The data were collected in the spring of 2010. In an effort to survey a representative sample of NEA members, both a telephone (63%) and an electronic survey (37%) were used. Specifically, we used a web-based electronic survey because of growing concerns that individuals are less inclined to participate in and/or be reached by phone surveys (Holbrook, Krosnick, & Pfent, 2007). In total, 1,601 teachers and 2,142 ESPs completed the telephone survey, whereas 562 teachers and 759 ESPs completed the electronic survey. The data collection activities were conducted by an external professional research firm contracted by the NEA; the subcontractor made the phone calls and administered the survey on behalf of the NEA. With regard to incentives for participation, a lottery was used, whereby all participants were informed that 20 participants would be selected at random for $100. Participants were told that the purpose of the study was to inform the NEA about members’ concerns and needs related to bullying and school climate. A sampling procedure was used to select participants that accounted for role and select demographics (e.g., age, region), thereby allowing the data to be weighted up to reflect the entire population of NEA members. Weighting is possible because of the known population distributions in the overall NEA membership database. Given the substantive interest in ESPs, they were over-sampled; however, the weighting procedure accounts for this oversampling and allows us to generalize to the full population of the NEA membership. As described in greater detail later, two weighting procedures were utilized on the data: a propensity score was used to adjust for the mode of survey administration (i.e., Web vs. phone) and a rim weight to weight the entire data set to the national population of NEA members (Watts, 2010).

The sample included 5,064 adults who were members of the NEA at the time of the data collection and were actively employed by a school or school system, which represented a 31% return rate for the survey. Just over half of the sample were ESPs (n = 2,901) and the remaining participants were “professional staff” as described next (n = 2,163). With regard to demographic characteristics of the weighted sample, the majority of professional staff were general education teachers (85%), followed by special educators (4%), remedial/ English as a second language teachers (2%), librarians (2%), counselors (3%), and other (4%). Thus, we refer to this group collectively as “teachers” in the current article. Of the ESPs, nearly half were paraprofessionals (49%), followed by maintenance (14%), cler ical (10%), school transportation (10%), food service (7%), health and student services (2%), technical and skilled trades (2%), security (1%), and other nonteaching support staff (6%). Women comprised 80% of the sample, and 89% self-identified as White, with 5% Black, 4% Hispanic, and 2% other. The participants were employed in a variety of school locations (suburban 34%, small town 24%, urban 24%, and rural areas 18%). Approximately 39% worked with students in elementary schools, 19% in middle schools, and 27% in high schools, with the remaining 16% working across multiple grade levels (see Watts, 2010).

The NEA Bullying Survey (Bradshaw, Waasdorp, & O'Brennan, 2010; see Appendix A) was developed by the research team in close collaboration with the NEA Research Department. The measure is an adapted version of a previously published, psychometrically sound measure of bullying (see Bradshaw et al., 2007). Consistent with previous research (Bradshaw et al., 2007; Centers for Disease Control and Prevention, in press; Nansel et al., 2001; Olweus, 1993), bullying was defined on the survey as “intentional and repeated aggressive acts that can be physical (such as hitting); verbal (such as threats or name calling); or relational (such as spreading rumors, or influencing social relationships). Bullying typically occurs in situations where there is a power or status difference.”

Exposure to bullying was assessed through three items (Bradshaw et al., 2007): (a) Did students report bullying to you within the past month? (yes/no); (b) Did parents report bullying to you within the past month? (yes/no); and (c) How often have you seen students being bullied at the school where you work, on school grounds, or on the bus? (dichotomized into 2–3 times a month vs. 1 time or less; Solberg & Olweus, 2004). Participants’ personal experiences with bullying were assessed through the question “Have you personally been bullied by someone else at the school where you work?” (yes/no). If they responded yes, they were asked by whom, with the response options being “students”, “parents,” “other staff,” and “administration.” This set of items is based on a previous study of staff members’ victimization (Bradshaw et al., 2007). Participants’ concerns about bullying were assessed through a question that read “How much of a problem is bullying at the school where you work?” (Bradshaw et al., 2007). For ease of interpretation, the response options were dichotomized into “not a problem or minor problem vs. moderate problem or major problem.”

Perceptions of different forms of bullying were assessed by three items from Bradshaw et al. (2007): (a) Did a student report that he or she experienced this form of bullying to you within the past month? (yes/no); (b) How much of a problem is this form of bullying at the school where you work? (4-point scale, from [1] not a problem to [4] major problem); and (c) How comfortable would you feel intervening or reprimanding a student who engaged in this form of bullying (4-point scale, from [1] very uncomfortable to [4] very comfortable). These three core items were asked for each specific form of bullying: physical (hitting, pushing, or kicking), verbal (general teasing or name calling), relational (rumor spreading or excluding someone from a group), cyberbullying (defined as “sending or posting harmful material or engaging in other forms of social aggression using the Internet or other digital devices, such as mobile phones”), and sexting (defined as “sending or forwarding sexually explicit photos, videos or messages from a mobile phone or other electronic device”). Given the emerging concerns regarding sexting (Mitchell, Finkelhor, Jones, & Wolak, 2011), it was specified separately from cyberbullying. The same three core items were utilized to assess the six special populations of students, including youth bullied because of their “perceived sexual orientation or gender-nonconformity,” because they are overweight, have a disability, and because of their sex, race, or religion.

Perceptions of the school's bullying policy and programming were assessed through four yes/no questions: (a) Does the school district have a bullying policy? (b) Are bullying problems adequately addressed by the bullying policy? (c) Is the policy clear and easy to implement? (d) Did you receive training on how to implement the policy? Perceptions of bullying prevention efforts were assessed through six items: (a) Does the school you work in most frequently have a formal prevention efforts—such as school teams, a committee, or prevention program that deals with bullying? (yes/no); (b) Are you currently involved in bullying prevention activities at the school you work in most frequently? (yes/no); (c) I have resources available to help me learn how to effectively intervene with bullying situations? (4-point, [1] agree strongly to [4] disagree strongly); (d) I have effective strategies for handling a bullying situation ([1] agree strongly to [4] disagree strongly); (e) When you try to intervene in bullying situations, does the situation tend to get (4-point, [1] much better to [4] much worse)?; and (f) It is your job to intervene when you see bullying happen (4-point, [1] agree strongly to [4] disagree strongly). Finally, patterns of need for additional training for intervening with bullying related to the five forms and six special populations were assessed utilizing yes/no questions (i.e., Do you think you could benefit from additional training on when and how to intervene with...). These items were adapted from the previously developed, psychometrically sound measure by Bradshaw et al. (2007). All items had “not sure” as an option, which was coded as missing in these analyses. The survey employed a skip pattern, whereby if a participant responded “no” to particular question, he/she would not be asked follow-up questions regarding that particular issue.

For the first set of analyses we utilized logistic regressions to compute odds ratios (OR) and multivariate analyses of variance, thereby comparing teachers and ESPs’ perceptions of their exposure to and concerns regarding bullying. In addition, variations in participants’ perceptions of the different forms of bullying (i.e., physical, verbal, relational, cyber, and sexting) were examined. We also contrasted teachers and ESPs’ perceptions of bullying among special populations (e.g., students who are gender nonconforming, over-weight, have a disability). In the second set of analyses, chi-squares, logistic regressions, and analyses of covariance were utilized where appropriate to contrast teachers and ESPs’ perceptions of bullying policies and involvement in programming efforts. Finally, we examined differences in perceptions of bullying prevention. Given the number of tests conducted, we applied a more conservative p value of .01 for statistical significance.

A set of covariates were adjusted for in the analyses, including amount of interaction between students and participants (with higher scores indicating more interaction with students) as it differed significantly between EPSs and teachers (i.e., more teachers [80%] reported having “constant” interaction with students as compared to ESPs [42%]). We also included school level (elementary and high school, with middle school as the reference category), school location (urban vs. suburban/rural), and survey modality (Web vs. phone) as covariates, and applied sampling weights in all analyses (see next section). Missing data were generally not a concern, as 93% of the sample had no missing data, and each item had less than 2% missing. Given the low rate of missing data, individuals without a response on a particular item were excluded from that analysis.

Two types of weights were applied to the data. First, we applied a propensity score weight to adjust for the mode of survey administration (i.e., Web vs. phone; Rosenbaum & Rubin, 1983; Schonlau, van Soest, Kapteyn, & Couper, 2009). The purpose of the propensity score weights was to make the Web-based survey comparable to the phone-based survey. Each participant was assigned a weight based on his/her propensity score, which was constructed based on 16 different demographic variables (e.g., full- vs. part-time worker, region of the country). These methods are com monly used in large-scale surveys that employ both phone- and Web-based assessments (for additional details, see Schonlau et al., 2009, and Taylor, 2000). Our decision to apply this type of weight was based on preliminary analyses of the data, which suggested that there were some systematic differences in the responses to select survey items based on the mode of survey administration. For example, phone respondents had a tendency to report greater comfort intervening in the different types of bullying situations assessed; this is likely from a social desirability bias among phone participants (Kreuter, Presser, & Tourangeau, 2009; Watts, 2010). As a result, the propensity score weights, along with controlling for survey administration as a covariate in the analyses, allowed us to account for potential bias associated with those respondents who completed the Web survey as compared to those who completed the phone survey. The second weight applied was a rim weight, which is a common weighting approach that enabled us to weight the entire data set to the national population of NEA members (Watts, 2010). Specifically, rim weighting was utilized to weight the sample that participated in the survey to those in the known NEA population. Therefore, the weighted sample reflects the full NEA membership.

Approximately 43% of participants reported that bullying was a moderate or major problem at their school, with teachers viewing bullying as a significantly greater problem than ESPs (OR [H11005] 1.38; p = .01). Participants reported witnessing bullying on their school campus quite frequently. In fact, 62% of the participants indicated that they witnessed two or more incidents of bullying in the last month, with 41% of these respondents having witnessed bullying once a week or more. Roughly half of the participants said students would often report bullying incidents to them. Not surprisingly, the more time participants spent interacting with students the more likely students were to report incidents to them (OR [H11005] 1.35; p < .01). Sixteen percent of participants (16% teachers, 15% ESPs) indicated that parents had reported bullying to them.

The survey also revealed that some participants had been bullied. Approximately 18% of teachers and 14% of ESPs reported being bullied on the job. There were no significant differences in the rates of workplace bullying among teachers as compared to ESPs. The most commonly reported perpetrator was another staff member (45.5% ESPs; 45.2% teachers; p > .01), followed by administrators (28.8% ESPs; 39.1% teachers; OR = 1.88, p > .01), students (28.6% ESPs; 29.5% teachers, p > .01), and parents (10.3% ESPs; 31.8% teachers; OR = 3.62, p < .01).

Teachers were more likely than ESPs to have verbal (OR = 1.48, p < .01) or relational (OR = 1.70, p < .01) bullying reported to them within the past month. The most common form of bullying reported to both teachers and ESPs was verbal bullying, whereas cyber-bullying and sexting were the least likely to be reported to participants. An overall significant difference was observed between teachers and ESPs’ perception of how problematic they viewed various forms of bullying, Wilks's Λ = .98, F(5, 4337) = 19.02, p < .001 η² = .02, d =.29. A follow-up analysis of covariance indicated that teachers viewed all five forms of bullying (physical, verbal, relational, cyberbullying, and sexting) to be more of a concern at their school than did ESPs ( p < .01). However, teachers and ESPs both viewed verbal bullying as the most problematic form and cyberbullying and sexting as the least problematic forms (see Figure 1).

Teachers generally reported feeling more comfortable intervening with different forms of bullying (physical, verbal, relational, cyber, and sexting) than did ESPs, Wilks's Λ = .99, F (5, 4515) = 11.08, p = .01, η² = .012, d = .22. The follow-up analysis of co variance indicated that teachers reported being more confident intervening with physical, verbal, and relational forms of bullying ( p < .01), but not cyberbullying and sexting. Both teachers and ESPs reported that they would be most comfortable intervening with verbal bullying and least comfortable intervening with cyber-bullying and sexting (see Figure 2).

A multivariate analysis of variance indicated that there were significant differences between ESPs and teachers in their perceptions that bullying among special populations was a problem at their school, F(3, 4602) = 14.10, p < .01, η² = .01, d = .20. Teachers perceived bullying related to students’ sexual orientation/gender nonconformity (referred to as sexual orientation; F = 34.5, p < .01) and weight (F = 15.9, p < .01) to be more of a problem than ESPs. Notably, both groups reported that bullying related to students’ weight was the most problematic among the three special populations examined (see Figure 1). Yet, ESPs and teachers differed significantly in their perception of bullying regarding students’ sex, race, and religion, F(3, 4599) = 16.7, p < .01, η² = .01, d = .20, such that teachers perceived sexist remarks (F = 21.9, p < .01), negative racial remarks (F = 48.5, p < .01), and negative religious remarks (F = 12.6, p < .01) to be more of a problem than did ESPs.

There were no significant differences between ESPs and teachers on students’ reporting bullying related to being overweight to them in the past month. However, teachers were more likely to report having students disclose bullying related to sexual orientation to them (OR = 1.54, p < .01). Teachers also were more likely to have bullying via negative sexist (OR = 1.40, p < .01) and racial remarks (OR = 1.54, p < .01) reported to them. The most commonly reported forms of bullying among special populations were comments about the students’ weight, followed by sexist remarks and comments about sexual orientation. The least common types of remarks were religious.

Similar to participants’ comfort intervening across the forms of bullying, teachers tended to feel more comfortable than ESPs intervening when bullying occurred among special populations. Teachers were more comfortable intervening when bullying was related to race, sex, or religion, F(3, 4641) = 14.1, p < .01, η² = .01, d = .20, and more comfortable intervening when bullying was related to sexual orientation, disability, or a students’ weight, F(3, 4641) = 10.04, p < .01, η² = .006, d = .16. Notably, both groups felt the most comfortable intervening with bullying among students with disabilities and the least comfortable intervening with bullying regarding sexual orientation (see Figure 2).

More ESPs (96%) than teachers (92%) reported that their school district had a bullying policy, χ² = 15.0, p < .01. Similarly, ESPs (88%) were more likely than teachers (80%) to report that the bullying problems are adequately addressed by policy, η² = 28.1, p < .001. More ESPs (87%) than teachers (75%) reported that their district's bullying policy is clear and easy to implement, η² = 45.1, p < .01. In fact, teachers were nearly 60% less likely than ESPs to report that the bullying policy is clear and easy to implement (OR = 0.39, p < .01). More teachers (55%) than ESPs (46%) reported that they had received training on how to implement the bullying policy, χ² = 22.5, p < .01. Specifically, teachers were 23% more likely than ESPs to report that that they had received training on their school's bullying policy (OR = 1.23, p < .01; see Figure 3).

ESPs (65%) were more likely than teachers (57%) to report that their school has formal prevention efforts, like program planning teams, committees, or programs, χ² = 16.8, p < .01, with teachers being 33% less likely to report formal prevention efforts at their school (OR = 0.67, p < .01). Yet, compared to ESPs (27%), a greater number of teachers (42%) reported that they were involved in bullying prevention at their school, χ² = 36.3, p < .01. Specifically, teachers were 65% more likely than ESPs to report that they were involved in bullying prevention (OR = 1.65, p < .01; see Figure 3).

An analysis of covariance indicated that ESPs were more likely than teachers to report that there are resources available to help them intervene with bullying, F(5, 4597) = 11.4, p < .01. However, there was no significant difference between the groups in their reports of having effective strategies for handling bullying or in their perceptions that these tactics make the situation worse ( p > .05), with 79% of ESPs and 75% of teachers reporting that they had access to resources to help them intervene. Teachers (90%) were more likely than ESPs (75%) to report that it was “their job” to intervene in bullying situations, F(5, 4642) = 164.5, p < .01.

There were significant differences in ESPs’ and teachers’ desire for additional training related to intervening in physical bullying (OR = 0.60, p < .01), verbal bullying (OR = 0.66, p < .01), and relational bullying (OR = 0.79, p < .01), with ESPs reporting a greater need than teachers (see Figure 4). Both ESPs and teachers equally were as likely to report needing additional training related to cyberbullying or sexting interventions. ESPs were significantly more likely to report need ing additional training in intervening in situations involving special populations examined: specifically, sexual orientation (OR = 0.70, p < .01), being overweight (OR = 0.57, p < .01), disability (OR = 0.58, p < .01), negative comments regarding sex (OR = 0.59, p < .01), race (OR = 0.61, p < .01), and religion (OR = 0.60, p < .01). Across the full sample, the areas of training identified as having the most need were interventions for bullying related to sexual orientation, gender issues, and racial issues (see Figure 5).

Although the research on bullying has increased over the past decade, there remain a number of gaps in the literature regarding potential differences in school staff members’ perceptions of the problem (Bradshaw et al., 2007). The current study aimed to specifically examine ESPs, which is an often neglected sector of school staff, and to contrast their perceptions with those of teachers. Our analyses revealed that teachers and ESPs have different perceptions of and concerns about bullying as well as differing involvement in prevention efforts. For example, it appears that teachers are more likely than ESPs to witness bullying and have students report bullying to them. It is important to note that these analyses adjusted for time spent with students; therefore, there is something unique about the student–teacher relationship beyond the amount of time they spend together that likely accounts for these differences. Teachers are able to get to know their students through their classes and have regular, if not daily, contact throughout the year. In contrast, many ESPs interact with just a few students (e.g., paraprofessional working one-on-one) or students in large groups (e.g., lunch room, riding the bus).

The results suggested that parents were equally likely to report bullying issues to ESPs and teachers. ESPs may be the first point of contact for families when they enter the school (e.g., clerical staff, bus drivers); consequently, parents may be prone to communicate their needs to ESPs in addition to contacting their child's teacher when a bullying situation arises. Thus, ESPs could be a valuable resource to extend home–school communication. Although ESPs may serve as an accessible avenue for reporting peer victimization, educational case law suggests the schools may only be held accountable if the bullying was reported by a full-time teacher or administrator as opposed to a paraprofessional (see Davis v. Monroe County (GA) Board of Education, 1999, and Rowinski v. Bryan (TX) Independent School District, 1996). Given the potential for liability in the case of a mishandled bullying incident, school districts are encouraged to clarify school-wide policies to all staff members (ESPs and teachers) regarding what to do when a parent and/or student reports bullying to them.

The results revealed that teachers were more comfortable than ESPs intervening in bullying situations. Surprisingly, ESPs were more likely to report that they had effective strategies for handling bullying, yet they reported a greater need for additional training. ESPs should be included in school prevention programming efforts because they oversee high-risk areas (e.g., playgrounds, buses, cafeteria). The current finding is consistent with DeLara's (2008) results, which revealed that bus drivers not only witness a considerable amount of bullying, but most of them also felt that they were not included in the district's school safety planning efforts. Current and previous research highlights the importance of conducting school-wide needs assessments across all school personnel because there are varying levels of knowledge and perceived support provided regarding bullying prevention and intervention strategies. Through a systematic evaluation, schools can streamline programming efforts by developing modularized bullying trainings that effectively address subgroups of staff as opposed to creating a one-size-fits-all professional development.

Lastly, the current study's findings indicate that bullying policies exist in many districts, but there seems to be a lack of sufficient instruction on the implementation of those policies. Moreover, ESPs were less likely than teachers to report that they had received training on the district policy. With most states requiring bullying prevention training for school staff as part of antibullying laws (Furlong, Morrison, & Greif, 2003; Srabstein, Berkman, & Pyntikova, 2008), there is a growing need to involve all staff members in bullying prevention programming. Although there are few preventative intervention programs for bullying designed to address the specific needs of ESPs, schools may want to include ESPs by adopting school-wide programs (e.g., Olweus, Limber, & Mihalic, 1999) and Positive Behavioral Interventions and Supports (Waasdorp, Bradshaw, & Leaf, 2012), which have been shown to reduce bullying. Training materials could be developed specifically for different types of ESPs who are most likely to encounter bullying, such as school nurses, bus drivers, cafeteria workers, and playground or hallway monitors.

It is important to note some limitations when interpreting these findings. For example, the data are all self-report, which includes some inherent limitations to their validity. Social desirability may play a role in participants’ responses, and this may vary by the mode of survey (i.e., Web vs. phone). The participation rate was 31%, which is acceptable for survey research, but suggests the potential for bias within the sample. For example, it is possible that staff more involved in bullying prevention efforts or were more concerned about the issue may have been more likely to agree to participate. Although telephone surveys remain the gold standard of surveys, fewer people are willing to take them than in years past (Holbrook et al., 2007), more households are able to block or screen their calls, and more households only have cell phones (Keeter, Kennedy, Clark, Tompson, & Mokrzycki, 2007). As a result, it is difficult and very costly to design a survey to reach a response rate over 50% (usually requiring incentives, prenotifications, 20 call backs, etc.), and it is not unusual for quality telephone surveys to have response rates below 20% (Holbrook et al., 2007). In fact, even after holding survey budget constant, telephone survey response rates have declined considerably in the past 30 years (Holbrook et al., 2007; Lavrakas, 1997).

Although the Internet does provide another venue to reach potential participants, spam blockers and time limitations are also barriers to reaching participants (Kreuter et al., 2009). Thus, we used both phone- and Web-based data collection procedures in the current study. Research does suggest, however, that a decline in response rate does not lead to tremendous error and the highly expensive efforts to achieve high response rates (e.g., 60%), such as continuous call backs and spe cial interviewers trained to convert refusals into completes, do not produce better results than those that have lower response rates (e.g., 30%; Keeter, Miller, Kohut, Groves, & Presser, 2000). However, it is important to note that the weights employed in our analyses allow for us to adjust for some of the potential bias associated with the less than optimal response rate.

There are also several strengths of this study, most notably the nationally representative design, the large sample size, the linkage with the NEA population, the use of propensity scores to address potential sampling biases, and the inclusion of teaching and non-teaching school staff. The weights were derived based on the NEA population, rather than all educators nationally. We lack comparable demographic data from all teachers in the United States upon which to derive weights to weight up the data to all educators nationally. However, we believe the availability to weight the data up to the known NEA membership is a major strength, given the organization's overall presence as the largest teachers’ union in the United States.

Because of the sampling strategy employed and use of self-report data, the respondents were not nested within schools; therefore, multilevel analysis was not warranted. There are several strengths of multilevel analyses and the integration of data on multiple aspects of the school (e.g., size, urbanicity, student population; Bradshaw, Sawyer, & O'Brennan, 2009)—thus future research should include additional school-level factors and student perspectives (e.g., school climate) to provide a more comprehensive view of staff perceptions and responses to bullying and possibly allow for exploration of school contextual factors. Additional research is also needed to examine other factors, such as teacher efficacy and burnout on teachers’ willingness to intervene, or differences by school demographic factors, such as urban versus rural, or elementary versus secondary schools. Given the focus of the current article on contrasting ESPs and teachers, we split the sample into these two broad groups, but there may be more heterogeneity within these two groups than implied by the names. For example, there were some nonclassroom teachers (i.e., other professional staff) in the teacher sample. Similarly, we did not further disaggregate the findings by the ESP roles, although there may be differences in perceptions among the subpopulations of EPSs (e.g., transportation workers vs. paraprofessionals). These subpopulation differences are potentially important, and thus warrant further investigation in future studies.