Abstract
Keywords
Introduction
School reforms often center their attention on the improvement of teaching quality because of the consensus among policymakers, researchers, and practitioners that high-quality teaching is the most critical school-related element for the learning of all students (Bryk et al., 2010; Darling-Hammond, 2000; Fauth et al., 2019; Leithwood et al., 2010; Thoonen et al., 2011). School principals, as leaders who shoulder responsibility for securing and sustaining the effective implementation of reforms within the school, face increasing pressure to create a school environment where teachers can enhance their professional knowledge and classroom practices (Hallinger, 2011; Leithwood et al., 2004). The recent literature provides supporting evidence that principal leadership can play an important role in encouraging teachers to modify their teaching practices to better fit the diverse needs of students (Bellibaş et al., 2022; Parise & Spillane, 2010).
Previous research has empirically established the link between school leadership and the quality of teaching (Bellibaş, Gümüş, et al., 2021; Bellibaş, Kılınç, et al., 2021; Sebastian & Allensworth, 2012; Supovitz et al., 2010); however, there remain two under-explored domains in the literature. First, although ardent supporters have long advocated the benefits of distributed leadership practices in building school capacity to achieve augmented student learning outcomes (Harris, 2005; Harris & Spillane, 2008; Spillane et al., 2007) and prior research provides promising evidence that verifies these assertions (Heck & Hallinger, 2010; Leithwood et al., 2020), there is still room for further research exploring whether and to what extent the distribution of leadership to a wider school community or shared decision making among teachers, leaders, students, and parents might influence the interaction among teachers and improve classroom practices (Amels et al., 2020; Heck & Hallinger, 2009, 2014). In addition, based on the previous research, we think teacher collaboration—a crucial competent of school improvement endeavors—could play a pivotal role in linking such efforts to classroom teaching (Geijsel et al., 2009; Hargreaves, 2019; Supovitz et al., 2010; Thoonen et al., 2011).
A second and more important caveat in the literature is that most existing studies use a single category of “instructional practices,” reflecting an assumption that a teacher would perform all sub-dimensions of instructional practice (e.g., classroom management, direct instruction, etc.) at a similar rate. However, schools employ teachers with varying teaching profiles, some of whom might try to deliver effective student-centered lessons by adopting a constructive approach, while others might spend more time on classroom management strategies (focusing on routines, rules, and student behaviors) to foster an orderly atmosphere (Blömeke et al., 2020). Therefore, we believe there is value in examining the extent to which there is a typology of teachers according to their instructional practices, as well as exploring how school leadership practices might influence such teacher profiles. We use a person-centered approach that unmasks the diversity of ways that teachers deliver instruction. Such an approach provides a more precise picture of the current situation of instruction in schools and holds the potential to shape policy-level interventions to address the varying instructional practice needs of the different typologies of teachers (Bowers et al., 2017).
In this study, we conducted Multilevel Latent Profile Analysis (MLPA) to examine the nature of teachers’ latent profiles concerning their instructional practices at both the individual and school level. We also employed mediation modeling to assess the extent to which the execution of distributed decision-making leadership practices might predict these profiles by encouraging collaboration among teachers. More specifically, by using a nationally representative secondary data set collected by the OECD (2019a), we investigated how distributed leadership predicts teacher and school profiles according to teachers’ instructional practices with the mediating role of teacher collaboration, while employing several key control variables at both the teacher level (gender, highest level of formal education completed, job experience, teaching career length, and enrollment in target class) and school level (location, access to resources, private or public, student enrollment, and diversity of student background). Consequently, our study aimed to address the following research questions: RQ1. Are teachers divided into different profiles in terms of their instructional practices? RQ2. What are the school- and teacher-level factors that predict the profiles? RQ3. Does distributed leadership influence the likelihood of teachers’ membership in different profiles of instructional practices indirectly through promoting professional collaboration in lessons among teachers?
Context and Background
Türkiye has a highly centralized and hierarchical educational bureaucracy. Sitting at the top of the hierarchy, the Ministry of National Education (MoNE) operates nearly all system-wide endeavors such as assigning teachers and school principals, designing curriculum, and budgeting. This naturally creates an educational context where schools have a limited amount of autonomy in decision-making on a range of issues such as teacher employment, professional development, and curriculum design, while teachers are viewed as mere implementers of a predetermined curriculum under the strict supervision of school principals (Şahin et al., 2017). Aligned with the main characteristics of the system, bureaucratic and administrative elements such as infrastructure, schedules, programs, and student discipline endeavors largely define the roles of school principals (Kılınç et al., 2021). School principals must then devote much of their time to managing administrative and bureaucratic tasks rather than establishing a strong vision and mission, focusing on teacher professional development, and building a healthy school environment conducive to high-quality teaching and learning (Kalman & Arslan, 2016; Koşar et al., 2013).
Since the establishment of the modern Turkish education system in 1923, the philosophy of educational essentialism, which regards the teacher as the sole authority and source of information, has dominated classroom teaching patterns across the country (Altınkurt et al., 2012). Since the turn of the millennium, however, the MoNE has undertaken several policy reforms focused on improving the quality of teachers’ instructional practices to produce larger gains in student learning outcomes. The first major reform initiative came in 2005 when the MoNE embraced the constructivist learning approach, which is seen as a panacea for more traditional teacher-centered instruction that pushes students to the position of passive receivers of information in the classroom (MoNE, 2005). In the ensuing years, with the recognition of school leadership's substantial contributions to the improvement of teaching and learning (e.g., Hallinger, 2011; Thoonen et al., 2011), Turkish policymakers paid growing attention to designing initiatives to secure and sustain the leadership capacity of schools. For instance, in 2014, the MoNE delegated national inspectors’ responsibilities for evaluating teacher performance to school administrators. This change exemplifies policymakers’ intentions to transform schools into professional learning communities where teachers and principals share the responsibility of providing high-quality education (Bowers, 2020; Thoonen et al., 2011). Moreover, the School-based Professional Development (MoNE, 2010) and Teacher Strategy Documents (MoNE, 2017) focus explicitly on establishing a professional culture that encourages teachers to collaborate on instructional issues and assume greater authority and responsibility for school decision-making processes. These policy papers are important for the Turkish educational setting because, for the first time, they announce a significant shift in the expected roles of teachers from solely conducting classroom teaching to operating as active agents in school improvement processes.
Overall, policymakers from around the world are looking for ways to improve the quality of schooling to address the diverse learning needs of students. As a result, scholars and practitioners are now more convinced that the contributions of school leadership to student learning outcomes expand when leadership is spread across multiple actors in the school (Harris & Spillane, 2008; Leithwood et al., 2020). Although the Turkish educational context continues to preserve its hierarchical and bureaucratic fashion to some extent, given the shifting research focus and climate in this area, we presume that the recent policy movement supporting shared decision-making could offer leverage to effect positive changes in the quality and methods of teachers’ instructional practices, through building a foundation for collaborative actions among school staff.
Theoretical Background
Teacher Instructional Practices
Scholars have long endeavored to delineate the nature and content of instructional practices that best predict student learning outcomes (Baumert et al., 2010; Darling-Hammond, 2000; Decristan et al., 2015; Hospel & Galand, 2016; Kane & Cantrell, 2010; Marzano et al., 2001; Rjosk et al., 2014; Scherer & Gustafsson, 2015). An important research focus in this area involves identifying the critical components of effective classroom teaching. Results from this research have revealed several mechanisms that might frame effective instructional practices, including classroom management, cognitive activation, tempo, supportive social environment (Klusmann et al., 2008), cognitive activation, classroom management (Lipowsky et al., 2009), student-oriented climate, and cognitively challenging learning opportunities (Rjosk et al., 2014).
This literature illustrates an emerging consensus that effective teaching requires implementing classroom management procedures that help teachers build a generative learning environment, linking cognitive activation strategies with teacher efforts to learn about students’ prior knowledge to provide them with challenging learning tasks, and ensuring the clarity of instruction by articulating the goals of classroom practices and identifying student expectations (Decristan et al., 2015; Scherer & Gustafsson, 2015). Consistent with this knowledge base, the current study borrowed the conceptualization of teacher instruction from the Teaching and Learning International Survey (TALIS), administered by the OECD, which assesses the construct under three dimensions: clarity of instruction, cognitive activation, and classroom management (OECD, 2018). Clarity of instruction refers to identifying clear teaching and learning goals, connecting new and old topics, and providing students with a summary of the lesson when it concludes (Hiebert & Grouws, 2007; Seidel et al., 2005). Cognitive activation is the process of evaluating, integrating, and applying students’ knowledge in the context of problem-solving (Lipowsky et al., 2009). Finally, classroom management encompasses teacher actions to ensure a consistent learning environment and effective use of classroom time (Klusmann et al., 2008).
The Profiling Approach to Identify Instructional Practice Configurations
This study was based on a recent genre of research that investigates the typology of respondents across a given survey instrument (Bowers & White, 2014; Duff & Bowers, 2022; Henry & Muthén, 2010; Urick, 2012). In this study, we performed MLPA to investigate the extent to which a group of teacher respondents with homogeneous instructional practices existed in a heterogeneous dataset (Henry & Muthén, 2010). Over the last decade, the field of educational administration has witnessed a growing corpus of scholarship that attempts to identify the different subgroups of schools, principals, or teachers within a given data set, typically by using Latent Class Analysis (LCA). For instance, Urick and Bowers (2014) identified different types of principals across U.S. schools based on their perceptions of their leadership styles. Two years later, Urick (2016) classified the different leadership types of teachers and principals, as well as the distribution of teacher types across principal types, to investigate the extent to which these types predicted teacher retention. Subsequently, Bowers et al. (2017) used the LCA approach on the Comprehensive Assessment of Leadership for Learning (CALL) to classify “leadership for learning” practices at high, moderate, and low levels. Furthermore, Agasisti et al. (2019) employed LCA to identify the subgroups of school leaders and assessed whether and to what extent individual characteristics and school context factors were associated with the probability of demonstrating a certain leadership style. A recent study by Duff and Bowers (2022) conducted a 3-step LCA to identify a typology of school organizational capacities using teachers’ perceptions of school organizational context.
A closer look at the literature also reveals recent work using the TALIS dataset to conduct MLPA. For instance, Bowers (2020) identified teacher and principal typologies based on the leadership for learning theory. To do so, Bowers borrowed a line of variables at both the school (e.g., focusing on instruction, monitoring teaching and learning, building nested learning communities, acquiring and allocating resources, and creating a safe and effective learning environment) and teacher level (e.g., student assessment, teacher feedback, teacher self-efficacy, professional development and trust, monitoring and induction, stakeholder engagement, and shared discipline climate). In addition, based on the leadership for learning theory and using LCA, Veletić and Olsen (2021) utilized more than a dozen variables from TALIS-2018 data (e.g., participation among stakeholders, professional collaboration in lessons among teachers, diversity practices, team innovativeness, etc.) to identify leadership typologies at the country level.
Despite growing interest in identifying the typology of schools, school principals, and teachers (e.g., Bowers et al., 2017), little research has attempted to explore teachers’ instructional practice profiles. The majority of the research on teaching has followed a traditional modeling approach contending that all the respondent teachers fit a single type (Bellibaş, Gümüş, et al., 2021; Doğan & Yurtseven, 2018; Özdemir, 2019). These studies often create a composite index by averaging the scores of each instructional practice dimension or item (Warwas & Helm, 2018). However, this approach fails to accurately depict real-life instructional practices (Bowers & White, 2014; Duff & Bowers, 2022). A profiling approach is much more likely to accurately identify and faithfully represent the types of instructional practices performed by teachers. Such an approach “assumes that between a fully-developed or ideal-type [instructional practices], represented by maximum scores on all its defining features, and the total absence of these features, several discrete scoring patterns coexist” (Warwas & Helm, 2018, p. 45), thus better characterizing the variations in teachers’ classroom practices. This is based on the notion that while there are teachers who might score highly on the three sub-groups of instructional practices (cognitive activation, classroom management, and clarity of instruction) and others who perform worse across all sub-dimensions, still more teachers may demonstrate mixed performance (e.g., highly effective in classroom management but less effective in cognitive activation). In this research, we employed MLPA to identify the key configuration of instructional practices and whether and to what extent the distribution of leadership practices and teacher collaboration predicted teachers’ membership in each profile.
Teacher Collaboration
Teacher collaboration refers to teachers’ collective efforts to contribute to the effectiveness of their organization (Bolman & Deal, 1997). Traditionally, teaching has been considered an individual endeavor requiring teacher autonomy, at least to some extent, to make critical decisions around curricular and instructional matters (Tschannen-Moran, 2001). Nevertheless, teachers also need to open the doors of their classrooms to establish professional bonds with their colleagues, and even with a larger professional community, to sustain professional learning and produce higher-quality classroom practices (Goddard et al., 2015). In this study, we define collaboration as teachers’ teaching jointly as a team, providing feedback to colleagues about their practice, engaging in cooperative activities across different classes and age groups, and participating in collaborative professional learning (OECD, 2019a).
Teacher collaboration is a well-established construct in school improvement research (Geijsel et al., 2009). Over the past two decades, a growing number of studies have been conducted to identify whether and to what extent collaboration influences school capacity for teaching and learning, as well as the factors that shape its nature and process (Duyar et al., 2013; Hargreaves, 2019; Tschannen-Moran, 2001). This research has concluded that teacher collaboration is one of the most crucial elements influencing the improvement of teachers’ learning and teaching (Bryk et al., 1999; Goddard & Kim, 2018; Stoll et al., 2006; Thoonen et al., 2011), thereby augmenting student learning outcomes (Goddard et al., 2007). More specifically, the evidence suggests that when teachers work in a school environment where they are provided with opportunities for sustained collaboration, they are more likely to make improvements in their instructional practices (Chen et al., 2020; Geijsel et al., 2009).
Distributed Leadership
Educational leadership scholars have devoted substantial effort to investigate the effects of school leadership practices on student learning outcomes. Largely dominated by instructional and transformational leadership emphases, this research has revealed positive and significant contributions of principals’ leadership practices to school improvement processes (Gumus et al., 2018; Robinson et al., 2008; Thoonen et al., 2011). However, over the past two decades, the distributed leadership perspective has received increasing interest from educational policymakers around the world, who have recently initiated several reform initiatives to promote schools’ academic capacity to achieve better student learning outcomes (Heck & Hallinger, 2009; Liu, 2021; Printy & Liu, 2021).
The growing interest of policymakers in the idea of leadership distribution as a lever to improve instruction has led to a relatively novel genre of research on understanding the potential contributions of distributed leadership to teaching and learning in schools (Heck & Hallinger, 2010; Liu & Werblow, 2019; Mascall et al., 2008). Though the empirical evidence is too limited to draw strong conclusions about the potential contributions of this line of inquiry to the improvement of teaching practices and thereby student learning, it is still promising. For instance, a decade ago, Heck and Hallinger (2009) provided empirical evidence of the positive effects of distributed leadership on school academic capacity, which in turn influences student growth rates in math. In addition, Malloy and Leithwood (2017) found a small and indirect effect of distributed leadership on student achievement through promoting academic press in schools. More recent research has also generated encouraging evidence that distributed leadership fosters teacher collaboration and job satisfaction, which in turn elevates instructional quality (Bellibaş et al., 2022).
Unlike traditional leadership approaches that equate leadership with the individual roles, responsibilities, or behaviors of school principals (Harris, 2005), distributed leadership advocates that leadership does not reside in one's individual capability—rather, it emerges from the collective action of various participants within a school, including principals, teachers, and even students (Gronn, 2002; Heck & Hallinger, 2009). The recent literature tends to conceptualize the construct according to two distinct lenses: analytical and normative. The analytical perspective focuses on understanding how leadership is extended to groups and networks to help modify teaching and learning in a school (Spillane et al., 2004). This lens emphasizes the synergistic relationships among school members (Spillane et al., 2001) and more importantly points to “the intersectionality of leaders, functions, and contexts” (Liu, 2021, p. 475). Framing their conceptual model around this perspective, several educational leadership scholars have assessed the extent to which school principals execute distributed leadership and investigated how this philosophy relates to critical school improvement phenomena such as academic press (Malloy & Leithwood, 2017) and academic optimism (Mascall et al., 2008). These researchers focused on patterns of leadership distribution and “ask[ed] respondents to identify the extent to which leadership distribution is planned or spontaneous, aligned or anarchic” (Mascall et al., 2008, p. 221). On the other hand, the normative perspective, from which the present study takes its direction, considers the goal of distributed leadership practices as encouraging teachers to participate in school decision-making processes to apply their knowledge and skills to curricular and instructional issues (Harris, 2005; OECD, 2019b). A growing body of research in this area has aligned distributed leadership practices with principals’ efforts to involve various stakeholders in school-wide decision-making processes (e.g., Bellibaş et al., 2022; Gumus et al., 2018). No matter which perspective researchers adopt when framing this construct, the literature still provides little guidance on how distributed leadership could function as leverage for improving teachers’ instructional practices. More specifically, previous studies fail to elaborate upon the potential influence of distributed leadership practices on teacher profiles according to their instructional practices.
Linking Distributed Leadership with Collaboration and Teaching Practice
Although the present study mainly rests upon the normative perspective to operationalize the construct of distributed leadership, we argue that a wider range of lenses should be borrowed from the existing literature to theorize the link between distributed leadership, collaboration, and teacher instructional practices (e.g., Louis et al., 2009; Mayrowetz, 2008; Mayrowetz & Weinstein, 1999). One such lens is the “distributed leadership as a work redesign” developed by Mayrowetz and colleagues (Mayrowetz et al., 2007; Mayrowetz & Smylie, 2004). This perspective acknowledges that distributed leadership encourages principals and teachers to perceive their work differently, largely beyond their traditional job descriptions. Based on this view, in this study, we approached distributed leadership as a pattern of interaction and influence among school members where principals are mainly responsible for creating the proper conditions for collaborative action around decision-making processes, and teachers function as active agents who engage in leadership for purposes of school-level improvement, beyond their formal work of classroom instruction (Harris, 2008). This perspective supports our proposition that implementing distributed leadership practices might help establish a school environment where teachers can collaborate and interact around instructional endeavors.
Our study is also grounded in the perspective of “distributed leadership as a human capacity-building process” (Harris, 2004; Mayrowetz, 2008). As stated by Gronn (2002), distributed leadership helps enlarge the leadership pool of a school by welcoming talented teachers as potential sources of leadership and expertise. Centering distributed leadership as the core of the school capacity-building process, Gronn furthered that implementing such practices might increase human capacity and potential, which are key factors for improved school performance and success. The literature has also indicated that such leadership is more likely to set the stage for meaningful change in the collective capacity of the school to improve teaching and learning (Goldstein, 2004; Harris et al., 2007). Overall, these theoretical debates and prevailing perspectives around distributed leadership provide us with a strong justification for linking distributed leadership to collaboration and thus teacher instructional practices.
Based on the arguments above, several influential leadership scholars have conducted empirical research on the potential role that distributed leadership might play in bolstering the quality of education in schools by activating untapped sources of influence and expertise (Bellibaş, Gümüş, et al., 2021; Harris et al., 2007; Leithwood et al., 2020; Malloy & Leithwood, 2017). These studies suggest that schools, which have faced unprecedented accountability pressures and demands for higher performance in recent years, can employ a more distributed form of leadership to foster a culture of collaboration and thus support the improvement of teaching and learning through higher quality decisions and changes around curricular and instructional issues.
The Present Study
Leithwood et al. (2020) proposed that “school leadership has a greater influence on schools and students when it is widely distributed” (p. 12). Although the existing research has yet to fully explore how and to what extent distributed leadership influences teaching, emerging evidence has shown a promising path between these two variables by providing evidence of the positive contributions of such leadership on several teacher-related factors, including teacher collaboration. While several studies have already provided some evidence of the link between distributed leadership and teacher instructional practices (e.g., Bellibaş et al., 2022; Spillane & Healey, 2010), the existing research lacks the potential to demonstrate how the distribution of leadership roles might play a role in determining the variation in teachers’ instructional practice profiles. Based on what we already know about the impact of distributed leadership, in the present study, we seek to test the proposition that distributed leadership fosters the establishment of a school climate that promotes teachers’ collective efforts to work together around curricular and instructional issues, which in turn determines teachers’ placement in a specific teaching profile (DeMatthews, 2014; Harris & Muijs, 2005). Therefore, using the TALIS 2018 dataset collected by the OECD, which includes 3,223 Turkish teachers within 196 schools, the purpose of the present research is threefold: (1) to identify the different configurations of teachers’ instructional practices at both the teacher and school level; (2) to test the school- and teacher-level factors that predict the profiles; and (3) to examine whether and to what extent distributed leadership predicts teachers’ membership in each profile indirectly through teacher professional collaboration. To reach the first goal, we performed a single-level and multilevel latent profile analysis respectively and assessed the fit indices of several profile combinations to identify patterns in teachers’ instructional practices at the teacher and school levels. Addressing the second purpose, we proposed a direct association between teacher collaboration and teacher profiles and employed MLPA with teacher- and school-level covariates. Finally, for the third purpose, we conducted a multilevel latent class mediation analysis to investigate whether and to what extent distributed leadership influences teacher and school-level profiles through promoting teacher collaboration (see Figure 1).
Conceptual model.
Method
This study employed quantitative research techniques, including a cross-sectional survey design with a secondary data analysis method.
Data Source
This study is based on a secondary analysis of the TALIS 2018 dataset. The TALIS is an international study that focuses on the working conditions of school principals and teachers and investigates the learning and teaching environments in schools. While the TALIS collects data from school principals regarding school leadership, teacher evaluations, school climate, and job satisfaction, the teacher questionnaires focus on professional development, feedback, instructional practices, school climate, job satisfaction, and self-efficacy (OECD, 2019a). The TALIS was first conducted in 2008, and again in 2013 and 2018. The TALIS 2018 was held at the lower secondary level (ISCED 2) with the participation of 48 countries. However, since 2013, the data has typically been collected in participating countries at the primary school (ISCED 1) and high school (ISCED 3) level. While Türkiye participated in the TALIS 2008 only at ISCED level 2, it did not participate in the TALIS 2013. However, Türkiye participated in the TALIS 2018 study at all school levels (ISCED 1, 2, and 3). A total of 825 school principals and 15,498 teachers working in the three different levels of Turkish primary, lower secondary, and high schools participated in the TALIS 2018 study (OECD, 2019b).
The present study used nationally generalizable ISCED Level 2 data from Türkiye. The TALIS 2018 employed a stratified two-stage probability sampling design. In the first stage, schools were randomly selected from public and private schools located across twelve regions (strata) in different parts of Türkiye by using the probability proportional to the size of the strata. In the second stage, approximately 20 teachers were randomly chosen within the selected schools. Therefore, the dataset had clustered a structure (teachers within schools) that allowed us to conduct a Multilevel Latent Profile Analysis (MLPA).
The study sample was comprised of a total of 196 schools with 3,223 teachers (with an average cluster size of 16), representing 16,228 schools and 310,932 teachers nationwide. In this sample, 58.3% of the teachers were female, 79.6% had completed a bachelor's degree program (ISCED level 6), and approximately 50% had ten or fewer years of teaching experience. We included sampling weights in the model at both the teacher (TCHWGT) and school level (SCHWGT) to reject the unequal probability of the selection of teachers and schools.
Instruments
This study used teachers’ self-reports on instructional practices, collaboration among teachers in lessons, and distributed leadership from the TALIS 2018 dataset.
Indicator variable: Teachers’ Instructional Practices
This scale encompasses three dimensions: clarity of instruction (CLAIN), classroom management (CLASM), and cognitive activation (COGAC). CLAIN (four items) represents setting clear learning goals, making a connection between old and new topics, and providing a summary of the current lesson. One sample item from this dimension is: “I present a summary of recently learned content.” The OECD (2019a) reported that the omega reliability coefficient for this dimension was high (ω = 0.925), and the CFA measurement model produced a good model fit for the Turkish data: CFI = 1.000, TLI = 1.003, RMSEA = 0.000, SRMR = 0.002.
CLASM (four items) measures teachers’ practices for ensuring an orderly classroom environment and effective use of time during lessons. One sample item from this dimension is: “I tell students to follow classroom rules.” The OECD (2019a) reported that the omega reliability coefficient was high (ω = 0.880), and the CFA measurement model produced a good model fit for the Turkish data: CFI = 0.996, TLI = 0.975, RMSEA = 0.049, SRMR = 0.006.
COGAC (four items) refers to student-centered activities that encourage students to apply, analyze, and evaluate their knowledge to reason and solve problems. One sample item from this dimension is: “I present tasks for which there is no obvious solution.” Teachers answered the items on a four-point Likert-type scale ranging from never (1) to almost never (4). The OECD (2019a) reported that the omega reliability coefficient was acceptable (ω = 0.771), and the CFA measurement model produced a good model fit for the Turkish data: CFI = 1.000, TLI = 1.001, RMSEA = 0.000, SRMR = 0.002.
Professional Collaboration among Teachers in Lessons (T3COLES)
This scale refers to collaborative practices with other teachers to plan, organize, and conduct classroom activities. One sample item from the scale is: “Teach jointly as a team in the same class.” Teachers answered the items on a six-point Likert-type scale ranging from never (1) to once a week or more (6). The OECD (2019a) reported that the omega reliability coefficient was acceptable (ω = 0.766), and the CFA measurement model produced a good model fit: CFI = 0.996, TLI = 0.988, RMSEA = 0.036, SRMR = 0.005.
Distributed Leadership (DL)
This scale involves three items (TT3G48A, C, D) measuring teacher opinions on school decision-making policies. One sample item from the scale is: “This school provides staff with opportunities to actively participate in school decisions.” Teachers answered the items on a four-point Likert-type scale ranging from strongly disagree (1) to strongly agree (4). The omega reliability coefficient was acceptable (ω = 0.949), and the CFA measurement model produced a good model fit for the Turkish data: CFI = 1.000, TLI = 1.000, RMSEA = 0.000, SRMR = 0.000.
Control Variables
At the teacher level, we included basic demographic characteristics including teacher gender (TT3G01, female = 1 and male = 0), the highest level of formal education completed (TT3G03, master's or doctoral degree = 1, else = 0), job experience (TT3G11B), teaching offered a steady career path (TT3G07A, moderate/high importance = 1, else = 0), and enrollment in target class (TT3G38) in our model to control for any effect they may have had on instructional practices. To incorporate these variables, we relied on several previous studies reporting that teacher gender (Lee et al., 2019; Thoonen et al., 2011), job experience (Thoonen et al., 2011), and class size (Greenwald et al., 1996) are related to teacher instructional practices.
At the school level, we included several variables as controls, such as school location (SCHLOC) (rural = 1, else = 0; and city = 1, else = 0), lack of resources (T3PLACRE, not a problem = 0, else = 1), public or private school (TC3G12, private = 1, else = 0), student enrollment (NENRSTUD), and diversity of student background (TC3G17A–E) by using principals’ responses (BCGTURT3). The diversity of student background variable encompasses schools with a high percentage of students with a first language different from instructional (TC3G17A), a high percentage of students with special needs (TC3G17B), a high percentage of students from socioeconomically disadvantaged homes (TC3G17C), a high percentage of students who are immigrants or with migrant backgrounds (TC3G17D), and a high percentage of students who are refugees (TC3G17E). We recoded them as 1 for 11% and above percent of students = 1, and 0 for else (Bowers, 2020).
Analytic Approach
This study benefited from a person-centered perspective in the multilevel context to differentiate between subgroups of teachers who produce similar patterns in their instructional practices. While variable-centered approaches (e.g., regression, SEM—Structural Equation Modeling, MSEM—Multilevel Structural Equation Modeling, etc.) employ data to estimate a single set of parameters, person-centered methods apply individual reports to identify homogeneous subgroups in heterogeneous data and estimate a distinct set of parameters (Mäkikangas et al., 2018). To accomplish this, in the first step, we conducted a preliminary analysis including descriptive statistics and correlations among variables. In the second step, we performed a single-level (unconditional) LPA to identify the profiles of teachers’ instructional practices in Türkiye based on their responses to the scales. Our data had a hierarchical (or clustered) structure in which teachers (Level 1) were nested within schools (Level 2). Therefore, after using single-level LPA, we employed unconditional MLPA to decide which school-level profiles were more appropriate when teacher-level profiles were fixed. As multi-group modeling does not need random start values (Asparouhov & Muthén, 2015), we specified the starting value as zero. In this step, we employed information criteria, including Log-likelihood (LL), Bayesian Information Criterion (BIC), Akaike Information Criterion (AIC), Sample-size Adjusted BIC (SABIC), entropy, Lo-Mendell-Rubin Adjusted Test (LRT), and Bootstrap Likelihood Ratio Test (BLRT) to choose the best fitting model. Lower values of LL, BIC, AIC, and SABIC point to a better model fit (Nylund et al., 2007). Entropy values that are closer to 1 (e.g., values > 0.80) indicate that the latent profiles are highly discriminating (Muthén & Muthén, 1998–2017). Moreover, LRT and BLRT with a small probability of likelihood (e.g., p <.05) illustrate that the k profile model represents a better fit than the k–1 profile model. Based on past research (e.g., Ahn et al., 2021; Boyce & Bowers, 2018; Marsh et al., 2012), we labeled our teacher and school profiles by considering different interpretations of the construct meanings at each level.
In the third step, we performed the effects of the Level 2 (distributed leadership and professional collaboration) and Level 1 covariates on the random intercepts of the teachers’ latent profiles. An important feature of multilevel mixture modeling is that “Level 2 predictors can be specified to predict the random intercepts of latent profiles at Level 1” (Van Horn et al., 2016). In the fourth step, we performed MLPA with covariates to examine the association of distributed leadership and professional collaboration with the school-level profiles, following the suggestions of Flunger et al. (2021). We performed all analyses using Mplus software (version 8.6) with default estimator MLR. To eliminate the problem of changing profiles, we used the starts values generated as a result of unconditional MLPA. Finally, following suggestions from several authors for performing mediation analysis (e.g., Baron & Kenny, 1986; Hayes & Preacher, 2014; Hayes & Rockwood, 2017; MacKinnon et al., 2002, 2007; Preacher & Hayes, 2004), we performed the “model constraint” feature of Mplus to calculate indirect effects by multiplying direct effects between variables (a x b). More specifically, we tested the “distributed leadership - professional collaboration” path (a), and “professional collaboration—teachers/school profiles” path (b) simultaneously. Mplus uses the delta method to construct standard errors. Finally, we used the 95% credibility interval (CI) of indirect effects to check whether 0 was not located between the lower and upper bound (Hayes, 2009). We did not use the bootstrap procedure, because it is not available for multilevel modeling in Mplus.
Results
Step 1: Preliminary Analysis
Table 1 shows the mean, standard deviation, skewness, kurtosis, and correlation coefficient results. These findings illustrate that the distribution of mean scores regarding instructional practices demonstrated a high level of clarity of instruction (M = 12.37) and classroom management (M = 10.85), but a surprisingly low level of cognitive activation (M = 9.92). This means that teachers spent less time assigning students tasks for critical thinking and creating small groups for generating a joint solution to a given problem. The skewness and kurtosis coefficients ranged between −3 and +3, pointing to a normal distribution. At the school level, there was a positive relationship between professional collaboration in lessons among teachers and cognitive activation (r =.395; p <.01). In addition, distributed leadership was positively correlated with professional collaboration in lessons among teachers (r =.110; p <.01).
Descriptive Statistics and Correlations Among Variables.
M = mean; SD = standard deviation; S = Skewness; K= Kurtosis; CLAIN = clarity of instruction; CLASM = classroom management; COGAC = cognitive activation; COLES = professional collaboration in lessons among teachers; DL = distributed leadership; JOBEXP = teacher job experience; CSIZE = enrolment in target class; GENDER = teacher gender (female=1); EDUCATION = master or doctoral degree is the reference group; CAREER = teaching offered a steady career path (moderate/high importance=1); RURAL = school location (rural=1); CITY= school location (city=1); LACK = lack of resources (not a problem = 0, else = 1); PRIVATE= private school; NENRSTUD = student enrolment, Diversity: high percentage of students with first language different from instructional (TC3G17A), special needs (TC3G17B), socioeconomically disadvantaged homes (TC3G17C), immigrants or with migrant backgrounds (TC3G17D), and refugees (TC3G17E).
** Correlation is significant at the 0.01 level.
Step 2: Identifying Latent Profiles at the Teacher- and School-Level
In this step, we aimed to estimate a single-level LPA (see the appendix for Mplus syntax; Model 1) to explore classroom instruction profiles at the teacher level (Level 1) by checking the information criteria. Table 2 presents the solutions of the LPA with seven latent profiles.
Fit Statistics for the Single-Level LPA.
Boldface font indicates the selected model.
LP = latent profile; LL = log-likelihood; P = parameters; BIC = Bayesian information criterion; AIC = Akaike information criterion; SABIC = sample-size adjusted BIC; LRT = Lo-Mendell-Rubin adjusted test; BLRT = Bootstrap likelihood-ratio test; smallest group = proportion of prescribers in the smallest group.
Table 2 demonstrates that the values of information criteria—namely BIC, AIC, and SABIC—continue to decline through the eight-profile model, and both the LRT and BLRT remain significant at the p<0.05 level through the seven-profile model. Furthermore, the entropy for the six latent profiles was .979, suggesting the six-profile model demonstrated greater precision in classification (Kaplan & Keller, 2011). We also checked elbow plots to identify the point at which the decrease flattened. In Figure 2, the slope of the line appears to increase somewhat between the four-profile model and the five-profile model. Model selection is an active area of debate in the mixture modeling literature (see Morgan et al., 2016). Numerous articles have demonstrated that fit criteria alone often fail to identify a single “best fitting” model, often forcing researchers to use their discretion and defend their choice of model (see Choi et al., 2020; Collie et al., 2021; Moore et al., 2019; Nylund-Gibson & Choi, 2018; Nylund-Gibson et al., 2014). Given that the fit indices did not unilaterally classify individuals’ responses as the correct one (Nylund-Gibson et al., 2014), we further examined the possible solutions for deciding the best model by considering the conceptual relevance and meaningfulness. As the analysis proceeded and new profiles emerged, we examined each profile to see whether it differed from others in terms of stability and interpretability.
Elbow plot for the information criteria (BIC, AIC, and SABIC) yielded with latent profile analyses.
We checked the latent class proportions of the smallest group, as previous studies have indicated a poor model fit if they are below 5% of the sample (Bowers, 2020). While the smallest group of our four-profile solution made up 11% of the sample, this proportion dropped to 2% in the five-profile solution and lower than 1% in the six- and seven-profile solutions. Accordingly, Nylund-Gibson and Choi (2018) suggest that researchers refrain from selecting an over-extracted and potentially unstable class solution. Additionally, Nylund-Gibson et al. (2022, p. 7) argue that “depending on sample size and power, a small class may not be identified in the context of mixture models because of the signal-to-noise ratio.” Since previous studies did not examine teacher profiles for instructional practices, we instead identified four profiles based on teachers’ classroom management styles (authoritarian, authoritative, permissive, and indulgent), which were originally sourced from Baumrind's parenting classification (1971) and then applied to the classroom environment in many previous studies (e.g., Bernstein, 2021; Thi & Nguyen, 2021). When we compared our profile solutions (see Supplementary Online Material), we identified that the five-, six-, and seven-profile solutions did not add any new profile that conceptually differed from others. Additional profiles appeared more like the division of one profile into two similar ones. For example, profile 2 in the four-profile solution seems to be divided into two distinct profiles (profile 3 and profile 4) in the five-profile solution. Consequently, we selected the four teacher-level profile solution as an optimal model.
Characteristics of the Teacher-Level Profiles
Figure 3 displays the standardized (s) scores regarding four latent profiles at the teacher level. The first profile contained 15% of the teachers (n = 476) and was characterized by low levels of classroom management (s = −1.792), clarity of instruction (s = −0.198), and cognitive activation (s = −0.225) practices. We labeled this profile as “laissez-faire.” This profile represents individual teachers who focus much less on practices to create an orderly learning environment, such as setting classroom rules and managing disruptive student behaviors, than practices related to the clarity of instruction and cognitive activation. The second profile included 45% of the teachers (n = 1464) and was characterized by a medium level of cognitive activation (s = −0.097) and low levels of classroom management (s = −0.358) and clarity of instruction (s = −0.264) practices. As the standardized scores are within 0.5 SD of the mean (slightly below), we labeled this profile as “typical” (see, for example, Klukas et al.'s [2021] labeling of parent profiles). This profile refers to individual teachers who pay an average level of attention to all three teaching practice areas. The third profile covered 29% of the teachers (n = 944) and was characterized by a high level of classroom management (s = 0.945), medium level of clarity of instruction (s = 0.093), and cognitive activation (s = −0.314) practices. We labeled this profile as “controlling,” because it represents individual teachers who focus more on ensuring an orderly classroom environment and using time during lessons effectively. The last profile included 11% of the teachers (n = 339) and was characterized by high levels of cognitive activation (s = 1.688), classroom management (s = 1.128), and clarity of instruction (s = 1.002) practices. We labeled this profile as “versatile,” representing individual teachers who pay significant attention to all three areas of teaching practices.
Standardized scores of instructional practices of the teacher level profiles.
After performing a single-level LPA, we employed MLPA to decide which school-level profiles were more appropriate when the teacher-level profiles were fixed at four (see the appendix for Mplus syntax; Model 2). Table 3 illustrates the solutions of the MLPA with six latent profiles.
Fit Statistics for the MLPA at the School Level.
Boldface font indicates the selected model.
LP = latent profile; LL = Log-likelihood; P = Parameters; BIC = Bayesian information criterion, AIC = Akaike information criterion; SABIC = sample-size adjusted BIC.
Table 3 demonstrates that the values of the information criteria—which included BIC, AIC, and SABIC—decreased when shifting from one to three latent profiles. Therefore, we rejected the four, five, and six latent profiles. Moreover, three latent profiles yielded an additional profile with very few cases and low classification probabilities (n = 41, 1%). Bowers (2020) noted that additional latent class proportions below 5% of the sample point to a poor model fit. A class size with fewer than 50 cases allows for a “trustworthy generalization” (Muthén & Muthén, 2000). The entropy for these two latent profiles was.941, suggesting more precision in classification and clearer distinction among profiles (Kaplan & Keller, 2011). We also inspected elbow plots to identify the point at which the decrease flattened (see Figure 4).
Elbow plot for the information criteria (BIC, AIC, and SABIC) yielded with multilevel latent profile analyses.
Figure 4 indicates that decreases in the information criteria in the school-level profiles tended to flatten around two profiles. Consequently, our criteria suggested a solution with two school-level profiles. Table 4 presents the frequencies of teachers in the teacher- and school-level profiles.
Frequencies of Teachers for Teacher- and School-Level Profiles.
P = profile.
Characteristics of the School-Level Profiles
The two school-level profiles identified in Table 4 indicated that more teachers were classified as Profile B (n = 1817) than Profile A (n = 1406). We checked the frequencies of teachers in both Profiles A and B to explore whether the school-level profiles were identified by the distributions of teachers’ instructional practice profiles. The findings of this analysis illustrated that the proportion of the second teacher profile (“typical”) is nearly the same in both school profiles. The distinction appears to be around the distribution of teacher profiles 1 and 3—those we have labeled “laissez-faire” and “controlling.” The majority of the laissez-faire teachers were assigned to Profile B (n = 406) rather than Profile A (n = 77). Thus, we called Profile B “high laissez-faire.” This profile represents the collective teaching body of a school that focuses less attention on practices that create an orderly learning environment in the classroom, such as setting classroom rules and managing disruptive student behaviors. Since the majority of the controlling profiles were assigned to Profile A (n = 487), we named this profile “high controlling.” This profile represents the collective teaching body of a school that focuses more on ensuring an orderly classroom environment and using time during lessons effectively.
Step 3: Investigating the Role of Level 1 and Level 2 Covariates in the Level 1 Profiles
In the third step, we performed MLPA with covariates, including teacher gender, the highest level of formal education, job experience, teaching offered a steady career path, and enrollment in target class as control variables (see the appendix for Mplus syntax; Model 3). By default, the four latent profiles served as the reference class against which the others were compared. When we added the covariates in the model, the LL (−12212.305), BIC (24617.573), AIC (24472.610), and SABIC (24541.315) values declined. This means that our model had a better fit (Nylund et al., 2007). Table 5 presents the effects of Level 1 and Level 2 covariates on the profiles of teachers’ instructional practices.
Direct Effects of the Covariates on Four Profiles Of Teachers’ Instructional Practices.
Professional collaboration is school level covariates; the others are teacher level covariates.
We used different starts values based on the PLA for reference profile.
Gender = Female is the reference group; b highest level of formal education completed = Master or doctoral degree is the reference group.
Teaching offered a steady career path = Moderate/high importance is the reference group.
Est. = standardized estimate; SE = standard error.
* Correlation is significant at the 0.05 level.
** Correlation is significant at the 0.01 level.
As for the control variables, teacher gender had a significant effect on instructional practices in profiles two and three. In this regard, female teachers were more likely to be classified in the “typical” (β = −0.162, SE = 0.058, p <.01) or “controlling” (β = −0.195, SE = 0.066, p <.01) profiles than “laissez-faire.” Teachers who completed master's or doctoral degrees were more likely to be classified in the “versatile” profile than “controlling” (β = −0.209, SE = 0.072, p <.01) or “typical” (β = −0.141, SE = 0.067, p <.05). While job experience only predicted the likelihood of “typical” profile membership rather than “versatile” (β = −0.159, SE = 0.077, p <.05), teachers who reported that teaching offered a steady career path were likely to be classified in the “versatile” profile rather than “controlling” (β = −0.277, SE = 0.082, p <.01), “typical” (β = 0.242, SE = 0.072, p <.01) or “laissez-faire” (β = −0.202, SE = 0.086, p <.05) profiles. Enrollment in target class positively predicted teachers’ classification into the “controlling” profile, rather than the “laissez-faire” (β = −0.227, SE = 0.073, p <.01) or “typical” (β = −0.140, SE = 0.048, p <.01) profiles.
As for the school-level covariates, the overall perception of a school's faculty that their school has a higher level of professional collaboration in lessons among teachers positively and significantly predicted the likelihood that individual teachers would be classified in the “versatile” teaching profile rather than “laissez-faire” (β = −0.351, SE = 0.113, p <.01), “typical” (β = −0.286, SE = 0.091, p <.01), or “controlling” (β = −0.241, SE = 0.097, p <.05) profiles. Table 6 presents the indirect effects of distributed leadership through professional collaboration on the four profiles of teachers’ instructional practices.
Indirect Effects of the Distributed Leadership on Four Profiles of Teachers’ Instructional Practices via Professional Collaboration.
Note (s): Est. = standardized estimate; SE = standard error.
* Correlation is significant at the 0.05 level.
** Correlation is significant at the 0.01 level.
Before calculating indirect effects, we found that distributed leadership had a positive and moderate direct effect on professional collaboration among teachers (β = 0.269, SE = 0.065, p <.01). Next, we multiplied the coefficients by using the “model constraint” feature of Mplus. The results showed that when a school's collective faculty perceived their school had a higher level of distributed leadership, they were more likely to collaborate around instructional matters, which in turn increased the likelihood that individual teachers would be classified in the “versatile” profile rather than “laissez-faire” (β = −0.094, SE = 0.011, p <.05), “typical” (β = −0.077, SE = 0.013, p <.05) or “controlling” (β = −0.065, SE = 0.040, p <.05).
Step 4: Investigating the Role of Level 2 Covariates on the Level 2 Profiles
In the final step, we conducted MLPA to test the effects of the Level 2 covariates on the Level 2 latent profiles (see the appendix for Mplus syntax; Model 4). Table 7 presents the results.
Direct Effects of the Covariates on Two School Profiles.
Private is the reference group; b %11 and above percent of students is the reference group.
Est. = standardized estimate; SE = standard error.
* Correlation is significant at the 0.05 level.
The results showed that schools where the collective faculty perceived a higher level of professional collaboration in lessons among teachers were more likely to be assigned to the “high controlling” profile rather than “high laissez-faire” (β = 0.606, SE = 0.249, p <.05). To calculate the indirect effect of distributed leadership on school profiles, we used the result of the school-level distributed leadership—professional collaboration path, and professional collaboration—school profiles path. The results showed that when a school's collective faculty perceived their school as having a higher level of distributed leadership, they were more likely to collaborate with each other around instructional matters, which in turn increased the likelihood of their schools being classified in the “high controlling” profile rather than “high laissez-faire” (β = 0.161, SE = 0.080, p <.05).
Limitations
The present study has several limitations that should be recognized when interpreting the findings and considering avenues for further research. First, our study depends upon the data collected through the TALIS (2018); thus, we followed the OECD's (2019a) conceptualizations of the study constructs. Although the OECD (2019a) has provided a sound conceptual and empirical framework for each of these constructs, it is still debatable whether each construct used in the TALIS dataset adequately covers its theoretical infrastructure. For instance, while the TALIS identifies distributed leadership as teachers, students, and parents’ participation in school decision-making processes, it does not capture the wider aspect described by Spillane et al. (2001, p. 27), who explain that “leadership practice emerges in and through the interaction of leaders, followers, and situation.” However, the other side of the coin is that TALIS provides us with a large-scale dataset, allowing the generalization of our results to the entire teacher population in Türkiye.
Second, since the TALIS does not provide longitudinal data, we were unable to make causal inferences regarding the associations among our study variables. Naturally, our data did not allow us to explore whether and to what extent the recent reform initiatives in Türkiye have been able to make the desired changes in teachers’ instructional practices, which is especially important to inform nationwide educational policymaking. Therefore, we believe that the literature could benefit from further longitudinal research to gain a deeper understanding of the nature of teachers’ instructional practices as well as how school leadership practices along with other teacher and school-level factors predict variations in teaching.
Third, our study was based solely on teacher-self reports to measure the covariates and identify the subgroups of teachers. Although several scholars have favored measuring principal leadership practices using teacher reports (e.g., Hallinger, 2011; Thoonen et al., 2011), concerns still arise over the objectivity of the data, as teachers might overestimate their own instructional practices. Besides, drawing solely upon teacher perceptions may provide a narrow view of educational endeavors, as schools include a variety of other stakeholders, such as parents and principals (Ahn et al., 2021; Boyce & Bowers, 2018; Duff & Bowers, 2022). We thus suggest that future research develop and test the congruency-typology model by using both teacher and principal ratings, which could yield deeper insights into the different types of teachers who perform instructional practices in their classrooms (Bowers et al., 2017).
Fourth, our study explores only a single dimension of leadership, while the literature has provided evidence of the significant effects of many other leadership types, such as instructional and transformational leadership, on teaching and learning (Gumus et al., 2018; Robinson et al., 2008; Thoonen et al., 2011). We also acknowledge the relevance of the integrated or hybrid model of leadership that challenges the traditional notion that school principals are either transformational or instructional leaders, an issue that has been extensively debated in this journal (e.g., Bellibaş, Kılınç, et al., 2021; Day et al., 2016; Kwan, 2020; Marks & Printy, 2003). Therefore, we encourage researchers to consider the multifaceted and contextual nature of leadership to advance our understanding of the link between school leadership and teaching.
Discussion
Grounded mainly on the normative conceptualization of distributed leadership and borrowing from other perspectives like “distributed leadership as work redesign” and “distributed leadership as a mechanism of building human capacity,” this study attempted to (1) discover whether there is a typology of teachers and schools according to teacher instructional practices, (2) examine whether and to what extent individual and school-level factors predicted teachers’ and schools’ membership in each profile, and (3) investigate the relationship between distributed leadership and teacher profiles with the mediating role of teacher collaboration.
The results of our research showed that labeling teachers as “good” or “bad,” or “effective” or “ineffective” based on their instructional practices does not fit well with real-world scenarios. Consistent with the recent research that supports a person-based approach rather than an item-based one when it comes to identifying the practices and behaviors of people through survey data (Bowers, 2020; Bowers et al., 2017), teachers in the present study were categorized in several groups based on their varying strengths and weaknesses related to their teaching practices. For instance, controlling teachers demonstrated high levels of classroom management, medium levels of clarity of instruction, and lower levels of cognitive activation, while typical teachers reported an average level of attention to all three areas of teaching practices. On the other hand, although laissez-faire teachers demonstrated a lower level of focus on clarity of instruction, cognitive activation, and classroom management, versatile teachers seem to be more skilled in these three areas.
Second, our examination of the factors predicting teachers’ instructional profiles produced several key findings. First, teachers who completed master's or doctoral degrees are more likely to be classified in the “versatile” profile than “controlling” or “typical.” Though it is unclear whether these teachers’ graduate degrees came from programs with an educational focus, this finding might suggest that teacher participation in qualification programs allows them to increase their content knowledge, possibly leading them to improve their classroom teaching practices. This finding found support from previous research suggesting that traditional professional development (i.e., degree programs) influences the quality of teachers’ instructional practices, such as managing the classroom, establishing a student-oriented classroom climate, and supporting cognitive activation by providing students with challenging content (e.g., Doğan & Yurtseven, 2018; Kowalczuk-Walędziak et al., 2017). Second, target class enrollment positively predicted teachers’ classification into the “controlling” profile, rather than the “laissez-faire” or “typical” profiles, meaning that teachers tend to prioritize managing the classroom effectively and articulating the goals of instruction along with expectations from students in classrooms with higher enrollment. Previous studies illustrated that large classes are one of the greatest obstacles to the application of student-centered pedagogy in Türkiye (Altinyelken, 2011; Gelbal & Kelecioglu, 2007). Third, teachers who reported that teaching offered a steady career path were more likely to be classified in the “versatile” profile rather than “controlling,” “typical,” or “laissez-faire.” This suggests that those teachers who chose the profession because it offers a steady career path tend to focus their attention on all three areas of instructional practice. This finding is noteworthy because, as discussed above, teaching is not viewed as a career profession in Türkiye; rather, egalitarian norms dominate the profession, making all teachers equal regardless of their education levels, experience, and seniority. However, we wish to note that Turkish policymakers have recently made significant progress in this direction by passing the Teaching Profession Law, which aims to transform teaching into a career profession (Official Gazette, 2022).
Finally, we investigated the relationship between distributed leadership and teachers’ instructional profiles with the mediating role of teacher collaboration at both the school and teacher level. While the school improvement literature has provided persuasive evidence that the distribution of leadership practice could improve collaboration and the quality of teaching (e.g., Bellibaş et al., 2022; Firestone & Cecilia Martinez, 2007; Spillane & Healey, 2010), these previous studies treated teaching as an item-based variable, indicating that teachers would demonstrate a similar pattern in all sub-dimensions of instructional practices (e.g., classroom management, cognitive activation, and clarity of instruction). The present study illustrates that the person-based approach is more appropriate for identifying the teaching characteristics, strategies, and practices of teachers since it reveals those teachers who demonstrate a mixed pattern in their teaching, rather than solely focusing on those with high or low scores across all sub-dimensions of classroom teaching (Bowers & White, 2014; Duff & Bowers, 2022). Unlike previous research, we found that each teacher is grouped into a different instructional profile, each of which has a complex teaching pattern. The present study is, therefore, the first of its kind to link leadership and collaboration with teaching profiles within and between schools. Our results illustrate that collective teacher perceptions of distributed leadership practices in a school influence individual teachers’ inclusion in different instructional practice profiles indirectly by encouraging the school's collective faculty perceptions of professional collaboration.
This finding shows the relevance of the work redesign perspective to distributed leadership (Mayrowetz et al., 2007; Mayrowetz & Smylie, 2004) and its linkage with collaboration and teacher practices, by suggesting that when principals include teachers in school decisions and foster collaboration around instructional issues, teachers are more likely to modify their teaching patterns. More specifically, individual teachers working in schools with a higher level of distributed leadership are more likely to be classified in the “versatile” profile than “laissez-faire” or “typical,” with the mediating function of professional collaboration. This finding adds new evidence to the school improvement literature (e.g., Bellibaş et al., 2022; Harris & DeFlaminis, 2016; Leithwood et al., 2020) by suggesting that the distribution of leadership practices among multiple stakeholders influences individual teachers’ likelihood of membership in different subgroups according to their classroom practices by supporting collaborative teamwork in schools. This finding also lends credence to the “distributed leadership as a mechanism of building human capacity perspective” (Harris, 2004; Mayrowetz, 2008) because, as the present study shows, the distributed leadership practices executed through this perspective supported variation in the profiles of teachers’ instructional practices through influencing their engagement in school decision-making processes and collaborative work.
Our analysis shows a more homogeneous distribution of teaching profiles across schools, supporting the previous research finding that there is more variation within schools than between schools in terms of teaching quality (Hanushek et al., 2005). Only two school-level profiles—high controlling (which included more teachers in the controlling profile) and high laissez-faire (which included more teachers in the laissez-faire profile)—emerged. In addition, leadership and collaboration were also significant predictors of these two profiles. We found that the perceptions of a collective body of teachers regarding the execution of distributed leadership practices had a direct effect on school-level teacher collaboration and an indirect, though small, effect on the “high controlling” school profile. This means that when principals favor teacher participation in school decision-making processes and seek to build and secure a culture of shared responsibility around instructional improvement, teachers are more likely to collaborate with colleagues, which in turn decreases the possibility of “laissez-faire” schools located in the “high controlling” school profile. This finding is noteworthy because an item-based approach could have found a significant relationship between distributed leadership and teacher instructional practices, which would have then been interpreted such that the more the school leadership practices are distributed the better instructional strategies teachers will have. Yet, such a clear conclusion may not be accurate, as the present study indicates. This research has demonstrated that greater distribution of leadership practices would increase the number of high-controlling teachers in a school; however, it remains unclear whether high-controlling teachers are truly “better” than laissez-faire teachers.
Conclusion
We believe our results can advance the field's knowledge of the link between leadership and teaching on several fronts. First, this study addresses the gap in the literature regarding the effects of distributed leadership, a highly relevant leadership emphasis for both policy and practice, on individual teachers’ and schools’ latent profiles concerning their instructional practices. We concluded that teachers working in schools where leadership is distributed among stakeholders are more likely to collaborate with colleagues around instructional and curricular issues, which in turn increases the likelihood of these teachers being classified as “versatile.” We also concluded that the implementation of distributed leadership practices promotes teacher collaboration, which ultimately influences the possibility of any given school being classified as “high controlling” rather than “high laissez-faire.” Thus, we provided empirical evidence to substantiate prior assertions that the enactment of distributed leadership could lead to an increase in school capacity to promote teacher collaboration and improve teaching practices (e.g., Hallinger, 2011; Liu, 2021). This finding echoes Mayrowetz's (2008) contention in this journal over a decade ago that treating distributed leadership as a process of building human capacity offers a stronger link with school improvement endeavors than any other perspectives on the construct (e.g., distributed leadership for democracy). Third, this study expands the boundaries of educational leadership research by confirming teacher collaboration as a significant mediator in the relationship between distributed leadership and both individual teacher- and school-level profiles.
Implications
Our study offers several implications for policy, practice, theory, and research. Beyond “average estimations of teacher instructional practices that do not appropriately capture teacher perceptions” (Duff & Bowers, 2022, p. 18), this study benefited from MLPA to identify heterogeneity in the way teachers perceive their instructional practices. At their core, our results suggest that the “one-size fits all” approach to improving teachers’ instructional practices may not work as intended, as there are different typologies of teachers according to their instructional practices, suggesting that each profile offers different needs and priorities. Therefore, we believe that policymakers could use our results to help teachers improve their instructional practices based on each profile's needs. For instance, investing in building “laissez-faire” and “typical” teachers’ instructional capacity could be a policy priority for policymakers in Türkiye and other developing nations.
In terms of practice, the difference between teachers’ instructional practices is more complex than simply labeling them as “effective” or “ineffective.” Schools do not have teachers who are either good or bad in all components of teaching; rather, teachers can be identified through profiles, representing different teaching configurations. Our results highlight the importance of the distribution of leadership for school collaboration and ultimately teachers’ instructional profiles. Leadership distribution encourages teachers to participate in school decision-making processes and build an environment where collaboration among colleagues is nurtured, which in turn can help secure high-quality teaching and augmented learning. Our results also imply that school principals should act responsibly to identify teachers’ instructional profiles, diagnose their learning needs, and provide them with meaningful opportunities for collaboration and in- and out-of-teachers’ professional development to support the teaching skills and content knowledge needed to teach their subjects effectively.
The contributions of the current study to the theoretical literature are twofold. First, this study applied a person-centered approach to uncover the variety of ways that teachers perform their instructional practices, challenging the conventional single category of “instructional practices” employed by most research in this area. This method allows us to identify subgroups of teachers who produce similar patterns in their instructional practices, which could take the existing knowledge base one step further by encouraging further research scrutinizing the link between various leadership emphases and teaching profiles. Second, our findings extended the existing knowledge base by showing the indirect effect of distributed leadership on variations in teachers’ teaching profiles as well as school profiles through promoting teacher collaboration.
Finally, we wish to provide important guidance for the direction of future research. First, it should be noted that the present study only focuses on how teachers perceive different profiles regarding their instructional practices and whether and to what extent teacher- and school-level covariates predict teacher and school membership in any given profile. Thus, we believe there is value in investigating whether and how any given typology of teachers and schools is associated with distal outcomes, such as student achievement. Further studies might consider linking the TALIS and PISA datasets to observe whether subgroup membership is associated with student achievement scores. Second, while we only used the distributed leadership model in this study, there is strong evidence that leadership practices can also be developed in the form of leadership profiles. An important research question would then ask how leadership profiles might be related to teacher profiles. Third, as our study represents one of the first empirical attempts to depict teaching profiles by bringing data from a non-Western and highly centralized educational context, we recommend that future studies seek to replicate our findings in different educational settings. Fourth, although our study explored the efficacy of collaboration as a significant pathway through which distributed leadership influences teacher instructional profiles, we wish to caution researchers that similar significant mechanisms might remain unexplored (e.g., professional learning, collective and self-efficacy, teacher agency, commitment, and satisfaction). Therefore, future research might refine the field by uncovering the potential mediators through which distributed leadership influences teacher instructional profiles. Finally, we encourage educational leadership scholars to conduct cross-cultural quantitative studies to examine whether and how profiles might differ across diverse educational contexts.
Supplemental Material
sj-docx-1-eaq-10.1177_0013161X231159092 - Supplemental material for Exploring Teachers’ Instructional Practice Profiles: Do Distributed Leadership and Teacher Collaboration Make a Difference?
Supplemental material, sj-docx-1-eaq-10.1177_0013161X231159092 for Exploring Teachers’ Instructional Practice Profiles: Do Distributed Leadership and Teacher Collaboration Make a Difference? by Nedim Özdemir, Ali Çağatay Kılınç, Mahmut Polatcan, Selçuk Turan and Mehmet Şükrü Bellibaş in Educational Administration Quarterly
Footnotes
Appendix
Mplus Syntax for the Tested Models
Step 2. Single-Level Latent Profile Analysis
Step 2. Multilevel Latent Profile Analysis
Step 3. Investigating the role of level 1 and level 2 covariates on the level 1 profiles
! The starts values of CW#1, CW#2, CW#3, and CW#4 are the comparison with the “versatile” profile. We performed different sets of starts values for each referent profile.
Step 4. Investigating the role of level 2 covariates on the Level 2 profiles
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Supplemental Material
Supplemental material for this article is available online.
Author biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
