Conceptualising principal resilience: Development and validation of the Principal Resilience Inventory

Introduction

The OECD 2030 Agenda for sustainable development also highlights that human resilience is needed for a better world (OECD/EC-JRC, 2021). This blueprint calls for building a more resilient society to incorporate constant change and unprecedented uncertainties for better well-being and outcomes. This message increases the importance of human individual resilience capacity building, especially the young generation's resilience building and collective resilience as well (Shina and Sangwook, 2021). Derived from the cascading resilience effect, the best way to foster student resilience is to promote teacher resilience; the best way to foster teacher resilience is to promote principal resilience (Beauchamp et al., 2021; Damico, 2020; Ungar, 2021). Haslam et al. (2020) go further and characterise the school leader as someone who must see themselves as one of a group, sharing an identity of ‘us-ness’ in pursuit of the collective good. The idea echoes the recent notion of ‘principal first’ which has been placed on the current agenda to cultivate principal resilience to lead, self-nurture, lead to nurture (Beauchamp et al., 2021; Klap et al., 2021). Principals have the unique opportunity to promote resilience in others by developing the belief that every individual has the capacity for resilience within the whole school (Truebridge, 2014). This will be beneficial for nurturing human capital and is one step forward towards responding to the OECD 2030 Agenda to build a more resilient society.

Principalship itself is generally considered a high-risk profession due to the high burden of the profession, increasing accountability and standardisation, challenging societal demands, complex management mechanism, constant change, and unprecedented uncertainties (Chen, Bellibaş et al., 2023; Chen, Waiker et al., 2023; Chen and Walker 2023; DeMatthews et al., 2021; Horwood et al., 2021). Paradoxically, leaders ‘spent putting out too many fires today (that) undermines high-leverage action and therefore creates crisis situations tomorrow’. Apparently, these ‘too many initiatives drain’ our leaders (Allison, 2012: 80). This phenomenon is given the metaphor ‘acceleration trap’ by Bruch and Menges (2010: 1). Principals are often to complain about being overwhelmed (Allison-Napolitano, 2022). This state of affairs is especially true for school principals in the post-pandemic era as they need not only to navigate the Covid-19 pandemic impact on top of their routine demands, but also ‘re-position their work’ and turn it ‘on its axis’ in order to maintain schools functioning and try to move their schools forward during this turbulent time (Harris and Jones, 2020: 245). Unavoidably, all these challenges have made principals highly vulnerable. Resilience, it turns out, is an inside job that begins with choosing to be resilient. In large part, resilience requires principals to take care of schools and themselves, while taking actions in the new normal (Allison-Napolitano, 2022).

Research has also shown that principal resilience is crucial not only for dealing with professional challenges and uncertainties but also for enhancing desirable individual and organizational well-being and performance (Glazzard and Stones, 2021; Lazaridou, 2021; Maulding et al., 2012; Papaioannou et al., 2021; Wells and Klocko, 2018). Despite its need and significance, research on principal resilience has been largely left behind and is hovering in its infancy. Particularly, it seems that no instrument that has been developed from and for measuring principal resilience using a robust validation procedure. The existing literature on principal resilience has always adopted instruments derived from general populations (e.g. Bumphus, 2008; Kiser, 2020; Offutt, 2011; Wescott, 2018). These adopted instruments may not best capture the unique nature of principalship and essence of principal resilience as individual principals contextualize ‘their experiences within broader social, cultural, and political arenas’ (Johnson et al., 2014: 533). Moreover, according to Southwick et al. (2014), ‘determinants of resilience include a host of biological, psychological, social and cultural factors that interact with one another to determine how one responds to stressful experiences’ (p. 2). Therefore, an instrument from and for principals is needed due to its context-dependent nature (Ungar, 2021). In addressing these concerns and gaps, the current study aims at developing and validating a concise Principal Resilience Inventory in order to prompt principal resilience research and practice.

Conceptualizing resilience and principal resilience

Resilience is a psychological construct leading to promising individual and organisational outcomes despite challenges and adversities (Masten, 2021; Richardson, 2017; Ungar, 2021). The word resilience derives from the Latin verb resilire, meaning to ‘jump back’ or ‘recoil’ and translating as ‘bounce back’ (Smith et al., 2013). As resilience is an elusive concept, no consensus on the definition has been reached so far. To better understand how resilience has been defined, conceptualised and theorised, we examine the four developmental waves of scientific inquiries on resilience research.

Resilience research began with studies examining how children overcome major or chronic adversity (Van Breda, 2018; Werner and Smith, 1982). The first wave of resilience research was a pursuit of ‘what’ questions to ask, seeking to define resilience with a trait-focused approach by identifying characteristics and their correlates of individuals who react positively to setbacks (Fletcher and Sarkar, 2013; Masten, 2021). Connor and Davidson (2003) defined resilience as a personal characteristic, i.e. ‘the personal qualities that enable one to thrive in the face of adversity’ (p. 76). Rather than simply identifying qualities and influential factors of resilience, the second wave denoted resilience as a process (Loomis, 2019). This wave shifted to a process-focused approach using ‘how’ questions, uncovering how potential assets or protective/promotive and risk factors worked to produce resilient outcomes. Luthar and Cicchetti (2000) regarded resilience as ‘a dynamic process encompassing positive adaptation within the context of significant adversity’ (p. 543). This definition highlights resilience as a dynamic process as well as being contextually based. However, Ungar (2021) claims resilience as linear causal linkages without accounting for mutual trade-offs among various systems. The third wave promoted resilience through prevention, intervention, and policy as a result of the concomitant rise in interest in the field. Masten et al. (1990) considered resilience as ‘the process of, capacity for, or outcome of successful adaptation despite challenging or threatening circumstances’ (p. 426). With the evolution of the literature, the current fourth wave focuses on the system approach which regards resilience as a dynamic, episodic, multiple-dimensional system, and investigates ‘where’ and ‘when’ questions based on the ‘what’ and ‘how’ questions of previous waves (Bryan et al., 2018; Vella and Pai, 2019). Van Breda (2018) considered resilience as ‘the multilevel processes that systems engage in to obtain better-than-expected outcomes in the face or wake of adversity’ (p. 4). Regardless of the definitions adopted across different disciplines, scholars currently share a common understanding of the key features of resilience in that it is ‘a process associated with change over time that produces a preferred outcome for one or more systems’ (Ungar, 2021: 8).

When the concept of resilience was first applied in the education field, the growth in interest and usage created a pressing need to define the traits, and identify the protective and risk factors, that facilitate and impede the resilience of educational leaders so that they can overcome challenges in the face of significant change, disruption, and adversity in schools (Reed and Reedman, 2020). However, scant research has conceptualised principal or leader resilience (Foerster and Duchek, 2018). Examining the teacher resilience literature, teacher resilience is defined as ‘the dynamic and complex interplay between individual, relational, and contextual conditions that either enable or constrain teachers’ power and agency’ (Johnson et al., 2014: 7). In addition, Day and Gu (2014) also proposed ‘everyday resilience’ rather than focusing on highly traumatic experiences (p. 5). This point echoes the observation on adversity by Davis et al. (2009) that ‘for most of us, the adversities we encounter do not constitute major disasters but rather are more modest disruptions that are embedded in our everyday lives’ (p. 1638). Although there is little principal resilience literature, Patterson et al. (2009) regarded a resilient educational leader as the one who demonstrates the capacity to recover and learn from, as well as grow stronger in the face of, hardship. Rather than referring only to the capacity to ‘bounce back’ (Wells and Klocko, 2018), one scholar in the principal resilience literature proposes the term to ‘bounce forward’ (Allison, 2012) and believes that leaders ‘not only do resilient leaders quickly get their mojo back, but because they understand that the status quo is unsustainable, they also use it to move mountains’ (p. 80).

Taken together, in this project, principal resilience is regarded as a multidimensional concept that represents a dynamic process on how an individual principal negotiates with resources and navigates everyday challenges and adversities to move over time from surviving to thriving and beyond across one or more systems leading to positive outcomes for quality education. This definition manifests seven generic principles of resilience with two major features (dynamic and contextualised) of resilience in the current literature drawn from seminal scholars (e.g. Masten, 2021; Ungar, 2021; van Breda, 2018). These principles are: (1) resilience is a process related to persistence, resistance, recovery, adaptation, and transformation in contexts of adversity; (2) resilience capacity can change, develop, and cascade; (3) a resilient system is open, dynamic, and complex; (4) a resilient system promotes connectivity; (5) a resilient system demonstrates experimentation and learning; (6) a resilient system includes diversity, redundancy, and participation; and (7) there are trade-offs between systems when a system experiences resilience. This operational definition and these seven generic principles will function to guide the research design and instrument development of this project.

Measurements of principal resilience

Influenced by the various factions offering resilience definitions, scholars have put different elements into the ‘resilience spin’ which makes it hard to measure resilience accurately. Van Breda (2018) described this phenomenon as ‘resilience, to some extent, has become an empty word that can be filled with almost any meaning’ (p. 2). Some measures use the term ‘resilience’, but assess separate constructs (Loomis, 2019). For example, the widely-cited Connor-Davidson resilience scale (2003) equates stress-coping ability with resilience. However, by definition, these proactive factors of resilience may not represent resilience itself. Some widely-cited instruments have been designed with multiple dimensions referencing resilient traits and/or protective factors. For example, the Child and Youth Resilience instrument developed by Liebenberg et al. (2012) covers personal skills, peer support, and social skills, physical caregiving, psychological caregiving, spirituality, culture, and education. However, these instruments may not be relevant for measuring principal resilience in a given context (e.g. the workplace).

Examining the principal resilience literature, three measurements originally developed for general populations in the US have been adopted to assess school principals. The first two focus on the traits and behaviours of resilient leaders or individuals. The first is the Personal Resilience Questionnaire with 70 items developed by Conner (1993) in the US to identify a combination of traits that enable people to assimilate change. The measurement evaluates flexible (social), flexible (thoughts), focused, organized, positive (yourself), positive (the world), and proactive. Scholars (e.g. Kiser, 2020; Offutt, 2011; Wescott, 2018) adopted this instrument for assessing principal resilience in the US setting. The second, the Assessment of Core Resilience questionnaire with 29 items developed by Shores (2004), also in the US, aims to investigate adult core resilience on three subscales, namely, love of self, love of others, and love of a higher power. This was adopted by studies (e.g. Bumphus, 2008) of school leader resilience in the US setting. The third one, the Leader Resilience Profile, was designed to assess the degree of engagement with resilient behaviours (Patterson et al., 2009; Reed and Reedman, 2020) that are more adaptive to the dynamic nature of resilience, covering process, capacity, and outcome. It consists of 44 items that can be allocated into 11 scales, namely, optimism, personal ethical values, personal efficacy/power, social support base, emotional, spiritual, and physical wellbeing, perseverance, adaptability, courageous decision making, and responsibility. Patterson's team (Reed and Reedman, 2020) in the US and other scholars (Lazaridou, 2021) in other contexts have used this instrument to assess school leader resilience.

These three frequently-cited instruments have significantly contributed to measuring principal resilience. However, due to the general neglection of the field and the blurred conceptualization of principal resilience, there is no specific instrument to measure principal resilience. Borrowing from other populations may lead to ill-measure of principal resilience, which may hold back the maturity of knowledge construction. Moreover, the first two were organically developed for a general population but not for school principals. The third was developed from school leaders but for application to leaders in general rather than especially to school leaders. The content of these instruments may not adequately represent and capture the essence of principal resilience. Furthermore, the third instrument focused particularly on leaders in the face of adversity. As ‘everyday resilience’ is incorporated into the notion of adversity in conceptualising resilience (Day and Gu, 2014), everyday challenges need to be embedded into the context (Ungar, 2021). In addition, as the resilience definition has evolved to incorporate a system approach and the multidimensional perspective in the fourth wave mentioned above (Masten, 2021; Ungar, 2021), an instrument that stays current with the resilience and principal resilience literature is also needed. ‘What we measure affects what we do; and if our measurements are flawed, decisions may be distorted’ (Stiglitz et al., 2009: 7). Therefore, the aim of this project is to develop and validate a concise multidimensional instrument on principal resilience that is developed from and for school principals. This PRI was expected to comprise four to six domains with four to five items for each dimension so that it can be used for various purposes (e.g. examining the nature of principal resilience and its associations, drivers, and consequences). The PRI was only concerned with certain salient resilience experience but skipped ones that might be also important in principal professional life.

Methods

Stage 2. Establishing construct validity

Phase 2 tested construct validity of the PRI using two samples of principals in two sequential empirical studies (Study 2 and Study 3).

Study 2

Study 2 tended to show preliminary results on the dimension structure, reduce the items which did not fit well, and unveil the latent dimensions of the PRI.

Sample. A sample of 216 principals from 115 (53.2%) primary and 101 (46.8%) secondary schools in Hong Kong through professional connections was approached using an on-line survey. 98 (45.4%) were females and 118 (54.6%) males with an average age of 44 years and 8 years of working experience. Approximately 29.2% (n = 63) of these principals had less than five years of working experience of being a principal, 32.3% (n = 70) had six to 10 years, 30.6% (n = 66) had 11 to 20 years, and 7.7% (n = 16) had more than twenty years. About 33.3% (n = 72) of them held a junior college degree or below, 66.2% (n = 143) held a bachelor degree, and 0.5% (n = 1) held a doctoral degree.

Procedure and Results. Besides demographic questions, the participants were asked to indicate their agreement with each item of the 48-item PRI. The study employed a symmetric agreement Likert scale with 6 points without a neutral point. Exploratory factor analysis (EFA) was employed to generate the model using SPSS 26. In the EFA procedure, the following criteria were employed: (1) all items with regression loadings greater than 0.60 on their intended conceptual factors; (2) three or more conceptual aligned items in one factor; (3) all items’ cross loadings fewer than 0.30; and (4) all items with no negative error variance (Costello and Osborne, 2005). During this process, 16 items were dropped and a 32-item PRI with the same six factors remained, namely, professional (5 items), physical (6 items), emotional (6 items), psychological (5 items), social (5 items), and spiritual resilience (5 items). Descriptive statistics also showed that these principals gave the highest ranking to spiritual resilience (M = 5.28 out of 6), followed by social (M = 5.25; SD = .69), emotional (M = 5.11; SD = .70), psychological (M = 5.07; SD = .70), and physical resilience (M = 4.69; SD = .78). A Multivariate analysis of variance (MANOVA) with main effects for school level, gender, age, and working experience on was performed. The results found that only school level main characteristic (School Level: F_6,2068 = 3.29, p < .01) had statistically significant mean differences in principal resilience with a higher resilience level of secondary school principals than their peers from primary schools in Hong Kong. When examining closely, Univariate analysis indicated that the significant school level effect applied to only physical dimension (F₂ = 3.15, p < .05; M_Secondary = 4.87 vs M_Primary = 4.51).

Study 3

Study 3 tended to confirm the latent structure and test the psychometric characteristics of the PRI with various samples of respondents.

Sample. A different sample of 698 principals, 310 (44.4%) from primary and 388 (55.6%) from secondary schools from Mainland China responded. Among these participants, 323 (46.3%) females and 375 (53.7%) were males with an average age of 44 years and 7 years of working experience. More than half (66.3%, n = 334) of these principals had less than ten years of working experience, 27.4% (n = 138) had ten to twenty years, and 6.3% (n = 32) had more than twenty years. The majority of them (66.9%, n = 467) held a bachelor degree, 32.1% (n = 224) of them held a junior college degree or below, and 1% (n = 7) held a master degree and above.

Procedure. The exploratory model developed from Study 2 was then assessed in confirmative factor analysis (CFA) which allowed better detection of items with weak loadings and items with strong covariance with other factors which is a violation of simple structure (i.e. items must belong to only one factor). In other words, all measured variables in EFA are related to every latent variable. For example, one item (measured variable) belongs to one factor (latent variable), but also connects with other factors. By contrast, items in CFA are only related to the designated factors but have zero relationships with other factors. The following criteria in the CFA procedure were employed for model fit evaluation: (1) Items with low factor loadings <0.5 may be considered for deletion (Hair et al., 2009); (2) Items with standardized residuals >2.58 were considered problematic items, which were further evaluated before removal (Brown, 2015); (3) Items with high Modification indices (usually above 10) can be considered for deletion, but this should be done in conjunction with substantive considerations on model fit indices (Brown, 2015; Marsh et al., 2004). All analyses were carried out in Mplus 8.

Results. This section reported the CFA model, reliability, and internal construct validity. First, following the criteria of the EFA procedure and CFA model fit index, and the original scope and principles of developing the PRI, the CFA procedure rejected 8 items. Three trimmed models were identified and compared. They consist of Model 1 with the one factor solution, Model 2 with the original six factor solution, and Model 3 with the 5-factor solution (Table 1). The 20-item model with five dimensions was given the same analysis and demonstrated a better model fit. Among the original six factors, the factor professional resilience was discarded and the rest of the five factors namely physical, emotional, psychological, social, and spiritual resilience with 4 items for each dimension were retained (Table 2).

OPEN IN VIEWER

Table 1.

Model fit comparison.

	Chi-square (df)	CFI	TLI	Gamma hat	RMSEA	90% CI	SRMR
Model 1	1596.21(170)	.877	.863		.110	.106∼.115	.057
Model 2	1614.27(237)	.912	.897		.091	.087∼.095	.050
Model 3	846.89(156)	.940	.927		.080	.074∼.085	.045

Note. Model 1 = the model with one factor; Model 2 = the model with six factors (solution is not admissible); Model 3 = the model with five factors; N = 698.

OPEN IN VIEWER

Table 2.

Principal Resilience Inventory (PRI) factors, items, and factor loadings.

PRI factors and items	Factor loading
F1. Physical Resilience
1. I seldom feel physical discomfort with my busy leadership routines	.60
2. I feel energized every day to do my leadership routines.	.77
3. I am a strong person with my busy leadership routines.	.78
4. I have tried the ways to craft physical energy so that I can handle my leadership work better.	.87
F2. Emotional Resilience
5. I have a positive influence in making things happen.	.92
6. I demonstrate an understanding of my emotions and how these emotions affect my leadership performance.	.83
7. I balance my professional life with other life dimensions.	.73
8. Coping with everyday challenges and adversities can make me stronger.	.89
F3. Psychological Resilience
9. I am at my best when I have a clear professional goal to strive for fulfilling my leadership role.	.86
10. I always accomplish professional but challenging tasks.	.84
11. I like changes and challenges and am willing to take risks.	.68
12. I demonstrate perseverance in overcoming everyday challenges and adversities.	.89
F4. Social Resilience
13. I can view situations from other people's perspectives.	.81
14. I try to learn from the experience of others who have a strong track record of demonstrating strong resilience.	.85
15. I’m the one that my colleagues can generally rely on.	.78
16. I am very willing to acknowledge others’ efforts and successes at school.	.74
F5. Spiritual Resilience
17. Ethics serve as the foundation for my leadership routines, even in the face of strong opposing forces.	.86
18. I make decisions based on professional values even in the face of strong opposing forces.	.66
19. I pay attention to the intention and process of every advancement of the profession.	.79
20. When facing changes in policy, I can pause without immediately rushing to make adjustments for the school, but understand the missions and values behind these.	.65

Second, internal reliability of each scale in the PRI was tested. The internal consistency analysis (Cronbach's alphas) ranged from .82 to .89 with an average value of .86 (Table 3). Descriptive statistics also showed that these principals gave the highest ranking to spiritual resilience (M = 5.25 out of 6), followed by social (M = 5.22; SD = .71), emotional (M = 5.08; SD = .73), psychological (M = 5.05; SD = .77), and physical resilience (M = 4.79; SD = .90). A Multivariate analysis of variance (MANOVA) with main effects for school level, gender, age, and working experience on was performed. The results found that only gender main characteristic (Gender: F_8,1492 = 3.31, p < .01) had statistically significant mean differences in principal resilience with a higher resilience level demonstrated by male principals than female principals. Univariate analysis indicated that the significant gender effect applied to three dimensions including physical (F₂ = 5.85, p < .01; M_Male = 4.85 vs M_Female = 4.73), emotional (F₂ = 5.74, p < .01; M_Male = 5.15 vs M_Female = 5.01), and social dimensions (F₂ = 3.68, p < .05; M_Male = 5.27 vs M_Female = 5.17).

OPEN IN VIEWER

Table 3.

PRI descriptive statistics, Cronbach α, and factor inter-correlations.

Resilience Scale	M1 (SD)	Cronbach (α)	Factor inter-correlations (r)					Psychological Contract	Work–family conflict
			F1	F2	F3	F4	F5
F1. Physical	4.79 (.90)	.82	—	.54**	.66**	.57**	.36**	.55**	−.78
F2. Emotional	5.08 (.73)	.89		—	.58**	.60**	.50**	.83**	−.64
F3. Psychological	5.05 (.77)	.88			—	.52**	.73**	.88**	−.53
F4. Social	5.22 (.71)	.86				—	.53**	.81**	−.33
F5. Spiritual	5.25 (.80)	.86					—	.46**	−.31

Note. N = 698.

**Correlation is significant at the 0.01 level (2-tailed).

Third, internal construct validity was also examined using correlation coefficient. Table 3 demonstrated that the inter-correlations ranged from .36 to .73 with an average value of .56. Latent inter-correlations between five dimensions also yielded a logical trend. These dimensions were moderately correlated with each other. Such findings demonstrated a good convergent validity of the five dimensions of principal resilience. These evidences revealed that separate resilience domains could be different even if they were moderately or strongly associated. To sum up, the initial test of internal validity of the PRI indicated a sufficient degree of convergent and divergent validity of its scales which gave implications for testing criterion validity next.

Stage 3. Evaluating criterion validity

This phase aimed at testing criterion validity including convergent, discriminate, and predictive validities between the PRI dimensions and the other two measures using the same sample as Study 3 (Krabbe, 2017).

Procedure and Measures. Convergent validity describes how closely the new scale is associated with other variables and other measures of the same construct, whilst divergent validity demonstrates that the two measures are believed to be negative or unrelated (Krabbe, 2017). Convergent and discriminant validity was examined using correlation coefficient with the psychological contract survey (Liu et al., 2008) and the work-family conflict survey (Netemeyer et al., 1996). Predictive validity refers to the degree to which a construct scale predicts a criterion variable measuring a future outcome, behaviour, or performance (Krabbe, 2017). Predictive validity was examined with the same two instruments using structural equation modelling (SEM). The 8-item psychological contract survey developed by Liu et al. (2008) measures the mutual expectation or agreement between individuals and organizations on mutual responsibility obligations. It reflects that school principals try to keep a balance between contribution and income with schools. The 5-item work-family conflict survey was developed by Netemeyer et al. (1996). The survey examined work influence on family life. These two surveys have been adopted with stable reliabilities in the Chinese context (e.g. Li et al., 2022). Previous research identified that human resilience may positively predict to psychological contract (Hind et al., 1996; Mccoy and Elwood, 2009) and negatively connect with the work-family conflict (Billing et al., 2023; Chen et al., 2021).

Results. As expected, all dimensions from the PRI are positively related to the psychological contract scale (r ranged from .46 to .88) and negatively related to the work-family conflict scale (r ranged from −.78 to −.31) (Table 3). These indicate that the PRI demonstrated a good convergent and discriminant validity. The SEM model showed that the dimensions of the PRI positively predicted the psychological contract construct and negatively predicted the work-family conflict construct (Figure 1). Descriptive statistics also showed that these principals gave the highest ranking to spiritual resilience (M = 5.27 out of 6), followed by social (M = 5.24; SD = .75), emotional (M = 5.10; SD = .82), psychological (M = 5.06; SD = .89), and physical resilience (M = 4.74; SD = .81). A Multivariate analysis of variance (MANOVA) with main effects for research location, school level, gender, age, and working experience on was performed. The results found that only sex main characteristic (Gender: F_8,1510 = 3.41, p < .01) had statistically significant mean differences in principal resilience with a higher resilience level demonstrated by male principals than female principals. Univariate analysis indicated that the significant gender effect applied to two dimensions including physical (F₂ = 5.87, p < .01; M_Male = 4.83 vs M_Female = 4.65) and emotional dimensions (F₂ = 3.42, p < .05; M_Male = 5.15 vs M_Female = 5.05).

OPEN IN VIEWER

Figure 1.

The model shows paths from principal resilience to psychological contract and work-family conflict.

Phase 4. Evaluating cross-cultural validity

Stage 4 aimed at establishing cross-cultural validity of the trimmed PRI using the two samples from two contexts of Hong Kong (the sample from Study 2) and Mainland China (the sample from Study 3). Cross-cultural validity refers to the degree to which the performance of the items in a translated or culturally adapted instrument adequately reflects the performance of the items in the original version of the instrument (Mokkink et al., 2018). It aimed at testing the generalizability of the PRI by fitting on two cross-cultural samples.

Procedure. Cross-cultural validation was performed by applying a commonly used stepwise procedure of model invariance analysis (Cheung and Rensvold, 2002; van de Schoot et al., 2012). This procedure involves inspecting a (1) configural invariance model which assumes that the same item is associated with the same factor in each group, (2) metric invariance model that additionally assumes equivalence of item factor loadings, and (3) scalar invariance model that additionally assumes equality of item intercepts. The nested models (i.e. configural, metric, and scalar) with increasingly constrained parameters were tested for model equivalency across principal samples from two different educational levels. As with the CFA procedure, the same criteria and model fit indices were applied in the model comparison procedure.

Results. The two models with a good fit were identified (Table 4). Overall, the configural invariance models had an acceptable fit and the changes of CFI (ΔCFI) were greater than .01 for equivalent regression weights, factor invariance, and metric invariance, indicating that models can be assumed to be equivalent across the two contexts. As such, it can be concluded that the PRI with two samples one from Hong Kong and one from Mainland China measure an equivalent set of latent variables. Hence, the PRI can be used across these two contexts, and possibly could be utilised in other educational contexts through the robust validation procedure.

OPEN IN VIEWER

Table 4.

Fit statistics and invariance results for two models.

Model	Chi-square	CFI	TLI	RMSEA	SRMR	ΔCFI
Mainland China (698)	χ2 = 846.89df = 156p < .001	.940	.927	.080	.045	-
Hong Kong (216)	χ2 = 371.84df = 157p < .001	.942	.929	.080	.041	-
Configural	χ2 = 1212.22df = 312p < .001	.941	.928	.079	.045	-
Metric	χ2 = 1230.07df = 324p < .001	.941	.930	.078	.050	.000
Scalar	χ2 = 1249.24df = 339p < .001	.940	.933	.077	.052	.001

Taken together, findings from these content validity analyses, accompanied by those from construct validity, criterion validity, and cross-cultural validation procedures described above, were considered to provide ample validity evidence in favour of the PRI as a promising, dependable measure developed based on theoretical and empirical evidences.

Discussion

Using a mixed-methods approach with different samples of school principals, this project developed and validated a concise multidimensional Principal Resilience Inventory. The major results on the development and validation of the instrument are discussed by intertwining with the existing literature in this section.

Based on existing work and theories and the evidence from this project, a theory- and conception-driven taxonomy was identified to categorize principal resilience and map dimension boundaries. The PRI embodies a general picture of salient resilience experience at work as a school principal from a multidimensional perspective. This measurement represents an extended view of principal resilience which hugs not only individual resilience literacy and sense-making of the principal role, but also the broader educational context (Allison, 2012; Allison-Napolitano, 2022; Day, 2014; Glazzard and Stones, 2021; Kiser, 2020; Lazaridou, 2021; Maulding et al., 2012; Papaioannou et al., 2021; Patterson et al., 2009; Reed and Reedman, 2020; Wells and Klocko, 2018). In other words, this instrument not only demonstrates the resilience experience of principals in the workplace but also supports the multidimensional views of principal resilience (Patterson et al., 2009; Viac and Fraser, 2020). Unexpectedly, the PRI encompasses just five dimensions that capture the most salient resilience experience of school principals namely physical, emotional, psychological, cognitive, and social resilience. Interestingly, the professional dimension was opted out by comparing the three models which included the one-factor solution and the six-factor solution. Possibly the items in professional resilience overlapped with those from the other five dimensions based on content meaning and inter-correlations with other dimensions. These five dimensions are associated with job nature, resilience literacy, students, colleagues, school, and beyond.

Most importantly, these five dimensions of principal resilience conceptualise the essentials of principal resilience, which may support concretely defining principal resilience in the literature. In particular, physical resilience refers to the process by which a principal's body rises to physical challenges or changes, maintains stamina, and recovers efficiently and effectively when damaged or to avoid damage. The elements consist of physical strength, recovery capacity, physical adaptability, and self-care. Emotional resilience refers to the process whereby principals emotionally respond to their jobs, and their emotional management and coping with changes and challenges. This dimension may include elements like positivity towards the future, reality and one's self, self-awareness, emotion regulation, and emotional sensitivity. Psychological resilience denotes the process of principals handling and recovering from changes and challenges through realistic evaluation of self-capacity and the obstacles, while maintaining a high level of motivation and engagement. This dimension consists of motivation, goal-orientation, job fulfilment, commitment, personal growth, perseverance, risk-taking, self-efficacy, initiative, and confidence. Social resilience refers to the process of principals tolerating, absorbing, coping with and adjusting to environmental and social interaction in the working environment with changes and challenges. This dimension may refer to elements including sympathy, social competence, social support seeking, social contribution, and social acceptance. Spiritual resilience refers to the process of principals observing, meditating, sustaining, and enhancing their sense of self and purpose by using internal and external spiritual resources, which will not be easily lost or hampered due to physical illness or adversity. This may include elements such as sense of mission, ethics, non-avoidance, harmony with surroundings, value-driven, acting with awareness, spiritual commitment, non-judgement, insight, and non-reaction (e.g. Conner, 1993; Friborg et al., 2005; Kiser, 2020; Lazaridou, 2021; Mahfouz and Richardson, 2021; Offutt, 2011; Patterson et al., 2009; Reed and Reedman, 2020; Skinner et al., 2020; van der Meer et al., 2018; Wescott, 2018).

These principals in the joint sample including the samples from Hong Kong and Mainland China gave the highest ranking to spiritual well-being, followed by social, emotional, psychological, and physical resilience. It is interesting that these participants ranked their spiritual resilience as the highest one which also significantly related the other three dimensions of resilience (emotional, psychological, and physical). When examining the correlations of the five dimensions of resilience indicate that the five dimensions of principal resilience are not only moderately interrelated but also varies in some extent. The significantly moderated correlations in this project between the five dimensions support the arguments on the nature of principal resilience as being multidimensional in one picture. Whist the discrepancies between dimensions indicate these five dimensions of principal resilience actually measured different aspects of principal resilience, whose psychometric properties are desirable to have. The results also showed that male principals had a higher resilience level than their female counterparts. Moreover, secondary principals from Hong Kong demonstrated a higher resilience level that female school principals and the statistically significant effect had applied for physical and emotional dimensions in particular. These findings will provide baseline data for future comparisons and implications for developing principal resilience interventions in future research.

Under the guidelines of the theoretical-empirical approach of test construction (Oosterveld et al., 2019), the PRI shows strong psychometric properties in terms of reliability, content validity, construct validity, criterion, and cross validity with different samples of principals. First, the PRI in this project confirmed a good content validity procedure. Fixed domains with concise items on principal resilience were identified and validated using the existing literature and responses of targeted groups, authentically covering relevant items and content domains in the PRI. This approach guarantees a greater content validity than using measures solely generated from existing literature or target respondents (Clark and Watson, 2019; Oosterveld et al., 2019; Ruscio, 2015). Second, the PRI also verified a good construct validity mechanism. All item loadings in the PRI were greater than .59 and all factor reliabilities were acceptable ranging from .56 to .88. The CFA and multiple group analysis procedures also revealed more detailed information on the construct validity of the different models of the PRI. Specifically, we tested various models and three models namely the one-factor solution, five-factor solution, and six-factor solution were constructed. Through CFA analysis, the five-factor solution with five separate latent dimensions was proven to be valid and sound based on the statistics and the current theories of principal resilience. This trimmed model demonstrated a better sense in terms of the model fit and theoretical framework. As such, the PRI is clearly suitable for assessing the five dimensions as distinct constructs of principal resilience but from a concise and multidimensional perspective. Third, the criterion validation provided evidence that the PRI shows a good external validity with two related constructs, namely, psychological contract to test convergent validity, work-family conflict to test discriminant validity. Moreover, predictive validity was also examined using these two constructs. This evidence warrants a clear conceptual integration and separation of these five dimensions of resilience as experienced by school principals in this project. Fourth, the cross-cultural validity was tested using two different samples, one from Hong Kong and one from Mainland China. The result revealed that the PRI with two independent samples measures an equivalent set of latent variables (Krabbe, 2017). Hence, the PRI can be used across these two contexts and possibly could be utilised in other educational contexts through a robust adoption procedure as well.

Overall, the PRI unveils that principals experience a sense of resilience at work, which are salient for these participants as school principals. Their sense of resilience seems strongly related to the notion of balance between work and life, intrapersonal and interpersonal, individual and context, which portrays a dynamic process of how school principals navigate and negotiate different demands and resources to deal with challenges and changes leading to desirable outcomes (Allison-Napolitano, 2022; Patterson et al., 2009; Steward, 2014; Ungar, 2021; van Breda, 2018; Wells and Klocko, 2018). One more bonus of this project was to identify two consequences (psychological contract and work-family conflict) of principal resilience. These findings may indicate a strengths-based, positive approach to principalship that offers an alternative perspective for supporting and encouraging principal resilience at work (Allison-Napolitano, 2022; Kiser, 2020; Wells and Klocko, 2018). Therefore, resilience seems a good business for every school principal.

Implications

This project makes three contributions. First, this project contributes to the advancement of knowledge production of principal resilience. Compared with the teacher resilience literature, it seems that the principal resilience literature is still in its infancy (Glazzard and Stones, 2021). The development of the multidimensional PRI will contribute to our understanding of the salient nature of principal resilience. Moreover, the PRI has been developed based on a comprehensive literature review and empirical evidence, covering nature (five dimensions) and consequences (psychological contract and work-family conflict) of principal resilience as a whole. This will encourage more scientific efforts and largely advance the maturity of knowledge production in the field (Edmondson and McManus, 2007).

Second, this project produces a concise multidimensional instrument that enables measuring the salient dimensions of principal resilience in the workplace. Our review of the current instruments measuring principal resilience indicated there is no instrument on principal resilience developed from and for principals. Therefore, the need for developing an instrument that directly measures principal resilience from and for principals through a robust validation procedure is particularly needed. The PRI will provide a validated instrument developed from and for school principals for future research. Moreover, the PRI can be used to establish test–retest reliability to assess responsiveness to change over time or in different targeted participants in future studies. Hence, the emergence of the PRI will encourage more authentic investigations of principal resilience in the future.

Third, this project is critical and timely particularly during the post-pandemic period in that the ways for enhancing the outcomes of school principals via the lens of resilience will be reinforced to help them cope with the everyday challenges and adversities that principals encounter (Day, 2014). This project will be informative for principal preparation and evaluation at the induction stage and the in-service stage. Principal preparation programs and continuous training programs might consider including the element of principal resilience extracted from empirical evidences in the project. Particularly, professional development programs can be designed to promote resilience literacy especially regarding the awareness and coping strategies of aspiring and current principals. The incorporation of resilience literacy should be regarded as an ongoing component of principal professional development and principal selection. Moreover, it is expected that this project will make policymakers, researchers, and school principals more aware of the importance of their own resilience literacy. The standards for school principals might include the resilience component as one of the professional capitals for developing principal capacity and contributing to principal evaluation mechanism.

Conclusion

Three major limitations of this project are addressed. First, the participating principals in this project were derived from convenience samples through professional networks due to the difficulties of gathering principal samples in the reality. This non-random approach may result in sampling bias (Gravetter and Forzano, 2011). Thus, results are not necessarily representative of the greater populations of interest. To correct any sampling bias, future studies may validate the PRI using random and weighted sampling techniques involving more diverse principal samples (Renshaw et al., 2015). Second, with hindsight, it is acknowledged that the decision to ask the participants to categorize their resilience experience was too simplistic, given that an individual resilience experience exists along a spectrum, is dynamic and dependent upon the contexts in which individuals operate. Moreover, principal resilience may fluctuate at specific time points during the academic year so that participating principals may not accurately report their resilience retrospectively. Hence, we wish that future research will cover where this project leaves off for the methodological issues, further validating the PRI as both a basic research instrument and means of promoting principal resilience, leading to individual and school outcomes.

To conclude, rather than using discrete constructs to measure resilience, this project reported a multidimensional instrument to better capture principal well-being at work. Leading during the uncertainties of a crisis requires school principals to respond and react effectively, cohesively and proactively using their resilience. On top of this, everyday resilience, as emphasized by Day (2014), needs to draw greater attention. Principals may infuse themselves into the scenarios by sharing an identity of ‘us-ness’ in pursuit of the ultimate goal of belonging to a coherent team in order to achieve the collective good (Haslam et al., 2020). As Day (2014) describes, everyday resilience involves the ‘processes of teaching, learning and leading, then, requires those who are engaged in them to have a resolute everyday persistence and commitment, which is much more than the ability to bounce back in adverse circumstances’ (p. 641). Not stopping here, bouncing forward will follow, which can be used to move mountains (Allison, 2012). In this regard, principal resilience is not just one of the qualities that leads to successful leadership, but is an indispensable essential of successful leadership. Resilience is an inside job that begins with choosing to be resilient. However, resilience is like a ‘fuel’ that needs to ‘be topped up’ to accumulate as a ‘fuel tank’ (Smith, 2017). In other words, support to be a resilient principal is imperative. It is not just adversities and challenges that damage resilience or impede resilience building, internal desire and internal support are necessary also.