−0.476	0.62
	(0.273)		(0.289)		(0.334)
Daily	−0.375	0.69	−0.557	0.57	−0.392	0.68
	(0.329)		(0.364)		(0.437)
n	912		903		716
Feels comfortable talking to SRO
Disagree	−1.025*	0.36	−0.916 †	0.40	−0.717	0.49
	(0.518)		(0.537)		(0.626)
Neutral	−0.963*	0.38	−0.738 †	0.48	−0.666	0.51
	(0.420)		(0.439)		(0.519)
Agree	−1.613**	0.20	−1.390**	0.25	−1.153*	0.32
	(0.420)		(0.444)		(0.526)
Strongly agree	−1.520**	0.22	−1.272**	0.28	−0.963 †	0.38
	(0.434)		(0.457)		(0.534)
n	906		897		710
Trusts the SRO
Disagree	1.403	4.07	1.350	3.86	1.348	3.85
	(0.907)		(0.921)		(1.093)
Neutral	0.648	1.91	0.653	1.92	0.871	2.39
	(0.805)		(0.820)		(0.947)
Agree	0.154	1.17	0.0586	1.06	0.189	1.21
	(0.796)		(0.816)		(0.946)
Strongly agree	0.0412	1.04	−0.120	0.89	0.198	1.22
	(0.796)		(0.815)		(0.940)
n	909		900		710
School fixed effects			Yes
Classroom fixed effects					Yes
Student race and gender controls					Yes

Note. Standard errors in parentheses. Omitted group is never interacted with SRO or strongly disagree. Models in middle and bottom panel include control for frequency of interactions with SRO. Sample sizes vary due to fixed effects groups with no variation within group as well as differential missing data patterns on primary variables. OR = odds ratio; SRO = school resource officers.

†

p < .1. *p < .05. **p < .01.

OPEN IN VIEWER

Table 5

Coefficients and Standard Errors From Logistic Regression Models Predicting Feeling Safe at School

Key independent variables	Feels safe
	(1)	OR	(2)	OR	(3)	OR
Frequency of interaction
Less than monthly	0.198	1.22	0.0420	1.04	0.0494	1.05
	(0.239)		(0.249)		(0.282)
Monthly	0.468	1.60	0.293	1.34	0.320	1.38
	(0.350)		(0.360)		(0.420)
Weekly	0.477	1.61	0.129	1.14	−0.0243	0.98
	(0.317)		(0.337)		(0.384)
Daily	0.103	1.11	−0.0871	0.92	0.215	1.24
	(0.357)		(0.376)		(0.452)
n	921		840		575
Feels comfortable talking to SRO
Disagree	0.723	2.06	0.977 †	2.66	0.932	2.54
	(0.507)		(0.541)		(0.603)
Neutral	1.292**	3.64	1.430**	4.18	1.713**	5.55
	(0.425)		(0.465)		(0.542)
Agree	2.045**	7.73	2.264**	9.62	2.339**	10.37
	(0.431)		(0.477)		(0.559)
Strongly agree	2.091**	8.09	2.246**	9.45	2.439**	11.46
	(0.455)		(0.502)		(0.587)
n	914		833		570
Trusts the SRO
Disagree	−0.584	0.56	−0.254	0.78	0.0422	1.04
	(0.733)		(0.758)		(0.948)
Neutral	0.368	1.44	0.783	2.19	0.291	1.34
	(0.601)		(0.651)		(0.795)
Agree	1.483*	4.41	1.950**	7.03	1.832*	6.25
	(0.596)		(0.656)		(0.815)
Strongly agree	2.418**	11.22	2.747**	15.60	2.723**	15.23
	(0.616)		(0.674)		(0.838)
n	918		837		574
School fixed effects			Yes
Classroom fixed effects					Yes
Student race and gender controls					Yes

Note. Standard errors in parentheses. Omitted group is never interacted with SRO or strongly disagree. Models in middle and bottom panel include control for frequency of interactions with SRO. Sample sizes vary due to fixed effects groups with no variation within group as well as differential missing data patterns on primary variables. OR = odds ratio; SRO = school resource officers.

†

p < .1. *p < .05. **p < .01.

Qualitatively, we found that trust and comfort, both of the SRO and in the school more broadly, were sometimes mentioned as components of students’ feelings of safety. Although several students noted that SROs could be intimidating, others explicitly said that such a view was dispelled by getting to know the SRO: “But he’s like—like if you get to know him a lot- like speak to him a lot, he’s really nice. Really nice—he’s like a really nice guy. Like he wouldn’t do anything to you” (SGI participant). Notably, this student reported that frequent interactions led to a greater comfort with the SRO, seeing him as “nice,” and that this mitigated any concern that the SRO would do something harmful to students. Another student noted that “she’s always really nice and kind, gives hugs, and, um, gives high fives and stuff. And, um, um, and when-when she’s a, our police officer it helps . . . it like . . . I feel more safe around her” (SGI participant). For these students, the SRO’s demeanor and level of comfort or trust that it established were directly related to their feelings of safety around the officer.

This view extended beyond the SRO to the broader school community. In one group interview, a student said, “Cause if you’ve been here for a long time, you know that you’re gonna be safe here with the teachers that are around here. You know that you can trust them, and be around them a lot to be safe” (SGI participant). Research on school climate supports our finding of the importance of relationships between students and SROs, finding that, where student/adult relationships are stronger, students feel safer (Fisher et al., 2018; Steinberg et al., 2011). Our findings suggest that a similar mechanism may underlie student–SRO relationships, with trust and comfort being important for the association between SROs and students’ perceptions of safety. In this way, the SROs in our context appeared to engage with students in ways that, on balance, represented a support model—one that built trust and comfort—rather than a criminalization model. Although not true of frequency of interactions, trust/comfort in SROs tended to predict greater feelings of safety in ways that were consistent with the hypotheses of the support model. These findings may again be understood in light of the schools’ context; to the extent that the SROs’ concern with external threats—particularly the potential for gun violence—was reflected in their daily activities, the students who had more positive interactions (but not more interactions) with the SROs may have felt safer at school.

Perceived Impact of SROs on Safety

Finally, we observed that both frequency of interactions and trust/comfort with SROs generally predicted a higher likelihood that students reported that the SRO himself or herself makes them feel safer. As shown in Table 6, estimates across all three key independent variables showed that a greater frequency of interaction, greater feelings of comfort talking with, and greater trust in the SRO was related to a greater likelihood of agreeing or strongly agreeing that SROs make them feel safer, a result that was robust to both school and classroom fixed effects. This is noteworthy, particularly with regard to frequency of interactions, given that the measure of frequency of interaction was not predictive of overall feelings of safety at school. To understand why this might be, it helps to consider SGI responses.

OPEN IN VIEWER

Table 6

Coefficients and Standard Errors From Logistic Regression Models Predicting Students Reporting That the SRO Makes Them Feel Safer

Key independent variables	SRO makes feel safer
	(1)	OR	(2)	OR	(3)	OR
Frequency of interaction
Less than monthly	0.918**	2.50	0.753**	2.12	0.857**	2.36
	(0.200)		(0.207)		(0.234)
Monthly	1.382**	3.98	1.218**	3.38	1.185**	3.27
	(0.316)		(0.325)		(0.365)
Weekly	1.015**	2.76	0.802**	2.23	0.714*	2.04
	(0.254)		(0.271)		(0.312)
Daily	1.583**	4.87	1.477**	4.38	1.318**	3.74
	(0.389)		(0.403)		(0.435)
n	921		913		745
Feels comfortable talking to SRO
Disagree	0.769	2.16	0.828	2.29	0.759	2.14
	(0.536)		(0.556)		(0.627)
Neutral	1.821**	6.18	1.847**	6.34	1.753**	5.77
	(0.468)		(0.489)		(0.560)
Agree	2.486**	12.01	2.533**	12.59	2.394**	10.96
	(0.466)		(0.492)		(0.563)
Strongly agree	2.779**	16.10	2.864**	17.53	2.694**	14.79
	(0.486)		(0.514)		(0.583)
n	915		907		740
Trusts the SRO
Disagree	−0.103	0.90	0.0245	1.02	−0.0831	0.92
	(0.872)		(0.871)		(1.286)
Neutral	0.702	2.02	0.657	1.93	1.142	3.13
	(0.687)		(0.697)		(0.931)
Agree	2.379**	10.79	2.295**	9.92	2.828**	16.91
	(0.677)		(0.691)		(0.941)
Strongly agree	3.368**	29.02	3.203**	24.61	3.640**	38.09
	(0.687)		(0.700)		(0.944)
n	918		910		744
School fixed effects			Yes
Classroom fixed effects					Yes
Student race and gender controls					Yes

Note. Standard errors in parentheses. Omitted group is never interacted with SRO or strongly disagree. Models in middle and bottom panel include control for frequency of interactions with SRO. Sample sizes vary due to fixed effects groups with no variation within group as well as differential missing data patterns on primary variables. OR = odds ratio; SRO = school resource officers.

†

p < .1. *p < .05. **p < .01.

With regard to feelings of safety, the qualitative group interviews revealed a multifaceted connection between SROs and feelings of safety. On one hand, students generally reported that their SRO makes them feel safer with many immediately mentioning the SRO when asked about safety. One student noted, “Um the difference of the SRO being here just makes me feel fine and happy whenever I come to school not feel worried” (SGI participant).

We found, however, that although students believed the SROs would protect them, the presence of an SRO may have contributed to a heightened overall sense of fear among students. The student who previously mentioned not feeling worried because of the SRO went on to say that “But then without the [SRO], it just makes me feel worried and don’t want to come to school because you don’t know what could happen when they aren’t here” (SGI participant). This concern was common and appeared to be related to an exaggerated sense of danger. Another student in the same group interview stated,

And um, my mom said she didn’t have a police officer when she was in school. Like, nothing really bad happened at her school . . . And then the, back then my mom’s day they didn’t think that they needed one. But now we need one. (SGI participant)

This perception of danger was notable given that the students never mentioned any incidents of serious violence and in light of a large body of research suggesting that schools are one of the safest locations for young people (Cornell, 2015; Musu-Gillette et al., 2018).

In fact, in cases where students expressed concerns about a lack of safety around an SRO, there was often a connection to their perceived sense of threats. Although several students mentioned that the SRO was “intimidating,” students also commonly mentioned fears related to the SRO’s potential to fail them in an incident of violence. One student began to describe the SRO’s gun belt, but rather than expressing a direct fear of the officer’s weapon, instead described a fear that the officer might have poor aim when encountering an intruder:

I just, his belt kind of scares me because if the like the . . . there was an intruder or something and then he pulls out a weapon and he could not have the best at aiming. So like if he hits someone else that was like, not the intruder. (SGI participant)

Another student noted how the SRO could not be in all places at once, stating,

Um so I kind of like feel scared because he’s not in all spots. If he’s in the third grade hallway or the second grade hallway, and then um and then something happened in fourth grade, he’s not in all spots so it’s going to take him a long time. (SGI participant)

Other students, when describing fearful situations in the school, recounted incidents of fire drills or active-shooter drills in which the SRO was involved—thereby identifying the SRO as a protective factor against threats whose likelihood of occurring are objectively very small. For example, in response to what the school would be like without an SRO, one student stated,

I’d be like . . . cause, like, we do have alarms, but there is not like teachers and people always like looking around the school and watching and stuff. Uh, I would kinda feel unsafe because there wouldn’t as many people in the office watching out for us like now and watching over the security cameras and making sure we’re all safe. (SGI participant)

Prior research has illustrated how the presence of visual security measures can heighten individuals’ concern about the relative risk of threats (Ferraro, 1995; Perumean-Chaney & Sutton, 2013; Schreck & Miller, 2003). Our qualitative results suggest that the mere presence of an SRO and possibly the frequency of interactions with SROs may operate similarly. Students who interact more frequently with an SRO may come to perceive the SRO as contributing to the safety of the school but may simultaneously be conditioned to believe that they face greater dangers at school. This may explain why interactions with SROs are not predictive of measures of students’ feeling of safety while nevertheless predicting a greater likelihood that they report the SRO making them feel safe. Such a finding could be consistent with both a criminalization and support theoretical perspective in which SROs simultaneously provide support (in the form of being a presence that promotes safety) while also creating an environment that is more aware of potential criminal acts (even if not directly criminalizing students).

Subgroup Analysis

Finally, we explored differences in our primary results across school level (elementary, middle, or high), student race/ethnicity, school racial composition, and whether schools qualified as Title 1 eligible (a proxy for poverty of students served). In these models, we interacted the primary independent variables of interest with indicators of each of these subgroups. In general, we found little evidence of heterogeneous relationships across these subgroups. Results of these analyses are shown in Supplemental Appendices C to F in the online version of the journal.

First, we found few differences in relationships by school level (Tables C1–C4). With regard to the level of school (elementary/middle/high), the lack of significant findings in the quantitative analyses was somewhat surprising in light of our qualitative data. In particular, in discussions with SROs themselves, we tended to find that high school and middle school SROs were more likely than elementary SROs to view themselves as playing a role in responding to student misconduct. In addition, high school and middle school SROs reported many more arrests than their elementary counterparts (though still relatively low overall numbers; Curran et al., 2020). We may have expected then that more frequent interactions with SROs at the middle or high school level would result in more disciplinary outcomes than in the elementary level. That this was not the case quantitatively may reflect the degree to which the district policy that SROs not engage in discipline shaped their actions or may indicate that any effect of SROs was not particular to students who interacted more frequently with them.

We examined potential heterogeneous relationships by race/ethnicity in two ways. First, we examined whether relationships differed based on students’ reported race/ethnicity. As shown in Tables D1 to D4, there was little evidence of a moderating effect of student race/ethnicity on the relationship between our independent and dependent variables. The one exception was that the relationship between reporting trust in the SRO and feeling that the SRO made students feel safer was larger for White students. In addition to examining potential moderating effects by students’ own race/ethnicity, we also examined whether relationships varied by the racial composition of the school as a whole. Prior work has suggested that school discipline and security approaches, including SROs, may vary as a function of school racial composition (Kupchik, 2010; Mukherjee, 2007; Nolan, 2011) in ways that may moderate our results even if relationships do not differ among students of different race/ethnicity within the school. As shown in Tables E1 to E4, we, however, find no evidence of a moderating effect of school racial composition. These findings contrast with hypotheses put forward in the criminalization model, which suggested that students of color or students in schools with a greater proportion of students of color might experience more negative associations with SRO interactions. We return to this lack of significant differential relationships by race/ethnicity in our discussion and suggest that specific characteristics of our study’s context may help explain why relationships did not vary by race/ethnicity.

Finally, we examined whether relationships differed significantly based on the socioeconomic composition of the schools. In Tables F1 to F4, we show results of models that interact our primary independent variables of interest with a binary indicator of a school being eligible for Title 1 (a measure of school-level poverty). In general, we find few significant differences, though we note that the relationship between trusting an SRO and school discipline appears larger in Title 1 schools, such that trust predicts a lower likelihood of disciplinary consequences in a Title 1 school. Similarly, trust appears more important for students’ feelings of safety in Title 1 schools, with the relationship between trusting the SRO and reports of feeling safe at school being larger in Title 1 schools than non-Title 1 schools. These results suggest the importance of relationships of trust, particularly in schools serving greater proportions of students from lower socioeconomic backgrounds.

Robustness Checks

In addition to these subgroup analyses, we also performed several sensitivity and robustness checks to our primary quantitative models. First, we found that our primary results were robust to the use of linear probability models to predict the binary outcomes rather than logistic regression. Details and results of these analyses can be found in Supplemental Appendix G (Tables G1–G4, in the online version of the journal).

We also confirmed that our results were robust to the decision to dichotomize the outcome variables. In our primary models, the outcome variables were dichotomized to ease interpretation, in each case representing students who reported receiving a consequence or not and, for the safety outcomes, feeling safe (free from threat of attack or that the SRO increased safety) or not. While such an approach is consistent with prior work in this area, we acknowledge that dichotomizing the dependent variables potentially masks important variation in the outcome variables. We re-estimated models using the original, ordinal measures of disciplinary consequences and safety (nondichotomized versions). As shown in Supplemental Appendix H Tables H1 to H4 in the online version of the journal, the results are substantively the same, with almost no changes in directionality or statistical significance of the results.

Next, we explored the relationships between our key independent (SRO interactions, trust, and comfort) and key dependent variables (disciplinary and safety outcomes) when their roles were swapped. In particular, we modeled SRO interactions and trust/comfort as outcome variables predicted by the disciplinary and safety variables. These models retained the binary disciplinary/safety measures and used the ordinal versions of the interaction and trust/comfort variables. Results are shown in Supplemental Appendix I in the online version of the journal. As expected, the relationships were generally consistent with those in the primary model, underscoring that, given the cross-sectional nature of our survey data, we cannot definitively establish the causal order of the relationships explored.

Finally, we addressed the potential threat of common methods bias through factor analysis. Common methods bias refers to the concern that relationships between our key independent and dependent variables could be driven by shared characteristics of the data collection process (i.e., both measures coming from self-reports on the same survey; Podsakoff et al., 2012). While the degree to which common methods bias is an issue in such research designs is debated (Spector, 2006; Spector & Brannick, 2009), one post hoc approach to assessing the extent to which it is an issue is to conduct a factor analysis on the variables used in the study. Evidence of a single factor or a factor that explains a majority of the variation would raise concerns about common methods bias. We estimated an exploratory principal components factor analysis without rotation on both our full set of student survey responses (minus demographic control variables) as well as on just the student survey responses used in our analysis. In the first case, we observed eight factors with an eigenvalue greater than one, the greatest of which accounted for only 28% of the variance. In the second case, we observed two factors with an eigenvalue greater than one, the greatest of which accounted for only 37% of the variance. While not definitive, we take this as evidence that our results are not driven by common methods bias.

SROs’ Lack of Involvement in Student Discipline

First, we find frequency of interactions, trust in, or comfort with SROs appears to have no discernible relationship with the likelihood of being disciplined. This contrasts with much of the prior quantitative research on SROs and discipline as well as the criminalization model posited in our theoretical framework section. With few exceptions (e.g., Na & Gottfredson, 2013), prior research has found that the presence of SROs increases student disciplinary consequences (Fisher & Hennessy, 2016; Weisburst, 2019). It is possible that the presence of SROs in our sample schools has, in fact, influenced school disciplinary environments in ways that may change the likelihood of disciplinary involvement beyond individual student interactions with SROs (such as suggested by Kupchik, 2010), but we see little relationship between the frequency of interactions or trust/comfort with SROs and disciplinary outcomes.

Prior qualitative work has found SROs commonly engage in rule enforcement and take on active roles in formal discipline (Curran et al., 2019; Kupchik, 2010; Mukherjee, 2007; Nolan, 2011). That our SGI participants reported observing little of these types of activities suggests that the types of interactions within our sample schools may differ meaningfully from those in other contexts. Indeed, in work elsewhere, we describe how the SROs in our sample schools approached their job with an emphasis on protecting students from external threats rather than policing internal threats in a way that may feed into school discipline, a difference that we attribute to the difference between SRO placement in an affluent, suburban setting versus an urban school system (Fisher et al., 2020). Although many SROs still engaged in forms of informal discipline, official policy was for SROs to avoid formal involvement in school discipline. This is important given that, nationally, almost two thirds of schools with SROs present either do not have a written policy about their involvement in discipline or the school leader is unaware of such a policy (Correa & Diliberti, 2020).

The lack of significant relationships with regard to discipline in this setting is in line with prior research that has pointed to the importance of considering contextual influences when examining the association between SROs and student outcomes, focusing in particular on school racial composition and the broader disciplinary climate of the school (Fisher, 2016). SROs serving schools with predominantly White student bodies are less likely to view the students as threats, compared with in settings with more students of color, potentially contributing to the lack of a relationship with discipline observed in this study (Fisher et al., 2020).

SROs Are Associated With Feelings of Safety and Potentially Increased Risk

Turning to outcomes of safety, our results demonstrate that SROs’ association with student safety is complex. We find that, although SROs may be seen as contributing to safety, their activities may not always result in overall improvements in students’ perceptions of safety. Prior work suggests the presence of SROs in schools is associated with higher feelings of safety among students (Reingle Gonzalez et al., 2016). On the intensive margin, our study suggests frequency of interaction with an SRO is not a significant predictor of either higher or lower feelings of overall safety. Specifically, we did not find frequency of interaction to be related to either students’ reports of feelings of safety or their fear of being attacked at school. At the same time, students who reported more comfort with and trust in SROs reported feeling safer at school, a finding that aligns with prior work that has found that students who report more positive perceptions of SROs generally feel safer at schools (Penteck & Eisenberg, 2018; Theriot & Orme, 2016).

Despite seeing no relationship between frequency of interactions with SROs and feelings of safety, we found that students who interacted with SROs more or reported greater comfort and trust in SROs were more likely to report that the SROs themselves improved their feelings of safety. In a prior study, Theriot (2016) found that more frequent student interactions with SROs predicted more positive attitudes about SROs but that such students were less connected to their schools. Our findings fit with such a phenomenon—one in which interactions with SROs alter views of SROs in positive ways but may have negative implications for other aspects of students’ views of the school. The findings from our SGIs suggest that the presence of SROs along with the security practices that SROs engage in (e.g., drills) may increase students’ perceived risk of threats. As a result, students who interact with SROs more may view SROs as increasing safety, but these interactions may lead students to feel at higher risk. This fits with prior literature that demonstrates how visual security measures can increase perceived risk (Ferraro, 1995; Perumean-Chaney & Sutton, 2013; Schreck & Miller, 2003) and is consistent with the presence of elements of both the criminalization and support theoretical model. It also points to the added value of examining within-school rather than between-school differences in SROs’ effects as well as the significance of including students’ perspectives in addition to official school records.

Interestingly, if SROs are indeed increasing students’ sense of danger but also being viewed as a mechanism for enhancing safety, there is the potential for a paradoxical feedback loop in which expanding interactions with SROs over time leads to students feeling more at risk which, in turn, leads to the need for more interactions with SROs to feel safe. In other words, interacting with SROs may reinforce the need for SROs, creating a cycle in which sense of risk and the need for interaction with SROs continually reinforce each other. This finding is especially noteworthy as the schools in this study were very safe prior to the introduction of SROs. Students were both easily socialized into expecting to interact with SROs as a normal function of schooling while justifying SRO presence through an increase in the students’ perception of risk. This leads to the concern that SROs can enter a very safe environment and introduce a false sense of danger, a phenomenon that only worsens with increased interaction with the SRO.

Few Differences in Findings by Race/Ethnicity in This Setting

Although one of our research questions focused on differences by student race/ethnicity, our findings indicated only very limited differences in the associations by either student-level race/ethnicity or school-level racial/ethnic composition. This finding is noteworthy given prior research showing that SROs (and their association with student outcomes) function differently by race/ethnicity both within and between schools (Fisher, 2016; Kupchik, 2010; Nolan, 2011). One potential reason we did not find differences by race/ethnicity is that the SROs had such limited engagement with monitoring student behavior, both as a function of (a) their mandate from the Sheriff to focus on external threats and not get involved with discipline and (b) their perception that the students in the schools were very well behaved. Prior research that has shown racial differences have largely focused on the broad theme of criminalization, and the current study found little evidence of criminalization occurring. Beyond differences by student/race ethnicity, considering the race/ethnicity of the SRO may also be important. Given the importance of the racial/ethnic congruence between students and teachers (Lindsay & Hart, 2017), the racial/ethnic congruence between students and SROs may also be important. The SROs in our sample were predominantly White, precluding any meaningful analyses by differences in SRO race/ethnicity or the racial/ethnic congruence between students and SROs, but this may be a useful direction for future research.

Implications for Policy and Practice

As the presence of SROs in schools expands nationwide, our findings have implications for the more routine day-to-day interactions between students and SROs. First, prior work and advocates have raised concerns about SROs’ influence on school disciplinary outcomes (Fisher & Hennessy, 2016). Our work suggests that the potential negative relationship between SROs and discipline suggested by prior work may be mitigated by careful attention to how policy defines SROs’ roles in discipline. As described by the participants qualitatively, the SROs in the participating schools did little formal discipline, and the quantitative results suggest that, perhaps as a result, the frequency and quality of interactions with students predicted few differences in disciplinary outcomes. By not prioritizing the policing of noncriminal misconduct within school walls, the schools studied may serve as an example for ways that other schools can mitigate SROs’ association with disciplinary outcomes.

At the same time, however, our findings suggest that the approach of emphasizing external threats may have unintended consequences. The finding that interactions with SROs predicted students’ views that SROs contributed to safety while not predicting greater feelings of safety, overall, coupled with students’ descriptions of perceived threats in conjunction with SRO interactions, suggests that an outward facing approach may unintentionally heighten students’ sense of fear. As efforts to expand SROs result in SROs in a greater number of elementary school classrooms, the implications of this potentially heightened sense of threats require particular attention among educators. For example, SROs might be instructed to explain their presence to elementary students in terms not related to school shootings or other threats to their safety. Educators might consider whether certain drills, including active-shooter drills, or components of drills that involve SROs are developmentally appropriate for younger learners or whether such drills could be as effectively completed without students present.

Limitations

It is important to note that our study has several limitations. First, the sample of schools, although diverse in terms of urbanicity and grade level, was nevertheless not representative of schools nationally or of all schools in similar urbanicities. We recognize that the results found in this context could vary in other settings (due to demographics, political leanings, or otherwise) or simply as a function of differential school or law enforcement agency leadership or structure. Nevertheless, the results do speak to contexts that, to this point, have been largely overlooked in the literature, but that (when one compares our results to prior work) seem to impact study results. As SROs increasingly expand to nonurban and elementary settings, this study provides an in-depth foundation for consideration of their role and relationship to discipline and safety.

A second limitation pertains to the internal validity of the study. As the study uses cross-sectional data, we were unable to disentangle time order between the independent and dependent variables. While we have framed this paper in terms of SRO interactions predicting discipline and safety outcomes, it is important to note that disciplinary interactions and feelings of safety may influence students’ interactions with the SROs. As discussed, we observe similar relationships in the quantitative data when modeling interactions with SROs as the dependent variable. Although the qualitative component of this study sheds light on the time order and mechanisms underlying the observed quantitative relationship, future work could expand on this study by using longitudinal data and increasing student participation in the qualitative work.

A third limitation has to do with the differences between the survey sample and the group interview sample. While we use the qualitative results to help explain the survey findings, these samples vary from each other in important ways. There were no high school group interviews and no group interviews in Fairfield County Schools. It is possible that high school students and students in more rural settings than Washington City have had substantially different experiences with their SRO that would have changed our interpretation of the results. However, we do note that this kind of arrangement with a countywide school district and a small, city-based school district is rather common in the southeastern United States. The setting itself likely generalizes to many similar areas, although the particular samples we have from these settings might limit our power to use student perspectives to explain quantitative findings.

Fourth, it is worth noting that our independent measures in the quantitative analysis focused on the frequency of interactions with SROs and student-reported trust and comfort in SROs. Future work could extend this analysis by developing more precise measures of the quality of interactions between SROs and students. Clearly, all interactions are not equal, and we would expect interactions that are supportive and developmentally appropriate to have differing relationships with outcomes than interactions that are more punitive or negative in tone. While our measures of trust and comfort talking to the SRO likely pick up some component of interaction quality and our qualitative data sheds light on the nature of the interactions, future work may nevertheless benefit from further probing the types and quality of SRO/student interactions.

Finally, we note that our study may be susceptible to common threats to survey research. For example, prior questions on the survey could have influenced students’ responses to later questions. Furthermore, our study relied on student-reported measures of discipline and safety. Feelings of safety is a multidimensional construct and, while we used multiple measures of the construct, it is possible that findings are particular to our chosen measures. For example, specific aspects of safety, such as experiences of bullying, were not included in our study. Likewise, our measure of disciplinary consequences was broad. It is possible that SROs may impact specific types of discipline (such as suspensions) while not impacting others. Future work might replicate this study with administrative data or measures specific to classroom-level consequences and responses like suspension.