Distributed Leadership Globally: The Interactive Nature of Principal and Teacher Leadership in 32 Countries

Abstract

Purpose: This study provides cross-country evidence regarding the extent to which distributed leadership operates in schools globally and offers preliminary insights about country education policies that explain the reported distributed leadership results. The researchers also investigate the relationship between principals’ reports of distributed leadership and teachers’ reports of the school culture, demonstrating the alignment of principal and teacher perspectives of distributed leadership. The conceptual framework highlights the interaction between leaders in varied situations characterized by leadership functions and country contexts. Research Design: Researchers use both principal and teacher surveys from the 2013 TALIS. In the first analytical stage, latent measures using confirmative factor analysis capture the extent to which principals and teachers were responsible for each of the three leadership functions; scatterplots explore distribution of interactive leadership, and further explanation is offered by document analysis of country policy profiles. In the second stage, the research uses hierarchical linear models for the effect of distributed leadership, specifically principal leadership, teacher leadership, and interactive leadership, on school culture for each country, synthesizing country results with meta-analysis. Findings: Distributed leadership varies by leadership function and appears to be influenced by country education policy. Teachers report that their school culture is conducive to distributed leadership when, in fact, they have the opportunity to lead. Conclusions: Findings add global evidence that country context is an important part of the situation for distributed leadership. The patterns of distributed leadership by function invite further research within each country, particularly to examine the influence of educational policy.

Keywords

distributed leadership leadership functions quantitative research TALIS 2013 principal leadership teacher leadership school culture country context education policy

Introduction

Researchers argue that schools where principals distribute leadership to appropriate stakeholders are more likely to achieve school success (Heck & Hallinger, 2009; Leithwood & Jantzi, 2012). While much of the existing research on distributed leadership (DL) has been conducted in Western contexts, such leadership is relevant in international school settings. Case studies in high-performing countries like Finland and Singapore, for instance, demonstrate strong school capacity due to leadership practices that involve a broad base of stakeholders (Hargreaves & Shirley, 2012).

Empirical evidence reported in the past decade has increasingly supported the relevance of DL to organizational culture, educator commitment and satisfaction, academic capacity, and teaching effectiveness (Angelle, 2010; Bellibas & Liu, 2018; Heck & Hallinger, 2009; Hulpia et al., 2012; Liu, 2016; Liu & Werblow, 2019; Sun & Xia, 2018). The leader who implements DL can intensify a school’s organizational capacity (Heck & Hallinger, 2009) by taking full advantage of intellectual capital and social capital through trust, respect, collegiality, and teamwork spirit (Hargreaves & Fink, 2006).

Using a large international data set for the study reported in this article, we add to the research base for DL with evidence from 32 countries. One aim is to inquire into the leadership distribution among principals and teachers within the specific leadership functions across 32 countries and understand how educational policy influences whether and how DL takes shape. A second aim is to determine whether principals’ reports of distributed leadership align with teachers’ perceptions of school culture as conductive to distributed leadership, and whether the interactive nature of leadership contributes significantly more than single source leadership to the development of this type of school culture.

Literature Review

Past research on DL focused on conceptual and theoretical frameworks (Spillane, 2012), the necessity for leadership distribution in schools, and the theoretical modes in which leadership distribution might operate (Gronn, 2002; Leithwood et al., 2007; Spillane & Diamond, 2007; Spillane et al., 2001). Recent studies explore the relationship of DL to various school aspects such as educator organizational commitment (Hulpia et al., 2012; Liu & Printy, 2017); increased trust and job satisfaction (Liu & Werblow, 2019); professional collaboration and teacher retention (Angelle, 2010; Sun & Xia, 2018); academic capacity (Heck & Hallinger, 2009); teachers’ academic optimism and teacher skills, knowledge, and beliefs (Leithwood & Jantzi, 2012; Liu, 2016; Liu et al., 2020); and student achievement (Heck & Hallinger, 2009; Leithwood & Jantzi, 2012; Malloy & Leithwood, 2017; Marks & Printy, 2003).

Conceptualization of Distributed Leadership

Leaders are those who engage in activities intended to influence the motivation, effort, skills, and engagement of their followers in order to accomplish organizational goals (Spillane, 2012). A distributed perspective of leadership assumes some level of interaction of formal leaders who have assigned roles and informal leaders who come from any positions within the organizational community (Spillane & Diamond, 2007).

Gronn (2002) advanced the idea that leadership is distributed when people come together with a common purpose. Individuals engage in some specific activities, and through the engagement, certain individuals are able to influence others. Spillane et al. (2001) state that the interaction with others in completion of tasks “ignites leadership practice so the result is more than any one leader could contribute” (p. 21). DL as described by Spillane (2012) is stretched over leaders, followers, and situations, with the passage of time; furthermore, it is in the interactions of these domains that distributed leadership practice emerges. Within Spillane’s model, there is an openness between leaders and followers, with changing relationships depending on the situation, that is, a leader in one situation might be a follower in another. Furthermore, Spillane et al. describe situations in terms of the routines, structures, tasks, and tools that illustrate what people do to accomplish purposeful activity toward a goal, the processes and organizational structures that enable their work, and the aspects of the context or environment that mediate the way that individuals interact (Spillane, 2012).

Principals can appoint teachers to lead in specific situations, but often teacher leaders step forward of their own volition or are thrust into leadership by circumstance, becoming informal leaders. Interaction of educators, both administrative and instructional personnel, encourages them to bring their expertise to carrying out designed initiatives and innovations (Harris & Muijs, 2005).

Leadership Is Contextual

Qualitative research evidence has identified additional elements of classroom or school environments beyond the aspects of situation referred to above (routines, structures, tasks, and tools). Under the umbrella of the Distributed Leadership study, Spillane et al. have explored leadership in the context of district policy, subject matter, school level and urbanicity, and racial composition. A recent article explores leadership practice in one elementary school in the context of governmental high-stakes accountability policy and shows how macro-level policy is negotiated at the micro-level of school leadership practice in formal school settings (Spillane & Anderson, 2019). Authors of the study demonstrate the ways in which school leaders communicate the legitimacy of the policy and mobilize teachers’ cooperation.

While similar aspects of situation are at play for leadership interactions in schools around the world, international evidence illustrates the ways in which organizational leadership, as a construct, is perceived differently in different societies (Day & Leithwood, 2007; Hofstede, 1984; Liu, 2020; Moos et al., 2011; Ylimaki & Jacobson, 2011). Building on prior research (e.g., Cyert & March, 2006; Hofstede, 1984; McClelland, 1967; Mulder, 1971), the large-scale, multiphased GLOBE (Global Leadership and Organizational Effectiveness) study has shown that leadership is contextual, embedded in societal and cultural values and norms, and that the fit between societal expectations and leadership styles is an important predictor of organizational success (House et al., 2013).

Given prior research, we propose that country context is an important consideration in a global investigation of DL, and we suggest that country context might be an important extension of the term “situation” as articulated by Spillane et al. We consider that “situation,” therefore, conveys layer of contexts: first, the immediate activity system of work; second, the organizational context encompassing goals, structures, resources; and third, the country, or governmental, context. In the current study, we investigate whether the country context illuminates variations in DL operation. A cross-context exploration is important for a baseline understanding of DL operation for relevant tasks in an international context.

To explore how a country context might influence DL, we include details of educational policy as an observable component of country context. As an example from existing research, the International Successful School Principalship Project studied essential educational contexts of countries. Focusing on instructional leadership, organizational capacity building, and culturally responsive practices, the study team found that policy factors such as increased accountability, decentralized educational system, along with demographic composition are related to school leadership priorities and effectiveness (Day & Leithwood, 2007; Moos et al., 2011; Ylimaki & Jacobson, 2011). Such findings for instructional leadership and organizational capacity-building suggest that examining educational policy across countries might be a direction for DL study because some policies at the national or regional level could potentially levy more distributed leadership, while other policies might hinder its occurrence (Liu, 2020).

Leadership Functions Distributed in Schools

Previous researchers have presented frameworks detailing leadership functions, often fulfilled by the principal (Hargreaves & Fink, 2006). Such frameworks have also been used to describe the ways in which leadership distribution can happen (Heck & Hallinger, 2009; Leithwood et al., 2007). For this study, we adopted the framework by Leithwood and his colleagues (Leithwood et al., 2007; Malloy & Leithwood, 2017) to categorize the tasks which are distributed to or shared with teachers. These original four functions include setting school direction, developing people, redesigning organizational structure, and managing instructional practice. Because the data in TALIS 2013 limits our exploration of direction setting, we focus on the remaining leadership functions and research indicating importance of teacher leaders.

Distributed leadership function: Developing people

Leithwood et al. (2007) defined the leadership function of developing people with three components as offering intellectual stimulation, providing individualized support, and modeling appropriate values and practices. Principals play an important role in structuring arenas for teacher interaction, thus creating occasions for teachers to learn with, and from, each other (Printy, 2008; Spillane et al., 2016). Other researchers also provide evidence regarding how teachers help develop other teachers. Through engaging in social interactions, for example, professional learning community and collaboration (Du Plessis & Eberlein, 2018), and information exchange, for example, observations of other teachers’ teaching, analyzing data from teacher evaluation, parent feedback, and student performance, experienced teachers help others learn and develop (Hunzicker, 2018). In environments of collaboration, teachers grow when they have opportunities to establish agendas for instructional improvement and professional learning, particularly when based on data (Ingersoll et al., 2018). These situations can be optimal DL opportunities for teachers to exercise their intellect and skills in developing the collective professional capacity in the school (Camburn et al., 2003; Copland, 2003). Thus, adopting DL in schools is likely to result in enhanced professional capacity and school outcomes (Marks & Printy, 2003; Printy, 2008).

Distributed leadership function: Instructional management

One of the most important school leadership functions is to manage instruction. Schools that are successful usually have an effective instructional leader or leaders (Goddard et al., 2015; Marks & Printy, 2003). The particular roles for instructional leaders are to supervise and evaluate instruction, coordinate the curriculum, and monitor student progress (Hallinger & Murphy, 1985). Early emphasis was primarily on school principals as the change agent for instructional improvement. Later empirical work revealed the importance of teacher leadership in managing instruction, sometimes in designated roles as instructional leaders who help classroom teachers improve practice(Camburn et al., 2003; Copland, 2003). Teachers remaining in the classroom are often active in instruction-related leadership functions such as coordinating the curriculum, selecting course materials, developing discipline, and evaluation policies (Ingersoll et al., 2018). Principals create the conditions within which teachers are able to learn with each other when adopting new materials, setting new targets, implementing new programs, or supporting student success (Printy, 2008). By providing time for meeting, ensuring access to consultants, attending professional developing with teachers, and staying connected to teachers’ work, principals actively engage with teachers in distributed instructional leadership. Research indicates that the interactive leadership of principals and teachers improves collective teacher-efficacy, instructional quality, and student performance (Goddard et al., 2015).

Distributed leadership function: Organizational decision making

Schools, as other organizations, have structures that can encourage interaction among members of a school, such as communication strategies or shared decision-making processes (Angelle, 2010). Principals, as head of school, generally have access to critical information and decision arenas. Teachers have direct knowledge of how students learn and play keys roles carrying out decisions reached about broader school decisions (Wenner & Campbell, 2017).

Teachers need opportunities as well as structure to participate in school decision making in order to contribute to school success (Harris & Muijs, 2005), indicating the importance of principals’ decisions to include teachers and facilitate decision-making interactions. When teachers participate in organizational decision making, such as regular participation on school improvement committees, they experience stronger authority while also gaining broader perspectives on professional responsibility. Evidence, further, shows a strong link of teacher participation in school improvement to student achievement (Ingersoll et al., 2018).

Principal and Teacher Perceptions of Distributed Leadership

As critics of DL have noted, the distribution of leadership influence or participation “exists uneasily” within the bureaucratic power relationships in a school (Hartley, 2010, p. 282). Research shows that principals retain considerable formal power and DL will not be manifest if principals do not support it. While distributed and hierarchical forms of leadership are not incompatible, DL is not a matter of course in any school. Harris (2005) notes that “distribution can work successfully only if formal leaders allow it to take root” (p. 167). As noted, because traditional power relationships are blurred, situations understood as DL might actually create instances where informal leaders are exploited (Law et al., 2010). In other cases, broad-based teacher leadership might have developed to fill a void left by the absence of formal leadership (Harris, 2011; Marks & Printy, 2003). In yet another possibility, school personnel might report indicators of DL with unrealistic enthusiasm in order to appear to be aligned with research or policy documents that tout the benefits of DL. In the United Kingdom, for example, OFSTED examiners expected to see evidence of DL as an indicator of excellent school leadership, so principals offered inflated reports of teachers’ leadership involvement (Bush & Glover, 2012).

This study adopted a measure of school culture as a way to explore the extent to which principals’ accounts of DL are representative of actual involvement (Bush et al., 2006). We follow the lead of other researchers, such as Harris (2005), who claimed that a school culture that supports various elements of DL needs to exist as a precondition to its development. Collegial norms and time to work with other adults are essential if DL is to take hold. Processes for participation need to be consistent and equitable. Relationships between teachers and principals need to be cordial, and formal leaders must not feel that their leadership is threatened (Harris, 2005; Law et al., 2010).

We emphasize the claim of researchers that while structural change, for example, for decision making, is necessary for the existence of DL, it is not sufficient to guarantee positive school outcomes (May & Supovitz, 2011). There is a need for interactive settings where participants can develop new relationships and new avenues of communication and benefit from the resulting professional exchanges (Murphy, 2017). Working in the spirit of mutuality, principals both monitor and support the distributed exercise of leadership occurring throughout the school (Murphy, 2017). Research, thus, points to the essential character of DL as interactive, with opportunities for collaborative decision making, sharing of knowledge and skills, and synergistic actions taking shape across the organization.

Conceptual Framework for This Study

The current study seeks to investigate the leadership engagement of principals and teachers for three specific leadership functions in order to highlight dynamic interactions among leaders and situations. Specifically, we offer evidence of “embedded” situations by expanding the concept of situation as tools, routines, structures. We consider leadership functions and related tasks as a first layer of situation illustrating the dynamic interactions of principals and teachers and our analyses account for schools, a second layer. Next, we introduce country as final layer of situation in order to investigate how distributed leadership patterns vary for the various leadership functions across countries/regions (Figure 1).

Figure 1.

Distributed leadership and school culture.

For the first stage of this study, we adapt a leadership model and analytic approach used in earlier research by one of the authors (Marks & Printy, 2003) where four quadrant scatterplots explore the ways principals and teachers lead interactively. The earlier study examined the relationship between principals enacting transformational leadership and principals and teachers sharing instructional leadership, while this study examines the extent to which principals and teachers contribute to leadership by completing the same tasks. For this study, the authors conceptualize two measures of leadership (principal and teacher leadership) as being on a continuum of low participation to high participation (see Figure 1). Intersecting these continua results in a two by two matrix, with four patterns of interaction of principal and teacher leadership: (Quadrant 1) low–low, low; (Quadrant 2) high–low, Hierarchical; (Quadrant 3) low–high, Lateral; and (Quadrant 4) high–high, Interactive. To explain, we conceive of Low as little active influence or participation from either principals or teachers. Hierarchical means primary influence by principals, generally understood as “top-down” influence. Lateral refers to a high level of teacher leadership with expected strong lateral relationships. Interactive indicates that both formal leaders and informal leaders participate on the relevant tasks and, thus, share influence.

The process above yields scatterplots displaying the country placement for each of the three leadership tasks: developing people, managing instruction, and organizational decision making. To explain why these patterns might exist, we discuss the variations through a consideration of country educational policy.

In the second stage of the study, we seek evidence of alignment between principals’ reports of leadership contributions and teachers’ perceptions of the school leadership culture. The TALIS survey for teachers did not ask questions parallel to the principal survey, limiting our ability to gauge how closely teachers’ perceptions of their participation in leadership functions match principals’ perceptions. We gain insight into the alignment or misalignment of principals and teachers’ perceptions by exploring the relationship of principals’ reports of DL operation to teachers’ reports of cultural conditions that are conducive to DL, specifically, the extent of collaboration and shared activity and whether teachers have opportunities to lead. After examining the relationship of principal leadership and teacher leadership to school culture separately, we further explore the dynamic effect of interactive leadership of principals and teachers together.

Specifically, we seek to answer:

How does distributed leadership vary by country for each leadership function?

What details of country educational policy explain the observed patterns of leadership?

To what extent are principal reports of principal and teacher leadership contributions aligned with teacher perceptions of school culture conducive to distributed leadership?

To what extent does interactive leadership enhance teacher perceptions of school culture?

Method

This study is a cross-sectional secondary analysis of 2013 Teaching and Learning Survey (TALIS) data which is augmented by a document analysis of country educational profiles prepared by Organization for Economic Cooperation and Development (OECD, 2014a). In the section that follows, we first present information about the 2013 TALIS and its sampling strategy, and then, for each research question, we discuss the specific variables from the 2013 TALIS and our analytical approaches. Before presenting details, we summarize the analytic approaches. For the first question, we generated the factor scores of both principal and teacher leadership for the three leadership functions, to reveal the DL patterns across countries. For the second question, we conducted document analysis of educational policy across countries to to inform our understanding of the DL patterns. To answer the third question, we employed hierarchical linear models (HLMs) to investigate how principal and teacher leadership (the factor scores generated from the first step) are related to school culture conductive to DL, respectively. For the last research question, we added the interactive effects of principal and teacher leadership to the above models. The HLMs for the second and third questions were fitted to each of the 32 countries’ data, then we applied meta-analysis to synthesize the findings.

Introduction of TALIS 2013 Study

This study used the 2013 administration of TALIS operated by the OECD. TALIS is the first descriptive, cross-national survey that allows policymakers and administrator to learn about leadership practices, professional development, and teaching in different national contexts (OECD, 2014c). The TALIS 2013 participation included 34 countries and regions, with 32 countries having public data (OECD, 2014b). TALIS required all countries to have the “core” surveys at the lower secondary level (Grades 7-9 generally) for both principals and teachers. The study adopted a stratified two-stage sampling method for data collection (OECD, 2014c); during the first stage, researchers drew 200 schools using systematic random sampling for each country, then selected 20 teachers randomly from each participating school. The TALIS 2013 technical report states that it is essential to apply weights using such a complex survey data. Thus, OECD provided two final weight indices (both for the principal survey and the teacher survey) in the data set, to adjust for the unequal probability of selection and varied response rates. Both weights were used for the corresponding level when conducting the analysis.

Stage 1, Research Question 1: Distributed Leadership for Each Leadership Function

The variables used for distributed leadership from the principal survey

Because exploring how leadership functions are interactively fulfilled by principal and teachers was not a primary focus of the TALIS 2013 study, measuring operational DL patterns was of paramount importance for this study. To have confidence in our measures, they needed to meet three requirements: (a) to include the appropriate stakeholders in the measurement for school leadership, (b) to reflect appropriately the conceptual basis for each of the leadership functions that were distributed in the school, and (c) as composite measures, to be reliable. The following section describes how each of the constructs meets our criteria.

In TALIS 2013, three sets of items inquired into administrators’ perceptions on varying degrees of leadership responsibilities fulfilled by a range of stakeholders. This study focuses on the action taken by principal and teachers.

Identifying leadership functions and verifying internal reliability

First, we used exploratory factor analysis (EFA; Stata 14) without the assumption of any underlying structure among observed variables to explore how the 17 items included in the TALIS principal survey for school principal and teacher leadership aligned with the school leadership functions. We employed principal-component factor analysis with the Promax (Oblique) rotation, in which the model assumed the underlying factors are correlated. The factor model explained approximately 71.8% of the total variance. The Kaiser–Meyer–Olkin measure of sampling adequacy was .885 and Bartlett’s test is significant (p < .001), so the EFA analysis is appropriate for the data (Fabrigar & Wegener, 2011). The EFA results revealed four factors that have the Eigenvalue larger than one: “developing people by the principal” and “developing people by teachers,” and “managing instruction by the principal” and “managing instruction by teachers.” The Cronbach α for all latent variables was then tested to be larger than .7 with the pooled data, which is a good sign of internal consistency of the construct (Creswell, 2014).

Factor scores

We used the software Mplus 7 to calculate factor scores (managing instruction and developing people by principals and teachers) inferred from the observed variables using confirmatory factor analysis (CFA) with categorical variables. CFA is a measurement model using multivariate regression that describes the relationships between a set of observed variables and a latent variable (Brown, 2015). We used CFA over EFA factor scores because researchers argue CFA has advantages over EFA, including distinguishing error components from a factor, conducting measurement at the factor level, including multiple fit indices, and allowing for flexibility in building a model (Bollen, 1989); the last feature also gave us the chance to verify the model of leadership functions fulfilled by the specific group one more time.

The TALIS 2013 used a two-stage cluster sampling design that collected the complex survey data. Therefore, the CFA model was specified as TYPE=COMPLEX for the ANALYSIS command in conjunction with the “Stratification=IDCNTRY” option, which indicated the schools were sampled within the country in order to take into account of the nonindependence of the school data for the stratified sampling effect. The observed variables were binary variables so the variables were indicated specifically as CATEGORICAL, which enabled the model to use weighted least squares mean and variance adjusted (WLSMV) estimator. WLSMV has been favored over robust maximum likelihood (Li, 2014) and Bayesian estimators (Liang & Yang, 2014) because researchers found it is less biased with categorical variables. In addition, Theta parameterization was used for this study, which allowed the model to obtain information on unexplained variance in the observed variables (Muthén & Satorra, 1995). Finally, the model used the school level weight, which is essential to compensate unequal selection probabilities of the sample and different response rates in the 2013 TALIS.

The factor scores were saved after the WLSMV estimation for each model. Factor scores account for varied coefficient strength between the latent construct and the observed variables. The factor score for individual principals was computed from the mean vector of $y$ variables μ_y, the factor-score coefficient matrix $C$ , the vector of intercepts, the vector of observed variables, and the matrix of factor loadings multiplied by the mean vector. The scores that are produced have a mean of 0 and a standard deviation of 1, and are uncorrelated.

{\hat{η}}_{i} = μ_{y} + C (υ_{i} - τ_{y} - Λ_{y} μ_{y})

The measurement of organizational decision making by the principal and teachers used the standardized score for TC2G22D and TC2G22A, which queried whether the principal made decisions on their own or schools provided staff with opportunities to actively participate in decision making. Summary details for the Developing People, Managing Instruction, and Organizational Decision Making variables are in Table 1.

Table 1.

Summary Details for Distributed Leadership Variables.

Leadership functions	Stakeholder	Survey questions	Coding	Factor loading	α
Developing People	Lead by principals	TC2G28A2 Direct observation of teaching	Yes 1	.620	.727
		TC2G28B2 Managing student surveys for teaching	No 0	.600
		TC2G28C2 Assessments of teachers’ knowledge		.603
		TC2G28D2 Analysis of students’ test scores		.626
		TC2G28E2 Managing teachers’ self-assessments		.718
		TC2G28F2 Discussion about feedback by parents		.696
	Lead by teachers	TC2G28A5 Direct observation of teaching	Yes 1	.628	.716
		TC2G28B5 Managing student surveys for teaching	No 0	.568
		TC2G28C5 Assessments of teachers’ knowledge		.697
		TC2G28D5 Analysis of students’ test scores		.660
		TC2G28E5 Teachers’ self-assessments		.703
		TC2G28F5 Discussion about feedback by parents		.645
Managing Instruction	Lead by principals	TC2G18F1 Student disciplinary policies	Yes 1	.717	.791
		TC2G18G1 Student assessment policies	No 0	.720
		TC2G18I1 Choosing learning materials		.848
		TC2G18J1 Determining course content		.766
		TC2G18K1 Deciding courses offered		.650
	Lead by teachers	TC2G18F3 Student disciplinary policies	Yes 1	.739	.770
		TC2G18G3 Student assessment policies	No 0	.765
		TC2G18I3 Choosing learning materials		.821
		TC2G18J3 Determining course content		.597
		TC2G18K3 Deciding courses offered		.579
Organization Decision Making	Lead by principals	TC2G22D, organizational decision making by principals	Strongly disagree 1Disagree 2Agree 3Strongly agree 4
Organization Decision Making	Lead by teachers	TC2G22A, organizational decision making by teachers	Strongly disagree 1Disagree 2Agree 3Strongly agree 4

Model fit indexes for distributed leadership variable construct

The model fit indices of each DL variable are in Table 2. Generally, the models met conventions of good model fit to the data: comparative fit index ≥ .90, Tucker–Lewis index ≥ .90, root mean square error of approximation ≤ .08, Chi square difference test is not used for this study because researchers argue Chi square difference test is sensitive to the sample size and the complexity of the model (Hu & Bentler, 1999).

Table 2.

Model Fit Indices and Descriptive Statistics for Each Distributed Leadership Factor.

Latent variable	CFI	TLI	RMSEA	M	SD	SE	Max	Min
Developing People: Principals	.964	.946	.029	−0.068	0.784	0.011	1.060	−1.555
Developing People: Teachers	.978	.968	.028	0.264	0.758	0.010	2.045	−0.566
Managing Instruction: Principals	.969	.948	.054	−0.010	0.734	0.009	0.975	−1.152
Managing Instruction: Teachers	.969	.949	.042	0.177	0.759	0.010	1.362	−0.892
Organizational DM: Principals				0.000	1.000	0.013	2.265	−1.319
Organizational DM: Teachers				0.000	1.000	0.013	1.281	−4.465

Note. Organizational decision making by principals and teachers are two single items (TC2G22D and TC2G22A) so the model fit indices are not applicable, the mean is the standardized mean. CFI = comparative fit index; TLI = Tucker–Lewis index; RMSEA = root mean square error of approximation; DM = decision making.

Correlation and scatterplots

The first research goal is to reveal the interactive pattern of DL among principal and teachers for each of the three leadership functions, and also generate a scatterplot in order to explore the position of the combined extent of principal and teacher leadership for each country. Correlation analysis revealed the relationship between principal and teacher leadership. We used Pearson correlation for Developing People and Managing Instruction (both continuous variables) and Spearman rank order correlation for Organizational Decision Making (ordinal variables).

The analysis using DL factor scores is at the country level using the country mean of the corresponding variables. Resulting scatterplots display the DL pattern in a two by two matrix according to the dimensions of principal and teacher leadership. We overlaid a quadrant on the scatterplot with the intersection of axes placed at the mean as 0. In this way, we situated principal leadership relative to teacher leadership, illustrating the interaction of the two as central to the concept of DL operation, which is similar as one previous study for the interaction between two different leadership styles (Marks & Printy, 2003). The quadrant shows a country’s placement relative to the other 31 countries as either low or high on the principal leadership and teacher leadership dimensions (refer to Figure 1). The quadrant labels reflect the authority dynamics in the specific quadrant.

Stage 1, Research Question 2: Country Educational Policy and DL Patterns

We have proposed that the country context might be considered part of the situation that shapes distributed leadership. To demonstrate this, and to explain how a country context might influence the country’s patterns of DL shown in the previous scatterplots, we include evidence from a document analysis of education policy as an observable component of country context. We drew on a series of Education Policy Outlook Country Profiles prepared by researchers at OECD. The profiles, prepared between 2013 and 2018 according to a standard template, incorporate findings from OECD surveys, including TALIS 2013 and a series of reports from OECD thematic studies, such as early childhood education or school leadership. Our analysis draws on three specific sections of the profiles, School Improvement, Evaluation and Assessment, and Governance, because these provide insight into policy related to the three complex leadership tasks highlighted in this study. We discuss only policies current when the TALIS survey was administered (OECD, 2014a).

Stage 2, Research Question 3: Distributed Leadership and School Culture

For the dependent variable of the second phase of this study, we selected three items from the TALIS teacher survey that reported on the kind of school culture that we argue is conducive to DL. High responses indicate teachers have structured processes and opportunities to engage with other adults to make decisions for the school, take a school-wide orientation and care for all students rather than focusing only on the students in their classroom. Words such as “active participation,” “shared responsibility,” and “mutual support” suggest the flattened hierarchy characteristic of DL. The factor loadings and the internal reliability alpha are listed in Table 3.

Table 3.

Summary Details for Culture Conducive to Distributed Leadership.

Survey items	Coding	Factor loading	Cronbach’s α
TT2G44A: This school provides staff with opportunities to actively participate in school decisions.	Strongly disagree 1Disagree 2Agree 3Strongly agree 4	.832	.824
TT2G44D: This school has a culture of shared responsibility for school issues.		.879
TT2G44E: There is a collaborative school culture which is characterized by mutual support.		.871

To answer the third research question, two-level HLM for each country reveal the relationships between principal leadership and teacher leadership for each leadership function and teacher perceived school culture conducive to DL. The HLM was applied using the software Stata 14. In the model, Level 1 represents teachers, and Level 2 represents schools/principals. Specifically, the teacher-level model (Model 1) is as follows:

\begin{array}{l} {(T e a c h e r p e r c e i v e d c u l t u r e)}_{i j} = β_{0 j} + r_{i j} \\ r_{i j} ~ N (0, σ^{2}) \end{array}

(1)

Where ${(T e a c h e r p e r c e i v e d c u l t u r e)}_{i j}$ is the value of individual teacher reported school culture for teacher i in school j, $β_{0 j}$ is the average score of school culture in school j, $r_{i j}$ is the teacher-level random effect, and $σ^{2}$ is the variability within schools.

The corresponding school-level model (Model 1) is as follows:

\begin{array}{l} β_{0 j} = γ_{00} + γ_{01} D P P + γ_{02} D P T + γ_{03} M I P \\ + γ_{04} M I T + γ_{05} O D P + γ_{06} O D T + μ_{0 j} \\ μ_{0 j} ~ N (0, τ_{00}) \end{array}

(2)

$γ_{01}$ to $γ_{06} a r e$ coefficients of three leadership functions (DP: Developing People; MI: Managing Instruction; OD: Organizational Decision Making) fulfilled by either the Principal (P) or Teachers (T). These items as independent variables show the separate effect of each source of leadership.

Stage 2, Research Question 4: The Interactive Effect of Principal and Teacher Leadership

For the fourth research question, we added the interactive effects of principal and teacher leadership for each leadership function to the previous two-level HLMs, which reported the single source leadership in relation to school culture. The interactive effects of developing people and managing instruction by principals and teachers were formed by multiplying the continuous factor scores for principal and teacher leadership (Jaccard et al., 2003). The measures of leadership for organizational decision making derive from categorical variables with four scales; these were recoded to dummy variable where 1 (strongly disagree) and 2 (disagree) were coded as 0, and 3 (agree) and 4 (strongly agree) were coded as 1. The interaction between two dummy variables then generated four conditions as low principal leadership (P0) and low teacher leadership (T0), high principal leadership (P1) and low teacher leadership (T0); high principal leadership (P1) and high teacher leadership (T1); low principal leadership (P0) and high teacher leadership (T1). These approaches to investigating an interactive effect of principals and teachers leading together adds to our understanding of DL.

\begin{array}{l} β_{0 j} = γ_{00} + γ_{01} D P P + γ_{02} D P T + γ_{03} M I P + γ_{04} M I T + γ_{05} O D P + γ_{06} O D T \\ + γ_{07} D P P * D P T + γ_{08} M I P * M I T + γ_{09} O D P * O D T + μ_{0 j} \\ μ_{0 j} ~ N (0, τ_{00}) \end{array}

(3)

Stage 2, Meta-Analysis for the Third and Fourth Research Questions

After running two-level HLMs for each of the 32 countries, we employed meta-analysis to synthesize the coefficient effects, in order to answer how principal and teacher leadership are respectively, and interactively related to the culture supporting DL across countries. We used the meta-analysis for two reasons: (a) educational learship practices vary significantly in different countries, so running one model for pooled data of the 32 countries would have missed an essential consideration of contextual variations (Bush, 2018); and (b) the effect sizes and the significance levels are different across countries, so the conclusion cannot be drawn from only one or some countries.

Meta-analysis is a statistical approach to synthesize results of findings from relevant studies, which is more rigorous than the simple arithmetic average of the results (Littell et al., 2008). And for this study, meta-analysis combined the results from the 32 countries, so the conclusion is statistically stronger, due to greater diversity among subjects, and accumulated effects (Lipsey & Wilson, 2001). Meta-analysis weights the results based on the sample size, in which the country with larger samples has more weighs. Furthermore, meta-analysis assigns weights to each country’s results using the inverse of the error variance (i.e., 1/variance), which provides a generic approach to combine estimates. Studies with a low variance or more precise estimate are generally assigned more weights (Littell et al., 2008).

The statistical techniques used in meta-analysis are classified into two models (Hedges & Vevea, 1998) as “fixed effects” or “random effects.” The “fixed effects” model assumes that the variability is exclusively due to a random reason (Borenstein et al., 2010). The “random effects” model, on the other hand, assumes there is a different underlying effect for each individual study (Egger et al., 1997; Hedges & Vevea, 1998; Littell et al., 2008). DerSimonian–Laird random-effects model is one of the most commonly used models, which assumes heterogeneity among the studies; specifically, it assumes that the true effect can be different for each study (DerSimonian & Kacker, 2007). For this study, DerSimonian–Laird random-effects model was used.

The classical measure of heterogeneity is Cochran’s Q, which is calculated as the weighted sum of squared differences between individual study effects and the pooled effect across studies. Q is criticized as a test of heterogeneity for too much power if the number of studies is large (Higgins & Thompson, 2002). The I² statistic describes the percentage of variation across studies that is due to heterogeneity rather than random chance (Egger et al., 1997; Higgins & Thompson, 2002). I² = 100% × (Q − df)/Q. Unlike Q, I² does not inherently depend on the number of studies. Stata 14, with the third-party program “Mettan” were used for the meta-analysis (Egger et al., 1997; Littell et al., 2008). The command “metaan effectsize SE, dl” was used to calculate the overall effect size with DerSimonian–Laird random effects.

Findings

Stage 1 Results: Distributed Leadership Patterns Within 32 Countries

The combinations of the two groups’ interactions in each of the three leadership functions generated three scatterplots to demonstrate the pattern of interactive leadership distribution.

Distributed Leadership Pattern: Developing People

Figure 2 reveals the leadership for developing people fulfilled by principals and teacher leaders, respectively, within each country. The distribution of 32 countries demonstrates significant variation, and there is no significant correlation found between the two groups’ leadership. Administrators’ reports place 17 countries in the Quadrant 4, where both principals and teacher leaders work interactively to develop human capital; primary responsibility is more likely taken by teacher leaders in 8 countries/regions (Quadrant 3); and a similar response indicated principals have primary responsibility for improving human capital in 2 countries (Quadrant 2). Five countries, then, have very low levels of school authority in efforts toward developing people (Quadrant 1). The findings of 25 countries in quadrant three and four resonate with the previous research that nonposition holders are often actively involved in developing people.

Figure 2.

Distributed leadership pattern for developing people.

Distributed Leadership Pattern: Managing Instruction

Figure 3 shows DL patterns for managing instruction. The scatterplot demonstrates quite a bit of variation across 32 countries, while the leadership in instructional management from principals and teacher leaders is significantly and positively related (r = .412, p < .001). The largest number of countries (16) are in Quadrant 4, indicating that these countries have principals who work proactively with teacher leaders. Seven countries are within Quadrant 3 that has teacher leaders be more likely responsible. Number of countries (7) in Quadrant 1, showing that these countries report low relative levels of leadership of any sort for instructional concerns within schools. Quadrant 2 has only two countries indicating that leadership contributed primarily by the principal is not predominant for the sample.

Figure 3.

Distributed leadership pattern for managing instruction.

Distributed Leadership Pattern in Organizational Decision Making

Figure 4 displays the DL patterns in organizational decision making. There is a significant negative relationship between principal and teacher leadership in organizational decision making ( $ρ$ [Spearman’s rho] = −.193, p < .001). Among the 32 countries, we found that only 5 countries reported that principals and teachers work interactively for decision making in their schools (Quadrant 4). Respondents in eight countries indicate that their schools are likely to have teacher leaders make decisions. The 12 countries that report organizational decision making in Quadrant 2 experience a traditional authority structure where principals have the major responsibility. The last group, within Quadrant 1, consists of seven countries that have low level of authority within schools in making decisions.

Figure 4.

Distributed leadership pattern for decision making.

As a summary comment about the leadership distribution for the tasks just discussed, the largest number of countries landed in Quadrant 4 for Developing People (17) and Managing Instruction (16), indicating that a large proportion of administrator respondents in each case perceived that both principals and teacher leaders held responsibilities. For organizational decision making, however, the number of countries (12) designated as Hierarchical (primarily formal leadership) is more than all other leadership patterns, which reveals in most countries, organizational decision making is the province of the principal.

Stage 1 Results: Educational Policy and Distributed Leadership

The OECD policy profiles offer insight into country conditions that explain the scatterplot patterns. We discuss how education policy, within the country context, might enable or constrain the development of distributed leadership and we include an illustrative display of policy influence in two countries (Table 4). A full cross-country comparison of this topic will follow in a subsequent publication.

Table 4.

Illustrative Sample: Country Policy Context by Leadership Function.

Developing People (DP)	Instructional Management (IM)	Organizational Decisions (OD)
Denmark (DP = 4, IM = 4, OD = 4)
• Recent policy to grow participation in professional development• Teacher appraisal is voluntary: self-appraisal plus principal feedback; 95% participate• Recent policy to improve national evaluation framework	• Recent policy to increase principals’ instructional leadership• Schools have increased autonomy but greater accountability demands• Teachers work in teams with support of special advisors	• Ministry of Education sets direction, national frameworks• Municipality responsible for national targets and operating schools within national frameworks• Schools run by board of governors with parents, teachers and student members
Japan (DP = 4, IM = 2, OD = 2)
• National data system and institute for policy research• Assessments of student learning guide teaching improvement• Schools conduct self-evaluation; parents and communities conduct external evaluation• Robust set of surveys provide broad view of influences on student learning	• Ministry of Education sets national curriculum standards• National education frameworks after earthquake set new directions for schooling (Zest for Life, Global leaders, project-based learning, student ownership)• Teachers have responsibility for how the curriculum is taught• Recent encouragement of bottom up innovation)• Teachers report being unprepared to teach	• Ministry of Education sets official requirements for schools• Municipalities responsible for operation of schools, appoint Board of Education• School board works with local governments to set policy to improve local systems

The Developing People construct is about teachers having the opportunity and motivation to improve the school’s instruction as a result of having information about their teaching practice, and the occasion to discuss the information with the principal and other teachers. Our analysis of country profiles indicates that information comes through teacher appraisals, student assessments, targeted assistance providers, and stakeholder surveys. The presence of national data systems and research agencies suggests an availability of additional sources of information. When teachers work in teams or schools have leadership teams or participation in professional development (both principal and teachers) is high, the likelihood of the country scoring a 3 or a 4 in Developing People appears greater.

Instructional Management, in this study, deals with occasions and responsibility for making decisions that influence what goes on in classrooms, including what courses are offered and the content of the courses, which materials are used, how students are assessed, and how students are expected to behave. A critical factor is the location in the country system at which decisions about teaching are made. The country may have a highly centralized system where the ministry or regional authority make all decisions, or it may have a centralized framework for education and decentralized plans for development and implementation. The extent to which educators are professionalized seems consequential, varying by required education, certification, induction and mentoring, and hiring practices. A number of profiles report increasing policy interest in instructional leadership by principals and enhanced accountability for student results, often accompanied by greater autonomy (and thus, the likelihood of higher DL, a score of 3 or 4).

The Organizational Decision-making measure tracks whether or not principals and teachers make such decisions. The most consequential policies assign decision-making authority, generally across three levels (the country ministry, the region or municipality, and the school). At the local level, the composition of school teams, school boards, and school councils have implications for DL, particularly whether principals and teachers are members.

The illustrative sample countries, Denmark and Japan, in Table 4, include reference to many factors just discussed. There are other details, however, that demonstrate the need for further information (and which is included in the profile). For instance, “A Denmark that Stands Together” (2011) laid out the country’s key education priorities, such as stronger early childhood education and reform of the primary and lower secondary schools in collaboration with teachers and parents (OECD, 2014a). Specific references to this document suggest an inclusive approach to education and the operation of schools, not only for students but also for all educators. A second point of interest is that teacher appraisal in Denmark is voluntary. The process involves a self-appraisal by the teacher and feedback from the principal. In this country, nearly all teachers participate. The policy conditions in Denmark line up in the three leadership tasks to suggest why the scatterplots place the country in the Interactive Leadership category, where principals and teachers work together.

Japan is placed in Quadrant 4, Interactive Leadership, only for Developing People. The policy guidance for schools requires the collection of data about students, teachers, and schools in multiple forms and from many stakeholders, and the principal indicates he or she participates in these activities equally to teachers. That the country has a national data system and shares research with schools and educators suggests access to valuable guidance at a higher grain size than the school. Principal leadership, however, is most consequential for Instructional Management and Organizational Decision Making, as each is placed in Quadrant 2. After the earthquake in 2011, the OECD Tohoku School Project was a small effort to promote innovation in order to foster resilience, creativity, and 21st-century skills, and these ideas then spread more broadly. New instructional methods are gaining favor because Japan is challenged by assuring students are ready for global pressure and to enter modern workplaces. Teachers, traditionally, organize instruction, but in 2013 they expressed being ill-prepared to do so. One can imagine that principals, as the responsible leaders, might exert hierarchical pressure in this context, to earn placement in Quadrant 2. Under Organizational Decision Making, teachers are not given a place at the local decision-making table.

Stage Two Results: Distributed Leadership and School Culture

The third research question inquires whether the administrator-reported distributed leadership fulfilled by principals and teachers, respectively, is related to teacher-reported culture conducive to DL. The results from the 32 independent HLMs (not included due to space limitations, but available from the corresponding author) suggested there are considerable variations existing among countries especially for organizational decision makings based on the heterogeneity test (please check the Cochrane Q and I² in Table 5); this test confirms the importance of considering country contextual variations in studying educational leadership or leadership in general.

Table 5.

Meta-analysis Results for Distributed Leadership and Teacher Reported School Culture.

Dependent variable: Teacher’s perceptions of culture that supports DL
	DPP	DPT	MIP	MIT	ODP	ODT
Overall effect (SE)	0.010* (0.004)	0.012* (0.005)	0.004 (0.006)	0.008 (0.005)	−0.008 (0.007)	0.062* (0.027)
N	32	32	32	32	32	32
Q	21.27	27.85	53.23	39.17	33.41	260.36
p	.905	.629	.008	.149	.351	.001
I ²	0.01	0.01	41.76	20.85	7.21	89.25

Note. Q is the Cochran’s Q, p is the heterogeneity significance test. DL = distributed leadership; DPP = developing people by principals; DPT = developing people by teachers; MIP = managing instruction by principals; MIT = managing instruction by teachers; ODP = organizational decision by principals; ODT = organizational decision by teachers.

p < .05. **p < .01. ***p < .001.

The meta-analysis results are listed in Table 5. Results indicate that teachers’ taking responsibility for developing people (β = .012, p < .05) and making organizational decisions (β = .062, p < .05) are both positively related to supportive school culture for DL. Meanwhile, principal leadership for developing people (β = .010, p < .05) is significantly related to the teacher-reported culture measure.

Stage Two Results: Interactive Effect and School Culture

We were further interested in whether the interactive effect, testing principal and teacher leadership working together, is related to teacher reports of supportive school culture for DL. The results in Table 6 show that, when adding the interactive leadership term, teacher leadership for Developing People (β = .011, p < .01) and for Organizational Decision Making (β = .061, p < .05) remain positively related to DL conducive school culture. Three interactions are significantly related to school culture. Principals and teachers together for Developing People (β = .015, p < .01) and Managing Instruction (β = .018, p < .05) have positive effects. The Organizational Decision-Making result, with a significant negative coefficient for Quadrant 1 (β = −.083, p < .05) indicates that hierarchical decisions by the principal do not support a culture where DL can take hold.

Table 6.

Meta-analysis Results for Interactive Effect and School Culture.

Independent variable: Teacher’s perceptions of culture supportive of DL
	DPP	DPT	MIP	MIT	ODP	ODT	DPPxDPT	MIPxMIT	OD_P1xT0	OD_P1xT1	OD_P0xT1
Overall effect (SE)	0.007 (0.004)	0.011** (0.005)	0.009 (0.008)	0.005 (0.005)	−0.026 (0.020)	0.061* (0.028)	0.015** (0.005)	0.018** (0.006)	−0.083* (0.038)	0.002 (0.039)	0.012 (0.026)
N	32	32	32	32	32	32	32	32	32	32	32
Q	30.92	27.82	67.74	40.84	141.82	229.47	34.88	54.04	440.59	403.12	608.91
p	.470	.631	.001	.111	.001	.001	.289	.006	.001	.001	.001
I ²	0.01	0.01	54.24	24.10	78.14	87.80	11.12	42.63	95.69	94.54	94.91

Note. Q is the Cochran’s Q, p is the heterogeneity significance test. SE = standard error; DL = distributed leadership;DPPXDPT = interaction of developing people led by principals and led by teachers; MIPXMIT = interaction of managing instruction led by principals and led by teachers; OD_P1XT0 = Organizational decision-making high principal leadership and low teacher leadership; OD_P1XT1 = Organizational decision-making high principal leadership and high teacher leadership; OD_P0XT1 = Organizational decision-making low principal leadership and high teacher leadership.

p < .05. **p < .01. ***p < .001.

Discussion

Distributed leadership is regarded as one promising way to improve school organizational capacity through involving teachers into decision making, and the evidence has been accumulating to support a positive relationship between distributed leadership and school outcomes (Hulpia et al., 2012; Liu & Printy, 2017; Malloy & Leithwood, 2017). The study extends current research and makes important contributions in the following areas.

Leadership Participation by Function Across Countries

The results reported in this study offer initial evidence that school leadership distribution is understood and practiced differently across countries. Within the sample of 32 countries in the TALIS 2013 data set, we found much variation of country placement on the two by two matrix for each leadership function. Furthermore, at the country level, the correlation of principals’ reports of their own leadership to teacher leadership varies by leadership function. The relationship is not significant for Developing People, is positive for Managing Instruction, and is negative for Organizational Decision Making. These findings further suggest that principals think differently about who participates in leadership activity according to the situation. Principals report that teachers more frequently engage with them in a set of activities related to assessing and enhancing teachers’ skills and in a set of activities related to policies and decisions relevant to instruction. Similar engagement is much less likely when it comes to making decisions generally for the school organization; in this situation, principals most often retain hierarchical authority.

The document analysis of educational policy from the OECD Educational Policy Outlook profiles provided exciting support to consider the country context when studying DL internationally. As demonstrated briefly in the findings, the country policies related to professional development, teacher and student appraisal, school evaluation, initiatives in instructional leadership, accountability demands, encouragement in team building and mentoring, and authority for various leadership responsibilities within an individual country could either encourage or constrain distributed leadership. Further examinations using the educational policy profile to study distributed leadership is needed. These results add preliminary confirmatory evidence that country is a critical level of context for distributed leadership.

Distributed Leadership and School Culture

Our examination of DL is limited by the data available in TALIS 2013, since the questions relevant to DL were asked only on the principal survey. To provide some indication of teachers’ perceptions, we developed a measure of school culture, specifically one that would empower or be conducive to teachers’ engagement in leadership activity. If principals report teachers lead in important leadership roles and if teachers report that their schools have an open and empowering culture, then we have some faith that principal reports are fairly accurate.

With HLMs, we explored this relationship for each of the 32 participating countries and conducted meta-analysis in order to produce general findings. The meta-analysis gives global evidence about the extent to which teachers’ reports of whether their school has a culture of broad participation aligns with the principals’ reports of who actually takes part in leadership activity within a given situation. Principal and teacher accounts appear to cohere when it comes to improve teachers’ instructional quality using a variety of approaches. Principal and teacher accounts do not seem to make any difference when it comes to managing instruction, particularly for assessment policies, choosing learning materials, and determining what courses and content are offered. For organizational decision making, the evidence is split. What the principals says about his or her decision involvement is unrelated to how teachers perceive their culture. Teacher-centered measures, however, do seem to agree, that when the principal reports primary decision-making authority, teachers do not report school culture that supports their engagement.

Interactive Effects on School Culture

The theory and research we review argues the importance of interaction among leaders, and suggests that distributed leadership emerges from this interaction. We include interaction terms to learn more about the situations in which principals and teachers add more value by leading interactively.

For Developing People, the interaction term is significant, suggesting a synergistic benefit when principals and teachers work together to make sense of varied sources of information in order to improve teachers’ professional skills. This finding adds evidence of an added value of interaction, beyond the positive effects of either teacher leadership or principal leadership alone on teachers’ reports of a culture conducive to DL. A similar effect results for Managing Instruction; principals and teachers working together to manage instruction contributes positively to school culture, even though contributions made by principal alone or teacher alone is insignificant, discussed earlier.

Due to the fact that the Organizational Decision-Making variable is categorical, we have a series of indicator variables, with the contrast representing Quadrant 1, Low leadership to explore the interactive influence of principals and teachers. We find that Quadrant 2, Hierarchical, shows a significant negative relationship to teachers reports of school culture that would support DL, though in the full meta-analysis model, teacher’s participation in Organizational Decision Making remains positive and significant. The indicators for Quadrant 3, Lateral and Quadrant 4, Interactive, are both insignificant relative to Quadrant 1.

Together, the positive result of primarily teacher involvement in organizational decisions and the negative result of primarily principal involvement in organizational decisions point to a specific situation when interactive leadership might be challenging to implement.

Meta-Analysis as an Innovative Approach for Multicountry Data

Methodologically, we contribute to the literature for evaluating large-scale and multicountry data with the meta-analysis approach. For multicountry data, controlling the country-level variations is imperative because country contexts are so different as proved by previous studies and this research. However, the 2013 TALIS (as well as PISA and other multicountry studies) did not include any country-level variables, therefore, researchers used different approaches to take into account the variance in the outcome that is related to country differences. Comparative studies have used multilevel models by adding the country-level random effect (Liu & Printy, 2017), or including country dummy variables as fixed effects to account for country variations (Chudgar et al., 2013). But both approaches would increase the estimation variance due to increased model complexity (Shatzer et al., 2014). Meanwhile, analyzing pooled data would not provide comprehensive comparisons among the participating countries, which misses the opportunity to take advantage of multicountry data for a more generalizable conclusion. Our approach of treating each country as an independent study, and synthesizing results from all countries is an innovative and statistically robust approach.

Conclusions

Increasingly, researchers regard DL as one promising way to improve school organizational capacity by engaging both principals and teachers in accomplishing tasks and in making decisions about critical school issues. Interest in DL is now a concept of international interest (Hargreaves & Shirley, 2012; Harris, 2011). The current study reports the country-level status of DL for three specific leadership tasks, using reports of school principals in the 2013 TALIS. By intersecting measures of principal leadership and teacher leadership, and displaying that relationship on scatterplots, we illustrate the relative placement of countries for DL in the sample of 32 countries for which we have appropriate data. The placement of countries within quadrants describes the likely authority dynamics predominant in the country for the specific tasks. We then draw on a document analysis of a set of Country Policy Outlook Profiles, with preliminary evidence about country education policies to support the resulting DL patterns.

As the policy profiles point out, there are generally three levels of authority with control over schools: the national level, a state, regional or municipal level, and the local-school level, and the amount of influence each level can exert varies by policy (The United States, without significant national control, has state level, district or operator level, and school level). Furthermore, upper secondary and tertiary education are often characterized by greater local control than is lower secondary education, which is the context for this work. Unobserved, in our study, though also likely impactful, is the cultural context of countries, including traditions and history (Hofstede, 1984; House et al., 2013).

As researchers, we regard DL, as a general proposition, to be beneficial for schools, teachers, and students if it is supported by a conducive school culture, and it is implemented with care and concern for those engaged in its tasks. Our awareness that “context matters” for leadership is stronger as a result of the findings reported here. Even more, we expect that DL of principals and teachers, as it exists or develops around the world, could look very different than the Western experience primarily represented in the research. Meanwhile, we acknowledge the powerful affect educational policy could have on the development and implementation of DL in schools; if the system supports information exchange, team building, boundary-spinning initiatives, it is likely to encourage DL.

Using the 2013 TALIS survey data, the study looks at DL from a global perspective, country by country. To our knowledge, our line of work is among the first to systematically investigate DL at the country level in a comparative way, with sufficient data from 32 countries using robust statistical approaches. This research calls for continuing work to better understand DL in countries around the world. As indicated earlier, our next stage of work will include a full discussion of global education policy and a replication of the DL classification using the recently published TALIS 2018 data. Individual researchers with high familiarity of particular countries can makes sense of the DL patterns in our results, based on deep understanding of culture or on country-specific policy initiatives or educational movements. Comparative studies of DL globally need to examine a range of other core leadership tasks that need to be fulfilled in schools. More important, comparative work is needed to explore the relationship of DL to valued outcomes, including teaching quality, adoption of innovative instructional methods, teacher satisfaction, well-being, self-efficacy, collective efficacy, and student achievement, persistence, and well-being. These possibilities should encourage additional research about the extent of and operation of DL in discrete contexts and specific testing in international contexts to determine whether and how DL actually makes a difference to educational or administrative processes.

Footnotes

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.

ORCID iDs

Susan Printy

Yan Liu

Author Biographies

Susan Printy recently retired as an associate professor of K-12 Educational Administration from Michigan State University. Her research explores the ways in which teachers engage in social learning within teacher communities and the organizational and leadership conditions that support that work. She seeks to better understand the leadership relationships between principals and teachers in order to provide educators with guidance for engaging in distributed or shared leadership as necessary for school improvement, for building a skilled and satisfied faculty of teachers, and for better student achievement.

Yan Liu is an assistant professor at Central Connecticut State University. Her research centers around the operational features of Distributed Leadership through quantitative investigations of the dynamic interactions among formal leaders, informal leaders, school leadership functions, and contexts.

References

A Denmark that stands together. (2011). http://stm.dk/multimedia/Regeringsgrundlag_uk_2011.pdf

Angelle

P. S.

(2010). An organizational perspective of distributed leadership: A portrait of a middle school. RMLE Online, 33(5), 1–16. https://doi.org/10.1080/19404476.2010.11462068

Bellibas

M. S.

Liu

(2018). The effects of principals’ perceived instructional and distributed leadership practices on their perceptions of school climate. International Journal of Leadership in Education, 21(2), 226–244. https://doi.org/10.1080/13603124.2016.1147608

Bollen

K. A.

(1989). A new incremental fit index for general structural equation models. Sociological Methods & Research, 17(3), 303–316. https://doi.org/10.1177/0049124189017003004

Borenstein

Hedges

L. V.

Higgins

J. P. T.

Rothstein

H. R.

(2010). A basic introduction to fixed-effect and random-effects models for meta-analysis. Research Synthesis Methods, 1(2), 97–111. https://doi.org/10.1002/jrsm.12

Brown

T. A.

(2015). Confirmatory factor analysis for applied research (2nd ed.). Guilford Press.

Bush

(2018). Leadership and context: Why one-size does not fit all. Educational Management Administration & Leadership, 46(1), 3–4. https://doi.org/10.1177/1741143217739543

Bush

Glover

(2012). Distributed leadership in action: Leading high-performing leadership teams in English schools. School Leadership & Management, 32(1), 21–36. https://doi.org/10.1080/13632434.2011.642354

Bush

Glover

Sood

(2006). Black and minority ethnic leaders in England: A portrait. School Leadership & Management, 26(3), 289–305. https://doi.org/10.1080/13632430600737140

10.

Camburn

E. M.

Rowan

Taylor

J. E.

(2003). Distributed leadership in schools: The case of elementary schools adopting comprehensive school reform models. Educational Evaluation and Policy Analysis, 25(4), 347–373. https://doi.org/10.3102/01623737025004347

11.

Chudgar

Luschei

T. F.

Zhou

(2013). Science and mathematics achievement and the importance of classroom composition: Multicountry analysis using TIMSS 2007. American Journal of Education, 119(2), 295–316. https://doi.org/10.1086/668764

12.

Copland

M. A.

(2003). Leadership of inquiry: Building and sustaining capacity for school improvement. Educational Evaluation and Policy Analysis, 25(4), 375–395. https://doi.org/10.3102/01623737025004375

13.

Creswell

J. W.

(2014). Research design: Qualitative, quantitative, and mixed methods approaches. Sage.

14.

Cyert

R. M.

March

J. G.

(2006). A behavioral theory of the firm. In Miner

J. B.

(Ed.), Organizational behavior 2: Essential theories of process and structure (Vol. 2, pp. 60–87). M.E. Sharpe.

15.

Day

Leithwood

K. A.

(2007). Successful principal leadership in times of change: An international perspective (Vol. 5). Springer. https://doi.org/10.1007/1-4020-5516-1

16.

DerSimonian

Kacker

(2007). Random-effects model for meta-analysis of clinical trials: An update. Contemporary Clinical Trials, 28(2), 105–114. https://doi.org/10.1016/j.cct.2006.04.004

17.

Du Plessis

Eberlein

(2018). The role of heads of department in the professional development of educators: A distributed leadership perspective. Africa Education Review, 15(1), 1–19. https://doi.org/10.1080/18146627.2016.1224583

18.

Egger

Smith

G. D.

Phillips

A. N.

(1997). Meta-analysis: principles and procedures. British Medical Journal, 315(7121), 1533-153 https://doi.org/10.1136/bmj.315.7121.1533

19.

Fabrigar

L. R.

Wegener

D. T.

(2011). Exploratory factor analysis. Oxford University Press. https://doi.org/10.1093/acprof:osobl/9780199734177.001.0001

20.

Goddard

R. D.

Goddard

Y. L.

Sook Kim

Miller

(2015). A theoretical and empirical analysis of the roles of instructional leadership, teacher collaboration, and collective efficacy beliefs in support of student learning. American Journal of Education, 121(4), 501–530. https://doi.org/10.1086/681925

21.

Gronn

(2002). Distributed leadership as a unit of analysis. Leadership Quarterly, 13(4), 423–451. https://doi.org/10.1016/S1048-9843(02)00120-0

22.

Hallinger

Murphy

(1985). Assessing the instructional management behavior of principals. Elementary School Journal, 86(2), 217–247. https://doi.org/10.1086/461445

23.

Hargreaves

Fink

(2006). Sustainable leadership. Jossey-Bass.

24.

Hargreaves

Shirley

D. L.

(2012). The global fourth way: The quest for educational excellence. Corwin Press.

25.

Harris

(2005). Distributed leadership. In Davies

(Ed.), The essentials of school leadership (pp. 160–172). Corwin Press/Paul Chapman Publishing.

26.

Harris

(2011). Distributed leadership: Implications for the role of the principal. Journal of Management Development, 31(1), 7–17. https://doi.org/10.1108/02621711211190961

27.

Harris

Muijs

(2005). Improving schools through teacher leadership. Open University Press.

28.

Hartley

(2010). Paradigms: How far does research in distributed leadership “stretch”? Educational Management Administration & Leadership, 38(3), 271–285. https://doi.org/10.1177/1741143209359716

29.

Heck

Hallinger

(2009). Assessing the contribution of distributed leadership to school improvement and growth in math achievement. American Educational Research Journal, 46(3), 659–689. https://doi.org/10.3102/0002831209340042

30.

Hedges

L. V.

Vevea

J. L.

(1998). Fixed- and random-effects models in meta-analysis. Psychological Methods, 3(4), 486–504. https://doi.org/10.1037/1082-989X.3.4.486

31.

Higgins

J. P. T.

Thompson

S. G.

(2002). Quantifying heterogeneity in a meta-analysis. Statistics in Medicine, 21(11), 1539–1558. https://doi.org/10.1002/sim.1186

32.

Hofstede

(1984). Culture’s consequences: International differences in work-related values. Sage.

33.

House

R. J.

Dorfman

P. W.

Javidan

Hanges

P. J.

de Luque

M. F. S.

(2013). Strategic leadership across cultures: GLOBE study of CEO leadership behavior and effectiveness in 24 countries. Sage. https://doi.org/10.4135/9781506374581

34.

Bentler

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1–55. https://doi.org/10.1080/10705519909540118

35.

Hulpia

Devos

Rosseel

Vlerick

(2012). Dimensions of distributed leadership and the impact on teachers’ organizational commitment: A study in secondary education. Journal of Applied Social Psychology, 42(7), 1745–1784. https://doi.org/10.1111/j.1559-1816.2012.00917.x

36.

Hunzicker

(2018). Teacher leadership in professional development schools. Emerald Group Publishing. https://doi.org/10.1108/9781787434035

37.

Ingersoll

R. M.

Sirinides

Dougherty

(2018). Leadership matters: Teachers’ roles in school decision making and school performance. American Educator, 42(1), 13–17.

38.

Jaccard

Turrisi

(2003). Interaction effects in multiple regression. Sage. https://doi.org/10.4135/9781412984522

39.

Law

Galton

Wan

(2010). Distributed curriculum leadership in action: A Hong Kong case study. Educational Management Administration & Leadership, 38(3), 286–303. https://doi.org/10.1177/1741143209359714

40.

Leithwood

K. A.

Jantzi

(2012). Collective leadership: The reality of leadership distribution within the school community. In Leithwood

K. A.

Seashore-Louis

(Eds.), Linking leadership to student learning (pp. 11–24). Jossey-Bass.

41.

Leithwood

K. A.

Mascall

Strauss

Sacks

Memon

Yashkina

(2007). Distributing leadership to make schools smarter: Taking the ego out of the system. Leadership and Policy in Schools, 6(1), 37–67. https://doi.org/10.1080/15700760601091267

42.

C. H.

(2014). The performance of MLR, USLMV, and WLSMV estimation in structural regression models with ordinal variables [Doctoral dissertation, Michigan State University]. https://d.lib.msu.edu/etd/3268

43.

Liang

Yang

(2014). An evaluation of WLSMV and Bayesian methods for confirmatory factor analysis with categorical indicators. International Journal of Quantitative Research in Education, 2(1), 17–38. https://doi.org/10.1504/IJQRE.2014.060972

44.

Lipsey

M. W.

Wilson

D. B.

(2001). Practical meta-analysis. Sage.

45.

Littell

J. H.

Corcoran

Pillai

V. K.

(2008). Systematic reviews and meta-analysis. Oxford University Press. https://doi.org/10.1093/acprof:oso/9780195326543.001.0001

46.

Liu

(2016). How leadership is distributed and how it is associated with teaching quality? A cross-country study with the TALIS 2013 [Doctoral dissertation, Michigan State University]. https://d.lib.msu.edu/etd/4309/datastream/OBJ/download/HOW_LEADERSHIP_IS_DISTRIBUTED_AND_HOW_IT_IS_ASSOCIATED_WITH_TEACHING_QUALITY__A_CROSS-COUNTRY_STUDY_WITH_THE_TALIS_2013.pdf

47.

Liu

(2020). Focusing on the practice of distributed leadership: The international evidence from the 2013 TALIS. Educational Administration Quarterly. Advance online publication. https://doi.org/10.1177/0013161X20907128

48.

Liu

Bellibas

M. S.

Gumus

(2020). The effect of instructional leadership and distributed leadership on teacher self-efficacy and job satisfaction: Mediating roles of supportive school culture and teacher collaboration. Educational Management Administration & Leadership. Advance online publication. https://doi.org/10.1177/1741143220910438

49.

Liu

Printy

S. M.

(2017). Distributed leadership and educator attitudes. In Ransaw

T. S.

Majors

(Eds.), Emerging issues and trends in education (pp. 143–180). Michigan State University Press. https://doi.org/10.14321/j.ctt1qd8zjk.11

50.

Liu

Werblow

(2019). The operation of distributed leadership and the relationship with organizational commitment and job satisfaction of principals and teachers: A multi-level model and meta-analysis using the 2013 TALIS data. International Journal of Educational Research, 96, 41–55. https://doi.org/10.1016/j.ijer.2019.05.005

51.

Malloy

Leithwood

K. A.

(2017). Effects of distributed leadership on school academic press and student achievement. In Leithwood

Sun

Pollock

(Eds.), How school leaders contribute to student success (pp. 69–91). Springer. https://doi.org/10.1007/978-3-319-50980-8_5

52.

Marks

H. M.

Printy

S. M.

(2003). Principal leadership and school performance: An integration of transformational and instructional leadership. Educational Administration Quarterly, 39(3), 370–397. https://doi.org/10.1177/0013161X03253412

53.

May

Supovitz

J. A.

(2011). The scope of principal efforts to improve instruction. Educational Administration Quarterly, 47(2), 332–352. https://doi.org/10.1177/0013161X10383411

54.

McClelland

D. C.

(1967). Achieving society. Free Press.

55.

Moos

Johansson

Day

(2011). How school principals sustain success over time: International perspectives (Vol. 14). Springer. https://doi.org/10.1007/978-94-007-1335-2

56.

Mulder

(1971). Power equalization through participation? Administrative Science Quarterly, 16(1), 31–38. https://doi.org/10.2307/2391284

57.

Murphy

(2017). Creating communities of professionalism: Addressing cultural and structural barriers: The four paths framework. In Leithwood

K. A.

Sun

Pollock

(Eds.), How school leaders contribute to student success (pp. 239–262). Springer International Publishing AG. https://doi.org/10.1007/978-3-319-50980-8_11

58.

Muthén

Satorra

(1995). Complex sample data in structural equation modeling. Sociological Methodology, 25, 267–316. https://doi.org/10.2307/271070

59.

OECD. (2014a). Education policy outlook country profiles. http://www.oecd.org/education/profiles.htm

60.

OECD. (2014b). TALIS 2013 results: An international perspective on teaching and learning. http://www.istruzione.it/allegati/2014/OCSE_TALIS_Rapporto_Internazionale_EN.pdf

61.

OECD. (2014c). TALIS 2013 technical report. https://www.oecd.org/edu/school/TALIS-technical-report-2013.pdf

62.

Printy

S. M.

(2008). Leadership for teacher learning: A community of practice perspective. Educational Administration Quarterly, 44(2), 187–226. https://doi.org/10.1177/0013161X07312958

63.

Shatzer

R. H.

Caldarella

Hallam

P. R.

Brown

B. L.

(2014). Comparing the effects of instructional and transformational leadership on student achievement: Implications for practice. Educational Management Administration & Leadership, 42(4), 445–459. https://doi.org/10.1177/1741143213502192

64.

Spillane

J. P.

(2012). Distributed leadership. John Wiley.

65.

Spillane

J. P.

Anderson

(2019). Negotiating policy meanings in school administrative practice: Practice, professionalism, and high stakes accountability in a shifting policy environment. https://distributedleadership.northwestern.edu/publications/

66.

Spillane

J. P.

Diamond

J. B.

(2007). Distributed leadership in practice. Teachers College, Columbia University.

67.

Spillane

J. P.

Halverson

Diamond

J. B.

(2001). Investigating school leadership practice: A distributed perspective. Educational Researcher, 30(3), 23–28. https://doi.org/10.3102/0013189X030003023

68.

Spillane

J. P.

Shirrell

Hopkins

(2016). Designing and deploying a professional learning community (PLC) organizational routine: Bureaucratic and collegial arrangements in tandem. Les Dossiers des Sciences de l’éducation, 35, 97–122. https://doi.org/10.4000/dse.1283

69.

Sun

Xia

(2018). Teacher-perceived distributed leadership, teacher self-efficacy and job satisfaction: A multilevel SEM approach using the 2013 TALIS data. International Journal of Educational Research, 92, 86–97. https://doi.org/10.1016/j.ijer.2018.09.006

70.

Wenner

J. A.

Campbell

(2017). The theoretical and empirical basis of teacher leadership: A review of the literature. Review of Educational Research, 87(1), 134–171. https://doi.org/10.3102/0034654316653478

71.

Ylimaki

R. M.

Jacobson

S. L.

(2011). US and cross-national policies, practices, and preparation: Implications for successful instructional leadership, organizational learning, and culturally responsive practices (Vol. 12). Springer Science & Business Media.