The Effect of Gender on Mobbing Experienced by Teachers at School: A Meta-Analysis Study

The aim of this research is to detect the effect of teachers’ genders on mobbing they experienced at schools. As one of the methods used to synthesize research results, meta-analysis method was used in this study. It was determined that there are 72 studies which are deemed appropriate for the inclusion criteria of meta-analysis study. In this meta-analysis study, 72 studies which form a sample consisting of 24954 people were calculated. As a result of the process of combining in random effects model, a statistically significant ES which was at the level of -0.02 was found in favour of male teachers. When results are evaluated together, it is observed that there is a difference, which may not be considered significant, between male and female teachers in terms of teachers’ perceptions about AM at school. As a result of the conducted moderator analysis, moderator effects of publishing type (p=0,03), of education level (p=0,03), and of gender of the researcher (p=0,02) were determined. It was observed that while mobbing perception of male teachers are higher in studies carried out in master’s and doctoral theses, mobbing perception of female teachers are higher in studies carried out in articles. Moderator effects of school type (public, private, and public/private) (p=0,63) and of the region where the research was carried out (p=0,17) could not be determined. It was observed that mobbing perception of female teachers who work in Central Anatolia and Eastern Anatolia are higher compared to other regions.


INTRODUCTION
Mobbing is a word derived from Latin which means intimidation, psychological violence, oppression, siege, harassment, disturbance or distress. Derived from the words "mobile vulgus" which mean "inconsistent crowd", the word "mob" means irregular crowd which commits illegal violence, "community of guilty", or "gang" in English (Aytaç 2017;Minibaş-Poussard and İdiğ-Çamuroğlu 2009). Translations which were adopted most by Turkish researchers for the term "mobbing" are "intimidation" (Aras 2012;Akkar 2010;Ertürk 2011), "psychological terror" (Özler and Mercan 2009), "emotional abuse" (Uzunçarşılı and Yoloğlu 2007), and "psychological violence" (Aydın 2009;Tutar 2004;Yaman 2009). In this study, (in Turkish version) instead of the term "mobbing", the term "intimidation" which is commonly used as equivalent of mobbing was adopted. Mobbing is generally considered psychological violence or harassment by one or more people that arises from systematically hostile and unethical practices to other person or people (Leyman, 1996).
scales and different independent variables of teachers (gender, branch, marital status, level of education, faculty of graduation, seniority etc.) were used. As a result of these studies, results which were statistically significant and insignificant and showing different findings in terms of subgroups of independent variables were obtained. Within the scope of these results, the effects of gender roles and responsibilities on acts of mobbing and perceptions about mobbing that teachers experienced at schools were revealed.
As a result of the fact that different studies carried on AM had different results in terms of gender variable, Ertürk (2013), Karabacak and Akın (2014), and Aytaç (2017) stated that metaanalysis studies on this issue need to be carried out. It is important to detect whether gender roles of teachers are effective in their experience of AM or in their perception of AM. When the literature is analysed, since there is not any study which analyses the effect of gender on AM that teachers experience at schools through meta-analysis methods, this study shall contribute to the field. The problem of this study is to determine if the gender (male/female) is effective in mobbing perception of teachers. The aim of this research is to detect the effect of teachers' genders on mobbing they experienced at schools.

METHOD
Research model, data collection and data analysis sections are given in this chapter.

Research Model
Meta-analysis method was used in the study. Meta-analysis method is a method of systematically analyzing and synthesizing the data of quantitative studies on the same subject independently. The purpose of a meta-analysis is to reveal facts on similar topics, combine findings of several studies conducted at various times and places, reach the most accurate quantitative results through increasing sample size. Group difference meta-analysis method was used in the analysis of the data. In a meta-analysis study, two models are used to calculate the general effect size: fixed and random effects models. At the stage of combining studies (general effect), which model shall be used is decided in accordance with these assumptions, and model can be selected either before the study or at the beginning of the study. While fixed effects model is selected in replication studies, random effects model is suggested especially in social sciences since operational and procedural variance is not present in most of the studies. As one of the comparative meta-analysis methods, Group Difference method was used in the analysis of data (Aytaç, 2019;Cumming, 2012;Ellis, 2012;Hedges and Vevea, 1998). Q and I 2 statistics are also used in alternative model selection, and model can be selected. However, especially in social sciences, since the aim of synthesizing is to make unconditional inferences for most researchers, the best option is to choose random effects model (Card, 2012;Cumming, 2012;Dinçer, 2014).

Data Collection Tool
Master's theses, doctoral theses, and research articles which discuss this issue in Turkey were taken into the scope of the study. In order to have access to the relevant studies, the keywords "intimidation", "psychological abuse", "mobbing", "psychological mobbing", and "psychological violence" were searched in YOK National Thesis Centre and various search engines (Ulakbim, Google Academic et al.). As a result of this search, it was observed that there are 95 studies in the scope of this study. It was determined that there are 72 studies which are deemed appropriate for the inclusion criteria of meta-analysis study. As one of the basic inclusion criteria, in Turkey addressing teachers' perceptions of mobbing field work were taken. Inclusion criteria used in the selection of the studies which would be included in the research are given below: (i) Criterion 1: Published sources: Master's theses, doctoral theses, and research articles published in the literature were taken into the scope.
(ii) Criterion 2: The appropriateness of dependent or independent variables in the studies for meta-analysis study: It was paid attention that studies included in meta-analysis studies to reach ES were empirical studies and that gender of teachers were taken as independent variable.
(iii) Criterion 3: Its inclusion of quantitative data which is necessary for meta-analysis: It was paid attention that it included quantitative data (average, standard deviation, number of samples, p value etc.) in order to calculate effect sizes which are necessary for meta-analysis, (iv) Criterion 4:Time/Years: It was paid attention that studies were carried out in Turkey between 2007 and 2019.

Exclusion Criteria, Reliability and Validity of the research
Twenty-three different studies obtained as a result of literature scan were excluded from the carried out meta-analysis study since they were not deemed appropriate for the inclusion criteria because they were carried out in different samples (school administrators and academic staff), they did not have necessary statistical data for meta-analysis, and they included only qualitative findings. In this study, data were coded by using two coders. Cohen's Kappa statistics was used in order to provide reliability between coders which process studies to coding protocol, and the reliability was found to be 0.92. This result indicates a good concordance between coders.
The fact that all accessible studies which are deemed appropriate for the inclusion criteria of meta-analysis were scanned by using all data bases is an indicator of the validity of the research (Petitti, 2000). In the context of accessing to all studies as a result of the scan, it can be stated that validity was ensured.

Analysis of the Data
In this meta-analysis study, random effects model was used in the calculation of overall effect size. In the study, females were included the experimental group and males were included the control group. Therefore, if calculated effect size was positive, then it was in favor of females, if calculated effect size was negative, then it was in favor of males. Current study is applied a statistical pocket programme of 2.2.064 version for Meta-Analysis [CMA-Comprehensive Meta Analysis] for comparing effect sizes of each study with variances and groups (Borenstein, Hedges, Higgins and Rothstein, 2009). SPSS version 20.0 pocket programme has been used from coder reliability. Studies, meeting the inclusion criteria and were included for the current study, had 0,05 significant level. Therefore, significance level of the statistical analysis was determined at 0,05 level.

RESULTS
Findings acquired from researches within the scope of meta-analysis study (publication bias, forest plot, random effects model, and moderator analysis) are given in this chapter.

Publication Bias
In order to understand whether the studies included in the meta-analysis cause publication bias or not, methods such as Funnel Plot, Rosenthal's fail-safe N, Orwin's fail-safe N number, Duval's and Tweedie's Trim and Fill, and Egger's tests are commonly used in the literature (Duval and Tweedie 2000;Sternee and Egger 2005). In this study, publication bias was tested by using these four methods.

Figure 1. Funnel Plot of Effect Sizes (ES) of Studies on Gender Variable
As observed in Figure 1, it is seen that majority of 72 studies included in the research are towards the top of the figure and close to the united ES. In meta-analysis studies, if there is not publication bias, it symmetrically expands to the both sides of the vertical line indicating the combined ES. If there is a publication bias in 72 studies included in the study, then majority of the studies shall unite through the bottom part of the funnel shape or only at one part of the vertical line (Borenstein et al., 2009). This funnel scatter plot ( Figure 1) is one of the indicators that there is not any publication bias for the studies included in the research. Other publication bias tests are given in Table 1. As a second test in testing publication bias, Orwin's Fail-Safe N calculation was also carried out. Orwin's Fail-Safe N gives the number of studies which might be missing in a meta-analysis synthesis (Borenstein et al. 2009: 285). As a result of this analysis, Orwin's Fail-Safe N was calculated as 138. The necessary number of studies for 0.01 average, which was found as a result of meta-analysis, to reach 0.01 level of ES, in other words to almost zero effect level is 138. However, 72 studies in this study are the whole number of studies which were carried out for this research question in Turkey and were deemed appropriate for the inclusion criteria. Since there is not any possibility to have access to 138 studies apart from these 72 studies, the acquired results are considered another indicator of that there is not publication bias in this meta-analysis. According to trim and fill method of Duval and Tweedie, when 7 equal studies were included, it was observed that average ES which was found to be -0.05 as a result of the meta-analysis changed to 0.07. Since this change is insignificant, it can be accepted that the reported ES is reliable. As another publication bias test, Egger's test (funnel plot asymmetry test) results were not significant (p=0.37); and it is considered another indicator that there is not publication bias in this meta-analysis (Table 1).

Uncombined Findings of ES Analysis in Accordance with Teachers' Gender
Effect size values of AM that male and female teachers experienced at schools are given in Table 2.  According to Table 2, it was observed in 72 studies that there is a low level of difference in favour of male teachers in the scope of ES. While statistically significant difference (p <0,05) was found in 25 studies, no significant difference was found in 47 studies.

Forest Plot of Studies Which Include Data about Gender
Forest plot of 72 studies which were included in the study and had data about gender is demonstrated in Figure 2.

Figure 2. Forest Plot of Effect Sizes of Studies about Gender Variable
When Figure 2 is analysed, it is observed that there is a difference in favour of male teachers; it can be interpreted as such that they experience more AM at school compared to female teachers.

Random Effects Model and Heterogeneity Test Results
ES values of AM that female and male teachers experienced at school are given in Table 3. ES values of studies included in the research in accordance with the gender variable in Table 3 were calculated; its average ES value in accordance with random effects model was calculated as d=-0,02, standard deviation of average ES was calculated as SE=0,04, upper limit of average ES's confidence interval was calculated as -0.10, and the lower limit was calculated as 0.06. In line with calculations, data in 72 studies which were included in meta-analysis revealed that according to random effects model, male teachers have more mobbing perception than female teachers. Since ES value was lower than 0.20 in this study, according to Cohen's (1998) classification, it was detected that it has a value which is even under the low level. Since ES is lower than 0.15 in Lipsey's classification, it was mentioned that ES is even under the low level. According to the classification carried out by Thalheimer and Cook (2002), -0.15 <d < 0.15 is insignificant, 0.15 < d < 0.40 is low, 0.40< d < 0.75 is medium, 0.75 < d < 1.10 is high, 1.10 < d < 1.45 is very high, and 1.45 < d means perfect level of ES. According to this classification, it was observed that there is an insignificant level of difference (-0.15 -0,15). When statistical significance was calculated in accordance with Z test, Z was found to be -0.50 (Z=-0,50).
Q was calculated as 744.02 (Q=744.02) for homogeneity test, in other words, for Q-statistics. From x 2 table, the level of 71 degrees of freedom was found to be 51.80 at 95% significance level. Since the Q-statistic value (Q=744.02) exceeds the critical value of the chi-square distribution (x 2 0.95 = 51.80) with the degree of freedom, the absence hypothesis of the distribution of effect sizes was rejected in the fixed effects model. It means that the distribution of effect sizes is heterogeneous according to the fixed effects model. Used in order to eliminate the lack of Q statistic from the sample, I 2 provides a clearer result regarding heterogeneity (Petticrew and Roberts, 2006). Since there is a high level of heterogeneity between studies as a result of homogeneity tests (Q and I 2 ) carried out for gender variable, moderator analyses were carried out to detect the possible reasons of this heterogeneity. Results of the moderator analysis carried out to detect the possible reasons of heterogeneity emerging in accordance with gender variable are given in Table 4. As a result of the conducted moderator analysis, moderator effects of publishing type (p=0,03), of education level (p=0,03), and of gender of the researcher (p=0,02) were determined. It was observed that while mobbing perception of male teachers are higher in studies carried out in master's and doctoral theses, mobbing perception of female teachers are higher in studies carried out in articles. Moderator effects of school type (public, private, and public/private) (p=0,63) and of the region where the research was carried out (p=0,17) could not be determined. It was observed that mobbing perception of female teachers who work in Central Anatolia and Eastern Anatolia are higher compared to other regions.

Figure 3. Meta-Regression Results of Effect Sizes by Years When Researches Carried out
As can be observed in Figure 3, it was observed that there was no significant difference in gender differences in terms of effect sizes of the studies.

DISCUSSION AND CONCLUSIONS
In this meta-analysis study, 72 ES from 72 studies which form a sample consisting of 24954 people were calculated. While statistically significant difference (p<0,05) was found in twenty five studies, no significant difference was found in 47 studies. As a result of the process of combining in random effects model, a statistically significant ES which was at the level of -0.02 was found in favour of male teachers. According to Cohen (1988) and Thalheimer and Cook (2002) classifications, it is an insignificant and low level result. When results are evaluated together, it is observed that there is a difference, which may not be considered significant, between male and female teachers in terms of teachers' perceptions about AM at school. This result shows that the male teachers perception mobbing relatively more than female ones. However, this ES difference may not have practical significance for social scientists. Yet, when vote-counting method was applied (Borenstein et al., 2009;Ellis, 2012) more studies indicate parallel findings to this research. Taken the average age of the male teachers, it can be said that they take everything they experience serious, feeling mobbing more than female teachers. Since there is no meta-analysis study in the literature to determine the effect of gender roles in teachers' perceptions about AM they experience at school, it has not been possible to compare these results.
In generally female employees are expected to accept the rule of institution in comparison with patriarchal domination in the context of gender stereotype. This acceptance is based on cultural, biological, economic, and religious factors (Tan, 1979). According to the research conducted by Gökce (2012) in educational organizations, while age, type of school and branch do not create a meaningful difference in mobbing, gender does. The findings of this meta-analysis study support this interpretation. It is also significant that the difference of opinion between male and female teachers is low.
Although some studies revealed different results in terms of gender, acts of mobbing are eperienced in every organization and culture (Ülbegi and Yalçın 2015). In some studies carried out in Turkey which were analysed within the scope of the research, for example, the results of the studies carried out by Aras (2012), Aydın (2009), Çelik (2011), Deniz and Ünsal (2010), Dilmaç (2009), and Ertürk (2015) revealed that there is not a significant difference between females and males in terms of experiencing acts of mobbing. Since there is a low level of difference between primary and secondary grade teachers' opinions about mobbing in the studies carried out by Nielsen and Einarsen (2012) and Russo, Miliç, Knežević, and Mulić and Mustajbegović (2008), they support the results of this study. In this meta-analysis study, low significant difference has been found in the comparison with the gender variable. Deniz and Unsal (2010) ;Hansen, Hogh, Persson, Karlson, Garde, and Orbaek (2006) and Rayner (1997) did not find any difference in the gender variable. In other words, Bıçakcı (2017) found that male teachers faced mobbing behavior more than female teachers. It is significant in the context of the results of this study that since male employees are more likely to aspire to career ladders at schools, especially managers apply more AM to male employees for various reasons (Apak, 2009;Ocak, 2008). Mobbing has negatively affected everyone in educational organizations especially at schools regardless of any gender difference (Nielsen and Einarsen 2012). Despite the general perception that gender roles in Turkey reflect on teachers' perception of mobbing in school organization, this meta-analysis study revealed that there is a low level of difference. The fact that there is a low level of difference between female and male teachers in terms of mobbing perception can be considered as an interesting result for the Turkish Education System. Female teachers' tendency to obey the rules can reduce their level of exposure to mobbing. The fact that the majority of school administrators are male can be considered as a factor in low teachers' perception of mobbing. Within the context of this study, it can be stated that mobbing perception at schools cannot be explained within the scope of gender variable or that this perception is at a low level. In this context, it could be suggested not to use teachers gender as an independent variable in studies on mobbing at school.
Within the scope of the results of this meta-analysis study, apart from the gender variable of teachers, meta-analysis studies can be carried out by using variables which are related to working conditions, school culture, and organizational behaviour elements which are predictors of AM.