Factorial Validity and Measurement Invariance of the Principal Leadership Questionnaire in Greek Educational Context

Abstract

The purpose of this study was to investigate the applicability of the Principal Leadership Questionnaire (PLQ) to Greek educational context and to present the most important aspects describing educational leadership in a centralized educational context. It was sought to examine; a) the factorial structure of PLQ b) its invariance across teachers’ levels of education and c) its concurrent validity. Data were collected from 730 Greek primary and secondary school teachers of 77 schools. Teachers were asked to fill in the PLQ and Teacher’s Satisfaction Inventory (TSI). A bi-factor model was selected as the most tenable among five completive PLQ structures to describe teachers’ responses. Moreover, PLQ was found to be invariant across primary and secondary school teachers. Multiple-group analysis results indicate that primary, compared to secondary school teachers, reported more often that their principals behave as a transformational leader. On the contrary, secondary school teachers’ perceptions revealed that their principal implemented practices related to factor “intellectual; stimulation” more often than primary school teachers. Structural equation modeling showed that the general factor of the PLQ significantly and substantially predicted the “principal” facet of teachers’ job satisfaction, providing evidence of concurrent validity of the Greek version of the PLQ. Implications and suggestions for future research and policy are discussed.

Keywords

Psychometrics measurement invariance factorial validity organizational psychology

Introduction

In the past three decades, interest in school leadership has increased. Research suggests that within the school context, successful principals usually implement the same basic leadership practices. For example, leaders could build a shared vision to motivate people to accept group goals and demonstrate high performance expectations. In addition, leaders need to understand staff by providing individualized support and intellectual stimulation to staff members (Leithwood et al., 2008). In 2019, Leithwood et al. published a follow-up paper in which they supported and confirmed the important role of leaders in successful schools by revising basic leadership practices through recent-empirical literature. Furthermore, school leaders could improve teaching and learning by reinforcing teachers’ motivation, commitment, and working conditions and consequently, inspire students’ learning (Berkovich & Eyal, 2017). Collective, different empirical studies, findings show that school leadership is an important variable that has a positively, direct or indirect, effect upon student learning and school improvement (Kyriakides & Creemers, 2012; Sebastian et al., 2017; Sun & Leithwood, 2012; Yeigh, et al., 2019). The present study builds upon previous research associated with the development and testing of Principal Leadership Questionnaire (PLQ) (Leithwood, 1994; Leithwood et al., 1999; Leithwood & Jantzi, 2000), which assess transformational leadership in the educational context. These studies revealed that transformational practices contribute to the development of commitment and trust on organizational conditions and could affect a variety of school, teacher variables and students’ learning outcomes. Most of the leadership practices is knowledge about a theoretical account of how teachers’ perceptions of leadership are formed (Bass, 1985; Hallinger, 1983; Jantzi & Leithwood,1996). Therefore, an accurate measurement (e.g. PLQ) seems to be a starting point to measure transformational school leadership behavior based on teachers’ perceptions.

Essential practices of transformational leadership

A plethora of findings indicate that the key to the success of the effectiveness of school organizations is school leadership (e.g. Brauckmann & Pashiardis, 2009; Kythreotis et al., 2010; Leithwood & Azah, 2016; Marzano et al., 2005; Menon Eliophotou, 2014; Yang, 2014). Leadership has to respond to innovative school challenges to support teachers following the best models of teaching and increase students’ achievements (Leithwood & Jantzi, 2000). Specifically, principals need to turn their attention to the conditions that can enhance teachers’ self-esteem in their skills (Day & Sammons, 2014) and teacher’s self-efficacy (Fackler & Malmberg, 2016; Gkolia et al., 2018; Sehgal et al., 2017). Furthermore, school principals could affect teachers’ commitment to their school (Dumay & Galand, 2012) by building trust between them and teachers (Browning, 2014). Last but not least, researchers support that school principals could improve teachers’ job satisfaction (Gkolia et al., 2014; Menon Eliophotou, 2014) by providing a shared vision, being an example based on their experience and by having high expectations of success for individuals staff (Dimopoulos et al., 2015; Leithwood et al., 1999; Nash & Bangert, 2014).

All the above leadership practices and behaviours describe a transformational leader rather than a transactional leader. Burns (1978) made a distinction between two types of leadership styles, transformational and transactional leadership. He identified that transactional leadership takes place when the leader motivates the followers through resources, rewards, and valued things, whereas the main goal of a transformational leadership is to increase a group’s level of motivation and morale, inspire, motivate and encourage followers in pursuit of a vision. Bass (1985) defined transformational leadership as the most effective leadership type for the success of an organization that may be learned and be the subject of management training. Leithwood (1994) used Burns (1978) definition and Bass (1985) two-factor theory of leadership (transactional and transformational leadership) and identified factors describing transformational leadership. Based on Bass’ two-factor theory of leadership, the principal who follows transformational leadership practices is not relying only on his or her charisma and on various intrinsic rewards (ex. salary, recognition) (transactional leadership), but is trying to empower teachers and share leadership practices and behaviours. The concept of transformational leadership, that has been used for this research is based on the aforementioned theory, has been adapted for schools but has also been developed in non-school contexts such as private companies (Bass, 1985; Burns, 1978; Leithwood, 1994).

Assessing levels of transformational leadership and its relationship to various aspects of education (e.g. teachers’ job satisfaction) is based on the existence of instruments with sound psychometric properties (Raykov & Marcoulides, 2011). Prior studies examined the validity and reliability of the PLQ. Their results showed the proposed factorial validity of PLQ and the satisfactory internal consistency of latent factors (Jantzi & Leithwood, 1996; Lane, 2016; Ngang, 2011).

To the best of our knowledge no study has thoroughly examined the psychometric properties of the PLQ in a cultural context different than the one that was developed. The usefulness of an instrument is enhanced if it is found to be culturally robust. Gkolia et al. (2018) provided initial evidence of the applicability of the PLQ in Greek educational context. Although, a bi-factor model was selected as the most tenable among four candidate models, two specific factors were dropped out from the original measurement (comprising of six latent factors) developed by Jantzi and Leithwood (1996). Moreover, because the main focus of the study was not to examine the psychometric properties of the PLQ, authors did not compare the general with the specific factors in terms of internal consistency, percent of common variance explained and whether the instrument is mainly a unidimensional or multidimensional. Therefore, additional research is needed to provide a deeper understanding of the PLQ structure.

Furthermore, its invariance across teachers’ important demographic characteristics, such as level of education, is also seldom explored. It is well known that educators in secondary schools have different working conditions (Van der Want et al., 2018) (e.g. they teach many classes and face frequent classroom management issues) in relation to their counterparts in primary schools (e.g. they usually teach only one class and have fewer discipline problems). This different working environment might suggest that principals need to adapt their leadership strategy and practice to their teachers’ specific needs (Jantzi & Leithwood, 1996).

The main purpose of this study was to explore the applicability of the PLQ in Greek educational settings providing the most important aspects describing educational leadership in a centralized educational context as is the Greek one. Towards this end it was decided to examine PLQ’s factorial validity by testing various candidate models, and its invariance across levels of education (primary versus secondary). An additional aim was to provide evidence of PLQ’s concurrent validity by testing its association with selected facets of teachers’ job satisfaction.

Methodology

Participants

Seven hundred and thirty Greek teachers employed in 77 primary and secondary schools in Central Greece were asked to evaluate the leadership skills of their principals. Out of the 730 teachers, 408 (56%) teachers were teaching in primary schools and 322 (44%) in secondary schools.

Measures

Transformational leadership

The Principal Leadership Questionnaire (PLQ) (Jantzi & Leithwood, 1996) was used to assess teachers’ perceptions of leadership of their principal. PLQ comprises 24 items to capture six dimensions of leadership practices. Item responses were given on a 5-point scale statements ranging from “strongly disagree” (1) to “strongly agree” (5). Higher scores denote that principals follow a model closer to transformational leader. A short description of the latent variables of the model is provided below.

Identifying and articulating a vision (5 items, α=.88). The leaders’ behaviour aims to identify new opportunities for their school, to develop, articulate and inspire others with vision for the future. Providing an appropriate model (3 items, α = .86). The leaders’ behaviour sets an example for staff members to follow, which is consistent with the values the leader espouses. Fostering the acceptance of group goals (5 items, α = .80). The leaders’ behaviour aims to promote cooperation among staff and to assist them in working collectively towards achieving common goals. Providing individualized support (5 items, α = .82). The leaders’ behaviour indicates respect for individual members of staff and concern regarding their personal feelings and needs. Intellectual stimulation. (3 items, α = .77). The leaders’ behaviour challenges the staff to re-examine some of the assumptions made regarding their work and rethink how it can be performed. High-performance expectations (3 items, α = .73). The leaders’ behaviour demonstrates the leaders’ expectations for excellence, quality and high performance of the staff.

Teachers job satisfaction

Teacher’s Satisfaction Inventory (TSI) (Gkolia & Koustelios, 2014) was employed to measure teachers’ job satisfaction. TSI consists of 20 items, which assess the following five facets; “principal”, “colleague”, “job itself”, “students” and “working conditions”. Prior studies provided strong evidence of its validity and reliability (Gkolia & Koustelios, 2014; Katsakioris, 2018). In the present study only the “principal” facet was used to test PLQ’s concurrent validity. Item responses were given on a 5-point scale statements ranging from “strongly disagree” (1) to “strongly agree” (5).

Adaptation of PLQ to the Greek educational context

Adaptation of the PLQ instrument into the Greek educational context conducted using procedures suggested by DeVellis (2003). The first step towards adapting the PLQ was its translation. Two English teaching experts, whose native language was Greek, translated the PLQ into Greek; then the Greek version of the instrument was translated back into English by another team of Greek teaching experts. Following translation in Greek, five school principals checked the instrument and proposed minor changes.

Data analysis

Generalisability theory

Since it is hypothesized that teachers observe their principal’s behaviours within a school, but as these differ from those teachers’ observations in other schools, a generalisability analysis (Shavelson et al., 1989) was initially performed. In contrast with classical test theory (CTT) where only ‘one type’ of error, ‘error score’, is concerned, in generalizability theory (GT), a ‘universe score’ is concerned. The notion of a universe score is at the heart of GT and basically corresponds to those explicit aspects of a measurement procedure that provide information about the observed scores. Specifically, whereas in CTT we are interested in ‘how accurately observed scores reflect corresponding true scores’ (p. 225), in GT we are focused on ‘how accurately observed scores permit us to generalize about persons’ (p. 225) (Raykov & Marcoulides, 2011). This approach has been followed by several researchers in the field of education (e.g. Creemers et al., 2010).

Confirmatory factor analysis

Then, Confirmatory Factor Analysis (CFA) procedures were employed to test the construct validity of the translated PLQ.

CFA is testing hypotheses of theories about the latent structure of studied sets of observed measures. This implies that in order to be able to apply CFA, the number of factors has been already defined by theory rather is determined from the collected data (Raykov & Marcoulides, 2011). Therefore, a first order six-factor model initially postulated and tested as has been defined by Jantzi and Leithwood (1996). Kline (2011) suggests that even if a first-order multidimensional factor model seems tenable, additional candidate models should also be tested (e.g. unidimensional model). For example, strong factor intercorrelations might indicate a higher order model (second-order model). A second-order model presume direct causal effects on the indicators of the first-order factors representing a general behaviour construct with no indicators (Kline, 2011). Last but not least, a bi-factor model was tested. The bi-factor model essentially hypothesizes that ‘‘(a) there is a general factor that accounts for the commonality shared by the facets, and (b) there are multiple specific factors, each of which account for the unique influence of the specific component over and above the general factor’’ (Chen et al., 2006, p. 223). EQS 6.1 (Bentler, 1990) was used to examine the tenability of the postulated models (unidimensional, a first-order multidimensional factor, second-order model and a bi-factor model). Models’ parameters were estimated employing the Robust Maximum Likelihood method. Multiple fit indices were utilized to evaluate the hypothesized measurement models. These include the scaled robust S-B χ², ratio S-B χ²/df, RMSEA, CFI and SRMR. The following values of the alternative fit indices were considered as evidence of a satisfactory fit, CFI close to 1, RMSEA less than .05, ratio S-B χ²/df less than 1.96, and SRMR less than 0.08 (Marcoulides & Schumacker, 2001; Raykov & Marcoulides, 2011).

Measurement invariance

After establishing the most tenable model, PLQ measurement invariance was examined between teachers serving in the primary education and their counterparts in the secondary. Measurement invariance has been discussed in more detail in different studies (e.g. Dimitrov, 2006; Sass, 2011).

For this study, participants were divided into two levels of education, according to level A (n₁ = 408 Greek primary school teachers) and B (n₂ = 322 Greek secondary school teachers). Initially the most tenable model derived from CFA procedures fitted separately for each level. Then the configural invariance was tested by constraining the factorial structure to be the same across levels without adding any constraint (configural invariance – unconstrained model). Next, the metric invariance was tested by constraining the factor loadings to be equal across levels. Finally, scalar invariance, was tested by setting items’ intercept to be the same across levels (scalar invariance). Since the increasingly restrictive models were nested, ΔS-B χ² was employed to assess their relative fit (Raykov, 2004).

Results

Generalisability theory

ANOVA analysis has been conducted with a single random factor. School was the independent variable and each item of the questionnaire was the dependent variable. It was found that only one item (p > .05) of the instrument may not be generalized over schools. The specific item was part of the factor “Identifying and articulating a vision” and asked if the headteacher “excites the faculty with visions of what they may be able to accomplish if they work together as a team”. Thus, for the purposes of the construct analysis 23 out of 24 PLQ scores was conducted for each model.

PLQ factorial validity

Initially a multidimensional first-order six factor model (Model 1) was examined. Fit indexes of the model provided mixed results (S-B χ²(215) = 748.6, S-B χ²/df =3.48, RMSEA = .058, CFI = .922 and SRMR = .041). Whereas RMSEA and SRMR showed good fit to the data, S-B χ²/df and CFI deviated from the suggested cut-off values. Another issue with this model was the high correlation between “Identifying and articulating vision” (F1) and “Providing an appropriate model” (F2) (r_F1F2 = .96). Although, strong associations among latent factors might lead to combine these factors into a single factor (Raykov & Marcoulides, 2011), CFA needs a correctly specified model that is arising from an existing theory (Raykov & Marcoulides, 2011). Having in mind the theoretical hypothesized model of PLQ upon which it was developed (Jantzi & Leithwood, 1996; Leithwood, 1994), it may be worthwhile to investigate further the structure of factors F1 and F2 as independent unidimensional models.

CFA for the F1 showed that model do not fit the data (S-B χ²(2) = 152.90, S-B χ²/df = 76.45, RMSEA = .344, CFI = .862 and SRMR = 0.071). On the other hand, CFA could not be conducted for F2, because it comprises only three items and thus it is just identified. (i.e., its degrees of freedom are 0). Therefore, for that factor (F2) exploratory factor analysis (Mulaik, 2009; Raykov & Marcoulides, 2011) was conducted with satisfactory results. Specifically, the first eigenvalue was equal to 2.431 and explained more than 50% of the variance (81.04%) whereas the second eigenvalue was much smaller than 1 (i.e., .38). These results show that we can treat these three items as belonging to one factor, especially since all three items had relatively high loadings (i.e., bigger than .859). Taking into account the above findings it was decided to exclude F1 from the final model and thus a five-factor structure seems to be a reasonable model and the analysis proceeded (Table 1). Furthermore, the question “provided for extended training to develop my knowledge and skills relevant to be a member of the school faculty” belonged to factor F4, had small and not significant loading (<.5) and was dropped out of the final model. For the items and factors that were excluded from the analyses some possible explanations are provided in the discussion.

Table 1.

Fit indices for the five candidates PLQ models.

Models	χ² (df)	χ²/df	CFI	SRMR	RMSEA	90% C.I. of RMSEA
M1	748.6 (215)*	3.48	.922	.041	.058	.05-.06
M2	385.3 (125)*	3.08	.950	.038	.053	.05-.06
M3	838.7 (134)*	6.25	.865	.050	.085	.08-.09
M4	403.3 (130)*	3.10	.948	.040	.054	.05-.06
M5	230.4 (116)*	1.98	.978	.030	.037	.03-.04

Note: M1 = First order 6-factor model, M2 = First order 5-factor model, M3 = Single factor model, M4 = 2nd order 5-factor model, Model 5 = bi-factor model.

*p < 0.01.

Table 1 presents fit indices for the four competitive models, namely, a first-order six factors model (M1), a first-order five factors model (M2), a single factor model (M3) and a second-order model (M4). None of the fit indices concerning the unidimensional model reached the accepted cut-off values and thus this model was rejected. The two multidimensional models (first and second order factors) yielded similar fit indices values. Moreover, it was noticed that latent factors intercorrelations were positively correlated ranging from .754 to .860, yielding strong values which might suggest an underlying general factor (Reise, 2012). Thus, it was decided to examine the applicability of a bi-factor model (M5) (Table 1). Fit indices’ results showed an excellent fit to the data.

Table 2 depicts the derived parameter estimates for the bi-factor model. It was noticed that items loading to the general factor yielded higher values in relation to the specific factors. All item loadings were statistically significant and so the bi-factor model was selected as the most tenable for representing PLQ factorial structure among both primary and secondary public-school teachers.

Table 2.

Item loadings of the PLQ bi-factor model to the general and specific factors.

Items	General factor	Model’s behavior	Fosters commitment	Provides individual support	Intellectual stimulation	Holds high performance expectations
item 5	.735	.212
Item 6	.804	.457
Item 7	.778	.401
Item 8	.742		.342
Item 9	.741		.443
Item 10	.715		.235
Item 11	.694		.141
Item 12	.729		.198
Item 14	.626			.137
Item 15	.643			.227
Item 16	.732			.334
Item 17	.738			.404
Item 18	.654				.370
Item 19	.615				.789
Item 20	.755				.205
Item 21	.745					.151
Item 22	.659					.696
Item 23	.620					.307

Note: item 5 – item 23 represents items order as they appear in the PLQ.

Reliability

Given that a bi-factor model was selected as the most tenable among the five candidate models, omega coefficient was used to examine the reliability of the latent factors. According to Rodriguez et al. (2016a) for a bi-factor model four omega coefficients can be calculated, omega (ω), omega subscale (ωS), omega hierarchical (ωH) and the omega hierarchical subscale (ωHS). Omega hierarchical denotes the percent of variance of the total score that can be attributed to the general factor, whereas omega hierarchical subscale reflects the reliability of the specific factor after controlling for the general factor. The general factor of the PLQ showed a very high ωΗ value of .92. On the other hand, the ωHS for the five specific factors yielded low to moderate values, ranging from .123 to .312 (Table 3).

Table 3.

Testing reliability and common variance of general and specific factors.

	Omega	Omega H	ECV
General Factor	.964	.922	.778
F1: Models behavior	.881	.176	.036
F2: Fosters commitment	.885	.123	.037
F3: Provides individuals support	.831	.139	.030
F4: Intellectual stimulation	.879	.312	.069
F5: Holds high performance expectations	.841	.245	.052

Another useful index for evaluating a bi-factor model is the explained common variance (ECV), which can be calculated for the general and the specific factors. ECV is usually used to judge the degree of the unidimensionality of the construct. Obviously the higher the ECV value for the genaral factor the better approximation of the unidimensional model is achieved. Table 3 presents the ECV for the general and the specific factors. From Table 3 is evident that the general factor explains the vast majority of the PLQ common variance. On the contrary, the five PLQ specific factors account only for a small fragment of the common variance.

Measurement invariance

Since the bi-factor model was the best fitting model, the measurement invariance was examined for this model for the two samples of teachers’ levels of education to investigate the factorial structure of the bi-factor model along different groups. The first step was to fit the PLQ separately for primary and secondary levels of education. Results (Table 4) showed satisfactory fit-indices and thus no modifications were introduced. Then, multiple-group CFA was tested to explore whether the structure of bi-factor model with no constraints was tenable across the two groups (primary and secondary) (Configural - Model 6-M6). Fit indexes for the configural model provided a reasonably good fit to the empirical data (Table 4) and so configural invariance was found tenable for teachers’ levels of education. Given that configural invariance (M6) hold, metric invariance model (M7) was tested. Findings showed (Table 4) similar fit indexes to the configural model. Moreover, deterioration of the model M7 was not statistically significant, as indicated by the ΔS-B χ² test, thus metric invariance was met. The scalar invariance model (M8) also provided a good fit to the data. In terms of model comparisons M8 was not statistically significant in relation to model M7, scalar invariance was met across education level. Thus, it was decided to accept PLQ’s scalar invariance. Given that measurement invariance of the bi-factor model across levels of education was hold, analysis could proceed by comparing the latent factor means.

Table 4.

PLQ invariance results across education level.

Models	χ² (df)	χ²/df	CFI	SRMR	RMSEA	90% C.I. of RMSEA	Models’ comparison	ΔS-B χ²(Δdf)
Primary	193.5 (116)*	1.66	.973	.032	.041	.03-.05	NA
Secondary	184.3 (116)*	1.58	.971	.040	.043	.03-.05	NA
M6	402.8 (232)*	1.74	.967	.036	.032	.03-.04	NA
M7	427.8 (263)*	1.62	.969	.049	.029	.02-.03	M6 vs M7	26.4 (31)^ns
M8	479.4 (275)*	1.74	.965	.049	.032	.03-.04	M7 vs M8	15.6 (12)^ns

Note: M6 = Configural model, M7 = Metric model, M8 = Scalar model, NA = not applicable.

*p < .01, ns = non-significant.

Multiple-group analysis

According to Byrne (2006) multiple-group structural equation model is more powerful than MANOVA for latent variables comparison and thus this approach was adopted. The groups involved were examined for differences by level of education (1 = primary, 0 = secondary: reference group). The equality constraint was correctly imposed in levels of education (M8, Table 4) models.

The latent mean of secondary school teachers’ responses for “intellectual stimulation” (F₅) factor ( $\hat{μ}$ _sec = .084), is higher than the one of primary school teachers. However, the latent mean of primary school teachers’ responses for general factor “transformational leadership” ( $\hat{μ}$ _sec = .188) is higher than the secondary school teachers. The rest of the latent factors of PLQ were not significantly different (p < .05) (Table 5).

Table 5.

PLQ latent mean comparison across educational stages (n = 730).

Models	Latent mean
Model 8 (focal group = primary schools)
Model’s behavior (F2)	.022 (.02)
Fosters commitment (F3)	.043 (.03)
Provides individual support (F4)	.033 (.02)
Intellectual stimulation (F5)	−.08 (.03)^*
Holds high performance expectations (F6)	−.01 (.01)
General Factor (GF)	.19 (.063)^*

Note: * p < .05.

Transformational leadership factors and teachers’ job satisfaction factor

The internal consistency of the “Principal” facet of TSI was satisfactory yielding a value of Cronbach α = .87 και Raykov’s RHO = .94. SEM analysis conducted to test whether general factor of PLQ have affect upon teachers’ satisfaction with their principal. The structural path from the PLQ general factor to teachers’ satisfaction with their principal was statistically significant, with β = 22.885 (.07) and R²=.753, (p < .05).

Discussion

The theoretical background, developed in this paper, of how teachers’ perceptions of principals’ transformational leadership are defined is based on the work of Jantzi and Leithwood (1996). Previous studies which investigated the effectiveness of transformational leadership in schools, found that almost all successful principals draw on the same basic transformational leadership practices, which positively affect the quality of learning and teaching (Damanik & Aldridge, 2017; Day & Sammons, 2014; Gkolia et al., 2014; 2018; Katsakioris, 2018; Lambersky, 2016; Leithwood et al., 2020; Yang, 2014).

This paper contributes to the existing body of knowledge on transformational leadership in the education context by testing the factorial validity of PLQ and its measurement invariance across levels of education in a different educational context that the one that PLQ was originally developed. Findings of the present study are encouraging and suggest that the PLQ can be applied to the Greek educational context to assess teacher’s perceptions of their principals.

A stringent methodology was followed to explore PLQ’s factorial validity. Several competitive models of the PLQ were examined based on prior studies, theoretical background (Jantzi & Leithwood, 1996) and suggestions about how to test an instrument’s structure (Kline, 2011). Results showed that a bi-factor model had the best fit to the data. Specifically, a bi-factor model examined not only influences on teachers’ responses of the five transformational leadership dimensions (“model behaviour”, “fosters commitment”, “provides individual support”, “intellectual stimulation” and “holds high expectations”) considered separately but also the effects on teachers’ overall perceptions of principals’ transformational leadership. In terms of internal consistency, omega coefficinet of the PLQ general factor as well as of the specific factrors showed acceptable values. When responses to an instrument are best described by a bi-factor model it is of interest to examine whether the construct is essentially unidimensional or multidimensional. Towards this end, several indices can inform such a decision. Rodriguez et al. (2016a) maintain that when ω_Η is above .80 total scores can be regarded as unidimensional. In the present study the omega hierarchical and omega hierarchical subscale showed that the reliability of the general factor was substantially higher than the reliability of the specific factors, yielding a value of .922. Thus, the vast majority of reliable variance is attributed to PLQ’s general factor, and only trivial systematic variance is accounted by the specific factors. Moreover, the ECV clearly showed that the general factor explaines a large portion of the variance. In a recent study, Rodriguez et al. (2016b) calculated the mean ECV for the general factor of fifty published studies, which was .67 (range .76 to .40). The ECV for the PLQ general factor is well above the maximum value of the published ECV (.78), which suggests a strongly unidimensional construct. Consequently, despite the multidimensional structure of the PLQ, total scores can be considered essentially unidimensional.

Measurement invariance (MI) of the bi-factor structure of PLQ across levels of education was further examined and supported. Findings show that the bi-factor model is invariant across primary and secondary education, indicating that teachers’ perceptions about their principal leadership using PLQ can be applied to both levels of education and their results can be meaningfully compared. To the best of our knowledge this is the first time in an educational research context that the structure of PLQ has been examined and confirmed by using MI analysis. Another implication of the PLQ measurement invariance is that outcomes of specifically designed interventions aiming to enhance leadership practices and/or polices can be reliably evaluated at the educators’ level.

It should be reminded that the PLQ subscale, “articulating a vision” was excluded from the analysis. These findings are in line with Kaparou and Bush (2016) comparative study in educational leadership of outstanding schools, in which they also found that in Greek schools creating a shared vision is not an official expectation. In addition, all the above was confirmed by Konidari and Abernot (2006) research in Greek secondary schools which argue that the lack of common vision in schools may be related to the lack of communication among headteachers, teachers and stakeholders. Although, one can only speculate, it seems that Greek educators do not perceive that their principals strive to identify and articulate a shared vision for learning within their schools. Without a doubt further research is needed to explore why this lack of specific management occurs consistently in Greek educational settings. Furthermore, one item as has been defined at “PLQ factorial validity” section (“provided for extended training… to be a member of the school faculty”) yielded low loading and was discarded from the analysis. It seems that teachers may need support to attend educational training programmes. Different studies supported the importance of lifelong professional development to enable teachers to refresh, develop and extended their knowledge on teaching methods as well as improve their skills and practices, and this is because teachers need to be always informed of the framework of established policies and regulations (Leithwood et al., 2020; Organisation for Economic Co-operation and Development, 2020).

Based on educational theory three important results arise from multiple-groups and reliability results, one general and two more practical. First, since scalar invariance was confirmed across education levels (primary and secondary) is likely to support to the proposed framework that a collective action of 18 transformational leadership practices which describe a transformational leader should be treated equal between primary and secondary teachers. This result confirms Jantzi and Leithwood (1996) implications who support the idea that the most important teachers’ perception is what their principal does and not who they are even if belong to primary or secondary school. First practical result is that, primary school teachers’ principal perceptions compared to secondary ones identified more often that primary school principals, can enhance teachers through various practices; by being an example for them, fostering commitment in the process of developing schools goals, providing individualized support to each teacher, promoting cooperation among the teachers, by expecting high quality and high levels of performance from the teacher and by challenging teachers to evaluate their work. It should be acknowledged that which level of education teachers’ work for is likely to influence teachers’ transformational perceptions because secondary schools are often larger than primary schools and so the latter is more personal and less bureaucratic. As a result, transformational leadership practices are easier to implement in primary schools. Second practical finding is that secondary school teachers, reported that their principal challenge and stimulate their effort to be innovative more often compared to primary school teachers.

The “Principal” facet of the TSI was used to examine the concurrent validity of the PLQ. Obviously, this facet was selected as being the most relevant to the concept of leadership. Results showed that the association between the “Principal” and the general factor of the PLQ was in the anticipated direction (positive) and strength (R² =.753) and provide further support of the validity of the Greek version of the PLQ. Findings are also in line with prior studies where transformational leadership practices had a positive and significant relationship with teachers’ job satisfaction (Menon Eliophotou, 2014; Tesfaw, 2014; Thomas et al., 2020). Future research could provide a more in-depth examination between each construct of transformational leadership and different types of teachers’ important aspects such as self-efficacy, burnout, school culture.

All the above imply that the education community needs to take a closer look at the principals’ reported behaviours as leaders and it may be worthwhile for the educational authorities and policy-makers, to design and apply appropriate policies and professional development training programmes in order to strengthen the position of principals with professional development leadership courses that promote transformational skills as well as to offer training teaching programmes in order to inform teachers and principals of new models and teaching methods regarding the improvement of student achievements and school effectiveness (Leithwood, 2012).

Implications of the results of this study for research could also be drawn. It should first be acknowledged that this study has been demonstrating in a centralized educational context, where there is no local control and important decisions are made by the educational authorities. Undoubtedly, teachers’ views on principals’ transformational leadership behaviours represent a concept with significant implications for the field of educational effectiveness; even though principals have limited control in curriculum implementation, salary scales, number, and staff appointments. However, principals’ behaviours in centralized educational contexts do not depend on the views of educational authorities, but on their own perspectives and willingness to engage and inspire teachers to share a common vision, to improve teachers’ job satisfaction and students’ learning outcomes (Kyriakides et al., 2018; Organisation for Economic Co-operation and Development, 2017). For example, central government control over the allocation of teachers to individual schools eliminates the accountability of school leaders but increases the school leaders’ available time to focus on leadership practices that can lead to educational improvement. This implies that the instrument which was tested can be used as a starting point for further studies within centralized educational contexts, which may try to measure all six transformational leadership factors.

There is also a need for many more longitudinal and empirical studies to investigate the impact of transformational school leadership factors and how directly and indirectly affect teachers' job satisfaction, students' learning and overall school improvement. Using carefully designed longitudinal studies one could enhance a consistent investigation or exploration of different outcomes across different school years and across the same individuals over time (Kyriakides & Creemers, 2008). Last, but not least, these studies may not only help the educational community to understand model transformational leadership through specific practices but also to bring a substantial change to improve school effectiveness.

Footnotes

Author notes

ORCID iDs

Aikaterini Gkolia

Nikolaos Tsigilis

Author Biographies

Aikaterini Gkolia is Adjunct professor at the School of education of the University of Nicosia in the section of Quantitative analysis in Education and a full employee in the Ministry of Justice, responsible of Justice Statistics. Her academic research interests focus in the field of educational leadership using advanced level of quantitative methods (SEM, Multilevel analysis) and in psychometric practice and theory.

Nikolaos Tsigilis is Associate Professor at the Department of Journalism and Mass Media of the Aristotle University of Thessaloniki. His field of expertise is “Methodology and Psychometry of Social Sciences”. His main research interests revolve around issues of research methods in the field of social sciences, the examination of instruments psychometric properties and the application of statistical techniques.

Maria Evangelou is currently an Honorary Research Fellow at the Department of Education, University of Oxford. Until the end of April 2018 she was an Associate Professor there, the Director for MSc Education; the pathway Leader for the MSc in Child Development and Education and the convener of the Families Effective Learning and Literacy (FELL) Research Group . Her research focuses on the design, implementation and evaluation of early childhood interventions; the development of children at risk of educational underachievement; the language and literacy development in the early years; parental engagement in children’s learning; the effects of home learning; early years professional development.

Athanasios Koustelios is a Professor in the Department of Business Administration at the University of Thessaly, Greece. He has published more than 120 papers in Internationals and Greek refereed journals and more than 160 conference presentations His research expertise is in the area of human resource management and organizational behaviour.

References

Bass

(1985). Leadership and Performance. Free Press.

Bentler

P. M.

(1990). Comparative fit indexes in structural models. Psychological Bulletin, 107(2), 238–246. https://doi.org/10.1037/0033-2909.107.2.238

Berkovich

Eyal

(2017). Emotional reframing as a mediator of the relationships between transformational school leadership and teachers’ motivation and commitment. Journal of Educational Administration, 55(5), 450–468. https://doi.org/10.1108/JEA-07-2016-0072

Brauckmann

Pashiardis

(2009). From PISA to LISA: New educational governance and school leadership: Εxploring the foundations of a new relationship in an international context [Paper presentation]. 90th Annual Meeting of the American Educational Research Association, San Diego, CA.

Browning

(2014). Why trust the head? Key practices for transformational school leaders to build a purposeful relationship of trust. International Journal of Leadership in Education, 17(4), 388–409. https://doi.org/10.1080/13603124.2013.844275

Burns

(1978). Leadership. Harper & Row.

Byrne

B. M.

(2006). Structural equation modelling with EQS: Basic concepts, applications, and programming (2nd ed.). Lawrence Erlbaum Associates, Inc.

Chen

F. F.

West

S. G.

Sousa

K. H.

(2006). A comparison of Bi-factor and second order models of quality of life. Multivariate Behavioral Research, 41(2), 189–225. https://doi.org/10.1207/s15327906mbr4102_5

Creemers

B. P. M.

Kyriakides

Sammons

(2010). Methodological advances in school effectiveness research. Routledge.

10.

Damanik

Aldridge

(2017). Transformational leadership and its impact on school climate and teachers’ self-efficacy in Indonesian high schools. Journal of School Leadership, 27(2), 269–296. https://doi.org/10.1177/105268461702700205

11.

Day

Sammons

(2014). Successful school leadership. Education Development Trust.

12.

DeVellis

R. F.

(2003). Scale development theory and applications (2nd ed.). SAGE Publications, Inc.

13.

Dimitrov

D. M.

(2006). Comparing groups on latent variables: A structural equation modeling approach. Work (Reading, Mass.), 26(4), 429–436.

14.

Dimopoulos

Dalkavouki

Koulaidis

(2015). Job realities of primary school principals in Greece: Similarities and variations in a highly centralized system. International Journal of Leadership in Education, 18(2), 197–224. https://doi.org/10.1080/13603124.2014.954627

15.

Dumay

Galand

(2012). The multilevel impact of transformational leadership on teacher commitment: Cognitive and motivational pathways. British Educational Research Journal, 38(5), 703–772. https://doi.org/10.1080/01411926.2011.577889

16.

Fackler

Malmberg

L. E.

(2016). Teachers’ self-efficacy in 14 OECD countries: Teacher, student group, school and leadership effects. Teaching and Teacher Education, 56, 185–195. https://doi.org/10.1016/j.tate.2016.03.002

17.

Gkolia

Koustelios

(2014). The development of a questionnaire measuring teachers’ job satisfaction (teacher’s satisfaction inventory-(TSI)). Journal of Education, 2–3, 195–214.

18.

Gkolia

A. K.

Belias

Koustelios

(2014). The impact of principals’ transformational leadership on teachers’ satisfaction: Evidence from Greece. European Journal of Business and Social Sciences, 3(6), 69–80.

19.

Gkolia

A. K.

Belias

D. A.

Koustelios

(2018). Exploring the association between transformational leadership and teacher’s self-efficacy in Greek education system: A multi-level SEM model. International Journal of Leadership in Education, 21(2), 176–196. https://doi.org/10.1080/13603124.2015.1094143

20.

Hallinger

(1983). Assessing the instructional management behavior of principals [PhD thesis]. Stanford University, USA.

21.

Jantzi

Leithwood

(1996). Toward an explanation of variation in teachers’ perceptions of transformational school leadership. Educational Administration Quarterly, 32(4), 512–538. https://doi.org/10.1177/0013161X9603200404

22.

Kaparou

Bush

(2016). Instructional leadership in Greek and English outstanding schools. International Journal of Educational Management, 30(6), 894–912. https://doi.org/10.1108/IJEM-03-2015-0025

23.

Katsakioris

(2018). Investigate the relationship among transformational leadership and teachers’ job satisfaction in primary schools: An empirical research. https://apothesis.eap.gr/handle/repo/3986

24.

Kline

B. R.

(2011). Principles and practice of structural equation modeling. The Guilford Press.

25.

Konidari

Abernot

(2006). From TQM to learning organisation: Another way for quality management in educational institutions. International Journal of Quality & Reliability Management, 23(1), 8–26. https://doi.org/10.1108/02656710610637523

26.

Kyriakides

Creemers

B. P.

(2008). A longitudinal study on the stability over time of school and teacher effects on student outcomes. Oxford Review of Education, 34(5), 521–545. https://doi.org/10.1080/03054980701782064

27.

Kyriakides

Creemers

B. P.

(2012). School policy on teaching and school learning environment: Direct and indirect effects upon student outcome measures. Educational Research and Evaluation, 18(5), 403–424. https://doi.org/10.1080/13803611.2012.689716

28.

Kyriakides

Georgiou

M. P.

Creemers

B. P.

Panayiotou

Reynolds

(2018). The impact of national educational policies on student achievement: A European study. School Effectiveness and School Improvement, 29(2), 171–203. https://doi.org/10.1080/09243453.2017.1398761

29.

Kythreotis

Pashiardis

Kyriakides

(2010). The influence of school leadership styles and culture on students’ achievement in Cyprus primary schools. Journal of Educational Administration, 48(2), 218–240. https://doi.org/10.1108/09578231011027860

30.

Lambersky

(2016). Understanding the human side of school leadership: Principals’ impact on teachers’ morale, self-efficacy, stress, and commitment. Leadership and Policy in Schools, 15(4), 379–405. https://doi.org/10.1080/15700763.2016.1181188

31.

Lane

E. S.

(2016). Examining the relationship between principal leadership and school climate [Electronic Theses and Dissertations, 1094]. https://digitalcommons.du.edu/etd/1094

32.

Leithwood

(1994). Leadership for school restructuring. Educational Administration Quarterly, 30(4), 498–518. https://doi.org/10.1177/0013161X94030004006

33.

Leithwood

(2012). The Ontario leadership. The Institute for Education Leadership.

34.

Leithwood

Azah

(2016). Characteristics of effective leadership networks. Journal of Educational Administration, 54(4), 409–433. https://doi.org/10.1108/JEA-08-2015-0068

35.

Leithwood

Jantzi

(2000). The effects of transformational leadership on organizational conditions and student engagement with school. Journal of Educational Administration, 38(2), 112–129. https://doi.org/10.1108/09578230010320064

36.

Leithwood

Harris

Hopkins

(2008). Seven strong claims about successful school leadership. School Leadership & Management, 28(1), 27–42. https://doi.org/10.1080/13632430701800060

37.

Leithwood

Harris

Hopkins

(2020). Seven strong claims about successful school leadership revisited. School Leadership & Management, 40(1), 5–22. https://doi.org/10.1080/13632434.2019.1596077

38.

Leithwood

Κ.

Jantzi

Steinbach

(1999). Changing leadership for changing times. Open University Press.

39.

Marcoulides

G. A.

Schumacker

R. E.

(2001). New developments and techniques in structural equation modeling. Lawrence Erlbaum Associates, Inc.

40.

Marzano

R. J.

Waters

McNulty

B. A.

(2005). School leadership that works. From research to results. ASCD and MCREL.

41.

Menon Eliophotou

(2014). The relationship between transformational leadership, perceived leader effectiveness and teachers’ job satisfaction. Journal of Educational Administration, 52(4), 509–528. https://doi.org/https://doi.org/10.1108/JEA-01-2013-0014

42.

Mulaik

(2009). Factor analysis. Chapman & Hall.

43.

Nash

Bangert

(2014). Exploring the relationships between principals’ life experiences and transformational leadership behaviours. International Journal of Leadership in Education, 17(4), 462–480. https://doi.org/10.1080/13603124.2013.858775

44.

Ngang

T. K.

(2011). The effect of transformational leadership on school culture in male’ primary schools Maldives. Procedia—Social and Behavioral Sciences, 30, 2575–2580. https://doi.org/10.1016/j.sbspro.2011.10.503

45.

Organisation for Economic Co-operation and Development. (2017). Education policy in Greece: A preliminary assessment.

46.

Organisation for Economic Co-operation and Development. (2020). Education GPS, the world of education at your fingertips. Teacher professional development. https://gpseducation.oecd.org/revieweducationpolicies/#!node=41732&filter=all

47.

Raykov

(2004). Behavioral scale reliability and measurement invariance evaluation using latent variable modeling. Behavior Therapy, 35(2), 299–331. https://doi.org/10.1016/S0005-7894(04)80041-8

48.

Raykov

Marcoulides

(2011). Introduction to psychometric theory. Routledge.

49.

Reise

S. P.

(2012). The rediscovery of bi-factor measurement models. Multivariate Behavioral Research, 47(5), 667–696. https://doi.org/10.1080/00273171.2012.715555

50.

Rodriguez

Reise

S. P.

Haviland

M. G.

(2016a). Evaluating bi-factor models: Calculating and interpreting statistical indices. Psychological Methods, 21(2), 137–150. https://doi.org/10.1037/met0000045

51.

Rodriguez

Reise

S. P.

Haviland

M. G.

(2016b). Applying bi-factor statistical indices in the evaluation of psychological measures. Journal of Personality Assessment, 98(3), 223–237. https://doi.org/10.1080/00223891.2015.1089249

52.

Sass

D. A.

(2011). Testing measurement invariance and comparing latent factor means within a confirmatory factor analysis framework. Journal of Psychoeducational Assessment, 29(4), 347–363. https://doi.org/10.1177/0734282911406661

53.

Sehgal

Nambudiri

Mishra

S. K.

(2017). Teacher effectiveness through self-efficacy, collaboration and principal leadership. International Journal of Educational Management, 31(4), 505–517. https://doi.org/10.1108/IJEM-05-2016-0090

54.

Sebastian

Huang

Allensworth

(2017). Examining integrated leadership systems in high schools: Connecting principal and teacher leadership to organizational processes and student outcomes. School Effectiveness and School Improvement, 28(3), 463–488. https://doi.org/10.1080/09243453.2017.1319392

55.

Shavelson

R. J.

Webb

N. M.

Rowley

G. L.

(1989). Generalizability theory. American Psychologist, 44(6), 922–932. https://doi.org/10.1037/0003-066X.44.6.922

56.

Sun

Leithwood

(2012). Transformational school leadership effects on student achievement. Leadership and Policy in Schools, 11(4), 418–451. https://doi.org/10.1080/15700763.2012.681001

57.

Tesfaw

T. A.

(2014). The relationship between transformational leadership and job satisfaction: The case of government secondary school teachers in Ethiopia. Educational Management Administration & Leadership, 42(6), 903–918. https://doi.org/10.1177/1741143214551948

58.

Thomas

Tuytens

Devos

Kelchtermans

Vanderlinde

(2020). Transformational school leadership as a key factor for teachers’ job attitudes during their first year in the profession. Educational Management Administration & Leadership, 48(1), 106–132. https://doi.org/10.1177/1741143218781064

59.

Van der Want

A. C.

Schellings

Mommers

(2018). Experienced teachers dealing with issues in education: A career perspective. Teachers and Teaching, 24(7), 802–824. https://doi.org/10.1080/13540602.2018.1481024

60.

Yeigh

Lynch

Turner

Provost

C. S.

Smith

Willis

L. R.

(2019). School leadership and school improvement: Αn examination of school readiness factors. School Leadership & Management, 39(5), 434–456. https://doi.org/10.1080/13632434.2018.1505718

61.

Yang

(2014). Principals’ transformational leadership in school improvement. International Journal of Educational Management, 28(3), 279–288. https://doi.org/10.1108/IJEM-04-2013-0063