The relation between school size and location and bullying

Размещено на http://www.allbest.ru/

FEDERAL STATE AUTONOMOUS EDUCATIONAL INSTITUTION

FOR HIGHER PROFESSIONAL EDUCATION

NATIONAL RESEARCH UNIVERSITY HIGHER SCHOOL OF ECONOMICS

St. Petersburg School of Social Sciences and Area Studies

Bachelor's project

The Relation between School Size and Location and Bullying

Petrova Anastasiia Alekseevna

Saint Petersburg 2020

Table of Contents

Introduction

1. Literature review

1.1 Definition of bullying

1.2 School characteristics

1.3 Individual characteristics

1.4 Present study

2. Method

2.1 Participants

2.2 School-level variables

2.3 Data analysis

3. Results

3.1 Descriptive results

3.2 Limitations

References

Appendix

Introduction

Today, bullying is one of the most widespread issues in terms of children`s well-being in schools. Dan Olweus was one of the first who described this problem in 1970s. Nowadays, bullying can be perceived as one of the most common indicator for measuring the quality of school life and climate (Due et al., 2005). Since there is higher prevalence of bullying in some schools than in others, authors consider the influence of school characteristics on bullying along with individual parameters (Wei et al., 2010; Sulak, 2018; Eisenberg et al., 2018).

The aim of this work was to study the relations between the main characteristics of school and the prevalence of bullying and victimization among students. The research question was “How do the school size and type of settlement affect the prevalence of bullying and victimization?”

There were several tasks in this research:

To analyze the difference between settlements according to the characteristics of settlements in Kaluga region (size of population, local budget revenues, families needed in houses, number of schools, libraries, health-care and cultural organizations);

to analyze the difference between schools in different types of settlements according to characteristics indicators from school social passports (percentage of poor families and families living in unsatisfactory conditions, children on intraschool accounting and in juvenile commission);

to analyze the effect of settlement characteristics and characteristics from social passports on bullying and victimization;

to analyze the effect of discipline on bullying and victimization;

to compare the effects of school characteristics and discipline in a general model.

This paper contains several parts. The first one is the literature review where an analytical and critical review of the literature is presented. There are three sections: definition of bullying, where the literature on the types of bullying, the sides of the process, the consequences and methods of studying bullying is analyzed. The second section contains an overview of the works about the influence of various school characteristics on the prevalence of bullying: the size and type of school, the type of settlement, the level of deprivation in the settlement, etc. The third part consists of an analysis of the literature, which highlights the impact of individual characteristics (gender, ethnicity, socio-economic status, and problems with the law).

The second one is the description of present study. There are suitable theoretical frameworks, the peculiarities of this work and formulated hypotheses are presented. For this study, Ecological concept was chosen as a framework as far as the aim of study is to describe the relations of school-level characteristics on the individual behavior. Moreover, the Authoritative School Climate Theory is a framework, too, since this theory suggests that the level of discipline determines the difference in bullying prevalence.

The third part contains the description of method: specification of databases, sample and variables. In this study, I used the latest wave of longitudinal research results that have been collected by the Higher School of Economics scientific and educational laboratory “Sociology of Education and Science”. The scientific interests of the laboratory include studying the education system (school segregation and differentiation), the life and well-being of schoolchildren (academic motivation, migration history), adolescent social networks and the prevalence of risky behaviour, in particular bullying. The goal of the longitudinal research was to study the school climate in Russian schools; the Kaluga Region became the site of the study. Data was collected with the support of organization “Teacher for Russia”. I participated in the collection and analysis of materials of a qualitative part of the study during the student scientific expedition in 2018; I am also a research assistant in laboratory and take part in other projects. I compiled a database on the characteristics of the settlement using the statistics resource of the Kaluga region, which was used in this study and can be used in future works. In this study, there are two dependent variables (status of aggressor and victim), control individual-level variables (gender, status of visible ethnic minority etc.) and school-level variables (type of settlement, size of school, the percentage of poor families etc.).

The fourth part is about data analysis where the chosen method and model evaluation are described. A multilevel logistic regression was chosen for conducting analysis. The first model is null model without variables. The second model contains control individual-level variables. The third model contains main school characteristics. The next one contains meaningful interactions, and the effect of discipline is added in the last model.

The fifth one contains the results of descriptive and main analysis and its interpretation. The preliminary data analysis shows that cities are different from other settlements, gymnasiums / lyceums are different from regular schools. This analysis allows to recode variables about school and settlement type. The HLM analysis shows that the effect of main school characteristics and discipline is stronger for bullying than for victimization. The effect of discipline is stronger than effects of other school-level variables. Schools with high level of discipline have lower prevalence of bullying and victimization than schools with low level of discipline.

The next part contains the critical discussion of results: summation of the results obtained in the framework of analyzed literature and a meaningful interpretation of the association of school characteristics, settlement context and individual characteristics and bullying prevalence. The findings confirm the previous results and are explained through chosen frameworks. The significance of the level of discipline is explained through the students' attitude to the existing school rules: the aggressor does not perceive the rules as fair, therefore, is not afraid of sanctions, and the victim does not trust the school structure because of a sense of insecurity.

The last part is the limitations that the study has, and their impact on the results. The main limitations relate to the collection of data on settlements in the Kaluga region and methods for measuring discipline and peer-reported bullying.

I would like to thank Ivanyushina Valeria Alexandrovna for her suggestions and comments on the conducting analysis. Her valuable and informative comments helped me in understanding the structure of the analytical part of the work.

1. Literature review

1.1 Definition of bullying

Bullying is one of the most common problems today (Due et al., 2005). Bullying and victimization are observed everywhere: in schools of different countries (Akiba, 2010; Adams & Mrug, 2018; Owoeye & Yara, 2011; Akpunne et al., 2019; Saarento et al., 2013); in big and small schools (Wei et al., 2010; Sulak, 2018; Olweus, 1994; Klein & Cornell, 2010; Welsh, Stokes & Greene, 2000; Ma, 2002); in urban and rural schools (Sulak, 2018; Olweus, 1994; Eisenberg et al., 2018; Leadbeater et al., 2013; McCaskill, 2013; Renfro et al., 2003).

The definition was given by Olweus (1994): bullying is a repetitive aggressive behaviour that is performed by one person or group towards a peer who is obviously weaker in the physical, emotional or status aspect within a particular group. There are different types of bullying: physical (kicks, theft of personal items, damage to property), verbal (insults, nicknames, curses), relational (exclusion from the group) and indirect (dissolving gossip and rumors) (Gredler, 2003; Borg, 1998; Fisher, 2015). Due to the widespread use of the Internet, a cyberbullying which is provided on various media platforms, social networks and forums became the subject of research, too (Menesini & Nocentini, 2009). In the process of bullying, there is always a victim who experiences victimization on himself or herself (Akiba, 2010) and an aggressor who initiates bullying (Cornell & Huang, 2016). In addition to these roles, the behaviour of bystander can be described (Midgett & Doumas, 2019; Ybarra et al., 2019).

Bullying and victimization have a tremendous impact on the well-being of children and adolescents. The negative effect of bullying has an influence on both victims (Tharp-Taylor et al., 2009) and aggressors (Juvonen & Graham, 2001, Chapter 14). Scholars mainly focus on the development of teenage depression and anxiety (Craig, 1998), suicidal thoughts (Wolke & Lereya, 2015), use of psychoactive substances (Tharp-Taylor et al., 2009) and development of various psychological disorders that manifest themselves in adulthood (Belfer, 2008).

Aggressors and victims perceive bullying differently: bully may not consider these actions as something traumatic, but the victim always feels victimization. Different perceptions lead to creating various tools for detecting bullying. First, self-reported bullying, i.e. bullying reported by aggressors and victims (Datta et al., 2017; Beckmann et al., 2019; Ma, 2002). Second, peer-reported bullying when students observed bullying in a group (Obermann, 2011; Branson & Cornell, 2009). Third, teacher-reported bullying; teachers describe relations between students and their attitudes to school policy (Muijs, 2017). Another one, parental-reported bullying, i.e. parental reports about whether their child is an aggressor or a victim (Stives et al., 2019). Comparison of peer-nominated and self-reported bullying often appears in studies, since participants identify such cases in different ways. Branson and Cornell (2009) found that there is a “poor correspondence” between self-reported and peer-reported cases. The authors argue that stigmatization and potential penalties for such behaviour can lead to students not reporting their involvement in bullying. The coincidence between reports from peers and children involved in bullying is affected by the overall level of moral disengagement in the class: if it is high, the reports will match (Obermann, 2011). In this work, I look at the bullying from the sides of aggressors and victims.

1.2 School characteristics

Bullying can occur variously in different schools depending on characteristics of a particular school. Private and public schools may differ in bullying level (Shujja et al., 2014). In public schools, there were more cases of bullying and victimization than in private ones (Machimbarrena & Garaigordobil, 2017). It can be explained through the socio-economic status of the child: as a rule, children with a high socio-economic status study in private schools. Such schools strive to create a favorable educational environment, and parents invest resources to get a quality education for children; parents of children in public schools are less interested in the ideal educational environment for their children (Shujja et al., 2014).

Gymnasiums / lyceums and regular schools differ in the prevalence of bullying. The lower the quality of education in school, the more often bullying occurs there. In Realschule / Werkklasse - schools with a low quality (main classes) - the level of bullying is higher than in Gymnasium - schools with a high quality of education, where students have a high level of involvement in educational process (classes of advanced level) (Perren & Hornung, 2005). Schools in which children are strongly involved in the academic process also demonstrate a high level of discipline. (Way, 2003). Perren and Hornung (2005) associate the high prevalence of bullying in `low-quality schools' with the low parental involvement and family support.

Schools located in different territorial units (city, suburb and village) may exhibit different spread of bullying. It can be explained, among other things, due to the characteristics of the territory. Rural schools may receive less financial support and have poor conditions, which leads not only to low teachers` motivation, but also to an unfavorable school climate (Owoeye & Yara, 2011). The type of community is related to the socio-economic status of the population: parents with a low socio-economic status are not interested in having children study well, as the family needs their help at home (Owoeye & Yara, 2011). An unfavorable educational atmosphere and school climate leads to a decrease in academic performance. A low level of academic aspiration and achievement may cause a student to be involved in bullying as a bully or a victim (Woods & Wolke, 2004; Strшm et al., 2013).

In rural schools, parents and teachers do not discuss social norms and, in particular, aggression with students (Sulak, 2018; Olweus, 1994). This leads to low student awareness of bullying and high prevalence of victimization. In rural areas, children may feel insecure (e.g., frightening adults or stray dogs) (Leadbeater et al., 2013). This leads to an increase of anxiety, which makes such children “visible” to aggressors as victims.

The presence of a “criminal” context around the school is associated with a high level of bullying in schools. The relationship between bullying prevalence and adolescent involvement in delinquency has been examined by researchers (Olweus, 1994; Perren & Hornung, 2005, Ttofi et al., 2011), as well as the relationship between bullying and other aggressive behaviours (Bandura, 1978). Bradshaw and colleagues (2013) showed that non-urban teens who were involved in the bullying process were more likely to get into gang. As a rule, teachers in such schools pay more attention to bad academic achievements than to bad behaviour; this allows students to feel unpunished. In large cities, the crime rate in the district where the school is located can play an important role. Schools located in unsafe criminal areas of the city show a level of bullying higher than rural schools: children reproduce among classmates the aggression that they observe outside the school (Thomas et al., 2006).

An equally important indicator is the socio-economic status of the school. One of the most used ways to determine the socio-economic status of a school is to describe the neighborhood in which the school is located. Researchers describe the school using the context of the district in which it is located: official statistics can determine which areas are low income or high income (through the number of people with a certain income), and then researchers choose educational institutions in each district, which allows them to achieve full view on the situation (Uludasdemie & Kucuk, 2019). Most authors get the same results: in schools that are located in areas with low socio-economic status, that is, schools that themselves have low SES, the prevalence of bullying is higher than in schools with medium or high levels of SES by neighborhood (Khoury-Kassabri et al., 2005; Jansen et al., 2012).

School size is one of the key indicators that can affect the prevalence of bullying and victimization. It is defined as the total number of students studying at the time of the study at school (Wei et al., 2010; Sulak, 2018). It is logical to assume that in large schools and classes the prevalence of bullying is higher, because more students account for more victims and aggressors. This may turn out to be a false conclusion, since it is necessary to take into account not the absolute number of aggressors and victims, but the prevalence of bullying (Olweus, 1994). Speaking about the prevalence of bullying, it is more difficult to build discipline and trusting relationships between teacher and students in large schools, and the school climate becomes unfavorable due to lack of resources to control students (Sulak, 2018). In large schools, ethnic diversity is higher than in small ones: the probability of conflict is higher when there are more groups of people who look not like you (Klein & Cornell, 2010; Welsh et al., 2000).

In small schools, it may be easier for aggressors to find a victim; because of the small number of students, a person has become a victim more than once, thereby becoming a visible target for the aggressor (Ma, 2002). In small schools, students can more clearly understand that bullying is a socially unacceptable action, and it is more difficult for an aggressor to hide bullying from others, and victims are ready to report such aggressive behaviour (Ma, 2001). The size of school may have no effect on the prevalence of bullying and victimization, since the individual characteristics of the student determine his or her involvement in bullying more than the characteristics of the school (Cornell & Huang, 2016; Wei et al., 2010).

Discipline can influence on the prevalence of bullying. There are many measures of this parameter (Socolar et al., 2004; Skiba et al., 2002; Welch & Payne, 2010), and one of them refers to a clarity of school rules. For example, Cornell and colleagues (2015) used this classification into their study, asked adolescents whether they suggested that school rules are honest and understandable, and sanctions are fair. Perceiving existence rules as fair system, students tend to be less aggressive with their peers because they know about possible punishments (Taylor, 2006). The low level of discipline also refers to a big prevalence of victimization, since victims do not trust school structure guided by their feeling of unprotected and insecure in school (Schreck et al., 2003; Khoury-Kassabri et al., 2004).

The deprivation level in the settlement can affect the students' school life. The deprivation level negatively affects the academic achievements of students (Garner & Raudenbush, 1991). In turn, Russian colleagues examined the impact of deprivation on schools` resources, but not on academic achievements (Yastrebov et al., 2013). There is no relevant literature about the influence of deprivation on bullying and victimization; nevertheless, I include these parameters in our study.

1.3 Individual characteristics

Many authors associate the individual characteristics of students with their involvement in the bullying process. Many studies suggest that boys are more involved in bullying than girls (Stubbs-Richardson et al., 2018; Grizel et al., 2012). Boys are more involved in physical and verbal bullying, and girls are more likely to use indirect bullying and cyberbullying (Wang et al., 2009; Kessel Schneider et al., 2015).

The social and economic status of the family is also used in studying bullying. There are various indicators of SES (Galobardes et al., 2006a, 2006b). First, parental education. A low level of parental education can contribute to the involvement of a teenager in the bullying process as victims and aggressors (Kavanagh et al., 2018; Knaappila et al., 2018). Second, parental occupation can also be significant: Knaappila and colleagues (2018) found that at least one parent's lack of work is positively associated with being bullied and bullying other students. The information about parental occupational status can be coded through the International Standard Classi?cation of Occupations (ISCO) and then transformed into an occupational scale the International Socioeconomic Index of Occupational Status (ISEI) (Ganzeboom et al., 1992). The relationship between different indicators of socio-economic status and involvement in bullying is demonstrated by Uludasdemir and Kucuk (2019): parents with a high education are most likely to have sufficient earnings to allow their child constant access to the Internet that can lead to a higher involvement in cyberbullying. Third, income of the family is used as indicator of SES, too: Due and colleagues (2009) found that adolescents whose families have a low level of income are more likely to be victims than adolescents with an average or high level of income in the family.

Many researchers suggest there is a connection between the behaviour of aggressors and victims and problems with the law (Olweus, 1994). Thus, bully perceives aggressive behaviour as normal and reproduces aggression outside of school (O`Brennan et al, 2009). Victims, feeling unprotected, practice substance use and have mental disorders (Belfer, 2008). Ttofi and colleagues (2011) conducted a meta-analysis and systematic review, concluding that adolescents who produce aggression towards peers are more likely to commit crimes than those who were not bully. Perren and Hornung (2005) found that adolescents who are both victims and bullies are more likely to be involved as victims and / or perpetrators than children not involved in bullying. The authors attribute this link to lack of family support and peer attention.

Ethnicity can also be a significant indicator for determining the level of involvement of a teenager in bullying or victimization. Cohen and colleagues (1990) examined that as far as one of the principles of bullying is an imbalance of power, proportions of ethnic majorities and minorities connected with the prevalence of bullying. Members of ethnic minorities can become victims (Mouttapa et al., 2004; Verkuyten & Thijs, 2002). There are results which examined the opposite situation when member of the ethnic majority become being bullied by representatives of ethnic minorities (Graham & Juvonen, 2002). Moreover, several researchers supposed that there is no relation between ethnicity and the level of victimization (Seals & Young, 2003).

1.4 Present study

Involvement in the bullying process is explained not only through individual characteristics, but also through relationships with peers, teachers, family and through neighborhood characteristics (Espelage & Swearer, 2009). Ecological framework (Bronfenbrenner, 1986; Swearer & Doll, 2001) suggests that each of these levels can influence behaviour. In this work I consider schools, taking into account the context of the settlement in which they are located. It is important to define a territorial context that is independent of schools for understanding conditions in which school creates an educational environment and climate for students (Yastrebov et al., 2013). The peculiarities of the Russian educational context imply that various indicators of the school are related to each other. A rural school in Russia is usually small; it is not a gymnasium or a lyceum; mostly, students have low SES; there is low percentage of highly educated mothers. The context the social composition of students (ISEI, family resources), school characteristics (size, type of location etc.) and settlement characteristics for determining the deprivation level (size of the population, local budget revenues etc.). There is a need to understand: does the fact that some schools have more bullying than others, depend on the main characteristics of schools or the context of the territory in which it is located? The work takes into account that part of the characteristics of the school is related to the characteristics of the village, so we take into account the general level of well-being of the village / city through the size of population, the number of schools, the number of libraries, etc. The information about additional school characteristics which was provided through schools` social passports is used (number of children in juvenile commission, number of poor families at school etc.).

I expect that:

Hypothesis 1: bullying and victimization are more common in large and / or urban schools than in small urban and rural schools.

The probable explanation of the connection between the effect of the school context and the aggressive behaviour of students is quite straightforward: it is difficult for schools with certain characteristics and / or in a certain context to maintain educational discipline and monitor student behaviour (Klein & Cornell, 2010). From the point of view of the Authoritative School Climate Theory, bullying and victimization will occur in any school (large / small, rural / urban, etc.) if it does not support a certain level of discipline. Pupils should feel that the school has strict rules that if they are violated, the punishment will be fair (Cornell et al., 2015; Jeong et al., 2013).

Hypothesis 2: high level of discipline at school is negatively associated with the frequency of bullying and victimization among students.

Since bullying is closely related to the individual characteristics of students, gender, GPA, visible ethnic minority status, maternal education and socio-economic status of the family are taken into account as control factors at the individual level.

2. Method

2.1 Participants

I used the data from a longitudinal survey in which 249 schools of the Kaluga Region participated - there are more than 80 students in these schools. The study was conducted by scientific-educational laboratory “Sociology of Education and Science” Higher School of Economics with the support of the Ministry of Education of Kaluga Region and organization “Teacher for Russia”. The School Climate Assessment Toolkit, which was developed by laboratory staff, was approved by the HSE Commission for Inter-University Surveys and the Ethical Assessment of Empirical Research Projects. Pupils from sixth to ninth grades (about 20,000 participants) participated in the survey, questionnaires were sent to the school in electronic form, and students filled them in computer classes under the supervision of teacher. For this work, forms of participants in the 9th grades of 2019 (4645 students) were selected.

After a multistage screening procedure, 4364 observations remained. The procedure had several reasons for dropping observations: completion of the survey too quickly, questionnaires with fake nationalities and languages (for example, Elven), questionnaires with unrealistic age (for example, 90 years), those who admitted their answers were not honest on the validity question “How many questions did you answer honestly.”

The final sample contained 4364 observations: 52% of female, 8.8% visible to ethnic minorities (Azerbaijanis, Armenians, Georgians, Uzbeks, Tatars, Tajiks, Moldovans and others), 63.9% of students have mothers with higher education. Among students, mean GPA in core subjects is 3.96.

Schools in the Kaluga Region provided social passport data. This database contained important information about the social well-being of children in school. The base with the characteristics of the settlement was collected using the Kalugastat statistics resource. This was done to distinguish between urban and rural settlements, since the Kaluga region is a heterogeneous region in which a small city may look like a large village. During the compilation of the database on settlements, information for three years (2016-2018) was collected to increase the chances of making complete data; for analysis, the fullest years were selected. There were 22.9% of missing data, and this information was mainly for small settlements.

Measures

Individual-level variables

Bullying behaviour. Pupils were asked how often they bullied one of their peers at school last year (for example, “You beat someone”). Pupils could choose one answer from 1 to 4, where 1 - never, 2 - rarely, 3 - often, 4 - very often. The variables were recoded. The final variable was binary, where 0 was the complete absence of the aggressor's experience and the cases when the adolescent had the aggressor's experience in only one of four questions (94.5%), and 1 was the presence of bullying experience in 2-4 questions (5.5%).

Bullying victimization. Teens were asked to rate on a scale of 1 (never) to 4 (very often) how often they were bullied over the past year (for example, “The whole class joked about you”). The final variable was also binary: 0 - the complete absence of the victim's experience and the cases when the adolescent had such experience in only one of six questions (93.6%), and 1 - the experience of victimization in 2-6 questions (6.4%).

Control variables

Students indicated their gender (0 = female, 1 = male), status of the visual ethnic minority (0 = visual ethnic majority, i.e. Russians, Ukrainians, and Belarusians, 1 = visual ethnic minority) and mother's level of education (0 = no higher education, 1 = higher education). Open questions about the parental occupations were coded according to the ISCO classifier, and then transcoded into ISEI (Ganzeboom et al., 1992). ISEI takes values from 10 to 75, where 10 was the most low-skilled specialty, and 75 was the most highly qualified specialty, and the variable was centered on the class level. The GPA score contained values from 2 to 5, and this score was calculated for each student based on grades on core subjects, and then was centered on the level of class.

2.2 School-level variables

The level of school discipline. The four-item scale was used to measure the level of school discipline through questions about the clarity of school rules (for example, “In our school, bullying is unacceptable,” “In our school, punishments are fair”). Each question was answered using the Likert scale (1 = strongly disagree, 2 = more likely disagree, 3 = more likely agree, 4 = strongly agree). The average value for each student was aggregated at the school level (the scale was centered by grand mean), the Cronbach's Alpha for this scale was 0.75.

School type. Schools were divided into three groups by type: 0 = school up to grade 11, 1 = school up to grade 9, 2 = gymnasiums, lyceums, schools with in-depth study of subjects that are located only in cities.

Location. All settlements were divided into several groups according to their type: 0 = city, 1 = township (поселок) and railway station, 2 = village (село) and village council (сельсовет), 3 = countryside (деревня) and state farm (совхоз).

The percentage of mothers with higher education at school became a control variable that was centered through grand-mean.

The ISEI score for schools was counted as a mean based on values of individual ISEI and than centered through grand-mean.

For analysis, the following variables were selected from social passports: school size, the percentage of poor families, families living in unsatisfactory conditions, adolescents on intraschool accounting and teens registered in juvenile commission. A student on account in juvenile commission must be on intraschool accounting, but not vice versa. All these variables were centered by grand-mean.

Settlement characteristics

Information on the number of schools (2017), health care organizations (2018), cultural organizations (2016), public libraries (2017), the number of population (2018), the number of families who need housing (2016), local budget revenues (2018) was selected as indicators of the deprivation level of the settlement. All these variables were also centered by grand-mean.

2.3 Data analysis

As a method of analysis, multilevel logistic regression was used. This method allows us to answer the research question about the relationship between the individual characteristics of the student and the characteristics of the school. Based on the results of comparing full and null models with different effects, a fixed effect was used in the work (better log likelihood indicators). Models were evaluated through Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), log likelihood, deviance and Pseudo R2. According to the Intraclass Correlation (ICC), 8.7% of variances for bullying and 5.7% for victimization occurs at the school level.

I created a null model on the two dependent variables. The second model contained all individual level variables (gender, ISEI of family, status of visible ethnic minority, GPA, mother`s higher education). I added the school level variables in the third level model (with the percentage of highly educated mothers as control variable). The fourth model contained the significant interactions between variables. I added the discipline in the fifth model to check this effect on control of other significant variables from two levels.

3. Results

3.1 Descriptive results

Variables on the number of schools, health care organizations, public libraries and cultural and leisure organizations are related to each other (see Appendix Table 1). These variables were collected through Exploratory Factor Analysis into one, which reflects the level of infrastructure development.

According to the results of correlations between settlement characteristics, including the new factor, showed that the higher the population, the higher all the parameters (see Appendix Table 2). victimization bullying harassment

The results of comparing the average settlement characteristics between the types of settlements are presented in Table 3. Kaluga and Obninsk were excluded as the two largest cities, which are very different from the rest. The cities significantly differed from the rest of the settlements, but the remaining settlements did not differ from each other, therefore the variable recoded: 0 = city, 1 = all rural settlements.

Table 3 Results of One-Way ANOVA between types of settlement and settlement`s characteristics (without Kaluga and Obninsk)

Almost all Pearson product-moment correlations are significant (Table 4.). All characteristics negatively correlated with school size: the more students, the lower the percentage of adolescents on juvenile commission and on intraschool accounting, the lower the percentage of poor families and families who live in unsatisfied conditions. The percentage of teens on juvenile commission and on intraschool accounting was positively correlated with each other: the more ones, the more others. Children on juvenile commission are automatically placed on intraschool accounting. The percentage of poor families was positively related to the percentage of children in intraschool accounting, and the percentage of families living in unsatisfactory conditions was related to the percentage of children on juvenile commission.

Table 4 Correlation among characteristics from schools` social passports

A significant difference in the average for all characteristics is observed both for different settlements and schools (see Table 5 and Table 6). The results lead to recoding the variable about the type of school into binary, since gymnasiums / lyceums are significantly different from other schools (others do not differ from each other): 0 = schools up to 11th and 9th grade, 1 = gymnasiums / lyceums. A new variable in the form of an interactive effect between the type of school and settlement was created: 0 = rural regular schools, 1 = city regular schools, 2 = city gymnasiums and lyceums. The results of ANOVA are presented in Table 7. All groups significantly differed from each other through almost all characteristics: there was no difference in the percentage of adolescents on juvenile commission in rural regular schools and urban regular schools.

Table 7 Results of One-Way ANOVA between types of school in different settlements and school characteristics

Model results

Results of HLM for the status of the aggressor are presented in Table 8. As we can see through measures of model fit (for example, lower AIC), each next model was better than the previous one. Model 2 contained control variables of the first level. Gender was significant: boys were more likely to be aggressors than girls, the effect ceased to be significant under control of other characteristics. Students who had higher GPA than the class average were less likely to be aggressors (B = -0.256, OR = 0.77, p <0.01). ISEI was not significant, so this was excluded from the model. Representatives of the visual minority were more likely to be aggressors (B = 0.988, OR = 2.26, p <0.01).

In model 3, all variables of the first level are significant. The effects of visual ethnic minority status and ratings are the same as in the previous model. Boys were more likely to be aggressors (B = 0.151, OR = 1.16, p <0.01). A child whose mother has a higher education is less likely to become an aggressor (B = -0.053, OR = 0.94, p <0.01). In this model, the percentage of educated mothers (control variable at the school level) and the type of school, locality and school size (variables of interest) were added.

School characteristics from social passports were not significant. A student in school with a high percentage of educated mothers was less likely to be an aggressor (B = -0.053, OR = 0.94, p <0.01). A student from an urban gymnasium/lyceum was less likely to be an aggressor compared to a student from a rural regular school (B = -1.25, OR = 0.28, p <0.01). A student from an urban regular school was less likely to be an aggressor (B = -0.26, OR = 0.77, p <0.01) than student from rural regular school. A student in big school was more likely to be an aggressor than a student from small one, but the size of the effect was very small (B = 0.001, OR = 1.001, p <0.05). Settlement characteristics were not significant.

Model 4 contained a significant interactive effect between school size and grades, although the size of the effect was small: students with high GPA in large schools were less likely to become aggressors due to their grades (B = -0.001, OR = 0.99, p <0.01). This effect did not change the significance and size of the remaining variables.

In model 5, the level of discipline in the school was added. Gender, GPA, and mother's education did not significant. The effect of visual ethnic minority status did not change in size, it was significant: representatives of the visual ethnic minority were more likely to be aggressors (B = 0.973, OR = 2.64, p <0.01). The school size and the percentage of educated mothers became not significant. The effect of urban gymnasiums was significant, but the effect of urban regular schools was not: students from urban gymnasium/lyceum were less likely to be aggressors than students from rural regular schools (B = -1.158, OR = 0.31, p<0.01). The interactive effect between GPA and school size remained the same. The discipline effect was the largest and most significant. Students in schools with a high level of discipline were less likely to be aggressors than students in schools with a low level of discipline (B = -1.565, OR = 0.20, p<0.01). This effect greatly influenced the significance of the remaining variables. Interactive effects between the type of school and community and the size of the school were not significant. The interactive effect between the school characteristics (type, location, size) and the level of discipline turned out to be not significant.

Table 8 Multivariate multilevel models for bullying

The results for victim status are presented in Table 10. There was gradual improvement of models compared to each previous one, however, in comparison with models for bullying, these models worked worse. Model 2 showed the significance of control variables of the first level. Boys were less likely to be victims compared to girls (B = -0.515, OR = 0.59, p <0.001). The status of a visible ethnic minority was not significant, unlike previous models. The mother's higher education was significant: teens whose mother had higher education were less likely to be victims (B = -0.224, OR = 0.79, p <0.001). GPA remained a significant variable: students with higher GPA than the class average were less likely to be victims (B = -0.285, OR = 0.75, p <0.001).

Model 3 contained variables for the school level which were not significant in model. Settlement characteristics were not significant for victimization. School characteristics from social passports were not significant, too.

An interactive effect between school size and GPA was added into Model 4, and it did not change the size and significance of other effects. The interaction was significant, but the size of the effect was very small with a significant effect of GPA, but not a significant effect of school size. Pupils with high GPA in large schools were less likely to become victims due to their grades (B = -0.0005, OR = 0.99, p <0.001).

The level of school discipline was added into Model 5. Adding this effect did not change the size and significance of the effects from Model 4. The effect of discipline was the greatest. In schools with high level of discipline, a student was less likely to be a victim than a student in schools with low level of discipline (B = -0.769, OR = 0.46, p <0.001). The size of the discipline effect for victimization (B = -0.769, OR = 0.46) was smaller than the size of this effect for bullying (B = -1.565, OR = 0.20). The interactive effect between school characteristics (type, location, size) and level of discipline was not significant, too.

Table 9 Multivariate multilevel models for victimization

Discussion

The findings related to an authoritative school climate theory as a theoretical framework that explains the low prevalence of bullying and victimization through a high level of discipline as one of the factors in a favorable school climate. Students in high-discipline schools were less likely to be aggressors and victims in both large and small schools, both rural and urban. The school size was significant only under control of GPA. School characteristics were not related to bullying behaviour and victimization under control of discipline.

Discipline

School discipline consists of a clear understanding of school rules by students. A high level of discipline means that students see the justice of punishment and the inadmissibility of bullying. Multilevel regression analysis showed that schools with a high level of discipline had a lower prevalence of bullying and victimization than schools with low level of discipline. This work confirms the results of a study conducted among seventh and eighth graders on several types of victimization (Cornell et al., 2015). These results confirm the authoritative school climate theory: maintaining a healthy school climate with rules which students understand and follow increases discipline in any school regardless of type and size (Cornell et al., 2015; Jeong et al., 2013). Discipline had a smaller effect on the prevalence of victimization compared to the prevalence of bullying. The aggressor does not consider the school rules as honest and fair; violating them, bully is not afraid sanctions, because he or she observes their injustice. If teenagers believe that school rules are fair, then they tend to trust the school and show less aggression (Tyler, 2006). Victims may be distrustful of existing rules because feel unsafe in school, which leads to feelings of alienation (Schreck et al., 2003; Khoury-Kassabri et al., 2004).

Settlement characteristics

There was a need to check the deprivation level of the settlement due to the heterogeneity of Kaluga region. Previous studies showed that the territorial context is important for studying the school climate (Uludasdemie & Kucuk, 2019). Our ANOVA analysis showed that the types of settlements differ in characteristics. There was an explanation through the socio-economic status of the settlement in literature. Owoeye and Yara (2011) found that the low socio-economic status of rural settlements, in contrast to urban ones, leads to a lack of resources at schools: financial support, quality teachers etc. Multilevel analysis showed that all settlement characteristics were not related to bullying and victimization under control of other variables. Yastrebov and colleagues (2013) obtained similar results about the lack of influence of the deprivation level on another aspect of students' life - academic performance. The authors explained that the availability of resources at the school and the social composition of students, depending on the type and location of the school, can affect student behaviour.

School characteristics

The effect of school size was very small. Nevertheless, bullying prevalence in large schools was higher than in small ones without taking into account discipline. The significance of school size was confirmed in other studies (Sulak, 2018; Olweus, 1994), where the authors explained this through the quality of the discipline: it is easier for teachers to control and build trust with students in small schools than in large ones. In large schools, the possibility of student heterogeneity increases, and it is easier for an aggressor to find a victim (Klein & Cornell, 2010; Welsh et al., 2000). School size was not related to the victimization prevalence, as confirmed by previous studies (Khoury-Kassabri et al., 2004). Being a victim can...