Teachers’ social self-efficacy: development and validation of a new scale

Abstract Teachers’ social self-efficacy is an important construct that has not been examined adequately in education yet. Developing a comprehensive understanding of teachers’ social self-efficacy requires measuring these beliefs in the social domain of teachers’ functioning. This study developed the Teachers’ Social Self-Efficacy Scale (TSSES) scale and investigated its psychometric properties in a total sample of 650 preschool teachers from Greece. Results from factor analysis revealed a bi-factor model underlying a global TSSE factor and five specific factors. Evidence of validity was obtained with the TSSES prediction on the closeness and conflict dimensions of teacher–student relationship quality. Findings showed that the TSSES can be considered as a valid and reliable instrument. Implications and directions for further research are discussed.

ABOUT THE AUTHORS ) is teaching at the International Hellenic University of Greece as an adjunct lecturer and she also has a PhD degree in Early Childhood Education. Her research interests focused on the classroom management, professional development of early childhood teachers and the development of teacher-child relationship quality. She has participated in several international research projects funded by the EU. Athanasios Gregoriadis (ORCID: 0000-0002-3026-6614) is an Associate Professor in the Department of Early Childhood Education of Aristotle University of Thessaloniki. In the past, he was a visiting professor in the University of Bielefeld and a visiting research fellow in the University of Oxford. His main research interests include teacher-child relationships, quality of early education environments, professional development, and the support of home learning environments. In August of 2019 he was the chair of the 29th EECERA conference. He has participated in several international research projects funded by the EU and has published several peer reviewed articles and books. Nikolaos Tsigilis (ORCID: 0000-0002-2388-959X) is an Associate professor at the Aristotle University of Thessaloniki. He holds two PhD's, one in Measurement and Evaluation from the National Sports Academy Sofia, Bulgaria and one in Organizational Behavior in Education from the University of Thessaly, Greece. He has published more than 70 papers in national and international journals. His main research interests revolve around research methods in social sciences, applied statistics and psychometrics. Vasilis Grammatikopoulos (ORCID: 0000-0001-7556-6162), Ph.D. is professor in educational assessment in early childhood and Head of the Department of Early Childhood Education and Care of International Hellenic University. He has vast experience in designing, implementing and evaluating professional development programs for teachers and an extensive record of publications in the domain. He served as coordinator in a EC funded project regarding professional development of early childhood teachers (KA2 -2016-1-EL01-KA201-023420) and also participated in several similar other projects. He authored extensively in teachers' professional development including books, journal articles, book chapters and conference proceedings.

Introduction
In the demanding classroom life of the twenty-first century, teachers have a prominent role in students' school life. They are called to promote not only students' academic achievement, but also students' socio-emotional skills, while facing a wide range of social challenges (e.g., peer rejection, socio-emotional problems, disruptive behavior; OECD, 2020). An extensive body of literature has shown that enhancing teacher-student interactions-known as process quality-are a primary contributor to student's learning and development, especially in early childhood education and care (ECEC; La Paro et al., 2014;White, 2020). The importance of positive teacher-student relationships is well established both for teachers' wellbeing and students' development (Hamre & Pianta, 2007;Mashburn et al., 2006;Roorda et al., 2017). Teachers who experience positive relationships with their students, provide more emotional support and attention to them, which in turn encourage students to develop their social and regulatory skills (Hartz et al., 2017). On the contrary, students who experience conflictual relationships with their teachers are at risk for negative outcomes, such as low engagement in activities or academic achievement (Roorda et al., 2017).
In the previous years, research attention has focused on the value of measuring and promoting high-quality teacher-student interactions (Mashburn et al., 2006;La Paro et al., 2014), indicating that the field could benefit from a deeper understanding of organizational conditions, practices, and attitudes that affect teachers' ability to effectively engage in interactions with students (OECD, 2020;Summers et al., 2017). Teachers' ability to enhance supportive relationships with students seems to be a crucial aspect of their professional role (Rimm-Kaufman et al., 2003). Recent findings showed that teachers perceive the part of interpersonal interactions with their students as one of the most difficult aspects of teaching (Vatou et al., 2021).
There is considerable evidence suggesting that teachers' self-efficacy is a contributor to successful professional practices (Zee & Koomen, 2016). The construct of teachers' self-efficacy has its roots in the Social Cognitive Theory (Bandura, 1977(Bandura, , 1997. Bandura (1997, p. 3) defined selfefficacy as "an individual's beliefs about his/her own capabilities to organize and execute courses of action required for accomplishing a specific task", suggesting that self-efficacy can have a positive or negative influence to individuals' beliefs about their abilities to perform a specific task. Teachers' self-efficacy is linked with students', teachers' and school-related outcomes including the development of the positive classroom climate, students' engagement in learning activities, improvement of academic achievement, teachers' wellbeing, job satisfaction and effects on burnout (Y. Guo et al., 2014;Höltge et al., 2019;Klassen & Chiu, 2010;Zee & Koomen, 2020). However, the teachers' self-efficacy regarding interactions with students is considered to be one of the most salient indicators of quality in supporting positive mutual interactions (De Jong et al., 2014;Summers et al., 2017;Veldman et al., 2017).

Teachers' efficacy in teacher-student relationships quality in ECEC
Much of the research on teacher-student relationships is guided by the attachment theory. Within this theory, teachers act as a "secure base" or "secure haven" for students that in turn feel safe to engage and explore the classroom environment . However, it is not always ensured that teachers manage to respond to students' needs. Teachers face multiple daily demands and must be attuned and well-prepared to respond to various social situations with students during the preschool day (OECD, 2020). For example, teachers tend to develop and follow personal hidden assumptions to hide their negative emotions when they experience challenging situations in which students exhibit inappropriate behaviors (De Ruiter et al., 2021). However, the need for the teacher to manage and re-direct students' inappropriate behavior and replace it with a positive and "expected" behavior still remains (Reinke et al., 2016). Therefore, developing positive teacher-student relationships can be a challenging task in ECEC. It seems that engaging in different situations during teacher-student interactions requires a different set of teachers' skills, practices and abilities to successfully accomplish them (Lippard et al., 2018).
As noted above, teachers' self-efficacy brings significant benefits for both teachers and students (Y. Guo et al., 2014;Höltge et al., 2019;Zee & Koomen, 2016); however, only limited research has investigated the association between teachers' self-efficacy and teacher-student relationships (Summers et al., 2017;Tsigilis et al., 2019;Zee et al., 2017). Recent research attention on selfefficacy has taken a more interpersonal turn (Reyhing & Perren, 2021;Veldman et al., 2017;Zee et al., 2017). While some studies found weak relations between teachers' self-efficacy and teacher-student relationships (Mashburn et al., 2006;Zee & Koomen, 2020), others have shown no relation at all (De Jong et al., 2014;Tsigilis et al., 2019). Tsigilis et al. (2019), for instance, found that teachers' self-efficacy dimensions did not contribute to any of the dimensions of the teacherstudent relationship quality (closeness, conflict and dependency) in elementary and secondary classrooms.
An interpretation of these mixed findings could be that teachers' self-efficacy reflects mainly teachers' self-perceptions regarding their teaching abilities (i.e., instructional strategies; Tschannen-Moran & Johnson, 2011;Zee & Koomen, 2016) rather than pedagogical or relational abilities. Teachers' self-efficacy consists of three different domains: instructional support, classroom management and student engagement (Tschannen-Moran & Hoy, 2001) and as such, it is likely to relate more with the teaching domain than the social or interpersonal domain (Rich et al., 1996). This line of research conceptualizes teachers' social efficacy as a different domain of teachers' self-efficacy for teaching (Rich et al., 1996;Veldman et al., 2017). The construct of social self-efficacy refers to a teachers' ability to develop positive relationships with their students. Such an ability entails skills like, for example, sensitivity, empathy, recognizing emotions, responding on time to students' needs, redirecting student's behavior, etc. (Di Giunta et al., 2010). Given the importance of teachers' self-efficacy across domains, Rich et al. (1996) have pointed out the distinction between academic and social domain of teachers' self-efficacy. However, till today little is known about how teachers view their own effectiveness in social interactions with students.

Gender differences in teacher-student relationships quality
Recent research showed the existence of gender differences in the teachers' perceptions of their relationships with their students (Horn et al., 2021). Teachers tend to have less favourable perceptions of their relationships with boys versus girls, and the gap is increasing over time (Buyse et al., 2011;Horn et al., 2021). Previous research indicates that teachers perceive their relationships with boys as conflictual and less supportive than their relationships with girls (Ewing & Taylor, 2009;Horn et al., 2021). It is unclear, however, whether teachers also experience different levels of social self-efficacy in relation to students' gender. Given the importance of teacher efficacy for high-quality education, a closer investigation of this question is timely and pertinent.

Understanding the construct of social self-efficacy
There is a growing interest in the literature to define social self-efficacy, based on Bandura's (1977) theory (Fan et al., 2013;Luo et al., 2019). Social self-efficacy has been widely studied in the organizational psychology and management research and efforts have been made to describe various aspects of this construct (Alessandri et al., 2021;Fan et al., 2013;Smith & Betz, 2000). An initial definition of the construct of social self-efficacy described it as "one's ability to engage in social interactions to establish or maintain interpersonal relationships" (Smith & Betz, 2000, p. 286). This conceptualization is "domain specific" and describes one's expectancies about his/ her performance on interpersonal demands (Bandura, 1997). Bandura (1997) argued that self-efficacy beliefs vary across different domains of an individual's functioning and he suggested researchers identify the degree of specificity of tasks or domains, when they develop self-efficacy measures. Recently, Lazarides et al. (2020) suggested that self-efficacy is dependent on the context and teachers' self-efficacy is formed based on resources and experiences in a specific setting. This line of research has revealed that teachers' self-efficacy beliefs may be captured as task-, context-or domain-specific (Zee & Koomen, 2016). However, the majority of the existing literature on teachers' self-efficacy focuses on the teaching domain rather than on the social domain (e.g., Tschannen-Moran & Johnson, 2011).
Previous literature on social self-efficacy has found that individual's expectations regarding the interpersonal domain are an important determinant of success in social relationships (Bilgin & Akkapulu, 2007). Social self-efficacy has also been considered as a key determinant of one's behavior in terms of social aspect in various environments (Q. Guo et al., 2018). Broadly defined, social self-efficacy has been linked with individuals' job-related well-being (Fan et al., 2013), life satisfaction (Wright & Perrone, 2010), team building (Loeb et al., 2016), prosocial behavior (Alessandri et al., 2021), and task performance (Luo et al., 2019). Furthermore, social selfefficacy beliefs encourage people to develop social relationships, which in turn are vital for themselves to satisfy their personal needs (e.g., self-esteem, sense of belonging; Akin & Akin, 2016).
Reflecting the importance of social interactions, the social self-efficacy construct has gradually become more prevalent. Researchers also developed multidimensional and unidimensional scales targeting on adults' or students' beliefs about their abilities in the social domain (Di Giunta et al., 2010;Grieve et al., 2014;Smith & Betz, 2000). Furthermore, several studies also investigated the construct of social self-efficacy in the workplace environment (Alessandri et al., 2021;Fan et al., 2013). However, there is a significant lack of knowledge about the social self-efficacy of early childhood teachers.
A literature review indicated several social self-efficacy scales: (1) the Perceived Social Self-Efficacy (Di Giunta et al., 2010); (2) the Workplace Social Self-Efficacy (Fan et al., 2013); (3) Social Efficacy and Social Outcome Expectations (Wright et al., 2013); and (4) the Social Understanding Self-Efficacy and Social Skill Self-Efficacy (Grieve et al., 2014). However, the most widely used social self-efficacy scale for adults is the Perceived Social Self-Efficacy (PSSE; Smith & Betz, 2000). The PSSE scale includes 25-items that demonstrate a single factor structure describing six aspects of social interactions (e.g., developing groups, making friends, offering help, participating in a group) and its psychometric properties have been confirmed in several studies (Fan et al., 2013).
Although most of these scales displayed excellent psychometrics properties, these scales reflect a more general social-self-efficacy measure and authors do not consider these measures specific enough to cover the full spectrum and the challenges of the teaching profession. The wide range of interactional tasks and the expectations a teacher has to fulfil are not necessarily examined in the existing dimensions of the above scales. Consequently, because of the importance for teachers to understand their beliefs regarding their interactions with students, it is necessary to consider the educational context in which teachers' social self-efficacy is developed.

The construct of teachers' social self-efficacy
In search of a more profound understanding of teachers' social self-efficacy, this study draws from the existing theory used in the field of self-efficacy (e.g., Bandura, 1997Bandura, , 2006 and the available research findings focused on the adults' social self-efficacy (e.g., Di Giunta et al., 2010), teacherstudent relationships (e.g., Hamre & Pianta, 2007) and teachers' competences to develop positive interpersonal relationships (e.g., Jennings & Greenberg, 2009). Within Bandura's theory (Bandura, 1997(Bandura, , 2006, the development of efficacy scales is based on a clear conceptual analysis of the relevant domain of an individual's functioning (e.g., context of ECEC). Therefore, for a comprehensive understanding of the teachers' ability to enhance social interactions, the awareness of the specific tasks of interactional domain and interpersonal abilities may allow researchers to identify the dimensions of the teachers' social self-efficacy (Bandura, 2006). In particular, our conceptual framework emphasizes the themes of teachers' sensitivity, supportive teacher-student relationships, taking students' perspectives into consideration, and students' behavior management (e.g., Hamre & Pianta, 2007;La Paro et al., 2014).
This study approaches the teachers' social self-efficacy as a multifaceted construct. We focus on teachers' beliefs regarding their interpersonal abilities to enhance affective teacher-student interactions in the ECEC. We view teacher-student relationships as more than just interactions. It has been suggested that recognizing and sharing others' emotions, offering help, and managing conflict situations are the main aspects of adults' social self-efficacy (Fan et al., 2013;Di Giunta et al., 2010). Similarly, according to attachment theory, positive teacher-student relationships are guided by a number of assumptions: teachers' respect for the students, teachers' sensitivity towards their needs, promoting students' autonomy and collaboration with other students, encouraging for appropriate behaviors and setting clear behavior expectations (Hamre & Pianta, 2007;La Paro et al., 2014;Roorda et al., 2017). These interactions frequently occur and constantly interchange during the school year. Thus, we paraphrase from social cognitive theory on selfefficacy to say that teachers' self-perceptions of social efficacy are shaped by specific events and experiences linked to social domain of functioning (Bandura, 1997).
Taken together, it appears that teachers' social self-efficacy relies solely on teachers' selfperceptions about interpersonal relationships (e.g., Di Giunta et al., 2010) rather than teaching practices (e.g., Tschannen-Moran & Hoy, 2001;Tschannen-Moran & Johnson, 2011). To better understand the connection between teachers' social self-efficacy and teacher-student relationships, it is critical to consider the teachers' interpersonal abilities (for a review see, Jennings & Greenberg, 2009). The majority of studies have investigated the association between teachers' interpersonal abilities and teacher-student relationships (Gunter et al., 2012;Jennings & Frank, 2015;Jennings & Greenberg, 2009). Findings of these studies showed that teachers' interpersonal abilities-self-awareness, social awareness, responsible decision-making, self-management, and relationship management-contribute to the development of teacher-student relationships, the classroom management and generally, the communication with students (Gunter et al., 2012;Jennings & Frank, 2015;Jennings & Greenberg, 2009). However, these studies often take for granted that teachers already have these interpersonal abilities and are capable to implement them in their interactions with students. Although researchers have recently recognized the importance of teachers' social-emotional competences (Jennings & Frank, 2015), understanding how teachers self-evaluate their beliefs about their interpersonal abilities may also influence teacher-student relationships.
In summary, there is a need for well-designed and validated measures that assess the teachers' social self-efficacy. Building on existing models and conceptualisations, we define teachers' social self-efficacy as the teachers' ability to perform specific interactional tasks, and to develop and maintain positive interpersonal relationships with students. Through an analysis of the role of teachers' interpersonal abilities and interactional domain, this study developed the Teachers' Social Self-Efficacy Scale (TSSES) which includes five aspects of teacher-student interactions: Teacher Sensitivity, Social Guidance, Teacher-Child Support, Classroom Climate-Children Engagement and Classroom Management-Conflict Resolution.

The present study
The main goal of this study is to introduce and validate the Teachers' Social Self-Efficacy Scale (TSSES), an efficacy instrument to prompt teachers' reflection upon their performance on teacherstudent interactions. Initially, this study describes the conceptual analysis of the teachers' social self-efficacy to provide a complete picture of teachers' beliefs regarding their effectiveness on social interactional domain. Our study is grounded in the following research questions: 1) Which is the factorial structure of the TSSES? 2) Does the TSSES demonstrate acceptable evidence of internal and external validity? 3) Does the teachers' social self-efficacy predict the teacher-student relationship quality? In particular, we expected positive association with close teacher-student relationships and negative association with conflictual relationships. For the last research question, we also included students' gender at the student level.

Participants
The present study is a quantitative cross-section research and examines data from early childhood teachers from different regions of Greece. Initially, 162 pre-kindergarten teachers (females, M age = 42.18 years, SD age = 9.6, M experience = 17.32 years, SD experience = 9.7) piloted the survey to ensure the initial content and construct validity of the TSSES (Study 1). A total of 488 early childhood teachers (females, M age = 45.34 years, SD age = 7.55, M experience = 20.7 years, SD experience = 7.1) were recruited to participate in Study 2 investigating the construct validity. A convenience sample of preschool teachers (N = 112) and their students (N = 897) was utilized to obtain evidence for the predictive validity. With regard to students' gender, 48.8% were boys and their average age was 5.11 years (SD age = .61).

Scale development process
To develop a reliable and valid measurement of teachers' social self-efficacy, we followed Bandura's (2006) guidelines and the design process detailed by DeVellis (2017): Step 1-Literature review: An overview of literature of both teachers' self-efficacy and social selfefficacy helped us to understand the range of conceptualizations and operationalizations of social self-efficacy as described above. This review highlighted the central tasks in teachers' daily interactions with students (e.g., teachers' sensitivity, behavior management), and shaped our definition. Table 1 provides the operational definitions and an example item for each teachers' social self-efficacy subscale, namely, Teacher Sensitivity, Social Guidance, Teacher-Child Support, Classroom Climate-Children Engagement and Classroom Management-Conflict Resolution.
Step 2-Interviews: We conducted open-ended interviews with preschool teachers to map current challenges and problems regarding teacher-student everyday interactions. The most common obstacle that teachers faced was mainly students' challenging behavior. Then, we compared responses from teachers' interviews against our literature review to indicate points of overlap, divergence, and disparities in terminology.
Step 3-Developing item pool generation: A systematic review was conducted using specific databases (e.g., Scopus, ERIC, and PsycINFO) to generate an item pool. The criteria used were the following a) studies in peer-reviewed journals, b) the research design (only empirical studies), and c) studies based on Bandura's theory (Bandura, 1977). Having our definition as a starting point, we created items following two paths. The first set of items reflected the teachers' interpersonal abilities whereas the second set of items reflected activities of teacher-student interactions. All items were positively worded, succinct and comprehensive. The resulting initial pool consisted of 45 items. Each dimension of teachers' social self-efficacy consisted of nine items.
Step 4-Expert review: The initial set of 45 items was reviewed by ten experts to assess their content validity. Experts had the necessary content expertise and theoretical background on the construct of self-efficacy and teacher-student interactions to provide a comprehensive assessment of the measure. Moreover, they held a PhD and had job affiliation on departments of education and early childhood education around the world (i.e. Oxford University, Aristotle University of Thessaloniki). Experts evaluated content validity of the items based on three criteria: a) whether each item is relevant to the dimension, which was designed to measure, b) clarity and conciseness of items, and c) whether each item is reflected our conceptualization of teachers' social self-efficacy. Experts were confused with eight items regarding the dimension to which they belonged. Although five items were relevant to the construct, its wording was problematic. Furthermore, experts advised us to remove three items that displayed various biased responses. Therefore, the initial number of 45 items was reduced to 29.
Step 5-Pre-testing items to teachers' sample: A series of small-scale pilot tests of the teachers' social self-efficacy measure were conducted to see whether teachers comprehend each item. Based on teachers' recommendation, we made minor additional corrections, mainly in wording.
Step 6-Final Scale: The teachers' social self-efficacy scale included the remaining 29 items that encompassed five dimensions: Teacher Sensitivity (6 items), Social Guidance (6 items), Teacher-Child Relationship (4 items), Classroom Climate-Children Engagement (8 items) and Classroom Management-Conflict Resolution (5 items). Finally, teachers were encouraged to respond to each item by using a 9-point Likert scale in order to rate the strength of their beliefs on their current abilities to perform each described behavior (1-not at all to 9-a great deal).

Student-Teacher Relationship Scale (STRS)
The STRS was used to examine TSSES concurrent validity . STRS assesses teachers' perceptions about their relationships with students in terms of Closeness (11 items), Conflict (11 items), and Dependency (6 items). Teachers rated the items through a 5-point Likert scale (1-definitely does not apply to 5-definitely applies). The Greek version of STRS (Tsigilis et al., 2018) was used in this study and results showed that the Cronbach's α of the three subscales were .83, .90 and .66, respectively.

Procedure
The research protocol and the ethics of the current studies were approved by the Greek National Educational Policy Institute. Protocol number of the licence approval: Φ15/72,273/136,296/Δ1. The data for Study 1 were collected through the National Early Childhood Educators' Conference (March 2019), using an electronic survey platform. Before the study, the researchers informed the organizational committee of the conference about the aims of the study and the overall research design. An invitation to voluntary participation in the study was sent to the enrolled teachers via email. The email included a) information about the aims of the study, b) guidance on how the research team would keep responses in strict confidence, and c) the measurement of TSSES. Teachers self-selected to participate; thus, we obtained a convenience sample.
A second group of 488 preschool teachers participated in the Study 2, 112 of whom participated in the second phase of this study. This study took place during the school year 2019-2020 and its purpose was to investigate both construct and concurrent validity of TSSES. A formal letter was sent through the Greek Ministry of Education to kindergartens all over Greece to inform teachers about the purpose of the study. Following, teachers participated in an online study, which consisted of the TSSES questions and demographic information.
Next, researchers established communication with kindergarten directors by email or telephone to arrange face-to-face visits, so as to discuss the teachers' participation in this study and research needs (e.g., parental form, questionnaires). Kindergartens were located in three prefectures in Northern Greece (Thessaloniki, Halkidiki, Kilkis). Researchers visited only the kindergartens whose directors consented to participate in the entire study, that is a subset of 112 teachers. A consent form which included ethical guidelines that explained the rights of the participants, the right to discontinue their participation at any time, and the confidentiality of their responses was signed by teachers and parents. Eight students from each classroom were randomly selected after collecting their parental permission to participate in the survey. Teachers filled in the STRS questionnaires for each of these students. The administration procedure did not last more than 40 minutes.

Analytic method
An exploratory factor analysis (EFA) was first conducted to investigate the factorial structure of the TSSES (Study 1). The selection of the derived factors was based on various criteria such as eigenvalues over 1.0, scree plot, and total variance explained using SPSS ver.25. At the second step, and following EFA results, a confirmatory factor analysis (CFA) was implemented to validate the measurement model of the TSSES (Study 2). The model was fitted in Mplus 8, using the maximum likelihood method of estimation (MLM; Muthén & Muthén, 1998-2017. Evaluation of the model's fit was examined using χ 2 statistic, the Comparative Fit Index (CFI > .95), the Root-Mean-Square Error of Approximation (RMSEA < .06) and the Standardized Root Mean Square Residual (SRMR < .06; Kline, 2015). Omega coefficient was used to calculate TSSES factors internal consistency (Hayes & Coutts, 2020).
The investigation of the linkage between teachers' social self-efficacy and the quality of teacherstudent relationships was achieved by fitting a series of multilevel models. In these analyses, STRS was the outcome variable (level 1-student level), whereas TSSES presents an explanatory variable at level 2 (teacher level). Multilevel models were systematically built starting with a model without predictors at any level (Null model) in order to compute the intraclass correlation coefficient (ICC). ICC values above .05 generally denote the necessity to take into account the hierarchical nature of the data. The Null model also provided important information about the variation in the within and between clusters (Geiser, 2013). Based on recommendations from other studies (Summers et al., 2017;Tsigilis et al., 2019), the ICC was calculated separately for each of the dimensions of STRS. Second, a conditional main effect model with student-level (e.g., gender) and teachers-level (e.g., TSSES) covariates were used in an attempt to explain the observed variability. The same procedure was followed for each dimension of the STRS (closeness, conflict and dependency). Finally, the TSSES entered into the model centred to the grand mean and the students' gender was dummy coded (0 = boys, 1 = girls; Geiser, 2013). This dataset does not suffer from missing values. Table 2 displays the means, and standard deviations for TSSES items. The descriptive statistics indicate that all item means were relatively high. The highest item means were found for items comprising the Classroom Climate-Children Engagement domain of teachers' social self-efficacy.

Preliminary factor analysis
EFA using a principal axis factor analysis followed by oblimin rotation revealed the presence of three eigenvalues above unity (14.56, 1.39 and 1.06), explaining 54% of the common variance (KMO = .97, Bartlett's test of sphericity = 12,235.1, df = 406, p < .001). However, the examination of the scree plot clearly showed that one factor should be retained. Thus, based on the scree plot and the large difference between the first and the second eigenvalue, it seemed wiser to retain one factor and the analysis was rerun. New results showed that TSSES can be initially described by one factor, explaining 49% of the variance. All 29 items had statistically significant loadings, ranging from .59 to .81. It is encouraging that no item is excluded from the scale due to low factor loading (< .30), supporting the clarity and the relevance of the items with the concept of social selfefficacy. The internal consistency of the 29 items was α = .97.

Main factor analysis of TSSES
Based on the EFA results, two candidate models were submitted to CFA using the second group. The first model (M1) was based on the EFA findings, suggesting a unidimensional structure. The second model (M2), hypothesized a five-factor structure. CFA results for the single-factor model (Μ1) and the five-factor model (Μ2) indicated that the M2 had a better fit to the empirical data (Table 3). The two models were compared by means of a Δχ 2 test. Results revealed that the five-factor model was significantly improved over the single-factor model (Δχ 2 = 224.86, df = 10, p < .001). Also, all factor loadings were statistically significant, ranging from .592 to .856. Although the five-factor model yielded acceptable fit indexes, high correlations were observed among subscales (< .82). Given the relatively high association among subscales and the suggestions of Kline (2015), it seemed necessary to further examine the TSSES dimensionality by testing the tenability of two alternative structures, a second order-factor (M3) and a bi-factor model (M4).
Application of a second-order analysis showed that all five of the first-order factors had statistically significant factor loadings with the second order. However, in terms of goodness-of-fit indices, M3 yielded mixed results. In particular, whereas RMSEA and SRMR showed an acceptable fit, CFI was rather low in relation to the guidelines provided by Hu and Bentler (1999). On the contrary, fit indices associated with the fourth model (M4) clearly showed an acceptable fit. Chi-square comparison between the five-factor model and the bi-factor model revealed a statistically significant difference (Δχ 2 = 107.56, df = 18, p < .001), in favor of the latter model. Similar results revealed for the chi-square comparison between M3 and M4 (Δχ 2 = 139.88, df = 24, p < .001). Thus, the bi-factor model was selected as the most tenable for describing teachers' response to TSSES. Notably, all factor loadings for general factor were statistically significant, ranging from .596 to .816, whereas the five specific latent factors demonstrated low to moderate factor loadings (Table 4).
Latent factors' internal consistency satisfactorily ranged from .86 to .97. Furthermore, omega hierarchical coefficient (ω Η ) was used to examine the reliability of the five specific latent factors after controlling them for the general factor. The general factor of the TSSES demonstrated a very high omegaΗ, whereas the five specific factors yielded low values (Table 4). Rodriguez et al. (2016) noted that high value of omegaH (> .80) means that the general factor "is the dominant source of systematic variance placing less influence on multidimensionality caused by the group factors" (p. 224). These results clearly indicate that the general factor of the TSSES was stronger in relation to the five specific factors. Finally, the vast majority of the TSSES common variance is explained by the general factor. The explained common variance (ECV) is another criterion to examine the strength of resulting general factor and specific factors (Rodriguez et al., 2016). In the present study ECV value is .835, which is an additional evidence supporting the dominance of the general factor of the TSSES over the specific ones.

TSSES concurrent validity
The ICCs for closeness, conflict and dependency were .354, .138 and .432, respectively. These values showed that a considerable amount of variance (greater than 5%) is attributed to differences among teachers, suggesting that a multilevel approach is meaningful (Table 5).
Overall, three multilevel analyses were conducted, one for each dimension of the STRS (Table 6). Results showed that teachers' social self-efficacy (β = .32, p < .001) and students' gender (β = .17, p < .001) significantly predicted the closeness dimension. Moreover, this model fits the data better than the null model based on the log-likelihood ratio test (Δχ 2 = 67.76, df = 2, p = .001). It should be noted that when gender was entered in the model the variance of the second level was .1113. This variance was reduced to .0736 when TSSES was introduced to the model. Thus, the variance reduction due to the TSSES was 34.2% [(.111 − .073)/.111 = .342] above and beyond the gender contribution.
A similar pattern merged from the results regarding the conflict dimension. In particular, teachers' social self-efficacy (β = −.06, p < .05) and students' gender (β = −.09, p < .001) were significant negative contributors to the conflict dimension. Chi-square difference based on loglikelihood showed that the main effect model yielded a better fit in comparison to the null model (Δχ 2 = 19.26, df = 2, p = .001). Additionally, the main effect model explains 10.5% of the variation of teachers' variability in conflict scores [(.019 − .017)/.019 = .105] above and beyond the gender contribution. It should be noted that teachers' social self-efficacy and students' gender did not significantly predict the dependency scores.

Discussion
Developing positive teacher-student interactions is one of the most important aspects of quality in education (OECD, 2020). As a consequence, having a clear understanding of how teachers perceive their abilities regarding the social interactional domain, and how these abilities are related with interpersonal relationships is an issue of great interest for researchers (Summers et al., 2017). Although the importance of teacher-student relationships and teachers' self-efficacy has been acknowledged (Klassen & Chiu, 2010;Roorda et al., 2017), till today there has been limited research on teachers' self-evaluation of their social effectiveness, like for example, their effectiveness in developing positive teacher-student relationships. Focusing on teachers, the purpose of this study was to introduce and validate the Teachers' Social Self-Efficacy Scale (TSSES); a construct not thoroughly examined insofar in the educational context.
To examine the teachers' social self-efficacy, we focused on analysis of the relevant domain of teachers' functioning (e.g., social domain in ECEC). Following a six-step survey design process, we developed the TSSES and attempted to obtain evidence of the scale's content validity. Specifically, we offered a foundation of evidence, suggesting the TSSES's content validity. Then, we further obtained additional evidence validity, including the TSSES factor structure (construct validity), its' reliability and the measurement of prediction of teacher-student relationships quality (concurrent validity). Preliminary findings suggest that teachers' social self-efficacy beliefs regarding their abilities in relation to teacher-student interactions may be considered as a specific efficacy domain.
In line with our hypothesis and findings emerging from previous research in social self-efficacy (Fan et al., 2013;Sherer et al., 1982), results from study 1 supported the appropriateness of using alternatives approaches to examine further the factor structure of TSSES. The second studied model was the one originally proposed by the research team-that is, the five-factor model. High correlations were yielded between the five factors, as has also been found in previous research on teachers' self-efficacy (e.g., Tschannen-Moran & Hoy, 2001). Like Tschannen-Moran and Hoy (2001), we then considered a second-order factor structure. That is, the correlations among the five dimensions (first-order factor) are accounted for by a higher, second-order factor (Brown, 2015). However, Mansolf and Reise (2017) also suggested that substantial interdependence among subscales might imply the existence of a bifactor model. In a bifactor model each item in the scale is influenced simultaneously by both a general factor and a specific one. In our case the bi-factor model demonstrated better fit than the alternative models. The results also showed that the variance of "teacher sensitivity", "social guidance", "teacher-child support", "classroom climatechildren engagement" and "classroom management-conflict resolution" was mainly explained by the general social self-efficacy factor. In this case, common variance of items was directly determined by the general factor while the five specific factors did not exist over and above the general factor. This finding supports that the five specific factors cannot be considered as unique contributors. Thus, it can be concluded that the plausibility of these five factors was decreased as the general factor demonstrated a large amount of variance (Reise et al., 2007). In terms of omega and omegaH values, results also showed that the general factor was held accountable for a considerable amount of variance. As such, the strong associations in our study among dimensions masked TSSE's multidimensional structure, which EFA procedures could not reveal.
The findings of the factor structure of TSSES suggest that the teachers' social self-efficacy construct can be considered as multidimensional. Previous work has shown that social self-efficacy can be conceptualized either as a multidimensional or a unidimensional construct (Fan et al., 2013;Sherer et al., 1982). It seems that teachers' social self-efficacy consists of both general and specific parts. Researchers can focus either on the general or the specific factors or on both parts. The TSSES general factor represents the overall perceived ability related to social self-efficacy, and the specific factors refer to the conceptually narrow subdomains of the social self-efficacy construct. The bi-factorial model allows us to better represent the TSSES factor by establishing that a uni-dimension and multiple specific dimensions are involved in TSSES scale answers (Reise, 2012).  At a classroom level, a multilevel modelling approach was used to examine the contribution of teachers' social self-efficacy on the quality of teacher-student relationships and its association with students' gender. This approach was suggested as necessary due to high ICC values (Tsigilis et al., 2019). The current results provide further support for the validity of the instrument, since TSSES seems to predict both the closeness and conflict dimension of STRS. This shows the importance of teachers' social selfefficacy for the development of positive interactions in the classroom and it offers new information regarding the relation between teachers' expectations and teacher-student interactions. Moreover, teachers seem to experience different levels of social self-efficacy regarding students' gender. Teachers tend to evaluate themselves to be more efficacious regarding their relationships with girls than boys. This finding is consistent with the literature which indicated that teachers experience more supportive relationships with girls rather than boys (Ewing & Taylor, 2009;Zee & Koomen, 2020).
In sum, evaluations of teachers' social self-efficacy not only can contribute to teacher-student relationships quality, but it can also be a valuable means for its improvement. Overall, improving the quality and efficiency of education is in line with one of four EU objectives of the ET2020 strategy (OECD, 2020).

Limitations and future research
A number of limitations should be considered when interpreting the results of this study. Firstly, this study includes only teachers reports and there is a risk of being biased because of common method variance. Although this method is frequently adopted in the self-efficacy research, it does not embrace the students' perspectives of their relationships with teachers. Future studies should also include students' reports to examine teacher-student relationships. Secondly, although our sample appeared to be large, it primarily consisted of female teachers. Future research should explore factors underlying the cluster bias by including teachers' or classroom's features as potentially biasing attribute. Third, this study has a cross-sectional nature and does not explore how teacher-student relationships change regarding teachers' social self-efficacy over time. Future research should apply experimental designs to determine cause-and-effect relations between teachers' and students' outcomes. Also, this study was conducted within the Greek ECEC setting, so the results cannot easily be generalized in other cultural contexts or in upper elementary grades. It would be interesting, if researchers could apply the TSSES in different ECEC settings to examine cultural issues or in upper elementary grades. The replicating analysis in other cultures and grades might be a beneficial future direction of the research.

Conclusions and implications
This study used the prior knowledge about adults' social self-efficacy research (Fan et al., 2013;Grieve et al., 2014), and followed specific steps-including extensive literature search, content analysis by experts and a pilot study-as a starting point for the TSSES development. A series of factor analyses were used to test the underlying structure of the TSSES. Moreover, MLM techniques were conducted to demonstrate the way in which teachers' social self-efficacy can predict the quality of teacher-student relationships. The findings were encouraging and seemed to support the reliability and validity of the TSSES.
Results of this study hold a number of important implications in the area of education. Since teachers' social self-efficacy is related to teacher-student relationships, improving the teachers' social self-efficacy may be a critical component of the quality of education. If teachers were able to assess their social selfefficacy regarding interactions with students, perhaps they could better reflect on their practice and then respond to positive relations with students. In this vein, having a measure for social self-efficacy is not only valuable for theoretical investigations, but it may also prove useful for school principals looking to evaluate their own school effectiveness on this domain. Moreover, academics and policymakers could provide teachers with the tools during the initial teacher education or training and professional development on how to develop, maintain and improve the quality of their relationships with students.